diff options
author | Dave Airlie <airlied@redhat.com> | 2016-08-25 12:59:50 +1000 |
---|---|---|
committer | Dave Airlie <airlied@redhat.com> | 2016-08-25 12:59:50 +1000 |
commit | e9c3ddee6a08c5b25cdb06b524320a5a98250513 (patch) | |
tree | 41cc2cc030ef965c1d34b336c994a3ac9c00c13e /drivers/gpu | |
parent | 51d6120792ab5f46d6f5f7f37b65d05cc1afc019 (diff) | |
parent | 7b4d3e297e8a7d3b82e68231ff077e891c370349 (diff) | |
download | linux-e9c3ddee6a08c5b25cdb06b524320a5a98250513.tar.bz2 |
Merge branch 'drm-next-4.9' of git://people.freedesktop.org/~agd5f/linux into drm-next
First drm-next pull for radeon and amdgpu for 4.9. Highlights:
- powerplay support for iceland asics
- improved GPU reset (both full asic and per block)
- UVD and VCE powergating for CZ and ST
- VCE clockgating for CZ and ST
- Support for pre-initialized (e.g., zeroed) vram buffers
- ttm cleanups
- virtual display support
- core and radeon/amdgpu support for page_flip_target
- lots of bug fixes and clean ups
* 'drm-next-4.9' of git://people.freedesktop.org/~agd5f/linux: (171 commits)
drm/amdgpu: use memcpy_toio for VCE firmware upload
drm/amdgpu: use memcpy_to/fromio for UVD fw upload
drm/amd/powerplay: delete useless code in iceland_hwmgr.c.
drm/radeon: switch UVD code to use UVD_NO_OP for padding
drm/amdgpu: switch UVD code to use UVD_NO_OP for padding
drm/radeon: add support for UVD_NO_OP register
drm/amdgpu: add support for UVD_NO_OP register
drm/amdgpu: fix VCE ib alignment value
drm/amdgpu: fix IB alignment for UVD
drm/amd/amdgpu: Print ring name in amdgpu_ib_schedule()
drm/radeon: remove dead code, si_mc_load_microcode (v2)
drm/radeon/cik: remove dead code (v2)
drm/amd/powerplay: avoid NULL dereference, cz_hwmgr.c
drm/amd/powerplay: avoid NULL pointer dereference
drm/amdgpu/gmc8: remove dead code (v2)
drm/amdgpu/gmc7: remove dead code (v2)
drm/amdgpu: Fix indentation in dce_v8_0_audio_write_sad_regs()
drm/amdgpu: Use correct mask in dce_v8_0_afmt_setmode() and fix comment typos.
drm/amdgpu: cleanup amdgpu_vm_bo_update params
drm/amdgpu: stop adding dummy entry in amdgpu_ttm_placement_init
...
Diffstat (limited to 'drivers/gpu')
121 files changed, 14404 insertions, 2014 deletions
diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig index fe1e86ea6d27..cbb64d9a82f8 100644 --- a/drivers/gpu/drm/Kconfig +++ b/drivers/gpu/drm/Kconfig @@ -163,9 +163,6 @@ config DRM_AMDGPU If M is selected, the module will be called amdgpu. source "drivers/gpu/drm/amd/amdgpu/Kconfig" -source "drivers/gpu/drm/amd/powerplay/Kconfig" - -source "drivers/gpu/drm/amd/acp/Kconfig" source "drivers/gpu/drm/nouveau/Kconfig" diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig b/drivers/gpu/drm/amd/amdgpu/Kconfig index 7335c0420c70..f3cb69de0c44 100644 --- a/drivers/gpu/drm/amd/amdgpu/Kconfig +++ b/drivers/gpu/drm/amd/amdgpu/Kconfig @@ -25,3 +25,5 @@ config DRM_AMDGPU_GART_DEBUGFS Selecting this option creates a debugfs file to inspect the mapped pages. Uses more memory for housekeeping, enable only for debugging. +source "drivers/gpu/drm/amd/powerplay/Kconfig" +source "drivers/gpu/drm/amd/acp/Kconfig" diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index c7fcdcedaadb..21dd7c00da15 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -58,7 +58,8 @@ amdgpu-y += \ # add DCE block amdgpu-y += \ dce_v10_0.o \ - dce_v11_0.o + dce_v11_0.o \ + dce_virtual.o # add GFX block amdgpu-y += \ diff --git a/drivers/gpu/drm/amd/amdgpu/ObjectID.h b/drivers/gpu/drm/amd/amdgpu/ObjectID.h index 06192698bd96..b8d66670bb17 100644 --- a/drivers/gpu/drm/amd/amdgpu/ObjectID.h +++ b/drivers/gpu/drm/amd/amdgpu/ObjectID.h @@ -90,6 +90,7 @@ #define ENCODER_OBJECT_ID_INTERNAL_VCE 0x24 #define ENCODER_OBJECT_ID_INTERNAL_UNIPHY3 0x25 #define ENCODER_OBJECT_ID_INTERNAL_AMCLK 0x27 +#define ENCODER_OBJECT_ID_VIRTUAL 0x28 #define ENCODER_OBJECT_ID_GENERAL_EXTERNAL_DVO 0xFF @@ -119,6 +120,7 @@ #define CONNECTOR_OBJECT_ID_eDP 0x14 #define CONNECTOR_OBJECT_ID_MXM 0x15 #define CONNECTOR_OBJECT_ID_LVDS_eDP 0x16 +#define CONNECTOR_OBJECT_ID_VIRTUAL 0x17 /* deleted */ @@ -147,6 +149,7 @@ #define GRAPH_OBJECT_ENUM_ID5 0x05 #define GRAPH_OBJECT_ENUM_ID6 0x06 #define GRAPH_OBJECT_ENUM_ID7 0x07 +#define GRAPH_OBJECT_ENUM_VIRTUAL 0x08 /****************************************************/ /* Graphics Object ID Bit definition */ @@ -408,6 +411,10 @@ GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\ ENCODER_OBJECT_ID_HDMI_ANX9805 << OBJECT_ID_SHIFT) +#define ENCODER_VIRTUAL_ENUM_VIRTUAL ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_VIRTUAL << ENUM_ID_SHIFT |\ + ENCODER_OBJECT_ID_VIRTUAL << OBJECT_ID_SHIFT) + /****************************************************/ /* Connector Object ID definition - Shared with BIOS */ /****************************************************/ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 8ebc5f1eb4c0..3cc2629eb158 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -51,6 +51,7 @@ #include "amdgpu_ih.h" #include "amdgpu_irq.h" #include "amdgpu_ucode.h" +#include "amdgpu_ttm.h" #include "amdgpu_gds.h" #include "amd_powerplay.h" #include "amdgpu_acp.h" @@ -91,6 +92,8 @@ extern unsigned amdgpu_pcie_lane_cap; extern unsigned amdgpu_cg_mask; extern unsigned amdgpu_pg_mask; extern char *amdgpu_disable_cu; +extern int amdgpu_sclk_deep_sleep_en; +extern char *amdgpu_virtual_display; #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS 3000 #define AMDGPU_MAX_USEC_TIMEOUT 100000 /* 100 ms */ @@ -248,10 +251,9 @@ struct amdgpu_vm_pte_funcs { uint64_t pe, uint64_t src, unsigned count); /* write pte one entry at a time with addr mapping */ - void (*write_pte)(struct amdgpu_ib *ib, - const dma_addr_t *pages_addr, uint64_t pe, - uint64_t addr, unsigned count, - uint32_t incr, uint32_t flags); + void (*write_pte)(struct amdgpu_ib *ib, uint64_t pe, + uint64_t value, unsigned count, + uint32_t incr); /* for linear pte/pde updates without addr mapping */ void (*set_pte_pde)(struct amdgpu_ib *ib, uint64_t pe, @@ -396,46 +398,9 @@ int amdgpu_fence_wait_empty(struct amdgpu_ring *ring); unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring); /* - * TTM. + * BO. */ -#define AMDGPU_TTM_LRU_SIZE 20 - -struct amdgpu_mman_lru { - struct list_head *lru[TTM_NUM_MEM_TYPES]; - struct list_head *swap_lru; -}; - -struct amdgpu_mman { - struct ttm_bo_global_ref bo_global_ref; - struct drm_global_reference mem_global_ref; - struct ttm_bo_device bdev; - bool mem_global_referenced; - bool initialized; - -#if defined(CONFIG_DEBUG_FS) - struct dentry *vram; - struct dentry *gtt; -#endif - - /* buffer handling */ - const struct amdgpu_buffer_funcs *buffer_funcs; - struct amdgpu_ring *buffer_funcs_ring; - /* Scheduler entity for buffer moves */ - struct amd_sched_entity entity; - - /* custom LRU management */ - struct amdgpu_mman_lru log2_size[AMDGPU_TTM_LRU_SIZE]; -}; - -int amdgpu_copy_buffer(struct amdgpu_ring *ring, - uint64_t src_offset, - uint64_t dst_offset, - uint32_t byte_count, - struct reservation_object *resv, - struct fence **fence); -int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma); - struct amdgpu_bo_list_entry { struct amdgpu_bo *robj; struct ttm_validate_buffer tv; @@ -498,10 +463,12 @@ struct amdgpu_bo { struct amdgpu_device *adev; struct drm_gem_object gem_base; struct amdgpu_bo *parent; + struct amdgpu_bo *shadow; struct ttm_bo_kmap_obj dma_buf_vmap; struct amdgpu_mn *mn; struct list_head mn_list; + struct list_head shadow_list; }; #define gem_to_amdgpu_bo(gobj) container_of((gobj), struct amdgpu_bo, gem_base) @@ -677,6 +644,8 @@ struct amdgpu_mc { uint32_t fw_version; struct amdgpu_irq_src vm_fault; uint32_t vram_type; + uint32_t srbm_soft_reset; + struct amdgpu_mode_mc_save save; }; /* @@ -721,10 +690,11 @@ void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev, */ struct amdgpu_flip_work { - struct work_struct flip_work; + struct delayed_work flip_work; struct work_struct unpin_work; struct amdgpu_device *adev; int crtc_id; + u32 target_vblank; uint64_t base; struct drm_pending_vblank_event *event; struct amdgpu_bo *old_rbo; @@ -815,13 +785,17 @@ struct amdgpu_ring { /* maximum number of VMIDs */ #define AMDGPU_NUM_VM 16 +/* Maximum number of PTEs the hardware can write with one command */ +#define AMDGPU_VM_MAX_UPDATE_SIZE 0x3FFFF + /* number of entries in page table */ #define AMDGPU_VM_PTE_COUNT (1 << amdgpu_vm_block_size) /* PTBs (Page Table Blocks) need to be aligned to 32K */ #define AMDGPU_VM_PTB_ALIGN_SIZE 32768 -#define AMDGPU_VM_PTB_ALIGN_MASK (AMDGPU_VM_PTB_ALIGN_SIZE - 1) -#define AMDGPU_VM_PTB_ALIGN(a) (((a) + AMDGPU_VM_PTB_ALIGN_MASK) & ~AMDGPU_VM_PTB_ALIGN_MASK) + +/* LOG2 number of continuous pages for the fragment field */ +#define AMDGPU_LOG2_PAGES_PER_FRAG 4 #define AMDGPU_PTE_VALID (1 << 0) #define AMDGPU_PTE_SYSTEM (1 << 1) @@ -833,10 +807,7 @@ struct amdgpu_ring { #define AMDGPU_PTE_READABLE (1 << 5) #define AMDGPU_PTE_WRITEABLE (1 << 6) -/* PTE (Page Table Entry) fragment field for different page sizes */ -#define AMDGPU_PTE_FRAG_4KB (0 << 7) -#define AMDGPU_PTE_FRAG_64KB (4 << 7) -#define AMDGPU_LOG2_PAGES_PER_FRAG 4 +#define AMDGPU_PTE_FRAG(x) ((x & 0x1f) << 7) /* How to programm VM fault handling */ #define AMDGPU_VM_FAULT_STOP_NEVER 0 @@ -846,6 +817,7 @@ struct amdgpu_ring { struct amdgpu_vm_pt { struct amdgpu_bo_list_entry entry; uint64_t addr; + uint64_t shadow_addr; }; struct amdgpu_vm { @@ -948,7 +920,6 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, struct amdgpu_job *job); int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job); void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vm_id); -uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr); int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, struct amdgpu_vm *vm); int amdgpu_vm_clear_freed(struct amdgpu_device *adev, @@ -957,7 +928,7 @@ int amdgpu_vm_clear_invalids(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct amdgpu_sync *sync); int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, - struct ttm_mem_reg *mem); + bool clear); void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, struct amdgpu_bo *bo); struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm, @@ -1195,6 +1166,10 @@ struct amdgpu_gfx { unsigned ce_ram_size; struct amdgpu_cu_info cu_info; const struct amdgpu_gfx_funcs *funcs; + + /* reset mask */ + uint32_t grbm_soft_reset; + uint32_t srbm_soft_reset; }; int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, @@ -1683,6 +1658,7 @@ struct amdgpu_uvd { bool address_64_bit; bool use_ctx_buf; struct amd_sched_entity entity; + uint32_t srbm_soft_reset; }; /* @@ -1709,6 +1685,7 @@ struct amdgpu_vce { struct amdgpu_irq_src irq; unsigned harvest_config; struct amd_sched_entity entity; + uint32_t srbm_soft_reset; }; /* @@ -1729,6 +1706,7 @@ struct amdgpu_sdma { struct amdgpu_irq_src trap_irq; struct amdgpu_irq_src illegal_inst_irq; int num_instances; + uint32_t srbm_soft_reset; }; /* @@ -1956,6 +1934,7 @@ struct amdgpu_ip_block_status { bool valid; bool sw; bool hw; + bool hang; }; struct amdgpu_device { @@ -2055,6 +2034,7 @@ struct amdgpu_device { atomic_t gpu_reset_counter; /* display */ + bool enable_virtual_display; struct amdgpu_mode_info mode_info; struct work_struct hotplug_work; struct amdgpu_irq_src crtc_irq; @@ -2117,6 +2097,10 @@ struct amdgpu_device { struct kfd_dev *kfd; struct amdgpu_virtualization virtualization; + + /* link all shadow bo */ + struct list_head shadow_list; + struct mutex shadow_list_lock; }; bool amdgpu_device_is_px(struct drm_device *dev); @@ -2192,6 +2176,9 @@ void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v); #define REG_GET_FIELD(value, reg, field) \ (((value) & REG_FIELD_MASK(reg, field)) >> REG_FIELD_SHIFT(reg, field)) +#define WREG32_FIELD(reg, field, val) \ + WREG32(mm##reg, (RREG32(mm##reg) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field)) + /* * BIOS helpers. */ @@ -2242,7 +2229,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_gart_flush_gpu_tlb(adev, vmid) (adev)->gart.gart_funcs->flush_gpu_tlb((adev), (vmid)) #define amdgpu_gart_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gart.gart_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags)) #define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count))) -#define amdgpu_vm_write_pte(adev, ib, pa, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pa), (pe), (addr), (count), (incr), (flags))) +#define amdgpu_vm_write_pte(adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr))) #define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags))) #define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib))) #define amdgpu_ring_test_ring(r) (r)->funcs->test_ring((r)) @@ -2387,6 +2374,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) /* Common functions */ int amdgpu_gpu_reset(struct amdgpu_device *adev); +bool amdgpu_need_backup(struct amdgpu_device *adev); void amdgpu_pci_config_reset(struct amdgpu_device *adev); bool amdgpu_card_posted(struct amdgpu_device *adev); void amdgpu_update_display_priority(struct amdgpu_device *adev); @@ -2412,6 +2400,8 @@ uint32_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, void amdgpu_vram_location(struct amdgpu_device *adev, struct amdgpu_mc *mc, u64 base); void amdgpu_gtt_location(struct amdgpu_device *adev, struct amdgpu_mc *mc); void amdgpu_ttm_set_active_vram_size(struct amdgpu_device *adev, u64 size); +u64 amdgpu_ttm_get_gtt_mem_size(struct amdgpu_device *adev); +int amdgpu_ttm_global_init(struct amdgpu_device *adev); void amdgpu_program_register_sequence(struct amdgpu_device *adev, const u32 *registers, const u32 array_size); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c index 983175363b06..1b621160b52e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c @@ -259,6 +259,33 @@ static const int object_connector_convert[] = { DRM_MODE_CONNECTOR_Unknown }; +bool amdgpu_atombios_has_dce_engine_info(struct amdgpu_device *adev) +{ + struct amdgpu_mode_info *mode_info = &adev->mode_info; + struct atom_context *ctx = mode_info->atom_context; + int index = GetIndexIntoMasterTable(DATA, Object_Header); + u16 size, data_offset; + u8 frev, crev; + ATOM_DISPLAY_OBJECT_PATH_TABLE *path_obj; + ATOM_OBJECT_HEADER *obj_header; + + if (!amdgpu_atom_parse_data_header(ctx, index, &size, &frev, &crev, &data_offset)) + return false; + + if (crev < 2) + return false; + + obj_header = (ATOM_OBJECT_HEADER *) (ctx->bios + data_offset); + path_obj = (ATOM_DISPLAY_OBJECT_PATH_TABLE *) + (ctx->bios + data_offset + + le16_to_cpu(obj_header->usDisplayPathTableOffset)); + + if (path_obj->ucNumOfDispPath) + return true; + else + return false; +} + bool amdgpu_atombios_get_connector_info_from_object_table(struct amdgpu_device *adev) { struct amdgpu_mode_info *mode_info = &adev->mode_info; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h index 8c2e69661799..15dd43ec38bb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h @@ -140,6 +140,8 @@ struct amdgpu_i2c_bus_rec amdgpu_atombios_lookup_i2c_gpio(struct amdgpu_device * uint8_t id); void amdgpu_atombios_i2c_init(struct amdgpu_device *adev); +bool amdgpu_atombios_has_dce_engine_info(struct amdgpu_device *adev); + bool amdgpu_atombios_get_connector_info_from_object_table(struct amdgpu_device *adev); int amdgpu_atombios_get_clock_info(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c index 33e47a43ae32..345305235349 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c @@ -39,7 +39,8 @@ static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size, start_jiffies = jiffies; for (i = 0; i < n; i++) { struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; - r = amdgpu_copy_buffer(ring, saddr, daddr, size, NULL, &fence); + r = amdgpu_copy_buffer(ring, saddr, daddr, size, NULL, &fence, + false); if (r) goto exit_do_move; r = fence_wait(fence, false); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index ff0b55a65ca3..319a5e1d9389 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -1504,6 +1504,86 @@ static const struct drm_connector_funcs amdgpu_connector_edp_funcs = { .force = amdgpu_connector_dvi_force, }; +static struct drm_encoder * +amdgpu_connector_virtual_encoder(struct drm_connector *connector) +{ + int enc_id = connector->encoder_ids[0]; + struct drm_encoder *encoder; + int i; + for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) { + if (connector->encoder_ids[i] == 0) + break; + + encoder = drm_encoder_find(connector->dev, connector->encoder_ids[i]); + if (!encoder) + continue; + + if (encoder->encoder_type == DRM_MODE_ENCODER_VIRTUAL) + return encoder; + } + + /* pick the first one */ + if (enc_id) + return drm_encoder_find(connector->dev, enc_id); + return NULL; +} + +static int amdgpu_connector_virtual_get_modes(struct drm_connector *connector) +{ + struct drm_encoder *encoder = amdgpu_connector_best_single_encoder(connector); + + if (encoder) { + amdgpu_connector_add_common_modes(encoder, connector); + } + + return 0; +} + +static int amdgpu_connector_virtual_mode_valid(struct drm_connector *connector, + struct drm_display_mode *mode) +{ + return MODE_OK; +} + +int amdgpu_connector_virtual_dpms(struct drm_connector *connector, int mode) +{ + return 0; +} + +static enum drm_connector_status + +amdgpu_connector_virtual_detect(struct drm_connector *connector, bool force) +{ + return connector_status_connected; +} + +int amdgpu_connector_virtual_set_property(struct drm_connector *connector, + struct drm_property *property, + uint64_t val) +{ + return 0; +} + +static void amdgpu_connector_virtual_force(struct drm_connector *connector) +{ + return; +} + +static const struct drm_connector_helper_funcs amdgpu_connector_virtual_helper_funcs = { + .get_modes = amdgpu_connector_virtual_get_modes, + .mode_valid = amdgpu_connector_virtual_mode_valid, + .best_encoder = amdgpu_connector_virtual_encoder, +}; + +static const struct drm_connector_funcs amdgpu_connector_virtual_funcs = { + .dpms = amdgpu_connector_virtual_dpms, + .detect = amdgpu_connector_virtual_detect, + .fill_modes = drm_helper_probe_single_connector_modes, + .set_property = amdgpu_connector_virtual_set_property, + .destroy = amdgpu_connector_destroy, + .force = amdgpu_connector_virtual_force, +}; + void amdgpu_connector_add(struct amdgpu_device *adev, uint32_t connector_id, @@ -1888,6 +1968,17 @@ amdgpu_connector_add(struct amdgpu_device *adev, connector->interlace_allowed = false; connector->doublescan_allowed = false; break; + case DRM_MODE_CONNECTOR_VIRTUAL: + amdgpu_dig_connector = kzalloc(sizeof(struct amdgpu_connector_atom_dig), GFP_KERNEL); + if (!amdgpu_dig_connector) + goto failed; + amdgpu_connector->con_priv = amdgpu_dig_connector; + drm_connector_init(dev, &amdgpu_connector->base, &amdgpu_connector_virtual_funcs, connector_type); + drm_connector_helper_add(&amdgpu_connector->base, &amdgpu_connector_virtual_helper_funcs); + subpixel_order = SubPixelHorizontalRGB; + connector->interlace_allowed = false; + connector->doublescan_allowed = false; + break; } } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 0307ff5887c5..d80e5d3a4add 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -287,18 +287,56 @@ static u64 amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev) return max(bytes_moved_threshold, 1024*1024ull); } +static int amdgpu_cs_bo_validate(struct amdgpu_cs_parser *p, + struct amdgpu_bo *bo) +{ + u64 initial_bytes_moved; + uint32_t domain; + int r; + + if (bo->pin_count) + return 0; + + /* Avoid moving this one if we have moved too many buffers + * for this IB already. + * + * Note that this allows moving at least one buffer of + * any size, because it doesn't take the current "bo" + * into account. We don't want to disallow buffer moves + * completely. + */ + if (p->bytes_moved <= p->bytes_moved_threshold) + domain = bo->prefered_domains; + else + domain = bo->allowed_domains; + +retry: + amdgpu_ttm_placement_from_domain(bo, domain); + initial_bytes_moved = atomic64_read(&bo->adev->num_bytes_moved); + r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + p->bytes_moved += atomic64_read(&bo->adev->num_bytes_moved) - + initial_bytes_moved; + + if (unlikely(r)) { + if (r != -ERESTARTSYS && domain != bo->allowed_domains) { + domain = bo->allowed_domains; + goto retry; + } + } + + return r; +} + int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, struct list_head *validated) { struct amdgpu_bo_list_entry *lobj; - u64 initial_bytes_moved; int r; list_for_each_entry(lobj, validated, tv.head) { struct amdgpu_bo *bo = lobj->robj; bool binding_userptr = false; struct mm_struct *usermm; - uint32_t domain; usermm = amdgpu_ttm_tt_get_usermm(bo->tbo.ttm); if (usermm && usermm != current->mm) @@ -313,35 +351,13 @@ int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, binding_userptr = true; } - if (bo->pin_count) - continue; - - /* Avoid moving this one if we have moved too many buffers - * for this IB already. - * - * Note that this allows moving at least one buffer of - * any size, because it doesn't take the current "bo" - * into account. We don't want to disallow buffer moves - * completely. - */ - if (p->bytes_moved <= p->bytes_moved_threshold) - domain = bo->prefered_domains; - else - domain = bo->allowed_domains; - - retry: - amdgpu_ttm_placement_from_domain(bo, domain); - initial_bytes_moved = atomic64_read(&bo->adev->num_bytes_moved); - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); - p->bytes_moved += atomic64_read(&bo->adev->num_bytes_moved) - - initial_bytes_moved; - - if (unlikely(r)) { - if (r != -ERESTARTSYS && domain != bo->allowed_domains) { - domain = bo->allowed_domains; - goto retry; - } + r = amdgpu_cs_bo_validate(p, bo); + if (r) return r; + if (bo->shadow) { + r = amdgpu_cs_bo_validate(p, bo); + if (r) + return r; } if (binding_userptr) { @@ -386,8 +402,10 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true, &duplicates); - if (unlikely(r != 0)) + if (unlikely(r != 0)) { + DRM_ERROR("ttm_eu_reserve_buffers failed.\n"); goto error_free_pages; + } /* Without a BO list we don't have userptr BOs */ if (!p->bo_list) @@ -427,9 +445,10 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, /* Unreserve everything again. */ ttm_eu_backoff_reservation(&p->ticket, &p->validated); - /* We tried to often, just abort */ + /* We tried too many times, just abort */ if (!--tries) { r = -EDEADLK; + DRM_ERROR("deadlock in %s\n", __func__); goto error_free_pages; } @@ -441,11 +460,13 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, sizeof(struct page*)); if (!e->user_pages) { r = -ENOMEM; + DRM_ERROR("calloc failure in %s\n", __func__); goto error_free_pages; } r = amdgpu_ttm_tt_get_user_pages(ttm, e->user_pages); if (r) { + DRM_ERROR("amdgpu_ttm_tt_get_user_pages failed.\n"); drm_free_large(e->user_pages); e->user_pages = NULL; goto error_free_pages; @@ -462,12 +483,16 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, p->bytes_moved = 0; r = amdgpu_cs_list_validate(p, &duplicates); - if (r) + if (r) { + DRM_ERROR("amdgpu_cs_list_validate(duplicates) failed.\n"); goto error_validate; + } r = amdgpu_cs_list_validate(p, &p->validated); - if (r) + if (r) { + DRM_ERROR("amdgpu_cs_list_validate(validated) failed.\n"); goto error_validate; + } fpriv->vm.last_eviction_counter = atomic64_read(&p->adev->num_evictions); @@ -617,7 +642,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p, if (bo_va == NULL) continue; - r = amdgpu_vm_bo_update(adev, bo_va, &bo->tbo.mem); + r = amdgpu_vm_bo_update(adev, bo_va, false); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index df7ab2458e50..c38dc47cd767 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -46,6 +46,7 @@ #endif #include "vi.h" #include "bif/bif_4_1_d.h" +#include <linux/pci.h> static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev); static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev); @@ -1181,10 +1182,38 @@ int amdgpu_ip_block_version_cmp(struct amdgpu_device *adev, return 1; } +static void amdgpu_whether_enable_virtual_display(struct amdgpu_device *adev) +{ + adev->enable_virtual_display = false; + + if (amdgpu_virtual_display) { + struct drm_device *ddev = adev->ddev; + const char *pci_address_name = pci_name(ddev->pdev); + char *pciaddstr, *pciaddstr_tmp, *pciaddname; + + pciaddstr = kstrdup(amdgpu_virtual_display, GFP_KERNEL); + pciaddstr_tmp = pciaddstr; + while ((pciaddname = strsep(&pciaddstr_tmp, ";"))) { + if (!strcmp(pci_address_name, pciaddname)) { + adev->enable_virtual_display = true; + break; + } + } + + DRM_INFO("virtual display string:%s, %s:virtual_display:%d\n", + amdgpu_virtual_display, pci_address_name, + adev->enable_virtual_display); + + kfree(pciaddstr); + } +} + static int amdgpu_early_init(struct amdgpu_device *adev) { int i, r; + amdgpu_whether_enable_virtual_display(adev); + switch (adev->asic_type) { case CHIP_TOPAZ: case CHIP_TONGA: @@ -1521,6 +1550,9 @@ int amdgpu_device_init(struct amdgpu_device *adev, spin_lock_init(&adev->gc_cac_idx_lock); spin_lock_init(&adev->audio_endpt_idx_lock); + INIT_LIST_HEAD(&adev->shadow_list); + mutex_init(&adev->shadow_list_lock); + adev->rmmio_base = pci_resource_start(adev->pdev, 5); adev->rmmio_size = pci_resource_len(adev->pdev, 5); adev->rmmio = ioremap(adev->rmmio_base, adev->rmmio_size); @@ -1937,6 +1969,126 @@ int amdgpu_resume_kms(struct drm_device *dev, bool resume, bool fbcon) return 0; } +static bool amdgpu_check_soft_reset(struct amdgpu_device *adev) +{ + int i; + bool asic_hang = false; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; + if (adev->ip_blocks[i].funcs->check_soft_reset) + adev->ip_blocks[i].funcs->check_soft_reset(adev); + if (adev->ip_block_status[i].hang) { + DRM_INFO("IP block:%d is hang!\n", i); + asic_hang = true; + } + } + return asic_hang; +} + +int amdgpu_pre_soft_reset(struct amdgpu_device *adev) +{ + int i, r = 0; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; + if (adev->ip_block_status[i].hang && + adev->ip_blocks[i].funcs->pre_soft_reset) { + r = adev->ip_blocks[i].funcs->pre_soft_reset(adev); + if (r) + return r; + } + } + + return 0; +} + +static bool amdgpu_need_full_reset(struct amdgpu_device *adev) +{ + if (adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang || + adev->ip_block_status[AMD_IP_BLOCK_TYPE_SMC].hang || + adev->ip_block_status[AMD_IP_BLOCK_TYPE_ACP].hang || + adev->ip_block_status[AMD_IP_BLOCK_TYPE_DCE].hang) { + DRM_INFO("Some block need full reset!\n"); + return true; + } + return false; +} + +static int amdgpu_soft_reset(struct amdgpu_device *adev) +{ + int i, r = 0; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; + if (adev->ip_block_status[i].hang && + adev->ip_blocks[i].funcs->soft_reset) { + r = adev->ip_blocks[i].funcs->soft_reset(adev); + if (r) + return r; + } + } + + return 0; +} + +static int amdgpu_post_soft_reset(struct amdgpu_device *adev) +{ + int i, r = 0; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; + if (adev->ip_block_status[i].hang && + adev->ip_blocks[i].funcs->post_soft_reset) + r = adev->ip_blocks[i].funcs->post_soft_reset(adev); + if (r) + return r; + } + + return 0; +} + +bool amdgpu_need_backup(struct amdgpu_device *adev) +{ + if (adev->flags & AMD_IS_APU) + return false; + + return amdgpu_lockup_timeout > 0 ? true : false; +} + +static int amdgpu_recover_vram_from_shadow(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_bo *bo, + struct fence **fence) +{ + uint32_t domain; + int r; + + if (!bo->shadow) + return 0; + + r = amdgpu_bo_reserve(bo, false); + if (r) + return r; + domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type); + /* if bo has been evicted, then no need to recover */ + if (domain == AMDGPU_GEM_DOMAIN_VRAM) { + r = amdgpu_bo_restore_from_shadow(adev, ring, bo, + NULL, fence, true); + if (r) { + DRM_ERROR("recover page table failed!\n"); + goto err; + } + } +err: + amdgpu_bo_unreserve(bo); + return r; +} + /** * amdgpu_gpu_reset - reset the asic * @@ -1949,6 +2101,12 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev) { int i, r; int resched; + bool need_full_reset; + + if (!amdgpu_check_soft_reset(adev)) { + DRM_INFO("No hardware hang detected. Did some blocks stall?\n"); + return 0; + } atomic_inc(&adev->gpu_reset_counter); @@ -1967,40 +2125,88 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev) /* after all hw jobs are reset, hw fence is meaningless, so force_completion */ amdgpu_fence_driver_force_completion(adev); - /* save scratch */ - amdgpu_atombios_scratch_regs_save(adev); - r = amdgpu_suspend(adev); + need_full_reset = amdgpu_need_full_reset(adev); -retry: - /* Disable fb access */ - if (adev->mode_info.num_crtc) { - struct amdgpu_mode_mc_save save; - amdgpu_display_stop_mc_access(adev, &save); - amdgpu_wait_for_idle(adev, AMD_IP_BLOCK_TYPE_GMC); + if (!need_full_reset) { + amdgpu_pre_soft_reset(adev); + r = amdgpu_soft_reset(adev); + amdgpu_post_soft_reset(adev); + if (r || amdgpu_check_soft_reset(adev)) { + DRM_INFO("soft reset failed, will fallback to full reset!\n"); + need_full_reset = true; + } } - r = amdgpu_asic_reset(adev); - /* post card */ - amdgpu_atom_asic_init(adev->mode_info.atom_context); + if (need_full_reset) { + /* save scratch */ + amdgpu_atombios_scratch_regs_save(adev); + r = amdgpu_suspend(adev); - if (!r) { - dev_info(adev->dev, "GPU reset succeeded, trying to resume\n"); - r = amdgpu_resume(adev); +retry: + /* Disable fb access */ + if (adev->mode_info.num_crtc) { + struct amdgpu_mode_mc_save save; + amdgpu_display_stop_mc_access(adev, &save); + amdgpu_wait_for_idle(adev, AMD_IP_BLOCK_TYPE_GMC); + } + + r = amdgpu_asic_reset(adev); + /* post card */ + amdgpu_atom_asic_init(adev->mode_info.atom_context); + + if (!r) { + dev_info(adev->dev, "GPU reset succeeded, trying to resume\n"); + r = amdgpu_resume(adev); + } + /* restore scratch */ + amdgpu_atombios_scratch_regs_restore(adev); } - /* restore scratch */ - amdgpu_atombios_scratch_regs_restore(adev); if (!r) { + amdgpu_irq_gpu_reset_resume_helper(adev); r = amdgpu_ib_ring_tests(adev); if (r) { dev_err(adev->dev, "ib ring test failed (%d).\n", r); r = amdgpu_suspend(adev); + need_full_reset = true; goto retry; } + /** + * recovery vm page tables, since we cannot depend on VRAM is + * consistent after gpu full reset. + */ + if (need_full_reset && amdgpu_need_backup(adev)) { + struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; + struct amdgpu_bo *bo, *tmp; + struct fence *fence = NULL, *next = NULL; + + DRM_INFO("recover vram bo from shadow\n"); + mutex_lock(&adev->shadow_list_lock); + list_for_each_entry_safe(bo, tmp, &adev->shadow_list, shadow_list) { + amdgpu_recover_vram_from_shadow(adev, ring, bo, &next); + if (fence) { + r = fence_wait(fence, false); + if (r) { + WARN(r, "recovery from shadow isn't comleted\n"); + break; + } + } + fence_put(fence); + fence = next; + } + mutex_unlock(&adev->shadow_list_lock); + if (fence) { + r = fence_wait(fence, false); + if (r) + WARN(r, "recovery from shadow isn't comleted\n"); + } + fence_put(fence); + } for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = adev->rings[i]; if (!ring) continue; + amd_sched_job_recovery(&ring->sched); kthread_unpark(ring->sched.thread); } @@ -2020,7 +2226,6 @@ retry: /* bad news, how to tell it to userspace ? */ dev_info(adev->dev, "GPU reset failed\n"); } - amdgpu_irq_gpu_reset_resume_helper(adev); return r; } @@ -2178,22 +2383,26 @@ static ssize_t amdgpu_debugfs_regs_read(struct file *f, char __user *buf, struct amdgpu_device *adev = f->f_inode->i_private; ssize_t result = 0; int r; - bool use_bank; + bool pm_pg_lock, use_bank; unsigned instance_bank, sh_bank, se_bank; if (size & 0x3 || *pos & 0x3) return -EINVAL; + /* are we reading registers for which a PG lock is necessary? */ + pm_pg_lock = (*pos >> 23) & 1; + if (*pos & (1ULL << 62)) { se_bank = (*pos >> 24) & 0x3FF; sh_bank = (*pos >> 34) & 0x3FF; instance_bank = (*pos >> 44) & 0x3FF; use_bank = 1; - *pos &= 0xFFFFFF; } else { use_bank = 0; } + *pos &= 0x3FFFF; + if (use_bank) { if (sh_bank >= adev->gfx.config.max_sh_per_se || se_bank >= adev->gfx.config.max_shader_engines) @@ -2203,6 +2412,9 @@ static ssize_t amdgpu_debugfs_regs_read(struct file *f, char __user *buf, sh_bank, instance_bank); } + if (pm_pg_lock) + mutex_lock(&adev->pm.mutex); + while (size) { uint32_t value; @@ -2228,6 +2440,9 @@ end: mutex_unlock(&adev->grbm_idx_mutex); } + if (pm_pg_lock) + mutex_unlock(&adev->pm.mutex); + return result; } @@ -2443,7 +2658,7 @@ static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf, return -ENOMEM; /* version, increment each time something is added */ - config[no_regs++] = 0; + config[no_regs++] = 2; config[no_regs++] = adev->gfx.config.max_shader_engines; config[no_regs++] = adev->gfx.config.max_tile_pipes; config[no_regs++] = adev->gfx.config.max_cu_per_sh; @@ -2468,6 +2683,15 @@ static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf, config[no_regs++] = adev->gfx.config.gb_addr_config; config[no_regs++] = adev->gfx.config.num_rbs; + /* rev==1 */ + config[no_regs++] = adev->rev_id; + config[no_regs++] = adev->pg_flags; + config[no_regs++] = adev->cg_flags; + + /* rev==2 */ + config[no_regs++] = adev->family; + config[no_regs++] = adev->external_rev_id; + while (size && (*pos < no_regs * 4)) { uint32_t value; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index 76f96028313d..9af8d3c7ae8b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -41,7 +41,7 @@ static void amdgpu_flip_callback(struct fence *f, struct fence_cb *cb) container_of(cb, struct amdgpu_flip_work, cb); fence_put(f); - schedule_work(&work->flip_work); + schedule_work(&work->flip_work.work); } static bool amdgpu_flip_handle_fence(struct amdgpu_flip_work *work, @@ -63,16 +63,17 @@ static bool amdgpu_flip_handle_fence(struct amdgpu_flip_work *work, static void amdgpu_flip_work_func(struct work_struct *__work) { + struct delayed_work *delayed_work = + container_of(__work, struct delayed_work, work); struct amdgpu_flip_work *work = - container_of(__work, struct amdgpu_flip_work, flip_work); + container_of(delayed_work, struct amdgpu_flip_work, flip_work); struct amdgpu_device *adev = work->adev; struct amdgpu_crtc *amdgpuCrtc = adev->mode_info.crtcs[work->crtc_id]; struct drm_crtc *crtc = &amdgpuCrtc->base; unsigned long flags; - unsigned i, repcnt = 4; - int vpos, hpos, stat, min_udelay = 0; - struct drm_vblank_crtc *vblank = &crtc->dev->vblank[work->crtc_id]; + unsigned i; + int vpos, hpos; if (amdgpu_flip_handle_fence(work, &work->excl)) return; @@ -81,55 +82,23 @@ static void amdgpu_flip_work_func(struct work_struct *__work) if (amdgpu_flip_handle_fence(work, &work->shared[i])) return; - /* We borrow the event spin lock for protecting flip_status */ - spin_lock_irqsave(&crtc->dev->event_lock, flags); - - /* If this happens to execute within the "virtually extended" vblank - * interval before the start of the real vblank interval then it needs - * to delay programming the mmio flip until the real vblank is entered. - * This prevents completing a flip too early due to the way we fudge - * our vblank counter and vblank timestamps in order to work around the - * problem that the hw fires vblank interrupts before actual start of - * vblank (when line buffer refilling is done for a frame). It - * complements the fudging logic in amdgpu_get_crtc_scanoutpos() for - * timestamping and amdgpu_get_vblank_counter_kms() for vblank counts. - * - * In practice this won't execute very often unless on very fast - * machines because the time window for this to happen is very small. + /* Wait until we're out of the vertical blank period before the one + * targeted by the flip */ - while (amdgpuCrtc->enabled && --repcnt) { - /* GET_DISTANCE_TO_VBLANKSTART returns distance to real vblank - * start in hpos, and to the "fudged earlier" vblank start in - * vpos. - */ - stat = amdgpu_get_crtc_scanoutpos(adev->ddev, work->crtc_id, - GET_DISTANCE_TO_VBLANKSTART, - &vpos, &hpos, NULL, NULL, - &crtc->hwmode); - - if ((stat & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_ACCURATE)) != - (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_ACCURATE) || - !(vpos >= 0 && hpos <= 0)) - break; - - /* Sleep at least until estimated real start of hw vblank */ - min_udelay = (-hpos + 1) * max(vblank->linedur_ns / 1000, 5); - if (min_udelay > vblank->framedur_ns / 2000) { - /* Don't wait ridiculously long - something is wrong */ - repcnt = 0; - break; - } - spin_unlock_irqrestore(&crtc->dev->event_lock, flags); - usleep_range(min_udelay, 2 * min_udelay); - spin_lock_irqsave(&crtc->dev->event_lock, flags); + if (amdgpuCrtc->enabled && + (amdgpu_get_crtc_scanoutpos(adev->ddev, work->crtc_id, 0, + &vpos, &hpos, NULL, NULL, + &crtc->hwmode) + & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK)) == + (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK) && + (int)(work->target_vblank - + amdgpu_get_vblank_counter_kms(adev->ddev, amdgpuCrtc->crtc_id)) > 0) { + schedule_delayed_work(&work->flip_work, usecs_to_jiffies(1000)); + return; } - if (!repcnt) - DRM_DEBUG_DRIVER("Delay problem on crtc %d: min_udelay %d, " - "framedur %d, linedur %d, stat %d, vpos %d, " - "hpos %d\n", work->crtc_id, min_udelay, - vblank->framedur_ns / 1000, - vblank->linedur_ns / 1000, stat, vpos, hpos); + /* We borrow the event spin lock for protecting flip_status */ + spin_lock_irqsave(&crtc->dev->event_lock, flags); /* Do the flip (mmio) */ adev->mode_info.funcs->page_flip(adev, work->crtc_id, work->base, work->async); @@ -169,10 +138,10 @@ static void amdgpu_unpin_work_func(struct work_struct *__work) kfree(work); } -int amdgpu_crtc_page_flip(struct drm_crtc *crtc, - struct drm_framebuffer *fb, - struct drm_pending_vblank_event *event, - uint32_t page_flip_flags) +int amdgpu_crtc_page_flip_target(struct drm_crtc *crtc, + struct drm_framebuffer *fb, + struct drm_pending_vblank_event *event, + uint32_t page_flip_flags, uint32_t target) { struct drm_device *dev = crtc->dev; struct amdgpu_device *adev = dev->dev_private; @@ -191,7 +160,7 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc, if (work == NULL) return -ENOMEM; - INIT_WORK(&work->flip_work, amdgpu_flip_work_func); + INIT_DELAYED_WORK(&work->flip_work, amdgpu_flip_work_func); INIT_WORK(&work->unpin_work, amdgpu_unpin_work_func); work->event = event; @@ -237,12 +206,8 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc, amdgpu_bo_unreserve(new_rbo); work->base = base; - - r = drm_crtc_vblank_get(crtc); - if (r) { - DRM_ERROR("failed to get vblank before flip\n"); - goto pflip_cleanup; - } + work->target_vblank = target - drm_crtc_vblank_count(crtc) + + amdgpu_get_vblank_counter_kms(dev, work->crtc_id); /* we borrow the event spin lock for protecting flip_wrok */ spin_lock_irqsave(&crtc->dev->event_lock, flags); @@ -250,7 +215,7 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc, DRM_DEBUG_DRIVER("flip queue: crtc already busy\n"); spin_unlock_irqrestore(&crtc->dev->event_lock, flags); r = -EBUSY; - goto vblank_cleanup; + goto pflip_cleanup; } amdgpu_crtc->pflip_status = AMDGPU_FLIP_PENDING; @@ -262,12 +227,9 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc, /* update crtc fb */ crtc->primary->fb = fb; spin_unlock_irqrestore(&crtc->dev->event_lock, flags); - amdgpu_flip_work_func(&work->flip_work); + amdgpu_flip_work_func(&work->flip_work.work); return 0; -vblank_cleanup: - drm_crtc_vblank_put(crtc); - pflip_cleanup: if (unlikely(amdgpu_bo_reserve(new_rbo, false) != 0)) { DRM_ERROR("failed to reserve new rbo in error path\n"); @@ -335,7 +297,7 @@ int amdgpu_crtc_set_config(struct drm_mode_set *set) return ret; } -static const char *encoder_names[38] = { +static const char *encoder_names[41] = { "NONE", "INTERNAL_LVDS", "INTERNAL_TMDS1", @@ -374,6 +336,9 @@ static const char *encoder_names[38] = { "TRAVIS", "INTERNAL_VCE", "INTERNAL_UNIPHY3", + "HDMI_ANX9805", + "INTERNAL_AMCLK", + "VIRTUAL", }; static const char *hpd_names[6] = { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 11263c5b9967..7c911d0be2b3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -53,9 +53,11 @@ * - 3.2.0 - GFX8: Uses EOP_TC_WB_ACTION_EN, so UMDs don't have to do the same * at the end of IBs. * - 3.3.0 - Add VM support for UVD on supported hardware. + * - 3.4.0 - Add AMDGPU_INFO_NUM_EVICTIONS. + * - 3.5.0 - Add support for new UVD_NO_OP register. */ #define KMS_DRIVER_MAJOR 3 -#define KMS_DRIVER_MINOR 3 +#define KMS_DRIVER_MINOR 5 #define KMS_DRIVER_PATCHLEVEL 0 int amdgpu_vram_limit = 0; @@ -84,11 +86,13 @@ int amdgpu_sched_jobs = 32; int amdgpu_sched_hw_submission = 2; int amdgpu_powerplay = -1; int amdgpu_powercontainment = 1; +int amdgpu_sclk_deep_sleep_en = 1; unsigned amdgpu_pcie_gen_cap = 0; unsigned amdgpu_pcie_lane_cap = 0; unsigned amdgpu_cg_mask = 0xffffffff; unsigned amdgpu_pg_mask = 0xffffffff; char *amdgpu_disable_cu = NULL; +char *amdgpu_virtual_display = NULL; MODULE_PARM_DESC(vramlimit, "Restrict VRAM for testing, in megabytes"); module_param_named(vramlimit, amdgpu_vram_limit, int, 0600); @@ -170,6 +174,9 @@ MODULE_PARM_DESC(powercontainment, "Power Containment (1 = enable (default), 0 = module_param_named(powercontainment, amdgpu_powercontainment, int, 0444); #endif +MODULE_PARM_DESC(sclkdeepsleep, "SCLK Deep Sleep (1 = enable (default), 0 = disable)"); +module_param_named(sclkdeepsleep, amdgpu_sclk_deep_sleep_en, int, 0444); + MODULE_PARM_DESC(pcie_gen_cap, "PCIE Gen Caps (0: autodetect (default))"); module_param_named(pcie_gen_cap, amdgpu_pcie_gen_cap, uint, 0444); @@ -185,6 +192,9 @@ module_param_named(pg_mask, amdgpu_pg_mask, uint, 0444); MODULE_PARM_DESC(disable_cu, "Disable CUs (se.sh.cu,...)"); module_param_named(disable_cu, amdgpu_disable_cu, charp, 0444); +MODULE_PARM_DESC(virtual_display, "Enable virtual display feature (the virtual_display will be set like xxxx:xx:xx.x;xxxx:xx:xx.x)"); +module_param_named(virtual_display, amdgpu_virtual_display, charp, 0444); + static const struct pci_device_id pciidlist[] = { #ifdef CONFIG_DRM_AMDGPU_CIK /* Kaveri */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h index 503d54098128..e73728d90388 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h @@ -31,14 +31,6 @@ #define AMDGPU_GWS_SHIFT PAGE_SHIFT #define AMDGPU_OA_SHIFT PAGE_SHIFT -#define AMDGPU_PL_GDS TTM_PL_PRIV0 -#define AMDGPU_PL_GWS TTM_PL_PRIV1 -#define AMDGPU_PL_OA TTM_PL_PRIV2 - -#define AMDGPU_PL_FLAG_GDS TTM_PL_FLAG_PRIV0 -#define AMDGPU_PL_FLAG_GWS TTM_PL_FLAG_PRIV1 -#define AMDGPU_PL_FLAG_OA TTM_PL_FLAG_PRIV2 - struct amdgpu_ring; struct amdgpu_bo; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c index 31a676376d73..c93a92a840ea 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c @@ -186,10 +186,8 @@ struct amdgpu_i2c_chan *amdgpu_i2c_create(struct drm_device *dev, "AMDGPU i2c hw bus %s", name); i2c->adapter.algo = &amdgpu_atombios_i2c_algo; ret = i2c_add_adapter(&i2c->adapter); - if (ret) { - DRM_ERROR("Failed to register hw i2c %s\n", name); + if (ret) goto out_free; - } } else { /* set the amdgpu bit adapter */ snprintf(i2c->adapter.name, sizeof(i2c->adapter.name), diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index a31d7ef3032c..f5810f700668 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -142,7 +142,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, } if (!ring->ready) { - dev_err(adev->dev, "couldn't schedule ib\n"); + dev_err(adev->dev, "couldn't schedule ib on ring <%s>\n", ring->name); return -EINVAL; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c index 534fc04e80fd..5ebb3f43feb6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c @@ -40,32 +40,15 @@ static int amdgpu_ih_ring_alloc(struct amdgpu_device *adev) /* Allocate ring buffer */ if (adev->irq.ih.ring_obj == NULL) { - r = amdgpu_bo_create(adev, adev->irq.ih.ring_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, - NULL, NULL, &adev->irq.ih.ring_obj); + r = amdgpu_bo_create_kernel(adev, adev->irq.ih.ring_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + &adev->irq.ih.ring_obj, + &adev->irq.ih.gpu_addr, + (void **)&adev->irq.ih.ring); if (r) { DRM_ERROR("amdgpu: failed to create ih ring buffer (%d).\n", r); return r; } - r = amdgpu_bo_reserve(adev->irq.ih.ring_obj, false); - if (unlikely(r != 0)) - return r; - r = amdgpu_bo_pin(adev->irq.ih.ring_obj, - AMDGPU_GEM_DOMAIN_GTT, - &adev->irq.ih.gpu_addr); - if (r) { - amdgpu_bo_unreserve(adev->irq.ih.ring_obj); - DRM_ERROR("amdgpu: failed to pin ih ring buffer (%d).\n", r); - return r; - } - r = amdgpu_bo_kmap(adev->irq.ih.ring_obj, - (void **)&adev->irq.ih.ring); - amdgpu_bo_unreserve(adev->irq.ih.ring_obj); - if (r) { - DRM_ERROR("amdgpu: failed to map ih ring buffer (%d).\n", r); - return r; - } } return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h index 7ef09352e534..f016464035b8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h @@ -70,6 +70,7 @@ struct amdgpu_irq { /* gen irq stuff */ struct irq_domain *domain; /* GPU irq controller domain */ unsigned virq[AMDGPU_MAX_IRQ_SRC_ID]; + uint32_t srbm_soft_reset; }; void amdgpu_irq_preinstall(struct drm_device *dev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index d942654a1de0..b78e74048f3d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -292,14 +292,14 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file type = AMD_IP_BLOCK_TYPE_UVD; ring_mask = adev->uvd.ring.ready ? 1 : 0; ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 8; + ib_size_alignment = 16; break; case AMDGPU_HW_IP_VCE: type = AMD_IP_BLOCK_TYPE_VCE; for (i = 0; i < AMDGPU_MAX_VCE_RINGS; i++) ring_mask |= ((adev->vce.ring[i].ready ? 1 : 0) << i); ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 8; + ib_size_alignment = 1; break; default: return -EINVAL; @@ -373,6 +373,9 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file case AMDGPU_INFO_NUM_BYTES_MOVED: ui64 = atomic64_read(&adev->num_bytes_moved); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; + case AMDGPU_INFO_NUM_EVICTIONS: + ui64 = atomic64_read(&adev->num_evictions); + return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; case AMDGPU_INFO_VRAM_USAGE: ui64 = atomic64_read(&adev->vram_usage); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index 6b1d7d306564..7b0eff7d060b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -39,6 +39,8 @@ #include <drm/drm_plane_helper.h> #include <linux/i2c.h> #include <linux/i2c-algo-bit.h> +#include <linux/hrtimer.h> +#include "amdgpu_irq.h" struct amdgpu_bo; struct amdgpu_device; @@ -339,6 +341,8 @@ struct amdgpu_mode_info { int num_dig; /* number of dig blocks */ int disp_priority; const struct amdgpu_display_funcs *funcs; + struct hrtimer vblank_timer; + enum amdgpu_interrupt_state vsync_timer_enabled; }; #define AMDGPU_MAX_BL_LEVEL 0xFF @@ -587,10 +591,10 @@ int amdgpu_align_pitch(struct amdgpu_device *adev, int width, int bpp, bool tile void amdgpu_print_display_setup(struct drm_device *dev); int amdgpu_modeset_create_props(struct amdgpu_device *adev); int amdgpu_crtc_set_config(struct drm_mode_set *set); -int amdgpu_crtc_page_flip(struct drm_crtc *crtc, - struct drm_framebuffer *fb, - struct drm_pending_vblank_event *event, - uint32_t page_flip_flags); +int amdgpu_crtc_page_flip_target(struct drm_crtc *crtc, + struct drm_framebuffer *fb, + struct drm_pending_vblank_event *event, + uint32_t page_flip_flags, uint32_t target); extern const struct drm_mode_config_funcs amdgpu_mode_funcs; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 6f0873c75a25..b17734e0ecc8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -44,14 +44,13 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev); static u64 amdgpu_get_vis_part_size(struct amdgpu_device *adev, struct ttm_mem_reg *mem) { - u64 ret = 0; - if (mem->start << PAGE_SHIFT < adev->mc.visible_vram_size) { - ret = (u64)((mem->start << PAGE_SHIFT) + mem->size) > - adev->mc.visible_vram_size ? - adev->mc.visible_vram_size - (mem->start << PAGE_SHIFT) : - mem->size; - } - return ret; + if (mem->start << PAGE_SHIFT >= adev->mc.visible_vram_size) + return 0; + + return ((mem->start << PAGE_SHIFT) + mem->size) > + adev->mc.visible_vram_size ? + adev->mc.visible_vram_size - (mem->start << PAGE_SHIFT) : + mem->size; } static void amdgpu_update_memory_usage(struct amdgpu_device *adev, @@ -99,6 +98,11 @@ static void amdgpu_ttm_bo_destroy(struct ttm_buffer_object *tbo) drm_gem_object_release(&bo->gem_base); amdgpu_bo_unref(&bo->parent); + if (!list_empty(&bo->shadow_list)) { + mutex_lock(&bo->adev->shadow_list_lock); + list_del_init(&bo->shadow_list); + mutex_unlock(&bo->adev->shadow_list_lock); + } kfree(bo->metadata); kfree(bo); } @@ -112,84 +116,93 @@ bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo) static void amdgpu_ttm_placement_init(struct amdgpu_device *adev, struct ttm_placement *placement, - struct ttm_place *placements, + struct ttm_place *places, u32 domain, u64 flags) { - u32 c = 0, i; - - placement->placement = placements; - placement->busy_placement = placements; + u32 c = 0; if (domain & AMDGPU_GEM_DOMAIN_VRAM) { + unsigned visible_pfn = adev->mc.visible_vram_size >> PAGE_SHIFT; + if (flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS && - adev->mc.visible_vram_size < adev->mc.real_vram_size) { - placements[c].fpfn = - adev->mc.visible_vram_size >> PAGE_SHIFT; - placements[c++].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | - TTM_PL_FLAG_VRAM | TTM_PL_FLAG_TOPDOWN; + !(flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) && + adev->mc.visible_vram_size < adev->mc.real_vram_size) { + places[c].fpfn = visible_pfn; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_WC | + TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_VRAM | + TTM_PL_FLAG_TOPDOWN; + c++; } - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | + + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_VRAM; - if (!(flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)) - placements[c - 1].flags |= TTM_PL_FLAG_TOPDOWN; + if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) + places[c].lpfn = visible_pfn; + else + places[c].flags |= TTM_PL_FLAG_TOPDOWN; + c++; } if (domain & AMDGPU_GEM_DOMAIN_GTT) { - if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_TT | + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_TT; + if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) + places[c].flags |= TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED; - } else { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_TT; - } + else + places[c].flags |= TTM_PL_FLAG_CACHED; + c++; } if (domain & AMDGPU_GEM_DOMAIN_CPU) { - if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_SYSTEM | + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_SYSTEM; + if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) + places[c].flags |= TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED; - } else { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_SYSTEM; - } + else + places[c].flags |= TTM_PL_FLAG_CACHED; + c++; } if (domain & AMDGPU_GEM_DOMAIN_GDS) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_UNCACHED | - AMDGPU_PL_FLAG_GDS; + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_UNCACHED | AMDGPU_PL_FLAG_GDS; + c++; } + if (domain & AMDGPU_GEM_DOMAIN_GWS) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_UNCACHED | - AMDGPU_PL_FLAG_GWS; + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_UNCACHED | AMDGPU_PL_FLAG_GWS; + c++; } + if (domain & AMDGPU_GEM_DOMAIN_OA) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_UNCACHED | - AMDGPU_PL_FLAG_OA; + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_UNCACHED | AMDGPU_PL_FLAG_OA; + c++; } if (!c) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_MASK_CACHING | - TTM_PL_FLAG_SYSTEM; + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM; + c++; } + placement->num_placement = c; - placement->num_busy_placement = c; + placement->placement = places; - for (i = 0; i < c; i++) { - if ((flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) && - (placements[i].flags & TTM_PL_FLAG_VRAM) && - !placements[i].fpfn) - placements[i].lpfn = - adev->mc.visible_vram_size >> PAGE_SHIFT; - else - placements[i].lpfn = 0; - } + placement->num_busy_placement = c; + placement->busy_placement = places; } void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *rbo, u32 domain) @@ -211,6 +224,69 @@ static void amdgpu_fill_placement_to_bo(struct amdgpu_bo *bo, bo->placement.busy_placement = bo->placements; } +/** + * amdgpu_bo_create_kernel - create BO for kernel use + * + * @adev: amdgpu device object + * @size: size for the new BO + * @align: alignment for the new BO + * @domain: where to place it + * @bo_ptr: resulting BO + * @gpu_addr: GPU addr of the pinned BO + * @cpu_addr: optional CPU address mapping + * + * Allocates and pins a BO for kernel internal use. + * + * Returns 0 on success, negative error code otherwise. + */ +int amdgpu_bo_create_kernel(struct amdgpu_device *adev, + unsigned long size, int align, + u32 domain, struct amdgpu_bo **bo_ptr, + u64 *gpu_addr, void **cpu_addr) +{ + int r; + + r = amdgpu_bo_create(adev, size, align, true, domain, + AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED, + NULL, NULL, bo_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to allocate kernel bo\n", r); + return r; + } + + r = amdgpu_bo_reserve(*bo_ptr, false); + if (r) { + dev_err(adev->dev, "(%d) failed to reserve kernel bo\n", r); + goto error_free; + } + + r = amdgpu_bo_pin(*bo_ptr, domain, gpu_addr); + if (r) { + dev_err(adev->dev, "(%d) kernel bo pin failed\n", r); + goto error_unreserve; + } + + if (cpu_addr) { + r = amdgpu_bo_kmap(*bo_ptr, cpu_addr); + if (r) { + dev_err(adev->dev, "(%d) kernel bo map failed\n", r); + goto error_unreserve; + } + } + + amdgpu_bo_unreserve(*bo_ptr); + + return 0; + +error_unreserve: + amdgpu_bo_unreserve(*bo_ptr); + +error_free: + amdgpu_bo_unref(bo_ptr); + + return r; +} + int amdgpu_bo_create_restricted(struct amdgpu_device *adev, unsigned long size, int byte_align, bool kernel, u32 domain, u64 flags, @@ -250,6 +326,7 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev, } bo->adev = adev; INIT_LIST_HEAD(&bo->list); + INIT_LIST_HEAD(&bo->shadow_list); INIT_LIST_HEAD(&bo->va); bo->prefered_domains = domain & (AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT | @@ -277,11 +354,79 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev, if (unlikely(r != 0)) { return r; } + + if (flags & AMDGPU_GEM_CREATE_VRAM_CLEARED && + bo->tbo.mem.placement & TTM_PL_FLAG_VRAM) { + struct fence *fence; + + if (adev->mman.buffer_funcs_ring == NULL || + !adev->mman.buffer_funcs_ring->ready) { + r = -EBUSY; + goto fail_free; + } + + r = amdgpu_bo_reserve(bo, false); + if (unlikely(r != 0)) + goto fail_free; + + amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_VRAM); + r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + if (unlikely(r != 0)) + goto fail_unreserve; + + amdgpu_fill_buffer(bo, 0, bo->tbo.resv, &fence); + amdgpu_bo_fence(bo, fence, false); + amdgpu_bo_unreserve(bo); + fence_put(bo->tbo.moving); + bo->tbo.moving = fence_get(fence); + fence_put(fence); + } *bo_ptr = bo; trace_amdgpu_bo_create(bo); return 0; + +fail_unreserve: + amdgpu_bo_unreserve(bo); +fail_free: + amdgpu_bo_unref(&bo); + return r; +} + +static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, + unsigned long size, int byte_align, + struct amdgpu_bo *bo) +{ + struct ttm_placement placement = {0}; + struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1]; + int r; + + if (bo->shadow) + return 0; + + bo->flags |= AMDGPU_GEM_CREATE_SHADOW; + memset(&placements, 0, + (AMDGPU_GEM_DOMAIN_MAX + 1) * sizeof(struct ttm_place)); + + amdgpu_ttm_placement_init(adev, &placement, + placements, AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_CREATE_CPU_GTT_USWC); + + r = amdgpu_bo_create_restricted(adev, size, byte_align, true, + AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_CREATE_CPU_GTT_USWC, + NULL, &placement, + bo->tbo.resv, + &bo->shadow); + if (!r) { + bo->shadow->parent = amdgpu_bo_ref(bo); + mutex_lock(&adev->shadow_list_lock); + list_add_tail(&bo->shadow_list, &adev->shadow_list); + mutex_unlock(&adev->shadow_list_lock); + } + + return r; } int amdgpu_bo_create(struct amdgpu_device *adev, @@ -293,6 +438,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev, { struct ttm_placement placement = {0}; struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1]; + int r; memset(&placements, 0, (AMDGPU_GEM_DOMAIN_MAX + 1) * sizeof(struct ttm_place)); @@ -300,9 +446,83 @@ int amdgpu_bo_create(struct amdgpu_device *adev, amdgpu_ttm_placement_init(adev, &placement, placements, domain, flags); - return amdgpu_bo_create_restricted(adev, size, byte_align, kernel, - domain, flags, sg, &placement, - resv, bo_ptr); + r = amdgpu_bo_create_restricted(adev, size, byte_align, kernel, + domain, flags, sg, &placement, + resv, bo_ptr); + if (r) + return r; + + if (amdgpu_need_backup(adev) && (flags & AMDGPU_GEM_CREATE_SHADOW)) { + r = amdgpu_bo_create_shadow(adev, size, byte_align, (*bo_ptr)); + if (r) + amdgpu_bo_unref(bo_ptr); + } + + return r; +} + +int amdgpu_bo_backup_to_shadow(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_bo *bo, + struct reservation_object *resv, + struct fence **fence, + bool direct) + +{ + struct amdgpu_bo *shadow = bo->shadow; + uint64_t bo_addr, shadow_addr; + int r; + + if (!shadow) + return -EINVAL; + + bo_addr = amdgpu_bo_gpu_offset(bo); + shadow_addr = amdgpu_bo_gpu_offset(bo->shadow); + + r = reservation_object_reserve_shared(bo->tbo.resv); + if (r) + goto err; + + r = amdgpu_copy_buffer(ring, bo_addr, shadow_addr, + amdgpu_bo_size(bo), resv, fence, + direct); + if (!r) + amdgpu_bo_fence(bo, *fence, true); + +err: + return r; +} + +int amdgpu_bo_restore_from_shadow(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_bo *bo, + struct reservation_object *resv, + struct fence **fence, + bool direct) + +{ + struct amdgpu_bo *shadow = bo->shadow; + uint64_t bo_addr, shadow_addr; + int r; + + if (!shadow) + return -EINVAL; + + bo_addr = amdgpu_bo_gpu_offset(bo); + shadow_addr = amdgpu_bo_gpu_offset(bo->shadow); + + r = reservation_object_reserve_shared(bo->tbo.resv); + if (r) + goto err; + + r = amdgpu_copy_buffer(ring, shadow_addr, bo_addr, + amdgpu_bo_size(bo), resv, fence, + direct); + if (!r) + amdgpu_bo_fence(bo, *fence, true); + +err: + return r; } int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr) @@ -380,16 +600,17 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, return -EINVAL; if (bo->pin_count) { + uint32_t mem_type = bo->tbo.mem.mem_type; + + if (domain != amdgpu_mem_type_to_domain(mem_type)) + return -EINVAL; + bo->pin_count++; if (gpu_addr) *gpu_addr = amdgpu_bo_gpu_offset(bo); if (max_offset != 0) { - u64 domain_start; - if (domain == AMDGPU_GEM_DOMAIN_VRAM) - domain_start = bo->adev->mc.vram_start; - else - domain_start = bo->adev->mc.gtt_start; + u64 domain_start = bo->tbo.bdev->man[mem_type].gpu_offset; WARN_ON_ONCE(max_offset < (amdgpu_bo_gpu_offset(bo) - domain_start)); } @@ -401,7 +622,8 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, /* force to pin into visible video ram */ if ((bo->placements[i].flags & TTM_PL_FLAG_VRAM) && !(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) && - (!max_offset || max_offset > bo->adev->mc.visible_vram_size)) { + (!max_offset || max_offset > + bo->adev->mc.visible_vram_size)) { if (WARN_ON_ONCE(min_offset > bo->adev->mc.visible_vram_size)) return -EINVAL; @@ -420,19 +642,23 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, } r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); - if (likely(r == 0)) { - bo->pin_count = 1; - if (gpu_addr != NULL) - *gpu_addr = amdgpu_bo_gpu_offset(bo); - if (domain == AMDGPU_GEM_DOMAIN_VRAM) { - bo->adev->vram_pin_size += amdgpu_bo_size(bo); - if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) - bo->adev->invisible_pin_size += amdgpu_bo_size(bo); - } else - bo->adev->gart_pin_size += amdgpu_bo_size(bo); - } else { + if (unlikely(r)) { dev_err(bo->adev->dev, "%p pin failed\n", bo); + goto error; + } + + bo->pin_count = 1; + if (gpu_addr != NULL) + *gpu_addr = amdgpu_bo_gpu_offset(bo); + if (domain == AMDGPU_GEM_DOMAIN_VRAM) { + bo->adev->vram_pin_size += amdgpu_bo_size(bo); + if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) + bo->adev->invisible_pin_size += amdgpu_bo_size(bo); + } else if (domain == AMDGPU_GEM_DOMAIN_GTT) { + bo->adev->gart_pin_size += amdgpu_bo_size(bo); } + +error: return r; } @@ -457,16 +683,20 @@ int amdgpu_bo_unpin(struct amdgpu_bo *bo) bo->placements[i].flags &= ~TTM_PL_FLAG_NO_EVICT; } r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); - if (likely(r == 0)) { - if (bo->tbo.mem.mem_type == TTM_PL_VRAM) { - bo->adev->vram_pin_size -= amdgpu_bo_size(bo); - if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) - bo->adev->invisible_pin_size -= amdgpu_bo_size(bo); - } else - bo->adev->gart_pin_size -= amdgpu_bo_size(bo); - } else { + if (unlikely(r)) { dev_err(bo->adev->dev, "%p validate failed for unpin\n", bo); + goto error; + } + + if (bo->tbo.mem.mem_type == TTM_PL_VRAM) { + bo->adev->vram_pin_size -= amdgpu_bo_size(bo); + if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) + bo->adev->invisible_pin_size -= amdgpu_bo_size(bo); + } else { + bo->adev->gart_pin_size -= amdgpu_bo_size(bo); } + +error: return r; } @@ -637,7 +867,8 @@ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) for (i = 0; i < abo->placement.num_placement; i++) { /* Force into visible VRAM */ if ((abo->placements[i].flags & TTM_PL_FLAG_VRAM) && - (!abo->placements[i].lpfn || abo->placements[i].lpfn > lpfn)) + (!abo->placements[i].lpfn || + abo->placements[i].lpfn > lpfn)) abo->placements[i].lpfn = lpfn; } r = ttm_bo_validate(bo, &abo->placement, false, false); @@ -674,3 +905,21 @@ void amdgpu_bo_fence(struct amdgpu_bo *bo, struct fence *fence, else reservation_object_add_excl_fence(resv, fence); } + +/** + * amdgpu_bo_gpu_offset - return GPU offset of bo + * @bo: amdgpu object for which we query the offset + * + * Returns current GPU offset of the object. + * + * Note: object should either be pinned or reserved when calling this + * function, it might be useful to add check for this for debugging. + */ +u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo) +{ + WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_SYSTEM); + WARN_ON_ONCE(!ww_mutex_is_locked(&bo->tbo.resv->lock) && + !bo->pin_count); + + return bo->tbo.offset; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index bdb01d932548..b6a27390ef88 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -85,21 +85,6 @@ static inline void amdgpu_bo_unreserve(struct amdgpu_bo *bo) ttm_bo_unreserve(&bo->tbo); } -/** - * amdgpu_bo_gpu_offset - return GPU offset of bo - * @bo: amdgpu object for which we query the offset - * - * Returns current GPU offset of the object. - * - * Note: object should either be pinned or reserved when calling this - * function, it might be useful to add check for this for debugging. - */ -static inline u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo) -{ - WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_SYSTEM); - return bo->tbo.offset; -} - static inline unsigned long amdgpu_bo_size(struct amdgpu_bo *bo) { return bo->tbo.num_pages << PAGE_SHIFT; @@ -139,6 +124,10 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev, struct ttm_placement *placement, struct reservation_object *resv, struct amdgpu_bo **bo_ptr); +int amdgpu_bo_create_kernel(struct amdgpu_device *adev, + unsigned long size, int align, + u32 domain, struct amdgpu_bo **bo_ptr, + u64 *gpu_addr, void **cpu_addr); int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr); void amdgpu_bo_kunmap(struct amdgpu_bo *bo); struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo); @@ -165,6 +154,19 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo); void amdgpu_bo_fence(struct amdgpu_bo *bo, struct fence *fence, bool shared); +u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo); +int amdgpu_bo_backup_to_shadow(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_bo *bo, + struct reservation_object *resv, + struct fence **fence, bool direct); +int amdgpu_bo_restore_from_shadow(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_bo *bo, + struct reservation_object *resv, + struct fence **fence, + bool direct); + /* * sub allocation diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c index 5cc7052e391d..d4ec3cb187a5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c @@ -1103,54 +1103,46 @@ force: void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable) { - if (adev->pp_enabled) + if (adev->pp_enabled || adev->pm.funcs->powergate_uvd) { + /* enable/disable UVD */ + mutex_lock(&adev->pm.mutex); amdgpu_dpm_powergate_uvd(adev, !enable); - else { - if (adev->pm.funcs->powergate_uvd) { + mutex_unlock(&adev->pm.mutex); + } else { + if (enable) { mutex_lock(&adev->pm.mutex); - /* enable/disable UVD */ - amdgpu_dpm_powergate_uvd(adev, !enable); + adev->pm.dpm.uvd_active = true; + adev->pm.dpm.state = POWER_STATE_TYPE_INTERNAL_UVD; mutex_unlock(&adev->pm.mutex); } else { - if (enable) { - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.uvd_active = true; - adev->pm.dpm.state = POWER_STATE_TYPE_INTERNAL_UVD; - mutex_unlock(&adev->pm.mutex); - } else { - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.uvd_active = false; - mutex_unlock(&adev->pm.mutex); - } - amdgpu_pm_compute_clocks(adev); + mutex_lock(&adev->pm.mutex); + adev->pm.dpm.uvd_active = false; + mutex_unlock(&adev->pm.mutex); } - + amdgpu_pm_compute_clocks(adev); } } void amdgpu_dpm_enable_vce(struct amdgpu_device *adev, bool enable) { - if (adev->pp_enabled) + if (adev->pp_enabled || adev->pm.funcs->powergate_vce) { + /* enable/disable VCE */ + mutex_lock(&adev->pm.mutex); amdgpu_dpm_powergate_vce(adev, !enable); - else { - if (adev->pm.funcs->powergate_vce) { + mutex_unlock(&adev->pm.mutex); + } else { + if (enable) { mutex_lock(&adev->pm.mutex); - amdgpu_dpm_powergate_vce(adev, !enable); + adev->pm.dpm.vce_active = true; + /* XXX select vce level based on ring/task */ + adev->pm.dpm.vce_level = AMDGPU_VCE_LEVEL_AC_ALL; mutex_unlock(&adev->pm.mutex); } else { - if (enable) { - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.vce_active = true; - /* XXX select vce level based on ring/task */ - adev->pm.dpm.vce_level = AMDGPU_VCE_LEVEL_AC_ALL; - mutex_unlock(&adev->pm.mutex); - } else { - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.vce_active = false; - mutex_unlock(&adev->pm.mutex); - } - amdgpu_pm_compute_clocks(adev); + mutex_lock(&adev->pm.mutex); + adev->pm.dpm.vce_active = false; + mutex_unlock(&adev->pm.mutex); } + amdgpu_pm_compute_clocks(adev); } } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c index c5738a22b690..545074479e1f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c @@ -52,7 +52,9 @@ static int amdgpu_powerplay_init(struct amdgpu_device *adev) pp_init->chip_family = adev->family; pp_init->chip_id = adev->asic_type; pp_init->device = amdgpu_cgs_create_device(adev); - pp_init->powercontainment_enabled = amdgpu_powercontainment; + pp_init->rev_id = adev->pdev->revision; + pp_init->sub_sys_id = adev->pdev->subsystem_device; + pp_init->sub_vendor_id = adev->pdev->subsystem_vendor; ret = amd_powerplay_init(pp_init, amd_pp); kfree(pp_init); @@ -106,11 +108,10 @@ static int amdgpu_pp_early_init(void *handle) break; case CHIP_TONGA: case CHIP_FIJI: - adev->pp_enabled = (amdgpu_powerplay == 0) ? false : true; - break; + case CHIP_TOPAZ: case CHIP_CARRIZO: case CHIP_STONEY: - adev->pp_enabled = (amdgpu_powerplay > 0) ? true : false; + adev->pp_enabled = (amdgpu_powerplay == 0) ? false : true; break; /* These chips don't have powerplay implemenations */ case CHIP_BONAIRE: @@ -118,7 +119,6 @@ static int amdgpu_pp_early_init(void *handle) case CHIP_KABINI: case CHIP_MULLINS: case CHIP_KAVERI: - case CHIP_TOPAZ: default: adev->pp_enabled = false; break; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 85aeb0a804bb..242ba04bfde6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -222,33 +222,16 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, /* Allocate ring buffer */ if (ring->ring_obj == NULL) { - r = amdgpu_bo_create(adev, ring->ring_size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, - NULL, NULL, &ring->ring_obj); + r = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &ring->ring_obj, + &ring->gpu_addr, + (void **)&ring->ring); if (r) { dev_err(adev->dev, "(%d) ring create failed\n", r); return r; } - r = amdgpu_bo_reserve(ring->ring_obj, false); - if (unlikely(r != 0)) - return r; - r = amdgpu_bo_pin(ring->ring_obj, AMDGPU_GEM_DOMAIN_GTT, - &ring->gpu_addr); - if (r) { - amdgpu_bo_unreserve(ring->ring_obj); - dev_err(adev->dev, "(%d) ring pin failed\n", r); - return r; - } - r = amdgpu_bo_kmap(ring->ring_obj, - (void **)&ring->ring); - memset((void *)ring->ring, 0, ring->ring_size); - - amdgpu_bo_unreserve(ring->ring_obj); - if (r) { - dev_err(adev->dev, "(%d) ring map failed\n", r); - return r; - } } ring->ptr_mask = (ring->ring_size / 4) - 1; ring->max_dw = max_dw; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c index 05a53f4fc334..b827c75e95de 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c @@ -111,7 +111,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) amdgpu_bo_kunmap(gtt_obj[i]); r = amdgpu_copy_buffer(ring, gtt_addr, vram_addr, - size, NULL, &fence); + size, NULL, &fence, false); if (r) { DRM_ERROR("Failed GTT->VRAM copy %d\n", i); @@ -156,7 +156,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) amdgpu_bo_kunmap(vram_obj); r = amdgpu_copy_buffer(ring, vram_addr, gtt_addr, - size, NULL, &fence); + size, NULL, &fence, false); if (r) { DRM_ERROR("Failed VRAM->GTT copy %d\n", i); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 9b61c8ba7aaf..5447973483ec 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -34,6 +34,7 @@ #include <ttm/ttm_placement.h> #include <ttm/ttm_module.h> #include <ttm/ttm_page_alloc.h> +#include <ttm/ttm_memory.h> #include <drm/drmP.h> #include <drm/amdgpu_drm.h> #include <linux/seq_file.h> @@ -74,7 +75,7 @@ static void amdgpu_ttm_mem_global_release(struct drm_global_reference *ref) ttm_mem_global_release(ref->object); } -static int amdgpu_ttm_global_init(struct amdgpu_device *adev) +int amdgpu_ttm_global_init(struct amdgpu_device *adev) { struct drm_global_reference *global_ref; struct amdgpu_ring *ring; @@ -256,10 +257,8 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo, switch (old_mem->mem_type) { case TTM_PL_VRAM: - old_start += adev->mc.vram_start; - break; case TTM_PL_TT: - old_start += adev->mc.gtt_start; + old_start += bo->bdev->man[old_mem->mem_type].gpu_offset; break; default: DRM_ERROR("Unknown placement %d\n", old_mem->mem_type); @@ -267,10 +266,8 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo, } switch (new_mem->mem_type) { case TTM_PL_VRAM: - new_start += adev->mc.vram_start; - break; case TTM_PL_TT: - new_start += adev->mc.gtt_start; + new_start += bo->bdev->man[new_mem->mem_type].gpu_offset; break; default: DRM_ERROR("Unknown placement %d\n", old_mem->mem_type); @@ -285,7 +282,7 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo, r = amdgpu_copy_buffer(ring, old_start, new_start, new_mem->num_pages * PAGE_SIZE, /* bytes */ - bo->resv, &fence); + bo->resv, &fence, false); if (r) return r; @@ -335,7 +332,7 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, if (unlikely(r)) { goto out_cleanup; } - r = ttm_bo_move_ttm(bo, true, interruptible, no_wait_gpu, new_mem); + r = ttm_bo_move_ttm(bo, interruptible, no_wait_gpu, new_mem); out_cleanup: ttm_bo_mem_put(bo, &tmp_mem); return r; @@ -368,7 +365,7 @@ static int amdgpu_move_ram_vram(struct ttm_buffer_object *bo, if (unlikely(r)) { return r; } - r = ttm_bo_move_ttm(bo, true, interruptible, no_wait_gpu, &tmp_mem); + r = ttm_bo_move_ttm(bo, interruptible, no_wait_gpu, &tmp_mem); if (unlikely(r)) { goto out_cleanup; } @@ -435,8 +432,7 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, if (r) { memcpy: - r = ttm_bo_move_memcpy(bo, evict, interruptible, - no_wait_gpu, new_mem); + r = ttm_bo_move_memcpy(bo, interruptible, no_wait_gpu, new_mem); if (r) { return r; } @@ -950,6 +946,8 @@ static struct list_head *amdgpu_ttm_lru_tail(struct ttm_buffer_object *tbo) struct list_head *res = lru->lru[tbo->mem.mem_type]; lru->lru[tbo->mem.mem_type] = &tbo->lru; + while ((++lru)->lru[tbo->mem.mem_type] == res) + lru->lru[tbo->mem.mem_type] = &tbo->lru; return res; } @@ -960,6 +958,8 @@ static struct list_head *amdgpu_ttm_swap_lru_tail(struct ttm_buffer_object *tbo) struct list_head *res = lru->swap_lru; lru->swap_lru = &tbo->swap; + while ((++lru)->swap_lru == res) + lru->swap_lru = &tbo->swap; return res; } @@ -987,10 +987,6 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) unsigned i, j; int r; - r = amdgpu_ttm_global_init(adev); - if (r) { - return r; - } /* No others user of address space so set it to 0 */ r = ttm_bo_device_init(&adev->mman.bdev, adev->mman.bo_global_ref.ref.object, @@ -1011,6 +1007,10 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) lru->swap_lru = &adev->mman.bdev.glob->swap_lru; } + for (j = 0; j < TTM_NUM_MEM_TYPES; ++j) + adev->mman.guard.lru[j] = NULL; + adev->mman.guard.swap_lru = NULL; + adev->mman.initialized = true; r = ttm_bo_init_mm(&adev->mman.bdev, TTM_PL_VRAM, adev->mc.real_vram_size >> PAGE_SHIFT); @@ -1151,7 +1151,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t dst_offset, uint32_t byte_count, struct reservation_object *resv, - struct fence **fence) + struct fence **fence, bool direct_submit) { struct amdgpu_device *adev = ring->adev; struct amdgpu_job *job; @@ -1195,8 +1195,79 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, amdgpu_ring_pad_ib(ring, &job->ibs[0]); WARN_ON(job->ibs[0].length_dw > num_dw); + if (direct_submit) { + r = amdgpu_ib_schedule(ring, job->num_ibs, job->ibs, + NULL, NULL, fence); + job->fence = fence_get(*fence); + if (r) + DRM_ERROR("Error scheduling IBs (%d)\n", r); + amdgpu_job_free(job); + } else { + r = amdgpu_job_submit(job, ring, &adev->mman.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, fence); + if (r) + goto error_free; + } + + return r; + +error_free: + amdgpu_job_free(job); + return r; +} + +int amdgpu_fill_buffer(struct amdgpu_bo *bo, + uint32_t src_data, + struct reservation_object *resv, + struct fence **fence) +{ + struct amdgpu_device *adev = bo->adev; + struct amdgpu_job *job; + struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; + + uint32_t max_bytes, byte_count; + uint64_t dst_offset; + unsigned int num_loops, num_dw; + unsigned int i; + int r; + + byte_count = bo->tbo.num_pages << PAGE_SHIFT; + max_bytes = adev->mman.buffer_funcs->fill_max_bytes; + num_loops = DIV_ROUND_UP(byte_count, max_bytes); + num_dw = num_loops * adev->mman.buffer_funcs->fill_num_dw; + + /* for IB padding */ + while (num_dw & 0x7) + num_dw++; + + r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, &job); + if (r) + return r; + + if (resv) { + r = amdgpu_sync_resv(adev, &job->sync, resv, + AMDGPU_FENCE_OWNER_UNDEFINED); + if (r) { + DRM_ERROR("sync failed (%d).\n", r); + goto error_free; + } + } + + dst_offset = bo->tbo.mem.start << PAGE_SHIFT; + for (i = 0; i < num_loops; i++) { + uint32_t cur_size_in_bytes = min(byte_count, max_bytes); + + amdgpu_emit_fill_buffer(adev, &job->ibs[0], src_data, + dst_offset, cur_size_in_bytes); + + dst_offset += cur_size_in_bytes; + byte_count -= cur_size_in_bytes; + } + + amdgpu_ring_pad_ib(ring, &job->ibs[0]); + WARN_ON(job->ibs[0].length_dw > num_dw); r = amdgpu_job_submit(job, ring, &adev->mman.entity, - AMDGPU_FENCE_OWNER_UNDEFINED, fence); + AMDGPU_FENCE_OWNER_UNDEFINED, fence); if (r) goto error_free; @@ -1387,3 +1458,8 @@ static void amdgpu_ttm_debugfs_fini(struct amdgpu_device *adev) #endif } + +u64 amdgpu_ttm_get_gtt_mem_size(struct amdgpu_device *adev) +{ + return ttm_get_kernel_zone_memory_size(adev->mman.mem_global_ref.object); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h new file mode 100644 index 000000000000..72f6bfc15d8f --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -0,0 +1,80 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __AMDGPU_TTM_H__ +#define __AMDGPU_TTM_H__ + +#include "gpu_scheduler.h" + +#define AMDGPU_PL_GDS TTM_PL_PRIV0 +#define AMDGPU_PL_GWS TTM_PL_PRIV1 +#define AMDGPU_PL_OA TTM_PL_PRIV2 + +#define AMDGPU_PL_FLAG_GDS TTM_PL_FLAG_PRIV0 +#define AMDGPU_PL_FLAG_GWS TTM_PL_FLAG_PRIV1 +#define AMDGPU_PL_FLAG_OA TTM_PL_FLAG_PRIV2 + +#define AMDGPU_TTM_LRU_SIZE 20 + +struct amdgpu_mman_lru { + struct list_head *lru[TTM_NUM_MEM_TYPES]; + struct list_head *swap_lru; +}; + +struct amdgpu_mman { + struct ttm_bo_global_ref bo_global_ref; + struct drm_global_reference mem_global_ref; + struct ttm_bo_device bdev; + bool mem_global_referenced; + bool initialized; + +#if defined(CONFIG_DEBUG_FS) + struct dentry *vram; + struct dentry *gtt; +#endif + + /* buffer handling */ + const struct amdgpu_buffer_funcs *buffer_funcs; + struct amdgpu_ring *buffer_funcs_ring; + /* Scheduler entity for buffer moves */ + struct amd_sched_entity entity; + + /* custom LRU management */ + struct amdgpu_mman_lru log2_size[AMDGPU_TTM_LRU_SIZE]; + /* guard for log2_size array, don't add anything in between */ + struct amdgpu_mman_lru guard; +}; + +int amdgpu_copy_buffer(struct amdgpu_ring *ring, + uint64_t src_offset, + uint64_t dst_offset, + uint32_t byte_count, + struct reservation_object *resv, + struct fence **fence, bool direct_submit); +int amdgpu_fill_buffer(struct amdgpu_bo *bo, + uint32_t src_data, + struct reservation_object *resv, + struct fence **fence); + +int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma); +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index b11f4e8868d7..cc766cc53a87 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -201,39 +201,14 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) bo_size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8) + AMDGPU_UVD_STACK_SIZE + AMDGPU_UVD_HEAP_SIZE + AMDGPU_UVD_SESSION_SIZE * adev->uvd.max_handles; - r = amdgpu_bo_create(adev, bo_size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED, - NULL, NULL, &adev->uvd.vcpu_bo); + r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, &adev->uvd.vcpu_bo, + &adev->uvd.gpu_addr, &adev->uvd.cpu_addr); if (r) { dev_err(adev->dev, "(%d) failed to allocate UVD bo\n", r); return r; } - r = amdgpu_bo_reserve(adev->uvd.vcpu_bo, false); - if (r) { - amdgpu_bo_unref(&adev->uvd.vcpu_bo); - dev_err(adev->dev, "(%d) failed to reserve UVD bo\n", r); - return r; - } - - r = amdgpu_bo_pin(adev->uvd.vcpu_bo, AMDGPU_GEM_DOMAIN_VRAM, - &adev->uvd.gpu_addr); - if (r) { - amdgpu_bo_unreserve(adev->uvd.vcpu_bo); - amdgpu_bo_unref(&adev->uvd.vcpu_bo); - dev_err(adev->dev, "(%d) UVD bo pin failed\n", r); - return r; - } - - r = amdgpu_bo_kmap(adev->uvd.vcpu_bo, &adev->uvd.cpu_addr); - if (r) { - dev_err(adev->dev, "(%d) UVD map failed\n", r); - return r; - } - - amdgpu_bo_unreserve(adev->uvd.vcpu_bo); - ring = &adev->uvd.ring; rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL]; r = amd_sched_entity_init(&ring->sched, &adev->uvd.entity, @@ -323,7 +298,7 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev) if (!adev->uvd.saved_bo) return -ENOMEM; - memcpy(adev->uvd.saved_bo, ptr, size); + memcpy_fromio(adev->uvd.saved_bo, ptr, size); return 0; } @@ -340,7 +315,7 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev) ptr = adev->uvd.cpu_addr; if (adev->uvd.saved_bo != NULL) { - memcpy(ptr, adev->uvd.saved_bo, size); + memcpy_toio(ptr, adev->uvd.saved_bo, size); kfree(adev->uvd.saved_bo); adev->uvd.saved_bo = NULL; } else { @@ -349,11 +324,11 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev) hdr = (const struct common_firmware_header *)adev->uvd.fw->data; offset = le32_to_cpu(hdr->ucode_array_offset_bytes); - memcpy(adev->uvd.cpu_addr, (adev->uvd.fw->data) + offset, - (adev->uvd.fw->size) - offset); + memcpy_toio(adev->uvd.cpu_addr, adev->uvd.fw->data + offset, + le32_to_cpu(hdr->ucode_size_bytes)); size -= le32_to_cpu(hdr->ucode_size_bytes); ptr += le32_to_cpu(hdr->ucode_size_bytes); - memset(ptr, 0, size); + memset_io(ptr, 0, size); } return 0; @@ -843,6 +818,7 @@ static int amdgpu_uvd_cs_reg(struct amdgpu_uvd_cs_ctx *ctx, return r; break; case mmUVD_ENGINE_CNTL: + case mmUVD_NO_OP: break; default: DRM_ERROR("Invalid reg 0x%X!\n", reg); @@ -981,8 +957,10 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, ib->ptr[3] = addr >> 32; ib->ptr[4] = PACKET0(mmUVD_GPCOM_VCPU_CMD, 0); ib->ptr[5] = 0; - for (i = 6; i < 16; ++i) - ib->ptr[i] = PACKET2(0); + for (i = 6; i < 16; i += 2) { + ib->ptr[i] = PACKET0(mmUVD_NO_OP, 0); + ib->ptr[i+1] = 0; + } ib->length_dw = 16; if (direct) { @@ -1114,15 +1092,9 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work) { struct amdgpu_device *adev = container_of(work, struct amdgpu_device, uvd.idle_work.work); - unsigned i, fences, handles = 0; - - fences = amdgpu_fence_count_emitted(&adev->uvd.ring); - - for (i = 0; i < adev->uvd.max_handles; ++i) - if (atomic_read(&adev->uvd.handles[i])) - ++handles; + unsigned fences = amdgpu_fence_count_emitted(&adev->uvd.ring); - if (fences == 0 && handles == 0) { + if (fences == 0) { if (adev->pm.dpm_enabled) { amdgpu_dpm_enable_uvd(adev, false); } else { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index 05865ce35351..da52af2a935a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -282,8 +282,8 @@ int amdgpu_vce_resume(struct amdgpu_device *adev) hdr = (const struct common_firmware_header *)adev->vce.fw->data; offset = le32_to_cpu(hdr->ucode_array_offset_bytes); - memcpy(cpu_addr, (adev->vce.fw->data) + offset, - (adev->vce.fw->size) - offset); + memcpy_toio(cpu_addr, adev->vce.fw->data + offset, + adev->vce.fw->size - offset); amdgpu_bo_kunmap(adev->vce.vcpu_bo); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 8e642fc48df4..bf56f1814437 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -51,19 +51,22 @@ * SI supports 16. */ -/* Special value that no flush is necessary */ -#define AMDGPU_VM_NO_FLUSH (~0ll) - /* Local structure. Encapsulate some VM table update parameters to reduce * the number of function parameters */ -struct amdgpu_vm_update_params { +struct amdgpu_pte_update_params { + /* amdgpu device we do this update for */ + struct amdgpu_device *adev; /* address where to copy page table entries from */ uint64_t src; - /* DMA addresses to use for mapping */ - dma_addr_t *pages_addr; /* indirect buffer to fill with commands */ struct amdgpu_ib *ib; + /* Function which actually does the update */ + void (*func)(struct amdgpu_pte_update_params *params, uint64_t pe, + uint64_t addr, unsigned count, uint32_t incr, + uint32_t flags); + /* indicate update pt or its shadow */ + bool shadow; }; /** @@ -467,10 +470,9 @@ struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm, } /** - * amdgpu_vm_update_pages - helper to call the right asic function + * amdgpu_vm_do_set_ptes - helper to call the right asic function * - * @adev: amdgpu_device pointer - * @vm_update_params: see amdgpu_vm_update_params definition + * @params: see amdgpu_pte_update_params definition * @pe: addr of the page entry * @addr: dst addr to write into pe * @count: number of page entries to update @@ -480,35 +482,47 @@ struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm, * Traces the parameters and calls the right asic functions * to setup the page table using the DMA. */ -static void amdgpu_vm_update_pages(struct amdgpu_device *adev, - struct amdgpu_vm_update_params - *vm_update_params, - uint64_t pe, uint64_t addr, - unsigned count, uint32_t incr, - uint32_t flags) +static void amdgpu_vm_do_set_ptes(struct amdgpu_pte_update_params *params, + uint64_t pe, uint64_t addr, + unsigned count, uint32_t incr, + uint32_t flags) { trace_amdgpu_vm_set_page(pe, addr, count, incr, flags); - if (vm_update_params->src) { - amdgpu_vm_copy_pte(adev, vm_update_params->ib, - pe, (vm_update_params->src + (addr >> 12) * 8), count); - - } else if (vm_update_params->pages_addr) { - amdgpu_vm_write_pte(adev, vm_update_params->ib, - vm_update_params->pages_addr, - pe, addr, count, incr, flags); - - } else if (count < 3) { - amdgpu_vm_write_pte(adev, vm_update_params->ib, NULL, pe, addr, - count, incr, flags); + if (count < 3) { + amdgpu_vm_write_pte(params->adev, params->ib, pe, + addr | flags, count, incr); } else { - amdgpu_vm_set_pte_pde(adev, vm_update_params->ib, pe, addr, + amdgpu_vm_set_pte_pde(params->adev, params->ib, pe, addr, count, incr, flags); } } /** + * amdgpu_vm_do_copy_ptes - copy the PTEs from the GART + * + * @params: see amdgpu_pte_update_params definition + * @pe: addr of the page entry + * @addr: dst addr to write into pe + * @count: number of page entries to update + * @incr: increase next addr by incr bytes + * @flags: hw access flags + * + * Traces the parameters and calls the DMA function to copy the PTEs. + */ +static void amdgpu_vm_do_copy_ptes(struct amdgpu_pte_update_params *params, + uint64_t pe, uint64_t addr, + unsigned count, uint32_t incr, + uint32_t flags) +{ + trace_amdgpu_vm_set_page(pe, addr, count, incr, flags); + + amdgpu_vm_copy_pte(params->adev, params->ib, pe, + (params->src + (addr >> 12) * 8), count); +} + +/** * amdgpu_vm_clear_bo - initially clear the page dir/table * * @adev: amdgpu_device pointer @@ -523,12 +537,11 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, struct amdgpu_ring *ring; struct fence *fence = NULL; struct amdgpu_job *job; - struct amdgpu_vm_update_params vm_update_params; + struct amdgpu_pte_update_params params; unsigned entries; uint64_t addr; int r; - memset(&vm_update_params, 0, sizeof(vm_update_params)); ring = container_of(vm->entity.sched, struct amdgpu_ring, sched); r = reservation_object_reserve_shared(bo->tbo.resv); @@ -546,9 +559,10 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, if (r) goto error; - vm_update_params.ib = &job->ibs[0]; - amdgpu_vm_update_pages(adev, &vm_update_params, addr, 0, entries, - 0, 0); + memset(¶ms, 0, sizeof(params)); + params.adev = adev; + params.ib = &job->ibs[0]; + amdgpu_vm_do_set_ptes(¶ms, addr, 0, entries, 0, 0); amdgpu_ring_pad_ib(ring, &job->ibs[0]); WARN_ON(job->ibs[0].length_dw > 64); @@ -577,55 +591,41 @@ error: * Look up the physical address of the page that the pte resolves * to and return the pointer for the page table entry. */ -uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr) +static uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr) { uint64_t result; - if (pages_addr) { - /* page table offset */ - result = pages_addr[addr >> PAGE_SHIFT]; - - /* in case cpu page size != gpu page size*/ - result |= addr & (~PAGE_MASK); + /* page table offset */ + result = pages_addr[addr >> PAGE_SHIFT]; - } else { - /* No mapping required */ - result = addr; - } + /* in case cpu page size != gpu page size*/ + result |= addr & (~PAGE_MASK); result &= 0xFFFFFFFFFFFFF000ULL; return result; } -/** - * amdgpu_vm_update_pdes - make sure that page directory is valid - * - * @adev: amdgpu_device pointer - * @vm: requested vm - * @start: start of GPU address range - * @end: end of GPU address range - * - * Allocates new page tables if necessary - * and updates the page directory. - * Returns 0 for success, error for failure. - */ -int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, - struct amdgpu_vm *vm) +static int amdgpu_vm_update_pd_or_shadow(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + bool shadow) { struct amdgpu_ring *ring; - struct amdgpu_bo *pd = vm->page_directory; - uint64_t pd_addr = amdgpu_bo_gpu_offset(pd); + struct amdgpu_bo *pd = shadow ? vm->page_directory->shadow : + vm->page_directory; + uint64_t pd_addr; uint32_t incr = AMDGPU_VM_PTE_COUNT * 8; uint64_t last_pde = ~0, last_pt = ~0; unsigned count = 0, pt_idx, ndw; struct amdgpu_job *job; - struct amdgpu_vm_update_params vm_update_params; + struct amdgpu_pte_update_params params; struct fence *fence = NULL; int r; - memset(&vm_update_params, 0, sizeof(vm_update_params)); + if (!pd) + return 0; + pd_addr = amdgpu_bo_gpu_offset(pd); ring = container_of(vm->entity.sched, struct amdgpu_ring, sched); /* padding, etc. */ @@ -638,7 +638,9 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, if (r) return r; - vm_update_params.ib = &job->ibs[0]; + memset(¶ms, 0, sizeof(params)); + params.adev = adev; + params.ib = &job->ibs[0]; /* walk over the address space and update the page directory */ for (pt_idx = 0; pt_idx <= vm->max_pde_used; ++pt_idx) { @@ -649,19 +651,25 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, continue; pt = amdgpu_bo_gpu_offset(bo); - if (vm->page_tables[pt_idx].addr == pt) - continue; - vm->page_tables[pt_idx].addr = pt; + if (!shadow) { + if (vm->page_tables[pt_idx].addr == pt) + continue; + vm->page_tables[pt_idx].addr = pt; + } else { + if (vm->page_tables[pt_idx].shadow_addr == pt) + continue; + vm->page_tables[pt_idx].shadow_addr = pt; + } pde = pd_addr + pt_idx * 8; if (((last_pde + 8 * count) != pde) || - ((last_pt + incr * count) != pt)) { + ((last_pt + incr * count) != pt) || + (count == AMDGPU_VM_MAX_UPDATE_SIZE)) { if (count) { - amdgpu_vm_update_pages(adev, &vm_update_params, - last_pde, last_pt, - count, incr, - AMDGPU_PTE_VALID); + amdgpu_vm_do_set_ptes(¶ms, last_pde, + last_pt, count, incr, + AMDGPU_PTE_VALID); } count = 1; @@ -673,15 +681,14 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, } if (count) - amdgpu_vm_update_pages(adev, &vm_update_params, - last_pde, last_pt, - count, incr, AMDGPU_PTE_VALID); + amdgpu_vm_do_set_ptes(¶ms, last_pde, last_pt, + count, incr, AMDGPU_PTE_VALID); - if (vm_update_params.ib->length_dw != 0) { - amdgpu_ring_pad_ib(ring, vm_update_params.ib); + if (params.ib->length_dw != 0) { + amdgpu_ring_pad_ib(ring, params.ib); amdgpu_sync_resv(adev, &job->sync, pd->tbo.resv, AMDGPU_FENCE_OWNER_VM); - WARN_ON(vm_update_params.ib->length_dw > ndw); + WARN_ON(params.ib->length_dw > ndw); r = amdgpu_job_submit(job, ring, &vm->entity, AMDGPU_FENCE_OWNER_VM, &fence); if (r) @@ -703,92 +710,33 @@ error_free: return r; } -/** - * amdgpu_vm_frag_ptes - add fragment information to PTEs +/* + * amdgpu_vm_update_pdes - make sure that page directory is valid * * @adev: amdgpu_device pointer - * @vm_update_params: see amdgpu_vm_update_params definition - * @pe_start: first PTE to handle - * @pe_end: last PTE to handle - * @addr: addr those PTEs should point to - * @flags: hw mapping flags + * @vm: requested vm + * @start: start of GPU address range + * @end: end of GPU address range + * + * Allocates new page tables if necessary + * and updates the page directory. + * Returns 0 for success, error for failure. */ -static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev, - struct amdgpu_vm_update_params - *vm_update_params, - uint64_t pe_start, uint64_t pe_end, - uint64_t addr, uint32_t flags) +int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, + struct amdgpu_vm *vm) { - /** - * The MC L1 TLB supports variable sized pages, based on a fragment - * field in the PTE. When this field is set to a non-zero value, page - * granularity is increased from 4KB to (1 << (12 + frag)). The PTE - * flags are considered valid for all PTEs within the fragment range - * and corresponding mappings are assumed to be physically contiguous. - * - * The L1 TLB can store a single PTE for the whole fragment, - * significantly increasing the space available for translation - * caching. This leads to large improvements in throughput when the - * TLB is under pressure. - * - * The L2 TLB distributes small and large fragments into two - * asymmetric partitions. The large fragment cache is significantly - * larger. Thus, we try to use large fragments wherever possible. - * Userspace can support this by aligning virtual base address and - * allocation size to the fragment size. - */ - - /* SI and newer are optimized for 64KB */ - uint64_t frag_flags = AMDGPU_PTE_FRAG_64KB; - uint64_t frag_align = 0x80; - - uint64_t frag_start = ALIGN(pe_start, frag_align); - uint64_t frag_end = pe_end & ~(frag_align - 1); - - unsigned count; - - /* Abort early if there isn't anything to do */ - if (pe_start == pe_end) - return; - - /* system pages are non continuously */ - if (vm_update_params->src || vm_update_params->pages_addr || - !(flags & AMDGPU_PTE_VALID) || (frag_start >= frag_end)) { - - count = (pe_end - pe_start) / 8; - amdgpu_vm_update_pages(adev, vm_update_params, pe_start, - addr, count, AMDGPU_GPU_PAGE_SIZE, - flags); - return; - } - - /* handle the 4K area at the beginning */ - if (pe_start != frag_start) { - count = (frag_start - pe_start) / 8; - amdgpu_vm_update_pages(adev, vm_update_params, pe_start, addr, - count, AMDGPU_GPU_PAGE_SIZE, flags); - addr += AMDGPU_GPU_PAGE_SIZE * count; - } - - /* handle the area in the middle */ - count = (frag_end - frag_start) / 8; - amdgpu_vm_update_pages(adev, vm_update_params, frag_start, addr, count, - AMDGPU_GPU_PAGE_SIZE, flags | frag_flags); + int r; - /* handle the 4K area at the end */ - if (frag_end != pe_end) { - addr += AMDGPU_GPU_PAGE_SIZE * count; - count = (pe_end - frag_end) / 8; - amdgpu_vm_update_pages(adev, vm_update_params, frag_end, addr, - count, AMDGPU_GPU_PAGE_SIZE, flags); - } + r = amdgpu_vm_update_pd_or_shadow(adev, vm, true); + if (r) + return r; + return amdgpu_vm_update_pd_or_shadow(adev, vm, false); } /** * amdgpu_vm_update_ptes - make sure that page tables are valid * - * @adev: amdgpu_device pointer - * @vm_update_params: see amdgpu_vm_update_params definition + * @params: see amdgpu_pte_update_params definition * @vm: requested vm * @start: start of GPU address range * @end: end of GPU address range @@ -797,16 +745,14 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev, * * Update the page tables in the range @start - @end. */ -static void amdgpu_vm_update_ptes(struct amdgpu_device *adev, - struct amdgpu_vm_update_params - *vm_update_params, +static void amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params, struct amdgpu_vm *vm, uint64_t start, uint64_t end, uint64_t dst, uint32_t flags) { const uint64_t mask = AMDGPU_VM_PTE_COUNT - 1; - uint64_t cur_pe_start, cur_pe_end, cur_dst; + uint64_t cur_pe_start, cur_nptes, cur_dst; uint64_t addr; /* next GPU address to be updated */ uint64_t pt_idx; struct amdgpu_bo *pt; @@ -817,7 +763,11 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev, addr = start; pt_idx = addr >> amdgpu_vm_block_size; pt = vm->page_tables[pt_idx].entry.robj; - + if (params->shadow) { + if (!pt->shadow) + return; + pt = vm->page_tables[pt_idx].entry.robj->shadow; + } if ((addr & ~mask) == (end & ~mask)) nptes = end - addr; else @@ -825,7 +775,7 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev, cur_pe_start = amdgpu_bo_gpu_offset(pt); cur_pe_start += (addr & mask) * 8; - cur_pe_end = cur_pe_start + 8 * nptes; + cur_nptes = nptes; cur_dst = dst; /* for next ptb*/ @@ -836,6 +786,11 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev, while (addr < end) { pt_idx = addr >> amdgpu_vm_block_size; pt = vm->page_tables[pt_idx].entry.robj; + if (params->shadow) { + if (!pt->shadow) + return; + pt = vm->page_tables[pt_idx].entry.robj->shadow; + } if ((addr & ~mask) == (end & ~mask)) nptes = end - addr; @@ -845,19 +800,19 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev, next_pe_start = amdgpu_bo_gpu_offset(pt); next_pe_start += (addr & mask) * 8; - if (cur_pe_end == next_pe_start) { + if ((cur_pe_start + 8 * cur_nptes) == next_pe_start && + ((cur_nptes + nptes) <= AMDGPU_VM_MAX_UPDATE_SIZE)) { /* The next ptb is consecutive to current ptb. - * Don't call amdgpu_vm_frag_ptes now. + * Don't call the update function now. * Will update two ptbs together in future. */ - cur_pe_end += 8 * nptes; + cur_nptes += nptes; } else { - amdgpu_vm_frag_ptes(adev, vm_update_params, - cur_pe_start, cur_pe_end, - cur_dst, flags); + params->func(params, cur_pe_start, cur_dst, cur_nptes, + AMDGPU_GPU_PAGE_SIZE, flags); cur_pe_start = next_pe_start; - cur_pe_end = next_pe_start + 8 * nptes; + cur_nptes = nptes; cur_dst = dst; } @@ -866,8 +821,79 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev, dst += nptes * AMDGPU_GPU_PAGE_SIZE; } - amdgpu_vm_frag_ptes(adev, vm_update_params, cur_pe_start, - cur_pe_end, cur_dst, flags); + params->func(params, cur_pe_start, cur_dst, cur_nptes, + AMDGPU_GPU_PAGE_SIZE, flags); +} + +/* + * amdgpu_vm_frag_ptes - add fragment information to PTEs + * + * @params: see amdgpu_pte_update_params definition + * @vm: requested vm + * @start: first PTE to handle + * @end: last PTE to handle + * @dst: addr those PTEs should point to + * @flags: hw mapping flags + */ +static void amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, + struct amdgpu_vm *vm, + uint64_t start, uint64_t end, + uint64_t dst, uint32_t flags) +{ + /** + * The MC L1 TLB supports variable sized pages, based on a fragment + * field in the PTE. When this field is set to a non-zero value, page + * granularity is increased from 4KB to (1 << (12 + frag)). The PTE + * flags are considered valid for all PTEs within the fragment range + * and corresponding mappings are assumed to be physically contiguous. + * + * The L1 TLB can store a single PTE for the whole fragment, + * significantly increasing the space available for translation + * caching. This leads to large improvements in throughput when the + * TLB is under pressure. + * + * The L2 TLB distributes small and large fragments into two + * asymmetric partitions. The large fragment cache is significantly + * larger. Thus, we try to use large fragments wherever possible. + * Userspace can support this by aligning virtual base address and + * allocation size to the fragment size. + */ + + const uint64_t frag_align = 1 << AMDGPU_LOG2_PAGES_PER_FRAG; + + uint64_t frag_start = ALIGN(start, frag_align); + uint64_t frag_end = end & ~(frag_align - 1); + + uint32_t frag; + + /* system pages are non continuously */ + if (params->src || !(flags & AMDGPU_PTE_VALID) || + (frag_start >= frag_end)) { + + amdgpu_vm_update_ptes(params, vm, start, end, dst, flags); + return; + } + + /* use more than 64KB fragment size if possible */ + frag = lower_32_bits(frag_start | frag_end); + frag = likely(frag) ? __ffs(frag) : 31; + + /* handle the 4K area at the beginning */ + if (start != frag_start) { + amdgpu_vm_update_ptes(params, vm, start, frag_start, + dst, flags); + dst += (frag_start - start) * AMDGPU_GPU_PAGE_SIZE; + } + + /* handle the area in the middle */ + amdgpu_vm_update_ptes(params, vm, frag_start, frag_end, dst, + flags | AMDGPU_PTE_FRAG(frag)); + + /* handle the 4K area at the end */ + if (frag_end != end) { + dst += (frag_end - frag_start) * AMDGPU_GPU_PAGE_SIZE; + amdgpu_vm_update_ptes(params, vm, frag_end, end, dst, flags); + } } /** @@ -900,14 +926,19 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, void *owner = AMDGPU_FENCE_OWNER_VM; unsigned nptes, ncmds, ndw; struct amdgpu_job *job; - struct amdgpu_vm_update_params vm_update_params; + struct amdgpu_pte_update_params params; struct fence *f = NULL; int r; + memset(¶ms, 0, sizeof(params)); + params.adev = adev; + params.src = src; + ring = container_of(vm->entity.sched, struct amdgpu_ring, sched); - memset(&vm_update_params, 0, sizeof(vm_update_params)); - vm_update_params.src = src; - vm_update_params.pages_addr = pages_addr; + + memset(¶ms, 0, sizeof(params)); + params.adev = adev; + params.src = src; /* sync to everything on unmapping */ if (!(flags & AMDGPU_PTE_VALID)) @@ -924,30 +955,52 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, /* padding, etc. */ ndw = 64; - if (vm_update_params.src) { + if (src) { /* only copy commands needed */ ndw += ncmds * 7; - } else if (vm_update_params.pages_addr) { - /* header for write data commands */ - ndw += ncmds * 4; + params.func = amdgpu_vm_do_copy_ptes; + + } else if (pages_addr) { + /* copy commands needed */ + ndw += ncmds * 7; - /* body of write data command */ + /* and also PTEs */ ndw += nptes * 2; + params.func = amdgpu_vm_do_copy_ptes; + } else { /* set page commands needed */ ndw += ncmds * 10; /* two extra commands for begin/end of fragment */ ndw += 2 * 10; + + params.func = amdgpu_vm_do_set_ptes; } r = amdgpu_job_alloc_with_ib(adev, ndw * 4, &job); if (r) return r; - vm_update_params.ib = &job->ibs[0]; + params.ib = &job->ibs[0]; + + if (!src && pages_addr) { + uint64_t *pte; + unsigned i; + + /* Put the PTEs at the end of the IB. */ + i = ndw - nptes * 2; + pte= (uint64_t *)&(job->ibs->ptr[i]); + params.src = job->ibs->gpu_addr + i * 4; + + for (i = 0; i < nptes; ++i) { + pte[i] = amdgpu_vm_map_gart(pages_addr, addr + i * + AMDGPU_GPU_PAGE_SIZE); + pte[i] |= flags; + } + } r = amdgpu_sync_fence(adev, &job->sync, exclusive); if (r) @@ -962,11 +1015,13 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (r) goto error_free; - amdgpu_vm_update_ptes(adev, &vm_update_params, vm, start, - last + 1, addr, flags); + params.shadow = true; + amdgpu_vm_frag_ptes(¶ms, vm, start, last + 1, addr, flags); + params.shadow = false; + amdgpu_vm_frag_ptes(¶ms, vm, start, last + 1, addr, flags); - amdgpu_ring_pad_ib(ring, vm_update_params.ib); - WARN_ON(vm_update_params.ib->length_dw > ndw); + amdgpu_ring_pad_ib(ring, params.ib); + WARN_ON(params.ib->length_dw > ndw); r = amdgpu_job_submit(job, ring, &vm->entity, AMDGPU_FENCE_OWNER_VM, &f); if (r) @@ -1062,28 +1117,32 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, * * @adev: amdgpu_device pointer * @bo_va: requested BO and VM object - * @mem: ttm mem + * @clear: if true clear the entries * * Fill in the page table entries for @bo_va. * Returns 0 for success, -EINVAL for failure. - * - * Object have to be reserved and mutex must be locked! */ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, - struct ttm_mem_reg *mem) + bool clear) { struct amdgpu_vm *vm = bo_va->vm; struct amdgpu_bo_va_mapping *mapping; dma_addr_t *pages_addr = NULL; uint32_t gtt_flags, flags; + struct ttm_mem_reg *mem; struct fence *exclusive; uint64_t addr; int r; - if (mem) { + if (clear) { + mem = NULL; + addr = 0; + exclusive = NULL; + } else { struct ttm_dma_tt *ttm; + mem = &bo_va->bo->tbo.mem; addr = (u64)mem->start << PAGE_SHIFT; switch (mem->mem_type) { case TTM_PL_TT: @@ -1101,9 +1160,6 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, } exclusive = reservation_object_get_excl(bo_va->bo->tbo.resv); - } else { - addr = 0; - exclusive = NULL; } flags = amdgpu_ttm_tt_pte_flags(adev, bo_va->bo->tbo.ttm, mem); @@ -1134,7 +1190,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, spin_lock(&vm->status_lock); list_splice_init(&bo_va->invalids, &bo_va->valids); list_del_init(&bo_va->vm_status); - if (!mem) + if (clear) list_add(&bo_va->vm_status, &vm->cleared); spin_unlock(&vm->status_lock); @@ -1197,7 +1253,7 @@ int amdgpu_vm_clear_invalids(struct amdgpu_device *adev, struct amdgpu_bo_va, vm_status); spin_unlock(&vm->status_lock); - r = amdgpu_vm_bo_update(adev, bo_va, NULL); + r = amdgpu_vm_bo_update(adev, bo_va, true); if (r) return r; @@ -1342,7 +1398,8 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev, r = amdgpu_bo_create(adev, AMDGPU_VM_PTE_COUNT * 8, AMDGPU_GPU_PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_NO_CPU_ACCESS, + AMDGPU_GEM_CREATE_NO_CPU_ACCESS | + AMDGPU_GEM_CREATE_SHADOW, NULL, resv, &pt); if (r) goto error_free; @@ -1541,7 +1598,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm) r = amdgpu_bo_create(adev, pd_size, align, true, AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_NO_CPU_ACCESS, + AMDGPU_GEM_CREATE_NO_CPU_ACCESS | + AMDGPU_GEM_CREATE_SHADOW, NULL, NULL, &vm->page_directory); if (r) goto error_free_sched_entity; @@ -1597,10 +1655,16 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) kfree(mapping); } - for (i = 0; i < amdgpu_vm_num_pdes(adev); i++) + for (i = 0; i < amdgpu_vm_num_pdes(adev); i++) { + if (vm->page_tables[i].entry.robj && + vm->page_tables[i].entry.robj->shadow) + amdgpu_bo_unref(&vm->page_tables[i].entry.robj->shadow); amdgpu_bo_unref(&vm->page_tables[i].entry.robj); + } drm_free_large(vm->page_tables); + if (vm->page_directory->shadow) + amdgpu_bo_unref(&vm->page_directory->shadow); amdgpu_bo_unref(&vm->page_directory); fence_put(vm->page_directory_fence); } diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c index 7f85c2c1d681..f81068ba4cc6 100644 --- a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c +++ b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c @@ -88,7 +88,6 @@ static int amdgpu_atombios_dp_process_aux_ch(struct amdgpu_i2c_chan *chan, /* timeout */ if (args.v2.ucReplyStatus == 1) { - DRM_DEBUG_KMS("dp_aux_ch timeout\n"); r = -ETIMEDOUT; goto done; } @@ -339,22 +338,21 @@ int amdgpu_atombios_dp_get_dpcd(struct amdgpu_connector *amdgpu_connector) { struct amdgpu_connector_atom_dig *dig_connector = amdgpu_connector->con_priv; u8 msg[DP_DPCD_SIZE]; - int ret, i; + int ret; - for (i = 0; i < 7; i++) { - ret = drm_dp_dpcd_read(&amdgpu_connector->ddc_bus->aux, DP_DPCD_REV, msg, - DP_DPCD_SIZE); - if (ret == DP_DPCD_SIZE) { - memcpy(dig_connector->dpcd, msg, DP_DPCD_SIZE); + ret = drm_dp_dpcd_read(&amdgpu_connector->ddc_bus->aux, DP_DPCD_REV, + msg, DP_DPCD_SIZE); + if (ret == DP_DPCD_SIZE) { + memcpy(dig_connector->dpcd, msg, DP_DPCD_SIZE); - DRM_DEBUG_KMS("DPCD: %*ph\n", (int)sizeof(dig_connector->dpcd), - dig_connector->dpcd); + DRM_DEBUG_KMS("DPCD: %*ph\n", (int)sizeof(dig_connector->dpcd), + dig_connector->dpcd); - amdgpu_atombios_dp_probe_oui(amdgpu_connector); + amdgpu_atombios_dp_probe_oui(amdgpu_connector); - return 0; - } + return 0; } + dig_connector->dpcd[0] = 0; return -EINVAL; } diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c index a5c94b482459..a0d63a293bb0 100644 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c @@ -5874,7 +5874,10 @@ static int ci_dpm_init(struct amdgpu_device *adev) pi->pcie_dpm_key_disabled = 0; pi->thermal_sclk_dpm_enabled = 0; - pi->caps_sclk_ds = true; + if (amdgpu_sclk_deep_sleep_en) + pi->caps_sclk_ds = true; + else + pi->caps_sclk_ds = false; pi->mclk_strobe_mode_threshold = 40000; pi->mclk_stutter_mode_threshold = 40000; diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c index 4efc901f658c..825de800b798 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik.c +++ b/drivers/gpu/drm/amd/amdgpu/cik.c @@ -67,6 +67,7 @@ #include "amdgpu_amdkfd.h" #include "amdgpu_powerplay.h" +#include "dce_virtual.h" /* * Indirect registers accessor @@ -1708,6 +1709,74 @@ static const struct amdgpu_ip_block_version bonaire_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version bonaire_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &cik_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 8, + .minor = 2, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 7, + .minor = 2, + .rev = 0, + .funcs = &gfx_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_sdma_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 4, + .minor = 2, + .rev = 0, + .funcs = &uvd_v4_2_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vce_v2_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version hawaii_ip_blocks[] = { /* ORDER MATTERS! */ @@ -1776,6 +1845,74 @@ static const struct amdgpu_ip_block_version hawaii_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version hawaii_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &cik_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 8, + .minor = 5, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 7, + .minor = 3, + .rev = 0, + .funcs = &gfx_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_sdma_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 4, + .minor = 2, + .rev = 0, + .funcs = &uvd_v4_2_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vce_v2_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version kabini_ip_blocks[] = { /* ORDER MATTERS! */ @@ -1844,6 +1981,74 @@ static const struct amdgpu_ip_block_version kabini_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version kabini_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &cik_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 8, + .minor = 3, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 7, + .minor = 2, + .rev = 0, + .funcs = &gfx_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_sdma_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 4, + .minor = 2, + .rev = 0, + .funcs = &uvd_v4_2_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vce_v2_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version mullins_ip_blocks[] = { /* ORDER MATTERS! */ @@ -1912,6 +2117,74 @@ static const struct amdgpu_ip_block_version mullins_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version mullins_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &cik_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 8, + .minor = 3, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 7, + .minor = 2, + .rev = 0, + .funcs = &gfx_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_sdma_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 4, + .minor = 2, + .rev = 0, + .funcs = &uvd_v4_2_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vce_v2_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version kaveri_ip_blocks[] = { /* ORDER MATTERS! */ @@ -1980,32 +2253,128 @@ static const struct amdgpu_ip_block_version kaveri_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version kaveri_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &cik_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 8, + .minor = 1, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 7, + .minor = 1, + .rev = 0, + .funcs = &gfx_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_sdma_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 4, + .minor = 2, + .rev = 0, + .funcs = &uvd_v4_2_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vce_v2_0_ip_funcs, + }, +}; + int cik_set_ip_blocks(struct amdgpu_device *adev) { - switch (adev->asic_type) { - case CHIP_BONAIRE: - adev->ip_blocks = bonaire_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(bonaire_ip_blocks); - break; - case CHIP_HAWAII: - adev->ip_blocks = hawaii_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(hawaii_ip_blocks); - break; - case CHIP_KAVERI: - adev->ip_blocks = kaveri_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(kaveri_ip_blocks); - break; - case CHIP_KABINI: - adev->ip_blocks = kabini_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(kabini_ip_blocks); - break; - case CHIP_MULLINS: - adev->ip_blocks = mullins_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(mullins_ip_blocks); - break; - default: - /* FIXME: not supported yet */ - return -EINVAL; + if (adev->enable_virtual_display) { + switch (adev->asic_type) { + case CHIP_BONAIRE: + adev->ip_blocks = bonaire_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(bonaire_ip_blocks_vd); + break; + case CHIP_HAWAII: + adev->ip_blocks = hawaii_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(hawaii_ip_blocks_vd); + break; + case CHIP_KAVERI: + adev->ip_blocks = kaveri_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(kaveri_ip_blocks_vd); + break; + case CHIP_KABINI: + adev->ip_blocks = kabini_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(kabini_ip_blocks_vd); + break; + case CHIP_MULLINS: + adev->ip_blocks = mullins_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(mullins_ip_blocks_vd); + break; + default: + /* FIXME: not supported yet */ + return -EINVAL; + } + } else { + switch (adev->asic_type) { + case CHIP_BONAIRE: + adev->ip_blocks = bonaire_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(bonaire_ip_blocks); + break; + case CHIP_HAWAII: + adev->ip_blocks = hawaii_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(hawaii_ip_blocks); + break; + case CHIP_KAVERI: + adev->ip_blocks = kaveri_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(kaveri_ip_blocks); + break; + case CHIP_KABINI: + adev->ip_blocks = kabini_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(kabini_ip_blocks); + break; + case CHIP_MULLINS: + adev->ip_blocks = mullins_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(mullins_ip_blocks); + break; + default: + /* FIXME: not supported yet */ + return -EINVAL; + } } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index ee6466912497..e71cd12104b3 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -694,24 +694,16 @@ static void cik_sdma_vm_copy_pte(struct amdgpu_ib *ib, uint64_t pe, uint64_t src, unsigned count) { - while (count) { - unsigned bytes = count * 8; - if (bytes > 0x1FFFF8) - bytes = 0x1FFFF8; - - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_COPY, - SDMA_WRITE_SUB_OPCODE_LINEAR, 0); - ib->ptr[ib->length_dw++] = bytes; - ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ - ib->ptr[ib->length_dw++] = lower_32_bits(src); - ib->ptr[ib->length_dw++] = upper_32_bits(src); - ib->ptr[ib->length_dw++] = lower_32_bits(pe); - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - - pe += bytes; - src += bytes; - count -= bytes / 8; - } + unsigned bytes = count * 8; + + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_COPY, + SDMA_WRITE_SUB_OPCODE_LINEAR, 0); + ib->ptr[ib->length_dw++] = bytes; + ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ + ib->ptr[ib->length_dw++] = lower_32_bits(src); + ib->ptr[ib->length_dw++] = upper_32_bits(src); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); } /** @@ -719,39 +711,27 @@ static void cik_sdma_vm_copy_pte(struct amdgpu_ib *ib, * * @ib: indirect buffer to fill with commands * @pe: addr of the page entry - * @addr: dst addr to write into pe + * @value: dst addr to write into pe * @count: number of page entries to update * @incr: increase next addr by incr bytes - * @flags: access flags * * Update PTEs by writing them manually using sDMA (CIK). */ -static void cik_sdma_vm_write_pte(struct amdgpu_ib *ib, - const dma_addr_t *pages_addr, uint64_t pe, - uint64_t addr, unsigned count, - uint32_t incr, uint32_t flags) +static void cik_sdma_vm_write_pte(struct amdgpu_ib *ib, uint64_t pe, + uint64_t value, unsigned count, + uint32_t incr) { - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - /* for non-physically contiguous pages (system) */ - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_WRITE, - SDMA_WRITE_SUB_OPCODE_LINEAR, 0); - ib->ptr[ib->length_dw++] = pe; - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = ndw; - for (; ndw > 0; ndw -= 2, --count, pe += 8) { - value = amdgpu_vm_map_gart(pages_addr, addr); - addr += incr; - value |= flags; - ib->ptr[ib->length_dw++] = value; - ib->ptr[ib->length_dw++] = upper_32_bits(value); - } + unsigned ndw = count * 2; + + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_WRITE, + SDMA_WRITE_SUB_OPCODE_LINEAR, 0); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = ndw; + for (; ndw > 0; ndw -= 2) { + ib->ptr[ib->length_dw++] = lower_32_bits(value); + ib->ptr[ib->length_dw++] = upper_32_bits(value); + value += incr; } } @@ -767,40 +747,21 @@ static void cik_sdma_vm_write_pte(struct amdgpu_ib *ib, * * Update the page tables using sDMA (CIK). */ -static void cik_sdma_vm_set_pte_pde(struct amdgpu_ib *ib, - uint64_t pe, +static void cik_sdma_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe, uint64_t addr, unsigned count, uint32_t incr, uint32_t flags) { - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count; - if (ndw > 0x7FFFF) - ndw = 0x7FFFF; - - if (flags & AMDGPU_PTE_VALID) - value = addr; - else - value = 0; - - /* for physically contiguous pages (vram) */ - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_GENERATE_PTE_PDE, 0, 0); - ib->ptr[ib->length_dw++] = pe; /* dst addr */ - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = flags; /* mask */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = value; /* value */ - ib->ptr[ib->length_dw++] = upper_32_bits(value); - ib->ptr[ib->length_dw++] = incr; /* increment size */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = ndw; /* number of entries */ - - pe += ndw * 8; - addr += ndw * incr; - count -= ndw; - } + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_GENERATE_PTE_PDE, 0, 0); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = flags; /* mask */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = lower_32_bits(addr); /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(addr); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = count; /* number of entries */ } /** diff --git a/drivers/gpu/drm/amd/amdgpu/cz_dpm.c b/drivers/gpu/drm/amd/amdgpu/cz_dpm.c index 2a11413ed54a..794c5f36ca68 100644 --- a/drivers/gpu/drm/amd/amdgpu/cz_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/cz_dpm.c @@ -435,7 +435,11 @@ static int cz_dpm_init(struct amdgpu_device *adev) pi->caps_td_ramping = true; pi->caps_tcp_ramping = true; } - pi->caps_sclk_ds = true; + if (amdgpu_sclk_deep_sleep_en) + pi->caps_sclk_ds = true; + else + pi->caps_sclk_ds = false; + pi->voting_clients = 0x00c00033; pi->auto_thermal_throttling_enabled = true; pi->bapm_enabled = false; @@ -2108,29 +2112,58 @@ static void cz_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate) /* disable clockgating so we can properly shut down the block */ ret = amdgpu_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_UVD, AMD_CG_STATE_UNGATE); + if (ret) { + DRM_ERROR("UVD DPM Power Gating failed to set clockgating state\n"); + return; + } + /* shutdown the UVD block */ ret = amdgpu_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_UVD, AMD_PG_STATE_GATE); - /* XXX: check for errors */ + + if (ret) { + DRM_ERROR("UVD DPM Power Gating failed to set powergating state\n"); + return; + } } cz_update_uvd_dpm(adev, gate); - if (pi->caps_uvd_pg) + if (pi->caps_uvd_pg) { /* power off the UVD block */ - cz_send_msg_to_smc(adev, PPSMC_MSG_UVDPowerOFF); + ret = cz_send_msg_to_smc(adev, PPSMC_MSG_UVDPowerOFF); + if (ret) { + DRM_ERROR("UVD DPM Power Gating failed to send SMU PowerOFF message\n"); + return; + } + } } else { if (pi->caps_uvd_pg) { /* power on the UVD block */ if (pi->uvd_dynamic_pg) - cz_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_UVDPowerON, 1); + ret = cz_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_UVDPowerON, 1); else - cz_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_UVDPowerON, 0); + ret = cz_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_UVDPowerON, 0); + + if (ret) { + DRM_ERROR("UVD DPM Power Gating Failed to send SMU PowerON message\n"); + return; + } + /* re-init the UVD block */ ret = amdgpu_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_UVD, AMD_PG_STATE_UNGATE); + + if (ret) { + DRM_ERROR("UVD DPM Power Gating Failed to set powergating state\n"); + return; + } + /* enable clockgating. hw will dynamically gate/ungate clocks on the fly */ ret = amdgpu_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_UVD, AMD_CG_STATE_GATE); - /* XXX: check for errors */ + if (ret) { + DRM_ERROR("UVD DPM Power Gating Failed to set clockgating state\n"); + return; + } } cz_update_uvd_dpm(adev, gate); } diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c index 741da36cd8b6..bc5bb4eb9625 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c @@ -646,8 +646,8 @@ static void dce_v10_0_resume_mc_access(struct amdgpu_device *adev, if (save->crtc_enabled[i]) { tmp = RREG32(mmMASTER_UPDATE_MODE + crtc_offsets[i]); - if (REG_GET_FIELD(tmp, MASTER_UPDATE_MODE, MASTER_UPDATE_MODE) != 3) { - tmp = REG_SET_FIELD(tmp, MASTER_UPDATE_MODE, MASTER_UPDATE_MODE, 3); + if (REG_GET_FIELD(tmp, MASTER_UPDATE_MODE, MASTER_UPDATE_MODE) != 0) { + tmp = REG_SET_FIELD(tmp, MASTER_UPDATE_MODE, MASTER_UPDATE_MODE, 0); WREG32(mmMASTER_UPDATE_MODE + crtc_offsets[i], tmp); } tmp = RREG32(mmGRPH_UPDATE + crtc_offsets[i]); @@ -712,6 +712,45 @@ static void dce_v10_0_set_vga_render_state(struct amdgpu_device *adev, WREG32(mmVGA_RENDER_CONTROL, tmp); } +static int dce_v10_0_get_num_crtc(struct amdgpu_device *adev) +{ + int num_crtc = 0; + + switch (adev->asic_type) { + case CHIP_FIJI: + case CHIP_TONGA: + num_crtc = 6; + break; + default: + num_crtc = 0; + } + return num_crtc; +} + +void dce_v10_0_disable_dce(struct amdgpu_device *adev) +{ + /*Disable VGA render and enabled crtc, if has DCE engine*/ + if (amdgpu_atombios_has_dce_engine_info(adev)) { + u32 tmp; + int crtc_enabled, i; + + dce_v10_0_set_vga_render_state(adev, false); + + /*Disable crtc*/ + for (i = 0; i < dce_v10_0_get_num_crtc(adev); i++) { + crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]), + CRTC_CONTROL, CRTC_MASTER_EN); + if (crtc_enabled) { + WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1); + tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]); + tmp = REG_SET_FIELD(tmp, CRTC_CONTROL, CRTC_MASTER_EN, 0); + WREG32(mmCRTC_CONTROL + crtc_offsets[i], tmp); + WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0); + } + } + } +} + static void dce_v10_0_program_fmt(struct drm_encoder *encoder) { struct drm_device *dev = encoder->dev; @@ -2277,8 +2316,8 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc, WREG32(mmVIEWPORT_SIZE + amdgpu_crtc->crtc_offset, (viewport_w << 16) | viewport_h); - /* set pageflip to happen only at start of vblank interval (front porch) */ - WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 3); + /* set pageflip to happen anywhere in vblank interval */ + WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { amdgpu_fb = to_amdgpu_framebuffer(fb); @@ -2700,7 +2739,7 @@ static const struct drm_crtc_funcs dce_v10_0_crtc_funcs = { .gamma_set = dce_v10_0_crtc_gamma_set, .set_config = amdgpu_crtc_set_config, .destroy = dce_v10_0_crtc_destroy, - .page_flip = amdgpu_crtc_page_flip, + .page_flip_target = amdgpu_crtc_page_flip_target, }; static void dce_v10_0_crtc_dpms(struct drm_crtc *crtc, int mode) @@ -2964,10 +3003,11 @@ static int dce_v10_0_early_init(void *handle) dce_v10_0_set_display_funcs(adev); dce_v10_0_set_irq_funcs(adev); + adev->mode_info.num_crtc = dce_v10_0_get_num_crtc(adev); + switch (adev->asic_type) { case CHIP_FIJI: case CHIP_TONGA: - adev->mode_info.num_crtc = 6; /* XXX 7??? */ adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 7; break; @@ -3143,11 +3183,26 @@ static int dce_v10_0_wait_for_idle(void *handle) return 0; } +static int dce_v10_0_check_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (dce_v10_0_is_display_hung(adev)) + adev->ip_block_status[AMD_IP_BLOCK_TYPE_DCE].hang = true; + else + adev->ip_block_status[AMD_IP_BLOCK_TYPE_DCE].hang = false; + + return 0; +} + static int dce_v10_0_soft_reset(void *handle) { u32 srbm_soft_reset = 0, tmp; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_DCE].hang) + return 0; + if (dce_v10_0_is_display_hung(adev)) srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK; @@ -3514,6 +3569,7 @@ const struct amd_ip_funcs dce_v10_0_ip_funcs = { .resume = dce_v10_0_resume, .is_idle = dce_v10_0_is_idle, .wait_for_idle = dce_v10_0_wait_for_idle, + .check_soft_reset = dce_v10_0_check_soft_reset, .soft_reset = dce_v10_0_soft_reset, .set_clockgating_state = dce_v10_0_set_clockgating_state, .set_powergating_state = dce_v10_0_set_powergating_state, diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.h b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.h index 1bfa48ddd8a6..e3dc04d293e4 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.h +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.h @@ -26,4 +26,6 @@ extern const struct amd_ip_funcs dce_v10_0_ip_funcs; +void dce_v10_0_disable_dce(struct amdgpu_device *adev); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c index 2282eb60aba6..b93eba077950 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c @@ -673,6 +673,53 @@ static void dce_v11_0_set_vga_render_state(struct amdgpu_device *adev, WREG32(mmVGA_RENDER_CONTROL, tmp); } +static int dce_v11_0_get_num_crtc (struct amdgpu_device *adev) +{ + int num_crtc = 0; + + switch (adev->asic_type) { + case CHIP_CARRIZO: + num_crtc = 3; + break; + case CHIP_STONEY: + num_crtc = 2; + break; + case CHIP_POLARIS10: + num_crtc = 6; + break; + case CHIP_POLARIS11: + num_crtc = 5; + break; + default: + num_crtc = 0; + } + return num_crtc; +} + +void dce_v11_0_disable_dce(struct amdgpu_device *adev) +{ + /*Disable VGA render and enabled crtc, if has DCE engine*/ + if (amdgpu_atombios_has_dce_engine_info(adev)) { + u32 tmp; + int crtc_enabled, i; + + dce_v11_0_set_vga_render_state(adev, false); + + /*Disable crtc*/ + for (i = 0; i < dce_v11_0_get_num_crtc(adev); i++) { + crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]), + CRTC_CONTROL, CRTC_MASTER_EN); + if (crtc_enabled) { + WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1); + tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]); + tmp = REG_SET_FIELD(tmp, CRTC_CONTROL, CRTC_MASTER_EN, 0); + WREG32(mmCRTC_CONTROL + crtc_offsets[i], tmp); + WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0); + } + } + } +} + static void dce_v11_0_program_fmt(struct drm_encoder *encoder) { struct drm_device *dev = encoder->dev; @@ -2252,8 +2299,8 @@ static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc, WREG32(mmVIEWPORT_SIZE + amdgpu_crtc->crtc_offset, (viewport_w << 16) | viewport_h); - /* set pageflip to happen only at start of vblank interval (front porch) */ - WREG32(mmCRTC_MASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 3); + /* set pageflip to happen anywhere in vblank interval */ + WREG32(mmCRTC_MASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { amdgpu_fb = to_amdgpu_framebuffer(fb); @@ -2710,7 +2757,7 @@ static const struct drm_crtc_funcs dce_v11_0_crtc_funcs = { .gamma_set = dce_v11_0_crtc_gamma_set, .set_config = amdgpu_crtc_set_config, .destroy = dce_v11_0_crtc_destroy, - .page_flip = amdgpu_crtc_page_flip, + .page_flip_target = amdgpu_crtc_page_flip_target, }; static void dce_v11_0_crtc_dpms(struct drm_crtc *crtc, int mode) @@ -3001,24 +3048,22 @@ static int dce_v11_0_early_init(void *handle) dce_v11_0_set_display_funcs(adev); dce_v11_0_set_irq_funcs(adev); + adev->mode_info.num_crtc = dce_v11_0_get_num_crtc(adev); + switch (adev->asic_type) { case CHIP_CARRIZO: - adev->mode_info.num_crtc = 3; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 9; break; case CHIP_STONEY: - adev->mode_info.num_crtc = 2; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 9; break; case CHIP_POLARIS10: - adev->mode_info.num_crtc = 6; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 6; break; case CHIP_POLARIS11: - adev->mode_info.num_crtc = 5; adev->mode_info.num_hpd = 5; adev->mode_info.num_dig = 5; break; diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h index 84e4618f5253..1f58a65ba2ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h @@ -26,4 +26,6 @@ extern const struct amd_ip_funcs dce_v11_0_ip_funcs; +void dce_v11_0_disable_dce(struct amdgpu_device *adev); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index 8b7ad345771f..abd5213dfe18 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -604,6 +604,52 @@ static void dce_v8_0_set_vga_render_state(struct amdgpu_device *adev, WREG32(mmVGA_RENDER_CONTROL, tmp); } +static int dce_v8_0_get_num_crtc(struct amdgpu_device *adev) +{ + int num_crtc = 0; + + switch (adev->asic_type) { + case CHIP_BONAIRE: + case CHIP_HAWAII: + num_crtc = 6; + break; + case CHIP_KAVERI: + num_crtc = 4; + break; + case CHIP_KABINI: + case CHIP_MULLINS: + num_crtc = 2; + break; + default: + num_crtc = 0; + } + return num_crtc; +} + +void dce_v8_0_disable_dce(struct amdgpu_device *adev) +{ + /*Disable VGA render and enabled crtc, if has DCE engine*/ + if (amdgpu_atombios_has_dce_engine_info(adev)) { + u32 tmp; + int crtc_enabled, i; + + dce_v8_0_set_vga_render_state(adev, false); + + /*Disable crtc*/ + for (i = 0; i < dce_v8_0_get_num_crtc(adev); i++) { + crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]), + CRTC_CONTROL, CRTC_MASTER_EN); + if (crtc_enabled) { + WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1); + tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]); + tmp = REG_SET_FIELD(tmp, CRTC_CONTROL, CRTC_MASTER_EN, 0); + WREG32(mmCRTC_CONTROL + crtc_offsets[i], tmp); + WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0); + } + } + } +} + static void dce_v8_0_program_fmt(struct drm_encoder *encoder) { struct drm_device *dev = encoder->dev; @@ -1501,13 +1547,13 @@ static void dce_v8_0_audio_write_sad_regs(struct drm_encoder *encoder) if (sad->format == eld_reg_to_type[i][1]) { if (sad->channels > max_channels) { - value = (sad->channels << - AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__MAX_CHANNELS__SHIFT) | - (sad->byte2 << - AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__DESCRIPTOR_BYTE_2__SHIFT) | - (sad->freq << - AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES__SHIFT); - max_channels = sad->channels; + value = (sad->channels << + AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__MAX_CHANNELS__SHIFT) | + (sad->byte2 << + AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__DESCRIPTOR_BYTE_2__SHIFT) | + (sad->freq << + AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES__SHIFT); + max_channels = sad->channels; } if (sad->format == HDMI_AUDIO_CODING_TYPE_PCM) @@ -1613,7 +1659,7 @@ static void dce_v8_0_afmt_update_ACR(struct drm_encoder *encoder, uint32_t clock struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv; uint32_t offset = dig->afmt->offset; - WREG32(mmHDMI_ACR_32_0 + offset, (acr.cts_32khz << HDMI_ACR_44_0__HDMI_ACR_CTS_44__SHIFT)); + WREG32(mmHDMI_ACR_32_0 + offset, (acr.cts_32khz << HDMI_ACR_32_0__HDMI_ACR_CTS_32__SHIFT)); WREG32(mmHDMI_ACR_32_1 + offset, acr.n_32khz); WREG32(mmHDMI_ACR_44_0 + offset, (acr.cts_44_1khz << HDMI_ACR_44_0__HDMI_ACR_CTS_44__SHIFT)); @@ -1693,6 +1739,7 @@ static void dce_v8_0_afmt_setmode(struct drm_encoder *encoder, /* Silent, r600_hdmi_enable will raise WARN for us */ if (!dig->afmt->enabled) return; + offset = dig->afmt->offset; /* hdmi deep color mode general control packets setup, if bpc > 8 */ @@ -1817,7 +1864,7 @@ static void dce_v8_0_afmt_setmode(struct drm_encoder *encoder, WREG32_OR(mmHDMI_INFOFRAME_CONTROL0 + offset, HDMI_INFOFRAME_CONTROL0__HDMI_AVI_INFO_SEND_MASK | /* enable AVI info frames */ - HDMI_INFOFRAME_CONTROL0__HDMI_AVI_INFO_SEND_MASK); /* required for audio info values to be updated */ + HDMI_INFOFRAME_CONTROL0__HDMI_AVI_INFO_CONT_MASK); /* required for audio info values to be updated */ WREG32_P(mmHDMI_INFOFRAME_CONTROL1 + offset, (2 << HDMI_INFOFRAME_CONTROL1__HDMI_AVI_INFO_LINE__SHIFT), /* anything other than 0 */ @@ -1826,13 +1873,12 @@ static void dce_v8_0_afmt_setmode(struct drm_encoder *encoder, WREG32_OR(mmAFMT_AUDIO_PACKET_CONTROL + offset, AFMT_AUDIO_PACKET_CONTROL__AFMT_AUDIO_SAMPLE_SEND_MASK); /* send audio packets */ - /* it's unknown what these bits do excatly, but it's indeed quite useful for debugging */ WREG32(mmAFMT_RAMP_CONTROL0 + offset, 0x00FFFFFF); WREG32(mmAFMT_RAMP_CONTROL1 + offset, 0x007FFFFF); WREG32(mmAFMT_RAMP_CONTROL2 + offset, 0x00000001); WREG32(mmAFMT_RAMP_CONTROL3 + offset, 0x00000001); - /* enable audio after to setting up hw */ + /* enable audio after setting up hw */ dce_v8_0_audio_enable(adev, dig->afmt->pin, true); } @@ -2000,7 +2046,7 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc, case DRM_FORMAT_XRGB4444: case DRM_FORMAT_ARGB4444: fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) | - (GRPH_FORMAT_ARGB1555 << GRPH_CONTROL__GRPH_FORMAT__SHIFT)); + (GRPH_FORMAT_ARGB4444 << GRPH_CONTROL__GRPH_FORMAT__SHIFT)); #ifdef __BIG_ENDIAN fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT); #endif @@ -2139,8 +2185,8 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc, WREG32(mmVIEWPORT_SIZE + amdgpu_crtc->crtc_offset, (viewport_w << 16) | viewport_h); - /* set pageflip to happen only at start of vblank interval (front porch) */ - WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 3); + /* set pageflip to happen anywhere in vblank interval */ + WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { amdgpu_fb = to_amdgpu_framebuffer(fb); @@ -2554,7 +2600,7 @@ static const struct drm_crtc_funcs dce_v8_0_crtc_funcs = { .gamma_set = dce_v8_0_crtc_gamma_set, .set_config = amdgpu_crtc_set_config, .destroy = dce_v8_0_crtc_destroy, - .page_flip = amdgpu_crtc_page_flip, + .page_flip_target = amdgpu_crtc_page_flip_target, }; static void dce_v8_0_crtc_dpms(struct drm_crtc *crtc, int mode) @@ -2655,7 +2701,7 @@ static void dce_v8_0_crtc_disable(struct drm_crtc *crtc) case ATOM_PPLL2: /* disable the ppll */ amdgpu_atombios_crtc_program_pll(crtc, amdgpu_crtc->crtc_id, amdgpu_crtc->pll_id, - 0, 0, ATOM_DISABLE, 0, 0, 0, 0, 0, false, &ss); + 0, 0, ATOM_DISABLE, 0, 0, 0, 0, 0, false, &ss); break; case ATOM_PPLL0: /* disable the ppll */ @@ -2805,21 +2851,20 @@ static int dce_v8_0_early_init(void *handle) dce_v8_0_set_display_funcs(adev); dce_v8_0_set_irq_funcs(adev); + adev->mode_info.num_crtc = dce_v8_0_get_num_crtc(adev); + switch (adev->asic_type) { case CHIP_BONAIRE: case CHIP_HAWAII: - adev->mode_info.num_crtc = 6; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 6; break; case CHIP_KAVERI: - adev->mode_info.num_crtc = 4; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 7; break; case CHIP_KABINI: case CHIP_MULLINS: - adev->mode_info.num_crtc = 2; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 6; /* ? */ break; @@ -3238,7 +3283,6 @@ static int dce_v8_0_crtc_irq(struct amdgpu_device *adev, drm_handle_vblank(adev->ddev, crtc); } DRM_DEBUG("IH: D%d vblank\n", crtc + 1); - break; case 1: /* vline */ if (disp_int & interrupt_status_offsets[crtc].vline) @@ -3247,7 +3291,6 @@ static int dce_v8_0_crtc_irq(struct amdgpu_device *adev, DRM_DEBUG("IH: IH event w/o asserted irq bit?\n"); DRM_DEBUG("IH: D%d vline\n", crtc + 1); - break; default: DRM_DEBUG("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data); diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.h b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.h index 77016852b252..7d0770c3a49b 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.h +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.h @@ -26,4 +26,6 @@ extern const struct amd_ip_funcs dce_v8_0_ip_funcs; +void dce_v8_0_disable_dce(struct amdgpu_device *adev); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c new file mode 100644 index 000000000000..00663a7b4053 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c @@ -0,0 +1,806 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "drmP.h" +#include "amdgpu.h" +#include "amdgpu_pm.h" +#include "amdgpu_i2c.h" +#include "atom.h" +#include "amdgpu_pll.h" +#include "amdgpu_connectors.h" +#ifdef CONFIG_DRM_AMDGPU_CIK +#include "dce_v8_0.h" +#endif +#include "dce_v10_0.h" +#include "dce_v11_0.h" +#include "dce_virtual.h" + +static void dce_virtual_set_display_funcs(struct amdgpu_device *adev); +static void dce_virtual_set_irq_funcs(struct amdgpu_device *adev); +static int dce_virtual_pageflip_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry); + +/** + * dce_virtual_vblank_wait - vblank wait asic callback. + * + * @adev: amdgpu_device pointer + * @crtc: crtc to wait for vblank on + * + * Wait for vblank on the requested crtc (evergreen+). + */ +static void dce_virtual_vblank_wait(struct amdgpu_device *adev, int crtc) +{ + return; +} + +static u32 dce_virtual_vblank_get_counter(struct amdgpu_device *adev, int crtc) +{ + return 0; +} + +static void dce_virtual_page_flip(struct amdgpu_device *adev, + int crtc_id, u64 crtc_base, bool async) +{ + return; +} + +static int dce_virtual_crtc_get_scanoutpos(struct amdgpu_device *adev, int crtc, + u32 *vbl, u32 *position) +{ + *vbl = 0; + *position = 0; + + return -EINVAL; +} + +static bool dce_virtual_hpd_sense(struct amdgpu_device *adev, + enum amdgpu_hpd_id hpd) +{ + return true; +} + +static void dce_virtual_hpd_set_polarity(struct amdgpu_device *adev, + enum amdgpu_hpd_id hpd) +{ + return; +} + +static u32 dce_virtual_hpd_get_gpio_reg(struct amdgpu_device *adev) +{ + return 0; +} + +static bool dce_virtual_is_display_hung(struct amdgpu_device *adev) +{ + return false; +} + +void dce_virtual_stop_mc_access(struct amdgpu_device *adev, + struct amdgpu_mode_mc_save *save) +{ + switch (adev->asic_type) { + case CHIP_BONAIRE: + case CHIP_HAWAII: + case CHIP_KAVERI: + case CHIP_KABINI: + case CHIP_MULLINS: +#ifdef CONFIG_DRM_AMDGPU_CIK + dce_v8_0_disable_dce(adev); +#endif + break; + case CHIP_FIJI: + case CHIP_TONGA: + dce_v10_0_disable_dce(adev); + break; + case CHIP_CARRIZO: + case CHIP_STONEY: + case CHIP_POLARIS11: + case CHIP_POLARIS10: + dce_v11_0_disable_dce(adev); + break; + case CHIP_TOPAZ: + /* no DCE */ + return; + default: + DRM_ERROR("Virtual display unsupported ASIC type: 0x%X\n", adev->asic_type); + } + + return; +} +void dce_virtual_resume_mc_access(struct amdgpu_device *adev, + struct amdgpu_mode_mc_save *save) +{ + return; +} + +void dce_virtual_set_vga_render_state(struct amdgpu_device *adev, + bool render) +{ + return; +} + +/** + * dce_virtual_bandwidth_update - program display watermarks + * + * @adev: amdgpu_device pointer + * + * Calculate and program the display watermarks and line + * buffer allocation (CIK). + */ +static void dce_virtual_bandwidth_update(struct amdgpu_device *adev) +{ + return; +} + +static int dce_virtual_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, + u16 *green, u16 *blue, uint32_t size) +{ + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + int i; + + /* userspace palettes are always correct as is */ + for (i = 0; i < size; i++) { + amdgpu_crtc->lut_r[i] = red[i] >> 6; + amdgpu_crtc->lut_g[i] = green[i] >> 6; + amdgpu_crtc->lut_b[i] = blue[i] >> 6; + } + + return 0; +} + +static void dce_virtual_crtc_destroy(struct drm_crtc *crtc) +{ + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + + drm_crtc_cleanup(crtc); + kfree(amdgpu_crtc); +} + +static const struct drm_crtc_funcs dce_virtual_crtc_funcs = { + .cursor_set2 = NULL, + .cursor_move = NULL, + .gamma_set = dce_virtual_crtc_gamma_set, + .set_config = amdgpu_crtc_set_config, + .destroy = dce_virtual_crtc_destroy, + .page_flip_target = amdgpu_crtc_page_flip_target, +}; + +static void dce_virtual_crtc_dpms(struct drm_crtc *crtc, int mode) +{ + struct drm_device *dev = crtc->dev; + struct amdgpu_device *adev = dev->dev_private; + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + unsigned type; + + switch (mode) { + case DRM_MODE_DPMS_ON: + amdgpu_crtc->enabled = true; + /* Make sure VBLANK and PFLIP interrupts are still enabled */ + type = amdgpu_crtc_idx_to_irq_type(adev, amdgpu_crtc->crtc_id); + amdgpu_irq_update(adev, &adev->crtc_irq, type); + amdgpu_irq_update(adev, &adev->pageflip_irq, type); + drm_vblank_on(dev, amdgpu_crtc->crtc_id); + break; + case DRM_MODE_DPMS_STANDBY: + case DRM_MODE_DPMS_SUSPEND: + case DRM_MODE_DPMS_OFF: + drm_vblank_off(dev, amdgpu_crtc->crtc_id); + amdgpu_crtc->enabled = false; + break; + } +} + + +static void dce_virtual_crtc_prepare(struct drm_crtc *crtc) +{ + dce_virtual_crtc_dpms(crtc, DRM_MODE_DPMS_OFF); +} + +static void dce_virtual_crtc_commit(struct drm_crtc *crtc) +{ + dce_virtual_crtc_dpms(crtc, DRM_MODE_DPMS_ON); +} + +static void dce_virtual_crtc_disable(struct drm_crtc *crtc) +{ + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + + dce_virtual_crtc_dpms(crtc, DRM_MODE_DPMS_OFF); + if (crtc->primary->fb) { + int r; + struct amdgpu_framebuffer *amdgpu_fb; + struct amdgpu_bo *rbo; + + amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); + rbo = gem_to_amdgpu_bo(amdgpu_fb->obj); + r = amdgpu_bo_reserve(rbo, false); + if (unlikely(r)) + DRM_ERROR("failed to reserve rbo before unpin\n"); + else { + amdgpu_bo_unpin(rbo); + amdgpu_bo_unreserve(rbo); + } + } + + amdgpu_crtc->pll_id = ATOM_PPLL_INVALID; + amdgpu_crtc->encoder = NULL; + amdgpu_crtc->connector = NULL; +} + +static int dce_virtual_crtc_mode_set(struct drm_crtc *crtc, + struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode, + int x, int y, struct drm_framebuffer *old_fb) +{ + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + + /* update the hw version fpr dpm */ + amdgpu_crtc->hw_mode = *adjusted_mode; + + return 0; +} + +static bool dce_virtual_crtc_mode_fixup(struct drm_crtc *crtc, + const struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode) +{ + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + struct drm_device *dev = crtc->dev; + struct drm_encoder *encoder; + + /* assign the encoder to the amdgpu crtc to avoid repeated lookups later */ + list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { + if (encoder->crtc == crtc) { + amdgpu_crtc->encoder = encoder; + amdgpu_crtc->connector = amdgpu_get_connector_for_encoder(encoder); + break; + } + } + if ((amdgpu_crtc->encoder == NULL) || (amdgpu_crtc->connector == NULL)) { + amdgpu_crtc->encoder = NULL; + amdgpu_crtc->connector = NULL; + return false; + } + + return true; +} + + +static int dce_virtual_crtc_set_base(struct drm_crtc *crtc, int x, int y, + struct drm_framebuffer *old_fb) +{ + return 0; +} + +static void dce_virtual_crtc_load_lut(struct drm_crtc *crtc) +{ + return; +} + +static int dce_virtual_crtc_set_base_atomic(struct drm_crtc *crtc, + struct drm_framebuffer *fb, + int x, int y, enum mode_set_atomic state) +{ + return 0; +} + +static const struct drm_crtc_helper_funcs dce_virtual_crtc_helper_funcs = { + .dpms = dce_virtual_crtc_dpms, + .mode_fixup = dce_virtual_crtc_mode_fixup, + .mode_set = dce_virtual_crtc_mode_set, + .mode_set_base = dce_virtual_crtc_set_base, + .mode_set_base_atomic = dce_virtual_crtc_set_base_atomic, + .prepare = dce_virtual_crtc_prepare, + .commit = dce_virtual_crtc_commit, + .load_lut = dce_virtual_crtc_load_lut, + .disable = dce_virtual_crtc_disable, +}; + +static int dce_virtual_crtc_init(struct amdgpu_device *adev, int index) +{ + struct amdgpu_crtc *amdgpu_crtc; + int i; + + amdgpu_crtc = kzalloc(sizeof(struct amdgpu_crtc) + + (AMDGPUFB_CONN_LIMIT * sizeof(struct drm_connector *)), GFP_KERNEL); + if (amdgpu_crtc == NULL) + return -ENOMEM; + + drm_crtc_init(adev->ddev, &amdgpu_crtc->base, &dce_virtual_crtc_funcs); + + drm_mode_crtc_set_gamma_size(&amdgpu_crtc->base, 256); + amdgpu_crtc->crtc_id = index; + adev->mode_info.crtcs[index] = amdgpu_crtc; + + for (i = 0; i < 256; i++) { + amdgpu_crtc->lut_r[i] = i << 2; + amdgpu_crtc->lut_g[i] = i << 2; + amdgpu_crtc->lut_b[i] = i << 2; + } + + amdgpu_crtc->pll_id = ATOM_PPLL_INVALID; + amdgpu_crtc->encoder = NULL; + amdgpu_crtc->connector = NULL; + drm_crtc_helper_add(&amdgpu_crtc->base, &dce_virtual_crtc_helper_funcs); + + return 0; +} + +static int dce_virtual_early_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + adev->mode_info.vsync_timer_enabled = AMDGPU_IRQ_STATE_DISABLE; + dce_virtual_set_display_funcs(adev); + dce_virtual_set_irq_funcs(adev); + + adev->mode_info.num_crtc = 1; + adev->mode_info.num_hpd = 1; + adev->mode_info.num_dig = 1; + return 0; +} + +static bool dce_virtual_get_connector_info(struct amdgpu_device *adev) +{ + struct amdgpu_i2c_bus_rec ddc_bus; + struct amdgpu_router router; + struct amdgpu_hpd hpd; + + /* look up gpio for ddc, hpd */ + ddc_bus.valid = false; + hpd.hpd = AMDGPU_HPD_NONE; + /* needed for aux chan transactions */ + ddc_bus.hpd = hpd.hpd; + + memset(&router, 0, sizeof(router)); + router.ddc_valid = false; + router.cd_valid = false; + amdgpu_display_add_connector(adev, + 0, + ATOM_DEVICE_CRT1_SUPPORT, + DRM_MODE_CONNECTOR_VIRTUAL, &ddc_bus, + CONNECTOR_OBJECT_ID_VIRTUAL, + &hpd, + &router); + + amdgpu_display_add_encoder(adev, ENCODER_VIRTUAL_ENUM_VIRTUAL, + ATOM_DEVICE_CRT1_SUPPORT, + 0); + + amdgpu_link_encoder_connector(adev->ddev); + + return true; +} + +static int dce_virtual_sw_init(void *handle) +{ + int r, i; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = amdgpu_irq_add_id(adev, 229, &adev->crtc_irq); + if (r) + return r; + + adev->ddev->max_vblank_count = 0; + + adev->ddev->mode_config.funcs = &amdgpu_mode_funcs; + + adev->ddev->mode_config.max_width = 16384; + adev->ddev->mode_config.max_height = 16384; + + adev->ddev->mode_config.preferred_depth = 24; + adev->ddev->mode_config.prefer_shadow = 1; + + adev->ddev->mode_config.fb_base = adev->mc.aper_base; + + r = amdgpu_modeset_create_props(adev); + if (r) + return r; + + adev->ddev->mode_config.max_width = 16384; + adev->ddev->mode_config.max_height = 16384; + + /* allocate crtcs */ + for (i = 0; i < adev->mode_info.num_crtc; i++) { + r = dce_virtual_crtc_init(adev, i); + if (r) + return r; + } + + dce_virtual_get_connector_info(adev); + amdgpu_print_display_setup(adev->ddev); + + drm_kms_helper_poll_init(adev->ddev); + + adev->mode_info.mode_config_initialized = true; + return 0; +} + +static int dce_virtual_sw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + kfree(adev->mode_info.bios_hardcoded_edid); + + drm_kms_helper_poll_fini(adev->ddev); + + drm_mode_config_cleanup(adev->ddev); + adev->mode_info.mode_config_initialized = false; + return 0; +} + +static int dce_virtual_hw_init(void *handle) +{ + return 0; +} + +static int dce_virtual_hw_fini(void *handle) +{ + return 0; +} + +static int dce_virtual_suspend(void *handle) +{ + return dce_virtual_hw_fini(handle); +} + +static int dce_virtual_resume(void *handle) +{ + int ret; + + ret = dce_virtual_hw_init(handle); + + return ret; +} + +static bool dce_virtual_is_idle(void *handle) +{ + return true; +} + +static int dce_virtual_wait_for_idle(void *handle) +{ + return 0; +} + +static int dce_virtual_soft_reset(void *handle) +{ + return 0; +} + +static int dce_virtual_set_clockgating_state(void *handle, + enum amd_clockgating_state state) +{ + return 0; +} + +static int dce_virtual_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + return 0; +} + +const struct amd_ip_funcs dce_virtual_ip_funcs = { + .name = "dce_virtual", + .early_init = dce_virtual_early_init, + .late_init = NULL, + .sw_init = dce_virtual_sw_init, + .sw_fini = dce_virtual_sw_fini, + .hw_init = dce_virtual_hw_init, + .hw_fini = dce_virtual_hw_fini, + .suspend = dce_virtual_suspend, + .resume = dce_virtual_resume, + .is_idle = dce_virtual_is_idle, + .wait_for_idle = dce_virtual_wait_for_idle, + .soft_reset = dce_virtual_soft_reset, + .set_clockgating_state = dce_virtual_set_clockgating_state, + .set_powergating_state = dce_virtual_set_powergating_state, +}; + +/* these are handled by the primary encoders */ +static void dce_virtual_encoder_prepare(struct drm_encoder *encoder) +{ + return; +} + +static void dce_virtual_encoder_commit(struct drm_encoder *encoder) +{ + return; +} + +static void +dce_virtual_encoder_mode_set(struct drm_encoder *encoder, + struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode) +{ + return; +} + +static void dce_virtual_encoder_disable(struct drm_encoder *encoder) +{ + return; +} + +static void +dce_virtual_encoder_dpms(struct drm_encoder *encoder, int mode) +{ + return; +} + +static bool dce_virtual_encoder_mode_fixup(struct drm_encoder *encoder, + const struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode) +{ + + /* set the active encoder to connector routing */ + amdgpu_encoder_set_active_device(encoder); + + return true; +} + +static const struct drm_encoder_helper_funcs dce_virtual_encoder_helper_funcs = { + .dpms = dce_virtual_encoder_dpms, + .mode_fixup = dce_virtual_encoder_mode_fixup, + .prepare = dce_virtual_encoder_prepare, + .mode_set = dce_virtual_encoder_mode_set, + .commit = dce_virtual_encoder_commit, + .disable = dce_virtual_encoder_disable, +}; + +static void dce_virtual_encoder_destroy(struct drm_encoder *encoder) +{ + struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder); + + kfree(amdgpu_encoder->enc_priv); + drm_encoder_cleanup(encoder); + kfree(amdgpu_encoder); +} + +static const struct drm_encoder_funcs dce_virtual_encoder_funcs = { + .destroy = dce_virtual_encoder_destroy, +}; + +static void dce_virtual_encoder_add(struct amdgpu_device *adev, + uint32_t encoder_enum, + uint32_t supported_device, + u16 caps) +{ + struct drm_device *dev = adev->ddev; + struct drm_encoder *encoder; + struct amdgpu_encoder *amdgpu_encoder; + + /* see if we already added it */ + list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { + amdgpu_encoder = to_amdgpu_encoder(encoder); + if (amdgpu_encoder->encoder_enum == encoder_enum) { + amdgpu_encoder->devices |= supported_device; + return; + } + + } + + /* add a new one */ + amdgpu_encoder = kzalloc(sizeof(struct amdgpu_encoder), GFP_KERNEL); + if (!amdgpu_encoder) + return; + + encoder = &amdgpu_encoder->base; + encoder->possible_crtcs = 0x1; + amdgpu_encoder->enc_priv = NULL; + amdgpu_encoder->encoder_enum = encoder_enum; + amdgpu_encoder->encoder_id = (encoder_enum & OBJECT_ID_MASK) >> OBJECT_ID_SHIFT; + amdgpu_encoder->devices = supported_device; + amdgpu_encoder->rmx_type = RMX_OFF; + amdgpu_encoder->underscan_type = UNDERSCAN_OFF; + amdgpu_encoder->is_ext_encoder = false; + amdgpu_encoder->caps = caps; + + drm_encoder_init(dev, encoder, &dce_virtual_encoder_funcs, + DRM_MODE_ENCODER_VIRTUAL, NULL); + drm_encoder_helper_add(encoder, &dce_virtual_encoder_helper_funcs); + DRM_INFO("[FM]encoder: %d is VIRTUAL\n", amdgpu_encoder->encoder_id); +} + +static const struct amdgpu_display_funcs dce_virtual_display_funcs = { + .set_vga_render_state = &dce_virtual_set_vga_render_state, + .bandwidth_update = &dce_virtual_bandwidth_update, + .vblank_get_counter = &dce_virtual_vblank_get_counter, + .vblank_wait = &dce_virtual_vblank_wait, + .is_display_hung = &dce_virtual_is_display_hung, + .backlight_set_level = NULL, + .backlight_get_level = NULL, + .hpd_sense = &dce_virtual_hpd_sense, + .hpd_set_polarity = &dce_virtual_hpd_set_polarity, + .hpd_get_gpio_reg = &dce_virtual_hpd_get_gpio_reg, + .page_flip = &dce_virtual_page_flip, + .page_flip_get_scanoutpos = &dce_virtual_crtc_get_scanoutpos, + .add_encoder = &dce_virtual_encoder_add, + .add_connector = &amdgpu_connector_add, + .stop_mc_access = &dce_virtual_stop_mc_access, + .resume_mc_access = &dce_virtual_resume_mc_access, +}; + +static void dce_virtual_set_display_funcs(struct amdgpu_device *adev) +{ + if (adev->mode_info.funcs == NULL) + adev->mode_info.funcs = &dce_virtual_display_funcs; +} + +static enum hrtimer_restart dce_virtual_vblank_timer_handle(struct hrtimer *vblank_timer) +{ + struct amdgpu_mode_info *mode_info = container_of(vblank_timer, struct amdgpu_mode_info ,vblank_timer); + struct amdgpu_device *adev = container_of(mode_info, struct amdgpu_device ,mode_info); + unsigned crtc = 0; + drm_handle_vblank(adev->ddev, crtc); + dce_virtual_pageflip_irq(adev, NULL, NULL); + hrtimer_start(vblank_timer, ktime_set(0, DCE_VIRTUAL_VBLANK_PERIOD), HRTIMER_MODE_REL); + return HRTIMER_NORESTART; +} + +static void dce_virtual_set_crtc_vblank_interrupt_state(struct amdgpu_device *adev, + int crtc, + enum amdgpu_interrupt_state state) +{ + if (crtc >= adev->mode_info.num_crtc) { + DRM_DEBUG("invalid crtc %d\n", crtc); + return; + } + + if (state && !adev->mode_info.vsync_timer_enabled) { + DRM_DEBUG("Enable software vsync timer\n"); + hrtimer_init(&adev->mode_info.vblank_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); + hrtimer_set_expires(&adev->mode_info.vblank_timer, ktime_set(0, DCE_VIRTUAL_VBLANK_PERIOD)); + adev->mode_info.vblank_timer.function = dce_virtual_vblank_timer_handle; + hrtimer_start(&adev->mode_info.vblank_timer, ktime_set(0, DCE_VIRTUAL_VBLANK_PERIOD), HRTIMER_MODE_REL); + } else if (!state && adev->mode_info.vsync_timer_enabled) { + DRM_DEBUG("Disable software vsync timer\n"); + hrtimer_cancel(&adev->mode_info.vblank_timer); + } + + adev->mode_info.vsync_timer_enabled = state; + DRM_DEBUG("[FM]set crtc %d vblank interrupt state %d\n", crtc, state); +} + + +static int dce_virtual_set_crtc_irq_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned type, + enum amdgpu_interrupt_state state) +{ + switch (type) { + case AMDGPU_CRTC_IRQ_VBLANK1: + dce_virtual_set_crtc_vblank_interrupt_state(adev, 0, state); + break; + default: + break; + } + return 0; +} + +static void dce_virtual_crtc_vblank_int_ack(struct amdgpu_device *adev, + int crtc) +{ + if (crtc >= adev->mode_info.num_crtc) { + DRM_DEBUG("invalid crtc %d\n", crtc); + return; + } +} + +static int dce_virtual_crtc_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + unsigned crtc = 0; + unsigned irq_type = AMDGPU_CRTC_IRQ_VBLANK1; + + dce_virtual_crtc_vblank_int_ack(adev, crtc); + + if (amdgpu_irq_enabled(adev, source, irq_type)) { + drm_handle_vblank(adev->ddev, crtc); + } + dce_virtual_pageflip_irq(adev, NULL, NULL); + DRM_DEBUG("IH: D%d vblank\n", crtc + 1); + return 0; +} + +static int dce_virtual_set_pageflip_irq_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *src, + unsigned type, + enum amdgpu_interrupt_state state) +{ + if (type >= adev->mode_info.num_crtc) { + DRM_ERROR("invalid pageflip crtc %d\n", type); + return -EINVAL; + } + DRM_DEBUG("[FM]set pageflip irq type %d state %d\n", type, state); + + return 0; +} + +static int dce_virtual_pageflip_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + unsigned long flags; + unsigned crtc_id = 0; + struct amdgpu_crtc *amdgpu_crtc; + struct amdgpu_flip_work *works; + + crtc_id = 0; + amdgpu_crtc = adev->mode_info.crtcs[crtc_id]; + + if (crtc_id >= adev->mode_info.num_crtc) { + DRM_ERROR("invalid pageflip crtc %d\n", crtc_id); + return -EINVAL; + } + + /* IRQ could occur when in initial stage */ + if (amdgpu_crtc == NULL) + return 0; + + spin_lock_irqsave(&adev->ddev->event_lock, flags); + works = amdgpu_crtc->pflip_works; + if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED) { + DRM_DEBUG_DRIVER("amdgpu_crtc->pflip_status = %d != " + "AMDGPU_FLIP_SUBMITTED(%d)\n", + amdgpu_crtc->pflip_status, + AMDGPU_FLIP_SUBMITTED); + spin_unlock_irqrestore(&adev->ddev->event_lock, flags); + return 0; + } + + /* page flip completed. clean up */ + amdgpu_crtc->pflip_status = AMDGPU_FLIP_NONE; + amdgpu_crtc->pflip_works = NULL; + + /* wakeup usersapce */ + if (works->event) + drm_crtc_send_vblank_event(&amdgpu_crtc->base, works->event); + + spin_unlock_irqrestore(&adev->ddev->event_lock, flags); + + drm_crtc_vblank_put(&amdgpu_crtc->base); + schedule_work(&works->unpin_work); + + return 0; +} + +static const struct amdgpu_irq_src_funcs dce_virtual_crtc_irq_funcs = { + .set = dce_virtual_set_crtc_irq_state, + .process = dce_virtual_crtc_irq, +}; + +static const struct amdgpu_irq_src_funcs dce_virtual_pageflip_irq_funcs = { + .set = dce_virtual_set_pageflip_irq_state, + .process = dce_virtual_pageflip_irq, +}; + +static void dce_virtual_set_irq_funcs(struct amdgpu_device *adev) +{ + adev->crtc_irq.num_types = AMDGPU_CRTC_IRQ_LAST; + adev->crtc_irq.funcs = &dce_virtual_crtc_irq_funcs; + + adev->pageflip_irq.num_types = AMDGPU_PAGEFLIP_IRQ_LAST; + adev->pageflip_irq.funcs = &dce_virtual_pageflip_irq_funcs; +} + diff --git a/drivers/gpu/drm/amd/amdgpu/dce_virtual.h b/drivers/gpu/drm/amd/amdgpu/dce_virtual.h new file mode 100644 index 000000000000..e239243f6ebc --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/dce_virtual.h @@ -0,0 +1,31 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __DCE_VIRTUAL_H__ +#define __DCE_VIRTUAL_H__ + +extern const struct amd_ip_funcs dce_virtual_ip_funcs; +#define DCE_VIRTUAL_VBLANK_PERIOD 16666666 + +#endif + diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index d869d058ef24..f4fbec3e224e 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -4465,24 +4465,21 @@ static int gfx_v7_0_sw_init(void *handle) } /* reserve GDS, GWS and OA resource for gfx */ - r = amdgpu_bo_create(adev, adev->gds.mem.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GDS, 0, - NULL, NULL, &adev->gds.gds_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.mem.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GDS, + &adev->gds.gds_gfx_bo, NULL, NULL); if (r) return r; - r = amdgpu_bo_create(adev, adev->gds.gws.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GWS, 0, - NULL, NULL, &adev->gds.gws_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.gws.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GWS, + &adev->gds.gws_gfx_bo, NULL, NULL); if (r) return r; - r = amdgpu_bo_create(adev, adev->gds.oa.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_OA, 0, - NULL, NULL, &adev->gds.oa_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.oa.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_OA, + &adev->gds.oa_gfx_bo, NULL, NULL); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index b8184617ca25..c6a63c2f91e3 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -703,7 +703,10 @@ static void gfx_v8_0_init_golden_registers(struct amdgpu_device *adev) polaris10_golden_common_all, (const u32)ARRAY_SIZE(polaris10_golden_common_all)); WREG32_SMC(ixCG_ACLK_CNTL, 0x0000001C); - if (adev->pdev->revision == 0xc7) { + if (adev->pdev->revision == 0xc7 && + ((adev->pdev->subsystem_device == 0xb37 && adev->pdev->subsystem_vendor == 0x1002) || + (adev->pdev->subsystem_device == 0x4a8 && adev->pdev->subsystem_vendor == 0x1043) || + (adev->pdev->subsystem_device == 0x9480 && adev->pdev->subsystem_vendor == 0x1682))) { amdgpu_atombios_i2c_channel_trans(adev, 0x10, 0x96, 0x1E, 0xDD); amdgpu_atombios_i2c_channel_trans(adev, 0x10, 0x96, 0x1F, 0xD0); } @@ -1233,10 +1236,9 @@ static void gfx_v8_0_rlc_fini(struct amdgpu_device *adev) if (adev->gfx.rlc.clear_state_obj) { r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false); if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC c bo failed\n", r); + dev_warn(adev->dev, "(%d) reserve RLC cbs bo failed\n", r); amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj); amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj); adev->gfx.rlc.clear_state_obj = NULL; } @@ -1248,7 +1250,6 @@ static void gfx_v8_0_rlc_fini(struct amdgpu_device *adev) dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r); amdgpu_bo_unpin(adev->gfx.rlc.cp_table_obj); amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); - amdgpu_bo_unref(&adev->gfx.rlc.cp_table_obj); adev->gfx.rlc.cp_table_obj = NULL; } @@ -1290,14 +1291,14 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) &adev->gfx.rlc.clear_state_gpu_addr); if (r) { amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); - dev_warn(adev->dev, "(%d) pin RLC c bo failed\n", r); + dev_warn(adev->dev, "(%d) pin RLC cbs bo failed\n", r); gfx_v8_0_rlc_fini(adev); return r; } r = amdgpu_bo_kmap(adev->gfx.rlc.clear_state_obj, (void **)&adev->gfx.rlc.cs_ptr); if (r) { - dev_warn(adev->dev, "(%d) map RLC c bo failed\n", r); + dev_warn(adev->dev, "(%d) map RLC cbs bo failed\n", r); gfx_v8_0_rlc_fini(adev); return r; } @@ -1332,7 +1333,7 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) &adev->gfx.rlc.cp_table_gpu_addr); if (r) { amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); - dev_warn(adev->dev, "(%d) pin RLC cp_table bo failed\n", r); + dev_warn(adev->dev, "(%d) pin RLC cp table bo failed\n", r); return r; } r = amdgpu_bo_kmap(adev->gfx.rlc.cp_table_obj, (void **)&adev->gfx.rlc.cp_table_ptr); @@ -1345,7 +1346,6 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj); amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); - } return 0; @@ -1361,7 +1361,6 @@ static void gfx_v8_0_mec_fini(struct amdgpu_device *adev) dev_warn(adev->dev, "(%d) reserve HPD EOP bo failed\n", r); amdgpu_bo_unpin(adev->gfx.mec.hpd_eop_obj); amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj); - amdgpu_bo_unref(&adev->gfx.mec.hpd_eop_obj); adev->gfx.mec.hpd_eop_obj = NULL; } @@ -2082,24 +2081,21 @@ static int gfx_v8_0_sw_init(void *handle) } /* reserve GDS, GWS and OA resource for gfx */ - r = amdgpu_bo_create(adev, adev->gds.mem.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GDS, 0, NULL, - NULL, &adev->gds.gds_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.mem.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GDS, + &adev->gds.gds_gfx_bo, NULL, NULL); if (r) return r; - r = amdgpu_bo_create(adev, adev->gds.gws.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GWS, 0, NULL, - NULL, &adev->gds.gws_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.gws.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GWS, + &adev->gds.gws_gfx_bo, NULL, NULL); if (r) return r; - r = amdgpu_bo_create(adev, adev->gds.oa.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_OA, 0, NULL, - NULL, &adev->gds.oa_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.oa.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_OA, + &adev->gds.oa_gfx_bo, NULL, NULL); if (r) return r; @@ -2127,9 +2123,7 @@ static int gfx_v8_0_sw_fini(void *handle) amdgpu_ring_fini(&adev->gfx.compute_ring[i]); gfx_v8_0_mec_fini(adev); - gfx_v8_0_rlc_fini(adev); - gfx_v8_0_free_microcode(adev); return 0; @@ -3465,19 +3459,16 @@ static void gfx_v8_0_select_se_sh(struct amdgpu_device *adev, else data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_INDEX, instance); - if ((se_num == 0xffffffff) && (sh_num == 0xffffffff)) { - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_BROADCAST_WRITES, 1); + if (se_num == 0xffffffff) data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_BROADCAST_WRITES, 1); - } else if (se_num == 0xffffffff) { - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_INDEX, sh_num); - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_BROADCAST_WRITES, 1); - } else if (sh_num == 0xffffffff) { - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_BROADCAST_WRITES, 1); + else data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX, se_num); - } else { + + if (sh_num == 0xffffffff) + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_BROADCAST_WRITES, 1); + else data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_INDEX, sh_num); - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX, se_num); - } + WREG32(mmGRBM_GFX_INDEX, data); } @@ -3490,11 +3481,10 @@ static u32 gfx_v8_0_get_rb_active_bitmap(struct amdgpu_device *adev) { u32 data, mask; - data = RREG32(mmCC_RB_BACKEND_DISABLE); - data |= RREG32(mmGC_USER_RB_BACKEND_DISABLE); + data = RREG32(mmCC_RB_BACKEND_DISABLE) | + RREG32(mmGC_USER_RB_BACKEND_DISABLE); - data &= CC_RB_BACKEND_DISABLE__BACKEND_DISABLE_MASK; - data >>= GC_USER_RB_BACKEND_DISABLE__BACKEND_DISABLE__SHIFT; + data = REG_GET_FIELD(data, GC_USER_RB_BACKEND_DISABLE, BACKEND_DISABLE); mask = gfx_v8_0_create_bitmask(adev->gfx.config.max_backends_per_se / adev->gfx.config.max_sh_per_se); @@ -3576,16 +3566,12 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) u32 tmp; int i; - tmp = RREG32(mmGRBM_CNTL); - tmp = REG_SET_FIELD(tmp, GRBM_CNTL, READ_TIMEOUT, 0xff); - WREG32(mmGRBM_CNTL, tmp); - + WREG32_FIELD(GRBM_CNTL, READ_TIMEOUT, 0xFF); WREG32(mmGB_ADDR_CONFIG, adev->gfx.config.gb_addr_config); WREG32(mmHDP_ADDR_CONFIG, adev->gfx.config.gb_addr_config); WREG32(mmDMIF_ADDR_CALC, adev->gfx.config.gb_addr_config); gfx_v8_0_tiling_mode_table_init(adev); - gfx_v8_0_setup_rb(adev); gfx_v8_0_get_cu_info(adev); @@ -3769,9 +3755,7 @@ static int gfx_v8_0_init_save_restore_list(struct amdgpu_device *adev) sizeof(indirect_start_offsets)/sizeof(int)); /* save and restore list */ - temp = RREG32(mmRLC_SRM_CNTL); - temp |= RLC_SRM_CNTL__AUTO_INCR_ADDR_MASK; - WREG32(mmRLC_SRM_CNTL, temp); + WREG32_FIELD(RLC_SRM_CNTL, AUTO_INCR_ADDR, 1); WREG32(mmRLC_SRM_ARAM_ADDR, 0); for (i = 0; i < adev->gfx.rlc.reg_list_size_bytes >> 2; i++) @@ -3808,11 +3792,7 @@ static int gfx_v8_0_init_save_restore_list(struct amdgpu_device *adev) static void gfx_v8_0_enable_save_restore_machine(struct amdgpu_device *adev) { - uint32_t data; - - data = RREG32(mmRLC_SRM_CNTL); - data |= RLC_SRM_CNTL__SRM_ENABLE_MASK; - WREG32(mmRLC_SRM_CNTL, data); + WREG32_FIELD(RLC_SRM_CNTL, SRM_ENABLE, 1); } static void gfx_v8_0_init_power_gating(struct amdgpu_device *adev) @@ -3822,75 +3802,34 @@ static void gfx_v8_0_init_power_gating(struct amdgpu_device *adev) if (adev->pg_flags & (AMD_PG_SUPPORT_GFX_PG | AMD_PG_SUPPORT_GFX_SMG | AMD_PG_SUPPORT_GFX_DMG)) { - data = RREG32(mmCP_RB_WPTR_POLL_CNTL); - data &= ~CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT_MASK; - data |= (0x60 << CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT__SHIFT); - WREG32(mmCP_RB_WPTR_POLL_CNTL, data); - - data = 0; - data |= (0x10 << RLC_PG_DELAY__POWER_UP_DELAY__SHIFT); - data |= (0x10 << RLC_PG_DELAY__POWER_DOWN_DELAY__SHIFT); - data |= (0x10 << RLC_PG_DELAY__CMD_PROPAGATE_DELAY__SHIFT); - data |= (0x10 << RLC_PG_DELAY__MEM_SLEEP_DELAY__SHIFT); - WREG32(mmRLC_PG_DELAY, data); + WREG32_FIELD(CP_RB_WPTR_POLL_CNTL, IDLE_POLL_COUNT, 0x60); - data = RREG32(mmRLC_PG_DELAY_2); - data &= ~RLC_PG_DELAY_2__SERDES_CMD_DELAY_MASK; - data |= (0x3 << RLC_PG_DELAY_2__SERDES_CMD_DELAY__SHIFT); - WREG32(mmRLC_PG_DELAY_2, data); + data = REG_SET_FIELD(0, RLC_PG_DELAY, POWER_UP_DELAY, 0x10); + data = REG_SET_FIELD(data, RLC_PG_DELAY, POWER_DOWN_DELAY, 0x10); + data = REG_SET_FIELD(data, RLC_PG_DELAY, CMD_PROPAGATE_DELAY, 0x10); + data = REG_SET_FIELD(data, RLC_PG_DELAY, MEM_SLEEP_DELAY, 0x10); + WREG32(mmRLC_PG_DELAY, data); - data = RREG32(mmRLC_AUTO_PG_CTRL); - data &= ~RLC_AUTO_PG_CTRL__GRBM_REG_SAVE_GFX_IDLE_THRESHOLD_MASK; - data |= (0x55f0 << RLC_AUTO_PG_CTRL__GRBM_REG_SAVE_GFX_IDLE_THRESHOLD__SHIFT); - WREG32(mmRLC_AUTO_PG_CTRL, data); + WREG32_FIELD(RLC_PG_DELAY_2, SERDES_CMD_DELAY, 0x3); + WREG32_FIELD(RLC_AUTO_PG_CTRL, GRBM_REG_SAVE_GFX_IDLE_THRESHOLD, 0x55f0); } } static void cz_enable_sck_slow_down_on_power_up(struct amdgpu_device *adev, bool enable) { - u32 data, orig; - - orig = data = RREG32(mmRLC_PG_CNTL); - - if (enable) - data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK; - - if (orig != data) - WREG32(mmRLC_PG_CNTL, data); + WREG32_FIELD(RLC_PG_CNTL, SMU_CLK_SLOWDOWN_ON_PU_ENABLE, enable ? 1 : 0); } static void cz_enable_sck_slow_down_on_power_down(struct amdgpu_device *adev, bool enable) { - u32 data, orig; - - orig = data = RREG32(mmRLC_PG_CNTL); - - if (enable) - data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK; - - if (orig != data) - WREG32(mmRLC_PG_CNTL, data); + WREG32_FIELD(RLC_PG_CNTL, SMU_CLK_SLOWDOWN_ON_PD_ENABLE, enable ? 1 : 0); } static void cz_enable_cp_power_gating(struct amdgpu_device *adev, bool enable) { - u32 data, orig; - - orig = data = RREG32(mmRLC_PG_CNTL); - - if (enable) - data &= ~RLC_PG_CNTL__CP_PG_DISABLE_MASK; - else - data |= RLC_PG_CNTL__CP_PG_DISABLE_MASK; - - if (orig != data) - WREG32(mmRLC_PG_CNTL, data); + WREG32_FIELD(RLC_PG_CNTL, CP_PG_DISABLE, enable ? 1 : 0); } static void gfx_v8_0_init_pg(struct amdgpu_device *adev) @@ -3929,34 +3868,24 @@ static void gfx_v8_0_init_pg(struct amdgpu_device *adev) void gfx_v8_0_rlc_stop(struct amdgpu_device *adev) { - u32 tmp = RREG32(mmRLC_CNTL); - - tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 0); - WREG32(mmRLC_CNTL, tmp); + WREG32_FIELD(RLC_CNTL, RLC_ENABLE_F32, 0); gfx_v8_0_enable_gui_idle_interrupt(adev, false); - gfx_v8_0_wait_for_rlc_serdes(adev); } static void gfx_v8_0_rlc_reset(struct amdgpu_device *adev) { - u32 tmp = RREG32(mmGRBM_SOFT_RESET); - - tmp = REG_SET_FIELD(tmp, GRBM_SOFT_RESET, SOFT_RESET_RLC, 1); - WREG32(mmGRBM_SOFT_RESET, tmp); + WREG32_FIELD(GRBM_SOFT_RESET, SOFT_RESET_RLC, 1); udelay(50); - tmp = REG_SET_FIELD(tmp, GRBM_SOFT_RESET, SOFT_RESET_RLC, 0); - WREG32(mmGRBM_SOFT_RESET, tmp); + + WREG32_FIELD(GRBM_SOFT_RESET, SOFT_RESET_RLC, 0); udelay(50); } static void gfx_v8_0_rlc_start(struct amdgpu_device *adev) { - u32 tmp = RREG32(mmRLC_CNTL); - - tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 1); - WREG32(mmRLC_CNTL, tmp); + WREG32_FIELD(RLC_CNTL, RLC_ENABLE_F32, 1); /* carrizo do enable cp interrupt after cp inited */ if (!(adev->flags & AMD_IS_APU)) @@ -3998,14 +3927,13 @@ static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev) /* disable CG */ WREG32(mmRLC_CGCG_CGLS_CTRL, 0); if (adev->asic_type == CHIP_POLARIS11 || - adev->asic_type == CHIP_POLARIS10) + adev->asic_type == CHIP_POLARIS10) WREG32(mmRLC_CGCG_CGLS_CTRL_3D, 0); /* disable PG */ WREG32(mmRLC_PG_CNTL, 0); gfx_v8_0_rlc_reset(adev); - gfx_v8_0_init_pg(adev); if (!adev->pp_enabled) { @@ -4300,12 +4228,10 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev) gfx_v8_0_cp_gfx_start(adev); ring->ready = true; r = amdgpu_ring_test_ring(ring); - if (r) { + if (r) ring->ready = false; - return r; - } - return 0; + return r; } static void gfx_v8_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) @@ -4980,7 +4906,6 @@ static int gfx_v8_0_hw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; gfx_v8_0_init_golden_registers(adev); - gfx_v8_0_gpu_init(adev); r = gfx_v8_0_rlc_resume(adev); @@ -4988,8 +4913,6 @@ static int gfx_v8_0_hw_init(void *handle) return r; r = gfx_v8_0_cp_resume(adev); - if (r) - return r; return r; } @@ -5037,25 +4960,22 @@ static bool gfx_v8_0_is_idle(void *handle) static int gfx_v8_0_wait_for_idle(void *handle) { unsigned i; - u32 tmp; struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->usec_timeout; i++) { - /* read MC_STATUS */ - tmp = RREG32(mmGRBM_STATUS) & GRBM_STATUS__GUI_ACTIVE_MASK; - - if (!REG_GET_FIELD(tmp, GRBM_STATUS, GUI_ACTIVE)) + if (gfx_v8_0_is_idle(handle)) return 0; + udelay(1); } return -ETIMEDOUT; } -static int gfx_v8_0_soft_reset(void *handle) +static int gfx_v8_0_check_soft_reset(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; u32 grbm_soft_reset = 0, srbm_soft_reset = 0; u32 tmp; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* GRBM_STATUS */ tmp = RREG32(mmGRBM_STATUS); @@ -5064,16 +4984,12 @@ static int gfx_v8_0_soft_reset(void *handle) GRBM_STATUS__TA_BUSY_MASK | GRBM_STATUS__VGT_BUSY_MASK | GRBM_STATUS__DB_BUSY_MASK | GRBM_STATUS__CB_BUSY_MASK | GRBM_STATUS__GDS_BUSY_MASK | GRBM_STATUS__SPI_BUSY_MASK | - GRBM_STATUS__IA_BUSY_MASK | GRBM_STATUS__IA_BUSY_NO_DMA_MASK)) { + GRBM_STATUS__IA_BUSY_MASK | GRBM_STATUS__IA_BUSY_NO_DMA_MASK | + GRBM_STATUS__CP_BUSY_MASK | GRBM_STATUS__CP_COHERENCY_BUSY_MASK)) { grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP, 1); grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX, 1); - } - - if (tmp & (GRBM_STATUS__CP_BUSY_MASK | GRBM_STATUS__CP_COHERENCY_BUSY_MASK)) { - grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, - GRBM_SOFT_RESET, SOFT_RESET_CP, 1); srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_GRBM, 1); } @@ -5084,73 +5000,199 @@ static int gfx_v8_0_soft_reset(void *handle) grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_RLC, 1); + if (REG_GET_FIELD(tmp, GRBM_STATUS2, CPF_BUSY) || + REG_GET_FIELD(tmp, GRBM_STATUS2, CPC_BUSY) || + REG_GET_FIELD(tmp, GRBM_STATUS2, CPG_BUSY)) { + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPF, 1); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPC, 1); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPG, 1); + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, + SOFT_RESET_GRBM, 1); + } + /* SRBM_STATUS */ tmp = RREG32(mmSRBM_STATUS); if (REG_GET_FIELD(tmp, SRBM_STATUS, GRBM_RQ_PENDING)) srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_GRBM, 1); + if (REG_GET_FIELD(tmp, SRBM_STATUS, SEM_BUSY)) + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, + SRBM_SOFT_RESET, SOFT_RESET_SEM, 1); if (grbm_soft_reset || srbm_soft_reset) { - /* stop the rlc */ - gfx_v8_0_rlc_stop(adev); + adev->ip_block_status[AMD_IP_BLOCK_TYPE_GFX].hang = true; + adev->gfx.grbm_soft_reset = grbm_soft_reset; + adev->gfx.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_GFX].hang = false; + adev->gfx.grbm_soft_reset = 0; + adev->gfx.srbm_soft_reset = 0; + } + + return 0; +} +static void gfx_v8_0_inactive_hqd(struct amdgpu_device *adev, + struct amdgpu_ring *ring) +{ + int i; + + vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + if (RREG32(mmCP_HQD_ACTIVE) & CP_HQD_ACTIVE__ACTIVE_MASK) { + u32 tmp; + tmp = RREG32(mmCP_HQD_DEQUEUE_REQUEST); + tmp = REG_SET_FIELD(tmp, CP_HQD_DEQUEUE_REQUEST, + DEQUEUE_REQ, 2); + WREG32(mmCP_HQD_DEQUEUE_REQUEST, tmp); + for (i = 0; i < adev->usec_timeout; i++) { + if (!(RREG32(mmCP_HQD_ACTIVE) & CP_HQD_ACTIVE__ACTIVE_MASK)) + break; + udelay(1); + } + } +} + +static int gfx_v8_0_pre_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GFX].hang) + return 0; + + grbm_soft_reset = adev->gfx.grbm_soft_reset; + srbm_soft_reset = adev->gfx.srbm_soft_reset; + + /* stop the rlc */ + gfx_v8_0_rlc_stop(adev); + + if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX)) /* Disable GFX parsing/prefetching */ gfx_v8_0_cp_gfx_enable(adev, false); + if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPF) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPC) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPG)) { + int i; + + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; + + gfx_v8_0_inactive_hqd(adev, ring); + } /* Disable MEC parsing/prefetching */ gfx_v8_0_cp_compute_enable(adev, false); + } - if (grbm_soft_reset || srbm_soft_reset) { - tmp = RREG32(mmGMCON_DEBUG); - tmp = REG_SET_FIELD(tmp, - GMCON_DEBUG, GFX_STALL, 1); - tmp = REG_SET_FIELD(tmp, - GMCON_DEBUG, GFX_CLEAR, 1); - WREG32(mmGMCON_DEBUG, tmp); + return 0; +} - udelay(50); - } +static int gfx_v8_0_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + u32 tmp; - if (grbm_soft_reset) { - tmp = RREG32(mmGRBM_SOFT_RESET); - tmp |= grbm_soft_reset; - dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp); - WREG32(mmGRBM_SOFT_RESET, tmp); - tmp = RREG32(mmGRBM_SOFT_RESET); + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GFX].hang) + return 0; - udelay(50); + grbm_soft_reset = adev->gfx.grbm_soft_reset; + srbm_soft_reset = adev->gfx.srbm_soft_reset; - tmp &= ~grbm_soft_reset; - WREG32(mmGRBM_SOFT_RESET, tmp); - tmp = RREG32(mmGRBM_SOFT_RESET); - } + if (grbm_soft_reset || srbm_soft_reset) { + tmp = RREG32(mmGMCON_DEBUG); + tmp = REG_SET_FIELD(tmp, GMCON_DEBUG, GFX_STALL, 1); + tmp = REG_SET_FIELD(tmp, GMCON_DEBUG, GFX_CLEAR, 1); + WREG32(mmGMCON_DEBUG, tmp); + udelay(50); + } - if (srbm_soft_reset) { - tmp = RREG32(mmSRBM_SOFT_RESET); - tmp |= srbm_soft_reset; - dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); - WREG32(mmSRBM_SOFT_RESET, tmp); - tmp = RREG32(mmSRBM_SOFT_RESET); + if (grbm_soft_reset) { + tmp = RREG32(mmGRBM_SOFT_RESET); + tmp |= grbm_soft_reset; + dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp); + WREG32(mmGRBM_SOFT_RESET, tmp); + tmp = RREG32(mmGRBM_SOFT_RESET); - udelay(50); + udelay(50); - tmp &= ~srbm_soft_reset; - WREG32(mmSRBM_SOFT_RESET, tmp); - tmp = RREG32(mmSRBM_SOFT_RESET); - } + tmp &= ~grbm_soft_reset; + WREG32(mmGRBM_SOFT_RESET, tmp); + tmp = RREG32(mmGRBM_SOFT_RESET); + } - if (grbm_soft_reset || srbm_soft_reset) { - tmp = RREG32(mmGMCON_DEBUG); - tmp = REG_SET_FIELD(tmp, - GMCON_DEBUG, GFX_STALL, 0); - tmp = REG_SET_FIELD(tmp, - GMCON_DEBUG, GFX_CLEAR, 0); - WREG32(mmGMCON_DEBUG, tmp); - } + if (srbm_soft_reset) { + tmp = RREG32(mmSRBM_SOFT_RESET); + tmp |= srbm_soft_reset; + dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); - /* Wait a little for things to settle down */ udelay(50); + + tmp &= ~srbm_soft_reset; + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); } + + if (grbm_soft_reset || srbm_soft_reset) { + tmp = RREG32(mmGMCON_DEBUG); + tmp = REG_SET_FIELD(tmp, GMCON_DEBUG, GFX_STALL, 0); + tmp = REG_SET_FIELD(tmp, GMCON_DEBUG, GFX_CLEAR, 0); + WREG32(mmGMCON_DEBUG, tmp); + } + + /* Wait a little for things to settle down */ + udelay(50); + + return 0; +} + +static void gfx_v8_0_init_hqd(struct amdgpu_device *adev, + struct amdgpu_ring *ring) +{ + vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + WREG32(mmCP_HQD_DEQUEUE_REQUEST, 0); + WREG32(mmCP_HQD_PQ_RPTR, 0); + WREG32(mmCP_HQD_PQ_WPTR, 0); + vi_srbm_select(adev, 0, 0, 0, 0); +} + +static int gfx_v8_0_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GFX].hang) + return 0; + + grbm_soft_reset = adev->gfx.grbm_soft_reset; + srbm_soft_reset = adev->gfx.srbm_soft_reset; + + if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX)) + gfx_v8_0_cp_gfx_resume(adev); + + if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPF) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPC) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPG)) { + int i; + + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; + + gfx_v8_0_init_hqd(adev, ring); + } + gfx_v8_0_cp_compute_resume(adev); + } + gfx_v8_0_rlc_start(adev); + return 0; } @@ -5269,8 +5311,6 @@ static int gfx_v8_0_late_init(void *handle) static void gfx_v8_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *adev, bool enable) { - uint32_t data, temp; - if (adev->asic_type == CHIP_POLARIS11) /* Send msg to SMU via Powerplay */ amdgpu_set_powergating_state(adev, @@ -5278,83 +5318,35 @@ static void gfx_v8_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *ade enable ? AMD_PG_STATE_GATE : AMD_PG_STATE_UNGATE); - temp = data = RREG32(mmRLC_PG_CNTL); - /* Enable static MGPG */ - if (enable) - data |= RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK; - - if (temp != data) - WREG32(mmRLC_PG_CNTL, data); + WREG32_FIELD(RLC_PG_CNTL, STATIC_PER_CU_PG_ENABLE, enable ? 1 : 0); } static void gfx_v8_0_enable_gfx_dynamic_mg_power_gating(struct amdgpu_device *adev, bool enable) { - uint32_t data, temp; - - temp = data = RREG32(mmRLC_PG_CNTL); - /* Enable dynamic MGPG */ - if (enable) - data |= RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK; - - if (temp != data) - WREG32(mmRLC_PG_CNTL, data); + WREG32_FIELD(RLC_PG_CNTL, DYN_PER_CU_PG_ENABLE, enable ? 1 : 0); } static void polaris11_enable_gfx_quick_mg_power_gating(struct amdgpu_device *adev, bool enable) { - uint32_t data, temp; - - temp = data = RREG32(mmRLC_PG_CNTL); - /* Enable quick PG */ - if (enable) - data |= RLC_PG_CNTL__QUICK_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__QUICK_PG_ENABLE_MASK; - - if (temp != data) - WREG32(mmRLC_PG_CNTL, data); + WREG32_FIELD(RLC_PG_CNTL, QUICK_PG_ENABLE, enable ? 1 : 0); } static void cz_enable_gfx_cg_power_gating(struct amdgpu_device *adev, bool enable) { - u32 data, orig; - - orig = data = RREG32(mmRLC_PG_CNTL); - - if (enable) - data |= RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK; - - if (orig != data) - WREG32(mmRLC_PG_CNTL, data); + WREG32_FIELD(RLC_PG_CNTL, GFX_POWER_GATING_ENABLE, enable ? 1 : 0); } static void cz_enable_gfx_pipeline_power_gating(struct amdgpu_device *adev, bool enable) { - u32 data, orig; - - orig = data = RREG32(mmRLC_PG_CNTL); - - if (enable) - data |= RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK; - - if (orig != data) - WREG32(mmRLC_PG_CNTL, data); + WREG32_FIELD(RLC_PG_CNTL, GFX_PIPELINE_PG_ENABLE, enable ? 1 : 0); /* Read any GFX register to wake up GFX. */ if (!enable) - data = RREG32(mmDB_RENDER_CONTROL); + RREG32(mmDB_RENDER_CONTROL); } static void cz_update_gfx_cg_power_gating(struct amdgpu_device *adev, @@ -5430,15 +5422,15 @@ static void gfx_v8_0_send_serdes_cmd(struct amdgpu_device *adev, data = RREG32(mmRLC_SERDES_WR_CTRL); if (adev->asic_type == CHIP_STONEY) - data &= ~(RLC_SERDES_WR_CTRL__WRITE_COMMAND_MASK | - RLC_SERDES_WR_CTRL__READ_COMMAND_MASK | - RLC_SERDES_WR_CTRL__P1_SELECT_MASK | - RLC_SERDES_WR_CTRL__P2_SELECT_MASK | - RLC_SERDES_WR_CTRL__RDDATA_RESET_MASK | - RLC_SERDES_WR_CTRL__POWER_DOWN_MASK | - RLC_SERDES_WR_CTRL__POWER_UP_MASK | - RLC_SERDES_WR_CTRL__SHORT_FORMAT_MASK | - RLC_SERDES_WR_CTRL__SRBM_OVERRIDE_MASK); + data &= ~(RLC_SERDES_WR_CTRL__WRITE_COMMAND_MASK | + RLC_SERDES_WR_CTRL__READ_COMMAND_MASK | + RLC_SERDES_WR_CTRL__P1_SELECT_MASK | + RLC_SERDES_WR_CTRL__P2_SELECT_MASK | + RLC_SERDES_WR_CTRL__RDDATA_RESET_MASK | + RLC_SERDES_WR_CTRL__POWER_DOWN_MASK | + RLC_SERDES_WR_CTRL__POWER_UP_MASK | + RLC_SERDES_WR_CTRL__SHORT_FORMAT_MASK | + RLC_SERDES_WR_CTRL__SRBM_OVERRIDE_MASK); else data &= ~(RLC_SERDES_WR_CTRL__WRITE_COMMAND_MASK | RLC_SERDES_WR_CTRL__READ_COMMAND_MASK | @@ -5461,10 +5453,10 @@ static void gfx_v8_0_send_serdes_cmd(struct amdgpu_device *adev, #define MSG_ENTER_RLC_SAFE_MODE 1 #define MSG_EXIT_RLC_SAFE_MODE 0 - -#define RLC_GPR_REG2__REQ_MASK 0x00000001 -#define RLC_GPR_REG2__MESSAGE__SHIFT 0x00000001 -#define RLC_GPR_REG2__MESSAGE_MASK 0x0000001e +#define RLC_GPR_REG2__REQ_MASK 0x00000001 +#define RLC_GPR_REG2__REQ__SHIFT 0 +#define RLC_GPR_REG2__MESSAGE__SHIFT 0x00000001 +#define RLC_GPR_REG2__MESSAGE_MASK 0x0000001e static void cz_enter_rlc_safe_mode(struct amdgpu_device *adev) { @@ -5494,7 +5486,7 @@ static void cz_enter_rlc_safe_mode(struct amdgpu_device *adev) } for (i = 0; i < adev->usec_timeout; i++) { - if ((RREG32(mmRLC_GPR_REG2) & RLC_GPR_REG2__REQ_MASK) == 0) + if (!REG_GET_FIELD(RREG32(mmRLC_GPR_REG2), RLC_GPR_REG2, REQ)) break; udelay(1); } @@ -5522,7 +5514,7 @@ static void cz_exit_rlc_safe_mode(struct amdgpu_device *adev) } for (i = 0; i < adev->usec_timeout; i++) { - if ((RREG32(mmRLC_GPR_REG2) & RLC_GPR_REG2__REQ_MASK) == 0) + if (!REG_GET_FIELD(RREG32(mmRLC_GPR_REG2), RLC_GPR_REG2, REQ)) break; udelay(1); } @@ -5554,7 +5546,7 @@ static void iceland_enter_rlc_safe_mode(struct amdgpu_device *adev) } for (i = 0; i < adev->usec_timeout; i++) { - if ((RREG32(mmRLC_SAFE_MODE) & RLC_SAFE_MODE__CMD_MASK) == 0) + if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) break; udelay(1); } @@ -5581,7 +5573,7 @@ static void iceland_exit_rlc_safe_mode(struct amdgpu_device *adev) } for (i = 0; i < adev->usec_timeout; i++) { - if ((RREG32(mmRLC_SAFE_MODE) & RLC_SAFE_MODE__CMD_MASK) == 0) + if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) break; udelay(1); } @@ -5622,21 +5614,12 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev /* It is disabled by HW by default */ if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) { if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGLS) { - if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) { + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) /* 1 - RLC memory Light sleep */ - temp = data = RREG32(mmRLC_MEM_SLP_CNTL); - data |= RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK; - if (temp != data) - WREG32(mmRLC_MEM_SLP_CNTL, data); - } + WREG32_FIELD(RLC_MEM_SLP_CNTL, RLC_MEM_LS_EN, 1); - if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) { - /* 2 - CP memory Light sleep */ - temp = data = RREG32(mmCP_MEM_SLP_CNTL); - data |= CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK; - if (temp != data) - WREG32(mmCP_MEM_SLP_CNTL, data); - } + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) + WREG32_FIELD(CP_MEM_SLP_CNTL, CP_MEM_LS_EN, 1); } /* 3 - RLC_CGTT_MGCG_OVERRIDE */ @@ -5854,25 +5837,18 @@ static int gfx_v8_0_set_clockgating_state(void *handle, static u32 gfx_v8_0_ring_get_rptr_gfx(struct amdgpu_ring *ring) { - u32 rptr; - - rptr = ring->adev->wb.wb[ring->rptr_offs]; - - return rptr; + return ring->adev->wb.wb[ring->rptr_offs]; } static u32 gfx_v8_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - u32 wptr; if (ring->use_doorbell) /* XXX check if swapping is necessary on BE */ - wptr = ring->adev->wb.wb[ring->wptr_offs]; + return ring->adev->wb.wb[ring->wptr_offs]; else - wptr = RREG32(mmCP_RB0_WPTR); - - return wptr; + return RREG32(mmCP_RB0_WPTR); } static void gfx_v8_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) @@ -5971,9 +5947,9 @@ static void gfx_v8_0_ring_emit_ib_compute(struct amdgpu_ring *ring, amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); amdgpu_ring_write(ring, #ifdef __BIG_ENDIAN - (2 << 0) | + (2 << 0) | #endif - (ib->gpu_addr & 0xFFFFFFFC)); + (ib->gpu_addr & 0xFFFFFFFC)); amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF); amdgpu_ring_write(ring, control); } @@ -6118,33 +6094,14 @@ static void gfx_v8_0_ring_emit_fence_compute(struct amdgpu_ring *ring, static void gfx_v8_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev, enum amdgpu_interrupt_state state) { - u32 cp_int_cntl; - - switch (state) { - case AMDGPU_IRQ_STATE_DISABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - TIME_STAMP_INT_ENABLE, 0); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - case AMDGPU_IRQ_STATE_ENABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = - REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - TIME_STAMP_INT_ENABLE, 1); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - default: - break; - } + WREG32_FIELD(CP_INT_CNTL_RING0, TIME_STAMP_INT_ENABLE, + state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); } static void gfx_v8_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev, int me, int pipe, enum amdgpu_interrupt_state state) { - u32 mec_int_cntl, mec_int_cntl_reg; - /* * amdgpu controls only pipe 0 of MEC1. That's why this function only * handles the setting of interrupts for this specific pipe. All other @@ -6154,7 +6111,6 @@ static void gfx_v8_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev, if (me == 1) { switch (pipe) { case 0: - mec_int_cntl_reg = mmCP_ME1_PIPE0_INT_CNTL; break; default: DRM_DEBUG("invalid pipe %d\n", pipe); @@ -6165,22 +6121,8 @@ static void gfx_v8_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev, return; } - switch (state) { - case AMDGPU_IRQ_STATE_DISABLE: - mec_int_cntl = RREG32(mec_int_cntl_reg); - mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, - TIME_STAMP_INT_ENABLE, 0); - WREG32(mec_int_cntl_reg, mec_int_cntl); - break; - case AMDGPU_IRQ_STATE_ENABLE: - mec_int_cntl = RREG32(mec_int_cntl_reg); - mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, - TIME_STAMP_INT_ENABLE, 1); - WREG32(mec_int_cntl_reg, mec_int_cntl); - break; - default: - break; - } + WREG32_FIELD(CP_ME1_PIPE0_INT_CNTL, TIME_STAMP_INT_ENABLE, + state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); } static int gfx_v8_0_set_priv_reg_fault_state(struct amdgpu_device *adev, @@ -6188,24 +6130,8 @@ static int gfx_v8_0_set_priv_reg_fault_state(struct amdgpu_device *adev, unsigned type, enum amdgpu_interrupt_state state) { - u32 cp_int_cntl; - - switch (state) { - case AMDGPU_IRQ_STATE_DISABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - PRIV_REG_INT_ENABLE, 0); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - case AMDGPU_IRQ_STATE_ENABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - PRIV_REG_INT_ENABLE, 1); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - default: - break; - } + WREG32_FIELD(CP_INT_CNTL_RING0, PRIV_REG_INT_ENABLE, + state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); return 0; } @@ -6215,24 +6141,8 @@ static int gfx_v8_0_set_priv_inst_fault_state(struct amdgpu_device *adev, unsigned type, enum amdgpu_interrupt_state state) { - u32 cp_int_cntl; - - switch (state) { - case AMDGPU_IRQ_STATE_DISABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - PRIV_INSTR_INT_ENABLE, 0); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - case AMDGPU_IRQ_STATE_ENABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - PRIV_INSTR_INT_ENABLE, 1); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - default: - break; - } + WREG32_FIELD(CP_INT_CNTL_RING0, PRIV_INSTR_INT_ENABLE, + state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); return 0; } @@ -6338,7 +6248,10 @@ const struct amd_ip_funcs gfx_v8_0_ip_funcs = { .resume = gfx_v8_0_resume, .is_idle = gfx_v8_0_is_idle, .wait_for_idle = gfx_v8_0_wait_for_idle, + .check_soft_reset = gfx_v8_0_check_soft_reset, + .pre_soft_reset = gfx_v8_0_pre_soft_reset, .soft_reset = gfx_v8_0_soft_reset, + .post_soft_reset = gfx_v8_0_post_soft_reset, .set_clockgating_state = gfx_v8_0_set_clockgating_state, .set_powergating_state = gfx_v8_0_set_powergating_state, }; @@ -6479,15 +6392,12 @@ static u32 gfx_v8_0_get_cu_active_bitmap(struct amdgpu_device *adev) { u32 data, mask; - data = RREG32(mmCC_GC_SHADER_ARRAY_CONFIG); - data |= RREG32(mmGC_USER_SHADER_ARRAY_CONFIG); - - data &= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK; - data >>= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT; + data = RREG32(mmCC_GC_SHADER_ARRAY_CONFIG) | + RREG32(mmGC_USER_SHADER_ARRAY_CONFIG); mask = gfx_v8_0_create_bitmask(adev->gfx.config.max_cu_per_sh); - return (~data) & mask; + return ~REG_GET_FIELD(data, CC_GC_SHADER_ARRAY_CONFIG, INACTIVE_CUS) & mask; } static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h index bc82c794312c..ebed1f829297 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h @@ -26,6 +26,4 @@ extern const struct amd_ip_funcs gfx_v8_0_ip_funcs; -void gfx_v8_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num); - #endif diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index 0b0f08641eed..aa0c4b964621 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -183,7 +183,7 @@ static int gmc_v7_0_mc_load_microcode(struct amdgpu_device *adev) const struct mc_firmware_header_v1_0 *hdr; const __le32 *fw_data = NULL; const __le32 *io_mc_regs = NULL; - u32 running, blackout = 0; + u32 running; int i, ucode_size, regs_size; if (!adev->mc.fw) @@ -203,11 +203,6 @@ static int gmc_v7_0_mc_load_microcode(struct amdgpu_device *adev) running = REG_GET_FIELD(RREG32(mmMC_SEQ_SUP_CNTL), MC_SEQ_SUP_CNTL, RUN); if (running == 0) { - if (running) { - blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL); - WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout | 1); - } - /* reset the engine and set to writable */ WREG32(mmMC_SEQ_SUP_CNTL, 0x00000008); WREG32(mmMC_SEQ_SUP_CNTL, 0x00000010); @@ -239,9 +234,6 @@ static int gmc_v7_0_mc_load_microcode(struct amdgpu_device *adev) break; udelay(1); } - - if (running) - WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout); } return 0; @@ -393,7 +385,7 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev) * size equal to the 1024 or vram, whichever is larger. */ if (amdgpu_gart_size == -1) - adev->mc.gtt_size = max((1024ULL << 20), adev->mc.mc_vram_size); + adev->mc.gtt_size = amdgpu_ttm_get_gtt_mem_size(adev); else adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20; @@ -953,6 +945,11 @@ static int gmc_v7_0_sw_init(void *handle) return r; } + r = amdgpu_ttm_global_init(adev); + if (r) { + return r; + } + r = gmc_v7_0_mc_init(adev); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index 2aee2c6f3cd5..84c10d5117a9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -261,7 +261,7 @@ static int gmc_v8_0_mc_load_microcode(struct amdgpu_device *adev) const struct mc_firmware_header_v1_0 *hdr; const __le32 *fw_data = NULL; const __le32 *io_mc_regs = NULL; - u32 running, blackout = 0; + u32 running; int i, ucode_size, regs_size; if (!adev->mc.fw) @@ -287,11 +287,6 @@ static int gmc_v8_0_mc_load_microcode(struct amdgpu_device *adev) running = REG_GET_FIELD(RREG32(mmMC_SEQ_SUP_CNTL), MC_SEQ_SUP_CNTL, RUN); if (running == 0) { - if (running) { - blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL); - WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout | 1); - } - /* reset the engine and set to writable */ WREG32(mmMC_SEQ_SUP_CNTL, 0x00000008); WREG32(mmMC_SEQ_SUP_CNTL, 0x00000010); @@ -323,9 +318,6 @@ static int gmc_v8_0_mc_load_microcode(struct amdgpu_device *adev) break; udelay(1); } - - if (running) - WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout); } return 0; @@ -477,7 +469,7 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev) * size equal to the 1024 or vram, whichever is larger. */ if (amdgpu_gart_size == -1) - adev->mc.gtt_size = max((1024ULL << 20), adev->mc.mc_vram_size); + adev->mc.gtt_size = amdgpu_ttm_get_gtt_mem_size(adev); else adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20; @@ -957,6 +949,11 @@ static int gmc_v8_0_sw_init(void *handle) return r; } + r = amdgpu_ttm_global_init(adev); + if (r) { + return r; + } + r = gmc_v8_0_mc_init(adev); if (r) return r; @@ -1100,9 +1097,8 @@ static int gmc_v8_0_wait_for_idle(void *handle) } -static int gmc_v8_0_soft_reset(void *handle) +static int gmc_v8_0_check_soft_reset(void *handle) { - struct amdgpu_mode_mc_save save; u32 srbm_soft_reset = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; u32 tmp = RREG32(mmSRBM_STATUS); @@ -1117,13 +1113,42 @@ static int gmc_v8_0_soft_reset(void *handle) srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_MC, 1); } - if (srbm_soft_reset) { - gmc_v8_0_mc_stop(adev, &save); - if (gmc_v8_0_wait_for_idle((void *)adev)) { - dev_warn(adev->dev, "Wait for GMC idle timed out !\n"); - } + adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang = true; + adev->mc.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang = false; + adev->mc.srbm_soft_reset = 0; + } + return 0; +} +static int gmc_v8_0_pre_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang) + return 0; + + gmc_v8_0_mc_stop(adev, &adev->mc.save); + if (gmc_v8_0_wait_for_idle(adev)) { + dev_warn(adev->dev, "Wait for GMC idle timed out !\n"); + } + + return 0; +} + +static int gmc_v8_0_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang) + return 0; + srbm_soft_reset = adev->mc.srbm_soft_reset; + + if (srbm_soft_reset) { + u32 tmp; tmp = RREG32(mmSRBM_SOFT_RESET); tmp |= srbm_soft_reset; @@ -1139,14 +1164,22 @@ static int gmc_v8_0_soft_reset(void *handle) /* Wait a little for things to settle down */ udelay(50); - - gmc_v8_0_mc_resume(adev, &save); - udelay(50); } return 0; } +static int gmc_v8_0_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang) + return 0; + + gmc_v8_0_mc_resume(adev, &adev->mc.save); + return 0; +} + static int gmc_v8_0_vm_fault_interrupt_state(struct amdgpu_device *adev, struct amdgpu_irq_src *src, unsigned type, @@ -1414,7 +1447,10 @@ const struct amd_ip_funcs gmc_v8_0_ip_funcs = { .resume = gmc_v8_0_resume, .is_idle = gmc_v8_0_is_idle, .wait_for_idle = gmc_v8_0_wait_for_idle, + .check_soft_reset = gmc_v8_0_check_soft_reset, + .pre_soft_reset = gmc_v8_0_pre_soft_reset, .soft_reset = gmc_v8_0_soft_reset, + .post_soft_reset = gmc_v8_0_post_soft_reset, .set_clockgating_state = gmc_v8_0_set_clockgating_state, .set_powergating_state = gmc_v8_0_set_powergating_state, }; diff --git a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c index a845e883f5fa..f8618a3881a8 100644 --- a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c @@ -2845,7 +2845,11 @@ static int kv_dpm_init(struct amdgpu_device *adev) pi->caps_tcp_ramping = true; } - pi->caps_sclk_ds = true; + if (amdgpu_sclk_deep_sleep_en) + pi->caps_sclk_ds = true; + else + pi->caps_sclk_ds = false; + pi->enable_auto_thermal_throttling = true; pi->disable_nb_ps3_in_battery = false; if (amdgpu_bapm == 0) diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index 1351c7e834a2..e82229686783 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -749,24 +749,16 @@ static void sdma_v2_4_vm_copy_pte(struct amdgpu_ib *ib, uint64_t pe, uint64_t src, unsigned count) { - while (count) { - unsigned bytes = count * 8; - if (bytes > 0x1FFFF8) - bytes = 0x1FFFF8; - - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) | - SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); - ib->ptr[ib->length_dw++] = bytes; - ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ - ib->ptr[ib->length_dw++] = lower_32_bits(src); - ib->ptr[ib->length_dw++] = upper_32_bits(src); - ib->ptr[ib->length_dw++] = lower_32_bits(pe); - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - - pe += bytes; - src += bytes; - count -= bytes / 8; - } + unsigned bytes = count * 8; + + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) | + SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); + ib->ptr[ib->length_dw++] = bytes; + ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ + ib->ptr[ib->length_dw++] = lower_32_bits(src); + ib->ptr[ib->length_dw++] = upper_32_bits(src); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); } /** @@ -774,39 +766,27 @@ static void sdma_v2_4_vm_copy_pte(struct amdgpu_ib *ib, * * @ib: indirect buffer to fill with commands * @pe: addr of the page entry - * @addr: dst addr to write into pe + * @value: dst addr to write into pe * @count: number of page entries to update * @incr: increase next addr by incr bytes - * @flags: access flags * * Update PTEs by writing them manually using sDMA (CIK). */ -static void sdma_v2_4_vm_write_pte(struct amdgpu_ib *ib, - const dma_addr_t *pages_addr, uint64_t pe, - uint64_t addr, unsigned count, - uint32_t incr, uint32_t flags) +static void sdma_v2_4_vm_write_pte(struct amdgpu_ib *ib, uint64_t pe, + uint64_t value, unsigned count, + uint32_t incr) { - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - /* for non-physically contiguous pages (system) */ - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | - SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); - ib->ptr[ib->length_dw++] = pe; - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = ndw; - for (; ndw > 0; ndw -= 2, --count, pe += 8) { - value = amdgpu_vm_map_gart(pages_addr, addr); - addr += incr; - value |= flags; - ib->ptr[ib->length_dw++] = value; - ib->ptr[ib->length_dw++] = upper_32_bits(value); - } + unsigned ndw = count * 2; + + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | + SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); + ib->ptr[ib->length_dw++] = pe; + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = ndw; + for (; ndw > 0; ndw -= 2, --count, pe += 8) { + ib->ptr[ib->length_dw++] = lower_32_bits(value); + ib->ptr[ib->length_dw++] = upper_32_bits(value); + value += incr; } } @@ -822,40 +802,21 @@ static void sdma_v2_4_vm_write_pte(struct amdgpu_ib *ib, * * Update the page tables using sDMA (CIK). */ -static void sdma_v2_4_vm_set_pte_pde(struct amdgpu_ib *ib, - uint64_t pe, +static void sdma_v2_4_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe, uint64_t addr, unsigned count, uint32_t incr, uint32_t flags) { - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count; - if (ndw > 0x7FFFF) - ndw = 0x7FFFF; - - if (flags & AMDGPU_PTE_VALID) - value = addr; - else - value = 0; - - /* for physically contiguous pages (vram) */ - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_GEN_PTEPDE); - ib->ptr[ib->length_dw++] = pe; /* dst addr */ - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = flags; /* mask */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = value; /* value */ - ib->ptr[ib->length_dw++] = upper_32_bits(value); - ib->ptr[ib->length_dw++] = incr; /* increment size */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = ndw; /* number of entries */ - - pe += ndw * 8; - addr += ndw * incr; - count -= ndw; - } + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_GEN_PTEPDE); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = flags; /* mask */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = lower_32_bits(addr); /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(addr); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = count; /* number of entries */ } /** diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 653ce5ed55ae..bee4978bec73 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -976,24 +976,16 @@ static void sdma_v3_0_vm_copy_pte(struct amdgpu_ib *ib, uint64_t pe, uint64_t src, unsigned count) { - while (count) { - unsigned bytes = count * 8; - if (bytes > 0x1FFFF8) - bytes = 0x1FFFF8; - - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) | - SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); - ib->ptr[ib->length_dw++] = bytes; - ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ - ib->ptr[ib->length_dw++] = lower_32_bits(src); - ib->ptr[ib->length_dw++] = upper_32_bits(src); - ib->ptr[ib->length_dw++] = lower_32_bits(pe); - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - - pe += bytes; - src += bytes; - count -= bytes / 8; - } + unsigned bytes = count * 8; + + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) | + SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); + ib->ptr[ib->length_dw++] = bytes; + ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ + ib->ptr[ib->length_dw++] = lower_32_bits(src); + ib->ptr[ib->length_dw++] = upper_32_bits(src); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); } /** @@ -1001,39 +993,27 @@ static void sdma_v3_0_vm_copy_pte(struct amdgpu_ib *ib, * * @ib: indirect buffer to fill with commands * @pe: addr of the page entry - * @addr: dst addr to write into pe + * @value: dst addr to write into pe * @count: number of page entries to update * @incr: increase next addr by incr bytes - * @flags: access flags * * Update PTEs by writing them manually using sDMA (CIK). */ -static void sdma_v3_0_vm_write_pte(struct amdgpu_ib *ib, - const dma_addr_t *pages_addr, uint64_t pe, - uint64_t addr, unsigned count, - uint32_t incr, uint32_t flags) -{ - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - /* for non-physically contiguous pages (system) */ - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | - SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); - ib->ptr[ib->length_dw++] = pe; - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = ndw; - for (; ndw > 0; ndw -= 2, --count, pe += 8) { - value = amdgpu_vm_map_gart(pages_addr, addr); - addr += incr; - value |= flags; - ib->ptr[ib->length_dw++] = value; - ib->ptr[ib->length_dw++] = upper_32_bits(value); - } +static void sdma_v3_0_vm_write_pte(struct amdgpu_ib *ib, uint64_t pe, + uint64_t value, unsigned count, + uint32_t incr) +{ + unsigned ndw = count * 2; + + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | + SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = ndw; + for (; ndw > 0; ndw -= 2, --count, pe += 8) { + ib->ptr[ib->length_dw++] = lower_32_bits(value); + ib->ptr[ib->length_dw++] = upper_32_bits(value); + value += incr; } } @@ -1049,40 +1029,21 @@ static void sdma_v3_0_vm_write_pte(struct amdgpu_ib *ib, * * Update the page tables using sDMA (CIK). */ -static void sdma_v3_0_vm_set_pte_pde(struct amdgpu_ib *ib, - uint64_t pe, +static void sdma_v3_0_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe, uint64_t addr, unsigned count, uint32_t incr, uint32_t flags) { - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count; - if (ndw > 0x7FFFF) - ndw = 0x7FFFF; - - if (flags & AMDGPU_PTE_VALID) - value = addr; - else - value = 0; - - /* for physically contiguous pages (vram) */ - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_GEN_PTEPDE); - ib->ptr[ib->length_dw++] = pe; /* dst addr */ - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = flags; /* mask */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = value; /* value */ - ib->ptr[ib->length_dw++] = upper_32_bits(value); - ib->ptr[ib->length_dw++] = incr; /* increment size */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = ndw; /* number of entries */ - - pe += ndw * 8; - addr += ndw * incr; - count -= ndw; - } + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_GEN_PTEPDE); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = flags; /* mask */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = lower_32_bits(addr); /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(addr); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = count; /* number of entries */ } /** @@ -1320,28 +1281,79 @@ static int sdma_v3_0_wait_for_idle(void *handle) return -ETIMEDOUT; } -static int sdma_v3_0_soft_reset(void *handle) +static int sdma_v3_0_check_soft_reset(void *handle) { - u32 srbm_soft_reset = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; u32 tmp = RREG32(mmSRBM_STATUS2); - if (tmp & SRBM_STATUS2__SDMA_BUSY_MASK) { - /* sdma0 */ - tmp = RREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET); - tmp = REG_SET_FIELD(tmp, SDMA0_F32_CNTL, HALT, 0); - WREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET, tmp); + if ((tmp & SRBM_STATUS2__SDMA_BUSY_MASK) || + (tmp & SRBM_STATUS2__SDMA1_BUSY_MASK)) { srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA_MASK; - } - if (tmp & SRBM_STATUS2__SDMA1_BUSY_MASK) { - /* sdma1 */ - tmp = RREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET); - tmp = REG_SET_FIELD(tmp, SDMA0_F32_CNTL, HALT, 0); - WREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET, tmp); srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA1_MASK; } if (srbm_soft_reset) { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_SDMA].hang = true; + adev->sdma.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_SDMA].hang = false; + adev->sdma.srbm_soft_reset = 0; + } + + return 0; +} + +static int sdma_v3_0_pre_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_SDMA].hang) + return 0; + + srbm_soft_reset = adev->sdma.srbm_soft_reset; + + if (REG_GET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_SDMA) || + REG_GET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_SDMA1)) { + sdma_v3_0_ctx_switch_enable(adev, false); + sdma_v3_0_enable(adev, false); + } + + return 0; +} + +static int sdma_v3_0_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_SDMA].hang) + return 0; + + srbm_soft_reset = adev->sdma.srbm_soft_reset; + + if (REG_GET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_SDMA) || + REG_GET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_SDMA1)) { + sdma_v3_0_gfx_resume(adev); + sdma_v3_0_rlc_resume(adev); + } + + return 0; +} + +static int sdma_v3_0_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; + u32 tmp; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_SDMA].hang) + return 0; + + srbm_soft_reset = adev->sdma.srbm_soft_reset; + + if (srbm_soft_reset) { tmp = RREG32(mmSRBM_SOFT_RESET); tmp |= srbm_soft_reset; dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); @@ -1559,6 +1571,9 @@ const struct amd_ip_funcs sdma_v3_0_ip_funcs = { .resume = sdma_v3_0_resume, .is_idle = sdma_v3_0_is_idle, .wait_for_idle = sdma_v3_0_wait_for_idle, + .check_soft_reset = sdma_v3_0_check_soft_reset, + .pre_soft_reset = sdma_v3_0_pre_soft_reset, + .post_soft_reset = sdma_v3_0_post_soft_reset, .soft_reset = sdma_v3_0_soft_reset, .set_clockgating_state = sdma_v3_0_set_clockgating_state, .set_powergating_state = sdma_v3_0_set_powergating_state, diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c index c92055805a45..d127d59f953a 100644 --- a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c @@ -373,10 +373,10 @@ static int tonga_ih_wait_for_idle(void *handle) return -ETIMEDOUT; } -static int tonga_ih_soft_reset(void *handle) +static int tonga_ih_check_soft_reset(void *handle) { - u32 srbm_soft_reset = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; u32 tmp = RREG32(mmSRBM_STATUS); if (tmp & SRBM_STATUS__IH_BUSY_MASK) @@ -384,6 +384,48 @@ static int tonga_ih_soft_reset(void *handle) SOFT_RESET_IH, 1); if (srbm_soft_reset) { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_IH].hang = true; + adev->irq.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_IH].hang = false; + adev->irq.srbm_soft_reset = 0; + } + + return 0; +} + +static int tonga_ih_pre_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_IH].hang) + return 0; + + return tonga_ih_hw_fini(adev); +} + +static int tonga_ih_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_IH].hang) + return 0; + + return tonga_ih_hw_init(adev); +} + +static int tonga_ih_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_IH].hang) + return 0; + srbm_soft_reset = adev->irq.srbm_soft_reset; + + if (srbm_soft_reset) { + u32 tmp; + tmp = RREG32(mmSRBM_SOFT_RESET); tmp |= srbm_soft_reset; dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); @@ -427,7 +469,10 @@ const struct amd_ip_funcs tonga_ih_ip_funcs = { .resume = tonga_ih_resume, .is_idle = tonga_ih_is_idle, .wait_for_idle = tonga_ih_wait_for_idle, + .check_soft_reset = tonga_ih_check_soft_reset, + .pre_soft_reset = tonga_ih_pre_soft_reset, .soft_reset = tonga_ih_soft_reset, + .post_soft_reset = tonga_ih_post_soft_reset, .set_clockgating_state = tonga_ih_set_clockgating_state, .set_powergating_state = tonga_ih_set_powergating_state, }; diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c index 132e613ed674..10c0407dcb6e 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c @@ -116,7 +116,7 @@ static int uvd_v4_2_sw_init(void *handle) ring = &adev->uvd.ring; sprintf(ring->name, "uvd"); - r = amdgpu_ring_init(adev, ring, 512, CP_PACKET2, 0xf, + r = amdgpu_ring_init(adev, ring, 512, PACKET0(mmUVD_NO_OP, 0), 0xf, &adev->uvd.irq, 0, AMDGPU_RING_TYPE_UVD); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c index 101de136ba63..8513376062c1 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c @@ -112,7 +112,7 @@ static int uvd_v5_0_sw_init(void *handle) ring = &adev->uvd.ring; sprintf(ring->name, "uvd"); - r = amdgpu_ring_init(adev, ring, 512, CP_PACKET2, 0xf, + r = amdgpu_ring_init(adev, ring, 512, PACKET0(mmUVD_NO_OP, 0), 0xf, &adev->uvd.irq, 0, AMDGPU_RING_TYPE_UVD); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index 7f21102bfb99..2abe8a93c99f 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -116,7 +116,7 @@ static int uvd_v6_0_sw_init(void *handle) ring = &adev->uvd.ring; sprintf(ring->name, "uvd"); - r = amdgpu_ring_init(adev, ring, 512, CP_PACKET2, 0xf, + r = amdgpu_ring_init(adev, ring, 512, PACKET0(mmUVD_NO_OP, 0), 0xf, &adev->uvd.irq, 0, AMDGPU_RING_TYPE_UVD); return r; @@ -396,21 +396,14 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) uvd_v6_0_mc_resume(adev); - /* Set dynamic clock gating in S/W control mode */ - if (adev->cg_flags & AMD_CG_SUPPORT_UVD_MGCG) { - uvd_v6_0_set_sw_clock_gating(adev); - } else { - /* disable clock gating */ - uint32_t data = RREG32(mmUVD_CGC_CTRL); - data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK; - WREG32(mmUVD_CGC_CTRL, data); - } + /* disable clock gating */ + WREG32_FIELD(UVD_CGC_CTRL, DYN_CLOCK_MODE, 0); /* disable interupt */ - WREG32_P(mmUVD_MASTINT_EN, 0, ~UVD_MASTINT_EN__VCPU_EN_MASK); + WREG32_FIELD(UVD_MASTINT_EN, VCPU_EN, 0); /* stall UMC and register bus before resetting VCPU */ - WREG32_P(mmUVD_LMI_CTRL2, UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); + WREG32_FIELD(UVD_LMI_CTRL2, STALL_ARB_UMC, 1); mdelay(1); /* put LMI, VCPU, RBC etc... into reset */ @@ -426,7 +419,7 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) mdelay(5); /* take UVD block out of reset */ - WREG32_P(mmSRBM_SOFT_RESET, 0, ~SRBM_SOFT_RESET__SOFT_RESET_UVD_MASK); + WREG32_FIELD(SRBM_SOFT_RESET, SOFT_RESET_UVD, 0); mdelay(5); /* initialize UVD memory controller */ @@ -461,7 +454,7 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) WREG32(mmUVD_VCPU_CNTL, UVD_VCPU_CNTL__CLK_EN_MASK); /* enable UMC */ - WREG32_P(mmUVD_LMI_CTRL2, 0, ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); + WREG32_FIELD(UVD_LMI_CTRL2, STALL_ARB_UMC, 0); /* boot up the VCPU */ WREG32(mmUVD_SOFT_RESET, 0); @@ -481,11 +474,9 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) break; DRM_ERROR("UVD not responding, trying to reset the VCPU!!!\n"); - WREG32_P(mmUVD_SOFT_RESET, UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK, - ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); + WREG32_FIELD(UVD_SOFT_RESET, VCPU_SOFT_RESET, 1); mdelay(10); - WREG32_P(mmUVD_SOFT_RESET, 0, - ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); + WREG32_FIELD(UVD_SOFT_RESET, VCPU_SOFT_RESET, 0); mdelay(10); r = -1; } @@ -502,15 +493,14 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) /* clear the bit 4 of UVD_STATUS */ WREG32_P(mmUVD_STATUS, 0, ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT)); + /* force RBC into idle state */ rb_bufsz = order_base_2(ring->ring_size); - tmp = 0; - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz); + tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz); tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1); tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1); tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_WPTR_POLL_EN, 0); tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1); tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1); - /* force RBC into idle state */ WREG32(mmUVD_RBC_RB_CNTL, tmp); /* set the write pointer delay */ @@ -531,7 +521,7 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) ring->wptr = RREG32(mmUVD_RBC_RB_RPTR); WREG32(mmUVD_RBC_RB_WPTR, ring->wptr); - WREG32_P(mmUVD_RBC_RB_CNTL, 0, ~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK); + WREG32_FIELD(UVD_RBC_RB_CNTL, RB_NO_FETCH, 0); return 0; } @@ -748,20 +738,82 @@ static int uvd_v6_0_wait_for_idle(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->usec_timeout; i++) { - if (!(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK)) + if (uvd_v6_0_is_idle(handle)) return 0; } return -ETIMEDOUT; } -static int uvd_v6_0_soft_reset(void *handle) +#define AMDGPU_UVD_STATUS_BUSY_MASK 0xfd +static int uvd_v6_0_check_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; + u32 tmp = RREG32(mmSRBM_STATUS); + + if (REG_GET_FIELD(tmp, SRBM_STATUS, UVD_RQ_PENDING) || + REG_GET_FIELD(tmp, SRBM_STATUS, UVD_BUSY) || + (RREG32(mmUVD_STATUS) & AMDGPU_UVD_STATUS_BUSY_MASK)) + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_UVD, 1); + + if (srbm_soft_reset) { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_UVD].hang = true; + adev->uvd.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_UVD].hang = false; + adev->uvd.srbm_soft_reset = 0; + } + return 0; +} +static int uvd_v6_0_pre_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_UVD].hang) + return 0; uvd_v6_0_stop(adev); + return 0; +} + +static int uvd_v6_0_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_UVD].hang) + return 0; + srbm_soft_reset = adev->uvd.srbm_soft_reset; + + if (srbm_soft_reset) { + u32 tmp; + + tmp = RREG32(mmSRBM_SOFT_RESET); + tmp |= srbm_soft_reset; + dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); + + udelay(50); + + tmp &= ~srbm_soft_reset; + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); + + /* Wait a little for things to settle down */ + udelay(50); + } + + return 0; +} + +static int uvd_v6_0_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_UVD].hang) + return 0; - WREG32_P(mmSRBM_SOFT_RESET, SRBM_SOFT_RESET__SOFT_RESET_UVD_MASK, - ~SRBM_SOFT_RESET__SOFT_RESET_UVD_MASK); mdelay(5); return uvd_v6_0_start(adev); @@ -902,21 +954,15 @@ static int uvd_v6_0_set_clockgating_state(void *handle, enum amd_clockgating_state state) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - bool enable = (state == AMD_CG_STATE_GATE) ? true : false; - static int curstate = -1; if (adev->asic_type == CHIP_FIJI || - adev->asic_type == CHIP_POLARIS10) - uvd_v6_set_bypass_mode(adev, enable); + adev->asic_type == CHIP_POLARIS10) + uvd_v6_set_bypass_mode(adev, state == AMD_CG_STATE_GATE ? true : false); if (!(adev->cg_flags & AMD_CG_SUPPORT_UVD_MGCG)) return 0; - if (curstate == state) - return 0; - - curstate = state; - if (enable) { + if (state == AMD_CG_STATE_GATE) { /* disable HW gating and enable Sw gating */ uvd_v6_0_set_sw_clock_gating(adev); } else { @@ -946,6 +992,8 @@ static int uvd_v6_0_set_powergating_state(void *handle, if (!(adev->pg_flags & AMD_PG_SUPPORT_UVD)) return 0; + WREG32(mmUVD_POWER_STATUS, UVD_POWER_STATUS__UVD_PG_EN_MASK); + if (state == AMD_PG_STATE_GATE) { uvd_v6_0_stop(adev); return 0; @@ -966,7 +1014,10 @@ const struct amd_ip_funcs uvd_v6_0_ip_funcs = { .resume = uvd_v6_0_resume, .is_idle = uvd_v6_0_is_idle, .wait_for_idle = uvd_v6_0_wait_for_idle, + .check_soft_reset = uvd_v6_0_check_soft_reset, + .pre_soft_reset = uvd_v6_0_pre_soft_reset, .soft_reset = uvd_v6_0_soft_reset, + .post_soft_reset = uvd_v6_0_post_soft_reset, .set_clockgating_state = uvd_v6_0_set_clockgating_state, .set_powergating_state = uvd_v6_0_set_powergating_state, }; diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c index 80a37a602181..5fa55b52c00e 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c @@ -40,6 +40,7 @@ #define VCE_V2_0_FW_SIZE (256 * 1024) #define VCE_V2_0_STACK_SIZE (64 * 1024) #define VCE_V2_0_DATA_SIZE (23552 * AMDGPU_MAX_VCE_HANDLES) +#define VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK 0x02 static void vce_v2_0_mc_resume(struct amdgpu_device *adev); static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev); @@ -96,6 +97,49 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring) WREG32(mmVCE_RB_WPTR2, ring->wptr); } +static int vce_v2_0_lmi_clean(struct amdgpu_device *adev) +{ + int i, j; + + for (i = 0; i < 10; ++i) { + for (j = 0; j < 100; ++j) { + uint32_t status = RREG32(mmVCE_LMI_STATUS); + + if (status & 0x337f) + return 0; + mdelay(10); + } + } + + return -ETIMEDOUT; +} + +static int vce_v2_0_firmware_loaded(struct amdgpu_device *adev) +{ + int i, j; + + for (i = 0; i < 10; ++i) { + for (j = 0; j < 100; ++j) { + uint32_t status = RREG32(mmVCE_STATUS); + + if (status & VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK) + return 0; + mdelay(10); + } + + DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n"); + WREG32_P(mmVCE_SOFT_RESET, + VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, + ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + mdelay(10); + WREG32_P(mmVCE_SOFT_RESET, 0, + ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + mdelay(10); + } + + return -ETIMEDOUT; +} + /** * vce_v2_0_start - start VCE block * @@ -106,7 +150,7 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring) static int vce_v2_0_start(struct amdgpu_device *adev) { struct amdgpu_ring *ring; - int i, j, r; + int r; vce_v2_0_mc_resume(adev); @@ -127,36 +171,12 @@ static int vce_v2_0_start(struct amdgpu_device *adev) WREG32(mmVCE_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); WREG32(mmVCE_RB_SIZE2, ring->ring_size / 4); - WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, ~VCE_VCPU_CNTL__CLK_EN_MASK); - - WREG32_P(mmVCE_SOFT_RESET, - VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - + WREG32_FIELD(VCE_VCPU_CNTL, CLK_EN, 1); + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 1); mdelay(100); + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 0); - WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - - for (i = 0; i < 10; ++i) { - uint32_t status; - for (j = 0; j < 100; ++j) { - status = RREG32(mmVCE_STATUS); - if (status & 2) - break; - mdelay(10); - } - r = 0; - if (status & 2) - break; - - DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n"); - WREG32_P(mmVCE_SOFT_RESET, VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - mdelay(10); - WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - mdelay(10); - r = -1; - } + r = vce_v2_0_firmware_loaded(adev); /* clear BUSY flag */ WREG32_P(mmVCE_STATUS, 0, ~1); @@ -338,47 +358,50 @@ static void vce_v2_0_set_sw_cg(struct amdgpu_device *adev, bool gated) static void vce_v2_0_set_dyn_cg(struct amdgpu_device *adev, bool gated) { - u32 orig, tmp; + if (vce_v2_0_wait_for_idle(adev)) { + DRM_INFO("VCE is busy, Can't set clock gateing"); + return; + } - if (gated) { - if (vce_v2_0_wait_for_idle(adev)) { - DRM_INFO("VCE is busy, Can't set clock gateing"); - return; - } - WREG32_P(mmVCE_VCPU_CNTL, 0, ~VCE_VCPU_CNTL__CLK_EN_MASK); - WREG32_P(mmVCE_SOFT_RESET, VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - mdelay(100); - WREG32(mmVCE_STATUS, 0); - } else { - WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, ~VCE_VCPU_CNTL__CLK_EN_MASK); - WREG32_P(mmVCE_SOFT_RESET, VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - mdelay(100); + WREG32_P(mmVCE_LMI_CTRL2, 0x100, ~0x100); + + if (vce_v2_0_lmi_clean(adev)) { + DRM_INFO("LMI is busy, Can't set clock gateing"); + return; } - tmp = RREG32(mmVCE_CLOCK_GATING_B); - tmp &= ~0x00060006; + WREG32_P(mmVCE_VCPU_CNTL, 0, ~VCE_VCPU_CNTL__CLK_EN_MASK); + WREG32_P(mmVCE_SOFT_RESET, + VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, + ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + WREG32(mmVCE_STATUS, 0); + + if (gated) + WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0); + /* LMI_MC/LMI_UMC always set in dynamic, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {0, 0} */ if (gated) { - tmp |= 0xe10000; + /* Force CLOCK OFF , set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {*, 1} */ + WREG32(mmVCE_CLOCK_GATING_B, 0xe90010); } else { - tmp |= 0xe1; - tmp &= ~0xe10000; + /* Force CLOCK ON, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {1, 0} */ + WREG32(mmVCE_CLOCK_GATING_B, 0x800f1); } - WREG32(mmVCE_CLOCK_GATING_B, tmp); - orig = tmp = RREG32(mmVCE_UENC_CLOCK_GATING); - tmp &= ~0x1fe000; - tmp &= ~0xff000000; - if (tmp != orig) - WREG32(mmVCE_UENC_CLOCK_GATING, tmp); + /* Set VCE_UENC_CLOCK_GATING always in dynamic mode {*_FORCE_ON, *_FORCE_OFF} = {0, 0}*/; + WREG32(mmVCE_UENC_CLOCK_GATING, 0x40); - orig = tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING); - tmp &= ~0x3fc; - if (tmp != orig) - WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp); + /* set VCE_UENC_REG_CLOCK_GATING always in dynamic mode */ + WREG32(mmVCE_UENC_REG_CLOCK_GATING, 0x00); - if (gated) - WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0); - WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + WREG32_P(mmVCE_LMI_CTRL2, 0, ~0x100); + if(!gated) { + WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, ~VCE_VCPU_CNTL__CLK_EN_MASK); + mdelay(100); + WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + + vce_v2_0_firmware_loaded(adev); + WREG32_P(mmVCE_STATUS, 0, ~VCE_STATUS__JOB_BUSY_MASK); + } } static void vce_v2_0_disable_cg(struct amdgpu_device *adev) @@ -458,9 +481,7 @@ static void vce_v2_0_mc_resume(struct amdgpu_device *adev) WREG32(mmVCE_VCPU_CACHE_SIZE2, size); WREG32_P(mmVCE_LMI_CTRL2, 0x0, ~0x100); - - WREG32_P(mmVCE_SYS_INT_EN, VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK, - ~VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK); + WREG32_FIELD(VCE_SYS_INT_EN, VCE_SYS_INT_TRAP_INTERRUPT_EN, 1); vce_v2_0_init_cg(adev); } @@ -474,11 +495,11 @@ static bool vce_v2_0_is_idle(void *handle) static int vce_v2_0_wait_for_idle(void *handle) { - unsigned i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + unsigned i; for (i = 0; i < adev->usec_timeout; i++) { - if (!(RREG32(mmSRBM_STATUS2) & SRBM_STATUS2__VCE_BUSY_MASK)) + if (vce_v2_0_is_idle(handle)) return 0; } return -ETIMEDOUT; @@ -488,8 +509,7 @@ static int vce_v2_0_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - WREG32_P(mmSRBM_SOFT_RESET, SRBM_SOFT_RESET__SOFT_RESET_VCE_MASK, - ~SRBM_SOFT_RESET__SOFT_RESET_VCE_MASK); + WREG32_FIELD(SRBM_SOFT_RESET, SOFT_RESET_VCE, 1); mdelay(5); return vce_v2_0_start(adev); @@ -516,10 +536,8 @@ static int vce_v2_0_process_interrupt(struct amdgpu_device *adev, DRM_DEBUG("IH: VCE\n"); switch (entry->src_data) { case 0: - amdgpu_fence_process(&adev->vce.ring[0]); - break; case 1: - amdgpu_fence_process(&adev->vce.ring[1]); + amdgpu_fence_process(&adev->vce.ring[entry->src_data]); break; default: DRM_ERROR("Unhandled interrupt: %d %d\n", diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c index c271abffd8dd..615b8b16ad04 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c @@ -37,6 +37,9 @@ #include "gca/gfx_8_0_d.h" #include "smu/smu_7_1_2_d.h" #include "smu/smu_7_1_2_sh_mask.h" +#include "gca/gfx_8_0_d.h" +#include "gca/gfx_8_0_sh_mask.h" + #define GRBM_GFX_INDEX__VCE_INSTANCE__SHIFT 0x04 #define GRBM_GFX_INDEX__VCE_INSTANCE_MASK 0x10 @@ -107,102 +110,72 @@ static void vce_v3_0_ring_set_wptr(struct amdgpu_ring *ring) static void vce_v3_0_override_vce_clock_gating(struct amdgpu_device *adev, bool override) { - u32 tmp, data; - - tmp = data = RREG32(mmVCE_RB_ARB_CTRL); - if (override) - data |= VCE_RB_ARB_CTRL__VCE_CGTT_OVERRIDE_MASK; - else - data &= ~VCE_RB_ARB_CTRL__VCE_CGTT_OVERRIDE_MASK; - - if (tmp != data) - WREG32(mmVCE_RB_ARB_CTRL, data); + WREG32_FIELD(VCE_RB_ARB_CTRL, VCE_CGTT_OVERRIDE, override ? 1 : 0); } static void vce_v3_0_set_vce_sw_clock_gating(struct amdgpu_device *adev, bool gated) { - u32 tmp, data; + u32 data; + /* Set Override to disable Clock Gating */ vce_v3_0_override_vce_clock_gating(adev, true); - if (!gated) { - /* Force CLOCK ON for VCE_CLOCK_GATING_B, - * {*_FORCE_ON, *_FORCE_OFF} = {1, 0} - * VREG can be FORCE ON or set to Dynamic, but can't be OFF - */ - tmp = data = RREG32(mmVCE_CLOCK_GATING_B); + /* This function enables MGCG which is controlled by firmware. + With the clocks in the gated state the core is still + accessible but the firmware will throttle the clocks on the + fly as necessary. + */ + if (gated) { + data = RREG32(mmVCE_CLOCK_GATING_B); data |= 0x1ff; data &= ~0xef0000; - if (tmp != data) - WREG32(mmVCE_CLOCK_GATING_B, data); + WREG32(mmVCE_CLOCK_GATING_B, data); - /* Force CLOCK ON for VCE_UENC_CLOCK_GATING, - * {*_FORCE_ON, *_FORCE_OFF} = {1, 0} - */ - tmp = data = RREG32(mmVCE_UENC_CLOCK_GATING); + data = RREG32(mmVCE_UENC_CLOCK_GATING); data |= 0x3ff000; data &= ~0xffc00000; - if (tmp != data) - WREG32(mmVCE_UENC_CLOCK_GATING, data); + WREG32(mmVCE_UENC_CLOCK_GATING, data); - /* set VCE_UENC_CLOCK_GATING_2 */ - tmp = data = RREG32(mmVCE_UENC_CLOCK_GATING_2); + data = RREG32(mmVCE_UENC_CLOCK_GATING_2); data |= 0x2; - data &= ~0x2; - if (tmp != data) - WREG32(mmVCE_UENC_CLOCK_GATING_2, data); + data &= ~0x00010000; + WREG32(mmVCE_UENC_CLOCK_GATING_2, data); - /* Force CLOCK ON for VCE_UENC_REG_CLOCK_GATING */ - tmp = data = RREG32(mmVCE_UENC_REG_CLOCK_GATING); + data = RREG32(mmVCE_UENC_REG_CLOCK_GATING); data |= 0x37f; - if (tmp != data) - WREG32(mmVCE_UENC_REG_CLOCK_GATING, data); + WREG32(mmVCE_UENC_REG_CLOCK_GATING, data); - /* Force VCE_UENC_DMA_DCLK_CTRL Clock ON */ - tmp = data = RREG32(mmVCE_UENC_DMA_DCLK_CTRL); + data = RREG32(mmVCE_UENC_DMA_DCLK_CTRL); data |= VCE_UENC_DMA_DCLK_CTRL__WRDMCLK_FORCEON_MASK | - VCE_UENC_DMA_DCLK_CTRL__RDDMCLK_FORCEON_MASK | - VCE_UENC_DMA_DCLK_CTRL__REGCLK_FORCEON_MASK | - 0x8; - if (tmp != data) - WREG32(mmVCE_UENC_DMA_DCLK_CTRL, data); + VCE_UENC_DMA_DCLK_CTRL__RDDMCLK_FORCEON_MASK | + VCE_UENC_DMA_DCLK_CTRL__REGCLK_FORCEON_MASK | + 0x8; + WREG32(mmVCE_UENC_DMA_DCLK_CTRL, data); } else { - /* Force CLOCK OFF for VCE_CLOCK_GATING_B, - * {*, *_FORCE_OFF} = {*, 1} - * set VREG to Dynamic, as it can't be OFF - */ - tmp = data = RREG32(mmVCE_CLOCK_GATING_B); + data = RREG32(mmVCE_CLOCK_GATING_B); data &= ~0x80010; data |= 0xe70008; - if (tmp != data) - WREG32(mmVCE_CLOCK_GATING_B, data); - /* Force CLOCK OFF for VCE_UENC_CLOCK_GATING, - * Force ClOCK OFF takes precedent over Force CLOCK ON setting. - * {*_FORCE_ON, *_FORCE_OFF} = {*, 1} - */ - tmp = data = RREG32(mmVCE_UENC_CLOCK_GATING); + WREG32(mmVCE_CLOCK_GATING_B, data); + + data = RREG32(mmVCE_UENC_CLOCK_GATING); data |= 0xffc00000; - if (tmp != data) - WREG32(mmVCE_UENC_CLOCK_GATING, data); - /* Set VCE_UENC_CLOCK_GATING_2 */ - tmp = data = RREG32(mmVCE_UENC_CLOCK_GATING_2); + WREG32(mmVCE_UENC_CLOCK_GATING, data); + + data = RREG32(mmVCE_UENC_CLOCK_GATING_2); data |= 0x10000; - if (tmp != data) - WREG32(mmVCE_UENC_CLOCK_GATING_2, data); - /* Set VCE_UENC_REG_CLOCK_GATING to dynamic */ - tmp = data = RREG32(mmVCE_UENC_REG_CLOCK_GATING); + WREG32(mmVCE_UENC_CLOCK_GATING_2, data); + + data = RREG32(mmVCE_UENC_REG_CLOCK_GATING); data &= ~0xffc00000; - if (tmp != data) - WREG32(mmVCE_UENC_REG_CLOCK_GATING, data); - /* Set VCE_UENC_DMA_DCLK_CTRL CG always in dynamic mode */ - tmp = data = RREG32(mmVCE_UENC_DMA_DCLK_CTRL); + WREG32(mmVCE_UENC_REG_CLOCK_GATING, data); + + data = RREG32(mmVCE_UENC_DMA_DCLK_CTRL); data &= ~(VCE_UENC_DMA_DCLK_CTRL__WRDMCLK_FORCEON_MASK | - VCE_UENC_DMA_DCLK_CTRL__RDDMCLK_FORCEON_MASK | - VCE_UENC_DMA_DCLK_CTRL__REGCLK_FORCEON_MASK | - 0x8); - if (tmp != data) - WREG32(mmVCE_UENC_DMA_DCLK_CTRL, data); + VCE_UENC_DMA_DCLK_CTRL__RDDMCLK_FORCEON_MASK | + VCE_UENC_DMA_DCLK_CTRL__REGCLK_FORCEON_MASK | + 0x8); + WREG32(mmVCE_UENC_DMA_DCLK_CTRL, data); } vce_v3_0_override_vce_clock_gating(adev, false); } @@ -221,12 +194,9 @@ static int vce_v3_0_firmware_loaded(struct amdgpu_device *adev) } DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n"); - WREG32_P(mmVCE_SOFT_RESET, - VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 1); mdelay(10); - WREG32_P(mmVCE_SOFT_RESET, 0, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 0); mdelay(10); } @@ -264,38 +234,22 @@ static int vce_v3_0_start(struct amdgpu_device *adev) if (adev->vce.harvest_config & (1 << idx)) continue; - if (idx == 0) - WREG32_P(mmGRBM_GFX_INDEX, 0, - ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); - else - WREG32_P(mmGRBM_GFX_INDEX, - GRBM_GFX_INDEX__VCE_INSTANCE_MASK, - ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); - + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, idx); vce_v3_0_mc_resume(adev, idx); - - WREG32_P(mmVCE_STATUS, VCE_STATUS__JOB_BUSY_MASK, - ~VCE_STATUS__JOB_BUSY_MASK); + WREG32_FIELD(VCE_STATUS, JOB_BUSY, 1); if (adev->asic_type >= CHIP_STONEY) WREG32_P(mmVCE_VCPU_CNTL, 1, ~0x200001); else - WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, - ~VCE_VCPU_CNTL__CLK_EN_MASK); - - WREG32_P(mmVCE_SOFT_RESET, 0, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + WREG32_FIELD(VCE_VCPU_CNTL, CLK_EN, 1); + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 0); mdelay(100); r = vce_v3_0_firmware_loaded(adev); /* clear BUSY flag */ - WREG32_P(mmVCE_STATUS, 0, ~VCE_STATUS__JOB_BUSY_MASK); - - /* Set Clock-Gating off */ - if (adev->cg_flags & AMD_CG_SUPPORT_VCE_MGCG) - vce_v3_0_set_vce_sw_clock_gating(adev, false); + WREG32_FIELD(VCE_STATUS, JOB_BUSY, 0); if (r) { DRM_ERROR("VCE not responding, giving up!!!\n"); @@ -304,7 +258,7 @@ static int vce_v3_0_start(struct amdgpu_device *adev) } } - WREG32_P(mmGRBM_GFX_INDEX, 0, ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, 0); mutex_unlock(&adev->grbm_idx_mutex); return 0; @@ -319,33 +273,25 @@ static int vce_v3_0_stop(struct amdgpu_device *adev) if (adev->vce.harvest_config & (1 << idx)) continue; - if (idx == 0) - WREG32_P(mmGRBM_GFX_INDEX, 0, - ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); - else - WREG32_P(mmGRBM_GFX_INDEX, - GRBM_GFX_INDEX__VCE_INSTANCE_MASK, - ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, idx); if (adev->asic_type >= CHIP_STONEY) WREG32_P(mmVCE_VCPU_CNTL, 0, ~0x200001); else - WREG32_P(mmVCE_VCPU_CNTL, 0, - ~VCE_VCPU_CNTL__CLK_EN_MASK); + WREG32_FIELD(VCE_VCPU_CNTL, CLK_EN, 0); + /* hold on ECPU */ - WREG32_P(mmVCE_SOFT_RESET, - VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 1); /* clear BUSY flag */ - WREG32_P(mmVCE_STATUS, 0, ~VCE_STATUS__JOB_BUSY_MASK); + WREG32_FIELD(VCE_STATUS, JOB_BUSY, 0); /* Set Clock-Gating off */ if (adev->cg_flags & AMD_CG_SUPPORT_VCE_MGCG) vce_v3_0_set_vce_sw_clock_gating(adev, false); } - WREG32_P(mmGRBM_GFX_INDEX, 0, ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, 0); mutex_unlock(&adev->grbm_idx_mutex); return 0; @@ -534,7 +480,7 @@ static void vce_v3_0_mc_resume(struct amdgpu_device *adev, int idx) WREG32_P(mmVCE_CLOCK_GATING_A, 0, ~(1 << 16)); WREG32_P(mmVCE_UENC_CLOCK_GATING, 0x1FF000, ~0xFF9FF000); WREG32_P(mmVCE_UENC_REG_CLOCK_GATING, 0x3F, ~0x3F); - WREG32(mmVCE_CLOCK_GATING_B, 0xf7); + WREG32(mmVCE_CLOCK_GATING_B, 0x1FF); WREG32(mmVCE_LMI_CTRL, 0x00398000); WREG32_P(mmVCE_LMI_CACHE_CTRL, 0x0, ~0x1); @@ -573,9 +519,7 @@ static void vce_v3_0_mc_resume(struct amdgpu_device *adev, int idx) } WREG32_P(mmVCE_LMI_CTRL2, 0x0, ~0x100); - - WREG32_P(mmVCE_SYS_INT_EN, VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK, - ~VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK); + WREG32_FIELD(VCE_SYS_INT_EN, VCE_SYS_INT_TRAP_INTERRUPT_EN, 1); } static bool vce_v3_0_is_idle(void *handle) @@ -601,20 +545,108 @@ static int vce_v3_0_wait_for_idle(void *handle) return -ETIMEDOUT; } +#define VCE_STATUS_VCPU_REPORT_AUTO_BUSY_MASK 0x00000008L /* AUTO_BUSY */ +#define VCE_STATUS_VCPU_REPORT_RB0_BUSY_MASK 0x00000010L /* RB0_BUSY */ +#define VCE_STATUS_VCPU_REPORT_RB1_BUSY_MASK 0x00000020L /* RB1_BUSY */ +#define AMDGPU_VCE_STATUS_BUSY_MASK (VCE_STATUS_VCPU_REPORT_AUTO_BUSY_MASK | \ + VCE_STATUS_VCPU_REPORT_RB0_BUSY_MASK) + +static int vce_v3_0_check_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; + + /* According to VCE team , we should use VCE_STATUS instead + * SRBM_STATUS.VCE_BUSY bit for busy status checking. + * GRBM_GFX_INDEX.INSTANCE_INDEX is used to specify which VCE + * instance's registers are accessed + * (0 for 1st instance, 10 for 2nd instance). + * + *VCE_STATUS + *|UENC|ACPI|AUTO ACTIVE|RB1 |RB0 |RB2 | |FW_LOADED|JOB | + *|----+----+-----------+----+----+----+----------+---------+----| + *|bit8|bit7| bit6 |bit5|bit4|bit3| bit2 | bit1 |bit0| + * + * VCE team suggest use bit 3--bit 6 for busy status check + */ + mutex_lock(&adev->grbm_idx_mutex); + WREG32_FIELD(GRBM_GFX_INDEX, INSTANCE_INDEX, 0); + if (RREG32(mmVCE_STATUS) & AMDGPU_VCE_STATUS_BUSY_MASK) { + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE0, 1); + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE1, 1); + } + WREG32_FIELD(GRBM_GFX_INDEX, INSTANCE_INDEX, 0x10); + if (RREG32(mmVCE_STATUS) & AMDGPU_VCE_STATUS_BUSY_MASK) { + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE0, 1); + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE1, 1); + } + WREG32_FIELD(GRBM_GFX_INDEX, INSTANCE_INDEX, 0); + + if (srbm_soft_reset) { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_VCE].hang = true; + adev->vce.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_VCE].hang = false; + adev->vce.srbm_soft_reset = 0; + } + mutex_unlock(&adev->grbm_idx_mutex); + return 0; +} + static int vce_v3_0_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - u32 mask = 0; + u32 srbm_soft_reset; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_VCE].hang) + return 0; + srbm_soft_reset = adev->vce.srbm_soft_reset; + + if (srbm_soft_reset) { + u32 tmp; + + tmp = RREG32(mmSRBM_SOFT_RESET); + tmp |= srbm_soft_reset; + dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); + + udelay(50); + + tmp &= ~srbm_soft_reset; + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); + + /* Wait a little for things to settle down */ + udelay(50); + } + + return 0; +} + +static int vce_v3_0_pre_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_VCE].hang) + return 0; + + mdelay(5); - mask |= (adev->vce.harvest_config & AMDGPU_VCE_HARVEST_VCE0) ? 0 : SRBM_SOFT_RESET__SOFT_RESET_VCE0_MASK; - mask |= (adev->vce.harvest_config & AMDGPU_VCE_HARVEST_VCE1) ? 0 : SRBM_SOFT_RESET__SOFT_RESET_VCE1_MASK; + return vce_v3_0_suspend(adev); +} + + +static int vce_v3_0_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_VCE].hang) + return 0; - WREG32_P(mmSRBM_SOFT_RESET, mask, - ~(SRBM_SOFT_RESET__SOFT_RESET_VCE0_MASK | - SRBM_SOFT_RESET__SOFT_RESET_VCE1_MASK)); mdelay(5); - return vce_v3_0_start(adev); + return vce_v3_0_resume(adev); } static int vce_v3_0_set_interrupt_state(struct amdgpu_device *adev, @@ -637,9 +669,7 @@ static int vce_v3_0_process_interrupt(struct amdgpu_device *adev, { DRM_DEBUG("IH: VCE\n"); - WREG32_P(mmVCE_SYS_INT_STATUS, - VCE_SYS_INT_STATUS__VCE_SYS_INT_TRAP_INTERRUPT_INT_MASK, - ~VCE_SYS_INT_STATUS__VCE_SYS_INT_TRAP_INTERRUPT_INT_MASK); + WREG32_FIELD(VCE_SYS_INT_STATUS, VCE_SYS_INT_TRAP_INTERRUPT_INT, 1); switch (entry->src_data) { case 0: @@ -686,13 +716,7 @@ static int vce_v3_0_set_clockgating_state(void *handle, if (adev->vce.harvest_config & (1 << i)) continue; - if (i == 0) - WREG32_P(mmGRBM_GFX_INDEX, 0, - ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); - else - WREG32_P(mmGRBM_GFX_INDEX, - GRBM_GFX_INDEX__VCE_INSTANCE_MASK, - ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, i); if (enable) { /* initialize VCE_CLOCK_GATING_A: Clock ON/OFF delay */ @@ -711,7 +735,7 @@ static int vce_v3_0_set_clockgating_state(void *handle, vce_v3_0_set_vce_sw_clock_gating(adev, enable); } - WREG32_P(mmGRBM_GFX_INDEX, 0, ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, 0); mutex_unlock(&adev->grbm_idx_mutex); return 0; @@ -751,7 +775,10 @@ const struct amd_ip_funcs vce_v3_0_ip_funcs = { .resume = vce_v3_0_resume, .is_idle = vce_v3_0_is_idle, .wait_for_idle = vce_v3_0_wait_for_idle, + .check_soft_reset = vce_v3_0_check_soft_reset, + .pre_soft_reset = vce_v3_0_pre_soft_reset, .soft_reset = vce_v3_0_soft_reset, + .post_soft_reset = vce_v3_0_post_soft_reset, .set_clockgating_state = vce_v3_0_set_clockgating_state, .set_powergating_state = vce_v3_0_set_powergating_state, }; diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index 03a31c53aec3..f2e8aa1a0dbd 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -77,6 +77,7 @@ #if defined(CONFIG_DRM_AMD_ACP) #include "amdgpu_acp.h" #endif +#include "dce_virtual.h" MODULE_FIRMWARE("amdgpu/polaris10_smc.bin"); MODULE_FIRMWARE("amdgpu/polaris10_smc_sk.bin"); @@ -822,6 +823,60 @@ static const struct amdgpu_ip_block_version topaz_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version topaz_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vi_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 4, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 4, + .rev = 0, + .funcs = &iceland_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 1, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gfx_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 4, + .rev = 0, + .funcs = &sdma_v2_4_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version tonga_ip_blocks[] = { /* ORDER MATTERS! */ @@ -890,6 +945,74 @@ static const struct amdgpu_ip_block_version tonga_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version tonga_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vi_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gmc_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &tonga_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 1, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 10, + .minor = 0, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gfx_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &sdma_v3_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 5, + .minor = 0, + .rev = 0, + .funcs = &uvd_v5_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &vce_v3_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version fiji_ip_blocks[] = { /* ORDER MATTERS! */ @@ -958,6 +1081,74 @@ static const struct amdgpu_ip_block_version fiji_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version fiji_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vi_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 8, + .minor = 5, + .rev = 0, + .funcs = &gmc_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &tonga_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 1, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 10, + .minor = 1, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gfx_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &sdma_v3_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 6, + .minor = 0, + .rev = 0, + .funcs = &uvd_v6_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &vce_v3_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version polaris11_ip_blocks[] = { /* ORDER MATTERS! */ @@ -1026,6 +1217,74 @@ static const struct amdgpu_ip_block_version polaris11_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version polaris11_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vi_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 8, + .minor = 1, + .rev = 0, + .funcs = &gmc_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 3, + .minor = 1, + .rev = 0, + .funcs = &tonga_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 2, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 11, + .minor = 2, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gfx_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 3, + .minor = 1, + .rev = 0, + .funcs = &sdma_v3_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 6, + .minor = 3, + .rev = 0, + .funcs = &uvd_v6_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 3, + .minor = 4, + .rev = 0, + .funcs = &vce_v3_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version cz_ip_blocks[] = { /* ORDER MATTERS! */ @@ -1103,34 +1362,142 @@ static const struct amdgpu_ip_block_version cz_ip_blocks[] = #endif }; +static const struct amdgpu_ip_block_version cz_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vi_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gmc_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &cz_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 11, + .minor = 0, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gfx_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &sdma_v3_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 6, + .minor = 0, + .rev = 0, + .funcs = &uvd_v6_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &vce_v3_0_ip_funcs, + }, +#if defined(CONFIG_DRM_AMD_ACP) + { + .type = AMD_IP_BLOCK_TYPE_ACP, + .major = 2, + .minor = 2, + .rev = 0, + .funcs = &acp_ip_funcs, + }, +#endif +}; + int vi_set_ip_blocks(struct amdgpu_device *adev) { - switch (adev->asic_type) { - case CHIP_TOPAZ: - adev->ip_blocks = topaz_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(topaz_ip_blocks); - break; - case CHIP_FIJI: - adev->ip_blocks = fiji_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(fiji_ip_blocks); - break; - case CHIP_TONGA: - adev->ip_blocks = tonga_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(tonga_ip_blocks); - break; - case CHIP_POLARIS11: - case CHIP_POLARIS10: - adev->ip_blocks = polaris11_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(polaris11_ip_blocks); - break; - case CHIP_CARRIZO: - case CHIP_STONEY: - adev->ip_blocks = cz_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(cz_ip_blocks); - break; - default: - /* FIXME: not supported yet */ - return -EINVAL; + if (adev->enable_virtual_display) { + switch (adev->asic_type) { + case CHIP_TOPAZ: + adev->ip_blocks = topaz_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(topaz_ip_blocks_vd); + break; + case CHIP_FIJI: + adev->ip_blocks = fiji_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(fiji_ip_blocks_vd); + break; + case CHIP_TONGA: + adev->ip_blocks = tonga_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(tonga_ip_blocks_vd); + break; + case CHIP_POLARIS11: + case CHIP_POLARIS10: + adev->ip_blocks = polaris11_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(polaris11_ip_blocks_vd); + break; + + case CHIP_CARRIZO: + case CHIP_STONEY: + adev->ip_blocks = cz_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(cz_ip_blocks_vd); + break; + default: + /* FIXME: not supported yet */ + return -EINVAL; + } + } else { + switch (adev->asic_type) { + case CHIP_TOPAZ: + adev->ip_blocks = topaz_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(topaz_ip_blocks); + break; + case CHIP_FIJI: + adev->ip_blocks = fiji_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(fiji_ip_blocks); + break; + case CHIP_TONGA: + adev->ip_blocks = tonga_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(tonga_ip_blocks); + break; + case CHIP_POLARIS11: + case CHIP_POLARIS10: + adev->ip_blocks = polaris11_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(polaris11_ip_blocks); + break; + case CHIP_CARRIZO: + case CHIP_STONEY: + adev->ip_blocks = cz_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(cz_ip_blocks); + break; + default: + /* FIXME: not supported yet */ + return -EINVAL; + } } return 0; @@ -1248,8 +1615,17 @@ static int vi_common_early_init(void *handle) AMD_CG_SUPPORT_HDP_MGCG | AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_SDMA_MGCG | - AMD_CG_SUPPORT_SDMA_LS; + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_VCE_MGCG; + /* rev0 hardware requires workarounds to support PG */ adev->pg_flags = 0; + if (adev->rev_id != 0x00) { + adev->pg_flags |= AMD_PG_SUPPORT_GFX_PG | + AMD_PG_SUPPORT_GFX_SMG | + AMD_PG_SUPPORT_GFX_PIPELINE | + AMD_PG_SUPPORT_UVD | + AMD_PG_SUPPORT_VCE; + } adev->external_rev_id = adev->rev_id + 0x1; break; case CHIP_STONEY: @@ -1267,7 +1643,13 @@ static int vi_common_early_init(void *handle) AMD_CG_SUPPORT_HDP_MGCG | AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_SDMA_MGCG | - AMD_CG_SUPPORT_SDMA_LS; + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_VCE_MGCG; + adev->pg_flags |= AMD_PG_SUPPORT_GFX_PG | + AMD_PG_SUPPORT_GFX_SMG | + AMD_PG_SUPPORT_GFX_PIPELINE | + AMD_PG_SUPPORT_UVD | + AMD_PG_SUPPORT_VCE; adev->external_rev_id = adev->rev_id + 0x1; break; default: diff --git a/drivers/gpu/drm/amd/include/amd_shared.h b/drivers/gpu/drm/amd/include/amd_shared.h index a74a0d2ff1ca..db710418f35f 100644 --- a/drivers/gpu/drm/amd/include/amd_shared.h +++ b/drivers/gpu/drm/amd/include/amd_shared.h @@ -159,8 +159,14 @@ struct amd_ip_funcs { bool (*is_idle)(void *handle); /* poll for idle */ int (*wait_for_idle)(void *handle); + /* check soft reset the IP block */ + int (*check_soft_reset)(void *handle); + /* pre soft reset the IP block */ + int (*pre_soft_reset)(void *handle); /* soft reset the IP block */ int (*soft_reset)(void *handle); + /* post soft reset the IP block */ + int (*post_soft_reset)(void *handle); /* enable/disable cg for the IP block */ int (*set_clockgating_state)(void *handle, enum amd_clockgating_state state); diff --git a/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_4_2_d.h b/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_4_2_d.h index f3e53b118361..19802e96417e 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_4_2_d.h +++ b/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_4_2_d.h @@ -34,6 +34,7 @@ #define mmUVD_UDEC_ADDR_CONFIG 0x3bd3 #define mmUVD_UDEC_DB_ADDR_CONFIG 0x3bd4 #define mmUVD_UDEC_DBW_ADDR_CONFIG 0x3bd5 +#define mmUVD_NO_OP 0x3bff #define mmUVD_SEMA_CNTL 0x3d00 #define mmUVD_LMI_EXT40_ADDR 0x3d26 #define mmUVD_CTX_INDEX 0x3d28 diff --git a/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_5_0_d.h b/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_5_0_d.h index eb4cf53427da..cc972d237a7e 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_5_0_d.h +++ b/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_5_0_d.h @@ -34,6 +34,7 @@ #define mmUVD_UDEC_ADDR_CONFIG 0x3bd3 #define mmUVD_UDEC_DB_ADDR_CONFIG 0x3bd4 #define mmUVD_UDEC_DBW_ADDR_CONFIG 0x3bd5 +#define mmUVD_NO_OP 0x3bff #define mmUVD_LMI_RBC_RB_64BIT_BAR_LOW 0x3c69 #define mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH 0x3c68 #define mmUVD_LMI_RBC_IB_64BIT_BAR_LOW 0x3c67 diff --git a/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_6_0_d.h b/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_6_0_d.h index ec69869c55ff..378f4b6b43da 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_6_0_d.h +++ b/drivers/gpu/drm/amd/include/asic_reg/uvd/uvd_6_0_d.h @@ -35,6 +35,7 @@ #define mmUVD_UDEC_DB_ADDR_CONFIG 0x3bd4 #define mmUVD_UDEC_DBW_ADDR_CONFIG 0x3bd5 #define mmUVD_POWER_STATUS_U 0x3bfd +#define mmUVD_NO_OP 0x3bff #define mmUVD_LMI_RBC_RB_64BIT_BAR_LOW 0x3c69 #define mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH 0x3c68 #define mmUVD_LMI_RBC_IB_64BIT_BAR_LOW 0x3c67 diff --git a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c index abbb658bdc1e..2de34a5a85c2 100644 --- a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c +++ b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c @@ -31,6 +31,7 @@ #include "eventmanager.h" #include "pp_debug.h" + #define PP_CHECK(handle) \ do { \ if ((handle) == NULL || (handle)->pp_valid != PP_VALID) \ @@ -162,12 +163,12 @@ static int pp_hw_fini(void *handle) pp_handle = (struct pp_instance *)handle; eventmgr = pp_handle->eventmgr; - if (eventmgr != NULL || eventmgr->pp_eventmgr_fini != NULL) + if (eventmgr != NULL && eventmgr->pp_eventmgr_fini != NULL) eventmgr->pp_eventmgr_fini(eventmgr); smumgr = pp_handle->smu_mgr; - if (smumgr != NULL || smumgr->smumgr_funcs != NULL || + if (smumgr != NULL && smumgr->smumgr_funcs != NULL && smumgr->smumgr_funcs->smu_fini != NULL) smumgr->smumgr_funcs->smu_fini(smumgr); diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/Makefile b/drivers/gpu/drm/amd/powerplay/hwmgr/Makefile index f7ce4cb71346..abbcbc9f6eca 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/Makefile +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/Makefile @@ -4,13 +4,15 @@ HARDWARE_MGR = hwmgr.o processpptables.o functiontables.o \ hardwaremanager.o pp_acpi.o cz_hwmgr.o \ - cz_clockpowergating.o \ + cz_clockpowergating.o tonga_powertune.o\ tonga_processpptables.o ppatomctrl.o \ tonga_hwmgr.o pppcielanes.o tonga_thermal.o\ fiji_powertune.o fiji_hwmgr.o tonga_clockpowergating.o \ fiji_clockpowergating.o fiji_thermal.o \ polaris10_hwmgr.o polaris10_powertune.o polaris10_thermal.o \ - polaris10_clockpowergating.o + polaris10_clockpowergating.o iceland_hwmgr.o \ + iceland_clockpowergating.o iceland_thermal.o \ + iceland_powertune.o AMD_PP_HWMGR = $(addprefix $(AMD_PP_PATH)/hwmgr/,$(HARDWARE_MGR)) diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/cz_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/cz_hwmgr.c index 8cc0df9b534a..5ecef1732e20 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/cz_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/cz_hwmgr.c @@ -178,7 +178,6 @@ static int cz_initialize_dpm_defaults(struct pp_hwmgr *hwmgr) int result; cz_hwmgr->gfx_ramp_step = 256*25/100; - cz_hwmgr->gfx_ramp_delay = 1; /* by default, we delay 1us */ for (i = 0; i < CZ_MAX_HARDWARE_POWERLEVELS; i++) @@ -186,33 +185,19 @@ static int cz_initialize_dpm_defaults(struct pp_hwmgr *hwmgr) cz_hwmgr->mgcg_cgtt_local0 = 0x00000000; cz_hwmgr->mgcg_cgtt_local1 = 0x00000000; - cz_hwmgr->clock_slow_down_freq = 25000; - cz_hwmgr->skip_clock_slow_down = 1; - cz_hwmgr->enable_nb_ps_policy = 1; /* disable until UNB is ready, Enabled */ - cz_hwmgr->voltage_drop_in_dce_power_gating = 0; /* disable until fully verified */ - cz_hwmgr->voting_rights_clients = 0x00C00033; - cz_hwmgr->static_screen_threshold = 8; - cz_hwmgr->ddi_power_gating_disabled = 0; - cz_hwmgr->bapm_enabled = 1; - cz_hwmgr->voltage_drop_threshold = 0; - cz_hwmgr->gfx_power_gating_threshold = 500; - cz_hwmgr->vce_slow_sclk_threshold = 20000; - cz_hwmgr->dce_slow_sclk_threshold = 30000; - cz_hwmgr->disable_driver_thermal_policy = 1; - cz_hwmgr->disable_nb_ps3_in_battery = 0; phm_cap_unset(hwmgr->platform_descriptor.platformCaps, @@ -221,9 +206,6 @@ static int cz_initialize_dpm_defaults(struct pp_hwmgr *hwmgr) phm_cap_set(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_NonABMSupportInPPLib); - phm_cap_set(hwmgr->platform_descriptor.platformCaps, - PHM_PlatformCaps_SclkDeepSleep); - phm_cap_unset(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_DynamicM3Arbiter); @@ -233,9 +215,7 @@ static int cz_initialize_dpm_defaults(struct pp_hwmgr *hwmgr) PHM_PlatformCaps_DynamicPatchPowerState); cz_hwmgr->thermal_auto_throttling_treshold = 0; - cz_hwmgr->tdr_clock = 0; - cz_hwmgr->disable_gfx_power_gating_in_uvd = 0; phm_cap_set(hwmgr->platform_descriptor.platformCaps, @@ -450,19 +430,12 @@ static int cz_construct_boot_state(struct pp_hwmgr *hwmgr) (uint8_t)cz_hwmgr->sys_info.bootup_nb_voltage_index; cz_hwmgr->boot_power_level.dsDividerIndex = 0; - cz_hwmgr->boot_power_level.ssDividerIndex = 0; - cz_hwmgr->boot_power_level.allowGnbSlow = 1; - cz_hwmgr->boot_power_level.forceNBPstate = 0; - cz_hwmgr->boot_power_level.hysteresis_up = 0; - cz_hwmgr->boot_power_level.numSIMDToPowerDown = 0; - cz_hwmgr->boot_power_level.display_wm = 0; - cz_hwmgr->boot_power_level.vce_wm = 0; return 0; @@ -749,7 +722,6 @@ static int cz_tf_update_sclk_limit(struct pp_hwmgr *hwmgr, cz_hwmgr->sclk_dpm.soft_max_clk = table->entries[table->count - 1].clk; clock = hwmgr->display_config.min_core_set_clock; -; if (clock == 0) printk(KERN_INFO "[ powerplay ] min_core_set_clock not set\n"); @@ -832,7 +804,7 @@ static int cz_tf_set_watermark_threshold(struct pp_hwmgr *hwmgr, smum_send_msg_to_smc_with_parameter(hwmgr->smumgr, PPSMC_MSG_SetWatermarkFrequency, - cz_hwmgr->sclk_dpm.soft_max_clk); + cz_hwmgr->sclk_dpm.soft_max_clk); return 0; } @@ -858,9 +830,9 @@ static int cz_tf_enable_nb_dpm(struct pp_hwmgr *hwmgr, PP_DBG_LOG("enabling ALL SMU features.\n"); dpm_features |= NB_DPM_MASK; ret = smum_send_msg_to_smc_with_parameter( - hwmgr->smumgr, - PPSMC_MSG_EnableAllSmuFeatures, - dpm_features); + hwmgr->smumgr, + PPSMC_MSG_EnableAllSmuFeatures, + dpm_features); if (ret == 0) cz_hwmgr->is_nb_dpm_enabled = true; } @@ -1246,7 +1218,7 @@ static int cz_hwmgr_backend_init(struct pp_hwmgr *hwmgr) static int cz_hwmgr_backend_fini(struct pp_hwmgr *hwmgr) { - if (hwmgr != NULL || hwmgr->backend != NULL) { + if (hwmgr != NULL && hwmgr->backend != NULL) { kfree(hwmgr->backend); kfree(hwmgr); } @@ -1402,10 +1374,12 @@ int cz_dpm_update_uvd_dpm(struct pp_hwmgr *hwmgr, bool bgate) PPSMC_MSG_SetUvdHardMin)); cz_enable_disable_uvd_dpm(hwmgr, true); - } else + } else { cz_enable_disable_uvd_dpm(hwmgr, true); - } else + } + } else { cz_enable_disable_uvd_dpm(hwmgr, false); + } return 0; } @@ -1690,13 +1664,10 @@ static int cz_store_cc6_data(struct pp_hwmgr *hwmgr, uint32_t separation_time, struct cz_hwmgr *hw_data = (struct cz_hwmgr *)(hwmgr->backend); if (separation_time != - hw_data->cc6_settings.cpu_pstate_separation_time - || cc6_disable != - hw_data->cc6_settings.cpu_cc6_disable - || pstate_disable != - hw_data->cc6_settings.cpu_pstate_disable - || pstate_switch_disable != - hw_data->cc6_settings.nb_pstate_switch_disable) { + hw_data->cc6_settings.cpu_pstate_separation_time || + cc6_disable != hw_data->cc6_settings.cpu_cc6_disable || + pstate_disable != hw_data->cc6_settings.cpu_pstate_disable || + pstate_switch_disable != hw_data->cc6_settings.nb_pstate_switch_disable) { hw_data->cc6_settings.cc6_setting_changed = true; @@ -1799,8 +1770,7 @@ static int cz_get_performance_level(struct pp_hwmgr *hwmgr, const struct pp_hw_p ps = cast_const_PhwCzPowerState(state); level_index = index > ps->level - 1 ? ps->level - 1 : index; - - level->coreClock = ps->levels[level_index].engineClock; + level->coreClock = ps->levels[level_index].engineClock; if (designation == PHM_PerformanceLevelDesignation_PowerContainment) { for (i = 1; i < ps->level; i++) { diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_hwmgr.c index 6483d680bbc8..9368e21f5695 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_hwmgr.c @@ -618,9 +618,6 @@ static int fiji_hwmgr_backend_init(struct pp_hwmgr *hwmgr) phm_cap_set(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_TablelessHardwareInterface); - phm_cap_set(hwmgr->platform_descriptor.platformCaps, - PHM_PlatformCaps_SclkDeepSleep); - data->gpio_debug = 0; phm_cap_set(hwmgr->platform_descriptor.platformCaps, diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_powertune.c b/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_powertune.c index 44658451a8d2..f5992ea0c56f 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_powertune.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_powertune.c @@ -57,8 +57,6 @@ void fiji_initialize_power_tune_defaults(struct pp_hwmgr *hwmgr) /* Assume disabled */ phm_cap_unset(hwmgr->platform_descriptor.platformCaps, - PHM_PlatformCaps_PowerContainment); - phm_cap_unset(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_CAC); phm_cap_unset(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_SQRamping); @@ -77,9 +75,8 @@ void fiji_initialize_power_tune_defaults(struct pp_hwmgr *hwmgr) fiji_hwmgr->fast_watermark_threshold = 100; - if (hwmgr->powercontainment_enabled) { - phm_cap_set(hwmgr->platform_descriptor.platformCaps, - PHM_PlatformCaps_PowerContainment); + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment)) { tmp = 1; fiji_hwmgr->enable_dte_feature = tmp ? false : true; fiji_hwmgr->enable_tdc_limit_feature = tmp ? true : false; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c index 27e07624ac28..d829076ed9ea 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c @@ -39,6 +39,26 @@ extern int cz_hwmgr_init(struct pp_hwmgr *hwmgr); extern int tonga_hwmgr_init(struct pp_hwmgr *hwmgr); extern int fiji_hwmgr_init(struct pp_hwmgr *hwmgr); extern int polaris10_hwmgr_init(struct pp_hwmgr *hwmgr); +extern int iceland_hwmgr_init(struct pp_hwmgr *hwmgr); + +static int hwmgr_set_features_platform_caps(struct pp_hwmgr *hwmgr) +{ + if (amdgpu_sclk_deep_sleep_en) + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SclkDeepSleep); + else + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SclkDeepSleep); + + if (amdgpu_powercontainment) + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment); + else + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment); + + return 0; +} int hwmgr_init(struct amd_pp_init *pp_init, struct pp_instance *handle) { @@ -57,9 +77,12 @@ int hwmgr_init(struct amd_pp_init *pp_init, struct pp_instance *handle) hwmgr->chip_family = pp_init->chip_family; hwmgr->chip_id = pp_init->chip_id; hwmgr->hw_revision = pp_init->rev_id; + hwmgr->sub_sys_id = pp_init->sub_sys_id; + hwmgr->sub_vendor_id = pp_init->sub_vendor_id; hwmgr->usec_timeout = AMD_MAX_USEC_TIMEOUT; hwmgr->power_source = PP_PowerSource_AC; - hwmgr->powercontainment_enabled = pp_init->powercontainment_enabled; + + hwmgr_set_features_platform_caps(hwmgr); switch (hwmgr->chip_family) { case AMDGPU_FAMILY_CZ: @@ -67,6 +90,9 @@ int hwmgr_init(struct amd_pp_init *pp_init, struct pp_instance *handle) break; case AMDGPU_FAMILY_VI: switch (hwmgr->chip_id) { + case CHIP_TOPAZ: + iceland_hwmgr_init(hwmgr); + break; case CHIP_TONGA: tonga_hwmgr_init(hwmgr); break; @@ -182,29 +208,7 @@ int phm_wait_on_register(struct pp_hwmgr *hwmgr, uint32_t index, return 0; } -int phm_wait_for_register_unequal(struct pp_hwmgr *hwmgr, - uint32_t index, uint32_t value, uint32_t mask) -{ - uint32_t i; - uint32_t cur_value; - if (hwmgr == NULL || hwmgr->device == NULL) { - printk(KERN_ERR "[ powerplay ] Invalid Hardware Manager!"); - return -EINVAL; - } - - for (i = 0; i < hwmgr->usec_timeout; i++) { - cur_value = cgs_read_register(hwmgr->device, index); - if ((cur_value & mask) != (value & mask)) - break; - udelay(1); - } - - /* timeout means wrong logic*/ - if (i == hwmgr->usec_timeout) - return -1; - return 0; -} /** @@ -227,21 +231,7 @@ void phm_wait_on_indirect_register(struct pp_hwmgr *hwmgr, phm_wait_on_register(hwmgr, indirect_port + 1, mask, value); } -void phm_wait_for_indirect_register_unequal(struct pp_hwmgr *hwmgr, - uint32_t indirect_port, - uint32_t index, - uint32_t value, - uint32_t mask) -{ - if (hwmgr == NULL || hwmgr->device == NULL) { - printk(KERN_ERR "[ powerplay ] Invalid Hardware Manager!"); - return; - } - cgs_write_register(hwmgr->device, indirect_port, index); - phm_wait_for_register_unequal(hwmgr, indirect_port + 1, - value, mask); -} bool phm_cf_want_uvd_power_gating(struct pp_hwmgr *hwmgr) { diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_clockpowergating.c b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_clockpowergating.c new file mode 100644 index 000000000000..47949f5cd073 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_clockpowergating.c @@ -0,0 +1,119 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Author: Huang Rui <ray.huang@amd.com> + * + */ + +#include "hwmgr.h" +#include "iceland_clockpowergating.h" +#include "ppsmc.h" +#include "iceland_hwmgr.h" + +int iceland_phm_powerdown_uvd(struct pp_hwmgr *hwmgr) +{ + /* iceland does not have MM hardware block */ + return 0; +} + +static int iceland_phm_powerup_uvd(struct pp_hwmgr *hwmgr) +{ + /* iceland does not have MM hardware block */ + return 0; +} + +static int iceland_phm_powerdown_vce(struct pp_hwmgr *hwmgr) +{ + /* iceland does not have MM hardware block */ + return 0; +} + +static int iceland_phm_powerup_vce(struct pp_hwmgr *hwmgr) +{ + /* iceland does not have MM hardware block */ + return 0; +} + +int iceland_phm_set_asic_block_gating(struct pp_hwmgr *hwmgr, enum + PHM_AsicBlock block, enum PHM_ClockGateSetting gating) +{ + int ret = 0; + + switch (block) { + case PHM_AsicBlock_UVD_MVC: + case PHM_AsicBlock_UVD: + case PHM_AsicBlock_UVD_HD: + case PHM_AsicBlock_UVD_SD: + if (gating == PHM_ClockGateSetting_StaticOff) + ret = iceland_phm_powerdown_uvd(hwmgr); + else + ret = iceland_phm_powerup_uvd(hwmgr); + break; + case PHM_AsicBlock_GFX: + default: + break; + } + + return ret; +} + +int iceland_phm_disable_clock_power_gating(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + data->uvd_power_gated = false; + data->vce_power_gated = false; + + iceland_phm_powerup_uvd(hwmgr); + iceland_phm_powerup_vce(hwmgr); + + return 0; +} + +int iceland_phm_powergate_uvd(struct pp_hwmgr *hwmgr, bool bgate) +{ + if (bgate) { + iceland_update_uvd_dpm(hwmgr, true); + iceland_phm_powerdown_uvd(hwmgr); + } else { + iceland_phm_powerup_uvd(hwmgr); + iceland_update_uvd_dpm(hwmgr, false); + } + + return 0; +} + +int iceland_phm_powergate_vce(struct pp_hwmgr *hwmgr, bool bgate) +{ + if (bgate) + return iceland_phm_powerdown_vce(hwmgr); + else + return iceland_phm_powerup_vce(hwmgr); + + return 0; +} + +int iceland_phm_update_clock_gatings(struct pp_hwmgr *hwmgr, + const uint32_t *msg_id) +{ + /* iceland does not have MM hardware block */ + return 0; +} diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_clockpowergating.h b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_clockpowergating.h new file mode 100644 index 000000000000..ff5ef00c7c68 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_clockpowergating.h @@ -0,0 +1,38 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Author: Huang Rui <ray.huang@amd.com> + * + */ + +#ifndef _ICELAND_CLOCK_POWER_GATING_H_ +#define _ICELAND_CLOCK_POWER_GATING_H_ + +#include "iceland_hwmgr.h" +#include "pp_asicblocks.h" + +extern int iceland_phm_set_asic_block_gating(struct pp_hwmgr *hwmgr, enum PHM_AsicBlock block, enum PHM_ClockGateSetting gating); +extern int iceland_phm_powergate_vce(struct pp_hwmgr *hwmgr, bool bgate); +extern int iceland_phm_powergate_uvd(struct pp_hwmgr *hwmgr, bool bgate); +extern int iceland_phm_powerdown_uvd(struct pp_hwmgr *hwmgr); +extern int iceland_phm_disable_clock_power_gating(struct pp_hwmgr *hwmgr); +extern int iceland_phm_update_clock_gatings(struct pp_hwmgr *hwmgr, const uint32_t *msg_id); +#endif /* _ICELAND_CLOCK_POWER_GATING_H_ */ diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_dyn_defaults.h b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_dyn_defaults.h new file mode 100644 index 000000000000..a7b4bc6caea2 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_dyn_defaults.h @@ -0,0 +1,41 @@ +#ifndef ICELAND_DYN_DEFAULTS_H +#define ICELAND_DYN_DEFAULTS_H + +enum ICELANDdpm_TrendDetection +{ + ICELANDdpm_TrendDetection_AUTO, + ICELANDdpm_TrendDetection_UP, + ICELANDdpm_TrendDetection_DOWN +}; +typedef enum ICELANDdpm_TrendDetection ICELANDdpm_TrendDetection; + + +#define PPICELAND_VOTINGRIGHTSCLIENTS_DFLT0 0x3FFFC102 +#define PPICELAND_VOTINGRIGHTSCLIENTS_DFLT1 0x000400 +#define PPICELAND_VOTINGRIGHTSCLIENTS_DFLT2 0xC00080 +#define PPICELAND_VOTINGRIGHTSCLIENTS_DFLT3 0xC00200 +#define PPICELAND_VOTINGRIGHTSCLIENTS_DFLT4 0xC01680 +#define PPICELAND_VOTINGRIGHTSCLIENTS_DFLT5 0xC00033 +#define PPICELAND_VOTINGRIGHTSCLIENTS_DFLT6 0xC00033 +#define PPICELAND_VOTINGRIGHTSCLIENTS_DFLT7 0x3FFFC000 + + +#define PPICELAND_THERMALPROTECTCOUNTER_DFLT 0x200 + +#define PPICELAND_STATICSCREENTHRESHOLDUNIT_DFLT 0 + +#define PPICELAND_STATICSCREENTHRESHOLD_DFLT 0x00C8 + +#define PPICELAND_GFXIDLECLOCKSTOPTHRESHOLD_DFLT 0x200 + +#define PPICELAND_REFERENCEDIVIDER_DFLT 4 + +#define PPICELAND_ULVVOLTAGECHANGEDELAY_DFLT 1687 + +#define PPICELAND_CGULVPARAMETER_DFLT 0x00040035 +#define PPICELAND_CGULVCONTROL_DFLT 0x00007450 +#define PPICELAND_TARGETACTIVITY_DFLT 30 +#define PPICELAND_MCLK_TARGETACTIVITY_DFLT 10 + +#endif + diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_hwmgr.c new file mode 100644 index 000000000000..8a7ada50551c --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_hwmgr.c @@ -0,0 +1,5692 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Author: Huang Rui <ray.huang@amd.com> + * + */ +#include <linux/module.h> +#include <linux/slab.h> +#include <linux/fb.h> +#include "linux/delay.h" +#include "pp_acpi.h" +#include "hwmgr.h" +#include <atombios.h> +#include "iceland_hwmgr.h" +#include "pptable.h" +#include "processpptables.h" +#include "pp_debug.h" +#include "ppsmc.h" +#include "cgs_common.h" +#include "pppcielanes.h" +#include "iceland_dyn_defaults.h" +#include "smumgr.h" +#include "iceland_smumgr.h" +#include "iceland_clockpowergating.h" +#include "iceland_thermal.h" +#include "iceland_powertune.h" + +#include "gmc/gmc_8_1_d.h" +#include "gmc/gmc_8_1_sh_mask.h" + +#include "bif/bif_5_0_d.h" +#include "bif/bif_5_0_sh_mask.h" + +#include "smu/smu_7_1_1_d.h" +#include "smu/smu_7_1_1_sh_mask.h" + +#include "cgs_linux.h" +#include "eventmgr.h" +#include "amd_pcie_helpers.h" + +#define MC_CG_ARB_FREQ_F0 0x0a +#define MC_CG_ARB_FREQ_F1 0x0b +#define MC_CG_ARB_FREQ_F2 0x0c +#define MC_CG_ARB_FREQ_F3 0x0d + +#define MC_CG_SEQ_DRAMCONF_S0 0x05 +#define MC_CG_SEQ_DRAMCONF_S1 0x06 +#define MC_CG_SEQ_YCLK_SUSPEND 0x04 +#define MC_CG_SEQ_YCLK_RESUME 0x0a + +#define PCIE_BUS_CLK 10000 +#define TCLK (PCIE_BUS_CLK / 10) + +#define SMC_RAM_END 0x40000 +#define SMC_CG_IND_START 0xc0030000 +#define SMC_CG_IND_END 0xc0040000 /* First byte after SMC_CG_IND*/ + +#define VOLTAGE_SCALE 4 +#define VOLTAGE_VID_OFFSET_SCALE1 625 +#define VOLTAGE_VID_OFFSET_SCALE2 100 + +const uint32_t iceland_magic = (uint32_t)(PHM_VIslands_Magic); + +#define MC_SEQ_MISC0_GDDR5_SHIFT 28 +#define MC_SEQ_MISC0_GDDR5_MASK 0xf0000000 +#define MC_SEQ_MISC0_GDDR5_VALUE 5 + +/** Values for the CG_THERMAL_CTRL::DPM_EVENT_SRC field. */ +enum DPM_EVENT_SRC { + DPM_EVENT_SRC_ANALOG = 0, /* Internal analog trip point */ + DPM_EVENT_SRC_EXTERNAL = 1, /* External (GPIO 17) signal */ + DPM_EVENT_SRC_DIGITAL = 2, /* Internal digital trip point (DIG_THERM_DPM) */ + DPM_EVENT_SRC_ANALOG_OR_EXTERNAL = 3, /* Internal analog or external */ + DPM_EVENT_SRC_DIGITAL_OR_EXTERNAL = 4 /* Internal digital or external */ +}; + +static int iceland_read_clock_registers(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + data->clock_registers.vCG_SPLL_FUNC_CNTL = + cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_SPLL_FUNC_CNTL); + data->clock_registers.vCG_SPLL_FUNC_CNTL_2 = + cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_SPLL_FUNC_CNTL_2); + data->clock_registers.vCG_SPLL_FUNC_CNTL_3 = + cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_SPLL_FUNC_CNTL_3); + data->clock_registers.vCG_SPLL_FUNC_CNTL_4 = + cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_SPLL_FUNC_CNTL_4); + data->clock_registers.vCG_SPLL_SPREAD_SPECTRUM = + cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_SPLL_SPREAD_SPECTRUM); + data->clock_registers.vCG_SPLL_SPREAD_SPECTRUM_2 = + cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_SPLL_SPREAD_SPECTRUM_2); + data->clock_registers.vDLL_CNTL = + cgs_read_register(hwmgr->device, mmDLL_CNTL); + data->clock_registers.vMCLK_PWRMGT_CNTL = + cgs_read_register(hwmgr->device, mmMCLK_PWRMGT_CNTL); + data->clock_registers.vMPLL_AD_FUNC_CNTL = + cgs_read_register(hwmgr->device, mmMPLL_AD_FUNC_CNTL); + data->clock_registers.vMPLL_DQ_FUNC_CNTL = + cgs_read_register(hwmgr->device, mmMPLL_DQ_FUNC_CNTL); + data->clock_registers.vMPLL_FUNC_CNTL = + cgs_read_register(hwmgr->device, mmMPLL_FUNC_CNTL); + data->clock_registers.vMPLL_FUNC_CNTL_1 = + cgs_read_register(hwmgr->device, mmMPLL_FUNC_CNTL_1); + data->clock_registers.vMPLL_FUNC_CNTL_2 = + cgs_read_register(hwmgr->device, mmMPLL_FUNC_CNTL_2); + data->clock_registers.vMPLL_SS1 = + cgs_read_register(hwmgr->device, mmMPLL_SS1); + data->clock_registers.vMPLL_SS2 = + cgs_read_register(hwmgr->device, mmMPLL_SS2); + + return 0; +} + +/** + * Find out if memory is GDDR5. + * + * @param hwmgr the address of the powerplay hardware manager. + * @return always 0 + */ +int iceland_get_memory_type(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + uint32_t temp; + + temp = cgs_read_register(hwmgr->device, mmMC_SEQ_MISC0); + + data->is_memory_GDDR5 = (MC_SEQ_MISC0_GDDR5_VALUE == + ((temp & MC_SEQ_MISC0_GDDR5_MASK) >> + MC_SEQ_MISC0_GDDR5_SHIFT)); + + return 0; +} + +int iceland_update_uvd_dpm(struct pp_hwmgr *hwmgr, bool bgate) +{ + /* iceland does not have MM hardware blocks */ + return 0; +} + +/** + * Enables Dynamic Power Management by SMC + * + * @param hwmgr the address of the powerplay hardware manager. + * @return always 0 + */ +int iceland_enable_acpi_power_management(struct pp_hwmgr *hwmgr) +{ + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, GENERAL_PWRMGT, STATIC_PM_EN, 1); + + return 0; +} + +/** + * Find the MC microcode version and store it in the HwMgr struct + * + * @param hwmgr the address of the powerplay hardware manager. + * @return always 0 + */ +int iceland_get_mc_microcode_version(struct pp_hwmgr *hwmgr) +{ + cgs_write_register(hwmgr->device, mmMC_SEQ_IO_DEBUG_INDEX, 0x9F); + + hwmgr->microcode_version_info.MC = cgs_read_register(hwmgr->device, mmMC_SEQ_IO_DEBUG_DATA); + + return 0; +} + +static int iceland_init_sclk_threshold(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + data->low_sclk_interrupt_threshold = 0; + + return 0; +} + + +static int iceland_setup_asic_task(struct pp_hwmgr *hwmgr) +{ + int tmp_result, result = 0; + + tmp_result = iceland_read_clock_registers(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to read clock registers!", result = tmp_result); + + tmp_result = iceland_get_memory_type(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to get memory type!", result = tmp_result); + + tmp_result = iceland_enable_acpi_power_management(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to enable ACPI power management!", result = tmp_result); + + tmp_result = iceland_get_mc_microcode_version(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to get MC microcode version!", result = tmp_result); + + tmp_result = iceland_init_sclk_threshold(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to init sclk threshold!", result = tmp_result); + + return result; +} + +static bool cf_iceland_voltage_control(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + return ICELAND_VOLTAGE_CONTROL_NONE != data->voltage_control; +} + +/* + * -------------- Voltage Tables ---------------------- + * If the voltage table would be bigger than what will fit into the + * state table on the SMC keep only the higher entries. + */ + +static void iceland_trim_voltage_table_to_fit_state_table( + struct pp_hwmgr *hwmgr, + uint32_t max_voltage_steps, + pp_atomctrl_voltage_table *voltage_table) +{ + unsigned int i, diff; + + if (voltage_table->count <= max_voltage_steps) { + return; + } + + diff = voltage_table->count - max_voltage_steps; + + for (i = 0; i < max_voltage_steps; i++) { + voltage_table->entries[i] = voltage_table->entries[i + diff]; + } + + voltage_table->count = max_voltage_steps; + + return; +} + +/** + * Enable voltage control + * + * @param hwmgr the address of the powerplay hardware manager. + * @return always 0 + */ +int iceland_enable_voltage_control(struct pp_hwmgr *hwmgr) +{ + /* enable voltage control */ + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, GENERAL_PWRMGT, VOLT_PWRMGT_EN, 1); + + return 0; +} + +static int iceland_get_svi2_voltage_table(struct pp_hwmgr *hwmgr, + struct phm_clock_voltage_dependency_table *voltage_dependency_table, + pp_atomctrl_voltage_table *voltage_table) +{ + uint32_t i; + + PP_ASSERT_WITH_CODE((NULL != voltage_table), + "Voltage Dependency Table empty.", return -EINVAL;); + + voltage_table->mask_low = 0; + voltage_table->phase_delay = 0; + voltage_table->count = voltage_dependency_table->count; + + for (i = 0; i < voltage_dependency_table->count; i++) { + voltage_table->entries[i].value = + voltage_dependency_table->entries[i].v; + voltage_table->entries[i].smio_low = 0; + } + + return 0; +} + +/** + * Create Voltage Tables. + * + * @param hwmgr the address of the powerplay hardware manager. + * @return always 0 + */ +int iceland_construct_voltage_tables(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + int result; + + /* GPIO voltage */ + if (ICELAND_VOLTAGE_CONTROL_BY_GPIO == data->voltage_control) { + result = atomctrl_get_voltage_table_v3(hwmgr, + VOLTAGE_TYPE_VDDC, VOLTAGE_OBJ_GPIO_LUT, + &data->vddc_voltage_table); + PP_ASSERT_WITH_CODE((0 == result), + "Failed to retrieve VDDC table.", return result;); + } else if (ICELAND_VOLTAGE_CONTROL_BY_SVID2 == data->voltage_control) { + /* SVI2 VDDC voltage */ + result = iceland_get_svi2_voltage_table(hwmgr, + hwmgr->dyn_state.vddc_dependency_on_mclk, + &data->vddc_voltage_table); + PP_ASSERT_WITH_CODE((0 == result), + "Failed to retrieve SVI2 VDDC table from dependancy table.", return result;); + } + + PP_ASSERT_WITH_CODE( + (data->vddc_voltage_table.count <= (SMU71_MAX_LEVELS_VDDC)), + "Too many voltage values for VDDC. Trimming to fit state table.", + iceland_trim_voltage_table_to_fit_state_table(hwmgr, + SMU71_MAX_LEVELS_VDDC, &(data->vddc_voltage_table)); + ); + + /* GPIO */ + if (ICELAND_VOLTAGE_CONTROL_BY_GPIO == data->vdd_ci_control) { + result = atomctrl_get_voltage_table_v3(hwmgr, + VOLTAGE_TYPE_VDDCI, VOLTAGE_OBJ_GPIO_LUT, &(data->vddci_voltage_table)); + PP_ASSERT_WITH_CODE((0 == result), + "Failed to retrieve VDDCI table.", return result;); + } + + /* SVI2 VDDCI voltage */ + if (ICELAND_VOLTAGE_CONTROL_BY_SVID2 == data->vdd_ci_control) { + result = iceland_get_svi2_voltage_table(hwmgr, + hwmgr->dyn_state.vddci_dependency_on_mclk, + &data->vddci_voltage_table); + PP_ASSERT_WITH_CODE((0 == result), + "Failed to retrieve SVI2 VDDCI table from dependancy table.", return result;); + } + + PP_ASSERT_WITH_CODE( + (data->vddci_voltage_table.count <= (SMU71_MAX_LEVELS_VDDCI)), + "Too many voltage values for VDDCI. Trimming to fit state table.", + iceland_trim_voltage_table_to_fit_state_table(hwmgr, + SMU71_MAX_LEVELS_VDDCI, &(data->vddci_voltage_table)); + ); + + + /* GPIO */ + if (ICELAND_VOLTAGE_CONTROL_BY_GPIO == data->mvdd_control) { + result = atomctrl_get_voltage_table_v3(hwmgr, + VOLTAGE_TYPE_MVDDC, VOLTAGE_OBJ_GPIO_LUT, &(data->mvdd_voltage_table)); + PP_ASSERT_WITH_CODE((0 == result), + "Failed to retrieve table.", return result;); + } + + /* SVI2 voltage control */ + if (ICELAND_VOLTAGE_CONTROL_BY_SVID2 == data->mvdd_control) { + result = iceland_get_svi2_voltage_table(hwmgr, + hwmgr->dyn_state.mvdd_dependency_on_mclk, + &data->mvdd_voltage_table); + PP_ASSERT_WITH_CODE((0 == result), + "Failed to retrieve SVI2 MVDD table from dependancy table.", return result;); + } + + PP_ASSERT_WITH_CODE( + (data->mvdd_voltage_table.count <= (SMU71_MAX_LEVELS_MVDD)), + "Too many voltage values for MVDD. Trimming to fit state table.", + iceland_trim_voltage_table_to_fit_state_table(hwmgr, + SMU71_MAX_LEVELS_MVDD, &(data->mvdd_voltage_table)); + ); + + return 0; +} + +/*---------------------------MC----------------------------*/ + +uint8_t iceland_get_memory_module_index(struct pp_hwmgr *hwmgr) +{ + return (uint8_t) (0xFF & (cgs_read_register(hwmgr->device, mmBIOS_SCRATCH_4) >> 16)); +} + +bool iceland_check_s0_mc_reg_index(uint16_t inReg, uint16_t *outReg) +{ + bool result = true; + + switch (inReg) { + case mmMC_SEQ_RAS_TIMING: + *outReg = mmMC_SEQ_RAS_TIMING_LP; + break; + + case mmMC_SEQ_DLL_STBY: + *outReg = mmMC_SEQ_DLL_STBY_LP; + break; + + case mmMC_SEQ_G5PDX_CMD0: + *outReg = mmMC_SEQ_G5PDX_CMD0_LP; + break; + + case mmMC_SEQ_G5PDX_CMD1: + *outReg = mmMC_SEQ_G5PDX_CMD1_LP; + break; + + case mmMC_SEQ_G5PDX_CTRL: + *outReg = mmMC_SEQ_G5PDX_CTRL_LP; + break; + + case mmMC_SEQ_CAS_TIMING: + *outReg = mmMC_SEQ_CAS_TIMING_LP; + break; + + case mmMC_SEQ_MISC_TIMING: + *outReg = mmMC_SEQ_MISC_TIMING_LP; + break; + + case mmMC_SEQ_MISC_TIMING2: + *outReg = mmMC_SEQ_MISC_TIMING2_LP; + break; + + case mmMC_SEQ_PMG_DVS_CMD: + *outReg = mmMC_SEQ_PMG_DVS_CMD_LP; + break; + + case mmMC_SEQ_PMG_DVS_CTL: + *outReg = mmMC_SEQ_PMG_DVS_CTL_LP; + break; + + case mmMC_SEQ_RD_CTL_D0: + *outReg = mmMC_SEQ_RD_CTL_D0_LP; + break; + + case mmMC_SEQ_RD_CTL_D1: + *outReg = mmMC_SEQ_RD_CTL_D1_LP; + break; + + case mmMC_SEQ_WR_CTL_D0: + *outReg = mmMC_SEQ_WR_CTL_D0_LP; + break; + + case mmMC_SEQ_WR_CTL_D1: + *outReg = mmMC_SEQ_WR_CTL_D1_LP; + break; + + case mmMC_PMG_CMD_EMRS: + *outReg = mmMC_SEQ_PMG_CMD_EMRS_LP; + break; + + case mmMC_PMG_CMD_MRS: + *outReg = mmMC_SEQ_PMG_CMD_MRS_LP; + break; + + case mmMC_PMG_CMD_MRS1: + *outReg = mmMC_SEQ_PMG_CMD_MRS1_LP; + break; + + case mmMC_SEQ_PMG_TIMING: + *outReg = mmMC_SEQ_PMG_TIMING_LP; + break; + + case mmMC_PMG_CMD_MRS2: + *outReg = mmMC_SEQ_PMG_CMD_MRS2_LP; + break; + + case mmMC_SEQ_WR_CTL_2: + *outReg = mmMC_SEQ_WR_CTL_2_LP; + break; + + default: + result = false; + break; + } + + return result; +} + +int iceland_set_s0_mc_reg_index(phw_iceland_mc_reg_table *table) +{ + uint32_t i; + uint16_t address; + + for (i = 0; i < table->last; i++) { + table->mc_reg_address[i].s0 = + iceland_check_s0_mc_reg_index(table->mc_reg_address[i].s1, &address) + ? address : table->mc_reg_address[i].s1; + } + return 0; +} + +int iceland_copy_vbios_smc_reg_table(const pp_atomctrl_mc_reg_table *table, phw_iceland_mc_reg_table *ni_table) +{ + uint8_t i, j; + + PP_ASSERT_WITH_CODE((table->last <= SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE), + "Invalid VramInfo table.", return -1); + PP_ASSERT_WITH_CODE((table->num_entries <= MAX_AC_TIMING_ENTRIES), + "Invalid VramInfo table.", return -1); + + for (i = 0; i < table->last; i++) { + ni_table->mc_reg_address[i].s1 = table->mc_reg_address[i].s1; + } + ni_table->last = table->last; + + for (i = 0; i < table->num_entries; i++) { + ni_table->mc_reg_table_entry[i].mclk_max = + table->mc_reg_table_entry[i].mclk_max; + for (j = 0; j < table->last; j++) { + ni_table->mc_reg_table_entry[i].mc_data[j] = + table->mc_reg_table_entry[i].mc_data[j]; + } + } + + ni_table->num_entries = table->num_entries; + + return 0; +} + +/** + * VBIOS omits some information to reduce size, we need to recover them here. + * 1. when we see mmMC_SEQ_MISC1, bit[31:16] EMRS1, need to be write to mmMC_PMG_CMD_EMRS /_LP[15:0]. + * Bit[15:0] MRS, need to be update mmMC_PMG_CMD_MRS/_LP[15:0] + * 2. when we see mmMC_SEQ_RESERVE_M, bit[15:0] EMRS2, need to be write to mmMC_PMG_CMD_MRS1/_LP[15:0]. + * 3. need to set these data for each clock range + * + * @param hwmgr the address of the powerplay hardware manager. + * @param table the address of MCRegTable + * @return always 0 + */ +static int iceland_set_mc_special_registers(struct pp_hwmgr *hwmgr, phw_iceland_mc_reg_table *table) +{ + uint8_t i, j, k; + uint32_t temp_reg; + const iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + for (i = 0, j = table->last; i < table->last; i++) { + PP_ASSERT_WITH_CODE((j < SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE), + "Invalid VramInfo table.", return -1); + switch (table->mc_reg_address[i].s1) { + /* + * mmMC_SEQ_MISC1, bit[31:16] EMRS1, need to be write + * to mmMC_PMG_CMD_EMRS/_LP[15:0]. Bit[15:0] MRS, need + * to be update mmMC_PMG_CMD_MRS/_LP[15:0] + */ + case mmMC_SEQ_MISC1: + temp_reg = cgs_read_register(hwmgr->device, mmMC_PMG_CMD_EMRS); + table->mc_reg_address[j].s1 = mmMC_PMG_CMD_EMRS; + table->mc_reg_address[j].s0 = mmMC_SEQ_PMG_CMD_EMRS_LP; + for (k = 0; k < table->num_entries; k++) { + table->mc_reg_table_entry[k].mc_data[j] = + ((temp_reg & 0xffff0000)) | + ((table->mc_reg_table_entry[k].mc_data[i] & 0xffff0000) >> 16); + } + j++; + PP_ASSERT_WITH_CODE((j < SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE), + "Invalid VramInfo table.", return -1); + + temp_reg = cgs_read_register(hwmgr->device, mmMC_PMG_CMD_MRS); + table->mc_reg_address[j].s1 = mmMC_PMG_CMD_MRS; + table->mc_reg_address[j].s0 = mmMC_SEQ_PMG_CMD_MRS_LP; + for (k = 0; k < table->num_entries; k++) { + table->mc_reg_table_entry[k].mc_data[j] = + (temp_reg & 0xffff0000) | + (table->mc_reg_table_entry[k].mc_data[i] & 0x0000ffff); + + if (!data->is_memory_GDDR5) { + table->mc_reg_table_entry[k].mc_data[j] |= 0x100; + } + } + j++; + PP_ASSERT_WITH_CODE((j <= SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE), + "Invalid VramInfo table.", return -1); + + if (!data->is_memory_GDDR5) { + table->mc_reg_address[j].s1 = mmMC_PMG_AUTO_CMD; + table->mc_reg_address[j].s0 = mmMC_PMG_AUTO_CMD; + for (k = 0; k < table->num_entries; k++) { + table->mc_reg_table_entry[k].mc_data[j] = + (table->mc_reg_table_entry[k].mc_data[i] & 0xffff0000) >> 16; + } + j++; + PP_ASSERT_WITH_CODE((j <= SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE), + "Invalid VramInfo table.", return -1); + } + + break; + + case mmMC_SEQ_RESERVE_M: + temp_reg = cgs_read_register(hwmgr->device, mmMC_PMG_CMD_MRS1); + table->mc_reg_address[j].s1 = mmMC_PMG_CMD_MRS1; + table->mc_reg_address[j].s0 = mmMC_SEQ_PMG_CMD_MRS1_LP; + for (k = 0; k < table->num_entries; k++) { + table->mc_reg_table_entry[k].mc_data[j] = + (temp_reg & 0xffff0000) | + (table->mc_reg_table_entry[k].mc_data[i] & 0x0000ffff); + } + j++; + PP_ASSERT_WITH_CODE((j <= SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE), + "Invalid VramInfo table.", return -1); + break; + + default: + break; + } + + } + + table->last = j; + + return 0; +} + + +static int iceland_set_valid_flag(phw_iceland_mc_reg_table *table) +{ + uint8_t i, j; + for (i = 0; i < table->last; i++) { + for (j = 1; j < table->num_entries; j++) { + if (table->mc_reg_table_entry[j-1].mc_data[i] != + table->mc_reg_table_entry[j].mc_data[i]) { + table->validflag |= (1<<i); + break; + } + } + } + + return 0; +} + +static int iceland_initialize_mc_reg_table(struct pp_hwmgr *hwmgr) +{ + int result; + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + pp_atomctrl_mc_reg_table *table; + phw_iceland_mc_reg_table *ni_table = &data->iceland_mc_reg_table; + uint8_t module_index = iceland_get_memory_module_index(hwmgr); + + table = kzalloc(sizeof(pp_atomctrl_mc_reg_table), GFP_KERNEL); + + if (NULL == table) + return -ENOMEM; + + /* Program additional LP registers that are no longer programmed by VBIOS */ + cgs_write_register(hwmgr->device, mmMC_SEQ_RAS_TIMING_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_RAS_TIMING)); + cgs_write_register(hwmgr->device, mmMC_SEQ_CAS_TIMING_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_CAS_TIMING)); + cgs_write_register(hwmgr->device, mmMC_SEQ_DLL_STBY_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_DLL_STBY)); + cgs_write_register(hwmgr->device, mmMC_SEQ_G5PDX_CMD0_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_G5PDX_CMD0)); + cgs_write_register(hwmgr->device, mmMC_SEQ_G5PDX_CMD1_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_G5PDX_CMD1)); + cgs_write_register(hwmgr->device, mmMC_SEQ_G5PDX_CTRL_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_G5PDX_CTRL)); + cgs_write_register(hwmgr->device, mmMC_SEQ_PMG_DVS_CMD_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_PMG_DVS_CMD)); + cgs_write_register(hwmgr->device, mmMC_SEQ_PMG_DVS_CTL_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_PMG_DVS_CTL)); + cgs_write_register(hwmgr->device, mmMC_SEQ_MISC_TIMING_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_MISC_TIMING)); + cgs_write_register(hwmgr->device, mmMC_SEQ_MISC_TIMING2_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_MISC_TIMING2)); + cgs_write_register(hwmgr->device, mmMC_SEQ_PMG_CMD_EMRS_LP, cgs_read_register(hwmgr->device, mmMC_PMG_CMD_EMRS)); + cgs_write_register(hwmgr->device, mmMC_SEQ_PMG_CMD_MRS_LP, cgs_read_register(hwmgr->device, mmMC_PMG_CMD_MRS)); + cgs_write_register(hwmgr->device, mmMC_SEQ_PMG_CMD_MRS1_LP, cgs_read_register(hwmgr->device, mmMC_PMG_CMD_MRS1)); + cgs_write_register(hwmgr->device, mmMC_SEQ_WR_CTL_D0_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_WR_CTL_D0)); + cgs_write_register(hwmgr->device, mmMC_SEQ_WR_CTL_D1_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_WR_CTL_D1)); + cgs_write_register(hwmgr->device, mmMC_SEQ_RD_CTL_D0_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_RD_CTL_D0)); + cgs_write_register(hwmgr->device, mmMC_SEQ_RD_CTL_D1_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_RD_CTL_D1)); + cgs_write_register(hwmgr->device, mmMC_SEQ_PMG_TIMING_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_PMG_TIMING)); + cgs_write_register(hwmgr->device, mmMC_SEQ_PMG_CMD_MRS2_LP, cgs_read_register(hwmgr->device, mmMC_PMG_CMD_MRS2)); + cgs_write_register(hwmgr->device, mmMC_SEQ_WR_CTL_2_LP, cgs_read_register(hwmgr->device, mmMC_SEQ_WR_CTL_2)); + + memset(table, 0x00, sizeof(pp_atomctrl_mc_reg_table)); + + result = atomctrl_initialize_mc_reg_table(hwmgr, module_index, table); + + if (0 == result) + result = iceland_copy_vbios_smc_reg_table(table, ni_table); + + if (0 == result) { + iceland_set_s0_mc_reg_index(ni_table); + result = iceland_set_mc_special_registers(hwmgr, ni_table); + } + + if (0 == result) + iceland_set_valid_flag(ni_table); + + kfree(table); + return result; +} + +/** + * Programs static screed detection parameters + * + * @param hwmgr the address of the powerplay hardware manager. + * @return always 0 + */ +int iceland_program_static_screen_threshold_parameters(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + /* Set static screen threshold unit*/ + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, + CGS_IND_REG__SMC, CG_STATIC_SCREEN_PARAMETER, STATIC_SCREEN_THRESHOLD_UNIT, + data->static_screen_threshold_unit); + /* Set static screen threshold*/ + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, + CGS_IND_REG__SMC, CG_STATIC_SCREEN_PARAMETER, STATIC_SCREEN_THRESHOLD, + data->static_screen_threshold); + + return 0; +} + +/** + * Setup display gap for glitch free memory clock switching. + * + * @param hwmgr the address of the powerplay hardware manager. + * @return always 0 + */ +int iceland_enable_display_gap(struct pp_hwmgr *hwmgr) +{ + uint32_t display_gap = cgs_read_ind_register(hwmgr->device, + CGS_IND_REG__SMC, ixCG_DISPLAY_GAP_CNTL); + + display_gap = PHM_SET_FIELD(display_gap, + CG_DISPLAY_GAP_CNTL, DISP_GAP, DISPLAY_GAP_IGNORE); + + display_gap = PHM_SET_FIELD(display_gap, + CG_DISPLAY_GAP_CNTL, DISP_GAP_MCHG, DISPLAY_GAP_VBLANK); + + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCG_DISPLAY_GAP_CNTL, display_gap); + + return 0; +} + +/** + * Programs activity state transition voting clients + * + * @param hwmgr the address of the powerplay hardware manager. + * @return always 0 + */ +int iceland_program_voting_clients(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + /* Clear reset for voting clients before enabling DPM */ + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + SCLK_PWRMGT_CNTL, RESET_SCLK_CNT, 0); + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + SCLK_PWRMGT_CNTL, RESET_BUSY_CNT, 0); + + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCG_FREQ_TRAN_VOTING_0, data->voting_rights_clients0); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCG_FREQ_TRAN_VOTING_1, data->voting_rights_clients1); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCG_FREQ_TRAN_VOTING_2, data->voting_rights_clients2); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCG_FREQ_TRAN_VOTING_3, data->voting_rights_clients3); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCG_FREQ_TRAN_VOTING_4, data->voting_rights_clients4); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCG_FREQ_TRAN_VOTING_5, data->voting_rights_clients5); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCG_FREQ_TRAN_VOTING_6, data->voting_rights_clients6); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCG_FREQ_TRAN_VOTING_7, data->voting_rights_clients7); + + return 0; +} + +static int iceland_upload_firmware(struct pp_hwmgr *hwmgr) +{ + int ret = 0; + + if (!iceland_is_smc_ram_running(hwmgr->smumgr)) + ret = iceland_smu_upload_firmware_image(hwmgr->smumgr); + + return ret; +} + +/** + * Get the location of various tables inside the FW image. + * + * @param hwmgr the address of the powerplay hardware manager. + * @return always 0 + */ +int iceland_process_firmware_header(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + uint32_t tmp; + int result; + bool error = 0; + + result = iceland_read_smc_sram_dword(hwmgr->smumgr, + SMU71_FIRMWARE_HEADER_LOCATION + + offsetof(SMU71_Firmware_Header, DpmTable), + &tmp, data->sram_end); + + if (0 == result) { + data->dpm_table_start = tmp; + } + + error |= (0 != result); + + result = iceland_read_smc_sram_dword(hwmgr->smumgr, + SMU71_FIRMWARE_HEADER_LOCATION + + offsetof(SMU71_Firmware_Header, SoftRegisters), + &tmp, data->sram_end); + + if (0 == result) { + data->soft_regs_start = tmp; + } + + error |= (0 != result); + + + result = iceland_read_smc_sram_dword(hwmgr->smumgr, + SMU71_FIRMWARE_HEADER_LOCATION + + offsetof(SMU71_Firmware_Header, mcRegisterTable), + &tmp, data->sram_end); + + if (0 == result) { + data->mc_reg_table_start = tmp; + } + + result = iceland_read_smc_sram_dword(hwmgr->smumgr, + SMU71_FIRMWARE_HEADER_LOCATION + + offsetof(SMU71_Firmware_Header, FanTable), + &tmp, data->sram_end); + + if (0 == result) { + data->fan_table_start = tmp; + } + + error |= (0 != result); + + result = iceland_read_smc_sram_dword(hwmgr->smumgr, + SMU71_FIRMWARE_HEADER_LOCATION + + offsetof(SMU71_Firmware_Header, mcArbDramTimingTable), + &tmp, data->sram_end); + + if (0 == result) { + data->arb_table_start = tmp; + } + + error |= (0 != result); + + + result = iceland_read_smc_sram_dword(hwmgr->smumgr, + SMU71_FIRMWARE_HEADER_LOCATION + + offsetof(SMU71_Firmware_Header, Version), + &tmp, data->sram_end); + + if (0 == result) { + hwmgr->microcode_version_info.SMC = tmp; + } + + error |= (0 != result); + + result = iceland_read_smc_sram_dword(hwmgr->smumgr, + SMU71_FIRMWARE_HEADER_LOCATION + + offsetof(SMU71_Firmware_Header, UlvSettings), + &tmp, data->sram_end); + + if (0 == result) { + data->ulv_settings_start = tmp; + } + + error |= (0 != result); + + return error ? 1 : 0; +} + +/* +* Copy one arb setting to another and then switch the active set. +* arbFreqSrc and arbFreqDest is one of the MC_CG_ARB_FREQ_Fx constants. +*/ +int iceland_copy_and_switch_arb_sets(struct pp_hwmgr *hwmgr, + uint32_t arbFreqSrc, uint32_t arbFreqDest) +{ + uint32_t mc_arb_dram_timing; + uint32_t mc_arb_dram_timing2; + uint32_t burst_time; + uint32_t mc_cg_config; + + switch (arbFreqSrc) { + case MC_CG_ARB_FREQ_F0: + mc_arb_dram_timing = cgs_read_register(hwmgr->device, mmMC_ARB_DRAM_TIMING); + mc_arb_dram_timing2 = cgs_read_register(hwmgr->device, mmMC_ARB_DRAM_TIMING2); + burst_time = PHM_READ_FIELD(hwmgr->device, MC_ARB_BURST_TIME, STATE0); + break; + + case MC_CG_ARB_FREQ_F1: + mc_arb_dram_timing = cgs_read_register(hwmgr->device, mmMC_ARB_DRAM_TIMING_1); + mc_arb_dram_timing2 = cgs_read_register(hwmgr->device, mmMC_ARB_DRAM_TIMING2_1); + burst_time = PHM_READ_FIELD(hwmgr->device, MC_ARB_BURST_TIME, STATE1); + break; + + default: + return -1; + } + + switch (arbFreqDest) { + case MC_CG_ARB_FREQ_F0: + cgs_write_register(hwmgr->device, mmMC_ARB_DRAM_TIMING, mc_arb_dram_timing); + cgs_write_register(hwmgr->device, mmMC_ARB_DRAM_TIMING2, mc_arb_dram_timing2); + PHM_WRITE_FIELD(hwmgr->device, MC_ARB_BURST_TIME, STATE0, burst_time); + break; + + case MC_CG_ARB_FREQ_F1: + cgs_write_register(hwmgr->device, mmMC_ARB_DRAM_TIMING_1, mc_arb_dram_timing); + cgs_write_register(hwmgr->device, mmMC_ARB_DRAM_TIMING2_1, mc_arb_dram_timing2); + PHM_WRITE_FIELD(hwmgr->device, MC_ARB_BURST_TIME, STATE1, burst_time); + break; + + default: + return -1; + } + + mc_cg_config = cgs_read_register(hwmgr->device, mmMC_CG_CONFIG); + mc_cg_config |= 0x0000000F; + cgs_write_register(hwmgr->device, mmMC_CG_CONFIG, mc_cg_config); + PHM_WRITE_FIELD(hwmgr->device, MC_ARB_CG, CG_ARB_REQ, arbFreqDest); + + return 0; +} + +/** + * Initial switch from ARB F0->F1 + * + * @param hwmgr the address of the powerplay hardware manager. + * @return always 0 + * This function is to be called from the SetPowerState table. + */ +int iceland_initial_switch_from_arb_f0_to_f1(struct pp_hwmgr *hwmgr) +{ + return iceland_copy_and_switch_arb_sets(hwmgr, MC_CG_ARB_FREQ_F0, MC_CG_ARB_FREQ_F1); +} + +/* ---------------------------------------- ULV related functions ----------------------------------------------------*/ + + +static int iceland_reset_single_dpm_table( + struct pp_hwmgr *hwmgr, + struct iceland_single_dpm_table *dpm_table, + uint32_t count) +{ + uint32_t i; + if (!(count <= MAX_REGULAR_DPM_NUMBER)) + printk(KERN_ERR "[ powerplay ] Fatal error, can not set up single DPM \ + table entries to exceed max number! \n"); + + dpm_table->count = count; + for (i = 0; i < MAX_REGULAR_DPM_NUMBER; i++) { + dpm_table->dpm_levels[i].enabled = 0; + } + + return 0; +} + +static void iceland_setup_pcie_table_entry( + struct iceland_single_dpm_table *dpm_table, + uint32_t index, uint32_t pcie_gen, + uint32_t pcie_lanes) +{ + dpm_table->dpm_levels[index].value = pcie_gen; + dpm_table->dpm_levels[index].param1 = pcie_lanes; + dpm_table->dpm_levels[index].enabled = 1; +} + +/* + * Set up the PCIe DPM table as follows: + * + * A = Performance State, Max, Gen Speed + * C = Performance State, Min, Gen Speed + * 1 = Performance State, Max, Lane # + * 3 = Performance State, Min, Lane # + * + * B = Power Saving State, Max, Gen Speed + * D = Power Saving State, Min, Gen Speed + * 2 = Power Saving State, Max, Lane # + * 4 = Power Saving State, Min, Lane # + * + * + * DPM Index Gen Speed Lane # + * 5 A 1 + * 4 B 2 + * 3 C 1 + * 2 D 2 + * 1 C 3 + * 0 D 4 + * + */ +static int iceland_setup_default_pcie_tables(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + PP_ASSERT_WITH_CODE((data->use_pcie_performance_levels || + data->use_pcie_power_saving_levels), + "No pcie performance levels!", return -EINVAL); + + if (data->use_pcie_performance_levels && !data->use_pcie_power_saving_levels) { + data->pcie_gen_power_saving = data->pcie_gen_performance; + data->pcie_lane_power_saving = data->pcie_lane_performance; + } else if (!data->use_pcie_performance_levels && data->use_pcie_power_saving_levels) { + data->pcie_gen_performance = data->pcie_gen_power_saving; + data->pcie_lane_performance = data->pcie_lane_power_saving; + } + + iceland_reset_single_dpm_table(hwmgr, &data->dpm_table.pcie_speed_table, SMU71_MAX_LEVELS_LINK); + + /* Hardcode Pcie Table */ + iceland_setup_pcie_table_entry(&data->dpm_table.pcie_speed_table, 0, + get_pcie_gen_support(data->pcie_gen_cap, PP_Min_PCIEGen), + get_pcie_lane_support(data->pcie_lane_cap, PP_Max_PCIELane)); + iceland_setup_pcie_table_entry(&data->dpm_table.pcie_speed_table, 1, + get_pcie_gen_support(data->pcie_gen_cap, PP_Min_PCIEGen), + get_pcie_lane_support(data->pcie_lane_cap, PP_Max_PCIELane)); + iceland_setup_pcie_table_entry(&data->dpm_table.pcie_speed_table, 2, + get_pcie_gen_support(data->pcie_gen_cap, PP_Max_PCIEGen), + get_pcie_lane_support(data->pcie_lane_cap, PP_Max_PCIELane)); + iceland_setup_pcie_table_entry(&data->dpm_table.pcie_speed_table, 3, + get_pcie_gen_support(data->pcie_gen_cap, PP_Max_PCIEGen), + get_pcie_lane_support(data->pcie_lane_cap, PP_Max_PCIELane)); + iceland_setup_pcie_table_entry(&data->dpm_table.pcie_speed_table, 4, + get_pcie_gen_support(data->pcie_gen_cap, PP_Max_PCIEGen), + get_pcie_lane_support(data->pcie_lane_cap, PP_Max_PCIELane)); + iceland_setup_pcie_table_entry(&data->dpm_table.pcie_speed_table, 5, + get_pcie_gen_support(data->pcie_gen_cap, PP_Max_PCIEGen), + get_pcie_lane_support(data->pcie_lane_cap, PP_Max_PCIELane)); + data->dpm_table.pcie_speed_table.count = 6; + + return 0; + +} + + +/* + * This function is to initalize all DPM state tables for SMU7 based on the dependency table. + * Dynamic state patching function will then trim these state tables to the allowed range based + * on the power policy or external client requests, such as UVD request, etc. + */ +static int iceland_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + uint32_t i; + + struct phm_clock_voltage_dependency_table *allowed_vdd_sclk_table = + hwmgr->dyn_state.vddc_dependency_on_sclk; + struct phm_clock_voltage_dependency_table *allowed_vdd_mclk_table = + hwmgr->dyn_state.vddc_dependency_on_mclk; + struct phm_cac_leakage_table *std_voltage_table = + hwmgr->dyn_state.cac_leakage_table; + + PP_ASSERT_WITH_CODE(allowed_vdd_sclk_table != NULL, + "SCLK dependency table is missing. This table is mandatory", return -1); + PP_ASSERT_WITH_CODE(allowed_vdd_sclk_table->count >= 1, + "SCLK dependency table has to have is missing. This table is mandatory", return -1); + + PP_ASSERT_WITH_CODE(allowed_vdd_mclk_table != NULL, + "MCLK dependency table is missing. This table is mandatory", return -1); + PP_ASSERT_WITH_CODE(allowed_vdd_mclk_table->count >= 1, + "VMCLK dependency table has to have is missing. This table is mandatory", return -1); + + /* clear the state table to reset everything to default */ + memset(&(data->dpm_table), 0x00, sizeof(data->dpm_table)); + iceland_reset_single_dpm_table(hwmgr, &data->dpm_table.sclk_table, SMU71_MAX_LEVELS_GRAPHICS); + iceland_reset_single_dpm_table(hwmgr, &data->dpm_table.mclk_table, SMU71_MAX_LEVELS_MEMORY); + iceland_reset_single_dpm_table(hwmgr, &data->dpm_table.vddc_table, SMU71_MAX_LEVELS_VDDC); + iceland_reset_single_dpm_table(hwmgr, &data->dpm_table.vdd_ci_table, SMU71_MAX_LEVELS_VDDCI); + iceland_reset_single_dpm_table(hwmgr, &data->dpm_table.mvdd_table, SMU71_MAX_LEVELS_MVDD); + + PP_ASSERT_WITH_CODE(allowed_vdd_sclk_table != NULL, + "SCLK dependency table is missing. This table is mandatory", return -1); + /* Initialize Sclk DPM table based on allow Sclk values*/ + data->dpm_table.sclk_table.count = 0; + + for (i = 0; i < allowed_vdd_sclk_table->count; i++) { + if (i == 0 || data->dpm_table.sclk_table.dpm_levels[data->dpm_table.sclk_table.count-1].value != + allowed_vdd_sclk_table->entries[i].clk) { + data->dpm_table.sclk_table.dpm_levels[data->dpm_table.sclk_table.count].value = + allowed_vdd_sclk_table->entries[i].clk; + data->dpm_table.sclk_table.dpm_levels[data->dpm_table.sclk_table.count].enabled = 1; /*(i==0) ? 1 : 0; to do */ + data->dpm_table.sclk_table.count++; + } + } + + PP_ASSERT_WITH_CODE(allowed_vdd_mclk_table != NULL, + "MCLK dependency table is missing. This table is mandatory", return -1); + /* Initialize Mclk DPM table based on allow Mclk values */ + data->dpm_table.mclk_table.count = 0; + for (i = 0; i < allowed_vdd_mclk_table->count; i++) { + if (i == 0 || data->dpm_table.mclk_table.dpm_levels[data->dpm_table.mclk_table.count-1].value != + allowed_vdd_mclk_table->entries[i].clk) { + data->dpm_table.mclk_table.dpm_levels[data->dpm_table.mclk_table.count].value = + allowed_vdd_mclk_table->entries[i].clk; + data->dpm_table.mclk_table.dpm_levels[data->dpm_table.mclk_table.count].enabled = 1; /*(i==0) ? 1 : 0; */ + data->dpm_table.mclk_table.count++; + } + } + + /* Initialize Vddc DPM table based on allow Vddc values. And populate corresponding std values. */ + for (i = 0; i < allowed_vdd_sclk_table->count; i++) { + data->dpm_table.vddc_table.dpm_levels[i].value = allowed_vdd_mclk_table->entries[i].v; + data->dpm_table.vddc_table.dpm_levels[i].param1 = std_voltage_table->entries[i].Leakage; + /* param1 is for corresponding std voltage */ + data->dpm_table.vddc_table.dpm_levels[i].enabled = 1; + } + + data->dpm_table.vddc_table.count = allowed_vdd_sclk_table->count; + allowed_vdd_mclk_table = hwmgr->dyn_state.vddci_dependency_on_mclk; + + if (NULL != allowed_vdd_mclk_table) { + /* Initialize Vddci DPM table based on allow Mclk values */ + for (i = 0; i < allowed_vdd_mclk_table->count; i++) { + data->dpm_table.vdd_ci_table.dpm_levels[i].value = allowed_vdd_mclk_table->entries[i].v; + data->dpm_table.vdd_ci_table.dpm_levels[i].enabled = 1; + } + data->dpm_table.vdd_ci_table.count = allowed_vdd_mclk_table->count; + } + + allowed_vdd_mclk_table = hwmgr->dyn_state.mvdd_dependency_on_mclk; + + if (NULL != allowed_vdd_mclk_table) { + /* + * Initialize MVDD DPM table based on allow Mclk + * values + */ + for (i = 0; i < allowed_vdd_mclk_table->count; i++) { + data->dpm_table.mvdd_table.dpm_levels[i].value = allowed_vdd_mclk_table->entries[i].v; + data->dpm_table.mvdd_table.dpm_levels[i].enabled = 1; + } + data->dpm_table.mvdd_table.count = allowed_vdd_mclk_table->count; + } + + /* setup PCIE gen speed levels*/ + iceland_setup_default_pcie_tables(hwmgr); + + /* save a copy of the default DPM table*/ + memcpy(&(data->golden_dpm_table), &(data->dpm_table), sizeof(struct iceland_dpm_table)); + + return 0; +} + +/** + * @brief PhwIceland_GetVoltageOrder + * Returns index of requested voltage record in lookup(table) + * @param hwmgr - pointer to hardware manager + * @param lookutab - lookup list to search in + * @param voltage - voltage to look for + * @return 0 on success + */ +uint8_t iceland_get_voltage_index(phm_ppt_v1_voltage_lookup_table *look_up_table, + uint16_t voltage) +{ + uint8_t count = (uint8_t) (look_up_table->count); + uint8_t i; + + PP_ASSERT_WITH_CODE((NULL != look_up_table), "Lookup Table empty.", return 0;); + PP_ASSERT_WITH_CODE((0 != count), "Lookup Table empty.", return 0;); + + for (i = 0; i < count; i++) { + /* find first voltage equal or bigger than requested */ + if (look_up_table->entries[i].us_vdd >= voltage) + return i; + } + + /* voltage is bigger than max voltage in the table */ + return i-1; +} + + +static int iceland_get_std_voltage_value_sidd(struct pp_hwmgr *hwmgr, + pp_atomctrl_voltage_table_entry *tab, uint16_t *hi, + uint16_t *lo) +{ + uint16_t v_index; + bool vol_found = false; + *hi = tab->value * VOLTAGE_SCALE; + *lo = tab->value * VOLTAGE_SCALE; + + /* SCLK/VDDC Dependency Table has to exist. */ + PP_ASSERT_WITH_CODE(NULL != hwmgr->dyn_state.vddc_dependency_on_sclk, + "The SCLK/VDDC Dependency Table does not exist.\n", + return -EINVAL); + + if (NULL == hwmgr->dyn_state.cac_leakage_table) { + pr_warning("CAC Leakage Table does not exist, using vddc.\n"); + return 0; + } + + /* + * Since voltage in the sclk/vddc dependency table is not + * necessarily in ascending order because of ELB voltage + * patching, loop through entire list to find exact voltage. + */ + for (v_index = 0; (uint32_t)v_index < hwmgr->dyn_state.vddc_dependency_on_sclk->count; v_index++) { + if (tab->value == hwmgr->dyn_state.vddc_dependency_on_sclk->entries[v_index].v) { + vol_found = true; + if ((uint32_t)v_index < hwmgr->dyn_state.cac_leakage_table->count) { + *lo = hwmgr->dyn_state.cac_leakage_table->entries[v_index].Vddc * VOLTAGE_SCALE; + *hi = (uint16_t)(hwmgr->dyn_state.cac_leakage_table->entries[v_index].Leakage * VOLTAGE_SCALE); + } else { + pr_warning("Index from SCLK/VDDC Dependency Table exceeds the CAC Leakage Table index, using maximum index from CAC table.\n"); + *lo = hwmgr->dyn_state.cac_leakage_table->entries[hwmgr->dyn_state.cac_leakage_table->count - 1].Vddc * VOLTAGE_SCALE; + *hi = (uint16_t)(hwmgr->dyn_state.cac_leakage_table->entries[hwmgr->dyn_state.cac_leakage_table->count - 1].Leakage * VOLTAGE_SCALE); + } + break; + } + } + + /* + * If voltage is not found in the first pass, loop again to + * find the best match, equal or higher value. + */ + if (!vol_found) { + for (v_index = 0; (uint32_t)v_index < hwmgr->dyn_state.vddc_dependency_on_sclk->count; v_index++) { + if (tab->value <= hwmgr->dyn_state.vddc_dependency_on_sclk->entries[v_index].v) { + vol_found = true; + if ((uint32_t)v_index < hwmgr->dyn_state.cac_leakage_table->count) { + *lo = hwmgr->dyn_state.cac_leakage_table->entries[v_index].Vddc * VOLTAGE_SCALE; + *hi = (uint16_t)(hwmgr->dyn_state.cac_leakage_table->entries[v_index].Leakage) * VOLTAGE_SCALE; + } else { + pr_warning("Index from SCLK/VDDC Dependency Table exceeds the CAC Leakage Table index in second look up, using maximum index from CAC table."); + *lo = hwmgr->dyn_state.cac_leakage_table->entries[hwmgr->dyn_state.cac_leakage_table->count - 1].Vddc * VOLTAGE_SCALE; + *hi = (uint16_t)(hwmgr->dyn_state.cac_leakage_table->entries[hwmgr->dyn_state.cac_leakage_table->count - 1].Leakage * VOLTAGE_SCALE); + } + break; + } + } + + if (!vol_found) + pr_warning("Unable to get std_vddc from SCLK/VDDC Dependency Table, using vddc.\n"); + } + + return 0; +} + +static int iceland_populate_smc_voltage_table(struct pp_hwmgr *hwmgr, + pp_atomctrl_voltage_table_entry *tab, + SMU71_Discrete_VoltageLevel *smc_voltage_tab) { + int result; + + + result = iceland_get_std_voltage_value_sidd(hwmgr, tab, + &smc_voltage_tab->StdVoltageHiSidd, + &smc_voltage_tab->StdVoltageLoSidd); + if (0 != result) { + smc_voltage_tab->StdVoltageHiSidd = tab->value * VOLTAGE_SCALE; + smc_voltage_tab->StdVoltageLoSidd = tab->value * VOLTAGE_SCALE; + } + + smc_voltage_tab->Voltage = PP_HOST_TO_SMC_US(tab->value * VOLTAGE_SCALE); + CONVERT_FROM_HOST_TO_SMC_US(smc_voltage_tab->StdVoltageHiSidd); + CONVERT_FROM_HOST_TO_SMC_US(smc_voltage_tab->StdVoltageHiSidd); + + return 0; +} + +/** + * Vddc table preparation for SMC. + * + * @param hwmgr the address of the hardware manager + * @param table the SMC DPM table structure to be populated + * @return always 0 + */ +static int iceland_populate_smc_vddc_table(struct pp_hwmgr *hwmgr, + SMU71_Discrete_DpmTable *table) +{ + unsigned int count; + int result; + + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + table->VddcLevelCount = data->vddc_voltage_table.count; + for (count = 0; count < table->VddcLevelCount; count++) { + result = iceland_populate_smc_voltage_table(hwmgr, + &data->vddc_voltage_table.entries[count], + &table->VddcLevel[count]); + PP_ASSERT_WITH_CODE(0 == result, "do not populate SMC VDDC voltage table", return -EINVAL); + + /* GPIO voltage control */ + if (ICELAND_VOLTAGE_CONTROL_BY_GPIO == data->voltage_control) + table->VddcLevel[count].Smio |= data->vddc_voltage_table.entries[count].smio_low; + else if (ICELAND_VOLTAGE_CONTROL_BY_SVID2 == data->voltage_control) + table->VddcLevel[count].Smio = 0; + } + + CONVERT_FROM_HOST_TO_SMC_UL(table->VddcLevelCount); + + return 0; +} + +/** + * Vddci table preparation for SMC. + * + * @param *hwmgr The address of the hardware manager. + * @param *table The SMC DPM table structure to be populated. + * @return 0 + */ +static int iceland_populate_smc_vdd_ci_table(struct pp_hwmgr *hwmgr, + SMU71_Discrete_DpmTable *table) +{ + int result; + uint32_t count; + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + table->VddciLevelCount = data->vddci_voltage_table.count; + for (count = 0; count < table->VddciLevelCount; count++) { + result = iceland_populate_smc_voltage_table(hwmgr, + &data->vddci_voltage_table.entries[count], + &table->VddciLevel[count]); + PP_ASSERT_WITH_CODE(0 == result, "do not populate SMC VDDCI voltage table", return -EINVAL); + + /* GPIO voltage control */ + if (ICELAND_VOLTAGE_CONTROL_BY_GPIO == data->vdd_ci_control) + table->VddciLevel[count].Smio |= data->vddci_voltage_table.entries[count].smio_low; + else + table->VddciLevel[count].Smio = 0; + } + + CONVERT_FROM_HOST_TO_SMC_UL(table->VddcLevelCount); + + return 0; +} + +/** + * Mvdd table preparation for SMC. + * + * @param *hwmgr The address of the hardware manager. + * @param *table The SMC DPM table structure to be populated. + * @return 0 + */ +static int iceland_populate_smc_mvdd_table(struct pp_hwmgr *hwmgr, + SMU71_Discrete_DpmTable *table) +{ + int result; + uint32_t count; + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + table->MvddLevelCount = data->mvdd_voltage_table.count; + for (count = 0; count < table->MvddLevelCount; count++) { + result = iceland_populate_smc_voltage_table(hwmgr, + &data->mvdd_voltage_table.entries[count], + &table->MvddLevel[count]); + PP_ASSERT_WITH_CODE(0 == result, "do not populate SMC VDDCI voltage table", return -EINVAL); + + /* GPIO voltage control */ + if (ICELAND_VOLTAGE_CONTROL_BY_GPIO == data->mvdd_control) + table->MvddLevel[count].Smio |= data->mvdd_voltage_table.entries[count].smio_low; + else + table->MvddLevel[count].Smio = 0; + } + + CONVERT_FROM_HOST_TO_SMC_UL(table->MvddLevelCount); + + return 0; +} + +/** + * Convert a voltage value in mv unit to VID number required by SMU firmware + */ +static uint8_t convert_to_vid(uint16_t vddc) +{ + return (uint8_t) ((6200 - (vddc * VOLTAGE_SCALE)) / 25); +} + +int iceland_populate_bapm_vddc_vid_sidd(struct pp_hwmgr *hwmgr) +{ + int i; + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + uint8_t * hi_vid = data->power_tune_table.BapmVddCVidHiSidd; + uint8_t * lo_vid = data->power_tune_table.BapmVddCVidLoSidd; + + PP_ASSERT_WITH_CODE(NULL != hwmgr->dyn_state.cac_leakage_table, + "The CAC Leakage table does not exist!", return -EINVAL); + PP_ASSERT_WITH_CODE(hwmgr->dyn_state.cac_leakage_table->count <= 8, + "There should never be more than 8 entries for BapmVddcVid!!!", return -EINVAL); + PP_ASSERT_WITH_CODE(hwmgr->dyn_state.cac_leakage_table->count == hwmgr->dyn_state.vddc_dependency_on_sclk->count, + "CACLeakageTable->count and VddcDependencyOnSCLk->count not equal", return -EINVAL); + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_EVV)) { + for (i = 0; (uint32_t) i < hwmgr->dyn_state.cac_leakage_table->count; i++) { + lo_vid[i] = convert_to_vid(hwmgr->dyn_state.cac_leakage_table->entries[i].Vddc1); + hi_vid[i] = convert_to_vid(hwmgr->dyn_state.cac_leakage_table->entries[i].Vddc2); + } + } else { + PP_ASSERT_WITH_CODE(false, "Iceland should always support EVV", return -EINVAL); + } + + return 0; +} + +int iceland_populate_vddc_vid(struct pp_hwmgr *hwmgr) +{ + int i; + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + uint8_t *vid = data->power_tune_table.VddCVid; + + PP_ASSERT_WITH_CODE(data->vddc_voltage_table.count <= 8, + "There should never be more than 8 entries for VddcVid!!!", + return -EINVAL); + + for (i = 0; i < (int)data->vddc_voltage_table.count; i++) { + vid[i] = convert_to_vid(data->vddc_voltage_table.entries[i].value); + } + + return 0; +} + +/** + * Preparation of voltage tables for SMC. + * + * @param hwmgr the address of the hardware manager + * @param table the SMC DPM table structure to be populated + * @return always 0 + */ + +int iceland_populate_smc_voltage_tables(struct pp_hwmgr *hwmgr, + SMU71_Discrete_DpmTable *table) +{ + int result; + + result = iceland_populate_smc_vddc_table(hwmgr, table); + PP_ASSERT_WITH_CODE(0 == result, + "can not populate VDDC voltage table to SMC", return -1); + + result = iceland_populate_smc_vdd_ci_table(hwmgr, table); + PP_ASSERT_WITH_CODE(0 == result, + "can not populate VDDCI voltage table to SMC", return -1); + + result = iceland_populate_smc_mvdd_table(hwmgr, table); + PP_ASSERT_WITH_CODE(0 == result, + "can not populate MVDD voltage table to SMC", return -1); + + return 0; +} + + +/** + * Re-generate the DPM level mask value + * @param hwmgr the address of the hardware manager + */ +static uint32_t iceland_get_dpm_level_enable_mask_value( + struct iceland_single_dpm_table * dpm_table) +{ + uint32_t i; + uint32_t mask_value = 0; + + for (i = dpm_table->count; i > 0; i--) { + mask_value = mask_value << 1; + + if (dpm_table->dpm_levels[i-1].enabled) + mask_value |= 0x1; + else + mask_value &= 0xFFFFFFFE; + } + return mask_value; +} + +int iceland_populate_memory_timing_parameters( + struct pp_hwmgr *hwmgr, + uint32_t engine_clock, + uint32_t memory_clock, + struct SMU71_Discrete_MCArbDramTimingTableEntry *arb_regs + ) +{ + uint32_t dramTiming; + uint32_t dramTiming2; + uint32_t burstTime; + int result; + + result = atomctrl_set_engine_dram_timings_rv770(hwmgr, + engine_clock, memory_clock); + + PP_ASSERT_WITH_CODE(result == 0, + "Error calling VBIOS to set DRAM_TIMING.", return result); + + dramTiming = cgs_read_register(hwmgr->device, mmMC_ARB_DRAM_TIMING); + dramTiming2 = cgs_read_register(hwmgr->device, mmMC_ARB_DRAM_TIMING2); + burstTime = PHM_READ_FIELD(hwmgr->device, MC_ARB_BURST_TIME, STATE0); + + arb_regs->McArbDramTiming = PP_HOST_TO_SMC_UL(dramTiming); + arb_regs->McArbDramTiming2 = PP_HOST_TO_SMC_UL(dramTiming2); + arb_regs->McArbBurstTime = (uint8_t)burstTime; + + return 0; +} + +/** + * Setup parameters for the MC ARB. + * + * @param hwmgr the address of the powerplay hardware manager. + * @return always 0 + * This function is to be called from the SetPowerState table. + */ +int iceland_program_memory_timing_parameters(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + int result = 0; + SMU71_Discrete_MCArbDramTimingTable arb_regs; + uint32_t i, j; + + memset(&arb_regs, 0x00, sizeof(SMU71_Discrete_MCArbDramTimingTable)); + + for (i = 0; i < data->dpm_table.sclk_table.count; i++) { + for (j = 0; j < data->dpm_table.mclk_table.count; j++) { + result = iceland_populate_memory_timing_parameters + (hwmgr, data->dpm_table.sclk_table.dpm_levels[i].value, + data->dpm_table.mclk_table.dpm_levels[j].value, + &arb_regs.entries[i][j]); + + if (0 != result) { + break; + } + } + } + + if (0 == result) { + result = iceland_copy_bytes_to_smc( + hwmgr->smumgr, + data->arb_table_start, + (uint8_t *)&arb_regs, + sizeof(SMU71_Discrete_MCArbDramTimingTable), + data->sram_end + ); + } + + return result; +} + +static int iceland_populate_smc_link_level(struct pp_hwmgr *hwmgr, SMU71_Discrete_DpmTable *table) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + struct iceland_dpm_table *dpm_table = &data->dpm_table; + uint32_t i; + + /* Index (dpm_table->pcie_speed_table.count) is reserved for PCIE boot level. */ + for (i = 0; i <= dpm_table->pcie_speed_table.count; i++) { + table->LinkLevel[i].PcieGenSpeed = + (uint8_t)dpm_table->pcie_speed_table.dpm_levels[i].value; + table->LinkLevel[i].PcieLaneCount = + (uint8_t)encode_pcie_lane_width(dpm_table->pcie_speed_table.dpm_levels[i].param1); + table->LinkLevel[i].EnabledForActivity = + 1; + table->LinkLevel[i].SPC = + (uint8_t)(data->pcie_spc_cap & 0xff); + table->LinkLevel[i].DownThreshold = + PP_HOST_TO_SMC_UL(5); + table->LinkLevel[i].UpThreshold = + PP_HOST_TO_SMC_UL(30); + } + + data->smc_state_table.LinkLevelCount = + (uint8_t)dpm_table->pcie_speed_table.count; + data->dpm_level_enable_mask.pcie_dpm_enable_mask = + iceland_get_dpm_level_enable_mask_value(&dpm_table->pcie_speed_table); + + return 0; +} + +static int iceland_populate_smc_uvd_level(struct pp_hwmgr *hwmgr, + SMU71_Discrete_DpmTable *table) +{ + return 0; +} + +uint8_t iceland_get_voltage_id(pp_atomctrl_voltage_table *voltage_table, + uint32_t voltage) +{ + uint8_t count = (uint8_t) (voltage_table->count); + uint8_t i = 0; + + PP_ASSERT_WITH_CODE((NULL != voltage_table), + "Voltage Table empty.", return 0;); + PP_ASSERT_WITH_CODE((0 != count), + "Voltage Table empty.", return 0;); + + for (i = 0; i < count; i++) { + /* find first voltage bigger than requested */ + if (voltage_table->entries[i].value >= voltage) + return i; + } + + /* voltage is bigger than max voltage in the table */ + return i - 1; +} + +static int iceland_populate_smc_vce_level(struct pp_hwmgr *hwmgr, + SMU71_Discrete_DpmTable *table) +{ + return 0; +} + +static int iceland_populate_smc_acp_level(struct pp_hwmgr *hwmgr, + SMU71_Discrete_DpmTable *table) +{ + return 0; +} + +static int iceland_populate_smc_samu_level(struct pp_hwmgr *hwmgr, + SMU71_Discrete_DpmTable *table) +{ + return 0; +} + + +static int iceland_populate_smc_svi2_config(struct pp_hwmgr *hwmgr, + SMU71_Discrete_DpmTable *tab) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + if(ICELAND_VOLTAGE_CONTROL_BY_SVID2 == data->voltage_control) + tab->SVI2Enable |= VDDC_ON_SVI2; + + if(ICELAND_VOLTAGE_CONTROL_BY_SVID2 == data->vdd_ci_control) + tab->SVI2Enable |= VDDCI_ON_SVI2; + else + tab->MergedVddci = 1; + + if(ICELAND_VOLTAGE_CONTROL_BY_SVID2 == data->mvdd_control) + tab->SVI2Enable |= MVDD_ON_SVI2; + + PP_ASSERT_WITH_CODE( tab->SVI2Enable != (VDDC_ON_SVI2 | VDDCI_ON_SVI2 | MVDD_ON_SVI2) && + (tab->SVI2Enable & VDDC_ON_SVI2), "SVI2 domain configuration is incorrect!", return -EINVAL); + + return 0; +} + +static int iceland_get_dependecy_volt_by_clk(struct pp_hwmgr *hwmgr, + struct phm_clock_voltage_dependency_table *allowed_clock_voltage_table, + uint32_t clock, uint32_t *vol) +{ + uint32_t i = 0; + + /* clock - voltage dependency table is empty table */ + if (allowed_clock_voltage_table->count == 0) + return -EINVAL; + + for (i = 0; i < allowed_clock_voltage_table->count; i++) { + /* find first sclk bigger than request */ + if (allowed_clock_voltage_table->entries[i].clk >= clock) { + *vol = allowed_clock_voltage_table->entries[i].v; + return 0; + } + } + + /* sclk is bigger than max sclk in the dependence table */ + *vol = allowed_clock_voltage_table->entries[i - 1].v; + + return 0; +} + +static uint8_t iceland_get_mclk_frequency_ratio(uint32_t memory_clock, + bool strobe_mode) +{ + uint8_t mc_para_index; + + if (strobe_mode) { + if (memory_clock < 12500) { + mc_para_index = 0x00; + } else if (memory_clock > 47500) { + mc_para_index = 0x0f; + } else { + mc_para_index = (uint8_t)((memory_clock - 10000) / 2500); + } + } else { + if (memory_clock < 65000) { + mc_para_index = 0x00; + } else if (memory_clock > 135000) { + mc_para_index = 0x0f; + } else { + mc_para_index = (uint8_t)((memory_clock - 60000) / 5000); + } + } + + return mc_para_index; +} + +static uint8_t iceland_get_ddr3_mclk_frequency_ratio(uint32_t memory_clock) +{ + uint8_t mc_para_index; + + if (memory_clock < 10000) { + mc_para_index = 0; + } else if (memory_clock >= 80000) { + mc_para_index = 0x0f; + } else { + mc_para_index = (uint8_t)((memory_clock - 10000) / 5000 + 1); + } + + return mc_para_index; +} + +static int iceland_populate_phase_value_based_on_sclk(struct pp_hwmgr *hwmgr, const struct phm_phase_shedding_limits_table *pl, + uint32_t sclk, uint32_t *p_shed) +{ + unsigned int i; + + /* use the minimum phase shedding */ + *p_shed = 1; + + /* + * PPGen ensures the phase shedding limits table is sorted + * from lowest voltage/sclk/mclk to highest voltage/sclk/mclk. + * VBIOS ensures the phase shedding masks table is sorted from + * least phases enabled (phase shedding on) to most phases + * enabled (phase shedding off). + */ + for (i = 0; i < pl->count; i++) { + if (sclk < pl->entries[i].Sclk) { + /* Enable phase shedding */ + *p_shed = i; + break; + } + } + + return 0; +} + +static int iceland_populate_phase_value_based_on_mclk(struct pp_hwmgr *hwmgr, const struct phm_phase_shedding_limits_table *pl, + uint32_t memory_clock, uint32_t *p_shed) +{ + unsigned int i; + + /* use the minimum phase shedding */ + *p_shed = 1; + + /* + * PPGen ensures the phase shedding limits table is sorted + * from lowest voltage/sclk/mclk to highest voltage/sclk/mclk. + * VBIOS ensures the phase shedding masks table is sorted from + * least phases enabled (phase shedding on) to most phases + * enabled (phase shedding off). + */ + for (i = 0; i < pl->count; i++) { + if (memory_clock < pl->entries[i].Mclk) { + /* Enable phase shedding */ + *p_shed = i; + break; + } + } + + return 0; +} + +/** + * Populates the SMC MCLK structure using the provided memory clock + * + * @param hwmgr the address of the hardware manager + * @param memory_clock the memory clock to use to populate the structure + * @param sclk the SMC SCLK structure to be populated + */ +static int iceland_calculate_mclk_params( + struct pp_hwmgr *hwmgr, + uint32_t memory_clock, + SMU71_Discrete_MemoryLevel *mclk, + bool strobe_mode, + bool dllStateOn + ) +{ + const iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + uint32_t dll_cntl = data->clock_registers.vDLL_CNTL; + uint32_t mclk_pwrmgt_cntl = data->clock_registers.vMCLK_PWRMGT_CNTL; + uint32_t mpll_ad_func_cntl = data->clock_registers.vMPLL_AD_FUNC_CNTL; + uint32_t mpll_dq_func_cntl = data->clock_registers.vMPLL_DQ_FUNC_CNTL; + uint32_t mpll_func_cntl = data->clock_registers.vMPLL_FUNC_CNTL; + uint32_t mpll_func_cntl_1 = data->clock_registers.vMPLL_FUNC_CNTL_1; + uint32_t mpll_func_cntl_2 = data->clock_registers.vMPLL_FUNC_CNTL_2; + uint32_t mpll_ss1 = data->clock_registers.vMPLL_SS1; + uint32_t mpll_ss2 = data->clock_registers.vMPLL_SS2; + + pp_atomctrl_memory_clock_param mpll_param; + int result; + + result = atomctrl_get_memory_pll_dividers_si(hwmgr, + memory_clock, &mpll_param, strobe_mode); + PP_ASSERT_WITH_CODE(0 == result, + "Error retrieving Memory Clock Parameters from VBIOS.", return result); + + /* MPLL_FUNC_CNTL setup*/ + mpll_func_cntl = PHM_SET_FIELD(mpll_func_cntl, MPLL_FUNC_CNTL, BWCTRL, mpll_param.bw_ctrl); + + /* MPLL_FUNC_CNTL_1 setup*/ + mpll_func_cntl_1 = PHM_SET_FIELD(mpll_func_cntl_1, + MPLL_FUNC_CNTL_1, CLKF, mpll_param.mpll_fb_divider.cl_kf); + mpll_func_cntl_1 = PHM_SET_FIELD(mpll_func_cntl_1, + MPLL_FUNC_CNTL_1, CLKFRAC, mpll_param.mpll_fb_divider.clk_frac); + mpll_func_cntl_1 = PHM_SET_FIELD(mpll_func_cntl_1, + MPLL_FUNC_CNTL_1, VCO_MODE, mpll_param.vco_mode); + + /* MPLL_AD_FUNC_CNTL setup*/ + mpll_ad_func_cntl = PHM_SET_FIELD(mpll_ad_func_cntl, + MPLL_AD_FUNC_CNTL, YCLK_POST_DIV, mpll_param.mpll_post_divider); + + if (data->is_memory_GDDR5) { + /* MPLL_DQ_FUNC_CNTL setup*/ + mpll_dq_func_cntl = PHM_SET_FIELD(mpll_dq_func_cntl, + MPLL_DQ_FUNC_CNTL, YCLK_SEL, mpll_param.yclk_sel); + mpll_dq_func_cntl = PHM_SET_FIELD(mpll_dq_func_cntl, + MPLL_DQ_FUNC_CNTL, YCLK_POST_DIV, mpll_param.mpll_post_divider); + } + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_MemorySpreadSpectrumSupport)) { + /* + ************************************ + Fref = Reference Frequency + NF = Feedback divider ratio + NR = Reference divider ratio + Fnom = Nominal VCO output frequency = Fref * NF / NR + Fs = Spreading Rate + D = Percentage down-spread / 2 + Fint = Reference input frequency to PFD = Fref / NR + NS = Spreading rate divider ratio = int(Fint / (2 * Fs)) + CLKS = NS - 1 = ISS_STEP_NUM[11:0] + NV = D * Fs / Fnom * 4 * ((Fnom/Fref * NR) ^ 2) + CLKV = 65536 * NV = ISS_STEP_SIZE[25:0] + ************************************* + */ + pp_atomctrl_internal_ss_info ss_info; + uint32_t freq_nom; + uint32_t tmp; + uint32_t reference_clock = atomctrl_get_mpll_reference_clock(hwmgr); + + /* for GDDR5 for all modes and DDR3 */ + if (1 == mpll_param.qdr) + freq_nom = memory_clock * 4 * (1 << mpll_param.mpll_post_divider); + else + freq_nom = memory_clock * 2 * (1 << mpll_param.mpll_post_divider); + + /* tmp = (freq_nom / reference_clock * reference_divider) ^ 2 Note: S.I. reference_divider = 1*/ + tmp = (freq_nom / reference_clock); + tmp = tmp * tmp; + + if (0 == atomctrl_get_memory_clock_spread_spectrum(hwmgr, freq_nom, &ss_info)) { + /* ss_info.speed_spectrum_percentage -- in unit of 0.01% */ + /* ss.Info.speed_spectrum_rate -- in unit of khz */ + /* CLKS = reference_clock / (2 * speed_spectrum_rate * reference_divider) * 10 */ + /* = reference_clock * 5 / speed_spectrum_rate */ + uint32_t clks = reference_clock * 5 / ss_info.speed_spectrum_rate; + + /* CLKV = 65536 * speed_spectrum_percentage / 2 * spreadSpecrumRate / freq_nom * 4 / 100000 * ((freq_nom / reference_clock) ^ 2) */ + /* = 131 * speed_spectrum_percentage * speed_spectrum_rate / 100 * ((freq_nom / reference_clock) ^ 2) / freq_nom */ + uint32_t clkv = + (uint32_t)((((131 * ss_info.speed_spectrum_percentage * + ss_info.speed_spectrum_rate) / 100) * tmp) / freq_nom); + + mpll_ss1 = PHM_SET_FIELD(mpll_ss1, MPLL_SS1, CLKV, clkv); + mpll_ss2 = PHM_SET_FIELD(mpll_ss2, MPLL_SS2, CLKS, clks); + } + } + + /* MCLK_PWRMGT_CNTL setup */ + mclk_pwrmgt_cntl = PHM_SET_FIELD(mclk_pwrmgt_cntl, + MCLK_PWRMGT_CNTL, DLL_SPEED, mpll_param.dll_speed); + mclk_pwrmgt_cntl = PHM_SET_FIELD(mclk_pwrmgt_cntl, + MCLK_PWRMGT_CNTL, MRDCK0_PDNB, dllStateOn); + mclk_pwrmgt_cntl = PHM_SET_FIELD(mclk_pwrmgt_cntl, + MCLK_PWRMGT_CNTL, MRDCK1_PDNB, dllStateOn); + + + /* Save the result data to outpupt memory level structure */ + mclk->MclkFrequency = memory_clock; + mclk->MpllFuncCntl = mpll_func_cntl; + mclk->MpllFuncCntl_1 = mpll_func_cntl_1; + mclk->MpllFuncCntl_2 = mpll_func_cntl_2; + mclk->MpllAdFuncCntl = mpll_ad_func_cntl; + mclk->MpllDqFuncCntl = mpll_dq_func_cntl; + mclk->MclkPwrmgtCntl = mclk_pwrmgt_cntl; + mclk->DllCntl = dll_cntl; + mclk->MpllSs1 = mpll_ss1; + mclk->MpllSs2 = mpll_ss2; + + return 0; +} + +static int iceland_populate_single_memory_level( + struct pp_hwmgr *hwmgr, + uint32_t memory_clock, + SMU71_Discrete_MemoryLevel *memory_level + ) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + int result = 0; + bool dllStateOn; + struct cgs_display_info info = {0}; + + + if (NULL != hwmgr->dyn_state.vddc_dependency_on_mclk) { + result = iceland_get_dependecy_volt_by_clk(hwmgr, + hwmgr->dyn_state.vddc_dependency_on_mclk, memory_clock, &memory_level->MinVddc); + PP_ASSERT_WITH_CODE((0 == result), + "can not find MinVddc voltage value from memory VDDC voltage dependency table", return result); + } + + if (data->vdd_ci_control == ICELAND_VOLTAGE_CONTROL_NONE) { + memory_level->MinVddci = memory_level->MinVddc; + } else if (NULL != hwmgr->dyn_state.vddci_dependency_on_mclk) { + result = iceland_get_dependecy_volt_by_clk(hwmgr, + hwmgr->dyn_state.vddci_dependency_on_mclk, + memory_clock, + &memory_level->MinVddci); + PP_ASSERT_WITH_CODE((0 == result), + "can not find MinVddci voltage value from memory VDDCI voltage dependency table", return result); + } + + if (NULL != hwmgr->dyn_state.mvdd_dependency_on_mclk) { + result = iceland_get_dependecy_volt_by_clk(hwmgr, + hwmgr->dyn_state.mvdd_dependency_on_mclk, memory_clock, &memory_level->MinMvdd); + PP_ASSERT_WITH_CODE((0 == result), + "can not find MinMVDD voltage value from memory MVDD voltage dependency table", return result); + } + + memory_level->MinVddcPhases = 1; + + if (data->vddc_phase_shed_control) { + iceland_populate_phase_value_based_on_mclk(hwmgr, hwmgr->dyn_state.vddc_phase_shed_limits_table, + memory_clock, &memory_level->MinVddcPhases); + } + + memory_level->EnabledForThrottle = 1; + memory_level->EnabledForActivity = 1; + memory_level->UpHyst = 0; + memory_level->DownHyst = 100; + memory_level->VoltageDownHyst = 0; + + /* Indicates maximum activity level for this performance level.*/ + memory_level->ActivityLevel = (uint16_t)data->mclk_activity_target; + memory_level->StutterEnable = 0; + memory_level->StrobeEnable = 0; + memory_level->EdcReadEnable = 0; + memory_level->EdcWriteEnable = 0; + memory_level->RttEnable = 0; + + /* default set to low watermark. Highest level will be set to high later.*/ + memory_level->DisplayWatermark = PPSMC_DISPLAY_WATERMARK_LOW; + + cgs_get_active_displays_info(hwmgr->device, &info); + data->display_timing.num_existing_displays = info.display_count; + + //if ((data->mclk_stutter_mode_threshold != 0) && + // (memory_clock <= data->mclk_stutter_mode_threshold) && + // (data->is_uvd_enabled == 0) + // && (PHM_READ_FIELD(hwmgr->device, DPG_PIPE_STUTTER_CONTROL, STUTTER_ENABLE) & 0x1) + // && (data->display_timing.num_existing_displays <= 2) + // && (data->display_timing.num_existing_displays != 0)) + // memory_level->StutterEnable = 1; + + /* decide strobe mode*/ + memory_level->StrobeEnable = (data->mclk_strobe_mode_threshold != 0) && + (memory_clock <= data->mclk_strobe_mode_threshold); + + /* decide EDC mode and memory clock ratio*/ + if (data->is_memory_GDDR5) { + memory_level->StrobeRatio = iceland_get_mclk_frequency_ratio(memory_clock, + memory_level->StrobeEnable); + + if ((data->mclk_edc_enable_threshold != 0) && + (memory_clock > data->mclk_edc_enable_threshold)) { + memory_level->EdcReadEnable = 1; + } + + if ((data->mclk_edc_wr_enable_threshold != 0) && + (memory_clock > data->mclk_edc_wr_enable_threshold)) { + memory_level->EdcWriteEnable = 1; + } + + if (memory_level->StrobeEnable) { + if (iceland_get_mclk_frequency_ratio(memory_clock, 1) >= + ((cgs_read_register(hwmgr->device, mmMC_SEQ_MISC7) >> 16) & 0xf)) { + dllStateOn = ((cgs_read_register(hwmgr->device, mmMC_SEQ_MISC5) >> 1) & 0x1) ? 1 : 0; + } else { + dllStateOn = ((cgs_read_register(hwmgr->device, mmMC_SEQ_MISC6) >> 1) & 0x1) ? 1 : 0; + } + + } else { + dllStateOn = data->dll_defaule_on; + } + } else { + memory_level->StrobeRatio = + iceland_get_ddr3_mclk_frequency_ratio(memory_clock); + dllStateOn = ((cgs_read_register(hwmgr->device, mmMC_SEQ_MISC5) >> 1) & 0x1) ? 1 : 0; + } + + result = iceland_calculate_mclk_params(hwmgr, + memory_clock, memory_level, memory_level->StrobeEnable, dllStateOn); + + if (0 == result) { + memory_level->MinVddc = PP_HOST_TO_SMC_UL(memory_level->MinVddc * VOLTAGE_SCALE); + CONVERT_FROM_HOST_TO_SMC_UL(memory_level->MinVddcPhases); + memory_level->MinVddci = PP_HOST_TO_SMC_UL(memory_level->MinVddci * VOLTAGE_SCALE); + memory_level->MinMvdd = PP_HOST_TO_SMC_UL(memory_level->MinMvdd * VOLTAGE_SCALE); + /* MCLK frequency in units of 10KHz*/ + CONVERT_FROM_HOST_TO_SMC_UL(memory_level->MclkFrequency); + /* Indicates maximum activity level for this performance level.*/ + CONVERT_FROM_HOST_TO_SMC_US(memory_level->ActivityLevel); + CONVERT_FROM_HOST_TO_SMC_UL(memory_level->MpllFuncCntl); + CONVERT_FROM_HOST_TO_SMC_UL(memory_level->MpllFuncCntl_1); + CONVERT_FROM_HOST_TO_SMC_UL(memory_level->MpllFuncCntl_2); + CONVERT_FROM_HOST_TO_SMC_UL(memory_level->MpllAdFuncCntl); + CONVERT_FROM_HOST_TO_SMC_UL(memory_level->MpllDqFuncCntl); + CONVERT_FROM_HOST_TO_SMC_UL(memory_level->MclkPwrmgtCntl); + CONVERT_FROM_HOST_TO_SMC_UL(memory_level->DllCntl); + CONVERT_FROM_HOST_TO_SMC_UL(memory_level->MpllSs1); + CONVERT_FROM_HOST_TO_SMC_UL(memory_level->MpllSs2); + } + + return result; +} + +/** + * Populates the SMC MVDD structure using the provided memory clock. + * + * @param hwmgr the address of the hardware manager + * @param mclk the MCLK value to be used in the decision if MVDD should be high or low. + * @param voltage the SMC VOLTAGE structure to be populated + */ +int iceland_populate_mvdd_value(struct pp_hwmgr *hwmgr, uint32_t mclk, SMU71_Discrete_VoltageLevel *voltage) +{ + const iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + uint32_t i = 0; + + if (ICELAND_VOLTAGE_CONTROL_NONE != data->mvdd_control) { + /* find mvdd value which clock is more than request */ + for (i = 0; i < hwmgr->dyn_state.mvdd_dependency_on_mclk->count; i++) { + if (mclk <= hwmgr->dyn_state.mvdd_dependency_on_mclk->entries[i].clk) { + /* Always round to higher voltage. */ + voltage->Voltage = data->mvdd_voltage_table.entries[i].value; + break; + } + } + + PP_ASSERT_WITH_CODE(i < hwmgr->dyn_state.mvdd_dependency_on_mclk->count, + "MVDD Voltage is outside the supported range.", return -1); + + } else { + return -1; + } + + return 0; +} + + +static int iceland_populate_smc_acpi_level(struct pp_hwmgr *hwmgr, + SMU71_Discrete_DpmTable *table) +{ + int result = 0; + const iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + pp_atomctrl_clock_dividers_vi dividers; + SMU71_Discrete_VoltageLevel voltage_level; + uint32_t spll_func_cntl = data->clock_registers.vCG_SPLL_FUNC_CNTL; + uint32_t spll_func_cntl_2 = data->clock_registers.vCG_SPLL_FUNC_CNTL_2; + uint32_t dll_cntl = data->clock_registers.vDLL_CNTL; + uint32_t mclk_pwrmgt_cntl = data->clock_registers.vMCLK_PWRMGT_CNTL; + + /* The ACPI state should not do DPM on DC (or ever).*/ + table->ACPILevel.Flags &= ~PPSMC_SWSTATE_FLAG_DC; + + if (data->acpi_vddc) + table->ACPILevel.MinVddc = PP_HOST_TO_SMC_UL(data->acpi_vddc * VOLTAGE_SCALE); + else + table->ACPILevel.MinVddc = PP_HOST_TO_SMC_UL(data->min_vddc_in_pp_table * VOLTAGE_SCALE); + + table->ACPILevel.MinVddcPhases = (data->vddc_phase_shed_control) ? 0 : 1; + + /* assign zero for now*/ + table->ACPILevel.SclkFrequency = atomctrl_get_reference_clock(hwmgr); + + /* get the engine clock dividers for this clock value*/ + result = atomctrl_get_engine_pll_dividers_vi(hwmgr, + table->ACPILevel.SclkFrequency, ÷rs); + + PP_ASSERT_WITH_CODE(result == 0, + "Error retrieving Engine Clock dividers from VBIOS.", return result); + + /* divider ID for required SCLK*/ + table->ACPILevel.SclkDid = (uint8_t)dividers.pll_post_divider; + table->ACPILevel.DisplayWatermark = PPSMC_DISPLAY_WATERMARK_LOW; + table->ACPILevel.DeepSleepDivId = 0; + + spll_func_cntl = PHM_SET_FIELD(spll_func_cntl, + CG_SPLL_FUNC_CNTL, SPLL_PWRON, 0); + spll_func_cntl = PHM_SET_FIELD(spll_func_cntl, + CG_SPLL_FUNC_CNTL, SPLL_RESET, 1); + spll_func_cntl_2 = PHM_SET_FIELD(spll_func_cntl_2, + CG_SPLL_FUNC_CNTL_2, SCLK_MUX_SEL, 4); + + table->ACPILevel.CgSpllFuncCntl = spll_func_cntl; + table->ACPILevel.CgSpllFuncCntl2 = spll_func_cntl_2; + table->ACPILevel.CgSpllFuncCntl3 = data->clock_registers.vCG_SPLL_FUNC_CNTL_3; + table->ACPILevel.CgSpllFuncCntl4 = data->clock_registers.vCG_SPLL_FUNC_CNTL_4; + table->ACPILevel.SpllSpreadSpectrum = data->clock_registers.vCG_SPLL_SPREAD_SPECTRUM; + table->ACPILevel.SpllSpreadSpectrum2 = data->clock_registers.vCG_SPLL_SPREAD_SPECTRUM_2; + table->ACPILevel.CcPwrDynRm = 0; + table->ACPILevel.CcPwrDynRm1 = 0; + + + /* For various features to be enabled/disabled while this level is active.*/ + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.Flags); + /* SCLK frequency in units of 10KHz*/ + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.SclkFrequency); + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.CgSpllFuncCntl); + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.CgSpllFuncCntl2); + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.CgSpllFuncCntl3); + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.CgSpllFuncCntl4); + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.SpllSpreadSpectrum); + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.SpllSpreadSpectrum2); + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.CcPwrDynRm); + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.CcPwrDynRm1); + + table->MemoryACPILevel.MinVddc = table->ACPILevel.MinVddc; + table->MemoryACPILevel.MinVddcPhases = table->ACPILevel.MinVddcPhases; + + /* CONVERT_FROM_HOST_TO_SMC_UL(table->MemoryACPILevel.MinVoltage);*/ + + if (0 == iceland_populate_mvdd_value(hwmgr, 0, &voltage_level)) + table->MemoryACPILevel.MinMvdd = + PP_HOST_TO_SMC_UL(voltage_level.Voltage * VOLTAGE_SCALE); + else + table->MemoryACPILevel.MinMvdd = 0; + + /* Force reset on DLL*/ + mclk_pwrmgt_cntl = PHM_SET_FIELD(mclk_pwrmgt_cntl, + MCLK_PWRMGT_CNTL, MRDCK0_RESET, 0x1); + mclk_pwrmgt_cntl = PHM_SET_FIELD(mclk_pwrmgt_cntl, + MCLK_PWRMGT_CNTL, MRDCK1_RESET, 0x1); + + /* Disable DLL in ACPIState*/ + mclk_pwrmgt_cntl = PHM_SET_FIELD(mclk_pwrmgt_cntl, + MCLK_PWRMGT_CNTL, MRDCK0_PDNB, 0); + mclk_pwrmgt_cntl = PHM_SET_FIELD(mclk_pwrmgt_cntl, + MCLK_PWRMGT_CNTL, MRDCK1_PDNB, 0); + + /* Enable DLL bypass signal*/ + dll_cntl = PHM_SET_FIELD(dll_cntl, + DLL_CNTL, MRDCK0_BYPASS, 0); + dll_cntl = PHM_SET_FIELD(dll_cntl, + DLL_CNTL, MRDCK1_BYPASS, 0); + + table->MemoryACPILevel.DllCntl = + PP_HOST_TO_SMC_UL(dll_cntl); + table->MemoryACPILevel.MclkPwrmgtCntl = + PP_HOST_TO_SMC_UL(mclk_pwrmgt_cntl); + table->MemoryACPILevel.MpllAdFuncCntl = + PP_HOST_TO_SMC_UL(data->clock_registers.vMPLL_AD_FUNC_CNTL); + table->MemoryACPILevel.MpllDqFuncCntl = + PP_HOST_TO_SMC_UL(data->clock_registers.vMPLL_DQ_FUNC_CNTL); + table->MemoryACPILevel.MpllFuncCntl = + PP_HOST_TO_SMC_UL(data->clock_registers.vMPLL_FUNC_CNTL); + table->MemoryACPILevel.MpllFuncCntl_1 = + PP_HOST_TO_SMC_UL(data->clock_registers.vMPLL_FUNC_CNTL_1); + table->MemoryACPILevel.MpllFuncCntl_2 = + PP_HOST_TO_SMC_UL(data->clock_registers.vMPLL_FUNC_CNTL_2); + table->MemoryACPILevel.MpllSs1 = + PP_HOST_TO_SMC_UL(data->clock_registers.vMPLL_SS1); + table->MemoryACPILevel.MpllSs2 = + PP_HOST_TO_SMC_UL(data->clock_registers.vMPLL_SS2); + + table->MemoryACPILevel.EnabledForThrottle = 0; + table->MemoryACPILevel.EnabledForActivity = 0; + table->MemoryACPILevel.UpHyst = 0; + table->MemoryACPILevel.DownHyst = 100; + table->MemoryACPILevel.VoltageDownHyst = 0; + /* Indicates maximum activity level for this performance level.*/ + table->MemoryACPILevel.ActivityLevel = PP_HOST_TO_SMC_US((uint16_t)data->mclk_activity_target); + + table->MemoryACPILevel.StutterEnable = 0; + table->MemoryACPILevel.StrobeEnable = 0; + table->MemoryACPILevel.EdcReadEnable = 0; + table->MemoryACPILevel.EdcWriteEnable = 0; + table->MemoryACPILevel.RttEnable = 0; + + return result; +} + +static int iceland_find_boot_level(struct iceland_single_dpm_table *table, uint32_t value, uint32_t *boot_level) +{ + int result = 0; + uint32_t i; + + for (i = 0; i < table->count; i++) { + if (value == table->dpm_levels[i].value) { + *boot_level = i; + result = 0; + } + } + return result; +} + +/** + * Calculates the SCLK dividers using the provided engine clock + * + * @param hwmgr the address of the hardware manager + * @param engine_clock the engine clock to use to populate the structure + * @param sclk the SMC SCLK structure to be populated + */ +int iceland_calculate_sclk_params(struct pp_hwmgr *hwmgr, + uint32_t engine_clock, SMU71_Discrete_GraphicsLevel *sclk) +{ + const iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + pp_atomctrl_clock_dividers_vi dividers; + uint32_t spll_func_cntl = data->clock_registers.vCG_SPLL_FUNC_CNTL; + uint32_t spll_func_cntl_3 = data->clock_registers.vCG_SPLL_FUNC_CNTL_3; + uint32_t spll_func_cntl_4 = data->clock_registers.vCG_SPLL_FUNC_CNTL_4; + uint32_t cg_spll_spread_spectrum = data->clock_registers.vCG_SPLL_SPREAD_SPECTRUM; + uint32_t cg_spll_spread_spectrum_2 = data->clock_registers.vCG_SPLL_SPREAD_SPECTRUM_2; + uint32_t reference_clock; + uint32_t reference_divider; + uint32_t fbdiv; + int result; + + /* get the engine clock dividers for this clock value*/ + result = atomctrl_get_engine_pll_dividers_vi(hwmgr, engine_clock, ÷rs); + + PP_ASSERT_WITH_CODE(result == 0, + "Error retrieving Engine Clock dividers from VBIOS.", return result); + + /* To get FBDIV we need to multiply this by 16384 and divide it by Fref.*/ + reference_clock = atomctrl_get_reference_clock(hwmgr); + + reference_divider = 1 + dividers.uc_pll_ref_div; + + /* low 14 bits is fraction and high 12 bits is divider*/ + fbdiv = dividers.ul_fb_div.ul_fb_divider & 0x3FFFFFF; + + /* SPLL_FUNC_CNTL setup*/ + spll_func_cntl = PHM_SET_FIELD(spll_func_cntl, + CG_SPLL_FUNC_CNTL, SPLL_REF_DIV, dividers.uc_pll_ref_div); + spll_func_cntl = PHM_SET_FIELD(spll_func_cntl, + CG_SPLL_FUNC_CNTL, SPLL_PDIV_A, dividers.uc_pll_post_div); + + /* SPLL_FUNC_CNTL_3 setup*/ + spll_func_cntl_3 = PHM_SET_FIELD(spll_func_cntl_3, + CG_SPLL_FUNC_CNTL_3, SPLL_FB_DIV, fbdiv); + + /* set to use fractional accumulation*/ + spll_func_cntl_3 = PHM_SET_FIELD(spll_func_cntl_3, + CG_SPLL_FUNC_CNTL_3, SPLL_DITHEN, 1); + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_EngineSpreadSpectrumSupport)) { + pp_atomctrl_internal_ss_info ss_info; + + uint32_t vcoFreq = engine_clock * dividers.uc_pll_post_div; + if (0 == atomctrl_get_engine_clock_spread_spectrum(hwmgr, vcoFreq, &ss_info)) { + /* + * ss_info.speed_spectrum_percentage -- in unit of 0.01% + * ss_info.speed_spectrum_rate -- in unit of khz + */ + /* clks = reference_clock * 10 / (REFDIV + 1) / speed_spectrum_rate / 2 */ + uint32_t clkS = reference_clock * 5 / (reference_divider * ss_info.speed_spectrum_rate); + + /* clkv = 2 * D * fbdiv / NS */ + uint32_t clkV = 4 * ss_info.speed_spectrum_percentage * fbdiv / (clkS * 10000); + + cg_spll_spread_spectrum = + PHM_SET_FIELD(cg_spll_spread_spectrum, CG_SPLL_SPREAD_SPECTRUM, CLKS, clkS); + cg_spll_spread_spectrum = + PHM_SET_FIELD(cg_spll_spread_spectrum, CG_SPLL_SPREAD_SPECTRUM, SSEN, 1); + cg_spll_spread_spectrum_2 = + PHM_SET_FIELD(cg_spll_spread_spectrum_2, CG_SPLL_SPREAD_SPECTRUM_2, CLKV, clkV); + } + } + + sclk->SclkFrequency = engine_clock; + sclk->CgSpllFuncCntl3 = spll_func_cntl_3; + sclk->CgSpllFuncCntl4 = spll_func_cntl_4; + sclk->SpllSpreadSpectrum = cg_spll_spread_spectrum; + sclk->SpllSpreadSpectrum2 = cg_spll_spread_spectrum_2; + sclk->SclkDid = (uint8_t)dividers.pll_post_divider; + + return 0; +} + +static uint8_t iceland_get_sleep_divider_id_from_clock(struct pp_hwmgr *hwmgr, + uint32_t engine_clock, uint32_t min_engine_clock_in_sr) +{ + uint32_t i, temp; + uint32_t min = (min_engine_clock_in_sr > ICELAND_MINIMUM_ENGINE_CLOCK) ? + min_engine_clock_in_sr : ICELAND_MINIMUM_ENGINE_CLOCK; + + PP_ASSERT_WITH_CODE((engine_clock >= min), + "Engine clock can't satisfy stutter requirement!", return 0); + + for (i = ICELAND_MAX_DEEPSLEEP_DIVIDER_ID;; i--) { + temp = engine_clock / (1 << i); + + if(temp >= min || i == 0) + break; + } + return (uint8_t)i; +} + +/** + * Populates single SMC SCLK structure using the provided engine clock + * + * @param hwmgr the address of the hardware manager + * @param engine_clock the engine clock to use to populate the structure + * @param sclk the SMC SCLK structure to be populated + */ +static int iceland_populate_single_graphic_level(struct pp_hwmgr *hwmgr, + uint32_t engine_clock, uint16_t sclk_activity_level_threshold, + SMU71_Discrete_GraphicsLevel *graphic_level) +{ + int result; + uint32_t threshold; + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + result = iceland_calculate_sclk_params(hwmgr, engine_clock, graphic_level); + + + /* populate graphics levels*/ + result = iceland_get_dependecy_volt_by_clk(hwmgr, + hwmgr->dyn_state.vddc_dependency_on_sclk, engine_clock, &graphic_level->MinVddc); + PP_ASSERT_WITH_CODE((0 == result), + "can not find VDDC voltage value for VDDC engine clock dependency table", return result); + + /* SCLK frequency in units of 10KHz*/ + graphic_level->SclkFrequency = engine_clock; + + /* + * Minimum VDDC phases required to support this level, it + * should get from dependence table. + */ + graphic_level->MinVddcPhases = 1; + + if (data->vddc_phase_shed_control) { + iceland_populate_phase_value_based_on_sclk(hwmgr, + hwmgr->dyn_state.vddc_phase_shed_limits_table, + engine_clock, + &graphic_level->MinVddcPhases); + } + + /* Indicates maximum activity level for this performance level. 50% for now*/ + graphic_level->ActivityLevel = sclk_activity_level_threshold; + + graphic_level->CcPwrDynRm = 0; + graphic_level->CcPwrDynRm1 = 0; + /* this level can be used if activity is high enough.*/ + graphic_level->EnabledForActivity = 1; + /* this level can be used for throttling.*/ + graphic_level->EnabledForThrottle = 1; + graphic_level->UpHyst = 0; + graphic_level->DownHyst = 100; + graphic_level->VoltageDownHyst = 0; + graphic_level->PowerThrottle = 0; + + threshold = engine_clock * data->fast_watermark_threshold / 100; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SclkDeepSleep)) { + graphic_level->DeepSleepDivId = + iceland_get_sleep_divider_id_from_clock(hwmgr, engine_clock, + data->display_timing.min_clock_insr); + } + + /* Default to slow, highest DPM level will be set to PPSMC_DISPLAY_WATERMARK_LOW later.*/ + graphic_level->DisplayWatermark = PPSMC_DISPLAY_WATERMARK_LOW; + + if (0 == result) { + graphic_level->MinVddc = PP_HOST_TO_SMC_UL(graphic_level->MinVddc * VOLTAGE_SCALE); + /* CONVERT_FROM_HOST_TO_SMC_UL(graphic_level->MinVoltage);*/ + CONVERT_FROM_HOST_TO_SMC_UL(graphic_level->MinVddcPhases); + CONVERT_FROM_HOST_TO_SMC_UL(graphic_level->SclkFrequency); + CONVERT_FROM_HOST_TO_SMC_US(graphic_level->ActivityLevel); + CONVERT_FROM_HOST_TO_SMC_UL(graphic_level->CgSpllFuncCntl3); + CONVERT_FROM_HOST_TO_SMC_UL(graphic_level->CgSpllFuncCntl4); + CONVERT_FROM_HOST_TO_SMC_UL(graphic_level->SpllSpreadSpectrum); + CONVERT_FROM_HOST_TO_SMC_UL(graphic_level->SpllSpreadSpectrum2); + CONVERT_FROM_HOST_TO_SMC_UL(graphic_level->CcPwrDynRm); + CONVERT_FROM_HOST_TO_SMC_UL(graphic_level->CcPwrDynRm1); + } + + return result; +} + +/** + * Populates all SMC SCLK levels' structure based on the trimmed allowed dpm engine clock states + * + * @param hwmgr the address of the hardware manager + */ +static int iceland_populate_all_graphic_levels(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + struct iceland_dpm_table *dpm_table = &data->dpm_table; + int result = 0; + uint32_t level_array_adress = data->dpm_table_start + + offsetof(SMU71_Discrete_DpmTable, GraphicsLevel); + + uint32_t level_array_size = sizeof(SMU71_Discrete_GraphicsLevel) * SMU71_MAX_LEVELS_GRAPHICS; + SMU71_Discrete_GraphicsLevel *levels = data->smc_state_table.GraphicsLevel; + uint32_t i; + uint8_t highest_pcie_level_enabled = 0, lowest_pcie_level_enabled = 0, mid_pcie_level_enabled = 0, count = 0; + memset(levels, 0x00, level_array_size); + + for (i = 0; i < dpm_table->sclk_table.count; i++) { + result = iceland_populate_single_graphic_level(hwmgr, + dpm_table->sclk_table.dpm_levels[i].value, + (uint16_t)data->activity_target[i], + &(data->smc_state_table.GraphicsLevel[i])); + if (0 != result) + return result; + + /* Making sure only DPM level 0-1 have Deep Sleep Div ID populated. */ + if (i > 1) + data->smc_state_table.GraphicsLevel[i].DeepSleepDivId = 0; + } + + /* set highest level watermark to high */ + if (dpm_table->sclk_table.count > 1) + data->smc_state_table.GraphicsLevel[dpm_table->sclk_table.count-1].DisplayWatermark = + PPSMC_DISPLAY_WATERMARK_HIGH; + + data->smc_state_table.GraphicsDpmLevelCount = + (uint8_t)dpm_table->sclk_table.count; + data->dpm_level_enable_mask.sclk_dpm_enable_mask = + iceland_get_dpm_level_enable_mask_value(&dpm_table->sclk_table); + + while ((data->dpm_level_enable_mask.pcie_dpm_enable_mask & + (1 << (highest_pcie_level_enabled + 1))) != 0) { + highest_pcie_level_enabled++; + } + + while ((data->dpm_level_enable_mask.pcie_dpm_enable_mask & + (1 << lowest_pcie_level_enabled)) == 0) { + lowest_pcie_level_enabled++; + } + + while ((count < highest_pcie_level_enabled) && + ((data->dpm_level_enable_mask.pcie_dpm_enable_mask & + (1 << (lowest_pcie_level_enabled + 1 + count))) == 0)) { + count++; + } + + mid_pcie_level_enabled = (lowest_pcie_level_enabled+1+count) < highest_pcie_level_enabled ? + (lowest_pcie_level_enabled + 1 + count) : highest_pcie_level_enabled; + + /* set pcieDpmLevel to highest_pcie_level_enabled*/ + for (i = 2; i < dpm_table->sclk_table.count; i++) { + data->smc_state_table.GraphicsLevel[i].pcieDpmLevel = highest_pcie_level_enabled; + } + + /* set pcieDpmLevel to lowest_pcie_level_enabled*/ + data->smc_state_table.GraphicsLevel[0].pcieDpmLevel = lowest_pcie_level_enabled; + + /* set pcieDpmLevel to mid_pcie_level_enabled*/ + data->smc_state_table.GraphicsLevel[1].pcieDpmLevel = mid_pcie_level_enabled; + + /* level count will send to smc once at init smc table and never change*/ + result = iceland_copy_bytes_to_smc(hwmgr->smumgr, level_array_adress, (uint8_t *)levels, (uint32_t)level_array_size, data->sram_end); + + if (0 != result) + return result; + + return 0; +} + +/** + * Populates all SMC MCLK levels' structure based on the trimmed allowed dpm memory clock states + * + * @param hwmgr the address of the hardware manager + */ + +static int iceland_populate_all_memory_levels(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + struct iceland_dpm_table *dpm_table = &data->dpm_table; + int result; + /* populate MCLK dpm table to SMU7 */ + uint32_t level_array_adress = data->dpm_table_start + offsetof(SMU71_Discrete_DpmTable, MemoryLevel); + uint32_t level_array_size = sizeof(SMU71_Discrete_MemoryLevel) * SMU71_MAX_LEVELS_MEMORY; + SMU71_Discrete_MemoryLevel *levels = data->smc_state_table.MemoryLevel; + uint32_t i; + + memset(levels, 0x00, level_array_size); + + for (i = 0; i < dpm_table->mclk_table.count; i++) { + PP_ASSERT_WITH_CODE((0 != dpm_table->mclk_table.dpm_levels[i].value), + "can not populate memory level as memory clock is zero", return -1); + result = iceland_populate_single_memory_level(hwmgr, dpm_table->mclk_table.dpm_levels[i].value, + &(data->smc_state_table.MemoryLevel[i])); + if (0 != result) { + return result; + } + } + + /* Only enable level 0 for now.*/ + data->smc_state_table.MemoryLevel[0].EnabledForActivity = 1; + + /* + * in order to prevent MC activity from stutter mode to push DPM up. + * the UVD change complements this by putting the MCLK in a higher state + * by default such that we are not effected by up threshold or and MCLK DPM latency. + */ + data->smc_state_table.MemoryLevel[0].ActivityLevel = 0x1F; + CONVERT_FROM_HOST_TO_SMC_US(data->smc_state_table.MemoryLevel[0].ActivityLevel); + + data->smc_state_table.MemoryDpmLevelCount = (uint8_t)dpm_table->mclk_table.count; + data->dpm_level_enable_mask.mclk_dpm_enable_mask = iceland_get_dpm_level_enable_mask_value(&dpm_table->mclk_table); + /* set highest level watermark to high*/ + data->smc_state_table.MemoryLevel[dpm_table->mclk_table.count-1].DisplayWatermark = PPSMC_DISPLAY_WATERMARK_HIGH; + + /* level count will send to smc once at init smc table and never change*/ + result = iceland_copy_bytes_to_smc(hwmgr->smumgr, + level_array_adress, (uint8_t *)levels, (uint32_t)level_array_size, data->sram_end); + + if (0 != result) { + return result; + } + + return 0; +} + +struct ICELAND_DLL_SPEED_SETTING +{ + uint16_t Min; /* Minimum Data Rate*/ + uint16_t Max; /* Maximum Data Rate*/ + uint32_t dll_speed; /* The desired DLL_SPEED setting*/ +}; + +static int iceland_populate_ulv_level(struct pp_hwmgr *hwmgr, SMU71_Discrete_Ulv *pstate) +{ + int result = 0; + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + uint32_t voltage_response_time, ulv_voltage; + + pstate->CcPwrDynRm = 0; + pstate->CcPwrDynRm1 = 0; + + //backbiasResponseTime is use for ULV state voltage value. + result = pp_tables_get_response_times(hwmgr, &voltage_response_time, &ulv_voltage); + PP_ASSERT_WITH_CODE((0 == result), "can not get ULV voltage value", return result;); + + if(!ulv_voltage) { + data->ulv.ulv_supported = false; + return 0; + } + + if (ICELAND_VOLTAGE_CONTROL_BY_SVID2 != data->voltage_control) { + /* use minimum voltage if ulv voltage in pptable is bigger than minimum voltage */ + if (ulv_voltage > hwmgr->dyn_state.vddc_dependency_on_sclk->entries[0].v) { + pstate->VddcOffset = 0; + } + else { + /* used in SMIO Mode. not implemented for now. this is backup only for CI. */ + pstate->VddcOffset = (uint16_t)(hwmgr->dyn_state.vddc_dependency_on_sclk->entries[0].v - ulv_voltage); + } + } else { + /* use minimum voltage if ulv voltage in pptable is bigger than minimum voltage */ + if(ulv_voltage > hwmgr->dyn_state.vddc_dependency_on_sclk->entries[0].v) { + pstate->VddcOffsetVid = 0; + } else { + /* used in SVI2 Mode */ + pstate->VddcOffsetVid = (uint8_t)((hwmgr->dyn_state.vddc_dependency_on_sclk->entries[0].v - ulv_voltage) * VOLTAGE_VID_OFFSET_SCALE2 / VOLTAGE_VID_OFFSET_SCALE1); + } + } + + /* used in SVI2 Mode to shed phase */ + pstate->VddcPhase = (data->vddc_phase_shed_control) ? 0 : 1; + + if (0 == result) { + CONVERT_FROM_HOST_TO_SMC_UL(pstate->CcPwrDynRm); + CONVERT_FROM_HOST_TO_SMC_UL(pstate->CcPwrDynRm1); + CONVERT_FROM_HOST_TO_SMC_US(pstate->VddcOffset); + } + + return result; +} + +static int iceland_populate_ulv_state(struct pp_hwmgr *hwmgr, SMU71_Discrete_Ulv *ulv) +{ + return iceland_populate_ulv_level(hwmgr, ulv); +} + +static int iceland_populate_smc_initial_state(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + uint8_t count, level; + + count = (uint8_t)(hwmgr->dyn_state.vddc_dependency_on_sclk->count); + + for (level = 0; level < count; level++) { + if (hwmgr->dyn_state.vddc_dependency_on_sclk->entries[level].clk + >= data->vbios_boot_state.sclk_bootup_value) { + data->smc_state_table.GraphicsBootLevel = level; + break; + } + } + + count = (uint8_t)(hwmgr->dyn_state.vddc_dependency_on_mclk->count); + + for (level = 0; level < count; level++) { + if (hwmgr->dyn_state.vddc_dependency_on_mclk->entries[level].clk + >= data->vbios_boot_state.mclk_bootup_value) { + data->smc_state_table.MemoryBootLevel = level; + break; + } + } + + return 0; +} + +/** + * Initializes the SMC table and uploads it + * + * @param hwmgr the address of the powerplay hardware manager. + * @param pInput the pointer to input data (PowerState) + * @return always 0 + */ +int iceland_init_smc_table(struct pp_hwmgr *hwmgr) +{ + int result; + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + SMU71_Discrete_DpmTable *table = &(data->smc_state_table); + const struct phw_iceland_ulv_parm *ulv = &(data->ulv); + + result = iceland_setup_default_dpm_tables(hwmgr); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to setup default DPM tables!", return result;); + memset(&(data->smc_state_table), 0x00, sizeof(data->smc_state_table)); + + if (ICELAND_VOLTAGE_CONTROL_NONE != data->voltage_control) { + iceland_populate_smc_voltage_tables(hwmgr, table); + } + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_AutomaticDCTransition)) { + table->SystemFlags |= PPSMC_SYSTEMFLAG_GPIO_DC; + } + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_StepVddc)) { + table->SystemFlags |= PPSMC_SYSTEMFLAG_STEPVDDC; + } + + if (data->is_memory_GDDR5) { + table->SystemFlags |= PPSMC_SYSTEMFLAG_GDDR5; + } + + if (ulv->ulv_supported) { + result = iceland_populate_ulv_state(hwmgr, &data->ulv_setting); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to initialize ULV state!", return result;); + + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCG_ULV_PARAMETER, ulv->ch_ulv_parameter); + } + + result = iceland_populate_smc_link_level(hwmgr, table); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to initialize Link Level!", return result;); + + result = iceland_populate_all_graphic_levels(hwmgr); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to initialize Graphics Level!", return result;); + + result = iceland_populate_all_memory_levels(hwmgr); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to initialize Memory Level!", return result;); + + result = iceland_populate_smc_acpi_level(hwmgr, table); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to initialize ACPI Level!", return result;); + + result = iceland_populate_smc_vce_level(hwmgr, table); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to initialize VCE Level!", return result;); + + result = iceland_populate_smc_acp_level(hwmgr, table); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to initialize ACP Level!", return result;); + + result = iceland_populate_smc_samu_level(hwmgr, table); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to initialize SAMU Level!", return result;); + + /* + * Since only the initial state is completely set up at this + * point (the other states are just copies of the boot state) + * we only need to populate the ARB settings for the initial + * state. + */ + result = iceland_program_memory_timing_parameters(hwmgr); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to Write ARB settings for the initial state.", return result;); + + result = iceland_populate_smc_uvd_level(hwmgr, table); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to initialize UVD Level!", return result;); + + table->GraphicsBootLevel = 0; + table->MemoryBootLevel = 0; + + /* find boot level from dpm table */ + result = iceland_find_boot_level(&(data->dpm_table.sclk_table), + data->vbios_boot_state.sclk_bootup_value, + (uint32_t *)&(data->smc_state_table.GraphicsBootLevel)); + + if (result) + pr_warning("VBIOS did not find boot engine clock value in dependency table.\n"); + + result = iceland_find_boot_level(&(data->dpm_table.mclk_table), + data->vbios_boot_state.mclk_bootup_value, + (uint32_t *)&(data->smc_state_table.MemoryBootLevel)); + + if (result) + pr_warning("VBIOS did not find boot memory clock value in dependency table.\n"); + + table->BootVddc = data->vbios_boot_state.vddc_bootup_value; + if (ICELAND_VOLTAGE_CONTROL_NONE == data->vdd_ci_control) { + table->BootVddci = table->BootVddc; + } + else { + table->BootVddci = data->vbios_boot_state.vddci_bootup_value; + } + table->BootMVdd = data->vbios_boot_state.mvdd_bootup_value; + + result = iceland_populate_smc_initial_state(hwmgr); + PP_ASSERT_WITH_CODE(0 == result, "Failed to initialize Boot State!", return result); + + result = iceland_populate_bapm_parameters_in_dpm_table(hwmgr); + PP_ASSERT_WITH_CODE(0 == result, "Failed to populate BAPM Parameters!", return result); + + table->GraphicsVoltageChangeEnable = 1; + table->GraphicsThermThrottleEnable = 1; + table->GraphicsInterval = 1; + table->VoltageInterval = 1; + table->ThermalInterval = 1; + table->TemperatureLimitHigh = + (data->thermal_temp_setting.temperature_high * + ICELAND_Q88_FORMAT_CONVERSION_UNIT) / PP_TEMPERATURE_UNITS_PER_CENTIGRADES; + table->TemperatureLimitLow = + (data->thermal_temp_setting.temperature_low * + ICELAND_Q88_FORMAT_CONVERSION_UNIT) / PP_TEMPERATURE_UNITS_PER_CENTIGRADES; + table->MemoryVoltageChangeEnable = 1; + table->MemoryInterval = 1; + table->VoltageResponseTime = 0; + table->PhaseResponseTime = 0; + table->MemoryThermThrottleEnable = 1; + table->PCIeBootLinkLevel = 0; + table->PCIeGenInterval = 1; + + result = iceland_populate_smc_svi2_config(hwmgr, table); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to populate SVI2 setting!", return result); + + table->ThermGpio = 17; + table->SclkStepSize = 0x4000; + + CONVERT_FROM_HOST_TO_SMC_UL(table->SystemFlags); + CONVERT_FROM_HOST_TO_SMC_UL(table->SmioMaskVddcVid); + CONVERT_FROM_HOST_TO_SMC_UL(table->SmioMaskVddcPhase); + CONVERT_FROM_HOST_TO_SMC_UL(table->SmioMaskVddciVid); + CONVERT_FROM_HOST_TO_SMC_UL(table->SmioMaskMvddVid); + CONVERT_FROM_HOST_TO_SMC_UL(table->SclkStepSize); + CONVERT_FROM_HOST_TO_SMC_US(table->TemperatureLimitHigh); + CONVERT_FROM_HOST_TO_SMC_US(table->TemperatureLimitLow); + CONVERT_FROM_HOST_TO_SMC_US(table->VoltageResponseTime); + CONVERT_FROM_HOST_TO_SMC_US(table->PhaseResponseTime); + + table->BootVddc = PP_HOST_TO_SMC_US(table->BootVddc * VOLTAGE_SCALE); + table->BootVddci = PP_HOST_TO_SMC_US(table->BootVddci * VOLTAGE_SCALE); + table->BootMVdd = PP_HOST_TO_SMC_US(table->BootMVdd * VOLTAGE_SCALE); + + /* Upload all dpm data to SMC memory.(dpm level, dpm level count etc) */ + result = iceland_copy_bytes_to_smc(hwmgr->smumgr, data->dpm_table_start + + offsetof(SMU71_Discrete_DpmTable, SystemFlags), + (uint8_t *)&(table->SystemFlags), + sizeof(SMU71_Discrete_DpmTable) - 3 * sizeof(SMU71_PIDController), + data->sram_end); + + PP_ASSERT_WITH_CODE(0 == result, + "Failed to upload dpm data to SMC memory!", return result); + + /* Upload all ulv setting to SMC memory.(dpm level, dpm level count etc) */ + result = iceland_copy_bytes_to_smc(hwmgr->smumgr, + data->ulv_settings_start, + (uint8_t *)&(data->ulv_setting), + sizeof(SMU71_Discrete_Ulv), + data->sram_end); + +#if 0 + /* Notify SMC to follow new GPIO scheme */ + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_AutomaticDCTransition)) { + if (0 == iceland_send_msg_to_smc(hwmgr->smumgr, PPSMC_MSG_UseNewGPIOScheme)) + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SMCtoPPLIBAcdcGpioScheme); + } +#endif + + return result; +} + +int iceland_populate_mc_reg_address(struct pp_hwmgr *hwmgr, SMU71_Discrete_MCRegisters *mc_reg_table) +{ + const struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + uint32_t i, j; + + for (i = 0, j = 0; j < data->iceland_mc_reg_table.last; j++) { + if (data->iceland_mc_reg_table.validflag & 1<<j) { + PP_ASSERT_WITH_CODE(i < SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE, + "Index of mc_reg_table->address[] array out of boundary", return -1); + mc_reg_table->address[i].s0 = + PP_HOST_TO_SMC_US(data->iceland_mc_reg_table.mc_reg_address[j].s0); + mc_reg_table->address[i].s1 = + PP_HOST_TO_SMC_US(data->iceland_mc_reg_table.mc_reg_address[j].s1); + i++; + } + } + + mc_reg_table->last = (uint8_t)i; + + return 0; +} + +/* convert register values from driver to SMC format */ +void iceland_convert_mc_registers( + const phw_iceland_mc_reg_entry * pEntry, + SMU71_Discrete_MCRegisterSet *pData, + uint32_t numEntries, uint32_t validflag) +{ + uint32_t i, j; + + for (i = 0, j = 0; j < numEntries; j++) { + if (validflag & 1<<j) { + pData->value[i] = PP_HOST_TO_SMC_UL(pEntry->mc_data[j]); + i++; + } + } +} + +/* find the entry in the memory range table, then populate the value to SMC's iceland_mc_reg_table */ +int iceland_convert_mc_reg_table_entry_to_smc( + struct pp_hwmgr *hwmgr, + const uint32_t memory_clock, + SMU71_Discrete_MCRegisterSet *mc_reg_table_data + ) +{ + const iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + uint32_t i = 0; + + for (i = 0; i < data->iceland_mc_reg_table.num_entries; i++) { + if (memory_clock <= + data->iceland_mc_reg_table.mc_reg_table_entry[i].mclk_max) { + break; + } + } + + if ((i == data->iceland_mc_reg_table.num_entries) && (i > 0)) + --i; + + iceland_convert_mc_registers(&data->iceland_mc_reg_table.mc_reg_table_entry[i], + mc_reg_table_data, data->iceland_mc_reg_table.last, data->iceland_mc_reg_table.validflag); + + return 0; +} + +int iceland_convert_mc_reg_table_to_smc(struct pp_hwmgr *hwmgr, + SMU71_Discrete_MCRegisters *mc_reg_table) +{ + int result = 0; + iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + int res; + uint32_t i; + + for (i = 0; i < data->dpm_table.mclk_table.count; i++) { + res = iceland_convert_mc_reg_table_entry_to_smc( + hwmgr, + data->dpm_table.mclk_table.dpm_levels[i].value, + &mc_reg_table->data[i] + ); + + if (0 != res) + result = res; + } + + return result; +} + +int iceland_populate_initial_mc_reg_table(struct pp_hwmgr *hwmgr) +{ + int result; + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + memset(&data->mc_reg_table, 0x00, sizeof(SMU71_Discrete_MCRegisters)); + result = iceland_populate_mc_reg_address(hwmgr, &(data->mc_reg_table)); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to initialize MCRegTable for the MC register addresses!", return result;); + + result = iceland_convert_mc_reg_table_to_smc(hwmgr, &data->mc_reg_table); + PP_ASSERT_WITH_CODE(0 == result, + "Failed to initialize MCRegTable for driver state!", return result;); + + return iceland_copy_bytes_to_smc(hwmgr->smumgr, data->mc_reg_table_start, + (uint8_t *)&data->mc_reg_table, sizeof(SMU71_Discrete_MCRegisters), data->sram_end); +} + +int iceland_notify_smc_display_change(struct pp_hwmgr *hwmgr, bool has_display) +{ + PPSMC_Msg msg = has_display? (PPSMC_Msg)PPSMC_HasDisplay : (PPSMC_Msg)PPSMC_NoDisplay; + + return (smum_send_msg_to_smc(hwmgr->smumgr, msg) == 0) ? 0 : -1; +} + +int iceland_enable_sclk_control(struct pp_hwmgr *hwmgr) +{ + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, SCLK_PWRMGT_CNTL, SCLK_PWRMGT_OFF, 0); + + return 0; +} + +int iceland_enable_sclk_mclk_dpm(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + /* enable SCLK dpm */ + if (0 == data->sclk_dpm_key_disabled) { + PP_ASSERT_WITH_CODE( + (0 == smum_send_msg_to_smc(hwmgr->smumgr, + PPSMC_MSG_DPM_Enable)), + "Failed to enable SCLK DPM during DPM Start Function!", + return -1); + } + + /* enable MCLK dpm */ + if (0 == data->mclk_dpm_key_disabled) { + PP_ASSERT_WITH_CODE( + (0 == smum_send_msg_to_smc(hwmgr->smumgr, + PPSMC_MSG_MCLKDPM_Enable)), + "Failed to enable MCLK DPM during DPM Start Function!", + return -1); + + PHM_WRITE_FIELD(hwmgr->device, MC_SEQ_CNTL_3, CAC_EN, 0x1); + + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixLCAC_MC0_CNTL, 0x05);/* CH0,1 read */ + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixLCAC_MC1_CNTL, 0x05);/* CH2,3 read */ + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixLCAC_CPL_CNTL, 0x100005);/*Read */ + + udelay(10); + + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixLCAC_MC0_CNTL, 0x400005);/* CH0,1 write */ + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixLCAC_MC1_CNTL, 0x400005);/* CH2,3 write */ + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixLCAC_CPL_CNTL, 0x500005);/* write */ + + } + + return 0; +} + +int iceland_start_dpm(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + /* enable general power management */ + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, GENERAL_PWRMGT, GLOBAL_PWRMGT_EN, 1); + /* enable sclk deep sleep */ + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, SCLK_PWRMGT_CNTL, DYNAMIC_PM_EN, 1); + + /* prepare for PCIE DPM */ + PHM_WRITE_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, SOFT_REGISTERS_TABLE_12, VoltageChangeTimeout, 0x1000); + + PHM_WRITE_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__PCIE, SWRST_COMMAND_1, RESETLC, 0x0); + + PP_ASSERT_WITH_CODE( + (0 == smum_send_msg_to_smc(hwmgr->smumgr, + PPSMC_MSG_Voltage_Cntl_Enable)), + "Failed to enable voltage DPM during DPM Start Function!", + return -1); + + if (0 != iceland_enable_sclk_mclk_dpm(hwmgr)) { + PP_ASSERT_WITH_CODE(0, "Failed to enable Sclk DPM and Mclk DPM!", return -1); + } + + /* enable PCIE dpm */ + if (0 == data->pcie_dpm_key_disabled) { + PP_ASSERT_WITH_CODE( + (0 == smum_send_msg_to_smc(hwmgr->smumgr, + PPSMC_MSG_PCIeDPM_Enable)), + "Failed to enable pcie DPM during DPM Start Function!", + return -1 + ); + } + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_Falcon_QuickTransition)) { + smum_send_msg_to_smc(hwmgr->smumgr, + PPSMC_MSG_EnableACDCGPIOInterrupt); + } + + return 0; +} + +static void iceland_set_dpm_event_sources(struct pp_hwmgr *hwmgr, + uint32_t sources) +{ + bool protection; + enum DPM_EVENT_SRC src; + + switch (sources) { + default: + printk(KERN_ERR "Unknown throttling event sources."); + /* fall through */ + case 0: + protection = false; + /* src is unused */ + break; + case (1 << PHM_AutoThrottleSource_Thermal): + protection = true; + src = DPM_EVENT_SRC_DIGITAL; + break; + case (1 << PHM_AutoThrottleSource_External): + protection = true; + src = DPM_EVENT_SRC_EXTERNAL; + break; + case (1 << PHM_AutoThrottleSource_External) | + (1 << PHM_AutoThrottleSource_Thermal): + protection = true; + src = DPM_EVENT_SRC_DIGITAL_OR_EXTERNAL; + break; + } + /* Order matters - don't enable thermal protection for the wrong source. */ + if (protection) { + PHM_WRITE_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_THERMAL_CTRL, + DPM_EVENT_SRC, src); + PHM_WRITE_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, GENERAL_PWRMGT, + THERMAL_PROTECTION_DIS, + !phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_ThermalController)); + } else + PHM_WRITE_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, GENERAL_PWRMGT, + THERMAL_PROTECTION_DIS, 1); +} + +static int iceland_enable_auto_throttle_source(struct pp_hwmgr *hwmgr, + PHM_AutoThrottleSource source) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + if (!(data->active_auto_throttle_sources & (1 << source))) { + data->active_auto_throttle_sources |= 1 << source; + iceland_set_dpm_event_sources(hwmgr, data->active_auto_throttle_sources); + } + return 0; +} + +static int iceland_enable_thermal_auto_throttle(struct pp_hwmgr *hwmgr) +{ + return iceland_enable_auto_throttle_source(hwmgr, PHM_AutoThrottleSource_Thermal); +} + +static int iceland_tf_start_smc(struct pp_hwmgr *hwmgr) +{ + int ret = 0; + + if (!iceland_is_smc_ram_running(hwmgr->smumgr)) + ret = iceland_smu_start_smc(hwmgr->smumgr); + + return ret; +} + +/** +* Programs the Deep Sleep registers +* +* @param pHwMgr the address of the powerplay hardware manager. +* @param pInput the pointer to input data (PhwEvergreen_DisplayConfiguration) +* @param pOutput the pointer to output data (unused) +* @param pStorage the pointer to temporary storage (unused) +* @param Result the last failure code (unused) +* @return always 0 +*/ +static int iceland_enable_deep_sleep_master_switch(struct pp_hwmgr *hwmgr) +{ + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SclkDeepSleep)) { + if (smum_send_msg_to_smc(hwmgr->smumgr, + PPSMC_MSG_MASTER_DeepSleep_ON) != 0) + PP_ASSERT_WITH_CODE(false, + "Attempt to enable Master Deep Sleep switch failed!", + return -EINVAL); + } else { + if (smum_send_msg_to_smc(hwmgr->smumgr, + PPSMC_MSG_MASTER_DeepSleep_OFF) != 0) + PP_ASSERT_WITH_CODE(false, + "Attempt to disable Master Deep Sleep switch failed!", + return -EINVAL); + } + + return 0; +} + +static int iceland_enable_dpm_tasks(struct pp_hwmgr *hwmgr) +{ + int tmp_result, result = 0; + + if (cf_iceland_voltage_control(hwmgr)) { + tmp_result = iceland_enable_voltage_control(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to enable voltage control!", return tmp_result); + + tmp_result = iceland_construct_voltage_tables(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to contruct voltage tables!", return tmp_result); + } + + tmp_result = iceland_initialize_mc_reg_table(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to initialize MC reg table!", return tmp_result); + + tmp_result = iceland_program_static_screen_threshold_parameters(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to program static screen threshold parameters!", return tmp_result); + + tmp_result = iceland_enable_display_gap(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to enable display gap!", return tmp_result); + + tmp_result = iceland_program_voting_clients(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to program voting clients!", return tmp_result); + + tmp_result = iceland_upload_firmware(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to upload firmware header!", return tmp_result); + + tmp_result = iceland_process_firmware_header(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to process firmware header!", return tmp_result); + + tmp_result = iceland_initial_switch_from_arb_f0_to_f1(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to initialize switch from ArbF0 to F1!", return tmp_result); + + tmp_result = iceland_init_smc_table(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to initialize SMC table!", return tmp_result); + + tmp_result = iceland_populate_initial_mc_reg_table(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to populate initialize MC Reg table!", return tmp_result); + + tmp_result = iceland_populate_pm_fuses(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to populate PM fuses!", return tmp_result); + + /* start SMC */ + tmp_result = iceland_tf_start_smc(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to start SMC!", return tmp_result); + + /* enable SCLK control */ + tmp_result = iceland_enable_sclk_control(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to enable SCLK control!", return tmp_result); + + tmp_result = iceland_enable_deep_sleep_master_switch(hwmgr); + PP_ASSERT_WITH_CODE((tmp_result == 0), + "Failed to enable deep sleep!", return tmp_result); + + /* enable DPM */ + tmp_result = iceland_start_dpm(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to start DPM!", return tmp_result); + + tmp_result = iceland_enable_smc_cac(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to enable SMC CAC!", return tmp_result); + + tmp_result = iceland_enable_power_containment(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to enable power containment!", return tmp_result); + + tmp_result = iceland_power_control_set_level(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to power control set level!", result = tmp_result); + + tmp_result = iceland_enable_thermal_auto_throttle(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to enable thermal auto throttle!", result = tmp_result); + + return result; +} + +static int iceland_hwmgr_backend_fini(struct pp_hwmgr *hwmgr) +{ + return phm_hwmgr_backend_fini(hwmgr); +} + +static void iceland_initialize_dpm_defaults(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + struct phw_iceland_ulv_parm *ulv; + + ulv = &data->ulv; + ulv->ch_ulv_parameter = PPICELAND_CGULVPARAMETER_DFLT; + data->voting_rights_clients0 = PPICELAND_VOTINGRIGHTSCLIENTS_DFLT0; + data->voting_rights_clients1 = PPICELAND_VOTINGRIGHTSCLIENTS_DFLT1; + data->voting_rights_clients2 = PPICELAND_VOTINGRIGHTSCLIENTS_DFLT2; + data->voting_rights_clients3 = PPICELAND_VOTINGRIGHTSCLIENTS_DFLT3; + data->voting_rights_clients4 = PPICELAND_VOTINGRIGHTSCLIENTS_DFLT4; + data->voting_rights_clients5 = PPICELAND_VOTINGRIGHTSCLIENTS_DFLT5; + data->voting_rights_clients6 = PPICELAND_VOTINGRIGHTSCLIENTS_DFLT6; + data->voting_rights_clients7 = PPICELAND_VOTINGRIGHTSCLIENTS_DFLT7; + + data->static_screen_threshold_unit = PPICELAND_STATICSCREENTHRESHOLDUNIT_DFLT; + data->static_screen_threshold = PPICELAND_STATICSCREENTHRESHOLD_DFLT; + + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_ABM); + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_NonABMSupportInPPLib); + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DynamicACTiming); + + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DisableMemoryTransition); + + iceland_initialize_power_tune_defaults(hwmgr); + + data->mclk_strobe_mode_threshold = 40000; + data->mclk_stutter_mode_threshold = 30000; + data->mclk_edc_enable_threshold = 40000; + data->mclk_edc_wr_enable_threshold = 40000; + + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DisableMCLS); + + data->pcie_gen_performance.max = PP_PCIEGen1; + data->pcie_gen_performance.min = PP_PCIEGen3; + data->pcie_gen_power_saving.max = PP_PCIEGen1; + data->pcie_gen_power_saving.min = PP_PCIEGen3; + + data->pcie_lane_performance.max = 0; + data->pcie_lane_performance.min = 16; + data->pcie_lane_power_saving.max = 0; + data->pcie_lane_power_saving.min = 16; + + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SclkThrottleLowNotification); +} + +static int iceland_get_evv_voltage(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + uint16_t virtual_voltage_id; + uint16_t vddc = 0; + uint16_t i; + + /* the count indicates actual number of entries */ + data->vddc_leakage.count = 0; + data->vddci_leakage.count = 0; + + if (!phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_EVV)) { + pr_err("Iceland should always support EVV\n"); + return -EINVAL; + } + + /* retrieve voltage for leakage ID (0xff01 + i) */ + for (i = 0; i < ICELAND_MAX_LEAKAGE_COUNT; i++) { + virtual_voltage_id = ATOM_VIRTUAL_VOLTAGE_ID0 + i; + + PP_ASSERT_WITH_CODE((0 == atomctrl_get_voltage_evv(hwmgr, virtual_voltage_id, &vddc)), + "Error retrieving EVV voltage value!\n", continue); + + if (vddc >= 2000) + pr_warning("Invalid VDDC value!\n"); + + if (vddc != 0 && vddc != virtual_voltage_id) { + data->vddc_leakage.actual_voltage[data->vddc_leakage.count] = vddc; + data->vddc_leakage.leakage_id[data->vddc_leakage.count] = virtual_voltage_id; + data->vddc_leakage.count++; + } + } + + return 0; +} + +static void iceland_patch_with_vddc_leakage(struct pp_hwmgr *hwmgr, + uint32_t *vddc) +{ + iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + uint32_t leakage_index; + struct phw_iceland_leakage_voltage *leakage_table = &data->vddc_leakage; + + /* search for leakage voltage ID 0xff01 ~ 0xff08 */ + for (leakage_index = 0; leakage_index < leakage_table->count; leakage_index++) { + /* + * If this voltage matches a leakage voltage ID, patch + * with actual leakage voltage. + */ + if (leakage_table->leakage_id[leakage_index] == *vddc) { + /* + * Need to make sure vddc is less than 2v or + * else, it could burn the ASIC. + */ + if (leakage_table->actual_voltage[leakage_index] >= 2000) + pr_warning("Invalid VDDC value!\n"); + *vddc = leakage_table->actual_voltage[leakage_index]; + /* we found leakage voltage */ + break; + } + } + + if (*vddc >= ATOM_VIRTUAL_VOLTAGE_ID0) + pr_warning("Voltage value looks like a Leakage ID but it's not patched\n"); +} + +static void iceland_patch_with_vddci_leakage(struct pp_hwmgr *hwmgr, + uint32_t *vddci) +{ + iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + uint32_t leakage_index; + struct phw_iceland_leakage_voltage *leakage_table = &data->vddci_leakage; + + /* search for leakage voltage ID 0xff01 ~ 0xff08 */ + for (leakage_index = 0; leakage_index < leakage_table->count; leakage_index++) { + /* + * If this voltage matches a leakage voltage ID, patch + * with actual leakage voltage. + */ + if (leakage_table->leakage_id[leakage_index] == *vddci) { + *vddci = leakage_table->actual_voltage[leakage_index]; + /* we found leakage voltage */ + break; + } + } + + if (*vddci >= ATOM_VIRTUAL_VOLTAGE_ID0) + pr_warning("Voltage value looks like a Leakage ID but it's not patched\n"); +} + +static int iceland_patch_vddc(struct pp_hwmgr *hwmgr, + struct phm_clock_voltage_dependency_table *tab) +{ + uint16_t i; + + if (tab) + for (i = 0; i < tab->count; i++) + iceland_patch_with_vddc_leakage(hwmgr, &tab->entries[i].v); + + return 0; +} + +static int iceland_patch_vddci(struct pp_hwmgr *hwmgr, + struct phm_clock_voltage_dependency_table *tab) +{ + uint16_t i; + + if (tab) + for (i = 0; i < tab->count; i++) + iceland_patch_with_vddci_leakage(hwmgr, &tab->entries[i].v); + + return 0; +} + +static int iceland_patch_vce_vddc(struct pp_hwmgr *hwmgr, + struct phm_vce_clock_voltage_dependency_table *tab) +{ + uint16_t i; + + if (tab) + for (i = 0; i < tab->count; i++) + iceland_patch_with_vddc_leakage(hwmgr, &tab->entries[i].v); + + return 0; +} + + +static int iceland_patch_uvd_vddc(struct pp_hwmgr *hwmgr, + struct phm_uvd_clock_voltage_dependency_table *tab) +{ + uint16_t i; + + if (tab) + for (i = 0; i < tab->count; i++) + iceland_patch_with_vddc_leakage(hwmgr, &tab->entries[i].v); + + return 0; +} + +static int iceland_patch_vddc_shed_limit(struct pp_hwmgr *hwmgr, + struct phm_phase_shedding_limits_table *tab) +{ + uint16_t i; + + if (tab) + for (i = 0; i < tab->count; i++) + iceland_patch_with_vddc_leakage(hwmgr, &tab->entries[i].Voltage); + + return 0; +} + +static int iceland_patch_samu_vddc(struct pp_hwmgr *hwmgr, + struct phm_samu_clock_voltage_dependency_table *tab) +{ + uint16_t i; + + if (tab) + for (i = 0; i < tab->count; i++) + iceland_patch_with_vddc_leakage(hwmgr, &tab->entries[i].v); + + return 0; +} + +static int iceland_patch_acp_vddc(struct pp_hwmgr *hwmgr, + struct phm_acp_clock_voltage_dependency_table *tab) +{ + uint16_t i; + + if (tab) + for (i = 0; i < tab->count; i++) + iceland_patch_with_vddc_leakage(hwmgr, &tab->entries[i].v); + + return 0; +} + +static int iceland_patch_limits_vddc(struct pp_hwmgr *hwmgr, + struct phm_clock_and_voltage_limits *tab) +{ + if (tab) { + iceland_patch_with_vddc_leakage(hwmgr, (uint32_t *)&tab->vddc); + iceland_patch_with_vddci_leakage(hwmgr, (uint32_t *)&tab->vddci); + } + + return 0; +} + +static int iceland_patch_cac_vddc(struct pp_hwmgr *hwmgr, struct phm_cac_leakage_table *tab) +{ + uint32_t i; + uint32_t vddc; + + if (tab) { + for (i = 0; i < tab->count; i++) { + vddc = (uint32_t)(tab->entries[i].Vddc); + iceland_patch_with_vddc_leakage(hwmgr, &vddc); + tab->entries[i].Vddc = (uint16_t)vddc; + } + } + + return 0; +} + +static int iceland_patch_dependency_tables_with_leakage(struct pp_hwmgr *hwmgr) +{ + int tmp; + + tmp = iceland_patch_vddc(hwmgr, hwmgr->dyn_state.vddc_dependency_on_sclk); + if(tmp) + return -EINVAL; + + tmp = iceland_patch_vddc(hwmgr, hwmgr->dyn_state.vddc_dependency_on_mclk); + if(tmp) + return -EINVAL; + + tmp = iceland_patch_vddc(hwmgr, hwmgr->dyn_state.vddc_dep_on_dal_pwrl); + if(tmp) + return -EINVAL; + + tmp = iceland_patch_vddci(hwmgr, hwmgr->dyn_state.vddci_dependency_on_mclk); + if(tmp) + return -EINVAL; + + tmp = iceland_patch_vce_vddc(hwmgr, hwmgr->dyn_state.vce_clock_voltage_dependency_table); + if(tmp) + return -EINVAL; + + tmp = iceland_patch_uvd_vddc(hwmgr, hwmgr->dyn_state.uvd_clock_voltage_dependency_table); + if(tmp) + return -EINVAL; + + tmp = iceland_patch_samu_vddc(hwmgr, hwmgr->dyn_state.samu_clock_voltage_dependency_table); + if(tmp) + return -EINVAL; + + tmp = iceland_patch_acp_vddc(hwmgr, hwmgr->dyn_state.acp_clock_voltage_dependency_table); + if(tmp) + return -EINVAL; + + tmp = iceland_patch_vddc_shed_limit(hwmgr, hwmgr->dyn_state.vddc_phase_shed_limits_table); + if(tmp) + return -EINVAL; + + tmp = iceland_patch_limits_vddc(hwmgr, &hwmgr->dyn_state.max_clock_voltage_on_ac); + if(tmp) + return -EINVAL; + + tmp = iceland_patch_limits_vddc(hwmgr, &hwmgr->dyn_state.max_clock_voltage_on_dc); + if(tmp) + return -EINVAL; + + tmp = iceland_patch_cac_vddc(hwmgr, hwmgr->dyn_state.cac_leakage_table); + if(tmp) + return -EINVAL; + + return 0; +} + +static int iceland_set_private_var_based_on_pptale(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + struct phm_clock_voltage_dependency_table *allowed_sclk_vddc_table = hwmgr->dyn_state.vddc_dependency_on_sclk; + struct phm_clock_voltage_dependency_table *allowed_mclk_vddc_table = hwmgr->dyn_state.vddc_dependency_on_mclk; + struct phm_clock_voltage_dependency_table *allowed_mclk_vddci_table = hwmgr->dyn_state.vddci_dependency_on_mclk; + + PP_ASSERT_WITH_CODE(allowed_sclk_vddc_table != NULL, + "VDDC dependency on SCLK table is missing. This table is mandatory\n", return -EINVAL); + PP_ASSERT_WITH_CODE(allowed_sclk_vddc_table->count >= 1, + "VDDC dependency on SCLK table has to have is missing. This table is mandatory\n", return -EINVAL); + + PP_ASSERT_WITH_CODE(allowed_mclk_vddc_table != NULL, + "VDDC dependency on MCLK table is missing. This table is mandatory\n", return -EINVAL); + PP_ASSERT_WITH_CODE(allowed_mclk_vddc_table->count >= 1, + "VDD dependency on MCLK table has to have is missing. This table is mandatory\n", return -EINVAL); + + data->min_vddc_in_pp_table = (uint16_t)allowed_sclk_vddc_table->entries[0].v; + data->max_vddc_in_pp_table = (uint16_t)allowed_sclk_vddc_table->entries[allowed_sclk_vddc_table->count - 1].v; + + hwmgr->dyn_state.max_clock_voltage_on_ac.sclk = + allowed_sclk_vddc_table->entries[allowed_sclk_vddc_table->count - 1].clk; + hwmgr->dyn_state.max_clock_voltage_on_ac.mclk = + allowed_mclk_vddc_table->entries[allowed_mclk_vddc_table->count - 1].clk; + hwmgr->dyn_state.max_clock_voltage_on_ac.vddc = + allowed_sclk_vddc_table->entries[allowed_sclk_vddc_table->count - 1].v; + + if (allowed_mclk_vddci_table != NULL && allowed_mclk_vddci_table->count >= 1) { + data->min_vddci_in_pp_table = (uint16_t)allowed_mclk_vddci_table->entries[0].v; + data->max_vddci_in_pp_table = (uint16_t)allowed_mclk_vddci_table->entries[allowed_mclk_vddci_table->count - 1].v; + } + + if (hwmgr->dyn_state.vddci_dependency_on_mclk != NULL && hwmgr->dyn_state.vddci_dependency_on_mclk->count > 1) + hwmgr->dyn_state.max_clock_voltage_on_ac.vddci = hwmgr->dyn_state.vddci_dependency_on_mclk->entries[hwmgr->dyn_state.vddci_dependency_on_mclk->count - 1].v; + + return 0; +} + +static int iceland_initializa_dynamic_state_adjustment_rule_settings(struct pp_hwmgr *hwmgr) +{ + uint32_t table_size; + struct phm_clock_voltage_dependency_table *table_clk_vlt; + + hwmgr->dyn_state.mclk_sclk_ratio = 4; + hwmgr->dyn_state.sclk_mclk_delta = 15000; /* 150 MHz */ + hwmgr->dyn_state.vddc_vddci_delta = 200; /* 200mV */ + + /* initialize vddc_dep_on_dal_pwrl table */ + table_size = sizeof(uint32_t) + 4 * sizeof(struct phm_clock_voltage_dependency_record); + table_clk_vlt = (struct phm_clock_voltage_dependency_table *)kzalloc(table_size, GFP_KERNEL); + + if (NULL == table_clk_vlt) { + pr_err("[ powerplay ] Can not allocate space for vddc_dep_on_dal_pwrl! \n"); + return -ENOMEM; + } else { + table_clk_vlt->count = 4; + table_clk_vlt->entries[0].clk = PP_DAL_POWERLEVEL_ULTRALOW; + table_clk_vlt->entries[0].v = 0; + table_clk_vlt->entries[1].clk = PP_DAL_POWERLEVEL_LOW; + table_clk_vlt->entries[1].v = 720; + table_clk_vlt->entries[2].clk = PP_DAL_POWERLEVEL_NOMINAL; + table_clk_vlt->entries[2].v = 810; + table_clk_vlt->entries[3].clk = PP_DAL_POWERLEVEL_PERFORMANCE; + table_clk_vlt->entries[3].v = 900; + hwmgr->dyn_state.vddc_dep_on_dal_pwrl = table_clk_vlt; + } + + return 0; +} + +/** + * Initializes the Volcanic Islands Hardware Manager + * + * @param hwmgr the address of the powerplay hardware manager. + * @return 1 if success; otherwise appropriate error code. + */ +static int iceland_hwmgr_backend_init(struct pp_hwmgr *hwmgr) +{ + int result = 0; + SMU71_Discrete_DpmTable *table = NULL; + iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + pp_atomctrl_gpio_pin_assignment gpio_pin_assignment; + bool stay_in_boot; + struct phw_iceland_ulv_parm *ulv; + struct cgs_system_info sys_info = {0}; + + PP_ASSERT_WITH_CODE((NULL != hwmgr), + "Invalid Parameter!", return -EINVAL;); + + data->dll_defaule_on = 0; + data->sram_end = SMC_RAM_END; + + data->activity_target[0] = PPICELAND_TARGETACTIVITY_DFLT; + data->activity_target[1] = PPICELAND_TARGETACTIVITY_DFLT; + data->activity_target[2] = PPICELAND_TARGETACTIVITY_DFLT; + data->activity_target[3] = PPICELAND_TARGETACTIVITY_DFLT; + data->activity_target[4] = PPICELAND_TARGETACTIVITY_DFLT; + data->activity_target[5] = PPICELAND_TARGETACTIVITY_DFLT; + data->activity_target[6] = PPICELAND_TARGETACTIVITY_DFLT; + data->activity_target[7] = PPICELAND_TARGETACTIVITY_DFLT; + + data->mclk_activity_target = PPICELAND_MCLK_TARGETACTIVITY_DFLT; + + data->sclk_dpm_key_disabled = 0; + data->mclk_dpm_key_disabled = 0; + data->pcie_dpm_key_disabled = 0; + data->pcc_monitor_enabled = 0; + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_UnTabledHardwareInterface); + + data->gpio_debug = 0; + data->engine_clock_data = 0; + data->memory_clock_data = 0; + + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SclkDeepSleepAboveLow); + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DynamicPatchPowerState); + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_TablelessHardwareInterface); + + /* Initializes DPM default values. */ + iceland_initialize_dpm_defaults(hwmgr); + + /* Enable Platform EVV support. */ + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_EVV); + + /* Get leakage voltage based on leakage ID. */ + result = iceland_get_evv_voltage(hwmgr); + if (result) + goto failed; + + /** + * Patch our voltage dependency table with actual leakage + * voltage. We need to perform leakage translation before it's + * used by other functions such as + * iceland_set_hwmgr_variables_based_on_pptable. + */ + result = iceland_patch_dependency_tables_with_leakage(hwmgr); + if (result) + goto failed; + + /* Parse pptable data read from VBIOS. */ + result = iceland_set_private_var_based_on_pptale(hwmgr); + if (result) + goto failed; + + /* ULV support */ + ulv = &(data->ulv); + ulv->ulv_supported = 1; + + /* Initalize Dynamic State Adjustment Rule Settings*/ + result = iceland_initializa_dynamic_state_adjustment_rule_settings(hwmgr); + if (result) { + pr_err("[ powerplay ] iceland_initializa_dynamic_state_adjustment_rule_settings failed!\n"); + goto failed; + } + + data->voltage_control = ICELAND_VOLTAGE_CONTROL_NONE; + data->vdd_ci_control = ICELAND_VOLTAGE_CONTROL_NONE; + data->mvdd_control = ICELAND_VOLTAGE_CONTROL_NONE; + + /* + * Hardcode thermal temperature settings for now, these will + * be overwritten if a custom policy exists. + */ + data->thermal_temp_setting.temperature_low = 99500; + data->thermal_temp_setting.temperature_high = 100000; + data->thermal_temp_setting.temperature_shutdown = 104000; + data->uvd_enabled = false; + + table = &data->smc_state_table; + + if (atomctrl_get_pp_assign_pin(hwmgr, VDDC_VRHOT_GPIO_PINID, + &gpio_pin_assignment)) { + table->VRHotGpio = gpio_pin_assignment.uc_gpio_pin_bit_shift; + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_RegulatorHot); + } else { + table->VRHotGpio = ICELAND_UNUSED_GPIO_PIN; + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_RegulatorHot); + } + + if (atomctrl_get_pp_assign_pin(hwmgr, PP_AC_DC_SWITCH_GPIO_PINID, + &gpio_pin_assignment)) { + table->AcDcGpio = gpio_pin_assignment.uc_gpio_pin_bit_shift; + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_AutomaticDCTransition); + } else { + table->AcDcGpio = ICELAND_UNUSED_GPIO_PIN; + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_AutomaticDCTransition); + } + + /* + * If ucGPIO_ID=VDDC_PCC_GPIO_PINID in GPIO_LUTable, Peak. + * Current Control feature is enabled and we should program + * PCC HW register + */ + if (atomctrl_get_pp_assign_pin(hwmgr, VDDC_PCC_GPIO_PINID, + &gpio_pin_assignment)) { + uint32_t temp_reg = cgs_read_ind_register(hwmgr->device, + CGS_IND_REG__SMC, + ixCNB_PWRMGT_CNTL); + + switch (gpio_pin_assignment.uc_gpio_pin_bit_shift) { + case 0: + temp_reg = PHM_SET_FIELD(temp_reg, + CNB_PWRMGT_CNTL, GNB_SLOW_MODE, 0x1); + break; + case 1: + temp_reg = PHM_SET_FIELD(temp_reg, + CNB_PWRMGT_CNTL, GNB_SLOW_MODE, 0x2); + break; + case 2: + temp_reg = PHM_SET_FIELD(temp_reg, + CNB_PWRMGT_CNTL, GNB_SLOW, 0x1); + break; + case 3: + temp_reg = PHM_SET_FIELD(temp_reg, + CNB_PWRMGT_CNTL, FORCE_NB_PS1, 0x1); + break; + case 4: + temp_reg = PHM_SET_FIELD(temp_reg, + CNB_PWRMGT_CNTL, DPM_ENABLED, 0x1); + break; + default: + pr_warning("[ powerplay ] Failed to setup PCC HW register! Wrong GPIO assigned for VDDC_PCC_GPIO_PINID!\n"); + break; + } + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCNB_PWRMGT_CNTL, temp_reg); + } + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_EnableSMU7ThermalManagement); + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SMU7); + + if (atomctrl_is_voltage_controled_by_gpio_v3(hwmgr, + VOLTAGE_TYPE_VDDC, + VOLTAGE_OBJ_GPIO_LUT)) + data->voltage_control = ICELAND_VOLTAGE_CONTROL_BY_GPIO; + else if (atomctrl_is_voltage_controled_by_gpio_v3(hwmgr, + VOLTAGE_TYPE_VDDC, + VOLTAGE_OBJ_SVID2)) + data->voltage_control = ICELAND_VOLTAGE_CONTROL_BY_SVID2; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_ControlVDDCI)) { + if (atomctrl_is_voltage_controled_by_gpio_v3(hwmgr, + VOLTAGE_TYPE_VDDCI, + VOLTAGE_OBJ_GPIO_LUT)) + data->vdd_ci_control = ICELAND_VOLTAGE_CONTROL_BY_GPIO; + else if (atomctrl_is_voltage_controled_by_gpio_v3(hwmgr, + VOLTAGE_TYPE_VDDCI, + VOLTAGE_OBJ_SVID2)) + data->vdd_ci_control = ICELAND_VOLTAGE_CONTROL_BY_SVID2; + } + + if (data->vdd_ci_control == ICELAND_VOLTAGE_CONTROL_NONE) + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_ControlVDDCI); + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_EnableMVDDControl)) { + if (atomctrl_is_voltage_controled_by_gpio_v3(hwmgr, + VOLTAGE_TYPE_MVDDC, + VOLTAGE_OBJ_GPIO_LUT)) + data->mvdd_control = ICELAND_VOLTAGE_CONTROL_BY_GPIO; + else if (atomctrl_is_voltage_controled_by_gpio_v3(hwmgr, + VOLTAGE_TYPE_MVDDC, + VOLTAGE_OBJ_SVID2)) + data->mvdd_control = ICELAND_VOLTAGE_CONTROL_BY_SVID2; + } + + if (data->mvdd_control == ICELAND_VOLTAGE_CONTROL_NONE) + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_EnableMVDDControl); + + data->vddc_phase_shed_control = false; + + stay_in_boot = phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_StayInBootState); + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DynamicPowerManagement); + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_ActivityReporting); + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_GFXClockGatingSupport); + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_MemorySpreadSpectrumSupport); + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_EngineSpreadSpectrumSupport); + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DynamicPCIEGen2Support); + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SMC); + + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DisablePowerGating); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_BACO); + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_ThermalAutoThrottling); + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DisableLSClockGating); + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SamuDPM); + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_AcpDPM); + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_OD6inACSupport); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_EnablePlatformPowerManagement); + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PauseMMSessions); + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_OD6PlusinACSupport); + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PauseMMSessions); + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_GFXClockGatingManagedInCAIL); + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_IcelandULPSSWWorkAround); + + + /* iceland doesn't support UVD and VCE */ + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_UVDPowerGating); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_VCEPowerGating); + + sys_info.size = sizeof(struct cgs_system_info); + sys_info.info_id = CGS_SYSTEM_INFO_PG_FLAGS; + result = cgs_query_system_info(hwmgr->device, &sys_info); + if (!result) { + if (sys_info.value & AMD_PG_SUPPORT_UVD) + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_UVDPowerGating); + if (sys_info.value & AMD_PG_SUPPORT_VCE) + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_VCEPowerGating); + + data->is_tlu_enabled = false; + hwmgr->platform_descriptor.hardwareActivityPerformanceLevels = + ICELAND_MAX_HARDWARE_POWERLEVELS; + hwmgr->platform_descriptor.hardwarePerformanceLevels = 2; + hwmgr->platform_descriptor.minimumClocksReductionPercentage = 50; + + sys_info.size = sizeof(struct cgs_system_info); + sys_info.info_id = CGS_SYSTEM_INFO_PCIE_GEN_INFO; + result = cgs_query_system_info(hwmgr->device, &sys_info); + if (result) + data->pcie_gen_cap = AMDGPU_DEFAULT_PCIE_GEN_MASK; + else + data->pcie_gen_cap = (uint32_t)sys_info.value; + if (data->pcie_gen_cap & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3) + data->pcie_spc_cap = 20; + sys_info.size = sizeof(struct cgs_system_info); + sys_info.info_id = CGS_SYSTEM_INFO_PCIE_MLW; + result = cgs_query_system_info(hwmgr->device, &sys_info); + if (result) + data->pcie_lane_cap = AMDGPU_DEFAULT_PCIE_MLW_MASK; + else + data->pcie_lane_cap = (uint32_t)sys_info.value; + } else { + /* Ignore return value in here, we are cleaning up a mess. */ + iceland_hwmgr_backend_fini(hwmgr); + } + + return 0; +failed: + return result; +} + +static int iceland_get_num_of_entries(struct pp_hwmgr *hwmgr) +{ + int result; + unsigned long ret = 0; + + result = pp_tables_get_num_of_entries(hwmgr, &ret); + + return result ? 0 : ret; +} + +static const unsigned long PhwIceland_Magic = (unsigned long)(PHM_VIslands_Magic); + +struct iceland_power_state *cast_phw_iceland_power_state( + struct pp_hw_power_state *hw_ps) +{ + if (hw_ps == NULL) + return NULL; + + PP_ASSERT_WITH_CODE((PhwIceland_Magic == hw_ps->magic), + "Invalid Powerstate Type!", + return NULL); + + return (struct iceland_power_state *)hw_ps; +} + +static int iceland_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, + struct pp_power_state *prequest_ps, + const struct pp_power_state *pcurrent_ps) +{ + struct iceland_power_state *iceland_ps = + cast_phw_iceland_power_state(&prequest_ps->hardware); + + uint32_t sclk; + uint32_t mclk; + struct PP_Clocks minimum_clocks = {0}; + bool disable_mclk_switching; + bool disable_mclk_switching_for_frame_lock; + struct cgs_display_info info = {0}; + const struct phm_clock_and_voltage_limits *max_limits; + uint32_t i; + iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + int32_t count; + int32_t stable_pstate_sclk = 0, stable_pstate_mclk = 0; + + data->battery_state = (PP_StateUILabel_Battery == prequest_ps->classification.ui_label); + + PP_ASSERT_WITH_CODE(iceland_ps->performance_level_count == 2, + "VI should always have 2 performance levels", + ); + + max_limits = (PP_PowerSource_AC == hwmgr->power_source) ? + &(hwmgr->dyn_state.max_clock_voltage_on_ac) : + &(hwmgr->dyn_state.max_clock_voltage_on_dc); + + if (PP_PowerSource_DC == hwmgr->power_source) { + for (i = 0; i < iceland_ps->performance_level_count; i++) { + if (iceland_ps->performance_levels[i].memory_clock > max_limits->mclk) + iceland_ps->performance_levels[i].memory_clock = max_limits->mclk; + if (iceland_ps->performance_levels[i].engine_clock > max_limits->sclk) + iceland_ps->performance_levels[i].engine_clock = max_limits->sclk; + } + } + + iceland_ps->vce_clocks.EVCLK = hwmgr->vce_arbiter.evclk; + iceland_ps->vce_clocks.ECCLK = hwmgr->vce_arbiter.ecclk; + + cgs_get_active_displays_info(hwmgr->device, &info); + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_StablePState)) { + + max_limits = &(hwmgr->dyn_state.max_clock_voltage_on_ac); + stable_pstate_sclk = (max_limits->sclk * 75) / 100; + + for (count = hwmgr->dyn_state.vddc_dependency_on_sclk->count-1; count >= 0; count--) { + if (stable_pstate_sclk >= hwmgr->dyn_state.vddc_dependency_on_sclk->entries[count].clk) { + stable_pstate_sclk = hwmgr->dyn_state.vddc_dependency_on_sclk->entries[count].clk; + break; + } + } + + if (count < 0) + stable_pstate_sclk = hwmgr->dyn_state.vddc_dependency_on_sclk->entries[0].clk; + + stable_pstate_mclk = max_limits->mclk; + + minimum_clocks.engineClock = stable_pstate_sclk; + minimum_clocks.memoryClock = stable_pstate_mclk; + } + + if (minimum_clocks.engineClock < hwmgr->gfx_arbiter.sclk) + minimum_clocks.engineClock = hwmgr->gfx_arbiter.sclk; + + if (minimum_clocks.memoryClock < hwmgr->gfx_arbiter.mclk) + minimum_clocks.memoryClock = hwmgr->gfx_arbiter.mclk; + + iceland_ps->sclk_threshold = hwmgr->gfx_arbiter.sclk_threshold; + + if (0 != hwmgr->gfx_arbiter.sclk_over_drive) { + PP_ASSERT_WITH_CODE((hwmgr->gfx_arbiter.sclk_over_drive <= hwmgr->platform_descriptor.overdriveLimit.engineClock), + "Overdrive sclk exceeds limit", + hwmgr->gfx_arbiter.sclk_over_drive = hwmgr->platform_descriptor.overdriveLimit.engineClock); + + if (hwmgr->gfx_arbiter.sclk_over_drive >= hwmgr->gfx_arbiter.sclk) + iceland_ps->performance_levels[1].engine_clock = hwmgr->gfx_arbiter.sclk_over_drive; + } + + if (0 != hwmgr->gfx_arbiter.mclk_over_drive) { + PP_ASSERT_WITH_CODE((hwmgr->gfx_arbiter.mclk_over_drive <= hwmgr->platform_descriptor.overdriveLimit.memoryClock), + "Overdrive mclk exceeds limit", + hwmgr->gfx_arbiter.mclk_over_drive = hwmgr->platform_descriptor.overdriveLimit.memoryClock); + + if (hwmgr->gfx_arbiter.mclk_over_drive >= hwmgr->gfx_arbiter.mclk) + iceland_ps->performance_levels[1].memory_clock = hwmgr->gfx_arbiter.mclk_over_drive; + } + + disable_mclk_switching_for_frame_lock = phm_cap_enabled( + hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DisableMclkSwitchingForFrameLock); + + disable_mclk_switching = (1 < info.display_count) || + disable_mclk_switching_for_frame_lock; + + sclk = iceland_ps->performance_levels[0].engine_clock; + mclk = iceland_ps->performance_levels[0].memory_clock; + + if (disable_mclk_switching) + mclk = iceland_ps->performance_levels[iceland_ps->performance_level_count - 1].memory_clock; + + if (sclk < minimum_clocks.engineClock) + sclk = (minimum_clocks.engineClock > max_limits->sclk) ? max_limits->sclk : minimum_clocks.engineClock; + + if (mclk < minimum_clocks.memoryClock) + mclk = (minimum_clocks.memoryClock > max_limits->mclk) ? max_limits->mclk : minimum_clocks.memoryClock; + + iceland_ps->performance_levels[0].engine_clock = sclk; + iceland_ps->performance_levels[0].memory_clock = mclk; + + iceland_ps->performance_levels[1].engine_clock = + (iceland_ps->performance_levels[1].engine_clock >= iceland_ps->performance_levels[0].engine_clock) ? + iceland_ps->performance_levels[1].engine_clock : + iceland_ps->performance_levels[0].engine_clock; + + if (disable_mclk_switching) { + if (mclk < iceland_ps->performance_levels[1].memory_clock) + mclk = iceland_ps->performance_levels[1].memory_clock; + + iceland_ps->performance_levels[0].memory_clock = mclk; + iceland_ps->performance_levels[1].memory_clock = mclk; + } else { + if (iceland_ps->performance_levels[1].memory_clock < iceland_ps->performance_levels[0].memory_clock) + iceland_ps->performance_levels[1].memory_clock = iceland_ps->performance_levels[0].memory_clock; + } + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_StablePState)) { + for (i=0; i < iceland_ps->performance_level_count; i++) { + iceland_ps->performance_levels[i].engine_clock = stable_pstate_sclk; + iceland_ps->performance_levels[i].memory_clock = stable_pstate_mclk; + iceland_ps->performance_levels[i].pcie_gen = data->pcie_gen_performance.max; + iceland_ps->performance_levels[i].pcie_lane = data->pcie_gen_performance.max; + } + } + + return 0; +} + +static bool iceland_is_dpm_running(struct pp_hwmgr *hwmgr) +{ + /* + * We return the status of Voltage Control instead of checking SCLK/MCLK DPM + * because we may have test scenarios that need us intentionly disable SCLK/MCLK DPM, + * whereas voltage control is a fundemental change that will not be disabled + */ + return (0 == PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + FEATURE_STATUS, VOLTAGE_CONTROLLER_ON) ? 1 : 0); +} + +/** + * force DPM power State + * + * @param hwmgr: the address of the powerplay hardware manager. + * @param n : DPM level + * @return The response that came from the SMC. + */ +int iceland_dpm_force_state(struct pp_hwmgr *hwmgr, uint32_t n) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + /* Checking if DPM is running. If we discover hang because of this, we should skip this message. */ + PP_ASSERT_WITH_CODE(0 == iceland_is_dpm_running(hwmgr), + "Trying to force SCLK when DPM is disabled", return -1;); + if (0 == data->sclk_dpm_key_disabled) + return (0 == smum_send_msg_to_smc_with_parameter( + hwmgr->smumgr, + PPSMC_MSG_DPM_ForceState, + n) ? 0 : 1); + + return 0; +} + +/** + * force DPM power State + * + * @param hwmgr: the address of the powerplay hardware manager. + * @param n : DPM level + * @return The response that came from the SMC. + */ +int iceland_dpm_force_state_mclk(struct pp_hwmgr *hwmgr, uint32_t n) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + /* Checking if DPM is running. If we discover hang because of this, we should skip this message. */ + PP_ASSERT_WITH_CODE(0 == iceland_is_dpm_running(hwmgr), + "Trying to Force MCLK when DPM is disabled", return -1;); + if (0 == data->mclk_dpm_key_disabled) + return (0 == smum_send_msg_to_smc_with_parameter( + hwmgr->smumgr, + PPSMC_MSG_MCLKDPM_ForceState, + n) ? 0 : 1); + + return 0; +} + +/** + * force DPM power State + * + * @param hwmgr: the address of the powerplay hardware manager. + * @param n : DPM level + * @return The response that came from the SMC. + */ +int iceland_dpm_force_state_pcie(struct pp_hwmgr *hwmgr, uint32_t n) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + /* Checking if DPM is running. If we discover hang because of this, we should skip this message.*/ + PP_ASSERT_WITH_CODE(0 == iceland_is_dpm_running(hwmgr), + "Trying to Force PCIE level when DPM is disabled", return -1;); + if (0 == data->pcie_dpm_key_disabled) + return (0 == smum_send_msg_to_smc_with_parameter( + hwmgr->smumgr, + PPSMC_MSG_PCIeDPM_ForceLevel, + n) ? 0 : 1); + + return 0; +} + +static int iceland_force_dpm_highest(struct pp_hwmgr *hwmgr) +{ + uint32_t level, tmp; + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + if (0 == data->sclk_dpm_key_disabled) { + /* SCLK */ + if (data->dpm_level_enable_mask.sclk_dpm_enable_mask != 0) { + level = 0; + tmp = data->dpm_level_enable_mask.sclk_dpm_enable_mask; + while (tmp >>= 1) + level++ ; + + if (0 != level) { + PP_ASSERT_WITH_CODE((0 == iceland_dpm_force_state(hwmgr, level)), + "force highest sclk dpm state failed!", return -1); + PHM_WAIT_INDIRECT_FIELD(hwmgr->device, + SMC_IND, TARGET_AND_CURRENT_PROFILE_INDEX, CURR_SCLK_INDEX, level); + } + } + } + + if (0 == data->mclk_dpm_key_disabled) { + /* MCLK */ + if (data->dpm_level_enable_mask.mclk_dpm_enable_mask != 0) { + level = 0; + tmp = data->dpm_level_enable_mask.mclk_dpm_enable_mask; + while (tmp >>= 1) + level++ ; + + if (0 != level) { + PP_ASSERT_WITH_CODE((0 == iceland_dpm_force_state_mclk(hwmgr, level)), + "force highest mclk dpm state failed!", return -1); + PHM_WAIT_INDIRECT_FIELD(hwmgr->device, SMC_IND, + TARGET_AND_CURRENT_PROFILE_INDEX, CURR_MCLK_INDEX, level); + } + } + } + + if (0 == data->pcie_dpm_key_disabled) { + /* PCIE */ + if (data->dpm_level_enable_mask.pcie_dpm_enable_mask != 0) { + level = 0; + tmp = data->dpm_level_enable_mask.pcie_dpm_enable_mask; + while (tmp >>= 1) + level++ ; + + if (0 != level) { + PP_ASSERT_WITH_CODE((0 == iceland_dpm_force_state_pcie(hwmgr, level)), + "force highest pcie dpm state failed!", return -1); + } + } + } + + return 0; +} + +static uint32_t iceland_get_lowest_enable_level(struct pp_hwmgr *hwmgr, + uint32_t level_mask) +{ + uint32_t level = 0; + + while (0 == (level_mask & (1 << level))) + level++; + + return level; +} + +static int iceland_force_dpm_lowest(struct pp_hwmgr *hwmgr) +{ + uint32_t level; + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + /* for now force only sclk */ + if (0 != data->dpm_level_enable_mask.sclk_dpm_enable_mask) { + level = iceland_get_lowest_enable_level(hwmgr, + data->dpm_level_enable_mask.sclk_dpm_enable_mask); + + PP_ASSERT_WITH_CODE((0 == iceland_dpm_force_state(hwmgr, level)), + "force sclk dpm state failed!", return -1); + + PHM_WAIT_INDIRECT_FIELD(hwmgr->device, SMC_IND, + TARGET_AND_CURRENT_PROFILE_INDEX, + CURR_SCLK_INDEX, + level); + } + + return 0; +} + +int iceland_unforce_dpm_levels(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + PP_ASSERT_WITH_CODE (0 == iceland_is_dpm_running(hwmgr), + "Trying to Unforce DPM when DPM is disabled. Returning without sending SMC message.", + return -1); + + if (0 == data->sclk_dpm_key_disabled) { + PP_ASSERT_WITH_CODE((0 == smum_send_msg_to_smc( + hwmgr->smumgr, + PPSMC_MSG_NoForcedLevel)), + "unforce sclk dpm state failed!", + return -1); + } + + if (0 == data->mclk_dpm_key_disabled) { + PP_ASSERT_WITH_CODE((0 == smum_send_msg_to_smc( + hwmgr->smumgr, + PPSMC_MSG_MCLKDPM_NoForcedLevel)), + "unforce mclk dpm state failed!", + return -1); + } + + if (0 == data->pcie_dpm_key_disabled) { + PP_ASSERT_WITH_CODE((0 == smum_send_msg_to_smc( + hwmgr->smumgr, + PPSMC_MSG_PCIeDPM_UnForceLevel)), + "unforce pcie level failed!", + return -1); + } + + return 0; +} + +static int iceland_force_dpm_level(struct pp_hwmgr *hwmgr, + enum amd_dpm_forced_level level) +{ + int ret = 0; + + switch (level) { + case AMD_DPM_FORCED_LEVEL_HIGH: + ret = iceland_force_dpm_highest(hwmgr); + if (ret) + return ret; + break; + case AMD_DPM_FORCED_LEVEL_LOW: + ret = iceland_force_dpm_lowest(hwmgr); + if (ret) + return ret; + break; + case AMD_DPM_FORCED_LEVEL_AUTO: + ret = iceland_unforce_dpm_levels(hwmgr); + if (ret) + return ret; + break; + default: + break; + } + + hwmgr->dpm_level = level; + return ret; +} + +const struct iceland_power_state *cast_const_phw_iceland_power_state( + const struct pp_hw_power_state *hw_ps) +{ + if (hw_ps == NULL) + return NULL; + + PP_ASSERT_WITH_CODE((PhwIceland_Magic == hw_ps->magic), + "Invalid Powerstate Type!", + return NULL); + + return (const struct iceland_power_state *)hw_ps; +} + +static int iceland_find_dpm_states_clocks_in_dpm_table(struct pp_hwmgr *hwmgr, const void *input) +{ + const struct phm_set_power_state_input *states = (const struct phm_set_power_state_input *)input; + const struct iceland_power_state *iceland_ps = cast_const_phw_iceland_power_state(states->pnew_state); + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + struct iceland_single_dpm_table *psclk_table = &(data->dpm_table.sclk_table); + uint32_t sclk = iceland_ps->performance_levels[iceland_ps->performance_level_count-1].engine_clock; + struct iceland_single_dpm_table *pmclk_table = &(data->dpm_table.mclk_table); + uint32_t mclk = iceland_ps->performance_levels[iceland_ps->performance_level_count-1].memory_clock; + struct PP_Clocks min_clocks = {0}; + uint32_t i; + struct cgs_display_info info = {0}; + + data->need_update_smu7_dpm_table = 0; + + for (i = 0; i < psclk_table->count; i++) { + if (sclk == psclk_table->dpm_levels[i].value) + break; + } + + if (i >= psclk_table->count) + data->need_update_smu7_dpm_table |= DPMTABLE_OD_UPDATE_SCLK; + else { + /* + * TODO: Check SCLK in DAL's minimum clocks in case DeepSleep + * divider update is required. + */ + if(data->display_timing.min_clock_insr != min_clocks.engineClockInSR) + data->need_update_smu7_dpm_table |= DPMTABLE_UPDATE_SCLK; + } + + for (i = 0; i < pmclk_table->count; i++) { + if (mclk == pmclk_table->dpm_levels[i].value) + break; + } + + if (i >= pmclk_table->count) + data->need_update_smu7_dpm_table |= DPMTABLE_OD_UPDATE_MCLK; + + cgs_get_active_displays_info(hwmgr->device, &info); + + if (data->display_timing.num_existing_displays != info.display_count) + data->need_update_smu7_dpm_table |= DPMTABLE_UPDATE_MCLK; + + return 0; +} + +static uint16_t iceland_get_maximum_link_speed(struct pp_hwmgr *hwmgr, const struct iceland_power_state *hw_ps) +{ + uint32_t i; + uint32_t pcie_speed, max_speed = 0; + + for (i = 0; i < hw_ps->performance_level_count; i++) { + pcie_speed = hw_ps->performance_levels[i].pcie_gen; + if (max_speed < pcie_speed) + max_speed = pcie_speed; + } + + return max_speed; +} + +static uint16_t iceland_get_current_pcie_speed(struct pp_hwmgr *hwmgr) +{ + uint32_t speed_cntl = 0; + + speed_cntl = cgs_read_ind_register(hwmgr->device, + CGS_IND_REG__PCIE, + ixPCIE_LC_SPEED_CNTL); + return((uint16_t)PHM_GET_FIELD(speed_cntl, + PCIE_LC_SPEED_CNTL, LC_CURRENT_DATA_RATE)); +} + + +static int iceland_request_link_speed_change_before_state_change(struct pp_hwmgr *hwmgr, const void *input) +{ + const struct phm_set_power_state_input *states = (const struct phm_set_power_state_input *)input; + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + const struct iceland_power_state *iceland_nps = cast_const_phw_iceland_power_state(states->pnew_state); + const struct iceland_power_state *iceland_cps = cast_const_phw_iceland_power_state(states->pcurrent_state); + + uint16_t target_link_speed = iceland_get_maximum_link_speed(hwmgr, iceland_nps); + uint16_t current_link_speed; + + if (data->force_pcie_gen == PP_PCIEGenInvalid) + current_link_speed = iceland_get_maximum_link_speed(hwmgr, iceland_cps); + else + current_link_speed = data->force_pcie_gen; + + data->force_pcie_gen = PP_PCIEGenInvalid; + data->pspp_notify_required = false; + if (target_link_speed > current_link_speed) { + switch(target_link_speed) { + case PP_PCIEGen3: + if (0 == acpi_pcie_perf_request(hwmgr->device, PCIE_PERF_REQ_GEN3, false)) + break; + data->force_pcie_gen = PP_PCIEGen2; + if (current_link_speed == PP_PCIEGen2) + break; + case PP_PCIEGen2: + if (0 == acpi_pcie_perf_request(hwmgr->device, PCIE_PERF_REQ_GEN2, false)) + break; + default: + data->force_pcie_gen = iceland_get_current_pcie_speed(hwmgr); + break; + } + } else { + if (target_link_speed < current_link_speed) + data->pspp_notify_required = true; + } + + return 0; +} + +static int iceland_freeze_sclk_mclk_dpm(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + if (0 == data->need_update_smu7_dpm_table) + return 0; + + if ((0 == data->sclk_dpm_key_disabled) && + (data->need_update_smu7_dpm_table & + (DPMTABLE_OD_UPDATE_SCLK + DPMTABLE_UPDATE_SCLK))) { + PP_ASSERT_WITH_CODE( + 0 == iceland_is_dpm_running(hwmgr), + "Trying to freeze SCLK DPM when DPM is disabled", + ); + PP_ASSERT_WITH_CODE( + 0 == smum_send_msg_to_smc(hwmgr->smumgr, + PPSMC_MSG_SCLKDPM_FreezeLevel), + "Failed to freeze SCLK DPM during FreezeSclkMclkDPM Function!", + return -1); + } + + if ((0 == data->mclk_dpm_key_disabled) && + (data->need_update_smu7_dpm_table & + DPMTABLE_OD_UPDATE_MCLK)) { + PP_ASSERT_WITH_CODE(0 == iceland_is_dpm_running(hwmgr), + "Trying to freeze MCLK DPM when DPM is disabled", + ); + PP_ASSERT_WITH_CODE( + 0 == smum_send_msg_to_smc(hwmgr->smumgr, + PPSMC_MSG_MCLKDPM_FreezeLevel), + "Failed to freeze MCLK DPM during FreezeSclkMclkDPM Function!", + return -1); + } + + return 0; +} + +static int iceland_populate_and_upload_sclk_mclk_dpm_levels(struct pp_hwmgr *hwmgr, const void *input) +{ + int result = 0; + + const struct phm_set_power_state_input *states = (const struct phm_set_power_state_input *)input; + const struct iceland_power_state *iceland_ps = cast_const_phw_iceland_power_state(states->pnew_state); + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + uint32_t sclk = iceland_ps->performance_levels[iceland_ps->performance_level_count-1].engine_clock; + uint32_t mclk = iceland_ps->performance_levels[iceland_ps->performance_level_count-1].memory_clock; + struct iceland_dpm_table *pdpm_table = &data->dpm_table; + + struct iceland_dpm_table *pgolden_dpm_table = &data->golden_dpm_table; + uint32_t dpm_count, clock_percent; + uint32_t i; + + if (0 == data->need_update_smu7_dpm_table) + return 0; + + if (data->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_SCLK) { + pdpm_table->sclk_table.dpm_levels[pdpm_table->sclk_table.count-1].value = sclk; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_OD6PlusinACSupport) || + phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_OD6PlusinDCSupport)) { + /* + * Need to do calculation based on the golden DPM table + * as the Heatmap GPU Clock axis is also based on the default values + */ + PP_ASSERT_WITH_CODE( + (pgolden_dpm_table->sclk_table.dpm_levels[pgolden_dpm_table->sclk_table.count-1].value != 0), + "Divide by 0!", + return -1); + dpm_count = pdpm_table->sclk_table.count < 2 ? 0 : pdpm_table->sclk_table.count-2; + for (i = dpm_count; i > 1; i--) { + if (sclk > pgolden_dpm_table->sclk_table.dpm_levels[pgolden_dpm_table->sclk_table.count-1].value) { + clock_percent = ((sclk - pgolden_dpm_table->sclk_table.dpm_levels[pgolden_dpm_table->sclk_table.count-1].value)*100) / + pgolden_dpm_table->sclk_table.dpm_levels[pgolden_dpm_table->sclk_table.count-1].value; + + pdpm_table->sclk_table.dpm_levels[i].value = + pgolden_dpm_table->sclk_table.dpm_levels[i].value + + (pgolden_dpm_table->sclk_table.dpm_levels[i].value * clock_percent)/100; + + } else if (pgolden_dpm_table->sclk_table.dpm_levels[pdpm_table->sclk_table.count-1].value > sclk) { + clock_percent = ((pgolden_dpm_table->sclk_table.dpm_levels[pgolden_dpm_table->sclk_table.count-1].value - sclk)*100) / + pgolden_dpm_table->sclk_table.dpm_levels[pgolden_dpm_table->sclk_table.count-1].value; + + pdpm_table->sclk_table.dpm_levels[i].value = + pgolden_dpm_table->sclk_table.dpm_levels[i].value - + (pgolden_dpm_table->sclk_table.dpm_levels[i].value * clock_percent)/100; + } else + pdpm_table->sclk_table.dpm_levels[i].value = + pgolden_dpm_table->sclk_table.dpm_levels[i].value; + } + } + } + + if (data->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_MCLK) { + pdpm_table->mclk_table.dpm_levels[pdpm_table->mclk_table.count-1].value = mclk; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_OD6PlusinACSupport) || + phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_OD6PlusinDCSupport)) { + + PP_ASSERT_WITH_CODE( + (pgolden_dpm_table->mclk_table.dpm_levels[pgolden_dpm_table->mclk_table.count-1].value != 0), + "Divide by 0!", + return -1); + dpm_count = pdpm_table->mclk_table.count < 2? 0 : pdpm_table->mclk_table.count-2; + for (i = dpm_count; i > 1; i--) { + if (mclk > pgolden_dpm_table->mclk_table.dpm_levels[pgolden_dpm_table->mclk_table.count-1].value) { + clock_percent = ((mclk - pgolden_dpm_table->mclk_table.dpm_levels[pgolden_dpm_table->mclk_table.count-1].value)*100) / + pgolden_dpm_table->mclk_table.dpm_levels[pgolden_dpm_table->mclk_table.count-1].value; + + pdpm_table->mclk_table.dpm_levels[i].value = + pgolden_dpm_table->mclk_table.dpm_levels[i].value + + (pgolden_dpm_table->mclk_table.dpm_levels[i].value * clock_percent)/100; + + } else if (pgolden_dpm_table->mclk_table.dpm_levels[pdpm_table->mclk_table.count-1].value > mclk) { + clock_percent = ((pgolden_dpm_table->mclk_table.dpm_levels[pgolden_dpm_table->mclk_table.count-1].value - mclk)*100) / + pgolden_dpm_table->mclk_table.dpm_levels[pgolden_dpm_table->mclk_table.count-1].value; + + pdpm_table->mclk_table.dpm_levels[i].value = + pgolden_dpm_table->mclk_table.dpm_levels[i].value - + (pgolden_dpm_table->mclk_table.dpm_levels[i].value * clock_percent)/100; + } else + pdpm_table->mclk_table.dpm_levels[i].value = pgolden_dpm_table->mclk_table.dpm_levels[i].value; + } + } + } + + + if (data->need_update_smu7_dpm_table & (DPMTABLE_OD_UPDATE_SCLK + DPMTABLE_UPDATE_SCLK)) { + result = iceland_populate_all_graphic_levels(hwmgr); + PP_ASSERT_WITH_CODE((0 == result), + "Failed to populate SCLK during PopulateNewDPMClocksStates Function!", + return result); + } + + if (data->need_update_smu7_dpm_table & (DPMTABLE_OD_UPDATE_MCLK + DPMTABLE_UPDATE_MCLK)) { + /*populate MCLK dpm table to SMU7 */ + result = iceland_populate_all_memory_levels(hwmgr); + PP_ASSERT_WITH_CODE((0 == result), + "Failed to populate MCLK during PopulateNewDPMClocksStates Function!", + return result); + } + + return result; +} + +static int iceland_trim_single_dpm_states(struct pp_hwmgr *hwmgr, + struct iceland_single_dpm_table *pdpm_table, + uint32_t low_limit, uint32_t high_limit) +{ + uint32_t i; + + for (i = 0; i < pdpm_table->count; i++) { + if ((pdpm_table->dpm_levels[i].value < low_limit) || + (pdpm_table->dpm_levels[i].value > high_limit)) + pdpm_table->dpm_levels[i].enabled = false; + else + pdpm_table->dpm_levels[i].enabled = true; + } + return 0; +} + +static int iceland_trim_dpm_states(struct pp_hwmgr *hwmgr, const struct iceland_power_state *hw_state) +{ + int result = 0; + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + uint32_t high_limit_count; + + PP_ASSERT_WITH_CODE((hw_state->performance_level_count >= 1), + "power state did not have any performance level", + return -1); + + high_limit_count = (1 == hw_state->performance_level_count) ? 0: 1; + + iceland_trim_single_dpm_states(hwmgr, &(data->dpm_table.sclk_table), + hw_state->performance_levels[0].engine_clock, + hw_state->performance_levels[high_limit_count].engine_clock); + + iceland_trim_single_dpm_states(hwmgr, &(data->dpm_table.mclk_table), + hw_state->performance_levels[0].memory_clock, + hw_state->performance_levels[high_limit_count].memory_clock); + + return result; +} + +static int iceland_generate_dpm_level_enable_mask(struct pp_hwmgr *hwmgr, const void *input) +{ + int result; + const struct phm_set_power_state_input *states = (const struct phm_set_power_state_input *)input; + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + const struct iceland_power_state *iceland_ps = cast_const_phw_iceland_power_state(states->pnew_state); + + result = iceland_trim_dpm_states(hwmgr, iceland_ps); + if (0 != result) + return result; + + data->dpm_level_enable_mask.sclk_dpm_enable_mask = iceland_get_dpm_level_enable_mask_value(&data->dpm_table.sclk_table); + data->dpm_level_enable_mask.mclk_dpm_enable_mask = iceland_get_dpm_level_enable_mask_value(&data->dpm_table.mclk_table); + data->last_mclk_dpm_enable_mask = data->dpm_level_enable_mask.mclk_dpm_enable_mask; + if (data->uvd_enabled && (data->dpm_level_enable_mask.mclk_dpm_enable_mask & 1)) + data->dpm_level_enable_mask.mclk_dpm_enable_mask &= 0xFFFFFFFE; + + data->dpm_level_enable_mask.pcie_dpm_enable_mask = iceland_get_dpm_level_enable_mask_value(&data->dpm_table.pcie_speed_table); + + return 0; +} + +static int iceland_update_vce_dpm(struct pp_hwmgr *hwmgr, const void *input) +{ + return 0; +} + +int iceland_update_sclk_threshold(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + int result = 0; + uint32_t low_sclk_interrupt_threshold = 0; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SclkThrottleLowNotification) + && (hwmgr->gfx_arbiter.sclk_threshold != data->low_sclk_interrupt_threshold)) { + data->low_sclk_interrupt_threshold = hwmgr->gfx_arbiter.sclk_threshold; + low_sclk_interrupt_threshold = data->low_sclk_interrupt_threshold; + + CONVERT_FROM_HOST_TO_SMC_UL(low_sclk_interrupt_threshold); + + result = iceland_copy_bytes_to_smc( + hwmgr->smumgr, + data->dpm_table_start + offsetof(SMU71_Discrete_DpmTable, + LowSclkInterruptThreshold), + (uint8_t *)&low_sclk_interrupt_threshold, + sizeof(uint32_t), + data->sram_end + ); + } + + return result; +} + +static int iceland_update_and_upload_mc_reg_table(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + uint32_t address; + int32_t result; + + if (0 == (data->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_MCLK)) + return 0; + + + memset(&data->mc_reg_table, 0, sizeof(SMU71_Discrete_MCRegisters)); + + result = iceland_convert_mc_reg_table_to_smc(hwmgr, &(data->mc_reg_table)); + + if(result != 0) + return result; + + + address = data->mc_reg_table_start + (uint32_t)offsetof(SMU71_Discrete_MCRegisters, data[0]); + + return iceland_copy_bytes_to_smc(hwmgr->smumgr, address, + (uint8_t *)&data->mc_reg_table.data[0], + sizeof(SMU71_Discrete_MCRegisterSet) * data->dpm_table.mclk_table.count, + data->sram_end); +} + +static int iceland_program_memory_timing_parameters_conditionally(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + if (data->need_update_smu7_dpm_table & + (DPMTABLE_OD_UPDATE_SCLK + DPMTABLE_OD_UPDATE_MCLK)) + return iceland_program_memory_timing_parameters(hwmgr); + + return 0; +} + +static int iceland_unfreeze_sclk_mclk_dpm(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + if (0 == data->need_update_smu7_dpm_table) + return 0; + + if ((0 == data->sclk_dpm_key_disabled) && + (data->need_update_smu7_dpm_table & + (DPMTABLE_OD_UPDATE_SCLK + DPMTABLE_UPDATE_SCLK))) { + + PP_ASSERT_WITH_CODE(0 == iceland_is_dpm_running(hwmgr), + "Trying to Unfreeze SCLK DPM when DPM is disabled", + ); + PP_ASSERT_WITH_CODE( + 0 == smum_send_msg_to_smc(hwmgr->smumgr, + PPSMC_MSG_SCLKDPM_UnfreezeLevel), + "Failed to unfreeze SCLK DPM during UnFreezeSclkMclkDPM Function!", + return -1); + } + + if ((0 == data->mclk_dpm_key_disabled) && + (data->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_MCLK)) { + + PP_ASSERT_WITH_CODE( + 0 == iceland_is_dpm_running(hwmgr), + "Trying to Unfreeze MCLK DPM when DPM is disabled", + ); + PP_ASSERT_WITH_CODE( + 0 == smum_send_msg_to_smc(hwmgr->smumgr, + PPSMC_MSG_MCLKDPM_UnfreezeLevel), + "Failed to unfreeze MCLK DPM during UnFreezeSclkMclkDPM Function!", + return -1); + } + + data->need_update_smu7_dpm_table = 0; + + return 0; +} + +static int iceland_notify_link_speed_change_after_state_change(struct pp_hwmgr *hwmgr, const void *input) +{ + const struct phm_set_power_state_input *states = (const struct phm_set_power_state_input *)input; + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + const struct iceland_power_state *iceland_ps = cast_const_phw_iceland_power_state(states->pnew_state); + uint16_t target_link_speed = iceland_get_maximum_link_speed(hwmgr, iceland_ps); + uint8_t request; + + if (data->pspp_notify_required || + data->pcie_performance_request) { + if (target_link_speed == PP_PCIEGen3) + request = PCIE_PERF_REQ_GEN3; + else if (target_link_speed == PP_PCIEGen2) + request = PCIE_PERF_REQ_GEN2; + else + request = PCIE_PERF_REQ_GEN1; + + if(request == PCIE_PERF_REQ_GEN1 && iceland_get_current_pcie_speed(hwmgr) > 0) { + data->pcie_performance_request = false; + return 0; + } + + if (0 != acpi_pcie_perf_request(hwmgr->device, request, false)) { + if (PP_PCIEGen2 == target_link_speed) + printk("PSPP request to switch to Gen2 from Gen3 Failed!"); + else + printk("PSPP request to switch to Gen1 from Gen2 Failed!"); + } + } + + data->pcie_performance_request = false; + return 0; +} + +int iceland_upload_dpm_level_enable_mask(struct pp_hwmgr *hwmgr) +{ + PPSMC_Result result; + iceland_hwmgr *data = (iceland_hwmgr *)(hwmgr->backend); + + if (0 == data->sclk_dpm_key_disabled) { + /* Checking if DPM is running. If we discover hang because of this, we should skip this message.*/ + if (0 != iceland_is_dpm_running(hwmgr)) + printk(KERN_ERR "[ powerplay ] Trying to set Enable Sclk Mask when DPM is disabled \n"); + + if (0 != data->dpm_level_enable_mask.sclk_dpm_enable_mask) { + result = smum_send_msg_to_smc_with_parameter( + hwmgr->smumgr, + (PPSMC_Msg)PPSMC_MSG_SCLKDPM_SetEnabledMask, + data->dpm_level_enable_mask.sclk_dpm_enable_mask); + PP_ASSERT_WITH_CODE((0 == result), + "Set Sclk Dpm enable Mask failed", return -1); + } + } + + if (0 == data->mclk_dpm_key_disabled) { + /* Checking if DPM is running. If we discover hang because of this, we should skip this message.*/ + if (0 != iceland_is_dpm_running(hwmgr)) + printk(KERN_ERR "[ powerplay ] Trying to set Enable Mclk Mask when DPM is disabled \n"); + + if (0 != data->dpm_level_enable_mask.mclk_dpm_enable_mask) { + result = smum_send_msg_to_smc_with_parameter( + hwmgr->smumgr, + (PPSMC_Msg)PPSMC_MSG_MCLKDPM_SetEnabledMask, + data->dpm_level_enable_mask.mclk_dpm_enable_mask); + PP_ASSERT_WITH_CODE((0 == result), + "Set Mclk Dpm enable Mask failed", return -1); + } + } + + return 0; +} + +static int iceland_set_power_state_tasks(struct pp_hwmgr *hwmgr, const void *input) +{ + int tmp_result, result = 0; + + tmp_result = iceland_find_dpm_states_clocks_in_dpm_table(hwmgr, input); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to find DPM states clocks in DPM table!", result = tmp_result); + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_PCIEPerformanceRequest)) { + tmp_result = iceland_request_link_speed_change_before_state_change(hwmgr, input); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to request link speed change before state change!", result = tmp_result); + } + + tmp_result = iceland_freeze_sclk_mclk_dpm(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to freeze SCLK MCLK DPM!", result = tmp_result); + + tmp_result = iceland_populate_and_upload_sclk_mclk_dpm_levels(hwmgr, input); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to populate and upload SCLK MCLK DPM levels!", result = tmp_result); + + tmp_result = iceland_generate_dpm_level_enable_mask(hwmgr, input); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to generate DPM level enabled mask!", result = tmp_result); + + tmp_result = iceland_update_vce_dpm(hwmgr, input); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to update VCE DPM!", result = tmp_result); + + tmp_result = iceland_update_sclk_threshold(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to update SCLK threshold!", result = tmp_result); + + tmp_result = iceland_update_and_upload_mc_reg_table(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to upload MC reg table!", result = tmp_result); + + tmp_result = iceland_program_memory_timing_parameters_conditionally(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to program memory timing parameters!", result = tmp_result); + + tmp_result = iceland_unfreeze_sclk_mclk_dpm(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to unfreeze SCLK MCLK DPM!", result = tmp_result); + + tmp_result = iceland_upload_dpm_level_enable_mask(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to upload DPM level enabled mask!", result = tmp_result); + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_PCIEPerformanceRequest)) { + tmp_result = iceland_notify_link_speed_change_after_state_change(hwmgr, input); + PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to notify link speed change after state change!", result = tmp_result); + } + + return result; +} + +static int iceland_get_power_state_size(struct pp_hwmgr *hwmgr) +{ + return sizeof(struct iceland_power_state); +} + +static int iceland_dpm_get_mclk(struct pp_hwmgr *hwmgr, bool low) +{ + struct pp_power_state *ps; + struct iceland_power_state *iceland_ps; + + if (hwmgr == NULL) + return -EINVAL; + + ps = hwmgr->request_ps; + + if (ps == NULL) + return -EINVAL; + + iceland_ps = cast_phw_iceland_power_state(&ps->hardware); + + if (low) + return iceland_ps->performance_levels[0].memory_clock; + else + return iceland_ps->performance_levels[iceland_ps->performance_level_count-1].memory_clock; +} + +static int iceland_dpm_get_sclk(struct pp_hwmgr *hwmgr, bool low) +{ + struct pp_power_state *ps; + struct iceland_power_state *iceland_ps; + + if (hwmgr == NULL) + return -EINVAL; + + ps = hwmgr->request_ps; + + if (ps == NULL) + return -EINVAL; + + iceland_ps = cast_phw_iceland_power_state(&ps->hardware); + + if (low) + return iceland_ps->performance_levels[0].engine_clock; + else + return iceland_ps->performance_levels[iceland_ps->performance_level_count-1].engine_clock; +} + +static int iceland_get_current_pcie_lane_number( + struct pp_hwmgr *hwmgr) +{ + uint32_t link_width; + + link_width = PHM_READ_INDIRECT_FIELD(hwmgr->device, + CGS_IND_REG__PCIE, + PCIE_LC_LINK_WIDTH_CNTL, + LC_LINK_WIDTH_RD); + + PP_ASSERT_WITH_CODE((7 >= link_width), + "Invalid PCIe lane width!", return 0); + + return decode_pcie_lane_width(link_width); +} + +static int iceland_dpm_patch_boot_state(struct pp_hwmgr *hwmgr, + struct pp_hw_power_state *hw_ps) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + struct iceland_power_state *ps = (struct iceland_power_state *)hw_ps; + ATOM_FIRMWARE_INFO_V2_2 *fw_info; + uint16_t size; + uint8_t frev, crev; + int index = GetIndexIntoMasterTable(DATA, FirmwareInfo); + + /* First retrieve the Boot clocks and VDDC from the firmware info table. + * We assume here that fw_info is unchanged if this call fails. + */ + fw_info = (ATOM_FIRMWARE_INFO_V2_2 *)cgs_atom_get_data_table( + hwmgr->device, index, + &size, &frev, &crev); + if (!fw_info) + /* During a test, there is no firmware info table. */ + return 0; + + /* Patch the state. */ + data->vbios_boot_state.sclk_bootup_value = le32_to_cpu(fw_info->ulDefaultEngineClock); + data->vbios_boot_state.mclk_bootup_value = le32_to_cpu(fw_info->ulDefaultMemoryClock); + data->vbios_boot_state.mvdd_bootup_value = le16_to_cpu(fw_info->usBootUpMVDDCVoltage); + data->vbios_boot_state.vddc_bootup_value = le16_to_cpu(fw_info->usBootUpVDDCVoltage); + data->vbios_boot_state.vddci_bootup_value = le16_to_cpu(fw_info->usBootUpVDDCIVoltage); + data->vbios_boot_state.pcie_gen_bootup_value = iceland_get_current_pcie_speed(hwmgr); + data->vbios_boot_state.pcie_lane_bootup_value = + (uint16_t)iceland_get_current_pcie_lane_number(hwmgr); + + /* set boot power state */ + ps->performance_levels[0].memory_clock = data->vbios_boot_state.mclk_bootup_value; + ps->performance_levels[0].engine_clock = data->vbios_boot_state.sclk_bootup_value; + ps->performance_levels[0].pcie_gen = data->vbios_boot_state.pcie_gen_bootup_value; + ps->performance_levels[0].pcie_lane = data->vbios_boot_state.pcie_lane_bootup_value; + + return 0; +} + +static int iceland_get_pp_table_entry_callback_func(struct pp_hwmgr *hwmgr, + struct pp_hw_power_state *power_state, + unsigned int index, const void *clock_info) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + struct iceland_power_state *iceland_power_state = cast_phw_iceland_power_state(power_state); + const ATOM_PPLIB_CI_CLOCK_INFO *visland_clk_info = clock_info; + struct iceland_performance_level *performance_level; + uint32_t engine_clock, memory_clock; + uint16_t pcie_gen_from_bios; + + engine_clock = visland_clk_info->ucEngineClockHigh << 16 | visland_clk_info->usEngineClockLow; + memory_clock = visland_clk_info->ucMemoryClockHigh << 16 | visland_clk_info->usMemoryClockLow; + + if (!(data->mc_micro_code_feature & DISABLE_MC_LOADMICROCODE) && memory_clock > data->highest_mclk) + data->highest_mclk = memory_clock; + + performance_level = &(iceland_power_state->performance_levels + [iceland_power_state->performance_level_count++]); + + PP_ASSERT_WITH_CODE( + (iceland_power_state->performance_level_count < SMU71_MAX_LEVELS_GRAPHICS), + "Performance levels exceeds SMC limit!", + return -1); + + PP_ASSERT_WITH_CODE( + (iceland_power_state->performance_level_count <= + hwmgr->platform_descriptor.hardwareActivityPerformanceLevels), + "Performance levels exceeds Driver limit!", + return -1); + + /* Performance levels are arranged from low to high. */ + performance_level->memory_clock = memory_clock; + performance_level->engine_clock = engine_clock; + + pcie_gen_from_bios = visland_clk_info->ucPCIEGen; + + performance_level->pcie_gen = get_pcie_gen_support(data->pcie_gen_cap, pcie_gen_from_bios); + performance_level->pcie_lane = get_pcie_lane_support(data->pcie_lane_cap, visland_clk_info->usPCIELane); + + return 0; +} + +static int iceland_get_pp_table_entry(struct pp_hwmgr *hwmgr, + unsigned long entry_index, struct pp_power_state *state) +{ + int result; + struct iceland_power_state *ps; + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + struct phm_clock_voltage_dependency_table *dep_mclk_table = + hwmgr->dyn_state.vddci_dependency_on_mclk; + + memset(&state->hardware, 0x00, sizeof(struct pp_hw_power_state)); + + state->hardware.magic = PHM_VIslands_Magic; + + ps = (struct iceland_power_state *)(&state->hardware); + + result = pp_tables_get_entry(hwmgr, entry_index, state, + iceland_get_pp_table_entry_callback_func); + + /* + * This is the earliest time we have all the dependency table + * and the VBIOS boot state as + * PP_Tables_GetPowerPlayTableEntry retrieves the VBIOS boot + * state if there is only one VDDCI/MCLK level, check if it's + * the same as VBIOS boot state + */ + if (dep_mclk_table != NULL && dep_mclk_table->count == 1) { + if (dep_mclk_table->entries[0].clk != + data->vbios_boot_state.mclk_bootup_value) + printk(KERN_ERR "Single MCLK entry VDDCI/MCLK dependency table " + "does not match VBIOS boot MCLK level"); + if (dep_mclk_table->entries[0].v != + data->vbios_boot_state.vddci_bootup_value) + printk(KERN_ERR "Single VDDCI entry VDDCI/MCLK dependency table " + "does not match VBIOS boot VDDCI level"); + } + + /* set DC compatible flag if this state supports DC */ + if (!state->validation.disallowOnDC) + ps->dc_compatible = true; + + if (state->classification.flags & PP_StateClassificationFlag_ACPI) + data->acpi_pcie_gen = ps->performance_levels[0].pcie_gen; + else if (0 != (state->classification.flags & PP_StateClassificationFlag_Boot)) { + if (data->bacos.best_match == 0xffff) { + /* For C.I. use boot state as base BACO state */ + data->bacos.best_match = PP_StateClassificationFlag_Boot; + data->bacos.performance_level = ps->performance_levels[0]; + } + } + + + ps->uvd_clocks.VCLK = state->uvd_clocks.VCLK; + ps->uvd_clocks.DCLK = state->uvd_clocks.DCLK; + + if (!result) { + uint32_t i; + + switch (state->classification.ui_label) { + case PP_StateUILabel_Performance: + data->use_pcie_performance_levels = true; + + for (i = 0; i < ps->performance_level_count; i++) { + if (data->pcie_gen_performance.max < + ps->performance_levels[i].pcie_gen) + data->pcie_gen_performance.max = + ps->performance_levels[i].pcie_gen; + + if (data->pcie_gen_performance.min > + ps->performance_levels[i].pcie_gen) + data->pcie_gen_performance.min = + ps->performance_levels[i].pcie_gen; + + if (data->pcie_lane_performance.max < + ps->performance_levels[i].pcie_lane) + data->pcie_lane_performance.max = + ps->performance_levels[i].pcie_lane; + + if (data->pcie_lane_performance.min > + ps->performance_levels[i].pcie_lane) + data->pcie_lane_performance.min = + ps->performance_levels[i].pcie_lane; + } + break; + case PP_StateUILabel_Battery: + data->use_pcie_power_saving_levels = true; + + for (i = 0; i < ps->performance_level_count; i++) { + if (data->pcie_gen_power_saving.max < + ps->performance_levels[i].pcie_gen) + data->pcie_gen_power_saving.max = + ps->performance_levels[i].pcie_gen; + + if (data->pcie_gen_power_saving.min > + ps->performance_levels[i].pcie_gen) + data->pcie_gen_power_saving.min = + ps->performance_levels[i].pcie_gen; + + if (data->pcie_lane_power_saving.max < + ps->performance_levels[i].pcie_lane) + data->pcie_lane_power_saving.max = + ps->performance_levels[i].pcie_lane; + + if (data->pcie_lane_power_saving.min > + ps->performance_levels[i].pcie_lane) + data->pcie_lane_power_saving.min = + ps->performance_levels[i].pcie_lane; + } + break; + default: + break; + } + } + return 0; +} + +static void +iceland_print_current_perforce_level(struct pp_hwmgr *hwmgr, struct seq_file *m) +{ + uint32_t sclk, mclk, activity_percent; + uint32_t offset; + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + smum_send_msg_to_smc(hwmgr->smumgr, (PPSMC_Msg)(PPSMC_MSG_API_GetSclkFrequency)); + + sclk = cgs_read_register(hwmgr->device, mmSMC_MSG_ARG_0); + + smum_send_msg_to_smc(hwmgr->smumgr, (PPSMC_Msg)(PPSMC_MSG_API_GetMclkFrequency)); + + mclk = cgs_read_register(hwmgr->device, mmSMC_MSG_ARG_0); + seq_printf(m, "\n [ mclk ]: %u MHz\n\n [ sclk ]: %u MHz\n", mclk/100, sclk/100); + + offset = data->soft_regs_start + offsetof(SMU71_SoftRegisters, AverageGraphicsActivity); + activity_percent = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, offset); + activity_percent += 0x80; + activity_percent >>= 8; + + seq_printf(m, "\n [GPU load]: %u%%\n\n", activity_percent > 100 ? 100 : activity_percent); + + seq_printf(m, "uvd %sabled\n", data->uvd_power_gated ? "dis" : "en"); + + seq_printf(m, "vce %sabled\n", data->vce_power_gated ? "dis" : "en"); +} + +int iceland_notify_smc_display_config_after_ps_adjustment(struct pp_hwmgr *hwmgr) +{ + uint32_t num_active_displays = 0; + struct cgs_display_info info = {0}; + info.mode_info = NULL; + + cgs_get_active_displays_info(hwmgr->device, &info); + + num_active_displays = info.display_count; + + if (num_active_displays > 1) /* to do && (pHwMgr->pPECI->displayConfiguration.bMultiMonitorInSync != TRUE)) */ + iceland_notify_smc_display_change(hwmgr, false); + else + iceland_notify_smc_display_change(hwmgr, true); + + return 0; +} + +/** +* Programs the display gap +* +* @param hwmgr the address of the powerplay hardware manager. +* @return always OK +*/ +int iceland_program_display_gap(struct pp_hwmgr *hwmgr) +{ + uint32_t num_active_displays = 0; + uint32_t display_gap = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_DISPLAY_GAP_CNTL); + uint32_t display_gap2; + uint32_t pre_vbi_time_in_us; + uint32_t frame_time_in_us; + uint32_t ref_clock; + uint32_t refresh_rate = 0; + struct cgs_display_info info = {0}; + struct cgs_mode_info mode_info; + + info.mode_info = &mode_info; + + cgs_get_active_displays_info(hwmgr->device, &info); + num_active_displays = info.display_count; + + display_gap = PHM_SET_FIELD(display_gap, CG_DISPLAY_GAP_CNTL, DISP_GAP, (num_active_displays > 0)? DISPLAY_GAP_VBLANK_OR_WM : DISPLAY_GAP_IGNORE); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_DISPLAY_GAP_CNTL, display_gap); + + ref_clock = mode_info.ref_clock; + refresh_rate = mode_info.refresh_rate; + + if(0 == refresh_rate) + refresh_rate = 60; + + frame_time_in_us = 1000000 / refresh_rate; + + pre_vbi_time_in_us = frame_time_in_us - 200 - mode_info.vblank_time_us; + display_gap2 = pre_vbi_time_in_us * (ref_clock / 100); + + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_DISPLAY_GAP_CNTL2, display_gap2); + + PHM_WRITE_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, SOFT_REGISTERS_TABLE_4, PreVBlankGap, 0x64); + + PHM_WRITE_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, SOFT_REGISTERS_TABLE_5, VBlankTimeout, (frame_time_in_us - pre_vbi_time_in_us)); + + if (num_active_displays == 1) + iceland_notify_smc_display_change(hwmgr, true); + + return 0; +} + +int iceland_display_configuration_changed_task(struct pp_hwmgr *hwmgr) +{ + iceland_program_display_gap(hwmgr); + + return 0; +} + +/** +* Set maximum target operating fan output PWM +* +* @param pHwMgr: the address of the powerplay hardware manager. +* @param usMaxFanPwm: max operating fan PWM in percents +* @return The response that came from the SMC. +*/ +static int iceland_set_max_fan_pwm_output(struct pp_hwmgr *hwmgr, uint16_t us_max_fan_pwm) +{ + hwmgr->thermal_controller.advanceFanControlParameters.usMaxFanPWM = us_max_fan_pwm; + + if (phm_is_hw_access_blocked(hwmgr)) + return 0; + + return (0 == smum_send_msg_to_smc_with_parameter(hwmgr->smumgr, PPSMC_MSG_SetFanPwmMax, us_max_fan_pwm) ? 0 : -1); +} + +/** +* Set maximum target operating fan output RPM +* +* @param pHwMgr: the address of the powerplay hardware manager. +* @param usMaxFanRpm: max operating fan RPM value. +* @return The response that came from the SMC. +*/ +static int iceland_set_max_fan_rpm_output(struct pp_hwmgr *hwmgr, uint16_t us_max_fan_pwm) +{ + hwmgr->thermal_controller.advanceFanControlParameters.usMaxFanRPM = us_max_fan_pwm; + + if (phm_is_hw_access_blocked(hwmgr)) + return 0; + + return (0 == smum_send_msg_to_smc_with_parameter(hwmgr->smumgr, PPSMC_MSG_SetFanRpmMax, us_max_fan_pwm) ? 0 : -1); +} + +static int iceland_dpm_set_interrupt_state(void *private_data, + unsigned src_id, unsigned type, + int enabled) +{ + uint32_t cg_thermal_int; + struct pp_hwmgr *hwmgr = ((struct pp_eventmgr *)private_data)->hwmgr; + + if (hwmgr == NULL) + return -EINVAL; + + switch (type) { + case AMD_THERMAL_IRQ_LOW_TO_HIGH: + if (enabled) { + cg_thermal_int = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_THERMAL_INT); + cg_thermal_int |= CG_THERMAL_INT_CTRL__THERM_INTH_MASK_MASK; + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_THERMAL_INT, cg_thermal_int); + } else { + cg_thermal_int = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_THERMAL_INT); + cg_thermal_int &= ~CG_THERMAL_INT_CTRL__THERM_INTH_MASK_MASK; + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_THERMAL_INT, cg_thermal_int); + } + break; + + case AMD_THERMAL_IRQ_HIGH_TO_LOW: + if (enabled) { + cg_thermal_int = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_THERMAL_INT); + cg_thermal_int |= CG_THERMAL_INT_CTRL__THERM_INTL_MASK_MASK; + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_THERMAL_INT, cg_thermal_int); + } else { + cg_thermal_int = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_THERMAL_INT); + cg_thermal_int &= ~CG_THERMAL_INT_CTRL__THERM_INTL_MASK_MASK; + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_THERMAL_INT, cg_thermal_int); + } + break; + default: + break; + } + return 0; +} + +static int iceland_register_internal_thermal_interrupt(struct pp_hwmgr *hwmgr, + const void *thermal_interrupt_info) +{ + int result; + const struct pp_interrupt_registration_info *info = + (const struct pp_interrupt_registration_info *)thermal_interrupt_info; + + if (info == NULL) + return -EINVAL; + + result = cgs_add_irq_source(hwmgr->device, 230, AMD_THERMAL_IRQ_LAST, + iceland_dpm_set_interrupt_state, + info->call_back, info->context); + + if (result) + return -EINVAL; + + result = cgs_add_irq_source(hwmgr->device, 231, AMD_THERMAL_IRQ_LAST, + iceland_dpm_set_interrupt_state, + info->call_back, info->context); + + if (result) + return -EINVAL; + + return 0; +} + + +static bool iceland_check_smc_update_required_for_display_configuration(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + bool is_update_required = false; + struct cgs_display_info info = {0,0,NULL}; + + cgs_get_active_displays_info(hwmgr->device, &info); + + if (data->display_timing.num_existing_displays != info.display_count) + is_update_required = true; +/* TO DO NEED TO GET DEEP SLEEP CLOCK FROM DAL + if (phm_cap_enabled(hwmgr->hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_SclkDeepSleep)) { + cgs_get_min_clock_settings(hwmgr->device, &min_clocks); + if(min_clocks.engineClockInSR != data->display_timing.minClockInSR) + is_update_required = true; +*/ + return is_update_required; +} + + +static inline bool iceland_are_power_levels_equal(const struct iceland_performance_level *pl1, + const struct iceland_performance_level *pl2) +{ + return ((pl1->memory_clock == pl2->memory_clock) && + (pl1->engine_clock == pl2->engine_clock) && + (pl1->pcie_gen == pl2->pcie_gen) && + (pl1->pcie_lane == pl2->pcie_lane)); +} + +int iceland_check_states_equal(struct pp_hwmgr *hwmgr, const struct pp_hw_power_state *pstate1, + const struct pp_hw_power_state *pstate2, bool *equal) +{ + const struct iceland_power_state *psa = cast_const_phw_iceland_power_state(pstate1); + const struct iceland_power_state *psb = cast_const_phw_iceland_power_state(pstate2); + int i; + + if (equal == NULL || psa == NULL || psb == NULL) + return -EINVAL; + + /* If the two states don't even have the same number of performance levels they cannot be the same state. */ + if (psa->performance_level_count != psb->performance_level_count) { + *equal = false; + return 0; + } + + for (i = 0; i < psa->performance_level_count; i++) { + if (!iceland_are_power_levels_equal(&(psa->performance_levels[i]), &(psb->performance_levels[i]))) { + /* If we have found even one performance level pair that is different the states are different. */ + *equal = false; + return 0; + } + } + + /* If all performance levels are the same try to use the UVD clocks to break the tie.*/ + *equal = ((psa->uvd_clocks.VCLK == psb->uvd_clocks.VCLK) && (psa->uvd_clocks.DCLK == psb->uvd_clocks.DCLK)); + *equal &= ((psa->vce_clocks.EVCLK == psb->vce_clocks.EVCLK) && (psa->vce_clocks.ECCLK == psb->vce_clocks.ECCLK)); + *equal &= (psa->sclk_threshold == psb->sclk_threshold); + *equal &= (psa->acp_clk == psb->acp_clk); + + return 0; +} + +static int iceland_set_fan_control_mode(struct pp_hwmgr *hwmgr, uint32_t mode) +{ + if (mode) { + /* stop auto-manage */ + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_MicrocodeFanControl)) + iceland_fan_ctrl_stop_smc_fan_control(hwmgr); + iceland_fan_ctrl_set_static_mode(hwmgr, mode); + } else + /* restart auto-manage */ + iceland_fan_ctrl_reset_fan_speed_to_default(hwmgr); + + return 0; +} + +static int iceland_get_fan_control_mode(struct pp_hwmgr *hwmgr) +{ + if (hwmgr->fan_ctrl_is_in_default_mode) + return hwmgr->fan_ctrl_default_mode; + else + return PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + CG_FDO_CTRL2, FDO_PWM_MODE); +} + +static int iceland_force_clock_level(struct pp_hwmgr *hwmgr, + enum pp_clock_type type, uint32_t mask) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + if (hwmgr->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL) + return -EINVAL; + + switch (type) { + case PP_SCLK: + if (!data->sclk_dpm_key_disabled) + smum_send_msg_to_smc_with_parameter(hwmgr->smumgr, + PPSMC_MSG_SCLKDPM_SetEnabledMask, + data->dpm_level_enable_mask.sclk_dpm_enable_mask & mask); + break; + case PP_MCLK: + if (!data->mclk_dpm_key_disabled) + smum_send_msg_to_smc_with_parameter(hwmgr->smumgr, + PPSMC_MSG_MCLKDPM_SetEnabledMask, + data->dpm_level_enable_mask.mclk_dpm_enable_mask & mask); + break; + case PP_PCIE: + { + uint32_t tmp = mask & data->dpm_level_enable_mask.pcie_dpm_enable_mask; + uint32_t level = 0; + + while (tmp >>= 1) + level++; + + if (!data->pcie_dpm_key_disabled) + smum_send_msg_to_smc_with_parameter(hwmgr->smumgr, + PPSMC_MSG_PCIeDPM_ForceLevel, + level); + break; + } + default: + break; + } + + return 0; +} + +static int iceland_print_clock_levels(struct pp_hwmgr *hwmgr, + enum pp_clock_type type, char *buf) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + struct iceland_single_dpm_table *sclk_table = &(data->dpm_table.sclk_table); + struct iceland_single_dpm_table *mclk_table = &(data->dpm_table.mclk_table); + struct iceland_single_dpm_table *pcie_table = &(data->dpm_table.pcie_speed_table); + int i, now, size = 0; + uint32_t clock, pcie_speed; + + switch (type) { + case PP_SCLK: + smum_send_msg_to_smc(hwmgr->smumgr, PPSMC_MSG_API_GetSclkFrequency); + clock = cgs_read_register(hwmgr->device, mmSMC_MSG_ARG_0); + + for (i = 0; i < sclk_table->count; i++) { + if (clock > sclk_table->dpm_levels[i].value) + continue; + break; + } + now = i; + + for (i = 0; i < sclk_table->count; i++) + size += sprintf(buf + size, "%d: %uMhz %s\n", + i, sclk_table->dpm_levels[i].value / 100, + (i == now) ? "*" : ""); + break; + case PP_MCLK: + smum_send_msg_to_smc(hwmgr->smumgr, PPSMC_MSG_API_GetMclkFrequency); + clock = cgs_read_register(hwmgr->device, mmSMC_MSG_ARG_0); + + for (i = 0; i < mclk_table->count; i++) { + if (clock > mclk_table->dpm_levels[i].value) + continue; + break; + } + now = i; + + for (i = 0; i < mclk_table->count; i++) + size += sprintf(buf + size, "%d: %uMhz %s\n", + i, mclk_table->dpm_levels[i].value / 100, + (i == now) ? "*" : ""); + break; + case PP_PCIE: + pcie_speed = iceland_get_current_pcie_speed(hwmgr); + for (i = 0; i < pcie_table->count; i++) { + if (pcie_speed != pcie_table->dpm_levels[i].value) + continue; + break; + } + now = i; + + for (i = 0; i < pcie_table->count; i++) + size += sprintf(buf + size, "%d: %s %s\n", i, + (pcie_table->dpm_levels[i].value == 0) ? "2.5GB, x8" : + (pcie_table->dpm_levels[i].value == 1) ? "5.0GB, x16" : + (pcie_table->dpm_levels[i].value == 2) ? "8.0GB, x16" : "", + (i == now) ? "*" : ""); + break; + default: + break; + } + return size; +} + +static int iceland_get_sclk_od(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + struct iceland_single_dpm_table *sclk_table = &(data->dpm_table.sclk_table); + struct iceland_single_dpm_table *golden_sclk_table = + &(data->golden_dpm_table.sclk_table); + int value; + + value = (sclk_table->dpm_levels[sclk_table->count - 1].value - + golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value) * + 100 / + golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value; + + return value; +} + +static int iceland_set_sclk_od(struct pp_hwmgr *hwmgr, uint32_t value) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + struct iceland_single_dpm_table *golden_sclk_table = + &(data->golden_dpm_table.sclk_table); + struct pp_power_state *ps; + struct iceland_power_state *iceland_ps; + + if (value > 20) + value = 20; + + ps = hwmgr->request_ps; + + if (ps == NULL) + return -EINVAL; + + iceland_ps = cast_phw_iceland_power_state(&ps->hardware); + + iceland_ps->performance_levels[iceland_ps->performance_level_count - 1].engine_clock = + golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value * + value / 100 + + golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value; + + return 0; +} + +static int iceland_get_mclk_od(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + struct iceland_single_dpm_table *mclk_table = &(data->dpm_table.mclk_table); + struct iceland_single_dpm_table *golden_mclk_table = + &(data->golden_dpm_table.mclk_table); + int value; + + value = (mclk_table->dpm_levels[mclk_table->count - 1].value - + golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value) * + 100 / + golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value; + + return value; +} + +uint32_t iceland_get_xclk(struct pp_hwmgr *hwmgr) +{ + uint32_t reference_clock; + uint32_t tc; + uint32_t divide; + + ATOM_FIRMWARE_INFO *fw_info; + uint16_t size; + uint8_t frev, crev; + int index = GetIndexIntoMasterTable(DATA, FirmwareInfo); + + tc = PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_CLKPIN_CNTL_2, MUX_TCLK_TO_XCLK); + + if (tc) + return TCLK; + + fw_info = (ATOM_FIRMWARE_INFO *)cgs_atom_get_data_table(hwmgr->device, index, + &size, &frev, &crev); + + if (!fw_info) + return 0; + + reference_clock = le16_to_cpu(fw_info->usReferenceClock); + + divide = PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_CLKPIN_CNTL, XTALIN_DIVIDE); + + if (0 != divide) + return reference_clock / 4; + + return reference_clock; +} + +static int iceland_set_mclk_od(struct pp_hwmgr *hwmgr, uint32_t value) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + struct iceland_single_dpm_table *golden_mclk_table = + &(data->golden_dpm_table.mclk_table); + struct pp_power_state *ps; + struct iceland_power_state *iceland_ps; + + if (value > 20) + value = 20; + + ps = hwmgr->request_ps; + + if (ps == NULL) + return -EINVAL; + + iceland_ps = cast_phw_iceland_power_state(&ps->hardware); + + iceland_ps->performance_levels[iceland_ps->performance_level_count - 1].memory_clock = + golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value * + value / 100 + + golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value; + + return 0; +} + +static const struct pp_hwmgr_func iceland_hwmgr_funcs = { + .backend_init = &iceland_hwmgr_backend_init, + .backend_fini = &iceland_hwmgr_backend_fini, + .asic_setup = &iceland_setup_asic_task, + .dynamic_state_management_enable = &iceland_enable_dpm_tasks, + .apply_state_adjust_rules = iceland_apply_state_adjust_rules, + .force_dpm_level = &iceland_force_dpm_level, + .power_state_set = iceland_set_power_state_tasks, + .get_power_state_size = iceland_get_power_state_size, + .get_mclk = iceland_dpm_get_mclk, + .get_sclk = iceland_dpm_get_sclk, + .patch_boot_state = iceland_dpm_patch_boot_state, + .get_pp_table_entry = iceland_get_pp_table_entry, + .get_num_of_pp_table_entries = iceland_get_num_of_entries, + .print_current_perforce_level = iceland_print_current_perforce_level, + .powerdown_uvd = iceland_phm_powerdown_uvd, + .powergate_uvd = iceland_phm_powergate_uvd, + .powergate_vce = iceland_phm_powergate_vce, + .disable_clock_power_gating = iceland_phm_disable_clock_power_gating, + .update_clock_gatings = iceland_phm_update_clock_gatings, + .notify_smc_display_config_after_ps_adjustment = iceland_notify_smc_display_config_after_ps_adjustment, + .display_config_changed = iceland_display_configuration_changed_task, + .set_max_fan_pwm_output = iceland_set_max_fan_pwm_output, + .set_max_fan_rpm_output = iceland_set_max_fan_rpm_output, + .get_temperature = iceland_thermal_get_temperature, + .stop_thermal_controller = iceland_thermal_stop_thermal_controller, + .get_fan_speed_info = iceland_fan_ctrl_get_fan_speed_info, + .get_fan_speed_percent = iceland_fan_ctrl_get_fan_speed_percent, + .set_fan_speed_percent = iceland_fan_ctrl_set_fan_speed_percent, + .reset_fan_speed_to_default = iceland_fan_ctrl_reset_fan_speed_to_default, + .get_fan_speed_rpm = iceland_fan_ctrl_get_fan_speed_rpm, + .set_fan_speed_rpm = iceland_fan_ctrl_set_fan_speed_rpm, + .uninitialize_thermal_controller = iceland_thermal_ctrl_uninitialize_thermal_controller, + .register_internal_thermal_interrupt = iceland_register_internal_thermal_interrupt, + .check_smc_update_required_for_display_configuration = iceland_check_smc_update_required_for_display_configuration, + .check_states_equal = iceland_check_states_equal, + .set_fan_control_mode = iceland_set_fan_control_mode, + .get_fan_control_mode = iceland_get_fan_control_mode, + .force_clock_level = iceland_force_clock_level, + .print_clock_levels = iceland_print_clock_levels, + .get_sclk_od = iceland_get_sclk_od, + .set_sclk_od = iceland_set_sclk_od, + .get_mclk_od = iceland_get_mclk_od, + .set_mclk_od = iceland_set_mclk_od, +}; + +int iceland_hwmgr_init(struct pp_hwmgr *hwmgr) +{ + iceland_hwmgr *data; + + data = kzalloc (sizeof(iceland_hwmgr), GFP_KERNEL); + if (data == NULL) + return -ENOMEM; + memset(data, 0x00, sizeof(iceland_hwmgr)); + + hwmgr->backend = data; + hwmgr->hwmgr_func = &iceland_hwmgr_funcs; + hwmgr->pptable_func = &pptable_funcs; + + /* thermal */ + pp_iceland_thermal_initialize(hwmgr); + return 0; +} diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_hwmgr.h b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_hwmgr.h new file mode 100644 index 000000000000..f253988de2d2 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_hwmgr.h @@ -0,0 +1,424 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Author: Huang Rui <ray.huang@amd.com> + * + */ +#ifndef ICELAND_HWMGR_H +#define ICELAND_HWMGR_H + +#include "hwmgr.h" +#include "ppatomctrl.h" +#include "ppinterrupt.h" +#include "ppsmc.h" +#include "iceland_powertune.h" +#include "pp_endian.h" +#include "smu71_discrete.h" + +#define ICELAND_MAX_HARDWARE_POWERLEVELS 2 +#define ICELAND_DYNCLK_NUMBER_OF_TREND_COEFFICIENTS 15 + +struct iceland_performance_level { + uint32_t memory_clock; + uint32_t engine_clock; + uint16_t pcie_gen; + uint16_t pcie_lane; +}; + +struct _phw_iceland_bacos { + uint32_t best_match; + uint32_t baco_flags; + struct iceland_performance_level performance_level; +}; +typedef struct _phw_iceland_bacos phw_iceland_bacos; + +struct _phw_iceland_uvd_clocks { + uint32_t VCLK; + uint32_t DCLK; +}; + +typedef struct _phw_iceland_uvd_clocks phw_iceland_uvd_clocks; + +struct _phw_iceland_vce_clocks { + uint32_t EVCLK; + uint32_t ECCLK; +}; + +typedef struct _phw_iceland_vce_clocks phw_iceland_vce_clocks; + +struct iceland_power_state { + uint32_t magic; + phw_iceland_uvd_clocks uvd_clocks; + phw_iceland_vce_clocks vce_clocks; + uint32_t sam_clk; + uint32_t acp_clk; + uint16_t performance_level_count; + bool dc_compatible; + uint32_t sclk_threshold; + struct iceland_performance_level performance_levels[ICELAND_MAX_HARDWARE_POWERLEVELS]; +}; + +struct _phw_iceland_dpm_level { + bool enabled; + uint32_t value; + uint32_t param1; +}; +typedef struct _phw_iceland_dpm_level phw_iceland_dpm_level; + +#define ICELAND_MAX_DEEPSLEEP_DIVIDER_ID 5 +#define MAX_REGULAR_DPM_NUMBER 8 +#define ICELAND_MINIMUM_ENGINE_CLOCK 5000 + +struct iceland_single_dpm_table { + uint32_t count; + phw_iceland_dpm_level dpm_levels[MAX_REGULAR_DPM_NUMBER]; +}; + +struct iceland_dpm_table { + struct iceland_single_dpm_table sclk_table; + struct iceland_single_dpm_table mclk_table; + struct iceland_single_dpm_table pcie_speed_table; + struct iceland_single_dpm_table vddc_table; + struct iceland_single_dpm_table vdd_gfx_table; + struct iceland_single_dpm_table vdd_ci_table; + struct iceland_single_dpm_table mvdd_table; +}; +typedef struct _phw_iceland_dpm_table phw_iceland_dpm_table; + + +struct _phw_iceland_clock_regisiters { + uint32_t vCG_SPLL_FUNC_CNTL; + uint32_t vCG_SPLL_FUNC_CNTL_2; + uint32_t vCG_SPLL_FUNC_CNTL_3; + uint32_t vCG_SPLL_FUNC_CNTL_4; + uint32_t vCG_SPLL_SPREAD_SPECTRUM; + uint32_t vCG_SPLL_SPREAD_SPECTRUM_2; + uint32_t vDLL_CNTL; + uint32_t vMCLK_PWRMGT_CNTL; + uint32_t vMPLL_AD_FUNC_CNTL; + uint32_t vMPLL_DQ_FUNC_CNTL; + uint32_t vMPLL_FUNC_CNTL; + uint32_t vMPLL_FUNC_CNTL_1; + uint32_t vMPLL_FUNC_CNTL_2; + uint32_t vMPLL_SS1; + uint32_t vMPLL_SS2; +}; +typedef struct _phw_iceland_clock_regisiters phw_iceland_clock_registers; + +struct _phw_iceland_voltage_smio_registers { + uint32_t vs0_vid_lower_smio_cntl; +}; +typedef struct _phw_iceland_voltage_smio_registers phw_iceland_voltage_smio_registers; + + +struct _phw_iceland_mc_reg_entry { + uint32_t mclk_max; + uint32_t mc_data[SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE]; +}; +typedef struct _phw_iceland_mc_reg_entry phw_iceland_mc_reg_entry; + +struct _phw_iceland_mc_reg_table { + uint8_t last; /* number of registers*/ + uint8_t num_entries; /* number of entries in mc_reg_table_entry used*/ + uint16_t validflag; /* indicate the corresponding register is valid or not. 1: valid, 0: invalid. bit0->address[0], bit1->address[1], etc.*/ + phw_iceland_mc_reg_entry mc_reg_table_entry[MAX_AC_TIMING_ENTRIES]; + SMU71_Discrete_MCRegisterAddress mc_reg_address[SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE]; +}; +typedef struct _phw_iceland_mc_reg_table phw_iceland_mc_reg_table; + +#define DISABLE_MC_LOADMICROCODE 1 +#define DISABLE_MC_CFGPROGRAMMING 2 + + +/*Ultra Low Voltage parameter structure */ +struct phw_iceland_ulv_parm{ + bool ulv_supported; + uint32_t ch_ulv_parameter; + uint32_t ulv_volt_change_delay; + struct iceland_performance_level ulv_power_level; +}; + +#define ICELAND_MAX_LEAKAGE_COUNT 8 + +struct phw_iceland_leakage_voltage { + uint16_t count; + uint16_t leakage_id[ICELAND_MAX_LEAKAGE_COUNT]; + uint16_t actual_voltage[ICELAND_MAX_LEAKAGE_COUNT]; +}; + +struct _phw_iceland_display_timing { + uint32_t min_clock_insr; + uint32_t num_existing_displays; +}; +typedef struct _phw_iceland_display_timing phw_iceland_display_timing; + + +struct phw_iceland_thermal_temperature_setting +{ + long temperature_low; + long temperature_high; + long temperature_shutdown; +}; + +struct _phw_iceland_dpmlevel_enable_mask { + uint32_t uvd_dpm_enable_mask; + uint32_t vce_dpm_enable_mask; + uint32_t acp_dpm_enable_mask; + uint32_t samu_dpm_enable_mask; + uint32_t sclk_dpm_enable_mask; + uint32_t mclk_dpm_enable_mask; + uint32_t pcie_dpm_enable_mask; +}; +typedef struct _phw_iceland_dpmlevel_enable_mask phw_iceland_dpmlevel_enable_mask; + +struct _phw_iceland_pcie_perf_range { + uint16_t max; + uint16_t min; +}; +typedef struct _phw_iceland_pcie_perf_range phw_iceland_pcie_perf_range; + +struct _phw_iceland_vbios_boot_state { + uint16_t mvdd_bootup_value; + uint16_t vddc_bootup_value; + uint16_t vddci_bootup_value; + uint16_t vddgfx_bootup_value; + uint32_t sclk_bootup_value; + uint32_t mclk_bootup_value; + uint16_t pcie_gen_bootup_value; + uint16_t pcie_lane_bootup_value; +}; +typedef struct _phw_iceland_vbios_boot_state phw_iceland_vbios_boot_state; + +#define DPMTABLE_OD_UPDATE_SCLK 0x00000001 +#define DPMTABLE_OD_UPDATE_MCLK 0x00000002 +#define DPMTABLE_UPDATE_SCLK 0x00000004 +#define DPMTABLE_UPDATE_MCLK 0x00000008 + +/* We need to review which fields are needed. */ +/* This is mostly a copy of the RV7xx/Evergreen structure which is close, but not identical to the N.Islands one. */ +struct iceland_hwmgr { + struct iceland_dpm_table dpm_table; + struct iceland_dpm_table golden_dpm_table; + + uint32_t voting_rights_clients0; + uint32_t voting_rights_clients1; + uint32_t voting_rights_clients2; + uint32_t voting_rights_clients3; + uint32_t voting_rights_clients4; + uint32_t voting_rights_clients5; + uint32_t voting_rights_clients6; + uint32_t voting_rights_clients7; + uint32_t static_screen_threshold_unit; + uint32_t static_screen_threshold; + uint32_t voltage_control; + uint32_t vdd_gfx_control; + + uint32_t vddc_vddci_delta; + uint32_t vddc_vddgfx_delta; + + struct pp_interrupt_registration_info internal_high_thermal_interrupt_info; + struct pp_interrupt_registration_info internal_low_thermal_interrupt_info; + struct pp_interrupt_registration_info smc_to_host_interrupt_info; + uint32_t active_auto_throttle_sources; + + struct pp_interrupt_registration_info external_throttle_interrupt; + irq_handler_func_t external_throttle_callback; + void *external_throttle_context; + + struct pp_interrupt_registration_info ctf_interrupt_info; + irq_handler_func_t ctf_callback; + void *ctf_context; + + phw_iceland_clock_registers clock_registers; + phw_iceland_voltage_smio_registers voltage_smio_registers; + + bool is_memory_GDDR5; + uint16_t acpi_vddc; + bool pspp_notify_required; /* Flag to indicate if PSPP notification to SBIOS is required */ + uint16_t force_pcie_gen; /* The forced PCI-E speed if not 0xffff */ + uint16_t acpi_pcie_gen; /* The PCI-E speed at ACPI time */ + uint32_t pcie_gen_cap; /* The PCI-E speed capabilities bitmap from CAIL */ + uint32_t pcie_lane_cap; /* The PCI-E lane capabilities bitmap from CAIL */ + uint32_t pcie_spc_cap; /* Symbol Per Clock Capabilities from registry */ + struct phw_iceland_leakage_voltage vddc_leakage; /* The Leakage VDDC supported (based on leakage ID).*/ + struct phw_iceland_leakage_voltage vddcgfx_leakage; /* The Leakage VDDC supported (based on leakage ID). */ + struct phw_iceland_leakage_voltage vddci_leakage; /* The Leakage VDDCI supported (based on leakage ID). */ + + uint32_t mvdd_control; + uint32_t vddc_mask_low; + uint32_t mvdd_mask_low; + uint16_t max_vddc_in_pp_table; /* the maximum VDDC value in the powerplay table*/ + uint16_t min_vddc_in_pp_table; + uint16_t max_vddci_in_pp_table; /* the maximum VDDCI value in the powerplay table */ + uint16_t min_vddci_in_pp_table; + uint32_t mclk_strobe_mode_threshold; + uint32_t mclk_stutter_mode_threshold; + uint32_t mclk_edc_enable_threshold; + uint32_t mclk_edc_wr_enable_threshold; + bool is_uvd_enabled; + bool is_xdma_enabled; + phw_iceland_vbios_boot_state vbios_boot_state; + + bool battery_state; + bool is_tlu_enabled; + bool pcie_performance_request; + + /* -------------- SMC SRAM Address of firmware header tables ----------------*/ + uint32_t sram_end; /* The first address after the SMC SRAM. */ + uint32_t dpm_table_start; /* The start of the dpm table in the SMC SRAM. */ + uint32_t soft_regs_start; /* The start of the soft registers in the SMC SRAM. */ + uint32_t mc_reg_table_start; /* The start of the mc register table in the SMC SRAM. */ + uint32_t fan_table_start; /* The start of the fan table in the SMC SRAM. */ + uint32_t arb_table_start; /* The start of the ARB setting table in the SMC SRAM. */ + uint32_t ulv_settings_start; + SMU71_Discrete_DpmTable smc_state_table; /* The carbon copy of the SMC state table. */ + SMU71_Discrete_MCRegisters mc_reg_table; + SMU71_Discrete_Ulv ulv_setting; /* The carbon copy of ULV setting. */ + + /* -------------- Stuff originally coming from Evergreen --------------------*/ + phw_iceland_mc_reg_table iceland_mc_reg_table; + uint32_t vdd_ci_control; + pp_atomctrl_voltage_table vddc_voltage_table; + pp_atomctrl_voltage_table vddci_voltage_table; + pp_atomctrl_voltage_table vddgfx_voltage_table; + pp_atomctrl_voltage_table mvdd_voltage_table; + + uint32_t mgcg_cgtt_local2; + uint32_t mgcg_cgtt_local3; + uint32_t gpio_debug; + uint32_t mc_micro_code_feature; + uint32_t highest_mclk; + uint16_t acpi_vdd_ci; + uint8_t mvdd_high_index; + uint8_t mvdd_low_index; + bool dll_defaule_on; + bool performance_request_registered; + + /* ----------------- Low Power Features ---------------------*/ + phw_iceland_bacos bacos; + struct phw_iceland_ulv_parm ulv; + + /* ----------------- CAC Stuff ---------------------*/ + uint32_t cac_table_start; + bool cac_configuration_required; /* TRUE if PP_CACConfigurationRequired == 1 */ + bool driver_calculate_cac_leakage; /* TRUE if PP_DriverCalculateCACLeakage == 1 */ + bool cac_enabled; + + /* ----------------- DPM2 Parameters ---------------------*/ + uint32_t power_containment_features; + bool enable_bapm_feature; + bool enable_dte_feature; + bool enable_tdc_limit_feature; + bool enable_pkg_pwr_tracking_feature; + bool disable_uvd_power_tune_feature; + struct iceland_pt_defaults *power_tune_defaults; + SMU71_Discrete_PmFuses power_tune_table; + uint32_t ul_dte_tj_offset; /* Fudge factor in DPM table to correct HW DTE errors */ + uint32_t fast_watermark_threshold; /* use fast watermark if clock is equal or above this. In percentage of the target high sclk. */ + + /* ----------------- Phase Shedding ---------------------*/ + bool vddc_phase_shed_control; + + /* --------------------- DI/DT --------------------------*/ + phw_iceland_display_timing display_timing; + + /* --------- ReadRegistry data for memory and engine clock margins ---- */ + uint32_t engine_clock_data; + uint32_t memory_clock_data; + + /* -------- Thermal Temperature Setting --------------*/ + struct phw_iceland_thermal_temperature_setting thermal_temp_setting; + phw_iceland_dpmlevel_enable_mask dpm_level_enable_mask; + + uint32_t need_update_smu7_dpm_table; + uint32_t sclk_dpm_key_disabled; + uint32_t mclk_dpm_key_disabled; + uint32_t pcie_dpm_key_disabled; + /* used to store the previous dal min sclock */ + uint32_t min_engine_clocks; + phw_iceland_pcie_perf_range pcie_gen_performance; + phw_iceland_pcie_perf_range pcie_lane_performance; + phw_iceland_pcie_perf_range pcie_gen_power_saving; + phw_iceland_pcie_perf_range pcie_lane_power_saving; + bool use_pcie_performance_levels; + bool use_pcie_power_saving_levels; + /* percentage value from 0-100, default 50 */ + uint32_t activity_target[SMU71_MAX_LEVELS_GRAPHICS]; + uint32_t mclk_activity_target; + uint32_t low_sclk_interrupt_threshold; + uint32_t last_mclk_dpm_enable_mask; + bool uvd_enabled; + uint32_t pcc_monitor_enabled; + + /* --------- Power Gating States ------------*/ + bool uvd_power_gated; /* 1: gated, 0:not gated */ + bool vce_power_gated; /* 1: gated, 0:not gated */ + bool samu_power_gated; /* 1: gated, 0:not gated */ + bool acp_power_gated; /* 1: gated, 0:not gated */ + bool pg_acp_init; + + /* soft pptable for re-uploading into smu */ + void *soft_pp_table; +}; + +typedef struct iceland_hwmgr iceland_hwmgr; + +int iceland_hwmgr_init(struct pp_hwmgr *hwmgr); +int iceland_update_uvd_dpm(struct pp_hwmgr *hwmgr, bool bgate); +uint32_t iceland_get_xclk(struct pp_hwmgr *hwmgr); +int iceland_populate_bapm_vddc_vid_sidd(struct pp_hwmgr *hwmgr); +int iceland_populate_vddc_vid(struct pp_hwmgr *hwmgr); + +#define ICELAND_DPM2_NEAR_TDP_DEC 10 +#define ICELAND_DPM2_ABOVE_SAFE_INC 5 +#define ICELAND_DPM2_BELOW_SAFE_INC 20 + +/* + * Log2 of the LTA window size (l2numWin_TDP). Eg. If LTA windows size + * is 128, then this value should be Log2(128) = 7. + */ +#define ICELAND_DPM2_LTA_WINDOW_SIZE 7 + +#define ICELAND_DPM2_LTS_TRUNCATE 0 + +#define ICELAND_DPM2_TDP_SAFE_LIMIT_PERCENT 80 // Maximum 100 + +#define ICELAND_DPM2_MAXPS_PERCENT_H 90 // Maximum 0xFF +#define ICELAND_DPM2_MAXPS_PERCENT_M 90 // Maximum 0xFF + +#define ICELAND_DPM2_PWREFFICIENCYRATIO_MARGIN 50 + +#define ICELAND_DPM2_SQ_RAMP_MAX_POWER 0x3FFF +#define ICELAND_DPM2_SQ_RAMP_MIN_POWER 0x12 +#define ICELAND_DPM2_SQ_RAMP_MAX_POWER_DELTA 0x15 +#define ICELAND_DPM2_SQ_RAMP_SHORT_TERM_INTERVAL_SIZE 0x1E +#define ICELAND_DPM2_SQ_RAMP_LONG_TERM_INTERVAL_RATIO 0xF + +#define ICELAND_VOLTAGE_CONTROL_NONE 0x0 +#define ICELAND_VOLTAGE_CONTROL_BY_GPIO 0x1 +#define ICELAND_VOLTAGE_CONTROL_BY_SVID2 0x2 + +/* convert to Q8.8 format for firmware */ +#define ICELAND_Q88_FORMAT_CONVERSION_UNIT 256 + +#define ICELAND_UNUSED_GPIO_PIN 0x7F + +#endif diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_powertune.c b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_powertune.c new file mode 100644 index 000000000000..041e9648e592 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_powertune.c @@ -0,0 +1,490 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Author: Huang Rui <ray.huang@amd.com> + * + */ + +#include "amdgpu.h" +#include "hwmgr.h" +#include "smumgr.h" +#include "iceland_hwmgr.h" +#include "iceland_powertune.h" +#include "iceland_smumgr.h" +#include "smu71_discrete.h" +#include "smu71.h" +#include "pp_debug.h" +#include "cgs_common.h" +#include "pp_endian.h" + +#include "bif/bif_5_0_d.h" +#include "bif/bif_5_0_sh_mask.h" + +#define VOLTAGE_SCALE 4 +#define POWERTUNE_DEFAULT_SET_MAX 1 + +#define DEVICE_ID_VI_ICELAND_M_6900 0x6900 +#define DEVICE_ID_VI_ICELAND_M_6901 0x6901 +#define DEVICE_ID_VI_ICELAND_M_6902 0x6902 +#define DEVICE_ID_VI_ICELAND_M_6903 0x6903 + + +struct iceland_pt_defaults defaults_iceland = +{ + /* + * sviLoadLIneEn, SviLoadLineVddC, TDC_VDDC_ThrottleReleaseLimitPerc, + * TDC_MAWt, TdcWaterfallCtl, DTEAmbientTempBase, DisplayCac, BAPM_TEMP_GRADIENT + */ + 1, 0xF, 0xFD, 0x19, 5, 45, 0, 0xB0000, + { 0x79, 0x253, 0x25D, 0xAE, 0x72, 0x80, 0x83, 0x86, 0x6F, 0xC8, 0xC9, 0xC9, 0x2F, 0x4D, 0x61 }, + { 0x17C, 0x172, 0x180, 0x1BC, 0x1B3, 0x1BD, 0x206, 0x200, 0x203, 0x25D, 0x25A, 0x255, 0x2C3, 0x2C5, 0x2B4 } +}; + +/* 35W - XT, XTL */ +struct iceland_pt_defaults defaults_icelandxt = +{ + /* + * sviLoadLIneEn, SviLoadLineVddC, + * TDC_VDDC_ThrottleReleaseLimitPerc, TDC_MAWt, + * TdcWaterfallCtl, DTEAmbientTempBase, DisplayCac, + * BAPM_TEMP_GRADIENT + */ + 1, 0xF, 0xFD, 0x19, 5, 45, 0, 0x0, + { 0xA7, 0x0, 0x0, 0xB5, 0x0, 0x0, 0x9F, 0x0, 0x0, 0xD6, 0x0, 0x0, 0xD7, 0x0, 0x0}, + { 0x1EA, 0x0, 0x0, 0x224, 0x0, 0x0, 0x25E, 0x0, 0x0, 0x28E, 0x0, 0x0, 0x2AB, 0x0, 0x0} +}; + +/* 25W - PRO, LE */ +struct iceland_pt_defaults defaults_icelandpro = +{ + /* + * sviLoadLIneEn, SviLoadLineVddC, + * TDC_VDDC_ThrottleReleaseLimitPerc, TDC_MAWt, + * TdcWaterfallCtl, DTEAmbientTempBase, DisplayCac, + * BAPM_TEMP_GRADIENT + */ + 1, 0xF, 0xFD, 0x19, 5, 45, 0, 0x0, + { 0xB7, 0x0, 0x0, 0xC3, 0x0, 0x0, 0xB5, 0x0, 0x0, 0xEA, 0x0, 0x0, 0xE6, 0x0, 0x0}, + { 0x1EA, 0x0, 0x0, 0x224, 0x0, 0x0, 0x25E, 0x0, 0x0, 0x28E, 0x0, 0x0, 0x2AB, 0x0, 0x0} +}; + +void iceland_initialize_power_tune_defaults(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + uint32_t tmp = 0; + struct cgs_system_info sys_info = {0}; + uint32_t pdev_id; + + sys_info.size = sizeof(struct cgs_system_info); + sys_info.info_id = CGS_SYSTEM_INFO_PCIE_DEV; + cgs_query_system_info(hwmgr->device, &sys_info); + pdev_id = (uint32_t)sys_info.value; + + switch (pdev_id) { + case DEVICE_ID_VI_ICELAND_M_6900: + case DEVICE_ID_VI_ICELAND_M_6903: + data->power_tune_defaults = &defaults_icelandxt; + break; + + case DEVICE_ID_VI_ICELAND_M_6901: + case DEVICE_ID_VI_ICELAND_M_6902: + data->power_tune_defaults = &defaults_icelandpro; + break; + default: + /* TODO: need to assign valid defaults */ + data->power_tune_defaults = &defaults_iceland; + pr_warning("Unknown V.I. Device ID.\n"); + break; + } + + /* Assume disabled */ + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_CAC); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SQRamping); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DBRamping); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_TDRamping); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_TCPRamping); + + data->ul_dte_tj_offset = tmp; + + if (!tmp) { + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_CAC); + + data->fast_watermark_threshold = 100; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment)) { + tmp = 1; + data->enable_dte_feature = tmp ? false : true; + data->enable_tdc_limit_feature = tmp ? true : false; + data->enable_pkg_pwr_tracking_feature = tmp ? true : false; + } + } +} + +int iceland_populate_bapm_parameters_in_dpm_table(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + struct iceland_pt_defaults *defaults = data->power_tune_defaults; + SMU71_Discrete_DpmTable *dpm_table = &(data->smc_state_table); + struct phm_cac_tdp_table *cac_dtp_table = hwmgr->dyn_state.cac_dtp_table; + struct phm_ppm_table *ppm = hwmgr->dyn_state.ppm_parameter_table; + uint16_t *def1, *def2; + int i, j, k; + + /* + * TDP number of fraction bits are changed from 8 to 7 for Iceland + * as requested by SMC team + */ + dpm_table->DefaultTdp = PP_HOST_TO_SMC_US((uint16_t)(cac_dtp_table->usTDP * 256)); + dpm_table->TargetTdp = PP_HOST_TO_SMC_US((uint16_t)(cac_dtp_table->usConfigurableTDP * 256)); + + dpm_table->DTETjOffset = (uint8_t)data->ul_dte_tj_offset; + + dpm_table->GpuTjMax = (uint8_t)(data->thermal_temp_setting.temperature_high / PP_TEMPERATURE_UNITS_PER_CENTIGRADES); + dpm_table->GpuTjHyst = 8; + + dpm_table->DTEAmbientTempBase = defaults->dte_ambient_temp_base; + + /* The following are for new Iceland Multi-input fan/thermal control */ + if(NULL != ppm) { + dpm_table->PPM_PkgPwrLimit = (uint16_t)ppm->dgpu_tdp * 256 / 1000; + dpm_table->PPM_TemperatureLimit = (uint16_t)ppm->tj_max * 256; + } else { + dpm_table->PPM_PkgPwrLimit = 0; + dpm_table->PPM_TemperatureLimit = 0; + } + + CONVERT_FROM_HOST_TO_SMC_US(dpm_table->PPM_PkgPwrLimit); + CONVERT_FROM_HOST_TO_SMC_US(dpm_table->PPM_TemperatureLimit); + + dpm_table->BAPM_TEMP_GRADIENT = PP_HOST_TO_SMC_UL(defaults->bamp_temp_gradient); + def1 = defaults->bapmti_r; + def2 = defaults->bapmti_rc; + + for (i = 0; i < SMU71_DTE_ITERATIONS; i++) { + for (j = 0; j < SMU71_DTE_SOURCES; j++) { + for (k = 0; k < SMU71_DTE_SINKS; k++) { + dpm_table->BAPMTI_R[i][j][k] = PP_HOST_TO_SMC_US(*def1); + dpm_table->BAPMTI_RC[i][j][k] = PP_HOST_TO_SMC_US(*def2); + def1++; + def2++; + } + } + } + + return 0; +} + +static int iceland_populate_svi_load_line(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + const struct iceland_pt_defaults *defaults = data->power_tune_defaults; + + data->power_tune_table.SviLoadLineEn = defaults->svi_load_line_en; + data->power_tune_table.SviLoadLineVddC = defaults->svi_load_line_vddc; + data->power_tune_table.SviLoadLineTrimVddC = 3; + data->power_tune_table.SviLoadLineOffsetVddC = 0; + + return 0; +} + +static int iceland_populate_tdc_limit(struct pp_hwmgr *hwmgr) +{ + uint16_t tdc_limit; + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + const struct iceland_pt_defaults *defaults = data->power_tune_defaults; + + /* TDC number of fraction bits are changed from 8 to 7 + * for Iceland as requested by SMC team + */ + tdc_limit = (uint16_t)(hwmgr->dyn_state.cac_dtp_table->usTDC * 256); + data->power_tune_table.TDC_VDDC_PkgLimit = + CONVERT_FROM_HOST_TO_SMC_US(tdc_limit); + data->power_tune_table.TDC_VDDC_ThrottleReleaseLimitPerc = + defaults->tdc_vddc_throttle_release_limit_perc; + data->power_tune_table.TDC_MAWt = defaults->tdc_mawt; + + return 0; +} + +static int iceland_populate_dw8(struct pp_hwmgr *hwmgr, uint32_t fuse_table_offset) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + const struct iceland_pt_defaults *defaults = data->power_tune_defaults; + uint32_t temp; + + if (iceland_read_smc_sram_dword(hwmgr->smumgr, + fuse_table_offset + + offsetof(SMU71_Discrete_PmFuses, TdcWaterfallCtl), + (uint32_t *)&temp, data->sram_end)) + PP_ASSERT_WITH_CODE(false, + "Attempt to read PmFuses.DW6 (SviLoadLineEn) from SMC Failed!", + return -EINVAL); + else + data->power_tune_table.TdcWaterfallCtl = defaults->tdc_waterfall_ctl; + + return 0; +} + +static int iceland_populate_temperature_scaler(struct pp_hwmgr *hwmgr) +{ + return 0; +} + +static int iceland_populate_gnb_lpml(struct pp_hwmgr *hwmgr) +{ + int i; + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + /* Currently not used. Set all to zero. */ + for (i = 0; i < 8; i++) + data->power_tune_table.GnbLPML[i] = 0; + + return 0; +} + +static int iceland_min_max_vgnb_lpml_id_from_bapm_vddc(struct pp_hwmgr *hwmgr) +{ + return 0; +} + +static int iceland_populate_bapm_vddc_base_leakage_sidd(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + uint16_t HiSidd = data->power_tune_table.BapmVddCBaseLeakageHiSidd; + uint16_t LoSidd = data->power_tune_table.BapmVddCBaseLeakageLoSidd; + struct phm_cac_tdp_table *cac_table = hwmgr->dyn_state.cac_dtp_table; + + HiSidd = (uint16_t)(cac_table->usHighCACLeakage / 100 * 256); + LoSidd = (uint16_t)(cac_table->usLowCACLeakage / 100 * 256); + + data->power_tune_table.BapmVddCBaseLeakageHiSidd = + CONVERT_FROM_HOST_TO_SMC_US(HiSidd); + data->power_tune_table.BapmVddCBaseLeakageLoSidd = + CONVERT_FROM_HOST_TO_SMC_US(LoSidd); + + return 0; +} + +int iceland_populate_pm_fuses(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + uint32_t pm_fuse_table_offset; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment)) { + if (iceland_read_smc_sram_dword(hwmgr->smumgr, + SMU71_FIRMWARE_HEADER_LOCATION + + offsetof(SMU71_Firmware_Header, PmFuseTable), + &pm_fuse_table_offset, data->sram_end)) + PP_ASSERT_WITH_CODE(false, + "Attempt to get pm_fuse_table_offset Failed!", + return -EINVAL); + + /* DW0 - DW3 */ + if (iceland_populate_bapm_vddc_vid_sidd(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate bapm vddc vid Failed!", + return -EINVAL); + + /* DW4 - DW5 */ + if (iceland_populate_vddc_vid(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate vddc vid Failed!", + return -EINVAL); + + /* DW6 */ + if (iceland_populate_svi_load_line(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate SviLoadLine Failed!", + return -EINVAL); + /* DW7 */ + if (iceland_populate_tdc_limit(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate TDCLimit Failed!", return -EINVAL); + /* DW8 */ + if (iceland_populate_dw8(hwmgr, pm_fuse_table_offset)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate TdcWaterfallCtl, " + "LPMLTemperature Min and Max Failed!", + return -EINVAL); + + /* DW9-DW12 */ + if (0 != iceland_populate_temperature_scaler(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate LPMLTemperatureScaler Failed!", + return -EINVAL); + + /* DW13-DW16 */ + if (iceland_populate_gnb_lpml(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate GnbLPML Failed!", + return -EINVAL); + + /* DW17 */ + if (iceland_min_max_vgnb_lpml_id_from_bapm_vddc(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate GnbLPML Min and Max Vid Failed!", + return -EINVAL); + + /* DW18 */ + if (iceland_populate_bapm_vddc_base_leakage_sidd(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate BapmVddCBaseLeakage Hi and Lo Sidd Failed!", + return -EINVAL); + + if (iceland_copy_bytes_to_smc(hwmgr->smumgr, pm_fuse_table_offset, + (uint8_t *)&data->power_tune_table, + sizeof(struct SMU71_Discrete_PmFuses), data->sram_end)) + PP_ASSERT_WITH_CODE(false, + "Attempt to download PmFuseTable Failed!", + return -EINVAL); + } + return 0; +} + +int iceland_enable_smc_cac(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + int result = 0; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_CAC)) { + int smc_result; + smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_EnableCac)); + PP_ASSERT_WITH_CODE((0 == smc_result), + "Failed to enable CAC in SMC.", result = -1); + + data->cac_enabled = (0 == smc_result) ? true : false; + } + return result; +} + +static int iceland_set_power_limit(struct pp_hwmgr *hwmgr, uint32_t n) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + + if(data->power_containment_features & + POWERCONTAINMENT_FEATURE_PkgPwrLimit) + return smum_send_msg_to_smc_with_parameter(hwmgr->smumgr, + PPSMC_MSG_PkgPwrSetLimit, n); + return 0; +} + +static int iceland_set_overdriver_target_tdp(struct pp_hwmgr *pHwMgr, uint32_t target_tdp) +{ + return smum_send_msg_to_smc_with_parameter(pHwMgr->smumgr, + PPSMC_MSG_OverDriveSetTargetTdp, target_tdp); +} + +int iceland_enable_power_containment(struct pp_hwmgr *hwmgr) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + SMU71_Discrete_DpmTable *dpm_table = &data->smc_state_table; + int smc_result; + int result = 0; + uint32_t is_asic_kicker; + + data->power_containment_features = 0; + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment)) { + is_asic_kicker = cgs_read_register(hwmgr->device, mmCC_BIF_BX_STRAP2); + is_asic_kicker = (is_asic_kicker >> 12) & 0x01; + + if (data->enable_bapm_feature && + (!is_asic_kicker || + phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DisableUsingActualTemperatureForPowerCalc))) { + smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_EnableDTE)); + PP_ASSERT_WITH_CODE((0 == smc_result), + "Failed to enable BAPM in SMC.", result = -1;); + if (0 == smc_result) + data->power_containment_features |= POWERCONTAINMENT_FEATURE_BAPM; + } + + if (is_asic_kicker && !phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DisableUsingActualTemperatureForPowerCalc)) + dpm_table->DTEMode = 2; + + if (data->enable_tdc_limit_feature) { + smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_TDCLimitEnable)); + PP_ASSERT_WITH_CODE((0 == smc_result), + "Failed to enable TDCLimit in SMC.", result = -1;); + if (0 == smc_result) + data->power_containment_features |= + POWERCONTAINMENT_FEATURE_TDCLimit; + } + + if (data->enable_pkg_pwr_tracking_feature) { + smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_PkgPwrLimitEnable)); + PP_ASSERT_WITH_CODE((0 == smc_result), + "Failed to enable PkgPwrTracking in SMC.", result = -1;); + if (0 == smc_result) { + struct phm_cac_tdp_table *cac_table = + hwmgr->dyn_state.cac_dtp_table; + uint32_t default_limit = + (uint32_t)(cac_table->usMaximumPowerDeliveryLimit * 256); + + data->power_containment_features |= + POWERCONTAINMENT_FEATURE_PkgPwrLimit; + + if (iceland_set_power_limit(hwmgr, default_limit)) + printk(KERN_ERR "Failed to set Default Power Limit in SMC!"); + } + } + } + return result; +} + +int iceland_power_control_set_level(struct pp_hwmgr *hwmgr) +{ + struct phm_cac_tdp_table *cac_table = hwmgr->dyn_state.cac_dtp_table; + int adjust_percent, target_tdp; + int result = 0; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment)) { + /* adjustment percentage has already been validated */ + adjust_percent = hwmgr->platform_descriptor.TDPAdjustmentPolarity ? + hwmgr->platform_descriptor.TDPAdjustment : + (-1 * hwmgr->platform_descriptor.TDPAdjustment); + /* + * SMC requested that target_tdp to be 7 bit fraction in DPM table + * but message to be 8 bit fraction for messages + */ + target_tdp = ((100 + adjust_percent) * (int)(cac_table->usTDP * 256)) / 100; + result = iceland_set_overdriver_target_tdp(hwmgr, (uint32_t)target_tdp); + } + + return result; +} diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_powertune.h b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_powertune.h new file mode 100644 index 000000000000..6c25ee139ca3 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_powertune.h @@ -0,0 +1,74 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Author: Huang Rui <ray.huang@amd.com> + * + */ +#ifndef ICELAND_POWERTUNE_H +#define ICELAND_POWERTUNE_H + +#include "smu71.h" + +enum iceland_pt_config_reg_type { + ICELAND_CONFIGREG_MMR = 0, + ICELAND_CONFIGREG_SMC_IND, + ICELAND_CONFIGREG_DIDT_IND, + ICELAND_CONFIGREG_CACHE, + ICELAND_CONFIGREG_MAX +}; + +/* PowerContainment Features */ +#define POWERCONTAINMENT_FEATURE_DTE 0x00000001 +#define POWERCONTAINMENT_FEATURE_TDCLimit 0x00000002 +#define POWERCONTAINMENT_FEATURE_PkgPwrLimit 0x00000004 +#define POWERCONTAINMENT_FEATURE_BAPM 0x00000001 + +struct iceland_pt_config_reg { + uint32_t offset; + uint32_t mask; + uint32_t shift; + uint32_t value; + enum iceland_pt_config_reg_type type; +}; + +struct iceland_pt_defaults +{ + uint8_t svi_load_line_en; + uint8_t svi_load_line_vddc; + uint8_t tdc_vddc_throttle_release_limit_perc; + uint8_t tdc_mawt; + uint8_t tdc_waterfall_ctl; + uint8_t dte_ambient_temp_base; + uint32_t display_cac; + uint32_t bamp_temp_gradient; + uint16_t bapmti_r[SMU71_DTE_ITERATIONS * SMU71_DTE_SOURCES * SMU71_DTE_SINKS]; + uint16_t bapmti_rc[SMU71_DTE_ITERATIONS * SMU71_DTE_SOURCES * SMU71_DTE_SINKS]; +}; + +void iceland_initialize_power_tune_defaults(struct pp_hwmgr *hwmgr); +int iceland_populate_bapm_parameters_in_dpm_table(struct pp_hwmgr *hwmgr); +int iceland_populate_pm_fuses(struct pp_hwmgr *hwmgr); +int iceland_enable_smc_cac(struct pp_hwmgr *hwmgr); +int iceland_enable_power_containment(struct pp_hwmgr *hwmgr); +int iceland_power_control_set_level(struct pp_hwmgr *hwmgr); + +#endif /* ICELAND_POWERTUNE_H */ + diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_thermal.c b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_thermal.c new file mode 100644 index 000000000000..527f37022424 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_thermal.c @@ -0,0 +1,595 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Author: Huang Rui <ray.huang@amd.com> + * + */ +#include <asm/div64.h> +#include "iceland_thermal.h" +#include "iceland_hwmgr.h" +#include "iceland_smumgr.h" +#include "atombios.h" +#include "ppsmc.h" + +#include "gmc/gmc_8_1_d.h" +#include "gmc/gmc_8_1_sh_mask.h" + +#include "bif/bif_5_0_d.h" +#include "bif/bif_5_0_sh_mask.h" + +#include "smu/smu_7_1_1_d.h" +#include "smu/smu_7_1_1_sh_mask.h" + + +/** +* Get Fan Speed Control Parameters. +* @param hwmgr the address of the powerplay hardware manager. +* @param pSpeed is the address of the structure where the result is to be placed. +* @exception Always succeeds except if we cannot zero out the output structure. +*/ +int iceland_fan_ctrl_get_fan_speed_info(struct pp_hwmgr *hwmgr, + struct phm_fan_speed_info *fan_speed_info) +{ + + if (hwmgr->thermal_controller.fanInfo.bNoFan) + return 0; + + fan_speed_info->supports_percent_read = true; + fan_speed_info->supports_percent_write = true; + fan_speed_info->min_percent = 0; + fan_speed_info->max_percent = 100; + + if (0 != hwmgr->thermal_controller.fanInfo.ucTachometerPulsesPerRevolution) { + fan_speed_info->supports_rpm_read = true; + fan_speed_info->supports_rpm_write = true; + fan_speed_info->min_rpm = hwmgr->thermal_controller.fanInfo.ulMinRPM; + fan_speed_info->max_rpm = hwmgr->thermal_controller.fanInfo.ulMaxRPM; + } else { + fan_speed_info->min_rpm = 0; + fan_speed_info->max_rpm = 0; + } + + return 0; +} + +/** +* Get Fan Speed in percent. +* @param hwmgr the address of the powerplay hardware manager. +* @param pSpeed is the address of the structure where the result is to be placed. +* @exception Fails is the 100% setting appears to be 0. +*/ +int iceland_fan_ctrl_get_fan_speed_percent(struct pp_hwmgr *hwmgr, uint32_t *speed) +{ + uint32_t duty100; + uint32_t duty; + uint64_t tmp64; + + if (hwmgr->thermal_controller.fanInfo.bNoFan) + return 0; + + duty100 = PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_FDO_CTRL1, FMAX_DUTY100); + duty = PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_THERMAL_STATUS, FDO_PWM_DUTY); + + if (0 == duty100) + return -EINVAL; + + + tmp64 = (uint64_t)duty * 100; + do_div(tmp64, duty100); + *speed = (uint32_t)tmp64; + + if (*speed > 100) + *speed = 100; + + return 0; +} + +/** +* Get Fan Speed in RPM. +* @param hwmgr the address of the powerplay hardware manager. +* @param speed is the address of the structure where the result is to be placed. +* @exception Returns not supported if no fan is found or if pulses per revolution are not set +*/ +int iceland_fan_ctrl_get_fan_speed_rpm(struct pp_hwmgr *hwmgr, uint32_t *speed) +{ + return 0; +} + +/** +* Set Fan Speed Control to static mode, so that the user can decide what speed to use. +* @param hwmgr the address of the powerplay hardware manager. +* mode the fan control mode, 0 default, 1 by percent, 5, by RPM +* @exception Should always succeed. +*/ +int iceland_fan_ctrl_set_static_mode(struct pp_hwmgr *hwmgr, uint32_t mode) +{ + + if (hwmgr->fan_ctrl_is_in_default_mode) { + hwmgr->fan_ctrl_default_mode = PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_FDO_CTRL2, FDO_PWM_MODE); + hwmgr->tmin = PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_FDO_CTRL2, TMIN); + hwmgr->fan_ctrl_is_in_default_mode = false; + } + + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_FDO_CTRL2, TMIN, 0); + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_FDO_CTRL2, FDO_PWM_MODE, mode); + + return 0; +} + +/** +* Reset Fan Speed Control to default mode. +* @param hwmgr the address of the powerplay hardware manager. +* @exception Should always succeed. +*/ +static int iceland_fan_ctrl_set_default_mode(struct pp_hwmgr *hwmgr) +{ + if (!hwmgr->fan_ctrl_is_in_default_mode) { + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_FDO_CTRL2, FDO_PWM_MODE, hwmgr->fan_ctrl_default_mode); + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_FDO_CTRL2, TMIN, hwmgr->tmin); + hwmgr->fan_ctrl_is_in_default_mode = true; + } + + return 0; +} + +int iceland_fan_ctrl_start_smc_fan_control(struct pp_hwmgr *hwmgr) +{ + return (smum_send_msg_to_smc(hwmgr->smumgr, PPSMC_StartFanControl) == 0) ? 0 : -EINVAL; +} + + +int iceland_fan_ctrl_stop_smc_fan_control(struct pp_hwmgr *hwmgr) +{ + return (smum_send_msg_to_smc(hwmgr->smumgr, PPSMC_StopFanControl) == 0) ? 0 : -EINVAL; +} + +/** +* Set Fan Speed in percent. +* @param hwmgr the address of the powerplay hardware manager. +* @param speed is the percentage value (0% - 100%) to be set. +* @exception Fails is the 100% setting appears to be 0. +*/ +int iceland_fan_ctrl_set_fan_speed_percent(struct pp_hwmgr *hwmgr, uint32_t speed) +{ + uint32_t duty100; + uint32_t duty; + uint64_t tmp64; + + if (hwmgr->thermal_controller.fanInfo.bNoFan) + return -EINVAL; + + if (speed > 100) { + pr_warning("Cannot set more than 100%% duty cycle. Set it to 100.\n"); + speed = 100; + } + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_MicrocodeFanControl)) + iceland_fan_ctrl_stop_smc_fan_control(hwmgr); + + duty100 = PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_FDO_CTRL1, FMAX_DUTY100); + + if (0 == duty100) + return -EINVAL; + + tmp64 = (uint64_t)speed * duty100; + do_div(tmp64, 100); + duty = (uint32_t)tmp64; + + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_FDO_CTRL0, FDO_STATIC_DUTY, duty); + + return iceland_fan_ctrl_set_static_mode(hwmgr, FDO_PWM_MODE_STATIC); +} + +/** +* Reset Fan Speed to default. +* @param hwmgr the address of the powerplay hardware manager. +* @exception Always succeeds. +*/ +int iceland_fan_ctrl_reset_fan_speed_to_default(struct pp_hwmgr *hwmgr) +{ + int result; + + if (hwmgr->thermal_controller.fanInfo.bNoFan) + return 0; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_MicrocodeFanControl)) { + result = iceland_fan_ctrl_set_static_mode(hwmgr, FDO_PWM_MODE_STATIC); + if (0 == result) + result = iceland_fan_ctrl_start_smc_fan_control(hwmgr); + } else + result = iceland_fan_ctrl_set_default_mode(hwmgr); + + return result; +} + +/** +* Set Fan Speed in RPM. +* @param hwmgr the address of the powerplay hardware manager. +* @param speed is the percentage value (min - max) to be set. +* @exception Fails is the speed not lie between min and max. +*/ +int iceland_fan_ctrl_set_fan_speed_rpm(struct pp_hwmgr *hwmgr, uint32_t speed) +{ + return 0; +} + +/** +* Reads the remote temperature from the SIslands thermal controller. +* +* @param hwmgr The address of the hardware manager. +*/ +int iceland_thermal_get_temperature(struct pp_hwmgr *hwmgr) +{ + int temp; + + temp = PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_MULT_THERMAL_STATUS, CTF_TEMP); + + /* + * Bit 9 means the reading is lower than the lowest usable + * value. + */ + if (0 != (0x200 & temp)) + temp = ICELAND_THERMAL_MAXIMUM_TEMP_READING; + else + temp = (temp & 0x1ff); + + temp = temp * PP_TEMPERATURE_UNITS_PER_CENTIGRADES; + + return temp; +} + +/** +* Set the requested temperature range for high and low alert signals +* +* @param hwmgr The address of the hardware manager. +* @param range Temperature range to be programmed for high and low alert signals +* @exception PP_Result_BadInput if the input data is not valid. +*/ +static int iceland_thermal_set_temperature_range(struct pp_hwmgr *hwmgr, uint32_t low_temp, uint32_t high_temp) +{ + uint32_t low = ICELAND_THERMAL_MINIMUM_ALERT_TEMP * PP_TEMPERATURE_UNITS_PER_CENTIGRADES; + uint32_t high = ICELAND_THERMAL_MAXIMUM_ALERT_TEMP * PP_TEMPERATURE_UNITS_PER_CENTIGRADES; + + if (low < low_temp) + low = low_temp; + if (high > high_temp) + high = high_temp; + + if (low > high) + return -EINVAL; + + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_THERMAL_INT, DIG_THERM_INTH, (high / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)); + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_THERMAL_INT, DIG_THERM_INTL, (low / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)); + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_THERMAL_CTRL, DIG_THERM_DPM, (high / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)); + + return 0; +} + +/** +* Programs thermal controller one-time setting registers +* +* @param hwmgr The address of the hardware manager. +*/ +static int iceland_thermal_initialize(struct pp_hwmgr *hwmgr) +{ + if (0 != hwmgr->thermal_controller.fanInfo.ucTachometerPulsesPerRevolution) + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + CG_TACH_CTRL, EDGE_PER_REV, + hwmgr->thermal_controller.fanInfo.ucTachometerPulsesPerRevolution - 1); + + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_FDO_CTRL2, TACH_PWM_RESP_RATE, 0x28); + + return 0; +} + +/** +* Enable thermal alerts on the RV770 thermal controller. +* +* @param hwmgr The address of the hardware manager. +*/ +static int iceland_thermal_enable_alert(struct pp_hwmgr *hwmgr) +{ + uint32_t alert; + + alert = PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_THERMAL_INT, THERM_INT_MASK); + alert &= ~(ICELAND_THERMAL_HIGH_ALERT_MASK | ICELAND_THERMAL_LOW_ALERT_MASK); + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_THERMAL_INT, THERM_INT_MASK, alert); + + /* send message to SMU to enable internal thermal interrupts */ + return (smum_send_msg_to_smc(hwmgr->smumgr, PPSMC_MSG_Thermal_Cntl_Enable) == 0) ? 0 : -1; +} + +/** +* Disable thermal alerts on the RV770 thermal controller. +* @param hwmgr The address of the hardware manager. +*/ +static int iceland_thermal_disable_alert(struct pp_hwmgr *hwmgr) +{ + uint32_t alert; + + alert = PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_THERMAL_INT, THERM_INT_MASK); + alert |= (ICELAND_THERMAL_HIGH_ALERT_MASK | ICELAND_THERMAL_LOW_ALERT_MASK); + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_THERMAL_INT, THERM_INT_MASK, alert); + + /* send message to SMU to disable internal thermal interrupts */ + return (smum_send_msg_to_smc(hwmgr->smumgr, PPSMC_MSG_Thermal_Cntl_Disable) == 0) ? 0 : -1; +} + +/** +* Uninitialize the thermal controller. +* Currently just disables alerts. +* @param hwmgr The address of the hardware manager. +*/ +int iceland_thermal_stop_thermal_controller(struct pp_hwmgr *hwmgr) +{ + int result = iceland_thermal_disable_alert(hwmgr); + + if (result) + pr_warning("Failed to disable thermal alerts!\n"); + + if (hwmgr->thermal_controller.fanInfo.bNoFan) + iceland_fan_ctrl_set_default_mode(hwmgr); + + return result; +} + +/** +* Set up the fan table to control the fan using the SMC. +* @param hwmgr the address of the powerplay hardware manager. +* @param pInput the pointer to input data +* @param pOutput the pointer to output data +* @param pStorage the pointer to temporary storage +* @param Result the last failure code +* @return result from set temperature range routine +*/ +int tf_iceland_thermal_setup_fan_table(struct pp_hwmgr *hwmgr, void *input, void *output, void *storage, int result) +{ + struct iceland_hwmgr *data = (struct iceland_hwmgr *)(hwmgr->backend); + SMU71_Discrete_FanTable fan_table = { FDO_MODE_HARDWARE }; + uint32_t duty100; + uint32_t t_diff1, t_diff2, pwm_diff1, pwm_diff2; + uint16_t fdo_min, slope1, slope2; + uint32_t reference_clock; + int res; + uint64_t tmp64; + + if (!phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_MicrocodeFanControl)) + return 0; + + if (0 == data->fan_table_start) { + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_MicrocodeFanControl); + return 0; + } + + duty100 = PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_FDO_CTRL1, FMAX_DUTY100); + + if (0 == duty100) { + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_MicrocodeFanControl); + return 0; + } + + tmp64 = hwmgr->thermal_controller.advanceFanControlParameters.usPWMMin * duty100; + do_div(tmp64, 10000); + fdo_min = (uint16_t)tmp64; + + t_diff1 = hwmgr->thermal_controller.advanceFanControlParameters.usTMed - hwmgr->thermal_controller.advanceFanControlParameters.usTMin; + t_diff2 = hwmgr->thermal_controller.advanceFanControlParameters.usTHigh - hwmgr->thermal_controller.advanceFanControlParameters.usTMed; + + pwm_diff1 = hwmgr->thermal_controller.advanceFanControlParameters.usPWMMed - hwmgr->thermal_controller.advanceFanControlParameters.usPWMMin; + pwm_diff2 = hwmgr->thermal_controller.advanceFanControlParameters.usPWMHigh - hwmgr->thermal_controller.advanceFanControlParameters.usPWMMed; + + slope1 = (uint16_t)((50 + ((16 * duty100 * pwm_diff1) / t_diff1)) / 100); + slope2 = (uint16_t)((50 + ((16 * duty100 * pwm_diff2) / t_diff2)) / 100); + + fan_table.TempMin = cpu_to_be16((50 + hwmgr->thermal_controller.advanceFanControlParameters.usTMin) / 100); + fan_table.TempMed = cpu_to_be16((50 + hwmgr->thermal_controller.advanceFanControlParameters.usTMed) / 100); + fan_table.TempMax = cpu_to_be16((50 + hwmgr->thermal_controller.advanceFanControlParameters.usTMax) / 100); + + fan_table.Slope1 = cpu_to_be16(slope1); + fan_table.Slope2 = cpu_to_be16(slope2); + + fan_table.FdoMin = cpu_to_be16(fdo_min); + + fan_table.HystDown = cpu_to_be16(hwmgr->thermal_controller.advanceFanControlParameters.ucTHyst); + + fan_table.HystUp = cpu_to_be16(1); + + fan_table.HystSlope = cpu_to_be16(1); + + fan_table.TempRespLim = cpu_to_be16(5); + + reference_clock = iceland_get_xclk(hwmgr); + + fan_table.RefreshPeriod = cpu_to_be32((hwmgr->thermal_controller.advanceFanControlParameters.ulCycleDelay * reference_clock) / 1600); + + fan_table.FdoMax = cpu_to_be16((uint16_t)duty100); + + fan_table.TempSrc = (uint8_t)PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, CG_MULT_THERMAL_CTRL, TEMP_SEL); + + //fan_table.FanControl_GL_Flag = 1; + + res = iceland_copy_bytes_to_smc(hwmgr->smumgr, data->fan_table_start, (uint8_t *)&fan_table, (uint32_t)sizeof(fan_table), data->sram_end); +/* TO DO FOR SOME DEVICE ID 0X692b, send this msg return invalid command. + if (res == 0 && hwmgr->thermal_controller.advanceFanControlParameters.ucMinimumPWMLimit != 0) + res = (0 == smum_send_msg_to_smc_with_parameter(hwmgr->smumgr, PPSMC_MSG_SetFanMinPwm, \ + hwmgr->thermal_controller.advanceFanControlParameters.ucMinimumPWMLimit) ? 0 : -1); + + if (res == 0 && hwmgr->thermal_controller.advanceFanControlParameters.ulMinFanSCLKAcousticLimit != 0) + res = (0 == smum_send_msg_to_smc_with_parameter(hwmgr->smumgr, PPSMC_MSG_SetFanSclkTarget, \ + hwmgr->thermal_controller.advanceFanControlParameters.ulMinFanSCLKAcousticLimit) ? 0 : -1); + + if (0 != res) + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_MicrocodeFanControl); +*/ + return 0; +} + +/** +* Start the fan control on the SMC. +* @param hwmgr the address of the powerplay hardware manager. +* @param pInput the pointer to input data +* @param pOutput the pointer to output data +* @param pStorage the pointer to temporary storage +* @param Result the last failure code +* @return result from set temperature range routine +*/ +int tf_iceland_thermal_start_smc_fan_control(struct pp_hwmgr *hwmgr, void *input, void *output, void *storage, int result) +{ +/* If the fantable setup has failed we could have disabled PHM_PlatformCaps_MicrocodeFanControl even after this function was included in the table. + * Make sure that we still think controlling the fan is OK. +*/ + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_MicrocodeFanControl)) { + iceland_fan_ctrl_start_smc_fan_control(hwmgr); + iceland_fan_ctrl_set_static_mode(hwmgr, FDO_PWM_MODE_STATIC); + } + + return 0; +} + +/** +* Set temperature range for high and low alerts +* @param hwmgr the address of the powerplay hardware manager. +* @param pInput the pointer to input data +* @param pOutput the pointer to output data +* @param pStorage the pointer to temporary storage +* @param Result the last failure code +* @return result from set temperature range routine +*/ +static int tf_iceland_thermal_set_temperature_range(struct pp_hwmgr *hwmgr, + void *input, void *output, void *storage, int result) +{ + struct PP_TemperatureRange *range = (struct PP_TemperatureRange *)input; + + if (range == NULL) + return -EINVAL; + + return iceland_thermal_set_temperature_range(hwmgr, range->min, range->max); +} + +/** +* Programs one-time setting registers +* @param hwmgr the address of the powerplay hardware manager. +* @param pInput the pointer to input data +* @param pOutput the pointer to output data +* @param pStorage the pointer to temporary storage +* @param Result the last failure code +* @return result from initialize thermal controller routine +*/ +static int tf_iceland_thermal_initialize(struct pp_hwmgr *hwmgr, void *input, + void *output, void *storage, int result) +{ + return iceland_thermal_initialize(hwmgr); +} + +/** +* Enable high and low alerts +* @param hwmgr the address of the powerplay hardware manager. +* @param pInput the pointer to input data +* @param pOutput the pointer to output data +* @param pStorage the pointer to temporary storage +* @param Result the last failure code +* @return result from enable alert routine +*/ +static int tf_iceland_thermal_enable_alert(struct pp_hwmgr *hwmgr, + void *input, void *output, void *storage, int result) +{ + return iceland_thermal_enable_alert(hwmgr); +} + +/** +* Disable high and low alerts +* @param hwmgr the address of the powerplay hardware manager. +* @param pInput the pointer to input data +* @param pOutput the pointer to output data +* @param pStorage the pointer to temporary storage +* @param Result the last failure code +* @return result from disable alert routine +*/ +static int tf_iceland_thermal_disable_alert(struct pp_hwmgr *hwmgr, void *input, void *output, void *storage, int result) +{ + return iceland_thermal_disable_alert(hwmgr); +} + +static const struct phm_master_table_item iceland_thermal_start_thermal_controller_master_list[] = { + { NULL, tf_iceland_thermal_initialize }, + { NULL, tf_iceland_thermal_set_temperature_range }, + { NULL, tf_iceland_thermal_enable_alert }, + /* + * We should restrict performance levels to low before we halt + * the SMC. On the other hand we are still in boot state when + * we do this so it would be pointless. If this assumption + * changes we have to revisit this table. + */ + { NULL, tf_iceland_thermal_setup_fan_table}, + { NULL, tf_iceland_thermal_start_smc_fan_control}, + { NULL, NULL } +}; + +static const struct phm_master_table_header iceland_thermal_start_thermal_controller_master = { + 0, + PHM_MasterTableFlag_None, + iceland_thermal_start_thermal_controller_master_list +}; + +static const struct phm_master_table_item iceland_thermal_set_temperature_range_master_list[] = { + { NULL, tf_iceland_thermal_disable_alert}, + { NULL, tf_iceland_thermal_set_temperature_range}, + { NULL, tf_iceland_thermal_enable_alert}, + { NULL, NULL } +}; + +static const struct phm_master_table_header iceland_thermal_set_temperature_range_master = { + 0, + PHM_MasterTableFlag_None, + iceland_thermal_set_temperature_range_master_list +}; + +int iceland_thermal_ctrl_uninitialize_thermal_controller(struct pp_hwmgr *hwmgr) +{ + if (!hwmgr->thermal_controller.fanInfo.bNoFan) + iceland_fan_ctrl_set_default_mode(hwmgr); + return 0; +} + +/** +* Initializes the thermal controller related functions in the Hardware Manager structure. +* @param hwmgr The address of the hardware manager. +* @exception Any error code from the low-level communication. +*/ +int pp_iceland_thermal_initialize(struct pp_hwmgr *hwmgr) +{ + int result; + + result = phm_construct_table(hwmgr, &iceland_thermal_set_temperature_range_master, &(hwmgr->set_temperature_range)); + + if (0 == result) { + result = phm_construct_table(hwmgr, + &iceland_thermal_start_thermal_controller_master, + &(hwmgr->start_thermal_controller)); + if (0 != result) + phm_destroy_table(hwmgr, &(hwmgr->set_temperature_range)); + } + + if (0 == result) + hwmgr->fan_ctrl_is_in_default_mode = true; + return result; +} + diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_thermal.h b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_thermal.h new file mode 100644 index 000000000000..267945f4df71 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/iceland_thermal.h @@ -0,0 +1,58 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Author: Huang Rui <ray.huang@amd.com> + * + */ + +#ifndef ICELAND_THERMAL_H +#define ICELAND_THERMAL_H + +#include "hwmgr.h" + +#define ICELAND_THERMAL_HIGH_ALERT_MASK 0x1 +#define ICELAND_THERMAL_LOW_ALERT_MASK 0x2 + +#define ICELAND_THERMAL_MINIMUM_TEMP_READING -256 +#define ICELAND_THERMAL_MAXIMUM_TEMP_READING 255 + +#define ICELAND_THERMAL_MINIMUM_ALERT_TEMP 0 +#define ICELAND_THERMAL_MAXIMUM_ALERT_TEMP 255 + +#define FDO_PWM_MODE_STATIC 1 +#define FDO_PWM_MODE_STATIC_RPM 5 + + +extern int iceland_thermal_get_temperature(struct pp_hwmgr *hwmgr); +extern int iceland_thermal_stop_thermal_controller(struct pp_hwmgr *hwmgr); +extern int iceland_fan_ctrl_get_fan_speed_info(struct pp_hwmgr *hwmgr, struct phm_fan_speed_info *fan_speed_info); +extern int iceland_fan_ctrl_get_fan_speed_percent(struct pp_hwmgr *hwmgr, uint32_t *speed); +extern int iceland_fan_ctrl_set_static_mode(struct pp_hwmgr *hwmgr, uint32_t mode); +extern int iceland_fan_ctrl_set_fan_speed_percent(struct pp_hwmgr *hwmgr, uint32_t speed); +extern int iceland_fan_ctrl_reset_fan_speed_to_default(struct pp_hwmgr *hwmgr); +extern int pp_iceland_thermal_initialize(struct pp_hwmgr *hwmgr); +extern int iceland_thermal_ctrl_uninitialize_thermal_controller(struct pp_hwmgr *hwmgr); +extern int iceland_fan_ctrl_set_fan_speed_rpm(struct pp_hwmgr *hwmgr, uint32_t speed); +extern int iceland_fan_ctrl_get_fan_speed_rpm(struct pp_hwmgr *hwmgr, uint32_t *speed); +extern int iceland_fan_ctrl_stop_smc_fan_control(struct pp_hwmgr *hwmgr); + +#endif + diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/polaris10_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/polaris10_hwmgr.c index 783dcc3384f1..b69132296672 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/polaris10_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/polaris10_hwmgr.c @@ -97,19 +97,6 @@ #define PCIE_BUS_CLK 10000 #define TCLK (PCIE_BUS_CLK / 10) - -static const uint16_t polaris10_clock_stretcher_lookup_table[2][4] = -{ {600, 1050, 3, 0}, {600, 1050, 6, 1} }; - -/* [FF, SS] type, [] 4 voltage ranges, and [Floor Freq, Boundary Freq, VID min , VID max] */ -static const uint32_t polaris10_clock_stretcher_ddt_table[2][4][4] = -{ { {265, 529, 120, 128}, {325, 650, 96, 119}, {430, 860, 32, 95}, {0, 0, 0, 31} }, - { {275, 550, 104, 112}, {319, 638, 96, 103}, {360, 720, 64, 95}, {384, 768, 32, 63} } }; - -/* [Use_For_Low_freq] value, [0%, 5%, 10%, 7.14%, 14.28%, 20%] (coming from PWR_CKS_CNTL.stretch_amount reg spec) */ -static const uint8_t polaris10_clock_stretch_amount_conversion[2][6] = -{ {0, 1, 3, 2, 4, 5}, {0, 2, 4, 5, 6, 5} }; - /** Values for the CG_THERMAL_CTRL::DPM_EVENT_SRC field. */ enum DPM_EVENT_SRC { DPM_EVENT_SRC_ANALOG = 0, @@ -2772,9 +2759,6 @@ int polaris10_set_features_platform_caps(struct pp_hwmgr *hwmgr) struct polaris10_hwmgr *data = (struct polaris10_hwmgr *)(hwmgr->backend); phm_cap_set(hwmgr->platform_descriptor.platformCaps, - PHM_PlatformCaps_SclkDeepSleep); - - phm_cap_set(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_DynamicPatchPowerState); if (data->mvdd_control == POLARIS10_VOLTAGE_CONTROL_NONE) @@ -2819,13 +2803,6 @@ int polaris10_set_features_platform_caps(struct pp_hwmgr *hwmgr) phm_cap_set(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_TCPRamping); - if (hwmgr->powercontainment_enabled) - phm_cap_set(hwmgr->platform_descriptor.platformCaps, - PHM_PlatformCaps_PowerContainment); - else - phm_cap_unset(hwmgr->platform_descriptor.platformCaps, - PHM_PlatformCaps_PowerContainment); - phm_cap_set(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_CAC); @@ -2904,8 +2881,8 @@ static int polaris10_get_evv_voltages(struct pp_hwmgr *hwmgr) continue; } - /* need to make sure vddc is less than 2v or else, it could burn the ASIC. - * real voltage level in unit of 0.01mv */ + /* need to make sure vddc is less than 2V or else, it could burn the ASIC. + * real voltage level in unit of 0.01mV */ PP_ASSERT_WITH_CODE((vddc < 200000 && vddc != 0), "Invalid VDDC value", result = -EINVAL;); @@ -3142,7 +3119,10 @@ int polaris10_patch_voltage_workaround(struct pp_hwmgr *hwmgr) table_info->vddc_lookup_table; uint32_t i; - if (hwmgr->chip_id == CHIP_POLARIS10 && hwmgr->hw_revision == 0xC7) { + if (hwmgr->chip_id == CHIP_POLARIS10 && hwmgr->hw_revision == 0xC7 && + ((hwmgr->sub_sys_id == 0xb37 && hwmgr->sub_vendor_id == 0x1002) || + (hwmgr->sub_sys_id == 0x4a8 && hwmgr->sub_vendor_id == 0x1043) || + (hwmgr->sub_sys_id == 0x9480 && hwmgr->sub_vendor_id == 0x1682))) { if (lookup_table->entries[dep_mclk_table->entries[dep_mclk_table->count-1].vddInd].us_vdd >= 1000) return 0; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.c index 7dc4afd19d26..7f9ba7f15e19 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.c @@ -301,6 +301,8 @@ void tonga_initialize_dpm_defaults(struct pp_hwmgr *hwmgr) phm_cap_set(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_DisableMemoryTransition); + tonga_initialize_power_tune_defaults(hwmgr); + data->mclk_strobe_mode_threshold = 40000; data->mclk_stutter_mode_threshold = 30000; data->mclk_edc_enable_threshold = 40000; @@ -2478,7 +2480,7 @@ static int tonga_populate_single_graphic_level(struct pp_hwmgr *hwmgr, uint32_t graphic_level->VoltageDownHyst = 0; graphic_level->PowerThrottle = 0; - threshold = engine_clock * data->fast_watemark_threshold / 100; + threshold = engine_clock * data->fast_watermark_threshold / 100; /* *get the DAL clock. do it in funture. PECI_GetMinClockSettings(hwmgr->peci, &minClocks); @@ -2981,6 +2983,10 @@ int tonga_init_smc_table(struct pp_hwmgr *hwmgr) PP_ASSERT_WITH_CODE(0 == result, "Failed to initialize Boot Level!", return result;); + result = tonga_populate_bapm_parameters_in_dpm_table(hwmgr); + PP_ASSERT_WITH_CODE(result == 0, + "Failed to populate BAPM Parameters!", return result); + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_ClockStretcher)) { result = tonga_populate_clock_stretcher_data_table(hwmgr); @@ -4369,6 +4375,10 @@ int tonga_enable_dpm_tasks(struct pp_hwmgr *hwmgr) PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to initialize ARB table index!", result = tmp_result); + tmp_result = tonga_populate_pm_fuses(hwmgr); + PP_ASSERT_WITH_CODE((tmp_result == 0), + "Failed to populate PM fuses!", result = tmp_result); + tmp_result = tonga_populate_initial_mc_reg_table(hwmgr); PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to populate initialize MC Reg table!", result = tmp_result); @@ -4387,6 +4397,18 @@ int tonga_enable_dpm_tasks(struct pp_hwmgr *hwmgr) PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to start DPM!", result = tmp_result); + tmp_result = tonga_enable_smc_cac(hwmgr); + PP_ASSERT_WITH_CODE((tmp_result == 0), + "Failed to enable SMC CAC!", result = tmp_result); + + tmp_result = tonga_enable_power_containment(hwmgr); + PP_ASSERT_WITH_CODE((tmp_result == 0), + "Failed to enable power containment!", result = tmp_result); + + tmp_result = tonga_power_control_set_level(hwmgr); + PP_ASSERT_WITH_CODE((tmp_result == 0), + "Failed to power control set level!", result = tmp_result); + return result; } diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.h b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.h index 3961884bfa9b..fcad9426d3c1 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.h @@ -300,6 +300,7 @@ struct tonga_hwmgr { bool dll_defaule_on; bool performance_request_registered; + /* ----------------- Low Power Features ---------------------*/ phw_tonga_bacos bacos; phw_tonga_ulv_parm ulv; @@ -314,10 +315,14 @@ struct tonga_hwmgr { bool enable_tdc_limit_feature; bool enable_pkg_pwr_tracking_feature; bool disable_uvd_power_tune_feature; - phw_tonga_pt_defaults *power_tune_defaults; + struct tonga_pt_defaults *power_tune_defaults; SMU72_Discrete_PmFuses power_tune_table; - uint32_t ul_dte_tj_offset; /* Fudge factor in DPM table to correct HW DTE errors */ - uint32_t fast_watemark_threshold; /* use fast watermark if clock is equal or above this. In percentage of the target high sclk. */ + uint32_t dte_tj_offset; /* Fudge factor in DPM table to correct HW DTE errors */ + uint32_t fast_watermark_threshold; /* use fast watermark if clock is equal or above this. In percentage of the target high sclk. */ + + + bool enable_dte_feature; + /* ----------------- Phase Shedding ---------------------*/ bool vddc_phase_shed_control; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_powertune.c b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_powertune.c new file mode 100644 index 000000000000..9496ade3247e --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_powertune.c @@ -0,0 +1,498 @@ +/* + * Copyright 2015 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "hwmgr.h" +#include "smumgr.h" +#include "tonga_hwmgr.h" +#include "tonga_powertune.h" +#include "tonga_smumgr.h" +#include "smu72_discrete.h" +#include "pp_debug.h" +#include "tonga_ppsmc.h" + +#define VOLTAGE_SCALE 4 +#define POWERTUNE_DEFAULT_SET_MAX 1 + +struct tonga_pt_defaults tonga_power_tune_data_set_array[POWERTUNE_DEFAULT_SET_MAX] = { +/* sviLoadLIneEn, SviLoadLineVddC, TDC_VDDC_ThrottleReleaseLimitPerc, TDC_MAWt, TdcWaterfallCtl, DTEAmbientTempBase, DisplayCac, BAPM_TEMP_GRADIENT */ + {1, 0xF, 0xFD, 0x19, 5, 45, 0, 0xB0000, + {0x79, 0x253, 0x25D, 0xAE, 0x72, 0x80, 0x83, 0x86, 0x6F, 0xC8, 0xC9, 0xC9, 0x2F, 0x4D, 0x61}, + {0x17C, 0x172, 0x180, 0x1BC, 0x1B3, 0x1BD, 0x206, 0x200, 0x203, 0x25D, 0x25A, 0x255, 0x2C3, 0x2C5, 0x2B4 } }, +}; + +void tonga_initialize_power_tune_defaults(struct pp_hwmgr *hwmgr) +{ + struct tonga_hwmgr *tonga_hwmgr = (struct tonga_hwmgr *)(hwmgr->backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + uint32_t tmp = 0; + + if (table_info && + table_info->cac_dtp_table->usPowerTuneDataSetID <= POWERTUNE_DEFAULT_SET_MAX && + table_info->cac_dtp_table->usPowerTuneDataSetID) + tonga_hwmgr->power_tune_defaults = + &tonga_power_tune_data_set_array + [table_info->cac_dtp_table->usPowerTuneDataSetID - 1]; + else + tonga_hwmgr->power_tune_defaults = &tonga_power_tune_data_set_array[0]; + + /* Assume disabled */ + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_CAC); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SQRamping); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_DBRamping); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_TDRamping); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_TCPRamping); + + tonga_hwmgr->dte_tj_offset = tmp; + + if (!tmp) { + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_CAC); + + tonga_hwmgr->fast_watermark_threshold = 100; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment)) { + tmp = 1; + tonga_hwmgr->enable_dte_feature = tmp ? false : true; + tonga_hwmgr->enable_tdc_limit_feature = tmp ? true : false; + tonga_hwmgr->enable_pkg_pwr_tracking_feature = tmp ? true : false; + } + } +} + + +int tonga_populate_bapm_parameters_in_dpm_table(struct pp_hwmgr *hwmgr) +{ + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + struct tonga_pt_defaults *defaults = data->power_tune_defaults; + SMU72_Discrete_DpmTable *dpm_table = &(data->smc_state_table); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + struct phm_cac_tdp_table *cac_dtp_table = table_info->cac_dtp_table; + int i, j, k; + uint16_t *pdef1; + uint16_t *pdef2; + + + /* TDP number of fraction bits are changed from 8 to 7 for Fiji + * as requested by SMC team + */ + dpm_table->DefaultTdp = PP_HOST_TO_SMC_US( + (uint16_t)(cac_dtp_table->usTDP * 256)); + dpm_table->TargetTdp = PP_HOST_TO_SMC_US( + (uint16_t)(cac_dtp_table->usConfigurableTDP * 256)); + + PP_ASSERT_WITH_CODE(cac_dtp_table->usTargetOperatingTemp <= 255, + "Target Operating Temp is out of Range!", + ); + + dpm_table->GpuTjMax = (uint8_t)(cac_dtp_table->usTargetOperatingTemp); + dpm_table->GpuTjHyst = 8; + + dpm_table->DTEAmbientTempBase = defaults->dte_ambient_temp_base; + + dpm_table->BAPM_TEMP_GRADIENT = PP_HOST_TO_SMC_UL(defaults->bamp_temp_gradient); + pdef1 = defaults->bapmti_r; + pdef2 = defaults->bapmti_rc; + + for (i = 0; i < SMU72_DTE_ITERATIONS; i++) { + for (j = 0; j < SMU72_DTE_SOURCES; j++) { + for (k = 0; k < SMU72_DTE_SINKS; k++) { + dpm_table->BAPMTI_R[i][j][k] = PP_HOST_TO_SMC_US(*pdef1); + dpm_table->BAPMTI_RC[i][j][k] = PP_HOST_TO_SMC_US(*pdef2); + pdef1++; + pdef2++; + } + } + } + + return 0; +} + +static int tonga_populate_svi_load_line(struct pp_hwmgr *hwmgr) +{ + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + const struct tonga_pt_defaults *defaults = data->power_tune_defaults; + + data->power_tune_table.SviLoadLineEn = defaults->svi_load_line_en; + data->power_tune_table.SviLoadLineVddC = defaults->svi_load_line_vddC; + data->power_tune_table.SviLoadLineTrimVddC = 3; + data->power_tune_table.SviLoadLineOffsetVddC = 0; + + return 0; +} + +static int tonga_populate_tdc_limit(struct pp_hwmgr *hwmgr) +{ + uint16_t tdc_limit; + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + const struct tonga_pt_defaults *defaults = data->power_tune_defaults; + + /* TDC number of fraction bits are changed from 8 to 7 + * for Fiji as requested by SMC team + */ + tdc_limit = (uint16_t)(table_info->cac_dtp_table->usTDC * 256); + data->power_tune_table.TDC_VDDC_PkgLimit = + CONVERT_FROM_HOST_TO_SMC_US(tdc_limit); + data->power_tune_table.TDC_VDDC_ThrottleReleaseLimitPerc = + defaults->tdc_vddc_throttle_release_limit_perc; + data->power_tune_table.TDC_MAWt = defaults->tdc_mawt; + + return 0; +} + +static int tonga_populate_dw8(struct pp_hwmgr *hwmgr, uint32_t fuse_table_offset) +{ + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + const struct tonga_pt_defaults *defaults = data->power_tune_defaults; + uint32_t temp; + + if (tonga_read_smc_sram_dword(hwmgr->smumgr, + fuse_table_offset + + offsetof(SMU72_Discrete_PmFuses, TdcWaterfallCtl), + (uint32_t *)&temp, data->sram_end)) + PP_ASSERT_WITH_CODE(false, + "Attempt to read PmFuses.DW6 (SviLoadLineEn) from SMC Failed!", + return -EINVAL); + else + data->power_tune_table.TdcWaterfallCtl = defaults->tdc_waterfall_ctl; + + return 0; +} + +static int tonga_populate_temperature_scaler(struct pp_hwmgr *hwmgr) +{ + int i; + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + + /* Currently not used. Set all to zero. */ + for (i = 0; i < 16; i++) + data->power_tune_table.LPMLTemperatureScaler[i] = 0; + + return 0; +} + +static int tonga_populate_fuzzy_fan(struct pp_hwmgr *hwmgr) +{ + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + + if ((hwmgr->thermal_controller.advanceFanControlParameters. + usFanOutputSensitivity & (1 << 15)) || + (hwmgr->thermal_controller.advanceFanControlParameters.usFanOutputSensitivity == 0)) + hwmgr->thermal_controller.advanceFanControlParameters. + usFanOutputSensitivity = hwmgr->thermal_controller. + advanceFanControlParameters.usDefaultFanOutputSensitivity; + + data->power_tune_table.FuzzyFan_PwmSetDelta = + PP_HOST_TO_SMC_US(hwmgr->thermal_controller. + advanceFanControlParameters.usFanOutputSensitivity); + return 0; +} + +static int tonga_populate_gnb_lpml(struct pp_hwmgr *hwmgr) +{ + int i; + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + + /* Currently not used. Set all to zero. */ + for (i = 0; i < 16; i++) + data->power_tune_table.GnbLPML[i] = 0; + + return 0; +} + +static int tonga_min_max_vgnb_lpml_id_from_bapm_vddc(struct pp_hwmgr *hwmgr) +{ + return 0; +} + +static int tonga_populate_bapm_vddc_base_leakage_sidd(struct pp_hwmgr *hwmgr) +{ + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + uint16_t hi_sidd = data->power_tune_table.BapmVddCBaseLeakageHiSidd; + uint16_t lo_sidd = data->power_tune_table.BapmVddCBaseLeakageLoSidd; + struct phm_cac_tdp_table *cac_table = table_info->cac_dtp_table; + + hi_sidd = (uint16_t)(cac_table->usHighCACLeakage / 100 * 256); + lo_sidd = (uint16_t)(cac_table->usLowCACLeakage / 100 * 256); + + data->power_tune_table.BapmVddCBaseLeakageHiSidd = + CONVERT_FROM_HOST_TO_SMC_US(hi_sidd); + data->power_tune_table.BapmVddCBaseLeakageLoSidd = + CONVERT_FROM_HOST_TO_SMC_US(lo_sidd); + + return 0; +} + +int tonga_populate_pm_fuses(struct pp_hwmgr *hwmgr) +{ + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + uint32_t pm_fuse_table_offset; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment)) { + if (tonga_read_smc_sram_dword(hwmgr->smumgr, + SMU72_FIRMWARE_HEADER_LOCATION + + offsetof(SMU72_Firmware_Header, PmFuseTable), + &pm_fuse_table_offset, data->sram_end)) + PP_ASSERT_WITH_CODE(false, + "Attempt to get pm_fuse_table_offset Failed!", + return -EINVAL); + + /* DW6 */ + if (tonga_populate_svi_load_line(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate SviLoadLine Failed!", + return -EINVAL); + /* DW7 */ + if (tonga_populate_tdc_limit(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate TDCLimit Failed!", return -EINVAL); + /* DW8 */ + if (tonga_populate_dw8(hwmgr, pm_fuse_table_offset)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate TdcWaterfallCtl Failed !", + return -EINVAL); + + /* DW9-DW12 */ + if (tonga_populate_temperature_scaler(hwmgr) != 0) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate LPMLTemperatureScaler Failed!", + return -EINVAL); + + /* DW13-DW14 */ + if (tonga_populate_fuzzy_fan(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate Fuzzy Fan Control parameters Failed!", + return -EINVAL); + + /* DW15-DW18 */ + if (tonga_populate_gnb_lpml(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate GnbLPML Failed!", + return -EINVAL); + + /* DW19 */ + if (tonga_min_max_vgnb_lpml_id_from_bapm_vddc(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate GnbLPML Min and Max Vid Failed!", + return -EINVAL); + + /* DW20 */ + if (tonga_populate_bapm_vddc_base_leakage_sidd(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate BapmVddCBaseLeakage Hi and Lo Sidd Failed!", + return -EINVAL); + + if (tonga_copy_bytes_to_smc(hwmgr->smumgr, pm_fuse_table_offset, + (uint8_t *)&data->power_tune_table, + sizeof(struct SMU72_Discrete_PmFuses), data->sram_end)) + PP_ASSERT_WITH_CODE(false, + "Attempt to download PmFuseTable Failed!", + return -EINVAL); + } + return 0; +} + +int tonga_enable_smc_cac(struct pp_hwmgr *hwmgr) +{ + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + int result = 0; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_CAC)) { + int smc_result; + + smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_EnableCac)); + PP_ASSERT_WITH_CODE((smc_result == 0), + "Failed to enable CAC in SMC.", result = -1); + + data->cac_enabled = (smc_result == 0) ? true : false; + } + return result; +} + +int tonga_disable_smc_cac(struct pp_hwmgr *hwmgr) +{ + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + int result = 0; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_CAC) && data->cac_enabled) { + int smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_DisableCac)); + PP_ASSERT_WITH_CODE((smc_result == 0), + "Failed to disable CAC in SMC.", result = -1); + + data->cac_enabled = false; + } + return result; +} + +int tonga_set_power_limit(struct pp_hwmgr *hwmgr, uint32_t n) +{ + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + + if (data->power_containment_features & + POWERCONTAINMENT_FEATURE_PkgPwrLimit) + return smum_send_msg_to_smc_with_parameter(hwmgr->smumgr, + PPSMC_MSG_PkgPwrSetLimit, n); + return 0; +} + +static int tonga_set_overdriver_target_tdp(struct pp_hwmgr *pHwMgr, uint32_t target_tdp) +{ + return smum_send_msg_to_smc_with_parameter(pHwMgr->smumgr, + PPSMC_MSG_OverDriveSetTargetTdp, target_tdp); +} + +int tonga_enable_power_containment(struct pp_hwmgr *hwmgr) +{ + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + int smc_result; + int result = 0; + + data->power_containment_features = 0; + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment)) { + if (data->enable_dte_feature) { + smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_EnableDTE)); + PP_ASSERT_WITH_CODE((smc_result == 0), + "Failed to enable DTE in SMC.", result = -1;); + if (smc_result == 0) + data->power_containment_features |= POWERCONTAINMENT_FEATURE_DTE; + } + + if (data->enable_tdc_limit_feature) { + smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_TDCLimitEnable)); + PP_ASSERT_WITH_CODE((smc_result == 0), + "Failed to enable TDCLimit in SMC.", result = -1;); + if (smc_result == 0) + data->power_containment_features |= + POWERCONTAINMENT_FEATURE_TDCLimit; + } + + if (data->enable_pkg_pwr_tracking_feature) { + smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_PkgPwrLimitEnable)); + PP_ASSERT_WITH_CODE((smc_result == 0), + "Failed to enable PkgPwrTracking in SMC.", result = -1;); + if (smc_result == 0) { + struct phm_cac_tdp_table *cac_table = + table_info->cac_dtp_table; + uint32_t default_limit = + (uint32_t)(cac_table->usMaximumPowerDeliveryLimit * 256); + + data->power_containment_features |= + POWERCONTAINMENT_FEATURE_PkgPwrLimit; + + if (tonga_set_power_limit(hwmgr, default_limit)) + printk(KERN_ERR "Failed to set Default Power Limit in SMC!"); + } + } + } + return result; +} + +int tonga_disable_power_containment(struct pp_hwmgr *hwmgr) +{ + struct tonga_hwmgr *data = (struct tonga_hwmgr *)(hwmgr->backend); + int result = 0; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment) && + data->power_containment_features) { + int smc_result; + + if (data->power_containment_features & + POWERCONTAINMENT_FEATURE_TDCLimit) { + smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_TDCLimitDisable)); + PP_ASSERT_WITH_CODE((smc_result == 0), + "Failed to disable TDCLimit in SMC.", + result = smc_result); + } + + if (data->power_containment_features & + POWERCONTAINMENT_FEATURE_DTE) { + smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_DisableDTE)); + PP_ASSERT_WITH_CODE((smc_result == 0), + "Failed to disable DTE in SMC.", + result = smc_result); + } + + if (data->power_containment_features & + POWERCONTAINMENT_FEATURE_PkgPwrLimit) { + smc_result = smum_send_msg_to_smc(hwmgr->smumgr, + (uint16_t)(PPSMC_MSG_PkgPwrLimitDisable)); + PP_ASSERT_WITH_CODE((smc_result == 0), + "Failed to disable PkgPwrTracking in SMC.", + result = smc_result); + } + data->power_containment_features = 0; + } + + return result; +} + +int tonga_power_control_set_level(struct pp_hwmgr *hwmgr) +{ + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + struct phm_cac_tdp_table *cac_table = table_info->cac_dtp_table; + int adjust_percent, target_tdp; + int result = 0; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment)) { + /* adjustment percentage has already been validated */ + adjust_percent = hwmgr->platform_descriptor.TDPAdjustmentPolarity ? + hwmgr->platform_descriptor.TDPAdjustment : + (-1 * hwmgr->platform_descriptor.TDPAdjustment); + /* SMC requested that target_tdp to be 7 bit fraction in DPM table + * but message to be 8 bit fraction for messages + */ + target_tdp = ((100 + adjust_percent) * (int)(cac_table->usTDP * 256)) / 100; + result = tonga_set_overdriver_target_tdp(hwmgr, (uint32_t)target_tdp); + } + + return result; +} diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_powertune.h b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_powertune.h index 8e6670b3cb67..c8bdb92d81f4 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_powertune.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_powertune.h @@ -35,20 +35,23 @@ enum _phw_tonga_ptc_config_reg_type { typedef enum _phw_tonga_ptc_config_reg_type phw_tonga_ptc_config_reg_type; /* PowerContainment Features */ +#define POWERCONTAINMENT_FEATURE_DTE 0x00000001 + + +/* PowerContainment Features */ #define POWERCONTAINMENT_FEATURE_BAPM 0x00000001 #define POWERCONTAINMENT_FEATURE_TDCLimit 0x00000002 #define POWERCONTAINMENT_FEATURE_PkgPwrLimit 0x00000004 -struct _phw_tonga_pt_config_reg { +struct tonga_pt_config_reg { uint32_t Offset; uint32_t Mask; uint32_t Shift; uint32_t Value; phw_tonga_ptc_config_reg_type Type; }; -typedef struct _phw_tonga_pt_config_reg phw_tonga_pt_config_reg; -struct _phw_tonga_pt_defaults { +struct tonga_pt_defaults { uint8_t svi_load_line_en; uint8_t svi_load_line_vddC; uint8_t tdc_vddc_throttle_release_limit_perc; @@ -60,7 +63,18 @@ struct _phw_tonga_pt_defaults { uint16_t bapmti_r[SMU72_DTE_ITERATIONS * SMU72_DTE_SOURCES * SMU72_DTE_SINKS]; uint16_t bapmti_rc[SMU72_DTE_ITERATIONS * SMU72_DTE_SOURCES * SMU72_DTE_SINKS]; }; -typedef struct _phw_tonga_pt_defaults phw_tonga_pt_defaults; + + + +void tonga_initialize_power_tune_defaults(struct pp_hwmgr *hwmgr); +int tonga_populate_bapm_parameters_in_dpm_table(struct pp_hwmgr *hwmgr); +int tonga_populate_pm_fuses(struct pp_hwmgr *hwmgr); +int tonga_enable_smc_cac(struct pp_hwmgr *hwmgr); +int tonga_disable_smc_cac(struct pp_hwmgr *hwmgr); +int tonga_enable_power_containment(struct pp_hwmgr *hwmgr); +int tonga_disable_power_containment(struct pp_hwmgr *hwmgr); +int tonga_set_power_limit(struct pp_hwmgr *hwmgr, uint32_t n); +int tonga_power_control_set_level(struct pp_hwmgr *hwmgr); #endif diff --git a/drivers/gpu/drm/amd/powerplay/inc/amd_powerplay.h b/drivers/gpu/drm/amd/powerplay/inc/amd_powerplay.h index b764c8c05ec8..3f8172f545b0 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/amd_powerplay.h +++ b/drivers/gpu/drm/amd/powerplay/inc/amd_powerplay.h @@ -132,8 +132,10 @@ struct amd_pp_init { uint32_t chip_family; uint32_t chip_id; uint32_t rev_id; - bool powercontainment_enabled; + uint16_t sub_sys_id; + uint16_t sub_vendor_id; }; + enum amd_pp_display_config_type{ AMD_PP_DisplayConfigType_None = 0, AMD_PP_DisplayConfigType_DP54 , diff --git a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h index bf0d2accf7bf..36b4ec9c9cb1 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h @@ -41,6 +41,9 @@ struct phm_fan_speed_info; struct pp_atomctrl_voltage_table; +extern int amdgpu_powercontainment; +extern int amdgpu_sclk_deep_sleep_en; + enum DISPLAY_GAP { DISPLAY_GAP_VBLANK_OR_WM = 0, /* Wait for vblank or MCHG watermark. */ DISPLAY_GAP_VBLANK = 1, /* Wait for vblank. */ @@ -614,7 +617,6 @@ struct pp_hwmgr { uint32_t num_ps; struct pp_thermal_controller_info thermal_controller; bool fan_ctrl_is_in_default_mode; - bool powercontainment_enabled; uint32_t fan_ctrl_default_mode; uint32_t tmin; struct phm_microcode_version_info microcode_version_info; @@ -637,16 +639,7 @@ extern int hw_init_power_state_table(struct pp_hwmgr *hwmgr); extern int phm_wait_on_register(struct pp_hwmgr *hwmgr, uint32_t index, uint32_t value, uint32_t mask); -extern int phm_wait_for_register_unequal(struct pp_hwmgr *hwmgr, - uint32_t index, uint32_t value, uint32_t mask); - -extern uint32_t phm_read_indirect_register(struct pp_hwmgr *hwmgr, - uint32_t indirect_port, uint32_t index); -extern void phm_write_indirect_register(struct pp_hwmgr *hwmgr, - uint32_t indirect_port, - uint32_t index, - uint32_t value); extern void phm_wait_on_indirect_register(struct pp_hwmgr *hwmgr, uint32_t indirect_port, @@ -654,12 +647,7 @@ extern void phm_wait_on_indirect_register(struct pp_hwmgr *hwmgr, uint32_t value, uint32_t mask); -extern void phm_wait_for_indirect_register_unequal( - struct pp_hwmgr *hwmgr, - uint32_t indirect_port, - uint32_t index, - uint32_t value, - uint32_t mask); + extern bool phm_cf_want_uvd_power_gating(struct pp_hwmgr *hwmgr); extern bool phm_cf_want_vce_power_gating(struct pp_hwmgr *hwmgr); @@ -697,43 +685,7 @@ extern void phm_apply_dal_min_voltage_request(struct pp_hwmgr *hwmgr); PHM_FIELD_SHIFT(reg, field)) -#define PHM_WAIT_REGISTER_GIVEN_INDEX(hwmgr, index, value, mask) \ - phm_wait_on_register(hwmgr, index, value, mask) - -#define PHM_WAIT_REGISTER_UNEQUAL_GIVEN_INDEX(hwmgr, index, value, mask) \ - phm_wait_for_register_unequal(hwmgr, index, value, mask) - -#define PHM_WAIT_INDIRECT_REGISTER_GIVEN_INDEX(hwmgr, port, index, value, mask) \ - phm_wait_on_indirect_register(hwmgr, mm##port##_INDEX, index, value, mask) - -#define PHM_WAIT_INDIRECT_REGISTER_UNEQUAL_GIVEN_INDEX(hwmgr, port, index, value, mask) \ - phm_wait_for_indirect_register_unequal(hwmgr, mm##port##_INDEX, index, value, mask) -#define PHM_WAIT_VFPF_INDIRECT_REGISTER_GIVEN_INDEX(hwmgr, port, index, value, mask) \ - phm_wait_on_indirect_register(hwmgr, mm##port##_INDEX_0, index, value, mask) - -#define PHM_WAIT_VFPF_INDIRECT_REGISTER_UNEQUAL_GIVEN_INDEX(hwmgr, port, index, value, mask) \ - phm_wait_for_indirect_register_unequal(hwmgr, mm##port##_INDEX_0, index, value, mask) - -/* Operations on named registers. */ - -#define PHM_WAIT_REGISTER(hwmgr, reg, value, mask) \ - PHM_WAIT_REGISTER_GIVEN_INDEX(hwmgr, mm##reg, value, mask) - -#define PHM_WAIT_REGISTER_UNEQUAL(hwmgr, reg, value, mask) \ - PHM_WAIT_REGISTER_UNEQUAL_GIVEN_INDEX(hwmgr, mm##reg, value, mask) - -#define PHM_WAIT_INDIRECT_REGISTER(hwmgr, port, reg, value, mask) \ - PHM_WAIT_INDIRECT_REGISTER_GIVEN_INDEX(hwmgr, port, ix##reg, value, mask) - -#define PHM_WAIT_INDIRECT_REGISTER_UNEQUAL(hwmgr, port, reg, value, mask) \ - PHM_WAIT_INDIRECT_REGISTER_UNEQUAL_GIVEN_INDEX(hwmgr, port, ix##reg, value, mask) - -#define PHM_WAIT_VFPF_INDIRECT_REGISTER(hwmgr, port, reg, value, mask) \ - PHM_WAIT_VFPF_INDIRECT_REGISTER_GIVEN_INDEX(hwmgr, port, ix##reg, value, mask) - -#define PHM_WAIT_VFPF_INDIRECT_REGISTER_UNEQUAL(hwmgr, port, reg, value, mask) \ - PHM_WAIT_VFPF_INDIRECT_REGISTER_UNEQUAL_GIVEN_INDEX(hwmgr, port, ix##reg, value, mask) /* Operations on named fields. */ @@ -762,60 +714,16 @@ extern void phm_apply_dal_min_voltage_request(struct pp_hwmgr *hwmgr); PHM_SET_FIELD(cgs_read_ind_register(device, port, ix##reg), \ reg, field, fieldval)) -#define PHM_WAIT_FIELD(hwmgr, reg, field, fieldval) \ - PHM_WAIT_REGISTER(hwmgr, reg, (fieldval) \ - << PHM_FIELD_SHIFT(reg, field), PHM_FIELD_MASK(reg, field)) +#define PHM_WAIT_INDIRECT_REGISTER_GIVEN_INDEX(hwmgr, port, index, value, mask) \ + phm_wait_on_indirect_register(hwmgr, mm##port##_INDEX, index, value, mask) -#define PHM_WAIT_INDIRECT_FIELD(hwmgr, port, reg, field, fieldval) \ - PHM_WAIT_INDIRECT_REGISTER(hwmgr, port, reg, (fieldval) \ - << PHM_FIELD_SHIFT(reg, field), PHM_FIELD_MASK(reg, field)) -#define PHM_WAIT_VFPF_INDIRECT_FIELD(hwmgr, port, reg, field, fieldval) \ - PHM_WAIT_VFPF_INDIRECT_REGISTER(hwmgr, port, reg, (fieldval) \ - << PHM_FIELD_SHIFT(reg, field), PHM_FIELD_MASK(reg, field)) +#define PHM_WAIT_INDIRECT_REGISTER(hwmgr, port, reg, value, mask) \ + PHM_WAIT_INDIRECT_REGISTER_GIVEN_INDEX(hwmgr, port, ix##reg, value, mask) -#define PHM_WAIT_FIELD_UNEQUAL(hwmgr, reg, field, fieldval) \ - PHM_WAIT_REGISTER_UNEQUAL(hwmgr, reg, (fieldval) \ - << PHM_FIELD_SHIFT(reg, field), PHM_FIELD_MASK(reg, field)) - -#define PHM_WAIT_INDIRECT_FIELD_UNEQUAL(hwmgr, port, reg, field, fieldval) \ - PHM_WAIT_INDIRECT_REGISTER_UNEQUAL(hwmgr, port, reg, (fieldval) \ - << PHM_FIELD_SHIFT(reg, field), PHM_FIELD_MASK(reg, field)) - -#define PHM_WAIT_VFPF_INDIRECT_FIELD_UNEQUAL(hwmgr, port, reg, field, fieldval) \ - PHM_WAIT_VFPF_INDIRECT_REGISTER_UNEQUAL(hwmgr, port, reg, (fieldval) \ +#define PHM_WAIT_INDIRECT_FIELD(hwmgr, port, reg, field, fieldval) \ + PHM_WAIT_INDIRECT_REGISTER(hwmgr, port, reg, (fieldval) \ << PHM_FIELD_SHIFT(reg, field), PHM_FIELD_MASK(reg, field)) -/* Operations on arrays of registers & fields. */ - -#define PHM_READ_ARRAY_REGISTER(device, reg, offset) \ - cgs_read_register(device, mm##reg + (offset)) - -#define PHM_WRITE_ARRAY_REGISTER(device, reg, offset, value) \ - cgs_write_register(device, mm##reg + (offset), value) - -#define PHM_WAIT_ARRAY_REGISTER(hwmgr, reg, offset, value, mask) \ - PHM_WAIT_REGISTER_GIVEN_INDEX(hwmgr, mm##reg + (offset), value, mask) - -#define PHM_WAIT_ARRAY_REGISTER_UNEQUAL(hwmgr, reg, offset, value, mask) \ - PHM_WAIT_REGISTER_UNEQUAL_GIVEN_INDEX(hwmgr, mm##reg + (offset), value, mask) - -#define PHM_READ_ARRAY_FIELD(hwmgr, reg, offset, field) \ - PHM_GET_FIELD(PHM_READ_ARRAY_REGISTER(hwmgr->device, reg, offset), reg, field) - -#define PHM_WRITE_ARRAY_FIELD(hwmgr, reg, offset, field, fieldvalue) \ - PHM_WRITE_ARRAY_REGISTER(hwmgr->device, reg, offset, \ - PHM_SET_FIELD(PHM_READ_ARRAY_REGISTER(hwmgr->device, reg, offset), \ - reg, field, fieldvalue)) - -#define PHM_WAIT_ARRAY_FIELD(hwmgr, reg, offset, field, fieldvalue) \ - PHM_WAIT_REGISTER_GIVEN_INDEX(hwmgr, mm##reg + (offset), \ - (fieldvalue) << PHM_FIELD_SHIFT(reg, field), \ - PHM_FIELD_MASK(reg, field)) - -#define PHM_WAIT_ARRAY_FIELD_UNEQUAL(hwmgr, reg, offset, field, fieldvalue) \ - PHM_WAIT_REGISTER_UNEQUAL_GIVEN_INDEX(hwmgr, mm##reg + (offset), \ - (fieldvalue) << PHM_FIELD_SHIFT(reg, field), \ - PHM_FIELD_MASK(reg, field)) #endif /* _HWMGR_H_ */ diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu71.h b/drivers/gpu/drm/amd/powerplay/inc/smu71.h new file mode 100644 index 000000000000..71c9b2d28640 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/inc/smu71.h @@ -0,0 +1,510 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef SMU71_H +#define SMU71_H + +#if !defined(SMC_MICROCODE) +#pragma pack(push, 1) +#endif + +#define SMU__NUM_PCIE_DPM_LEVELS 8 +#define SMU__NUM_SCLK_DPM_STATE 8 +#define SMU__NUM_MCLK_DPM_LEVELS 4 +#define SMU__VARIANT__ICELAND 1 +#define SMU__DGPU_ONLY 1 +#define SMU__DYNAMIC_MCARB_SETTINGS 1 + +enum SID_OPTION { + SID_OPTION_HI, + SID_OPTION_LO, + SID_OPTION_COUNT +}; + +typedef struct { + uint32_t high; + uint32_t low; +} data_64_t; + +typedef struct { + data_64_t high; + data_64_t low; +} data_128_t; + +#define SMU7_CONTEXT_ID_SMC 1 +#define SMU7_CONTEXT_ID_VBIOS 2 + +#define SMU71_MAX_LEVELS_VDDC 8 +#define SMU71_MAX_LEVELS_VDDCI 4 +#define SMU71_MAX_LEVELS_MVDD 4 +#define SMU71_MAX_LEVELS_VDDNB 8 + +#define SMU71_MAX_LEVELS_GRAPHICS SMU__NUM_SCLK_DPM_STATE +#define SMU71_MAX_LEVELS_MEMORY SMU__NUM_MCLK_DPM_LEVELS +#define SMU71_MAX_LEVELS_GIO SMU__NUM_LCLK_DPM_LEVELS +#define SMU71_MAX_LEVELS_LINK SMU__NUM_PCIE_DPM_LEVELS +#define SMU71_MAX_ENTRIES_SMIO 32 + +#define DPM_NO_LIMIT 0 +#define DPM_NO_UP 1 +#define DPM_GO_DOWN 2 +#define DPM_GO_UP 3 + +#define SMU7_FIRST_DPM_GRAPHICS_LEVEL 0 +#define SMU7_FIRST_DPM_MEMORY_LEVEL 0 + +#define GPIO_CLAMP_MODE_VRHOT 1 +#define GPIO_CLAMP_MODE_THERM 2 +#define GPIO_CLAMP_MODE_DC 4 + +#define SCRATCH_B_TARG_PCIE_INDEX_SHIFT 0 +#define SCRATCH_B_TARG_PCIE_INDEX_MASK (0x7<<SCRATCH_B_TARG_PCIE_INDEX_SHIFT) +#define SCRATCH_B_CURR_PCIE_INDEX_SHIFT 3 +#define SCRATCH_B_CURR_PCIE_INDEX_MASK (0x7<<SCRATCH_B_CURR_PCIE_INDEX_SHIFT) +#define SCRATCH_B_TARG_UVD_INDEX_SHIFT 6 +#define SCRATCH_B_TARG_UVD_INDEX_MASK (0x7<<SCRATCH_B_TARG_UVD_INDEX_SHIFT) +#define SCRATCH_B_CURR_UVD_INDEX_SHIFT 9 +#define SCRATCH_B_CURR_UVD_INDEX_MASK (0x7<<SCRATCH_B_CURR_UVD_INDEX_SHIFT) +#define SCRATCH_B_TARG_VCE_INDEX_SHIFT 12 +#define SCRATCH_B_TARG_VCE_INDEX_MASK (0x7<<SCRATCH_B_TARG_VCE_INDEX_SHIFT) +#define SCRATCH_B_CURR_VCE_INDEX_SHIFT 15 +#define SCRATCH_B_CURR_VCE_INDEX_MASK (0x7<<SCRATCH_B_CURR_VCE_INDEX_SHIFT) +#define SCRATCH_B_TARG_ACP_INDEX_SHIFT 18 +#define SCRATCH_B_TARG_ACP_INDEX_MASK (0x7<<SCRATCH_B_TARG_ACP_INDEX_SHIFT) +#define SCRATCH_B_CURR_ACP_INDEX_SHIFT 21 +#define SCRATCH_B_CURR_ACP_INDEX_MASK (0x7<<SCRATCH_B_CURR_ACP_INDEX_SHIFT) +#define SCRATCH_B_TARG_SAMU_INDEX_SHIFT 24 +#define SCRATCH_B_TARG_SAMU_INDEX_MASK (0x7<<SCRATCH_B_TARG_SAMU_INDEX_SHIFT) +#define SCRATCH_B_CURR_SAMU_INDEX_SHIFT 27 +#define SCRATCH_B_CURR_SAMU_INDEX_MASK (0x7<<SCRATCH_B_CURR_SAMU_INDEX_SHIFT) + + +#if defined SMU__DGPU_ONLY +#define SMU71_DTE_ITERATIONS 5 +#define SMU71_DTE_SOURCES 3 +#define SMU71_DTE_SINKS 1 +#define SMU71_NUM_CPU_TES 0 +#define SMU71_NUM_GPU_TES 1 +#define SMU71_NUM_NON_TES 2 + +#endif + +#if defined SMU__FUSION_ONLY +#define SMU7_DTE_ITERATIONS 5 +#define SMU7_DTE_SOURCES 5 +#define SMU7_DTE_SINKS 3 +#define SMU7_NUM_CPU_TES 2 +#define SMU7_NUM_GPU_TES 1 +#define SMU7_NUM_NON_TES 2 + +#endif + +struct SMU71_PIDController +{ + uint32_t Ki; + int32_t LFWindupUpperLim; + int32_t LFWindupLowerLim; + uint32_t StatePrecision; + uint32_t LfPrecision; + uint32_t LfOffset; + uint32_t MaxState; + uint32_t MaxLfFraction; + uint32_t StateShift; +}; + +typedef struct SMU71_PIDController SMU71_PIDController; + +struct SMU7_LocalDpmScoreboard +{ + uint32_t PercentageBusy; + + int32_t PIDError; + int32_t PIDIntegral; + int32_t PIDOutput; + + uint32_t SigmaDeltaAccum; + uint32_t SigmaDeltaOutput; + uint32_t SigmaDeltaLevel; + + uint32_t UtilizationSetpoint; + + uint8_t TdpClampMode; + uint8_t TdcClampMode; + uint8_t ThermClampMode; + uint8_t VoltageBusy; + + int8_t CurrLevel; + int8_t TargLevel; + uint8_t LevelChangeInProgress; + uint8_t UpHyst; + + uint8_t DownHyst; + uint8_t VoltageDownHyst; + uint8_t DpmEnable; + uint8_t DpmRunning; + + uint8_t DpmForce; + uint8_t DpmForceLevel; + uint8_t DisplayWatermark; + uint8_t McArbIndex; + + uint32_t MinimumPerfSclk; + + uint8_t AcpiReq; + uint8_t AcpiAck; + uint8_t GfxClkSlow; + uint8_t GpioClampMode; + + uint8_t FpsFilterWeight; + uint8_t EnabledLevelsChange; + uint8_t DteClampMode; + uint8_t FpsClampMode; + + uint16_t LevelResidencyCounters [SMU71_MAX_LEVELS_GRAPHICS]; + uint16_t LevelSwitchCounters [SMU71_MAX_LEVELS_GRAPHICS]; + + void (*TargetStateCalculator)(uint8_t); + void (*SavedTargetStateCalculator)(uint8_t); + + uint16_t AutoDpmInterval; + uint16_t AutoDpmRange; + + uint8_t FpsEnabled; + uint8_t MaxPerfLevel; + uint8_t AllowLowClkInterruptToHost; + uint8_t FpsRunning; + + uint32_t MaxAllowedFrequency; +}; + +typedef struct SMU7_LocalDpmScoreboard SMU7_LocalDpmScoreboard; + +#define SMU7_MAX_VOLTAGE_CLIENTS 12 + +struct SMU7_VoltageScoreboard +{ + uint16_t CurrentVoltage; + uint16_t HighestVoltage; + uint16_t MaxVid; + uint8_t HighestVidOffset; + uint8_t CurrentVidOffset; +#if defined (SMU__DGPU_ONLY) + uint8_t CurrentPhases; + uint8_t HighestPhases; +#else + uint8_t AvsOffset; + uint8_t AvsOffsetApplied; +#endif + uint8_t ControllerBusy; + uint8_t CurrentVid; + uint16_t RequestedVoltage[SMU7_MAX_VOLTAGE_CLIENTS]; +#if defined (SMU__DGPU_ONLY) + uint8_t RequestedPhases[SMU7_MAX_VOLTAGE_CLIENTS]; +#endif + uint8_t EnabledRequest[SMU7_MAX_VOLTAGE_CLIENTS]; + uint8_t TargetIndex; + uint8_t Delay; + uint8_t ControllerEnable; + uint8_t ControllerRunning; + uint16_t CurrentStdVoltageHiSidd; + uint16_t CurrentStdVoltageLoSidd; +#if defined (SMU__DGPU_ONLY) + uint16_t RequestedVddci; + uint16_t CurrentVddci; + uint16_t HighestVddci; + uint8_t CurrentVddciVid; + uint8_t TargetVddciIndex; +#endif +}; + +typedef struct SMU7_VoltageScoreboard SMU7_VoltageScoreboard; + +// ------------------------------------------------------------------------------------------------------------------------- +#define SMU7_MAX_PCIE_LINK_SPEEDS 3 /* 0:Gen1 1:Gen2 2:Gen3 */ + +struct SMU7_PCIeLinkSpeedScoreboard +{ + uint8_t DpmEnable; + uint8_t DpmRunning; + uint8_t DpmForce; + uint8_t DpmForceLevel; + + uint8_t CurrentLinkSpeed; + uint8_t EnabledLevelsChange; + uint16_t AutoDpmInterval; + + uint16_t AutoDpmRange; + uint16_t AutoDpmCount; + + uint8_t DpmMode; + uint8_t AcpiReq; + uint8_t AcpiAck; + uint8_t CurrentLinkLevel; + +}; + +typedef struct SMU7_PCIeLinkSpeedScoreboard SMU7_PCIeLinkSpeedScoreboard; + +// -------------------------------------------------------- CAC table ------------------------------------------------------ +#define SMU7_LKGE_LUT_NUM_OF_TEMP_ENTRIES 16 +#define SMU7_LKGE_LUT_NUM_OF_VOLT_ENTRIES 16 + +#define SMU7_SCALE_I 7 +#define SMU7_SCALE_R 12 + +struct SMU7_PowerScoreboard +{ + uint16_t MinVoltage; + uint16_t MaxVoltage; + + uint32_t AvgGpuPower; + + uint16_t VddcLeakagePower[SID_OPTION_COUNT]; + uint16_t VddcSclkConstantPower[SID_OPTION_COUNT]; + uint16_t VddcSclkDynamicPower[SID_OPTION_COUNT]; + uint16_t VddcNonSclkDynamicPower[SID_OPTION_COUNT]; + uint16_t VddcTotalPower[SID_OPTION_COUNT]; + uint16_t VddcTotalCurrent[SID_OPTION_COUNT]; + uint16_t VddcLoadVoltage[SID_OPTION_COUNT]; + uint16_t VddcNoLoadVoltage[SID_OPTION_COUNT]; + + uint16_t DisplayPhyPower; + uint16_t PciePhyPower; + + uint16_t VddciTotalPower; + uint16_t Vddr1TotalPower; + + uint32_t RocPower; + + uint32_t last_power; + uint32_t enableWinAvg; + + uint32_t lkg_acc; + uint16_t VoltLkgeScaler; + uint16_t TempLkgeScaler; + + uint32_t uvd_cac_dclk; + uint32_t uvd_cac_vclk; + uint32_t vce_cac_eclk; + uint32_t samu_cac_samclk; + uint32_t display_cac_dispclk; + uint32_t acp_cac_aclk; + uint32_t unb_cac; + + uint32_t WinTime; + + uint16_t GpuPwr_MAWt; + uint16_t FilteredVddcTotalPower; + + uint8_t CalculationRepeats; + uint8_t WaterfallUp; + uint8_t WaterfallDown; + uint8_t WaterfallLimit; +}; + +typedef struct SMU7_PowerScoreboard SMU7_PowerScoreboard; + +// -------------------------------------------------------------------------------------------------- + +struct SMU7_ThermalScoreboard +{ + int16_t GpuLimit; + int16_t GpuHyst; + uint16_t CurrGnbTemp; + uint16_t FilteredGnbTemp; + uint8_t ControllerEnable; + uint8_t ControllerRunning; + uint8_t WaterfallUp; + uint8_t WaterfallDown; + uint8_t WaterfallLimit; + uint8_t padding[3]; +}; + +typedef struct SMU7_ThermalScoreboard SMU7_ThermalScoreboard; + +// For FeatureEnables: +#define SMU7_SCLK_DPM_CONFIG_MASK 0x01 +#define SMU7_VOLTAGE_CONTROLLER_CONFIG_MASK 0x02 +#define SMU7_THERMAL_CONTROLLER_CONFIG_MASK 0x04 +#define SMU7_MCLK_DPM_CONFIG_MASK 0x08 +#define SMU7_UVD_DPM_CONFIG_MASK 0x10 +#define SMU7_VCE_DPM_CONFIG_MASK 0x20 +#define SMU7_ACP_DPM_CONFIG_MASK 0x40 +#define SMU7_SAMU_DPM_CONFIG_MASK 0x80 +#define SMU7_PCIEGEN_DPM_CONFIG_MASK 0x100 + +#define SMU7_ACP_MCLK_HANDSHAKE_DISABLE 0x00000001 +#define SMU7_ACP_SCLK_HANDSHAKE_DISABLE 0x00000002 +#define SMU7_UVD_MCLK_HANDSHAKE_DISABLE 0x00000100 +#define SMU7_UVD_SCLK_HANDSHAKE_DISABLE 0x00000200 +#define SMU7_VCE_MCLK_HANDSHAKE_DISABLE 0x00010000 +#define SMU7_VCE_SCLK_HANDSHAKE_DISABLE 0x00020000 + +// All 'soft registers' should be uint32_t. +struct SMU71_SoftRegisters +{ + uint32_t RefClockFrequency; + uint32_t PmTimerPeriod; + uint32_t FeatureEnables; +#if defined (SMU__DGPU_ONLY) + uint32_t PreVBlankGap; + uint32_t VBlankTimeout; + uint32_t TrainTimeGap; + uint32_t MvddSwitchTime; + uint32_t LongestAcpiTrainTime; + uint32_t AcpiDelay; + uint32_t G5TrainTime; + uint32_t DelayMpllPwron; + uint32_t VoltageChangeTimeout; +#endif + uint32_t HandshakeDisables; + + uint8_t DisplayPhy1Config; + uint8_t DisplayPhy2Config; + uint8_t DisplayPhy3Config; + uint8_t DisplayPhy4Config; + + uint8_t DisplayPhy5Config; + uint8_t DisplayPhy6Config; + uint8_t DisplayPhy7Config; + uint8_t DisplayPhy8Config; + + uint32_t AverageGraphicsActivity; + uint32_t AverageMemoryActivity; + uint32_t AverageGioActivity; + + uint8_t SClkDpmEnabledLevels; + uint8_t MClkDpmEnabledLevels; + uint8_t LClkDpmEnabledLevels; + uint8_t PCIeDpmEnabledLevels; + + uint32_t DRAM_LOG_ADDR_H; + uint32_t DRAM_LOG_ADDR_L; + uint32_t DRAM_LOG_PHY_ADDR_H; + uint32_t DRAM_LOG_PHY_ADDR_L; + uint32_t DRAM_LOG_BUFF_SIZE; + uint32_t UlvEnterCount; + uint32_t UlvTime; + uint32_t UcodeLoadStatus; + uint8_t DPMFreezeAndForced; + uint8_t Activity_Weight; + uint8_t Reserved8[2]; + uint32_t Reserved; +}; + +typedef struct SMU71_SoftRegisters SMU71_SoftRegisters; + +struct SMU71_Firmware_Header +{ + uint32_t Digest[5]; + uint32_t Version; + uint32_t HeaderSize; + uint32_t Flags; + uint32_t EntryPoint; + uint32_t CodeSize; + uint32_t ImageSize; + + uint32_t Rtos; + uint32_t SoftRegisters; + uint32_t DpmTable; + uint32_t FanTable; + uint32_t CacConfigTable; + uint32_t CacStatusTable; + + uint32_t mcRegisterTable; + + uint32_t mcArbDramTimingTable; + + uint32_t PmFuseTable; + uint32_t Globals; + uint32_t UvdDpmTable; + uint32_t AcpDpmTable; + uint32_t VceDpmTable; + uint32_t SamuDpmTable; + uint32_t UlvSettings; + uint32_t Reserved[37]; + uint32_t Signature; +}; + +typedef struct SMU71_Firmware_Header SMU71_Firmware_Header; + +struct SMU7_HystController_Data +{ + uint8_t waterfall_up; + uint8_t waterfall_down; + uint8_t pstate; + uint8_t clamp_mode; +}; + +typedef struct SMU7_HystController_Data SMU7_HystController_Data; + +#define SMU71_FIRMWARE_HEADER_LOCATION 0x20000 + +enum DisplayConfig { + PowerDown = 1, + DP54x4, + DP54x2, + DP54x1, + DP27x4, + DP27x2, + DP27x1, + HDMI297, + HDMI162, + LVDS, + DP324x4, + DP324x2, + DP324x1 +}; + +//#define SX_BLOCK_COUNT 8 +//#define MC_BLOCK_COUNT 1 +//#define CPL_BLOCK_COUNT 27 + +#if defined SMU__VARIANT__ICELAND + #define SX_BLOCK_COUNT 8 + #define MC_BLOCK_COUNT 1 + #define CPL_BLOCK_COUNT 29 +#endif + +struct SMU7_Local_Cac { + uint8_t BlockId; + uint8_t SignalId; + uint8_t Threshold; + uint8_t Padding; +}; + +typedef struct SMU7_Local_Cac SMU7_Local_Cac; + +struct SMU7_Local_Cac_Table { + SMU7_Local_Cac SxLocalCac[SX_BLOCK_COUNT]; + SMU7_Local_Cac CplLocalCac[CPL_BLOCK_COUNT]; + SMU7_Local_Cac McLocalCac[MC_BLOCK_COUNT]; +}; + +typedef struct SMU7_Local_Cac_Table SMU7_Local_Cac_Table; + +#if !defined(SMC_MICROCODE) +#pragma pack(pop) +#endif + +#endif + diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu71_discrete.h b/drivers/gpu/drm/amd/powerplay/inc/smu71_discrete.h new file mode 100644 index 000000000000..c0e3936d5c2e --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/inc/smu71_discrete.h @@ -0,0 +1,631 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef SMU71_DISCRETE_H +#define SMU71_DISCRETE_H + +#include "smu71.h" + +#if !defined(SMC_MICROCODE) +#pragma pack(push, 1) +#endif + +#define VDDC_ON_SVI2 0x1 +#define VDDCI_ON_SVI2 0x2 +#define MVDD_ON_SVI2 0x4 + +struct SMU71_Discrete_VoltageLevel +{ + uint16_t Voltage; + uint16_t StdVoltageHiSidd; + uint16_t StdVoltageLoSidd; + uint8_t Smio; + uint8_t padding; +}; + +typedef struct SMU71_Discrete_VoltageLevel SMU71_Discrete_VoltageLevel; + +struct SMU71_Discrete_GraphicsLevel +{ + uint32_t MinVddc; + uint32_t MinVddcPhases; + + uint32_t SclkFrequency; + + uint8_t pcieDpmLevel; + uint8_t DeepSleepDivId; + uint16_t ActivityLevel; + + uint32_t CgSpllFuncCntl3; + uint32_t CgSpllFuncCntl4; + uint32_t SpllSpreadSpectrum; + uint32_t SpllSpreadSpectrum2; + uint32_t CcPwrDynRm; + uint32_t CcPwrDynRm1; + uint8_t SclkDid; + uint8_t DisplayWatermark; + uint8_t EnabledForActivity; + uint8_t EnabledForThrottle; + uint8_t UpHyst; + uint8_t DownHyst; + uint8_t VoltageDownHyst; + uint8_t PowerThrottle; +}; + +typedef struct SMU71_Discrete_GraphicsLevel SMU71_Discrete_GraphicsLevel; + +struct SMU71_Discrete_ACPILevel +{ + uint32_t Flags; + uint32_t MinVddc; + uint32_t MinVddcPhases; + uint32_t SclkFrequency; + uint8_t SclkDid; + uint8_t DisplayWatermark; + uint8_t DeepSleepDivId; + uint8_t padding; + uint32_t CgSpllFuncCntl; + uint32_t CgSpllFuncCntl2; + uint32_t CgSpllFuncCntl3; + uint32_t CgSpllFuncCntl4; + uint32_t SpllSpreadSpectrum; + uint32_t SpllSpreadSpectrum2; + uint32_t CcPwrDynRm; + uint32_t CcPwrDynRm1; +}; + +typedef struct SMU71_Discrete_ACPILevel SMU71_Discrete_ACPILevel; + +struct SMU71_Discrete_Ulv +{ + uint32_t CcPwrDynRm; + uint32_t CcPwrDynRm1; + uint16_t VddcOffset; + uint8_t VddcOffsetVid; + uint8_t VddcPhase; + uint32_t Reserved; +}; + +typedef struct SMU71_Discrete_Ulv SMU71_Discrete_Ulv; + +struct SMU71_Discrete_MemoryLevel +{ + uint32_t MinVddc; + uint32_t MinVddcPhases; + uint32_t MinVddci; + uint32_t MinMvdd; + + uint32_t MclkFrequency; + + uint8_t EdcReadEnable; + uint8_t EdcWriteEnable; + uint8_t RttEnable; + uint8_t StutterEnable; + + uint8_t StrobeEnable; + uint8_t StrobeRatio; + uint8_t EnabledForThrottle; + uint8_t EnabledForActivity; + + uint8_t UpHyst; + uint8_t DownHyst; + uint8_t VoltageDownHyst; + uint8_t padding; + + uint16_t ActivityLevel; + uint8_t DisplayWatermark; + uint8_t padding1; + + uint32_t MpllFuncCntl; + uint32_t MpllFuncCntl_1; + uint32_t MpllFuncCntl_2; + uint32_t MpllAdFuncCntl; + uint32_t MpllDqFuncCntl; + uint32_t MclkPwrmgtCntl; + uint32_t DllCntl; + uint32_t MpllSs1; + uint32_t MpllSs2; +}; + +typedef struct SMU71_Discrete_MemoryLevel SMU71_Discrete_MemoryLevel; + +struct SMU71_Discrete_LinkLevel +{ + uint8_t PcieGenSpeed; ///< 0:PciE-gen1 1:PciE-gen2 2:PciE-gen3 + uint8_t PcieLaneCount; ///< 1=x1, 2=x2, 3=x4, 4=x8, 5=x12, 6=x16 + uint8_t EnabledForActivity; + uint8_t SPC; + uint32_t DownThreshold; + uint32_t UpThreshold; + uint32_t Reserved; +}; + +typedef struct SMU71_Discrete_LinkLevel SMU71_Discrete_LinkLevel; + + +#ifdef SMU__DYNAMIC_MCARB_SETTINGS +// MC ARB DRAM Timing registers. +struct SMU71_Discrete_MCArbDramTimingTableEntry +{ + uint32_t McArbDramTiming; + uint32_t McArbDramTiming2; + uint8_t McArbBurstTime; + uint8_t padding[3]; +}; + +typedef struct SMU71_Discrete_MCArbDramTimingTableEntry SMU71_Discrete_MCArbDramTimingTableEntry; + +struct SMU71_Discrete_MCArbDramTimingTable +{ + SMU71_Discrete_MCArbDramTimingTableEntry entries[SMU__NUM_SCLK_DPM_STATE][SMU__NUM_MCLK_DPM_LEVELS]; +}; + +typedef struct SMU71_Discrete_MCArbDramTimingTable SMU71_Discrete_MCArbDramTimingTable; +#endif + +// UVD VCLK/DCLK state (level) definition. +struct SMU71_Discrete_UvdLevel +{ + uint32_t VclkFrequency; + uint32_t DclkFrequency; + uint16_t MinVddc; + uint8_t MinVddcPhases; + uint8_t VclkDivider; + uint8_t DclkDivider; + uint8_t padding[3]; +}; + +typedef struct SMU71_Discrete_UvdLevel SMU71_Discrete_UvdLevel; + +// Clocks for other external blocks (VCE, ACP, SAMU). +struct SMU71_Discrete_ExtClkLevel +{ + uint32_t Frequency; + uint16_t MinVoltage; + uint8_t MinPhases; + uint8_t Divider; +}; + +typedef struct SMU71_Discrete_ExtClkLevel SMU71_Discrete_ExtClkLevel; + +// Everything that we need to keep track of about the current state. +// Use this instead of copies of the GraphicsLevel and MemoryLevel structures to keep track of state parameters +// that need to be checked later. +// We don't need to cache everything about a state, just a few parameters. +struct SMU71_Discrete_StateInfo +{ + uint32_t SclkFrequency; + uint32_t MclkFrequency; + uint32_t VclkFrequency; + uint32_t DclkFrequency; + uint32_t SamclkFrequency; + uint32_t AclkFrequency; + uint32_t EclkFrequency; + uint16_t MvddVoltage; + uint16_t padding16; + uint8_t DisplayWatermark; + uint8_t McArbIndex; + uint8_t McRegIndex; + uint8_t SeqIndex; + uint8_t SclkDid; + int8_t SclkIndex; + int8_t MclkIndex; + uint8_t PCIeGen; + +}; + +typedef struct SMU71_Discrete_StateInfo SMU71_Discrete_StateInfo; + + +struct SMU71_Discrete_DpmTable +{ + // Multi-DPM controller settings + SMU71_PIDController GraphicsPIDController; + SMU71_PIDController MemoryPIDController; + SMU71_PIDController LinkPIDController; + + uint32_t SystemFlags; + + // SMIO masks for voltage and phase controls + uint32_t SmioMaskVddcVid; + uint32_t SmioMaskVddcPhase; + uint32_t SmioMaskVddciVid; + uint32_t SmioMaskMvddVid; + + uint32_t VddcLevelCount; + uint32_t VddciLevelCount; + uint32_t MvddLevelCount; + + SMU71_Discrete_VoltageLevel VddcLevel [SMU71_MAX_LEVELS_VDDC]; + SMU71_Discrete_VoltageLevel VddciLevel [SMU71_MAX_LEVELS_VDDCI]; + SMU71_Discrete_VoltageLevel MvddLevel [SMU71_MAX_LEVELS_MVDD]; + + uint8_t GraphicsDpmLevelCount; + uint8_t MemoryDpmLevelCount; + uint8_t LinkLevelCount; + uint8_t MasterDeepSleepControl; + + uint32_t Reserved[5]; + + // State table entries for each DPM state + SMU71_Discrete_GraphicsLevel GraphicsLevel [SMU71_MAX_LEVELS_GRAPHICS]; + SMU71_Discrete_MemoryLevel MemoryACPILevel; + SMU71_Discrete_MemoryLevel MemoryLevel [SMU71_MAX_LEVELS_MEMORY]; + SMU71_Discrete_LinkLevel LinkLevel [SMU71_MAX_LEVELS_LINK]; + SMU71_Discrete_ACPILevel ACPILevel; + + uint32_t SclkStepSize; + uint32_t Smio [SMU71_MAX_ENTRIES_SMIO]; + + uint8_t GraphicsBootLevel; + uint8_t GraphicsVoltageChangeEnable; + uint8_t GraphicsThermThrottleEnable; + uint8_t GraphicsInterval; + + uint8_t VoltageInterval; + uint8_t ThermalInterval; + uint16_t TemperatureLimitHigh; + + uint16_t TemperatureLimitLow; + uint8_t MemoryBootLevel; + uint8_t MemoryVoltageChangeEnable; + + uint8_t MemoryInterval; + uint8_t MemoryThermThrottleEnable; + uint8_t MergedVddci; + uint8_t padding2; + + uint16_t VoltageResponseTime; + uint16_t PhaseResponseTime; + + uint8_t PCIeBootLinkLevel; + uint8_t PCIeGenInterval; + uint8_t DTEInterval; + uint8_t DTEMode; + + uint8_t SVI2Enable; + uint8_t VRHotGpio; + uint8_t AcDcGpio; + uint8_t ThermGpio; + + uint32_t DisplayCac; + + uint16_t MaxPwr; + uint16_t NomPwr; + + uint16_t FpsHighThreshold; + uint16_t FpsLowThreshold; + + uint16_t BAPMTI_R [SMU71_DTE_ITERATIONS][SMU71_DTE_SOURCES][SMU71_DTE_SINKS]; + uint16_t BAPMTI_RC [SMU71_DTE_ITERATIONS][SMU71_DTE_SOURCES][SMU71_DTE_SINKS]; + + uint8_t DTEAmbientTempBase; + uint8_t DTETjOffset; + uint8_t GpuTjMax; + uint8_t GpuTjHyst; + + uint16_t BootVddc; + uint16_t BootVddci; + + uint16_t BootMVdd; + uint16_t padding; + + uint32_t BAPM_TEMP_GRADIENT; + + uint32_t LowSclkInterruptThreshold; + uint32_t VddGfxReChkWait; + + uint16_t PPM_PkgPwrLimit; + uint16_t PPM_TemperatureLimit; + + uint16_t DefaultTdp; + uint16_t TargetTdp; +}; + +typedef struct SMU71_Discrete_DpmTable SMU71_Discrete_DpmTable; + +// --------------------------------------------------- AC Timing Parameters ------------------------------------------------ +#define SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE 16 +#define SMU71_DISCRETE_MC_REGISTER_ARRAY_SET_COUNT SMU71_MAX_LEVELS_MEMORY + +struct SMU71_Discrete_MCRegisterAddress +{ + uint16_t s0; + uint16_t s1; +}; + +typedef struct SMU71_Discrete_MCRegisterAddress SMU71_Discrete_MCRegisterAddress; + +struct SMU71_Discrete_MCRegisterSet +{ + uint32_t value[SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE]; +}; + +typedef struct SMU71_Discrete_MCRegisterSet SMU71_Discrete_MCRegisterSet; + +struct SMU71_Discrete_MCRegisters +{ + uint8_t last; + uint8_t reserved[3]; + SMU71_Discrete_MCRegisterAddress address[SMU71_DISCRETE_MC_REGISTER_ARRAY_SIZE]; + SMU71_Discrete_MCRegisterSet data[SMU71_DISCRETE_MC_REGISTER_ARRAY_SET_COUNT]; +}; + +typedef struct SMU71_Discrete_MCRegisters SMU71_Discrete_MCRegisters; + + +// --------------------------------------------------- Fan Table ----------------------------------------------------------- +struct SMU71_Discrete_FanTable +{ + uint16_t FdoMode; + int16_t TempMin; + int16_t TempMed; + int16_t TempMax; + int16_t Slope1; + int16_t Slope2; + int16_t FdoMin; + int16_t HystUp; + int16_t HystDown; + int16_t HystSlope; + int16_t TempRespLim; + int16_t TempCurr; + int16_t SlopeCurr; + int16_t PwmCurr; + uint32_t RefreshPeriod; + int16_t FdoMax; + uint8_t TempSrc; + int8_t Padding; +}; + +typedef struct SMU71_Discrete_FanTable SMU71_Discrete_FanTable; + +#define SMU7_DISCRETE_GPIO_SCLK_DEBUG 4 +#define SMU7_DISCRETE_GPIO_SCLK_DEBUG_BIT (0x1 << SMU7_DISCRETE_GPIO_SCLK_DEBUG) + +struct SMU71_MclkDpmScoreboard +{ + + uint32_t PercentageBusy; + + int32_t PIDError; + int32_t PIDIntegral; + int32_t PIDOutput; + + uint32_t SigmaDeltaAccum; + uint32_t SigmaDeltaOutput; + uint32_t SigmaDeltaLevel; + + uint32_t UtilizationSetpoint; + + uint8_t TdpClampMode; + uint8_t TdcClampMode; + uint8_t ThermClampMode; + uint8_t VoltageBusy; + + int8_t CurrLevel; + int8_t TargLevel; + uint8_t LevelChangeInProgress; + uint8_t UpHyst; + + uint8_t DownHyst; + uint8_t VoltageDownHyst; + uint8_t DpmEnable; + uint8_t DpmRunning; + + uint8_t DpmForce; + uint8_t DpmForceLevel; + uint8_t DisplayWatermark; + uint8_t McArbIndex; + + uint32_t MinimumPerfMclk; + + uint8_t AcpiReq; + uint8_t AcpiAck; + uint8_t MclkSwitchInProgress; + uint8_t MclkSwitchCritical; + + uint8_t TargetMclkIndex; + uint8_t TargetMvddIndex; + uint8_t MclkSwitchResult; + + uint8_t EnabledLevelsChange; + + uint16_t LevelResidencyCounters [SMU71_MAX_LEVELS_MEMORY]; + uint16_t LevelSwitchCounters [SMU71_MAX_LEVELS_MEMORY]; + + void (*TargetStateCalculator)(uint8_t); + void (*SavedTargetStateCalculator)(uint8_t); + + uint16_t AutoDpmInterval; + uint16_t AutoDpmRange; + + uint16_t MclkSwitchingTime; + uint8_t padding[2]; +}; + +typedef struct SMU71_MclkDpmScoreboard SMU71_MclkDpmScoreboard; + +struct SMU71_UlvScoreboard +{ + uint8_t EnterUlv; + uint8_t ExitUlv; + uint8_t UlvActive; + uint8_t WaitingForUlv; + uint8_t UlvEnable; + uint8_t UlvRunning; + uint8_t UlvMasterEnable; + uint8_t padding; + uint32_t UlvAbortedCount; + uint32_t UlvTimeStamp; +}; + +typedef struct SMU71_UlvScoreboard SMU71_UlvScoreboard; + +struct SMU71_VddGfxScoreboard +{ + uint8_t VddGfxEnable; + uint8_t VddGfxActive; + uint8_t padding[2]; + + uint32_t VddGfxEnteredCount; + uint32_t VddGfxAbortedCount; +}; + +typedef struct SMU71_VddGfxScoreboard SMU71_VddGfxScoreboard; + +struct SMU71_AcpiScoreboard { + uint32_t SavedInterruptMask[2]; + uint8_t LastACPIRequest; + uint8_t CgBifResp; + uint8_t RequestType; + uint8_t Padding; + SMU71_Discrete_ACPILevel D0Level; +}; + +typedef struct SMU71_AcpiScoreboard SMU71_AcpiScoreboard; + + +struct SMU71_Discrete_PmFuses { + // dw0-dw1 + uint8_t BapmVddCVidHiSidd[8]; + + // dw2-dw3 + uint8_t BapmVddCVidLoSidd[8]; + + // dw4-dw5 + uint8_t VddCVid[8]; + + // dw6 + uint8_t SviLoadLineEn; + uint8_t SviLoadLineVddC; + uint8_t SviLoadLineTrimVddC; + uint8_t SviLoadLineOffsetVddC; + + // dw7 + uint16_t TDC_VDDC_PkgLimit; + uint8_t TDC_VDDC_ThrottleReleaseLimitPerc; + uint8_t TDC_MAWt; + + // dw8 + uint8_t TdcWaterfallCtl; + uint8_t LPMLTemperatureMin; + uint8_t LPMLTemperatureMax; + uint8_t Reserved; + + // dw9-dw12 + uint8_t LPMLTemperatureScaler[16]; + + // dw13-dw14 + int16_t FuzzyFan_ErrorSetDelta; + int16_t FuzzyFan_ErrorRateSetDelta; + int16_t FuzzyFan_PwmSetDelta; + uint16_t Reserved6; + + // dw15 + uint8_t GnbLPML[16]; + + // dw15 + uint8_t GnbLPMLMaxVid; + uint8_t GnbLPMLMinVid; + uint8_t Reserved1[2]; + + // dw16 + uint16_t BapmVddCBaseLeakageHiSidd; + uint16_t BapmVddCBaseLeakageLoSidd; +}; + +typedef struct SMU71_Discrete_PmFuses SMU71_Discrete_PmFuses; + +struct SMU71_Discrete_Log_Header_Table { + uint32_t version; + uint32_t asic_id; + uint16_t flags; + uint16_t entry_size; + uint32_t total_size; + uint32_t num_of_entries; + uint8_t type; + uint8_t mode; + uint8_t filler_0[2]; + uint32_t filler_1[2]; +}; + +typedef struct SMU71_Discrete_Log_Header_Table SMU71_Discrete_Log_Header_Table; + +struct SMU71_Discrete_Log_Cntl { + uint8_t Enabled; + uint8_t Type; + uint8_t padding[2]; + uint32_t BufferSize; + uint32_t SamplesLogged; + uint32_t SampleSize; + uint32_t AddrL; + uint32_t AddrH; +}; + +typedef struct SMU71_Discrete_Log_Cntl SMU71_Discrete_Log_Cntl; + +#if defined SMU__DGPU_ONLY + #define CAC_ACC_NW_NUM_OF_SIGNALS 83 +#endif + + +struct SMU71_Discrete_Cac_Collection_Table { + uint32_t temperature; + uint32_t cac_acc_nw[CAC_ACC_NW_NUM_OF_SIGNALS]; + uint32_t filler[4]; +}; + +typedef struct SMU71_Discrete_Cac_Collection_Table SMU71_Discrete_Cac_Collection_Table; + +struct SMU71_Discrete_Cac_Verification_Table { + uint32_t VddcTotalPower; + uint32_t VddcLeakagePower; + uint32_t VddcConstantPower; + uint32_t VddcGfxDynamicPower; + uint32_t VddcUvdDynamicPower; + uint32_t VddcVceDynamicPower; + uint32_t VddcAcpDynamicPower; + uint32_t VddcPcieDynamicPower; + uint32_t VddcDceDynamicPower; + uint32_t VddcCurrent; + uint32_t VddcVoltage; + uint32_t VddciTotalPower; + uint32_t VddciLeakagePower; + uint32_t VddciConstantPower; + uint32_t VddciDynamicPower; + uint32_t Vddr1TotalPower; + uint32_t Vddr1LeakagePower; + uint32_t Vddr1ConstantPower; + uint32_t Vddr1DynamicPower; + uint32_t spare[8]; + uint32_t temperature; +}; + +typedef struct SMU71_Discrete_Cac_Verification_Table SMU71_Discrete_Cac_Verification_Table; + +#if !defined(SMC_MICROCODE) +#pragma pack(pop) +#endif + + +#endif + diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/Makefile b/drivers/gpu/drm/amd/powerplay/smumgr/Makefile index f10fb64ef981..19e79469f6bc 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/Makefile +++ b/drivers/gpu/drm/amd/powerplay/smumgr/Makefile @@ -2,7 +2,8 @@ # Makefile for the 'smu manager' sub-component of powerplay. # It provides the smu management services for the driver. -SMU_MGR = smumgr.o cz_smumgr.o tonga_smumgr.o fiji_smumgr.o polaris10_smumgr.o +SMU_MGR = smumgr.o cz_smumgr.o tonga_smumgr.o fiji_smumgr.o \ + polaris10_smumgr.o iceland_smumgr.o AMD_PP_SMUMGR = $(addprefix $(AMD_PP_PATH)/smumgr/,$(SMU_MGR)) diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/iceland_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/iceland_smumgr.c new file mode 100644 index 000000000000..f50658332d9d --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/smumgr/iceland_smumgr.c @@ -0,0 +1,713 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Author: Huang Rui <ray.huang@amd.com> + * + */ +#include <linux/types.h> +#include <linux/kernel.h> +#include <linux/slab.h> +#include <linux/gfp.h> + +#include "smumgr.h" +#include "iceland_smumgr.h" +#include "pp_debug.h" +#include "smu_ucode_xfer_vi.h" +#include "ppsmc.h" +#include "smu/smu_7_1_1_d.h" +#include "smu/smu_7_1_1_sh_mask.h" +#include "cgs_common.h" + +#define ICELAND_SMC_SIZE 0x20000 +#define BUFFER_SIZE 80000 +#define MAX_STRING_SIZE 15 +#define BUFFER_SIZETWO 131072 /*128 *1024*/ + +/** + * Set the address for reading/writing the SMC SRAM space. + * @param smumgr the address of the powerplay hardware manager. + * @param smcAddress the address in the SMC RAM to access. + */ +static int iceland_set_smc_sram_address(struct pp_smumgr *smumgr, + uint32_t smcAddress, uint32_t limit) +{ + if (smumgr == NULL || smumgr->device == NULL) + return -EINVAL; + PP_ASSERT_WITH_CODE((0 == (3 & smcAddress)), + "SMC address must be 4 byte aligned.", + return -1;); + + PP_ASSERT_WITH_CODE((limit > (smcAddress + 3)), + "SMC address is beyond the SMC RAM area.", + return -1;); + + cgs_write_register(smumgr->device, mmSMC_IND_INDEX_0, smcAddress); + SMUM_WRITE_FIELD(smumgr->device, SMC_IND_ACCESS_CNTL, AUTO_INCREMENT_IND_0, 0); + + return 0; +} + +/** + * Copy bytes from an array into the SMC RAM space. + * + * @param smumgr the address of the powerplay SMU manager. + * @param smcStartAddress the start address in the SMC RAM to copy bytes to. + * @param src the byte array to copy the bytes from. + * @param byteCount the number of bytes to copy. + */ +int iceland_copy_bytes_to_smc(struct pp_smumgr *smumgr, + uint32_t smcStartAddress, const uint8_t *src, + uint32_t byteCount, uint32_t limit) +{ + uint32_t addr; + uint32_t data, orig_data; + int result = 0; + uint32_t extra_shift; + + if (smumgr == NULL || smumgr->device == NULL) + return -EINVAL; + PP_ASSERT_WITH_CODE((0 == (3 & smcStartAddress)), + "SMC address must be 4 byte aligned.", + return 0;); + + PP_ASSERT_WITH_CODE((limit > (smcStartAddress + byteCount)), + "SMC address is beyond the SMC RAM area.", + return 0;); + + addr = smcStartAddress; + + while (byteCount >= 4) { + /* + * Bytes are written into the + * SMC address space with the MSB first + */ + data = (src[0] << 24) + (src[1] << 16) + (src[2] << 8) + src[3]; + + result = iceland_set_smc_sram_address(smumgr, addr, limit); + + if (result) + goto out; + + cgs_write_register(smumgr->device, mmSMC_IND_DATA_0, data); + + src += 4; + byteCount -= 4; + addr += 4; + } + + if (0 != byteCount) { + /* Now write odd bytes left, do a read modify write cycle */ + data = 0; + + result = iceland_set_smc_sram_address(smumgr, addr, limit); + if (result) + goto out; + + orig_data = cgs_read_register(smumgr->device, + mmSMC_IND_DATA_0); + extra_shift = 8 * (4 - byteCount); + + while (byteCount > 0) { + data = (data << 8) + *src++; + byteCount--; + } + + data <<= extra_shift; + data |= (orig_data & ~((~0UL) << extra_shift)); + + result = iceland_set_smc_sram_address(smumgr, addr, limit); + if (result) + goto out; + + cgs_write_register(smumgr->device, mmSMC_IND_DATA_0, data); + } + +out: + return result; +} + +/** + * Deassert the reset'pin' (set it to high). + * + * @param smumgr the address of the powerplay hardware manager. + */ +static int iceland_start_smc(struct pp_smumgr *smumgr) +{ + SMUM_WRITE_INDIRECT_FIELD(smumgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_RESET_CNTL, rst_reg, 0); + + return 0; +} + +static void iceland_pp_reset_smc(struct pp_smumgr *smumgr) +{ + SMUM_WRITE_INDIRECT_FIELD(smumgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_RESET_CNTL, + rst_reg, 1); +} + +int iceland_program_jump_on_start(struct pp_smumgr *smumgr) +{ + static const unsigned char pData[] = { 0xE0, 0x00, 0x80, 0x40 }; + + iceland_copy_bytes_to_smc(smumgr, 0x0, pData, 4, sizeof(pData)+1); + + return 0; +} + +/** + * Return if the SMC is currently running. + * + * @param smumgr the address of the powerplay hardware manager. + */ +bool iceland_is_smc_ram_running(struct pp_smumgr *smumgr) +{ + uint32_t val1, val2; + + val1 = SMUM_READ_INDIRECT_FIELD(smumgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_CLOCK_CNTL_0, ck_disable); + val2 = cgs_read_ind_register(smumgr->device, CGS_IND_REG__SMC, + ixSMC_PC_C); + + return ((0 == val1) && (0x20100 <= val2)); +} + +/** + * Send a message to the SMC, and wait for its response. + * + * @param smumgr the address of the powerplay hardware manager. + * @param msg the message to send. + * @return The response that came from the SMC. + */ +static int iceland_send_msg_to_smc(struct pp_smumgr *smumgr, uint16_t msg) +{ + if (smumgr == NULL || smumgr->device == NULL) + return -EINVAL; + + if (!iceland_is_smc_ram_running(smumgr)) + return -EINVAL; + + SMUM_WAIT_FIELD_UNEQUAL(smumgr, SMC_RESP_0, SMC_RESP, 0); + PP_ASSERT_WITH_CODE( + 1 == SMUM_READ_FIELD(smumgr->device, SMC_RESP_0, SMC_RESP), + "Failed to send Previous Message.", + ); + + cgs_write_register(smumgr->device, mmSMC_MESSAGE_0, msg); + + SMUM_WAIT_FIELD_UNEQUAL(smumgr, SMC_RESP_0, SMC_RESP, 0); + PP_ASSERT_WITH_CODE( + 1 == SMUM_READ_FIELD(smumgr->device, SMC_RESP_0, SMC_RESP), + "Failed to send Message.", + ); + + return 0; +} + +/** + * Send a message to the SMC with parameter + * + * @param smumgr: the address of the powerplay hardware manager. + * @param msg: the message to send. + * @param parameter: the parameter to send + * @return The response that came from the SMC. + */ +static int iceland_send_msg_to_smc_with_parameter(struct pp_smumgr *smumgr, + uint16_t msg, uint32_t parameter) +{ + if (smumgr == NULL || smumgr->device == NULL) + return -EINVAL; + + cgs_write_register(smumgr->device, mmSMC_MSG_ARG_0, parameter); + + return iceland_send_msg_to_smc(smumgr, msg); +} + +/* + * Read a 32bit value from the SMC SRAM space. + * ALL PARAMETERS ARE IN HOST BYTE ORDER. + * @param smumgr the address of the powerplay hardware manager. + * @param smcAddress the address in the SMC RAM to access. + * @param value and output parameter for the data read from the SMC SRAM. + */ +int iceland_read_smc_sram_dword(struct pp_smumgr *smumgr, + uint32_t smcAddress, uint32_t *value, + uint32_t limit) +{ + int result; + + result = iceland_set_smc_sram_address(smumgr, smcAddress, limit); + + if (0 != result) + return result; + + *value = cgs_read_register(smumgr->device, mmSMC_IND_DATA_0); + + return 0; +} + +/* + * Write a 32bit value to the SMC SRAM space. + * ALL PARAMETERS ARE IN HOST BYTE ORDER. + * @param smumgr the address of the powerplay hardware manager. + * @param smcAddress the address in the SMC RAM to access. + * @param value to write to the SMC SRAM. + */ +int iceland_write_smc_sram_dword(struct pp_smumgr *smumgr, + uint32_t smcAddress, uint32_t value, + uint32_t limit) +{ + int result; + + result = iceland_set_smc_sram_address(smumgr, smcAddress, limit); + + if (0 != result) + return result; + + cgs_write_register(smumgr->device, mmSMC_IND_DATA_0, value); + + return 0; +} + +static int iceland_smu_fini(struct pp_smumgr *smumgr) +{ + struct iceland_smumgr *priv = (struct iceland_smumgr *)(smumgr->backend); + + smu_free_memory(smumgr->device, (void *)priv->header_buffer.handle); + + if (smumgr->backend != NULL) { + kfree(smumgr->backend); + smumgr->backend = NULL; + } + + cgs_rel_firmware(smumgr->device, CGS_UCODE_ID_SMU); + return 0; +} + +static enum cgs_ucode_id iceland_convert_fw_type_to_cgs(uint32_t fw_type) +{ + enum cgs_ucode_id result = CGS_UCODE_ID_MAXIMUM; + + switch (fw_type) { + case UCODE_ID_SMU: + result = CGS_UCODE_ID_SMU; + break; + case UCODE_ID_SDMA0: + result = CGS_UCODE_ID_SDMA0; + break; + case UCODE_ID_SDMA1: + result = CGS_UCODE_ID_SDMA1; + break; + case UCODE_ID_CP_CE: + result = CGS_UCODE_ID_CP_CE; + break; + case UCODE_ID_CP_PFP: + result = CGS_UCODE_ID_CP_PFP; + break; + case UCODE_ID_CP_ME: + result = CGS_UCODE_ID_CP_ME; + break; + case UCODE_ID_CP_MEC: + result = CGS_UCODE_ID_CP_MEC; + break; + case UCODE_ID_CP_MEC_JT1: + result = CGS_UCODE_ID_CP_MEC_JT1; + break; + case UCODE_ID_CP_MEC_JT2: + result = CGS_UCODE_ID_CP_MEC_JT2; + break; + case UCODE_ID_RLC_G: + result = CGS_UCODE_ID_RLC_G; + break; + default: + break; + } + + return result; +} + +/** + * Convert the PPIRI firmware type to SMU type mask. + * For MEC, we need to check all MEC related type + */ +static uint16_t iceland_get_mask_for_firmware_type(uint16_t firmwareType) +{ + uint16_t result = 0; + + switch (firmwareType) { + case UCODE_ID_SDMA0: + result = UCODE_ID_SDMA0_MASK; + break; + case UCODE_ID_SDMA1: + result = UCODE_ID_SDMA1_MASK; + break; + case UCODE_ID_CP_CE: + result = UCODE_ID_CP_CE_MASK; + break; + case UCODE_ID_CP_PFP: + result = UCODE_ID_CP_PFP_MASK; + break; + case UCODE_ID_CP_ME: + result = UCODE_ID_CP_ME_MASK; + break; + case UCODE_ID_CP_MEC: + case UCODE_ID_CP_MEC_JT1: + case UCODE_ID_CP_MEC_JT2: + result = UCODE_ID_CP_MEC_MASK; + break; + case UCODE_ID_RLC_G: + result = UCODE_ID_RLC_G_MASK; + break; + default: + break; + } + + return result; +} + +/** + * Check if the FW has been loaded, + * SMU will not return if loading has not finished. +*/ +static int iceland_check_fw_load_finish(struct pp_smumgr *smumgr, uint32_t fwType) +{ + uint16_t fwMask = iceland_get_mask_for_firmware_type(fwType); + + if (0 != SMUM_WAIT_VFPF_INDIRECT_REGISTER(smumgr, SMC_IND, + SOFT_REGISTERS_TABLE_27, fwMask, fwMask)) { + pr_err("[ powerplay ] check firmware loading failed\n"); + return -EINVAL; + } + + return 0; +} + +/* Populate one firmware image to the data structure */ +static int iceland_populate_single_firmware_entry(struct pp_smumgr *smumgr, + uint16_t firmware_type, + struct SMU_Entry *pentry) +{ + int result; + struct cgs_firmware_info info = {0}; + + result = cgs_get_firmware_info( + smumgr->device, + iceland_convert_fw_type_to_cgs(firmware_type), + &info); + + if (result == 0) { + pentry->version = 0; + pentry->id = (uint16_t)firmware_type; + pentry->image_addr_high = smu_upper_32_bits(info.mc_addr); + pentry->image_addr_low = smu_lower_32_bits(info.mc_addr); + pentry->meta_data_addr_high = 0; + pentry->meta_data_addr_low = 0; + pentry->data_size_byte = info.image_size; + pentry->num_register_entries = 0; + + if (firmware_type == UCODE_ID_RLC_G) + pentry->flags = 1; + else + pentry->flags = 0; + } else { + return result; + } + + return result; +} + +static void iceland_pp_stop_smc_clock(struct pp_smumgr *smumgr) +{ + SMUM_WRITE_INDIRECT_FIELD(smumgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_CLOCK_CNTL_0, + ck_disable, 1); +} + +static void iceland_start_smc_clock(struct pp_smumgr *smumgr) +{ + SMUM_WRITE_INDIRECT_FIELD(smumgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_CLOCK_CNTL_0, + ck_disable, 0); +} + +int iceland_smu_start_smc(struct pp_smumgr *smumgr) +{ + /* set smc instruct start point at 0x0 */ + iceland_program_jump_on_start(smumgr); + + /* enable smc clock */ + iceland_start_smc_clock(smumgr); + + /* de-assert reset */ + iceland_start_smc(smumgr); + + SMUM_WAIT_INDIRECT_FIELD(smumgr, SMC_IND, FIRMWARE_FLAGS, + INTERRUPTS_ENABLED, 1); + + return 0; +} + +/** + * Upload the SMC firmware to the SMC microcontroller. + * + * @param smumgr the address of the powerplay hardware manager. + * @param pFirmware the data structure containing the various sections of the firmware. + */ +int iceland_smu_upload_firmware_image(struct pp_smumgr *smumgr) +{ + const uint8_t *src; + uint32_t byte_count, val; + uint32_t data; + struct cgs_firmware_info info = {0}; + + if (smumgr == NULL || smumgr->device == NULL) + return -EINVAL; + + /* load SMC firmware */ + cgs_get_firmware_info(smumgr->device, + iceland_convert_fw_type_to_cgs(UCODE_ID_SMU), &info); + + if (info.image_size & 3) { + pr_err("[ powerplay ] SMC ucode is not 4 bytes aligned\n"); + return -EINVAL; + } + + if (info.image_size > ICELAND_SMC_SIZE) { + pr_err("[ powerplay ] SMC address is beyond the SMC RAM area\n"); + return -EINVAL; + } + + /* wait for smc boot up */ + SMUM_WAIT_INDIRECT_FIELD_UNEQUAL(smumgr, SMC_IND, + RCU_UC_EVENTS, boot_seq_done, 0); + + /* clear firmware interrupt enable flag */ + val = cgs_read_ind_register(smumgr->device, CGS_IND_REG__SMC, + ixSMC_SYSCON_MISC_CNTL); + cgs_write_ind_register(smumgr->device, CGS_IND_REG__SMC, + ixSMC_SYSCON_MISC_CNTL, val | 1); + + /* stop smc clock */ + iceland_pp_stop_smc_clock(smumgr); + + /* reset smc */ + iceland_pp_reset_smc(smumgr); + + cgs_write_register(smumgr->device, mmSMC_IND_INDEX_0, + info.ucode_start_address); + + SMUM_WRITE_FIELD(smumgr->device, SMC_IND_ACCESS_CNTL, + AUTO_INCREMENT_IND_0, 1); + + byte_count = info.image_size; + src = (const uint8_t *)info.kptr; + + while (byte_count >= 4) { + data = (src[0] << 24) + (src[1] << 16) + (src[2] << 8) + src[3]; + cgs_write_register(smumgr->device, mmSMC_IND_DATA_0, data); + src += 4; + byte_count -= 4; + } + + SMUM_WRITE_FIELD(smumgr->device, SMC_IND_ACCESS_CNTL, + AUTO_INCREMENT_IND_0, 0); + + return 0; +} + +static int iceland_request_smu_reload_fw(struct pp_smumgr *smumgr) +{ + struct iceland_smumgr *iceland_smu = + (struct iceland_smumgr *)(smumgr->backend); + uint16_t fw_to_load; + int result = 0; + struct SMU_DRAMData_TOC *toc; + + toc = (struct SMU_DRAMData_TOC *)iceland_smu->pHeader; + toc->num_entries = 0; + toc->structure_version = 1; + + PP_ASSERT_WITH_CODE( + 0 == iceland_populate_single_firmware_entry(smumgr, + UCODE_ID_RLC_G, + &toc->entry[toc->num_entries++]), + "Failed to Get Firmware Entry.\n", + return -1); + PP_ASSERT_WITH_CODE( + 0 == iceland_populate_single_firmware_entry(smumgr, + UCODE_ID_CP_CE, + &toc->entry[toc->num_entries++]), + "Failed to Get Firmware Entry.\n", + return -1); + PP_ASSERT_WITH_CODE( + 0 == iceland_populate_single_firmware_entry + (smumgr, UCODE_ID_CP_PFP, &toc->entry[toc->num_entries++]), + "Failed to Get Firmware Entry.\n", return -1); + PP_ASSERT_WITH_CODE( + 0 == iceland_populate_single_firmware_entry + (smumgr, UCODE_ID_CP_ME, &toc->entry[toc->num_entries++]), + "Failed to Get Firmware Entry.\n", return -1); + PP_ASSERT_WITH_CODE( + 0 == iceland_populate_single_firmware_entry + (smumgr, UCODE_ID_CP_MEC, &toc->entry[toc->num_entries++]), + "Failed to Get Firmware Entry.\n", return -1); + PP_ASSERT_WITH_CODE( + 0 == iceland_populate_single_firmware_entry + (smumgr, UCODE_ID_CP_MEC_JT1, &toc->entry[toc->num_entries++]), + "Failed to Get Firmware Entry.\n", return -1); + PP_ASSERT_WITH_CODE( + 0 == iceland_populate_single_firmware_entry + (smumgr, UCODE_ID_CP_MEC_JT2, &toc->entry[toc->num_entries++]), + "Failed to Get Firmware Entry.\n", return -1); + PP_ASSERT_WITH_CODE( + 0 == iceland_populate_single_firmware_entry + (smumgr, UCODE_ID_SDMA0, &toc->entry[toc->num_entries++]), + "Failed to Get Firmware Entry.\n", return -1); + PP_ASSERT_WITH_CODE( + 0 == iceland_populate_single_firmware_entry + (smumgr, UCODE_ID_SDMA1, &toc->entry[toc->num_entries++]), + "Failed to Get Firmware Entry.\n", return -1); + + if (!iceland_is_smc_ram_running(smumgr)) { + result = iceland_smu_upload_firmware_image(smumgr); + if (result) + return result; + + result = iceland_smu_start_smc(smumgr); + if (result) + return result; + } + + iceland_send_msg_to_smc_with_parameter(smumgr, + PPSMC_MSG_DRV_DRAM_ADDR_HI, + iceland_smu->header_buffer.mc_addr_high); + + iceland_send_msg_to_smc_with_parameter(smumgr, + PPSMC_MSG_DRV_DRAM_ADDR_LO, + iceland_smu->header_buffer.mc_addr_low); + + fw_to_load = UCODE_ID_RLC_G_MASK + + UCODE_ID_SDMA0_MASK + + UCODE_ID_SDMA1_MASK + + UCODE_ID_CP_CE_MASK + + UCODE_ID_CP_ME_MASK + + UCODE_ID_CP_PFP_MASK + + UCODE_ID_CP_MEC_MASK + + UCODE_ID_CP_MEC_JT1_MASK + + UCODE_ID_CP_MEC_JT2_MASK; + + PP_ASSERT_WITH_CODE( + 0 == iceland_send_msg_to_smc_with_parameter( + smumgr, PPSMC_MSG_LoadUcodes, fw_to_load), + "Fail to Request SMU Load uCode", return 0); + + return result; +} + +static int iceland_request_smu_load_specific_fw(struct pp_smumgr *smumgr, + uint32_t firmwareType) +{ + return 0; +} + +static int iceland_start_smu(struct pp_smumgr *smumgr) +{ + int result; + + result = iceland_smu_upload_firmware_image(smumgr); + if (result) + return result; + + result = iceland_smu_start_smc(smumgr); + if (result) + return result; + + result = iceland_request_smu_reload_fw(smumgr); + + return result; +} + +/** + * Write a 32bit value to the SMC SRAM space. + * ALL PARAMETERS ARE IN HOST BYTE ORDER. + * @param smumgr the address of the powerplay hardware manager. + * @param smcAddress the address in the SMC RAM to access. + * @param value to write to the SMC SRAM. + */ +static int iceland_smu_init(struct pp_smumgr *smumgr) +{ + struct iceland_smumgr *iceland_smu; + uint64_t mc_addr = 0; + + /* Allocate memory for backend private data */ + iceland_smu = (struct iceland_smumgr *)(smumgr->backend); + iceland_smu->header_buffer.data_size = + ((sizeof(struct SMU_DRAMData_TOC) / 4096) + 1) * 4096; + + smu_allocate_memory(smumgr->device, + iceland_smu->header_buffer.data_size, + CGS_GPU_MEM_TYPE__VISIBLE_CONTIG_FB, + PAGE_SIZE, + &mc_addr, + &iceland_smu->header_buffer.kaddr, + &iceland_smu->header_buffer.handle); + + iceland_smu->pHeader = iceland_smu->header_buffer.kaddr; + iceland_smu->header_buffer.mc_addr_high = smu_upper_32_bits(mc_addr); + iceland_smu->header_buffer.mc_addr_low = smu_lower_32_bits(mc_addr); + + PP_ASSERT_WITH_CODE((NULL != iceland_smu->pHeader), + "Out of memory.", + kfree(smumgr->backend); + cgs_free_gpu_mem(smumgr->device, + (cgs_handle_t)iceland_smu->header_buffer.handle); + return -1); + + return 0; +} + +static const struct pp_smumgr_func iceland_smu_funcs = { + .smu_init = &iceland_smu_init, + .smu_fini = &iceland_smu_fini, + .start_smu = &iceland_start_smu, + .check_fw_load_finish = &iceland_check_fw_load_finish, + .request_smu_load_fw = &iceland_request_smu_reload_fw, + .request_smu_load_specific_fw = &iceland_request_smu_load_specific_fw, + .send_msg_to_smc = &iceland_send_msg_to_smc, + .send_msg_to_smc_with_parameter = &iceland_send_msg_to_smc_with_parameter, + .download_pptable_settings = NULL, + .upload_pptable_settings = NULL, +}; + +int iceland_smum_init(struct pp_smumgr *smumgr) +{ + struct iceland_smumgr *iceland_smu = NULL; + + iceland_smu = kzalloc(sizeof(struct iceland_smumgr), GFP_KERNEL); + + if (iceland_smu == NULL) + return -ENOMEM; + + smumgr->backend = iceland_smu; + smumgr->smumgr_funcs = &iceland_smu_funcs; + + return 0; +} diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/iceland_smumgr.h b/drivers/gpu/drm/amd/powerplay/smumgr/iceland_smumgr.h new file mode 100644 index 000000000000..62009a7ae827 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/smumgr/iceland_smumgr.h @@ -0,0 +1,64 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Author: Huang Rui <ray.huang@amd.com> + * + */ + +#ifndef _ICELAND_SMUMGR_H_ +#define _ICELAND_SMUMGR_H_ + +struct iceland_buffer_entry { + uint32_t data_size; + uint32_t mc_addr_low; + uint32_t mc_addr_high; + void *kaddr; + unsigned long handle; +}; + +/* Iceland only has header_buffer, don't have smu buffer. */ +struct iceland_smumgr { + uint8_t *pHeader; + uint8_t *pMecImage; + uint32_t ulSoftRegsStart; + + struct iceland_buffer_entry header_buffer; +}; + +extern int iceland_smum_init(struct pp_smumgr *smumgr); +extern int iceland_copy_bytes_to_smc(struct pp_smumgr *smumgr, + uint32_t smcStartAddress, + const uint8_t *src, + uint32_t byteCount, uint32_t limit); + +extern int iceland_smu_start_smc(struct pp_smumgr *smumgr); + +extern int iceland_read_smc_sram_dword(struct pp_smumgr *smumgr, + uint32_t smcAddress, + uint32_t *value, uint32_t limit); +extern int iceland_write_smc_sram_dword(struct pp_smumgr *smumgr, + uint32_t smcAddress, + uint32_t value, uint32_t limit); + +extern bool iceland_is_smc_ram_running(struct pp_smumgr *smumgr); +extern int iceland_smu_upload_firmware_image(struct pp_smumgr *smumgr); + +#endif diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c index 5dba7c509710..704ff4cc0023 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c @@ -978,7 +978,7 @@ static int polaris10_smu_init(struct pp_smumgr *smumgr) return 0; } -static const struct pp_smumgr_func ellsemere_smu_funcs = { +static const struct pp_smumgr_func polaris10_smu_funcs = { .smu_init = polaris10_smu_init, .smu_fini = polaris10_smu_fini, .start_smu = polaris10_start_smu, @@ -1001,7 +1001,7 @@ int polaris10_smum_init(struct pp_smumgr *smumgr) return -1; smumgr->backend = polaris10_smu; - smumgr->smumgr_funcs = &ellsemere_smu_funcs; + smumgr->smumgr_funcs = &polaris10_smu_funcs; return 0; } diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c index 7723473e51a0..cf3cabee8918 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c @@ -30,6 +30,7 @@ #include "linux/delay.h" #include "cz_smumgr.h" #include "tonga_smumgr.h" +#include "iceland_smumgr.h" #include "fiji_smumgr.h" #include "polaris10_smumgr.h" @@ -58,6 +59,9 @@ int smum_init(struct amd_pp_init *pp_init, struct pp_instance *handle) break; case AMDGPU_FAMILY_VI: switch (smumgr->chip_id) { + case CHIP_TOPAZ: + iceland_smum_init(smumgr); + break; case CHIP_TONGA: tonga_smum_init(smumgr); break; diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index 7f2510524f09..7b21281c4b78 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -3480,14 +3480,23 @@ out: int drm_mode_page_flip_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) { - struct drm_mode_crtc_page_flip *page_flip = data; + struct drm_mode_crtc_page_flip_target *page_flip = data; struct drm_crtc *crtc; struct drm_framebuffer *fb = NULL; struct drm_pending_vblank_event *e = NULL; + u32 target_vblank = page_flip->sequence; int ret = -EINVAL; - if (page_flip->flags & ~DRM_MODE_PAGE_FLIP_FLAGS || - page_flip->reserved != 0) + if (page_flip->flags & ~DRM_MODE_PAGE_FLIP_FLAGS) + return -EINVAL; + + if (page_flip->sequence != 0 && !(page_flip->flags & DRM_MODE_PAGE_FLIP_TARGET)) + return -EINVAL; + + /* Only one of the DRM_MODE_PAGE_FLIP_TARGET_ABSOLUTE/RELATIVE flags + * can be specified + */ + if ((page_flip->flags & DRM_MODE_PAGE_FLIP_TARGET) == DRM_MODE_PAGE_FLIP_TARGET) return -EINVAL; if ((page_flip->flags & DRM_MODE_PAGE_FLIP_ASYNC) && !dev->mode_config.async_page_flip) @@ -3497,6 +3506,45 @@ int drm_mode_page_flip_ioctl(struct drm_device *dev, if (!crtc) return -ENOENT; + if (crtc->funcs->page_flip_target) { + u32 current_vblank; + int r; + + r = drm_crtc_vblank_get(crtc); + if (r) + return r; + + current_vblank = drm_crtc_vblank_count(crtc); + + switch (page_flip->flags & DRM_MODE_PAGE_FLIP_TARGET) { + case DRM_MODE_PAGE_FLIP_TARGET_ABSOLUTE: + if ((int)(target_vblank - current_vblank) > 1) { + DRM_DEBUG("Invalid absolute flip target %u, " + "must be <= %u\n", target_vblank, + current_vblank + 1); + drm_crtc_vblank_put(crtc); + return -EINVAL; + } + break; + case DRM_MODE_PAGE_FLIP_TARGET_RELATIVE: + if (target_vblank != 0 && target_vblank != 1) { + DRM_DEBUG("Invalid relative flip target %u, " + "must be 0 or 1\n", target_vblank); + drm_crtc_vblank_put(crtc); + return -EINVAL; + } + target_vblank += current_vblank; + break; + default: + target_vblank = current_vblank + + !(page_flip->flags & DRM_MODE_PAGE_FLIP_ASYNC); + break; + } + } else if (crtc->funcs->page_flip == NULL || + (page_flip->flags & DRM_MODE_PAGE_FLIP_TARGET)) { + return -EINVAL; + } + drm_modeset_lock_crtc(crtc, crtc->primary); if (crtc->primary->fb == NULL) { /* The framebuffer is currently unbound, presumably @@ -3507,9 +3555,6 @@ int drm_mode_page_flip_ioctl(struct drm_device *dev, goto out; } - if (crtc->funcs->page_flip == NULL) - goto out; - fb = drm_framebuffer_lookup(dev, page_flip->fb_id); if (!fb) { ret = -ENOENT; @@ -3550,7 +3595,12 @@ int drm_mode_page_flip_ioctl(struct drm_device *dev, } crtc->primary->old_fb = crtc->primary->fb; - ret = crtc->funcs->page_flip(crtc, fb, e, page_flip->flags); + if (crtc->funcs->page_flip_target) + ret = crtc->funcs->page_flip_target(crtc, fb, e, + page_flip->flags, + target_vblank); + else + ret = crtc->funcs->page_flip(crtc, fb, e, page_flip->flags); if (ret) { if (page_flip->flags & DRM_MODE_PAGE_FLIP_EVENT) drm_event_cancel_free(dev, &e->base); @@ -3563,6 +3613,8 @@ int drm_mode_page_flip_ioctl(struct drm_device *dev, } out: + if (ret) + drm_crtc_vblank_put(crtc); if (fb) drm_framebuffer_unreference(fb); if (crtc->primary->old_fb) diff --git a/drivers/gpu/drm/drm_ioctl.c b/drivers/gpu/drm/drm_ioctl.c index bb51ee97956d..12b7753a0d27 100644 --- a/drivers/gpu/drm/drm_ioctl.c +++ b/drivers/gpu/drm/drm_ioctl.c @@ -228,6 +228,7 @@ static int drm_getstats(struct drm_device *dev, void *data, static int drm_getcap(struct drm_device *dev, void *data, struct drm_file *file_priv) { struct drm_get_cap *req = data; + struct drm_crtc *crtc; req->value = 0; switch (req->capability) { @@ -254,6 +255,13 @@ static int drm_getcap(struct drm_device *dev, void *data, struct drm_file *file_ case DRM_CAP_ASYNC_PAGE_FLIP: req->value = dev->mode_config.async_page_flip; break; + case DRM_CAP_PAGE_FLIP_TARGET: + req->value = 1; + drm_for_each_crtc(crtc, dev) { + if (!crtc->funcs->page_flip_target) + req->value = 0; + } + break; case DRM_CAP_CURSOR_WIDTH: if (dev->mode_config.cursor_width) req->value = dev->mode_config.cursor_width; diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c index 6190035edfea..8ab9ce5089fe 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bo.c +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c @@ -1151,7 +1151,7 @@ nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict, bool intr, if (ret) goto out; - ret = ttm_bo_move_ttm(bo, true, intr, no_wait_gpu, new_mem); + ret = ttm_bo_move_ttm(bo, intr, no_wait_gpu, new_mem); out: ttm_bo_mem_put(bo, &tmp_mem); return ret; @@ -1179,7 +1179,7 @@ nouveau_bo_move_flips(struct ttm_buffer_object *bo, bool evict, bool intr, if (ret) return ret; - ret = ttm_bo_move_ttm(bo, true, intr, no_wait_gpu, &tmp_mem); + ret = ttm_bo_move_ttm(bo, intr, no_wait_gpu, &tmp_mem); if (ret) goto out; @@ -1297,7 +1297,7 @@ nouveau_bo_move(struct ttm_buffer_object *bo, bool evict, bool intr, /* Fallback to software copy. */ ret = ttm_bo_wait(bo, intr, no_wait_gpu); if (ret == 0) - ret = ttm_bo_move_memcpy(bo, evict, intr, no_wait_gpu, new_mem); + ret = ttm_bo_move_memcpy(bo, intr, no_wait_gpu, new_mem); out: if (drm->device.info.family < NV_DEVICE_INFO_V0_TESLA) { diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c index d50c9679e631..6a22de045cb5 100644 --- a/drivers/gpu/drm/qxl/qxl_ttm.c +++ b/drivers/gpu/drm/qxl/qxl_ttm.c @@ -361,8 +361,8 @@ static int qxl_bo_move(struct ttm_buffer_object *bo, qxl_move_null(bo, new_mem); return 0; } - return ttm_bo_move_memcpy(bo, evict, interruptible, - no_wait_gpu, new_mem); + return ttm_bo_move_memcpy(bo, interruptible, no_wait_gpu, + new_mem); } static void qxl_bo_move_notify(struct ttm_buffer_object *bo, diff --git a/drivers/gpu/drm/radeon/atombios_crtc.c b/drivers/gpu/drm/radeon/atombios_crtc.c index a89c4803aced..4824f70b0258 100644 --- a/drivers/gpu/drm/radeon/atombios_crtc.c +++ b/drivers/gpu/drm/radeon/atombios_crtc.c @@ -1435,8 +1435,8 @@ static int dce4_crtc_do_set_base(struct drm_crtc *crtc, WREG32(EVERGREEN_VIEWPORT_SIZE + radeon_crtc->crtc_offset, (viewport_w << 16) | viewport_h); - /* set pageflip to happen only at start of vblank interval (front porch) */ - WREG32(EVERGREEN_MASTER_UPDATE_MODE + radeon_crtc->crtc_offset, 3); + /* set pageflip to happen anywhere in vblank interval */ + WREG32(EVERGREEN_MASTER_UPDATE_MODE + radeon_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { radeon_fb = to_radeon_framebuffer(fb); @@ -1636,8 +1636,8 @@ static int avivo_crtc_do_set_base(struct drm_crtc *crtc, WREG32(AVIVO_D1MODE_VIEWPORT_SIZE + radeon_crtc->crtc_offset, (viewport_w << 16) | viewport_h); - /* set pageflip to happen only at start of vblank interval (front porch) */ - WREG32(AVIVO_D1MODE_MASTER_UPDATE_MODE + radeon_crtc->crtc_offset, 3); + /* set pageflip to happen anywhere in vblank interval */ + WREG32(AVIVO_D1MODE_MASTER_UPDATE_MODE + radeon_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { radeon_fb = to_radeon_framebuffer(fb); diff --git a/drivers/gpu/drm/radeon/atombios_dp.c b/drivers/gpu/drm/radeon/atombios_dp.c index cead089a9e7d..432cb46f6a34 100644 --- a/drivers/gpu/drm/radeon/atombios_dp.c +++ b/drivers/gpu/drm/radeon/atombios_dp.c @@ -389,22 +389,21 @@ bool radeon_dp_getdpcd(struct radeon_connector *radeon_connector) { struct radeon_connector_atom_dig *dig_connector = radeon_connector->con_priv; u8 msg[DP_DPCD_SIZE]; - int ret, i; + int ret; - for (i = 0; i < 7; i++) { - ret = drm_dp_dpcd_read(&radeon_connector->ddc_bus->aux, DP_DPCD_REV, msg, - DP_DPCD_SIZE); - if (ret == DP_DPCD_SIZE) { - memcpy(dig_connector->dpcd, msg, DP_DPCD_SIZE); + ret = drm_dp_dpcd_read(&radeon_connector->ddc_bus->aux, DP_DPCD_REV, msg, + DP_DPCD_SIZE); + if (ret == DP_DPCD_SIZE) { + memcpy(dig_connector->dpcd, msg, DP_DPCD_SIZE); - DRM_DEBUG_KMS("DPCD: %*ph\n", (int)sizeof(dig_connector->dpcd), - dig_connector->dpcd); + DRM_DEBUG_KMS("DPCD: %*ph\n", (int)sizeof(dig_connector->dpcd), + dig_connector->dpcd); - radeon_dp_probe_oui(radeon_connector); + radeon_dp_probe_oui(radeon_connector); - return true; - } + return true; } + dig_connector->dpcd[0] = 0; return false; } diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 0c1b9ff433af..b1784a1b482a 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -1871,7 +1871,7 @@ int ci_mc_load_microcode(struct radeon_device *rdev) { const __be32 *fw_data = NULL; const __le32 *new_fw_data = NULL; - u32 running, blackout = 0, tmp; + u32 running, tmp; u32 *io_mc_regs = NULL; const __le32 *new_io_mc_regs = NULL; int i, regs_size, ucode_size; @@ -1912,11 +1912,6 @@ int ci_mc_load_microcode(struct radeon_device *rdev) running = RREG32(MC_SEQ_SUP_CNTL) & RUN_MASK; if (running == 0) { - if (running) { - blackout = RREG32(MC_SHARED_BLACKOUT_CNTL); - WREG32(MC_SHARED_BLACKOUT_CNTL, blackout | 1); - } - /* reset the engine and set to writable */ WREG32(MC_SEQ_SUP_CNTL, 0x00000008); WREG32(MC_SEQ_SUP_CNTL, 0x00000010); @@ -1964,9 +1959,6 @@ int ci_mc_load_microcode(struct radeon_device *rdev) break; udelay(1); } - - if (running) - WREG32(MC_SHARED_BLACKOUT_CNTL, blackout); } return 0; @@ -8215,7 +8207,7 @@ static void cik_uvd_resume(struct radeon_device *rdev) return; ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; - r = radeon_ring_init(rdev, ring, ring->ring_size, 0, RADEON_CP_PACKET2); + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, PACKET0(UVD_NO_OP, 0)); if (r) { dev_err(rdev->dev, "failed initializing UVD ring (%d).\n", r); return; diff --git a/drivers/gpu/drm/radeon/cikd.h b/drivers/gpu/drm/radeon/cikd.h index cead2284fd79..48db93577c1d 100644 --- a/drivers/gpu/drm/radeon/cikd.h +++ b/drivers/gpu/drm/radeon/cikd.h @@ -2069,6 +2069,7 @@ #define UVD_UDEC_ADDR_CONFIG 0xef4c #define UVD_UDEC_DB_ADDR_CONFIG 0xef50 #define UVD_UDEC_DBW_ADDR_CONFIG 0xef54 +#define UVD_NO_OP 0xeffc #define UVD_LMI_EXT40_ADDR 0xf498 #define UVD_GP_SCRATCH4 0xf4e0 diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index db275b7ed34a..0b6b5766216f 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -2878,9 +2878,8 @@ void evergreen_mc_resume(struct radeon_device *rdev, struct evergreen_mc_save *s for (i = 0; i < rdev->num_crtc; i++) { if (save->crtc_enabled[i]) { tmp = RREG32(EVERGREEN_MASTER_UPDATE_MODE + crtc_offsets[i]); - if ((tmp & 0x7) != 3) { + if ((tmp & 0x7) != 0) { tmp &= ~0x7; - tmp |= 0x3; WREG32(EVERGREEN_MASTER_UPDATE_MODE + crtc_offsets[i], tmp); } tmp = RREG32(EVERGREEN_GRPH_UPDATE + crtc_offsets[i]); @@ -5580,7 +5579,7 @@ static void evergreen_uvd_resume(struct radeon_device *rdev) return; ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; - r = radeon_ring_init(rdev, ring, ring->ring_size, 0, RADEON_CP_PACKET2); + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, PACKET0(UVD_NO_OP, 0)); if (r) { dev_err(rdev->dev, "failed initializing UVD ring (%d).\n", r); return; diff --git a/drivers/gpu/drm/radeon/evergreend.h b/drivers/gpu/drm/radeon/evergreend.h index c8e3d394cde7..f3d88ca2aa8f 100644 --- a/drivers/gpu/drm/radeon/evergreend.h +++ b/drivers/gpu/drm/radeon/evergreend.h @@ -1523,6 +1523,7 @@ #define UVD_UDEC_ADDR_CONFIG 0xef4c #define UVD_UDEC_DB_ADDR_CONFIG 0xef50 #define UVD_UDEC_DBW_ADDR_CONFIG 0xef54 +#define UVD_NO_OP 0xeffc #define UVD_RBC_RB_RPTR 0xf690 #define UVD_RBC_RB_WPTR 0xf694 #define UVD_STATUS 0xf6bc diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index 4a3d7cab83f7..103fc8650197 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -2062,7 +2062,7 @@ static void cayman_uvd_resume(struct radeon_device *rdev) return; ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; - r = radeon_ring_init(rdev, ring, ring->ring_size, 0, RADEON_CP_PACKET2); + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, PACKET0(UVD_NO_OP, 0)); if (r) { dev_err(rdev->dev, "failed initializing UVD ring (%d).\n", r); return; diff --git a/drivers/gpu/drm/radeon/nid.h b/drivers/gpu/drm/radeon/nid.h index 47eb49b77d32..3c9fec88ea44 100644 --- a/drivers/gpu/drm/radeon/nid.h +++ b/drivers/gpu/drm/radeon/nid.h @@ -1137,6 +1137,7 @@ #define UVD_UDEC_ADDR_CONFIG 0xEF4C #define UVD_UDEC_DB_ADDR_CONFIG 0xEF50 #define UVD_UDEC_DBW_ADDR_CONFIG 0xEF54 +#define UVD_NO_OP 0xEFFC #define UVD_RBC_RB_RPTR 0xF690 #define UVD_RBC_RB_WPTR 0xF694 #define UVD_STATUS 0xf6bc diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 9247e7d207fe..640653606f86 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -3097,7 +3097,7 @@ static void r600_uvd_resume(struct radeon_device *rdev) return; ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; - r = radeon_ring_init(rdev, ring, ring->ring_size, 0, RADEON_CP_PACKET2); + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, PACKET0(UVD_NO_OP, 0)); if (r) { dev_err(rdev->dev, "failed initializing UVD ring (%d).\n", r); return; diff --git a/drivers/gpu/drm/radeon/r600d.h b/drivers/gpu/drm/radeon/r600d.h index 1e8495cca41e..2e00a5287bd2 100644 --- a/drivers/gpu/drm/radeon/r600d.h +++ b/drivers/gpu/drm/radeon/r600d.h @@ -1490,6 +1490,7 @@ #define UVD_GPCOM_VCPU_DATA0 0xef10 #define UVD_GPCOM_VCPU_DATA1 0xef14 #define UVD_ENGINE_CNTL 0xef18 +#define UVD_NO_OP 0xeffc #define UVD_SEMA_CNTL 0xf400 #define UVD_RB_ARB_CTRL 0xf480 diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 5633ee3eb46e..1b0dcad916b0 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -742,6 +742,7 @@ struct radeon_flip_work { struct work_struct unpin_work; struct radeon_device *rdev; int crtc_id; + u32 target_vblank; uint64_t base; struct drm_pending_vblank_event *event; struct radeon_bo *old_rbo; diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c index c3206fb8f4cf..890171f08987 100644 --- a/drivers/gpu/drm/radeon/radeon_display.c +++ b/drivers/gpu/drm/radeon/radeon_display.c @@ -400,14 +400,13 @@ static void radeon_flip_work_func(struct work_struct *__work) struct radeon_flip_work *work = container_of(__work, struct radeon_flip_work, flip_work); struct radeon_device *rdev = work->rdev; + struct drm_device *dev = rdev->ddev; struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[work->crtc_id]; struct drm_crtc *crtc = &radeon_crtc->base; unsigned long flags; int r; - int vpos, hpos, stat, min_udelay = 0; - unsigned repcnt = 4; - struct drm_vblank_crtc *vblank = &crtc->dev->vblank[work->crtc_id]; + int vpos, hpos; down_read(&rdev->exclusive_lock); if (work->fence) { @@ -438,59 +437,25 @@ static void radeon_flip_work_func(struct work_struct *__work) work->fence = NULL; } + /* Wait until we're out of the vertical blank period before the one + * targeted by the flip + */ + while (radeon_crtc->enabled && + (radeon_get_crtc_scanoutpos(dev, work->crtc_id, 0, + &vpos, &hpos, NULL, NULL, + &crtc->hwmode) + & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK)) == + (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK) && + (int)(work->target_vblank - + dev->driver->get_vblank_counter(dev, work->crtc_id)) > 0) + usleep_range(1000, 2000); + /* We borrow the event spin lock for protecting flip_status */ spin_lock_irqsave(&crtc->dev->event_lock, flags); /* set the proper interrupt */ radeon_irq_kms_pflip_irq_get(rdev, radeon_crtc->crtc_id); - /* If this happens to execute within the "virtually extended" vblank - * interval before the start of the real vblank interval then it needs - * to delay programming the mmio flip until the real vblank is entered. - * This prevents completing a flip too early due to the way we fudge - * our vblank counter and vblank timestamps in order to work around the - * problem that the hw fires vblank interrupts before actual start of - * vblank (when line buffer refilling is done for a frame). It - * complements the fudging logic in radeon_get_crtc_scanoutpos() for - * timestamping and radeon_get_vblank_counter_kms() for vblank counts. - * - * In practice this won't execute very often unless on very fast - * machines because the time window for this to happen is very small. - */ - while (radeon_crtc->enabled && --repcnt) { - /* GET_DISTANCE_TO_VBLANKSTART returns distance to real vblank - * start in hpos, and to the "fudged earlier" vblank start in - * vpos. - */ - stat = radeon_get_crtc_scanoutpos(rdev->ddev, work->crtc_id, - GET_DISTANCE_TO_VBLANKSTART, - &vpos, &hpos, NULL, NULL, - &crtc->hwmode); - - if ((stat & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_ACCURATE)) != - (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_ACCURATE) || - !(vpos >= 0 && hpos <= 0)) - break; - - /* Sleep at least until estimated real start of hw vblank */ - min_udelay = (-hpos + 1) * max(vblank->linedur_ns / 1000, 5); - if (min_udelay > vblank->framedur_ns / 2000) { - /* Don't wait ridiculously long - something is wrong */ - repcnt = 0; - break; - } - spin_unlock_irqrestore(&crtc->dev->event_lock, flags); - usleep_range(min_udelay, 2 * min_udelay); - spin_lock_irqsave(&crtc->dev->event_lock, flags); - }; - - if (!repcnt) - DRM_DEBUG_DRIVER("Delay problem on crtc %d: min_udelay %d, " - "framedur %d, linedur %d, stat %d, vpos %d, " - "hpos %d\n", work->crtc_id, min_udelay, - vblank->framedur_ns / 1000, - vblank->linedur_ns / 1000, stat, vpos, hpos); - /* do the flip (mmio) */ radeon_page_flip(rdev, radeon_crtc->crtc_id, work->base, work->async); @@ -499,10 +464,11 @@ static void radeon_flip_work_func(struct work_struct *__work) up_read(&rdev->exclusive_lock); } -static int radeon_crtc_page_flip(struct drm_crtc *crtc, - struct drm_framebuffer *fb, - struct drm_pending_vblank_event *event, - uint32_t page_flip_flags) +static int radeon_crtc_page_flip_target(struct drm_crtc *crtc, + struct drm_framebuffer *fb, + struct drm_pending_vblank_event *event, + uint32_t page_flip_flags, + uint32_t target) { struct drm_device *dev = crtc->dev; struct radeon_device *rdev = dev->dev_private; @@ -599,12 +565,8 @@ static int radeon_crtc_page_flip(struct drm_crtc *crtc, base &= ~7; } work->base = base; - - r = drm_crtc_vblank_get(crtc); - if (r) { - DRM_ERROR("failed to get vblank before flip\n"); - goto pflip_cleanup; - } + work->target_vblank = target - drm_crtc_vblank_count(crtc) + + dev->driver->get_vblank_counter(dev, work->crtc_id); /* We borrow the event spin lock for protecting flip_work */ spin_lock_irqsave(&crtc->dev->event_lock, flags); @@ -613,7 +575,7 @@ static int radeon_crtc_page_flip(struct drm_crtc *crtc, DRM_DEBUG_DRIVER("flip queue: crtc already busy\n"); spin_unlock_irqrestore(&crtc->dev->event_lock, flags); r = -EBUSY; - goto vblank_cleanup; + goto pflip_cleanup; } radeon_crtc->flip_status = RADEON_FLIP_PENDING; radeon_crtc->flip_work = work; @@ -626,9 +588,6 @@ static int radeon_crtc_page_flip(struct drm_crtc *crtc, queue_work(radeon_crtc->flip_queue, &work->flip_work); return 0; -vblank_cleanup: - drm_crtc_vblank_put(crtc); - pflip_cleanup: if (unlikely(radeon_bo_reserve(new_rbo, false) != 0)) { DRM_ERROR("failed to reserve new rbo in error path\n"); @@ -697,7 +656,7 @@ static const struct drm_crtc_funcs radeon_crtc_funcs = { .gamma_set = radeon_crtc_gamma_set, .set_config = radeon_crtc_set_config, .destroy = radeon_crtc_destroy, - .page_flip = radeon_crtc_page_flip, + .page_flip_target = radeon_crtc_page_flip_target, }; static void radeon_crtc_init(struct drm_device *dev, int index) diff --git a/drivers/gpu/drm/radeon/radeon_dp_auxch.c b/drivers/gpu/drm/radeon/radeon_dp_auxch.c index db64e0062689..2d465648856a 100644 --- a/drivers/gpu/drm/radeon/radeon_dp_auxch.c +++ b/drivers/gpu/drm/radeon/radeon_dp_auxch.c @@ -164,7 +164,6 @@ radeon_dp_aux_transfer_native(struct drm_dp_aux *aux, struct drm_dp_aux_msg *msg } if (tmp & AUX_SW_RX_TIMEOUT) { - DRM_DEBUG_KMS("dp_aux_ch timed out\n"); ret = -ETIMEDOUT; goto done; } diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c index 90f2ff217b31..07e44931f1f1 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.c +++ b/drivers/gpu/drm/radeon/radeon_drv.c @@ -95,9 +95,10 @@ * 2.44.0 - SET_APPEND_CNT packet3 support * 2.45.0 - Allow setting shader registers using DMA/COPY packet3 on SI * 2.46.0 - Add PFP_SYNC_ME support on evergreen + * 2.47.0 - Add UVD_NO_OP register support */ #define KMS_DRIVER_MAJOR 2 -#define KMS_DRIVER_MINOR 46 +#define KMS_DRIVER_MINOR 47 #define KMS_DRIVER_PATCHLEVEL 0 int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags); int radeon_driver_unload_kms(struct drm_device *dev); diff --git a/drivers/gpu/drm/radeon/radeon_i2c.c b/drivers/gpu/drm/radeon/radeon_i2c.c index 9590bcd321c0..021aa005623f 100644 --- a/drivers/gpu/drm/radeon/radeon_i2c.c +++ b/drivers/gpu/drm/radeon/radeon_i2c.c @@ -938,10 +938,8 @@ struct radeon_i2c_chan *radeon_i2c_create(struct drm_device *dev, "Radeon i2c hw bus %s", name); i2c->adapter.algo = &radeon_i2c_algo; ret = i2c_add_adapter(&i2c->adapter); - if (ret) { - DRM_ERROR("Failed to register hw i2c %s\n", name); + if (ret) goto out_free; - } } else if (rec->hw_capable && radeon_hw_i2c && ASIC_IS_DCE3(rdev)) { @@ -950,10 +948,8 @@ struct radeon_i2c_chan *radeon_i2c_create(struct drm_device *dev, "Radeon i2c hw bus %s", name); i2c->adapter.algo = &radeon_atom_i2c_algo; ret = i2c_add_adapter(&i2c->adapter); - if (ret) { - DRM_ERROR("Failed to register hw i2c %s\n", name); + if (ret) goto out_free; - } } else { /* set the radeon bit adapter */ snprintf(i2c->adapter.name, sizeof(i2c->adapter.name), diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c index 0c00e192c845..93414aca60d6 100644 --- a/drivers/gpu/drm/radeon/radeon_ttm.c +++ b/drivers/gpu/drm/radeon/radeon_ttm.c @@ -346,7 +346,7 @@ static int radeon_move_vram_ram(struct ttm_buffer_object *bo, if (unlikely(r)) { goto out_cleanup; } - r = ttm_bo_move_ttm(bo, true, interruptible, no_wait_gpu, new_mem); + r = ttm_bo_move_ttm(bo, interruptible, no_wait_gpu, new_mem); out_cleanup: ttm_bo_mem_put(bo, &tmp_mem); return r; @@ -379,7 +379,7 @@ static int radeon_move_ram_vram(struct ttm_buffer_object *bo, if (unlikely(r)) { return r; } - r = ttm_bo_move_ttm(bo, true, interruptible, no_wait_gpu, &tmp_mem); + r = ttm_bo_move_ttm(bo, interruptible, no_wait_gpu, &tmp_mem); if (unlikely(r)) { goto out_cleanup; } @@ -444,8 +444,7 @@ static int radeon_bo_move(struct ttm_buffer_object *bo, if (r) { memcpy: - r = ttm_bo_move_memcpy(bo, evict, interruptible, - no_wait_gpu, new_mem); + r = ttm_bo_move_memcpy(bo, interruptible, no_wait_gpu, new_mem); if (r) { return r; } diff --git a/drivers/gpu/drm/radeon/radeon_uvd.c b/drivers/gpu/drm/radeon/radeon_uvd.c index 73dfe01435ea..0cd0e7bdee55 100644 --- a/drivers/gpu/drm/radeon/radeon_uvd.c +++ b/drivers/gpu/drm/radeon/radeon_uvd.c @@ -669,6 +669,7 @@ static int radeon_uvd_cs_reg(struct radeon_cs_parser *p, return r; break; case UVD_ENGINE_CNTL: + case UVD_NO_OP: break; default: DRM_ERROR("Invalid reg 0x%X!\n", @@ -753,8 +754,10 @@ static int radeon_uvd_send_msg(struct radeon_device *rdev, ib.ptr[3] = addr >> 32; ib.ptr[4] = PACKET0(UVD_GPCOM_VCPU_CMD, 0); ib.ptr[5] = 0; - for (i = 6; i < 16; ++i) - ib.ptr[i] = PACKET2(0); + for (i = 6; i < 16; i += 2) { + ib.ptr[i] = PACKET0(UVD_NO_OP, 0); + ib.ptr[i+1] = 0; + } ib.length_dw = 16; r = radeon_ib_schedule(rdev, &ib, NULL, false); diff --git a/drivers/gpu/drm/radeon/rv515.c b/drivers/gpu/drm/radeon/rv515.c index c55d653aaf5f..76c55c5d11ec 100644 --- a/drivers/gpu/drm/radeon/rv515.c +++ b/drivers/gpu/drm/radeon/rv515.c @@ -406,9 +406,8 @@ void rv515_mc_resume(struct radeon_device *rdev, struct rv515_mc_save *save) for (i = 0; i < rdev->num_crtc; i++) { if (save->crtc_enabled[i]) { tmp = RREG32(AVIVO_D1MODE_MASTER_UPDATE_MODE + crtc_offsets[i]); - if ((tmp & 0x7) != 3) { + if ((tmp & 0x7) != 0) { tmp &= ~0x7; - tmp |= 0x3; WREG32(AVIVO_D1MODE_MASTER_UPDATE_MODE + crtc_offsets[i], tmp); } tmp = RREG32(AVIVO_D1GRPH_UPDATE + crtc_offsets[i]); diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index 1c120a4c3c97..729ae588c970 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -1738,7 +1738,7 @@ static void rv770_uvd_resume(struct radeon_device *rdev) return; ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; - r = radeon_ring_init(rdev, ring, ring->ring_size, 0, RADEON_CP_PACKET2); + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, PACKET0(UVD_NO_OP, 0)); if (r) { dev_err(rdev->dev, "failed initializing UVD ring (%d).\n", r); return; diff --git a/drivers/gpu/drm/radeon/rv770d.h b/drivers/gpu/drm/radeon/rv770d.h index 9ef2064b1c9c..0271f4c559ae 100644 --- a/drivers/gpu/drm/radeon/rv770d.h +++ b/drivers/gpu/drm/radeon/rv770d.h @@ -387,6 +387,7 @@ #define UVD_UDEC_TILING_CONFIG 0xef40 #define UVD_UDEC_DB_TILING_CONFIG 0xef44 #define UVD_UDEC_DBW_TILING_CONFIG 0xef48 +#define UVD_NO_OP 0xeffc #define GC_USER_SHADER_PIPE_CONFIG 0x8954 #define INACTIVE_QD_PIPES(x) ((x) << 8) diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 2523ca96c6c7..7ee9aafbdf74 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -1547,7 +1547,7 @@ int si_mc_load_microcode(struct radeon_device *rdev) { const __be32 *fw_data = NULL; const __le32 *new_fw_data = NULL; - u32 running, blackout = 0; + u32 running; u32 *io_mc_regs = NULL; const __le32 *new_io_mc_regs = NULL; int i, regs_size, ucode_size; @@ -1598,11 +1598,6 @@ int si_mc_load_microcode(struct radeon_device *rdev) running = RREG32(MC_SEQ_SUP_CNTL) & RUN_MASK; if (running == 0) { - if (running) { - blackout = RREG32(MC_SHARED_BLACKOUT_CNTL); - WREG32(MC_SHARED_BLACKOUT_CNTL, blackout | 1); - } - /* reset the engine and set to writable */ WREG32(MC_SEQ_SUP_CNTL, 0x00000008); WREG32(MC_SEQ_SUP_CNTL, 0x00000010); @@ -1641,9 +1636,6 @@ int si_mc_load_microcode(struct radeon_device *rdev) break; udelay(1); } - - if (running) - WREG32(MC_SHARED_BLACKOUT_CNTL, blackout); } return 0; @@ -6928,7 +6920,7 @@ static void si_uvd_resume(struct radeon_device *rdev) return; ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; - r = radeon_ring_init(rdev, ring, ring->ring_size, 0, RADEON_CP_PACKET2); + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, PACKET0(UVD_NO_OP, 0)); if (r) { dev_err(rdev->dev, "failed initializing UVD ring (%d).\n", r); return; diff --git a/drivers/gpu/drm/radeon/sid.h b/drivers/gpu/drm/radeon/sid.h index d1a7b58dd291..eb220eecba78 100644 --- a/drivers/gpu/drm/radeon/sid.h +++ b/drivers/gpu/drm/radeon/sid.h @@ -1559,6 +1559,7 @@ #define UVD_UDEC_ADDR_CONFIG 0xEF4C #define UVD_UDEC_DB_ADDR_CONFIG 0xEF50 #define UVD_UDEC_DBW_ADDR_CONFIG 0xEF54 +#define UVD_NO_OP 0xEFFC #define UVD_RBC_RB_RPTR 0xF690 #define UVD_RBC_RB_WPTR 0xF694 #define UVD_STATUS 0xf6bc diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index 42c074a9c955..c2a30bdc8a01 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -354,14 +354,12 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo, if (!(old_man->flags & TTM_MEMTYPE_FLAG_FIXED) && !(new_man->flags & TTM_MEMTYPE_FLAG_FIXED)) - ret = ttm_bo_move_ttm(bo, evict, interruptible, no_wait_gpu, - mem); + ret = ttm_bo_move_ttm(bo, interruptible, no_wait_gpu, mem); else if (bdev->driver->move) ret = bdev->driver->move(bo, evict, interruptible, no_wait_gpu, mem); else - ret = ttm_bo_move_memcpy(bo, evict, interruptible, - no_wait_gpu, mem); + ret = ttm_bo_move_memcpy(bo, interruptible, no_wait_gpu, mem); if (ret) { if (bdev->driver->move_notify) { diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c index f157a9efd220..bf6e21655c57 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_util.c +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c @@ -45,8 +45,8 @@ void ttm_bo_free_old_node(struct ttm_buffer_object *bo) } int ttm_bo_move_ttm(struct ttm_buffer_object *bo, - bool evict, bool interruptible, - bool no_wait_gpu, struct ttm_mem_reg *new_mem) + bool interruptible, bool no_wait_gpu, + struct ttm_mem_reg *new_mem) { struct ttm_tt *ttm = bo->ttm; struct ttm_mem_reg *old_mem = &bo->mem; @@ -329,8 +329,7 @@ static int ttm_copy_ttm_io_page(struct ttm_tt *ttm, void *dst, } int ttm_bo_move_memcpy(struct ttm_buffer_object *bo, - bool evict, bool interruptible, - bool no_wait_gpu, + bool interruptible, bool no_wait_gpu, struct ttm_mem_reg *new_mem) { struct ttm_bo_device *bdev = bo->bdev; diff --git a/drivers/gpu/drm/ttm/ttm_memory.c b/drivers/gpu/drm/ttm/ttm_memory.c index a1803fbcc898..29855be96be0 100644 --- a/drivers/gpu/drm/ttm/ttm_memory.c +++ b/drivers/gpu/drm/ttm/ttm_memory.c @@ -600,3 +600,9 @@ size_t ttm_round_pot(size_t size) return 0; } EXPORT_SYMBOL(ttm_round_pot); + +uint64_t ttm_get_kernel_zone_memory_size(struct ttm_mem_global *glob) +{ + return glob->zone_kernel->max_mem; +} +EXPORT_SYMBOL(ttm_get_kernel_zone_memory_size); |