diff options
author | Dave Airlie <airlied@redhat.com> | 2014-02-27 14:39:30 +1000 |
---|---|---|
committer | Dave Airlie <airlied@redhat.com> | 2014-02-27 14:39:30 +1000 |
commit | 4d538b79197901fecc42e746d515d07fd1089b62 (patch) | |
tree | 0551acabf260fcd7c0189a4e06bb608e31d1206e /drivers/gpu | |
parent | c48cdd23ea5006c377b670bc3056fa7b63f92574 (diff) | |
parent | 82dc62a31ce3ed7b4eeea9c65a3b69e81e2ea688 (diff) | |
download | linux-4d538b79197901fecc42e746d515d07fd1089b62.tar.bz2 |
Merge branch 'drm-next-3.15' of git://people.freedesktop.org/~deathsimple/linux into drm-next
So this is the initial pull request for radeon drm-next 3.15. Highlights:
- VCE bringup including DPM support
- Few cleanups for the ring handling code
* 'drm-next-3.15' of git://people.freedesktop.org/~deathsimple/linux:
drm/radeon: cleanup false positive lockup handling
drm/radeon: drop radeon_ring_force_activity
drm/radeon: drop drivers copy of the rptr
drm/radeon/cik: enable/disable vce cg when encoding v2
drm/radeon: add support for vce 2.0 clock gating
drm/radeon/dpm: properly enable/disable vce when vce pg is enabled
drm/radeon/dpm: enable dynamic vce state switching v2
drm/radeon: add vce dpm support for KV/KB
drm/radeon: enable vce dpm on CI
drm/radeon: add vce dpm support for CI
drm/radeon: fill in set_vce_clocks for CIK asics
drm/radeon/dpm: fetch vce states from the vbios
drm/radeon/dpm: fill in some initial vce infrastructure
drm/radeon/dpm: move platform caps fetching to a separate function
drm/radeon: add callback for setting vce clocks
drm/radeon: add VCE version parsing and checking
drm/radeon: add VCE ring query
drm/radeon: initial VCE support v4
drm/radeon: fix CP semaphores on CIK
Diffstat (limited to 'drivers/gpu')
39 files changed, 1686 insertions, 175 deletions
diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile index 306364a1ecda..ed60caa32518 100644 --- a/drivers/gpu/drm/radeon/Makefile +++ b/drivers/gpu/drm/radeon/Makefile @@ -99,6 +99,12 @@ radeon-y += \ uvd_v3_1.o \ uvd_v4_2.o +# add VCE block +radeon-y += \ + radeon_vce.o \ + vce_v1_0.o \ + vce_v2_0.o \ + radeon-$(CONFIG_COMPAT) += radeon_ioc32.o radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o radeon-$(CONFIG_ACPI) += radeon_acpi.o diff --git a/drivers/gpu/drm/radeon/btc_dpm.c b/drivers/gpu/drm/radeon/btc_dpm.c index ea103ccdf4bd..f81d7ca134db 100644 --- a/drivers/gpu/drm/radeon/btc_dpm.c +++ b/drivers/gpu/drm/radeon/btc_dpm.c @@ -2601,6 +2601,10 @@ int btc_dpm_init(struct radeon_device *rdev) pi->min_vddc_in_table = 0; pi->max_vddc_in_table = 0; + ret = r600_get_platform_caps(rdev); + if (ret) + return ret; + ret = rv7xx_parse_power_table(rdev); if (ret) return ret; diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index 8d49104ca6c2..cad89a977527 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -172,6 +172,8 @@ extern void si_trim_voltage_table_to_fit_state_table(struct radeon_device *rdev, extern void cik_enter_rlc_safe_mode(struct radeon_device *rdev); extern void cik_exit_rlc_safe_mode(struct radeon_device *rdev); extern int ci_mc_load_microcode(struct radeon_device *rdev); +extern void cik_update_cg(struct radeon_device *rdev, + u32 block, bool enable); static int ci_get_std_voltage_value_sidd(struct radeon_device *rdev, struct atom_voltage_table_entry *voltage_table, @@ -746,6 +748,14 @@ static void ci_apply_state_adjust_rules(struct radeon_device *rdev, u32 max_sclk_vddc, max_mclk_vddci, max_mclk_vddc; int i; + if (rps->vce_active) { + rps->evclk = rdev->pm.dpm.vce_states[rdev->pm.dpm.vce_level].evclk; + rps->ecclk = rdev->pm.dpm.vce_states[rdev->pm.dpm.vce_level].ecclk; + } else { + rps->evclk = 0; + rps->ecclk = 0; + } + if ((rdev->pm.dpm.new_active_crtc_count > 1) || ci_dpm_vblank_too_short(rdev)) disable_mclk_switching = true; @@ -804,6 +814,13 @@ static void ci_apply_state_adjust_rules(struct radeon_device *rdev, sclk = ps->performance_levels[0].sclk; } + if (rps->vce_active) { + if (sclk < rdev->pm.dpm.vce_states[rdev->pm.dpm.vce_level].sclk) + sclk = rdev->pm.dpm.vce_states[rdev->pm.dpm.vce_level].sclk; + if (mclk < rdev->pm.dpm.vce_states[rdev->pm.dpm.vce_level].mclk) + mclk = rdev->pm.dpm.vce_states[rdev->pm.dpm.vce_level].mclk; + } + ps->performance_levels[0].sclk = sclk; ps->performance_levels[0].mclk = mclk; @@ -3468,7 +3485,6 @@ static int ci_enable_uvd_dpm(struct radeon_device *rdev, bool enable) 0 : -EINVAL; } -#if 0 static int ci_enable_vce_dpm(struct radeon_device *rdev, bool enable) { struct ci_power_info *pi = ci_get_pi(rdev); @@ -3501,6 +3517,7 @@ static int ci_enable_vce_dpm(struct radeon_device *rdev, bool enable) 0 : -EINVAL; } +#if 0 static int ci_enable_samu_dpm(struct radeon_device *rdev, bool enable) { struct ci_power_info *pi = ci_get_pi(rdev); @@ -3587,7 +3604,6 @@ static int ci_update_uvd_dpm(struct radeon_device *rdev, bool gate) return ci_enable_uvd_dpm(rdev, !gate); } -#if 0 static u8 ci_get_vce_boot_level(struct radeon_device *rdev) { u8 i; @@ -3608,15 +3624,15 @@ static int ci_update_vce_dpm(struct radeon_device *rdev, struct radeon_ps *radeon_current_state) { struct ci_power_info *pi = ci_get_pi(rdev); - bool new_vce_clock_non_zero = (radeon_new_state->evclk != 0); - bool old_vce_clock_non_zero = (radeon_current_state->evclk != 0); int ret = 0; u32 tmp; - if (new_vce_clock_non_zero != old_vce_clock_non_zero) { - if (new_vce_clock_non_zero) { - pi->smc_state_table.VceBootLevel = ci_get_vce_boot_level(rdev); + if (radeon_current_state->evclk != radeon_new_state->evclk) { + if (radeon_new_state->evclk) { + /* turn the clocks on when encoding */ + cik_update_cg(rdev, RADEON_CG_BLOCK_VCE, false); + pi->smc_state_table.VceBootLevel = ci_get_vce_boot_level(rdev); tmp = RREG32_SMC(DPM_TABLE_475); tmp &= ~VceBootLevel_MASK; tmp |= VceBootLevel(pi->smc_state_table.VceBootLevel); @@ -3624,12 +3640,16 @@ static int ci_update_vce_dpm(struct radeon_device *rdev, ret = ci_enable_vce_dpm(rdev, true); } else { + /* turn the clocks off when not encoding */ + cik_update_cg(rdev, RADEON_CG_BLOCK_VCE, true); + ret = ci_enable_vce_dpm(rdev, false); } } return ret; } +#if 0 static int ci_update_samu_dpm(struct radeon_device *rdev, bool gate) { return ci_enable_samu_dpm(rdev, gate); @@ -4752,13 +4772,13 @@ int ci_dpm_set_power_state(struct radeon_device *rdev) DRM_ERROR("ci_generate_dpm_level_enable_mask failed\n"); return ret; } -#if 0 + ret = ci_update_vce_dpm(rdev, new_ps, old_ps); if (ret) { DRM_ERROR("ci_update_vce_dpm failed\n"); return ret; } -#endif + ret = ci_update_sclk_t(rdev); if (ret) { DRM_ERROR("ci_update_sclk_t failed\n"); @@ -4959,9 +4979,6 @@ static int ci_parse_power_table(struct radeon_device *rdev) if (!rdev->pm.dpm.ps) return -ENOMEM; power_state_offset = (u8 *)state_array->states; - rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); - rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); - rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); for (i = 0; i < state_array->ucNumEntries; i++) { u8 *idx; power_state = (union pplib_power_state *)power_state_offset; @@ -4998,6 +5015,21 @@ static int ci_parse_power_table(struct radeon_device *rdev) power_state_offset += 2 + power_state->v2.ucNumDPMLevels; } rdev->pm.dpm.num_ps = state_array->ucNumEntries; + + /* fill in the vce power states */ + for (i = 0; i < RADEON_MAX_VCE_LEVELS; i++) { + u32 sclk, mclk; + clock_array_index = rdev->pm.dpm.vce_states[i].clk_idx; + clock_info = (union pplib_clock_info *) + &clock_info_array->clockInfo[clock_array_index * clock_info_array->ucEntrySize]; + sclk = le16_to_cpu(clock_info->ci.usEngineClockLow); + sclk |= clock_info->ci.ucEngineClockHigh << 16; + mclk = le16_to_cpu(clock_info->ci.usMemoryClockLow); + mclk |= clock_info->ci.ucMemoryClockHigh << 16; + rdev->pm.dpm.vce_states[i].sclk = sclk; + rdev->pm.dpm.vce_states[i].mclk = mclk; + } + return 0; } @@ -5077,17 +5109,25 @@ int ci_dpm_init(struct radeon_device *rdev) ci_dpm_fini(rdev); return ret; } - ret = ci_parse_power_table(rdev); + + ret = r600_get_platform_caps(rdev); if (ret) { ci_dpm_fini(rdev); return ret; } + ret = r600_parse_extended_power_table(rdev); if (ret) { ci_dpm_fini(rdev); return ret; } + ret = ci_parse_power_table(rdev); + if (ret) { + ci_dpm_fini(rdev); + return ret; + } + pi->dll_default_on = false; pi->sram_end = SMC_RAM_END; @@ -5120,6 +5160,7 @@ int ci_dpm_init(struct radeon_device *rdev) pi->caps_sclk_throttle_low_notification = false; pi->caps_uvd_dpm = true; + pi->caps_vce_dpm = true; ci_get_leakage_voltages(rdev); ci_patch_dependency_tables_with_leakage(rdev); diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index e6419ca7cd37..92e38b54efb9 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -75,6 +75,7 @@ extern void si_init_uvd_internal_cg(struct radeon_device *rdev); extern int cik_sdma_resume(struct radeon_device *rdev); extern void cik_sdma_enable(struct radeon_device *rdev, bool enable); extern void cik_sdma_fini(struct radeon_device *rdev); +extern void vce_v2_0_enable_mgcg(struct radeon_device *rdev, bool enable); static void cik_rlc_stop(struct radeon_device *rdev); static void cik_pcie_gen3_enable(struct radeon_device *rdev); static void cik_program_aspm(struct radeon_device *rdev); @@ -4030,8 +4031,6 @@ static int cik_cp_gfx_resume(struct radeon_device *rdev) WREG32(CP_RB0_BASE, rb_addr); WREG32(CP_RB0_BASE_HI, upper_32_bits(rb_addr)); - ring->rptr = RREG32(CP_RB0_RPTR); - /* start the ring */ cik_cp_gfx_start(rdev); rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = true; @@ -4586,8 +4585,7 @@ static int cik_cp_compute_resume(struct radeon_device *rdev) rdev->ring[idx].wptr = 0; mqd->queue_state.cp_hqd_pq_wptr = rdev->ring[idx].wptr; WREG32(CP_HQD_PQ_WPTR, mqd->queue_state.cp_hqd_pq_wptr); - rdev->ring[idx].rptr = RREG32(CP_HQD_PQ_RPTR); - mqd->queue_state.cp_hqd_pq_rptr = rdev->ring[idx].rptr; + mqd->queue_state.cp_hqd_pq_rptr = RREG32(CP_HQD_PQ_RPTR); /* set the vmid for the queue */ mqd->queue_state.cp_hqd_vmid = 0; @@ -5117,11 +5115,9 @@ bool cik_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) if (!(reset_mask & (RADEON_RESET_GFX | RADEON_RESET_COMPUTE | RADEON_RESET_CP))) { - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - /* force CP activities */ - radeon_ring_force_activity(rdev, ring); return radeon_ring_test_lockup(rdev, ring); } @@ -6141,6 +6137,10 @@ void cik_update_cg(struct radeon_device *rdev, cik_enable_hdp_mgcg(rdev, enable); cik_enable_hdp_ls(rdev, enable); } + + if (block & RADEON_CG_BLOCK_VCE) { + vce_v2_0_enable_mgcg(rdev, enable); + } } static void cik_init_cg(struct radeon_device *rdev) @@ -7490,6 +7490,20 @@ restart_ih: /* reset addr and status */ WREG32_P(VM_CONTEXT1_CNTL2, 1, ~1); break; + case 167: /* VCE */ + DRM_DEBUG("IH: VCE int: 0x%08x\n", src_data); + switch (src_data) { + case 0: + radeon_fence_process(rdev, TN_RING_TYPE_VCE1_INDEX); + break; + case 1: + radeon_fence_process(rdev, TN_RING_TYPE_VCE2_INDEX); + break; + default: + DRM_ERROR("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; case 176: /* GFX RB CP_INT */ case 177: /* GFX IB CP_INT */ radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX); @@ -7789,6 +7803,22 @@ static int cik_startup(struct radeon_device *rdev) if (r) rdev->ring[R600_RING_TYPE_UVD_INDEX].ring_size = 0; + r = radeon_vce_resume(rdev); + if (!r) { + r = vce_v2_0_resume(rdev); + if (!r) + r = radeon_fence_driver_start_ring(rdev, + TN_RING_TYPE_VCE1_INDEX); + if (!r) + r = radeon_fence_driver_start_ring(rdev, + TN_RING_TYPE_VCE2_INDEX); + } + if (r) { + dev_err(rdev->dev, "VCE init error (%d).\n", r); + rdev->ring[TN_RING_TYPE_VCE1_INDEX].ring_size = 0; + rdev->ring[TN_RING_TYPE_VCE2_INDEX].ring_size = 0; + } + /* Enable IRQ */ if (!rdev->irq.installed) { r = radeon_irq_kms_init(rdev); @@ -7864,6 +7894,23 @@ static int cik_startup(struct radeon_device *rdev) DRM_ERROR("radeon: failed initializing UVD (%d).\n", r); } + r = -ENOENT; + + ring = &rdev->ring[TN_RING_TYPE_VCE1_INDEX]; + if (ring->ring_size) + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, + VCE_CMD_NO_OP); + + ring = &rdev->ring[TN_RING_TYPE_VCE2_INDEX]; + if (ring->ring_size) + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, + VCE_CMD_NO_OP); + + if (!r) + r = vce_v1_0_init(rdev); + else if (r != -ENOENT) + DRM_ERROR("radeon: failed initializing VCE (%d).\n", r); + r = radeon_ib_pool_init(rdev); if (r) { dev_err(rdev->dev, "IB initialization failed (%d).\n", r); @@ -7934,6 +7981,7 @@ int cik_suspend(struct radeon_device *rdev) cik_sdma_enable(rdev, false); uvd_v1_0_fini(rdev); radeon_uvd_suspend(rdev); + radeon_vce_suspend(rdev); cik_fini_pg(rdev); cik_fini_cg(rdev); cik_irq_suspend(rdev); @@ -8066,6 +8114,17 @@ int cik_init(struct radeon_device *rdev) r600_ring_init(rdev, ring, 4096); } + r = radeon_vce_init(rdev); + if (!r) { + ring = &rdev->ring[TN_RING_TYPE_VCE1_INDEX]; + ring->ring_obj = NULL; + r600_ring_init(rdev, ring, 4096); + + ring = &rdev->ring[TN_RING_TYPE_VCE2_INDEX]; + ring->ring_obj = NULL; + r600_ring_init(rdev, ring, 4096); + } + rdev->ih.ring_obj = NULL; r600_ih_ring_init(rdev, 64 * 1024); @@ -8127,6 +8186,7 @@ void cik_fini(struct radeon_device *rdev) radeon_irq_kms_fini(rdev); uvd_v1_0_fini(rdev); radeon_uvd_fini(rdev); + radeon_vce_fini(rdev); cik_pcie_gart_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); @@ -8865,6 +8925,41 @@ int cik_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk) return r; } +int cik_set_vce_clocks(struct radeon_device *rdev, u32 evclk, u32 ecclk) +{ + int r, i; + struct atom_clock_dividers dividers; + u32 tmp; + + r = radeon_atom_get_clock_dividers(rdev, COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK, + ecclk, false, ÷rs); + if (r) + return r; + + for (i = 0; i < 100; i++) { + if (RREG32_SMC(CG_ECLK_STATUS) & ECLK_STATUS) + break; + mdelay(10); + } + if (i == 100) + return -ETIMEDOUT; + + tmp = RREG32_SMC(CG_ECLK_CNTL); + tmp &= ~(ECLK_DIR_CNTL_EN|ECLK_DIVIDER_MASK); + tmp |= dividers.post_divider; + WREG32_SMC(CG_ECLK_CNTL, tmp); + + for (i = 0; i < 100; i++) { + if (RREG32_SMC(CG_ECLK_STATUS) & ECLK_STATUS) + break; + mdelay(10); + } + if (i == 100) + return -ETIMEDOUT; + + return 0; +} + static void cik_pcie_gen3_enable(struct radeon_device *rdev) { struct pci_dev *root = rdev->pdev->bus->self; diff --git a/drivers/gpu/drm/radeon/cik_sdma.c b/drivers/gpu/drm/radeon/cik_sdma.c index 1ecb3f1070e3..00150ac49cd2 100644 --- a/drivers/gpu/drm/radeon/cik_sdma.c +++ b/drivers/gpu/drm/radeon/cik_sdma.c @@ -362,8 +362,6 @@ static int cik_sdma_gfx_resume(struct radeon_device *rdev) ring->wptr = 0; WREG32(SDMA0_GFX_RB_WPTR + reg_offset, ring->wptr << 2); - ring->rptr = RREG32(SDMA0_GFX_RB_RPTR + reg_offset) >> 2; - /* enable DMA RB */ WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl | SDMA_RB_ENABLE); @@ -713,11 +711,9 @@ bool cik_sdma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) mask = RADEON_RESET_DMA1; if (!(reset_mask & mask)) { - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - /* force ring activities */ - radeon_ring_force_activity(rdev, ring); return radeon_ring_test_lockup(rdev, ring); } diff --git a/drivers/gpu/drm/radeon/cikd.h b/drivers/gpu/drm/radeon/cikd.h index 98bae9d7b74d..213873270d5f 100644 --- a/drivers/gpu/drm/radeon/cikd.h +++ b/drivers/gpu/drm/radeon/cikd.h @@ -203,6 +203,12 @@ #define CTF_TEMP_MASK 0x0003fe00 #define CTF_TEMP_SHIFT 9 +#define CG_ECLK_CNTL 0xC05000AC +# define ECLK_DIVIDER_MASK 0x7f +# define ECLK_DIR_CNTL_EN (1 << 8) +#define CG_ECLK_STATUS 0xC05000B0 +# define ECLK_STATUS (1 << 0) + #define CG_SPLL_FUNC_CNTL 0xC0500140 #define SPLL_RESET (1 << 0) #define SPLL_PWRON (1 << 1) @@ -2010,4 +2016,47 @@ /* UVD CTX indirect */ #define UVD_CGC_MEM_CTRL 0xC0 +/* VCE */ + +#define VCE_VCPU_CACHE_OFFSET0 0x20024 +#define VCE_VCPU_CACHE_SIZE0 0x20028 +#define VCE_VCPU_CACHE_OFFSET1 0x2002c +#define VCE_VCPU_CACHE_SIZE1 0x20030 +#define VCE_VCPU_CACHE_OFFSET2 0x20034 +#define VCE_VCPU_CACHE_SIZE2 0x20038 +#define VCE_RB_RPTR2 0x20178 +#define VCE_RB_WPTR2 0x2017c +#define VCE_RB_RPTR 0x2018c +#define VCE_RB_WPTR 0x20190 +#define VCE_CLOCK_GATING_A 0x202f8 +# define CGC_CLK_GATE_DLY_TIMER_MASK (0xf << 0) +# define CGC_CLK_GATE_DLY_TIMER(x) ((x) << 0) +# define CGC_CLK_GATER_OFF_DLY_TIMER_MASK (0xff << 4) +# define CGC_CLK_GATER_OFF_DLY_TIMER(x) ((x) << 4) +# define CGC_UENC_WAIT_AWAKE (1 << 18) +#define VCE_CLOCK_GATING_B 0x202fc +#define VCE_CGTT_CLK_OVERRIDE 0x207a0 +#define VCE_UENC_CLOCK_GATING 0x207bc +# define CLOCK_ON_DELAY_MASK (0xf << 0) +# define CLOCK_ON_DELAY(x) ((x) << 0) +# define CLOCK_OFF_DELAY_MASK (0xff << 4) +# define CLOCK_OFF_DELAY(x) ((x) << 4) +#define VCE_UENC_REG_CLOCK_GATING 0x207c0 +#define VCE_SYS_INT_EN 0x21300 +# define VCE_SYS_INT_TRAP_INTERRUPT_EN (1 << 3) +#define VCE_LMI_CTRL2 0x21474 +#define VCE_LMI_CTRL 0x21498 +#define VCE_LMI_VM_CTRL 0x214a0 +#define VCE_LMI_SWAP_CNTL 0x214b4 +#define VCE_LMI_SWAP_CNTL1 0x214b8 +#define VCE_LMI_CACHE_CTRL 0x214f4 + +#define VCE_CMD_NO_OP 0x00000000 +#define VCE_CMD_END 0x00000001 +#define VCE_CMD_IB 0x00000002 +#define VCE_CMD_FENCE 0x00000003 +#define VCE_CMD_TRAP 0x00000004 +#define VCE_CMD_IB_AUTO 0x00000005 +#define VCE_CMD_SEMAPHORE 0x00000006 + #endif diff --git a/drivers/gpu/drm/radeon/cypress_dpm.c b/drivers/gpu/drm/radeon/cypress_dpm.c index cf783fc0ef21..5a9a5f4d7888 100644 --- a/drivers/gpu/drm/radeon/cypress_dpm.c +++ b/drivers/gpu/drm/radeon/cypress_dpm.c @@ -2036,6 +2036,10 @@ int cypress_dpm_init(struct radeon_device *rdev) pi->min_vddc_in_table = 0; pi->max_vddc_in_table = 0; + ret = r600_get_platform_caps(rdev); + if (ret) + return ret; + ret = rv7xx_parse_power_table(rdev); if (ret) return ret; diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 5623e7542d99..b1f1253e2ced 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -2990,8 +2990,6 @@ static int evergreen_cp_resume(struct radeon_device *rdev) WREG32(CP_RB_BASE, ring->gpu_addr >> 8); WREG32(CP_DEBUG, (1 << 27) | (1 << 28)); - ring->rptr = RREG32(CP_RB_RPTR); - evergreen_cp_start(rdev); ring->ready = true; r = radeon_ring_test(rdev, RADEON_RING_TYPE_GFX_INDEX, ring); @@ -3952,11 +3950,9 @@ bool evergreen_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *rin if (!(reset_mask & (RADEON_RESET_GFX | RADEON_RESET_COMPUTE | RADEON_RESET_CP))) { - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - /* force CP activities */ - radeon_ring_force_activity(rdev, ring); return radeon_ring_test_lockup(rdev, ring); } diff --git a/drivers/gpu/drm/radeon/evergreen_dma.c b/drivers/gpu/drm/radeon/evergreen_dma.c index a37b54436382..287fe966d7de 100644 --- a/drivers/gpu/drm/radeon/evergreen_dma.c +++ b/drivers/gpu/drm/radeon/evergreen_dma.c @@ -174,11 +174,9 @@ bool evergreen_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *rin u32 reset_mask = evergreen_gpu_check_soft_reset(rdev); if (!(reset_mask & RADEON_RESET_DMA)) { - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - /* force ring activities */ - radeon_ring_force_activity(rdev, ring); return radeon_ring_test_lockup(rdev, ring); } diff --git a/drivers/gpu/drm/radeon/kv_dpm.c b/drivers/gpu/drm/radeon/kv_dpm.c index 351db361239d..16ec9d56a234 100644 --- a/drivers/gpu/drm/radeon/kv_dpm.c +++ b/drivers/gpu/drm/radeon/kv_dpm.c @@ -1338,13 +1338,11 @@ static int kv_enable_uvd_dpm(struct radeon_device *rdev, bool enable) PPSMC_MSG_UVDDPM_Enable : PPSMC_MSG_UVDDPM_Disable); } -#if 0 static int kv_enable_vce_dpm(struct radeon_device *rdev, bool enable) { return kv_notify_message_to_smu(rdev, enable ? PPSMC_MSG_VCEDPM_Enable : PPSMC_MSG_VCEDPM_Disable); } -#endif static int kv_enable_samu_dpm(struct radeon_device *rdev, bool enable) { @@ -1389,7 +1387,6 @@ static int kv_update_uvd_dpm(struct radeon_device *rdev, bool gate) return kv_enable_uvd_dpm(rdev, !gate); } -#if 0 static u8 kv_get_vce_boot_level(struct radeon_device *rdev) { u8 i; @@ -1414,6 +1411,9 @@ static int kv_update_vce_dpm(struct radeon_device *rdev, int ret; if (radeon_new_state->evclk > 0 && radeon_current_state->evclk == 0) { + kv_dpm_powergate_vce(rdev, false); + /* turn the clocks on when encoding */ + cik_update_cg(rdev, RADEON_CG_BLOCK_VCE, false); if (pi->caps_stable_p_state) pi->vce_boot_level = table->count - 1; else @@ -1436,11 +1436,13 @@ static int kv_update_vce_dpm(struct radeon_device *rdev, kv_enable_vce_dpm(rdev, true); } else if (radeon_new_state->evclk == 0 && radeon_current_state->evclk > 0) { kv_enable_vce_dpm(rdev, false); + /* turn the clocks off when not encoding */ + cik_update_cg(rdev, RADEON_CG_BLOCK_VCE, true); + kv_dpm_powergate_vce(rdev, true); } return 0; } -#endif static int kv_update_samu_dpm(struct radeon_device *rdev, bool gate) { @@ -1575,11 +1577,16 @@ static void kv_dpm_powergate_vce(struct radeon_device *rdev, bool gate) pi->vce_power_gated = gate; if (gate) { - if (pi->caps_vce_pg) + if (pi->caps_vce_pg) { + /* XXX do we need a vce_v1_0_stop() ? */ kv_notify_message_to_smu(rdev, PPSMC_MSG_VCEPowerOFF); + } } else { - if (pi->caps_vce_pg) + if (pi->caps_vce_pg) { kv_notify_message_to_smu(rdev, PPSMC_MSG_VCEPowerON); + vce_v2_0_resume(rdev); + vce_v1_0_start(rdev); + } } } @@ -1768,7 +1775,7 @@ int kv_dpm_set_power_state(struct radeon_device *rdev) { struct kv_power_info *pi = kv_get_pi(rdev); struct radeon_ps *new_ps = &pi->requested_rps; - /*struct radeon_ps *old_ps = &pi->current_rps;*/ + struct radeon_ps *old_ps = &pi->current_rps; int ret; if (pi->bapm_enable) { @@ -1798,13 +1805,12 @@ int kv_dpm_set_power_state(struct radeon_device *rdev) kv_set_enabled_levels(rdev); kv_force_lowest_valid(rdev); kv_unforce_levels(rdev); -#if 0 + ret = kv_update_vce_dpm(rdev, new_ps, old_ps); if (ret) { DRM_ERROR("kv_update_vce_dpm failed\n"); return ret; } -#endif kv_update_sclk_t(rdev); } } else { @@ -1823,13 +1829,11 @@ int kv_dpm_set_power_state(struct radeon_device *rdev) kv_program_nbps_index_settings(rdev, new_ps); kv_freeze_sclk_dpm(rdev, false); kv_set_enabled_levels(rdev); -#if 0 ret = kv_update_vce_dpm(rdev, new_ps, old_ps); if (ret) { DRM_ERROR("kv_update_vce_dpm failed\n"); return ret; } -#endif kv_update_acp_boot_level(rdev); kv_update_sclk_t(rdev); kv_enable_nb_dpm(rdev); @@ -2037,6 +2041,14 @@ static void kv_apply_state_adjust_rules(struct radeon_device *rdev, struct radeon_clock_and_voltage_limits *max_limits = &rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac; + if (new_rps->vce_active) { + new_rps->evclk = rdev->pm.dpm.vce_states[rdev->pm.dpm.vce_level].evclk; + new_rps->ecclk = rdev->pm.dpm.vce_states[rdev->pm.dpm.vce_level].ecclk; + } else { + new_rps->evclk = 0; + new_rps->ecclk = 0; + } + mclk = max_limits->mclk; sclk = min_sclk; @@ -2056,6 +2068,11 @@ static void kv_apply_state_adjust_rules(struct radeon_device *rdev, sclk = stable_p_state_sclk; } + if (new_rps->vce_active) { + if (sclk < rdev->pm.dpm.vce_states[rdev->pm.dpm.vce_level].sclk) + sclk = rdev->pm.dpm.vce_states[rdev->pm.dpm.vce_level].sclk; + } + ps->need_dfs_bypass = true; for (i = 0; i < ps->num_levels; i++) { @@ -2092,7 +2109,8 @@ static void kv_apply_state_adjust_rules(struct radeon_device *rdev, } } - pi->video_start = new_rps->dclk || new_rps->vclk; + pi->video_start = new_rps->dclk || new_rps->vclk || + new_rps->evclk || new_rps->ecclk; if ((new_rps->class & ATOM_PPLIB_CLASSIFICATION_UI_MASK) == ATOM_PPLIB_CLASSIFICATION_UI_BATTERY) @@ -2538,9 +2556,6 @@ static int kv_parse_power_table(struct radeon_device *rdev) if (!rdev->pm.dpm.ps) return -ENOMEM; power_state_offset = (u8 *)state_array->states; - rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); - rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); - rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); for (i = 0; i < state_array->ucNumEntries; i++) { u8 *idx; power_state = (union pplib_power_state *)power_state_offset; @@ -2577,6 +2592,19 @@ static int kv_parse_power_table(struct radeon_device *rdev) power_state_offset += 2 + power_state->v2.ucNumDPMLevels; } rdev->pm.dpm.num_ps = state_array->ucNumEntries; + + /* fill in the vce power states */ + for (i = 0; i < RADEON_MAX_VCE_LEVELS; i++) { + u32 sclk; + clock_array_index = rdev->pm.dpm.vce_states[i].clk_idx; + clock_info = (union pplib_clock_info *) + &clock_info_array->clockInfo[clock_array_index * clock_info_array->ucEntrySize]; + sclk = le16_to_cpu(clock_info->sumo.usEngineClockLow); + sclk |= clock_info->sumo.ucEngineClockHigh << 16; + rdev->pm.dpm.vce_states[i].sclk = sclk; + rdev->pm.dpm.vce_states[i].mclk = 0; + } + return 0; } @@ -2590,6 +2618,10 @@ int kv_dpm_init(struct radeon_device *rdev) return -ENOMEM; rdev->pm.dpm.priv = pi; + ret = r600_get_platform_caps(rdev); + if (ret) + return ret; + ret = r600_parse_extended_power_table(rdev); if (ret) return ret; @@ -2623,7 +2655,7 @@ int kv_dpm_init(struct radeon_device *rdev) pi->caps_fps = false; /* true? */ pi->caps_uvd_pg = true; pi->caps_uvd_dpm = true; - pi->caps_vce_pg = false; + pi->caps_vce_pg = false; /* XXX true */ pi->caps_samu_pg = false; pi->caps_acp_pg = false; pi->caps_stable_p_state = false; diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index ea932ac66fc6..85168ecd216b 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -1642,8 +1642,8 @@ static int cayman_cp_resume(struct radeon_device *rdev) ring = &rdev->ring[ridx[i]]; WREG32_P(cp_rb_cntl[i], RB_RPTR_WR_ENA, ~RB_RPTR_WR_ENA); - ring->rptr = ring->wptr = 0; - WREG32(cp_rb_rptr[i], ring->rptr); + ring->wptr = 0; + WREG32(cp_rb_rptr[i], 0); WREG32(cp_rb_wptr[i], ring->wptr); mdelay(1); @@ -1917,11 +1917,9 @@ bool cayman_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) if (!(reset_mask & (RADEON_RESET_GFX | RADEON_RESET_COMPUTE | RADEON_RESET_CP))) { - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - /* force CP activities */ - radeon_ring_force_activity(rdev, ring); return radeon_ring_test_lockup(rdev, ring); } diff --git a/drivers/gpu/drm/radeon/ni_dma.c b/drivers/gpu/drm/radeon/ni_dma.c index 7cf96b15377f..6378e0276691 100644 --- a/drivers/gpu/drm/radeon/ni_dma.c +++ b/drivers/gpu/drm/radeon/ni_dma.c @@ -248,8 +248,6 @@ int cayman_dma_resume(struct radeon_device *rdev) ring->wptr = 0; WREG32(DMA_RB_WPTR + reg_offset, ring->wptr << 2); - ring->rptr = RREG32(DMA_RB_RPTR + reg_offset) >> 2; - WREG32(DMA_RB_CNTL + reg_offset, rb_cntl | DMA_RB_ENABLE); ring->ready = true; @@ -302,11 +300,9 @@ bool cayman_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) mask = RADEON_RESET_DMA1; if (!(reset_mask & mask)) { - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - /* force ring activities */ - radeon_ring_force_activity(rdev, ring); return radeon_ring_test_lockup(rdev, ring); } diff --git a/drivers/gpu/drm/radeon/ni_dpm.c b/drivers/gpu/drm/radeon/ni_dpm.c index ca814276b075..004c931606c4 100644 --- a/drivers/gpu/drm/radeon/ni_dpm.c +++ b/drivers/gpu/drm/radeon/ni_dpm.c @@ -4025,9 +4025,6 @@ static int ni_parse_power_table(struct radeon_device *rdev) power_info->pplib.ucNumStates, GFP_KERNEL); if (!rdev->pm.dpm.ps) return -ENOMEM; - rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); - rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); - rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); for (i = 0; i < power_info->pplib.ucNumStates; i++) { power_state = (union pplib_power_state *) @@ -4089,6 +4086,10 @@ int ni_dpm_init(struct radeon_device *rdev) pi->min_vddc_in_table = 0; pi->max_vddc_in_table = 0; + ret = r600_get_platform_caps(rdev); + if (ret) + return ret; + ret = ni_parse_power_table(rdev); if (ret) return ret; diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c index ef024ce3f7cc..1690a2dc0721 100644 --- a/drivers/gpu/drm/radeon/r100.c +++ b/drivers/gpu/drm/radeon/r100.c @@ -1193,7 +1193,6 @@ int r100_cp_init(struct radeon_device *rdev, unsigned ring_size) WREG32(RADEON_CP_RB_CNTL, tmp); udelay(10); - ring->rptr = RREG32(RADEON_CP_RB_RPTR); /* Set cp mode to bus mastering & enable cp*/ WREG32(RADEON_CP_CSQ_MODE, REG_SET(RADEON_INDIRECT2_START, indirect2_start) | @@ -2523,11 +2522,9 @@ bool r100_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) rbbm_status = RREG32(R_000E40_RBBM_STATUS); if (!G_000E40_GUI_ACTIVE(rbbm_status)) { - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - /* force CP activities */ - radeon_ring_force_activity(rdev, ring); return radeon_ring_test_lockup(rdev, ring); } diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index cdbc4171fe73..0f4ab928a15a 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -1748,11 +1748,9 @@ bool r600_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) if (!(reset_mask & (RADEON_RESET_GFX | RADEON_RESET_COMPUTE | RADEON_RESET_CP))) { - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - /* force CP activities */ - radeon_ring_force_activity(rdev, ring); return radeon_ring_test_lockup(rdev, ring); } @@ -2604,8 +2602,6 @@ int r600_cp_resume(struct radeon_device *rdev) WREG32(CP_RB_BASE, ring->gpu_addr >> 8); WREG32(CP_DEBUG, (1 << 27) | (1 << 28)); - ring->rptr = RREG32(CP_RB_RPTR); - r600_cp_start(rdev); ring->ready = true; r = radeon_ring_test(rdev, RADEON_RING_TYPE_GFX_INDEX, ring); diff --git a/drivers/gpu/drm/radeon/r600_dma.c b/drivers/gpu/drm/radeon/r600_dma.c index b2d4c91e6272..53fcb28f5578 100644 --- a/drivers/gpu/drm/radeon/r600_dma.c +++ b/drivers/gpu/drm/radeon/r600_dma.c @@ -176,8 +176,6 @@ int r600_dma_resume(struct radeon_device *rdev) ring->wptr = 0; WREG32(DMA_RB_WPTR, ring->wptr << 2); - ring->rptr = RREG32(DMA_RB_RPTR) >> 2; - WREG32(DMA_RB_CNTL, rb_cntl | DMA_RB_ENABLE); ring->ready = true; @@ -221,11 +219,9 @@ bool r600_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) u32 reset_mask = r600_gpu_check_soft_reset(rdev); if (!(reset_mask & RADEON_RESET_DMA)) { - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - /* force ring activities */ - radeon_ring_force_activity(rdev, ring); return radeon_ring_test_lockup(rdev, ring); } diff --git a/drivers/gpu/drm/radeon/r600_dpm.c b/drivers/gpu/drm/radeon/r600_dpm.c index e4cc9b314ce9..cbf7e3269f84 100644 --- a/drivers/gpu/drm/radeon/r600_dpm.c +++ b/drivers/gpu/drm/radeon/r600_dpm.c @@ -834,6 +834,26 @@ static int r600_parse_clk_voltage_dep_table(struct radeon_clock_voltage_dependen return 0; } +int r600_get_platform_caps(struct radeon_device *rdev) +{ + struct radeon_mode_info *mode_info = &rdev->mode_info; + union power_info *power_info; + int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo); + u16 data_offset; + u8 frev, crev; + + if (!atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) + return -EINVAL; + power_info = (union power_info *)(mode_info->atom_context->bios + data_offset); + + rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); + rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); + rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); + + return 0; +} + /* sizeof(ATOM_PPLIB_EXTENDEDHEADER) */ #define SIZE_OF_ATOM_PPLIB_EXTENDEDHEADER_V2 12 #define SIZE_OF_ATOM_PPLIB_EXTENDEDHEADER_V3 14 @@ -1043,7 +1063,15 @@ int r600_parse_extended_power_table(struct radeon_device *rdev) (mode_info->atom_context->bios + data_offset + le16_to_cpu(ext_hdr->usVCETableOffset) + 1 + 1 + array->ucNumEntries * sizeof(VCEClockInfo)); + ATOM_PPLIB_VCE_State_Table *states = + (ATOM_PPLIB_VCE_State_Table *) + (mode_info->atom_context->bios + data_offset + + le16_to_cpu(ext_hdr->usVCETableOffset) + 1 + + 1 + (array->ucNumEntries * sizeof (VCEClockInfo)) + + 1 + (limits->numEntries * sizeof(ATOM_PPLIB_VCE_Clock_Voltage_Limit_Record))); ATOM_PPLIB_VCE_Clock_Voltage_Limit_Record *entry; + ATOM_PPLIB_VCE_State_Record *state_entry; + VCEClockInfo *vce_clk; u32 size = limits->numEntries * sizeof(struct radeon_vce_clock_voltage_dependency_entry); rdev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table.entries = @@ -1055,8 +1083,9 @@ int r600_parse_extended_power_table(struct radeon_device *rdev) rdev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table.count = limits->numEntries; entry = &limits->entries[0]; + state_entry = &states->entries[0]; for (i = 0; i < limits->numEntries; i++) { - VCEClockInfo *vce_clk = (VCEClockInfo *) + vce_clk = (VCEClockInfo *) ((u8 *)&array->entries[0] + (entry->ucVCEClockInfoIndex * sizeof(VCEClockInfo))); rdev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table.entries[i].evclk = @@ -1068,6 +1097,23 @@ int r600_parse_extended_power_table(struct radeon_device *rdev) entry = (ATOM_PPLIB_VCE_Clock_Voltage_Limit_Record *) ((u8 *)entry + sizeof(ATOM_PPLIB_VCE_Clock_Voltage_Limit_Record)); } + for (i = 0; i < states->numEntries; i++) { + if (i >= RADEON_MAX_VCE_LEVELS) + break; + vce_clk = (VCEClockInfo *) + ((u8 *)&array->entries[0] + + (state_entry->ucVCEClockInfoIndex * sizeof(VCEClockInfo))); + rdev->pm.dpm.vce_states[i].evclk = + le16_to_cpu(vce_clk->usEVClkLow) | (vce_clk->ucEVClkHigh << 16); + rdev->pm.dpm.vce_states[i].ecclk = + le16_to_cpu(vce_clk->usECClkLow) | (vce_clk->ucECClkHigh << 16); + rdev->pm.dpm.vce_states[i].clk_idx = + state_entry->ucClockInfoIndex & 0x3f; + rdev->pm.dpm.vce_states[i].pstate = + (state_entry->ucClockInfoIndex & 0xc0) >> 6; + state_entry = (ATOM_PPLIB_VCE_State_Record *) + ((u8 *)state_entry + sizeof(ATOM_PPLIB_VCE_State_Record)); + } } if ((le16_to_cpu(ext_hdr->usSize) >= SIZE_OF_ATOM_PPLIB_EXTENDEDHEADER_V3) && ext_hdr->usUVDTableOffset) { diff --git a/drivers/gpu/drm/radeon/r600_dpm.h b/drivers/gpu/drm/radeon/r600_dpm.h index 07eab2b04e81..46b9d2a03018 100644 --- a/drivers/gpu/drm/radeon/r600_dpm.h +++ b/drivers/gpu/drm/radeon/r600_dpm.h @@ -215,6 +215,8 @@ void r600_stop_dpm(struct radeon_device *rdev); bool r600_is_internal_thermal_sensor(enum radeon_int_thermal_type sensor); +int r600_get_platform_caps(struct radeon_device *rdev); + int r600_parse_extended_power_table(struct radeon_device *rdev); void r600_free_extended_power_table(struct radeon_device *rdev); diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 024db37b1832..4581df193932 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -113,19 +113,16 @@ extern int radeon_hard_reset; #define RADEONFB_CONN_LIMIT 4 #define RADEON_BIOS_NUM_SCRATCH 8 -/* max number of rings */ -#define RADEON_NUM_RINGS 6 - /* fence seq are set to this number when signaled */ #define RADEON_FENCE_SIGNALED_SEQ 0LL /* internal ring indices */ /* r1xx+ has gfx CP ring */ -#define RADEON_RING_TYPE_GFX_INDEX 0 +#define RADEON_RING_TYPE_GFX_INDEX 0 /* cayman has 2 compute CP rings */ -#define CAYMAN_RING_TYPE_CP1_INDEX 1 -#define CAYMAN_RING_TYPE_CP2_INDEX 2 +#define CAYMAN_RING_TYPE_CP1_INDEX 1 +#define CAYMAN_RING_TYPE_CP2_INDEX 2 /* R600+ has an async dma ring */ #define R600_RING_TYPE_DMA_INDEX 3 @@ -133,7 +130,17 @@ extern int radeon_hard_reset; #define CAYMAN_RING_TYPE_DMA1_INDEX 4 /* R600+ */ -#define R600_RING_TYPE_UVD_INDEX 5 +#define R600_RING_TYPE_UVD_INDEX 5 + +/* TN+ */ +#define TN_RING_TYPE_VCE1_INDEX 6 +#define TN_RING_TYPE_VCE2_INDEX 7 + +/* max number of rings */ +#define RADEON_NUM_RINGS 8 + +/* number of hw syncs before falling back on blocking */ +#define RADEON_NUM_SYNCS 4 /* number of hw syncs before falling back on blocking */ #define RADEON_NUM_SYNCS 4 @@ -789,7 +796,6 @@ struct radeon_ib { struct radeon_ring { struct radeon_bo *ring_obj; volatile uint32_t *ring; - unsigned rptr; unsigned rptr_offs; unsigned rptr_save_reg; u64 next_rptr_gpu_addr; @@ -953,8 +959,8 @@ void radeon_ring_unlock_commit(struct radeon_device *rdev, struct radeon_ring *c void radeon_ring_undo(struct radeon_ring *ring); void radeon_ring_unlock_undo(struct radeon_device *rdev, struct radeon_ring *cp); int radeon_ring_test(struct radeon_device *rdev, struct radeon_ring *cp); -void radeon_ring_force_activity(struct radeon_device *rdev, struct radeon_ring *ring); -void radeon_ring_lockup_update(struct radeon_ring *ring); +void radeon_ring_lockup_update(struct radeon_device *rdev, + struct radeon_ring *ring); bool radeon_ring_test_lockup(struct radeon_device *rdev, struct radeon_ring *ring); unsigned radeon_ring_backup(struct radeon_device *rdev, struct radeon_ring *ring, uint32_t **data); @@ -1255,6 +1261,17 @@ enum radeon_dpm_event_src { RADEON_DPM_EVENT_SRC_DIGIAL_OR_EXTERNAL = 4 }; +#define RADEON_MAX_VCE_LEVELS 6 + +enum radeon_vce_level { + RADEON_VCE_LEVEL_AC_ALL = 0, /* AC, All cases */ + RADEON_VCE_LEVEL_DC_EE = 1, /* DC, entropy encoding */ + RADEON_VCE_LEVEL_DC_LL_LOW = 2, /* DC, low latency queue, res <= 720 */ + RADEON_VCE_LEVEL_DC_LL_HIGH = 3, /* DC, low latency queue, 1080 >= res > 720 */ + RADEON_VCE_LEVEL_DC_GP_LOW = 4, /* DC, general purpose queue, res <= 720 */ + RADEON_VCE_LEVEL_DC_GP_HIGH = 5, /* DC, general purpose queue, 1080 >= res > 720 */ +}; + struct radeon_ps { u32 caps; /* vbios flags */ u32 class; /* vbios flags */ @@ -1265,6 +1282,8 @@ struct radeon_ps { /* VCE clocks */ u32 evclk; u32 ecclk; + bool vce_active; + enum radeon_vce_level vce_level; /* asic priv */ void *ps_priv; }; @@ -1439,6 +1458,17 @@ enum radeon_dpm_forced_level { RADEON_DPM_FORCED_LEVEL_HIGH = 2, }; +struct radeon_vce_state { + /* vce clocks */ + u32 evclk; + u32 ecclk; + /* gpu clocks */ + u32 sclk; + u32 mclk; + u8 clk_idx; + u8 pstate; +}; + struct radeon_dpm { struct radeon_ps *ps; /* number of valid power states */ @@ -1451,6 +1481,9 @@ struct radeon_dpm { struct radeon_ps *boot_ps; /* default uvd power state */ struct radeon_ps *uvd_ps; + /* vce requirements */ + struct radeon_vce_state vce_states[RADEON_MAX_VCE_LEVELS]; + enum radeon_vce_level vce_level; enum radeon_pm_state_type state; enum radeon_pm_state_type user_state; u32 platform_caps; @@ -1476,6 +1509,7 @@ struct radeon_dpm { /* special states active */ bool thermal_active; bool uvd_active; + bool vce_active; /* thermal handling */ struct radeon_dpm_thermal thermal; /* forced levels */ @@ -1486,6 +1520,7 @@ struct radeon_dpm { }; void radeon_dpm_enable_uvd(struct radeon_device *rdev, bool enable); +void radeon_dpm_enable_vce(struct radeon_device *rdev, bool enable); struct radeon_pm { struct mutex mutex; @@ -1591,6 +1626,46 @@ int radeon_uvd_calc_upll_dividers(struct radeon_device *rdev, int radeon_uvd_send_upll_ctlreq(struct radeon_device *rdev, unsigned cg_upll_func_cntl); +/* + * VCE + */ +#define RADEON_MAX_VCE_HANDLES 16 +#define RADEON_VCE_STACK_SIZE (1024*1024) +#define RADEON_VCE_HEAP_SIZE (4*1024*1024) + +struct radeon_vce { + struct radeon_bo *vcpu_bo; + void *cpu_addr; + uint64_t gpu_addr; + unsigned fw_version; + unsigned fb_version; + atomic_t handles[RADEON_MAX_VCE_HANDLES]; + struct drm_file *filp[RADEON_MAX_VCE_HANDLES]; + struct delayed_work idle_work; +}; + +int radeon_vce_init(struct radeon_device *rdev); +void radeon_vce_fini(struct radeon_device *rdev); +int radeon_vce_suspend(struct radeon_device *rdev); +int radeon_vce_resume(struct radeon_device *rdev); +int radeon_vce_get_create_msg(struct radeon_device *rdev, int ring, + uint32_t handle, struct radeon_fence **fence); +int radeon_vce_get_destroy_msg(struct radeon_device *rdev, int ring, + uint32_t handle, struct radeon_fence **fence); +void radeon_vce_free_handles(struct radeon_device *rdev, struct drm_file *filp); +void radeon_vce_note_usage(struct radeon_device *rdev); +int radeon_vce_cs_reloc(struct radeon_cs_parser *p, int lo, int hi); +int radeon_vce_cs_parse(struct radeon_cs_parser *p); +bool radeon_vce_semaphore_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait); +void radeon_vce_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib); +void radeon_vce_fence_emit(struct radeon_device *rdev, + struct radeon_fence *fence); +int radeon_vce_ring_test(struct radeon_device *rdev, struct radeon_ring *ring); +int radeon_vce_ib_test(struct radeon_device *rdev, struct radeon_ring *ring); + struct r600_audio_pin { int channels; int rate; @@ -1780,6 +1855,7 @@ struct radeon_asic { void (*set_pcie_lanes)(struct radeon_device *rdev, int lanes); void (*set_clock_gating)(struct radeon_device *rdev, int enable); int (*set_uvd_clocks)(struct radeon_device *rdev, u32 vclk, u32 dclk); + int (*set_vce_clocks)(struct radeon_device *rdev, u32 evclk, u32 ecclk); int (*get_temperature)(struct radeon_device *rdev); } pm; /* dynamic power management */ @@ -2186,6 +2262,7 @@ struct radeon_device { struct radeon_gem gem; struct radeon_pm pm; struct radeon_uvd uvd; + struct radeon_vce vce; uint32_t bios_scratch[RADEON_BIOS_NUM_SCRATCH]; struct radeon_wb wb; struct radeon_dummy_page dummy_page; @@ -2205,6 +2282,7 @@ struct radeon_device { const struct firmware *sdma_fw; /* CIK SDMA firmware */ const struct firmware *smc_fw; /* SMC firmware */ const struct firmware *uvd_fw; /* UVD firmware */ + const struct firmware *vce_fw; /* VCE firmware */ struct r600_vram_scratch vram_scratch; int msi_enabled; /* msi enabled */ struct r600_ih ih; /* r6/700 interrupt ring */ @@ -2639,6 +2717,7 @@ void radeon_ring_write(struct radeon_ring *ring, uint32_t v); #define radeon_set_pcie_lanes(rdev, l) (rdev)->asic->pm.set_pcie_lanes((rdev), (l)) #define radeon_set_clock_gating(rdev, e) (rdev)->asic->pm.set_clock_gating((rdev), (e)) #define radeon_set_uvd_clocks(rdev, v, d) (rdev)->asic->pm.set_uvd_clocks((rdev), (v), (d)) +#define radeon_set_vce_clocks(rdev, ev, ec) (rdev)->asic->pm.set_vce_clocks((rdev), (ev), (ec)) #define radeon_get_temperature(rdev) (rdev)->asic->pm.get_temperature((rdev)) #define radeon_set_surface_reg(rdev, r, f, p, o, s) ((rdev)->asic->surface.set_reg((rdev), (r), (f), (p), (o), (s))) #define radeon_clear_surface_reg(rdev, r) ((rdev)->asic->surface.clear_reg((rdev), (r))) diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c index dda02bfc10a4..b8a24a75d4ff 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.c +++ b/drivers/gpu/drm/radeon/radeon_asic.c @@ -1987,6 +1987,19 @@ static struct radeon_asic_ring ci_dma_ring = { .set_wptr = &cik_sdma_set_wptr, }; +static struct radeon_asic_ring ci_vce_ring = { + .ib_execute = &radeon_vce_ib_execute, + .emit_fence = &radeon_vce_fence_emit, + .emit_semaphore = &radeon_vce_semaphore_emit, + .cs_parse = &radeon_vce_cs_parse, + .ring_test = &radeon_vce_ring_test, + .ib_test = &radeon_vce_ib_test, + .is_lockup = &radeon_ring_test_lockup, + .get_rptr = &vce_v1_0_get_rptr, + .get_wptr = &vce_v1_0_get_wptr, + .set_wptr = &vce_v1_0_set_wptr, +}; + static struct radeon_asic ci_asic = { .init = &cik_init, .fini = &cik_fini, @@ -2015,6 +2028,8 @@ static struct radeon_asic ci_asic = { [R600_RING_TYPE_DMA_INDEX] = &ci_dma_ring, [CAYMAN_RING_TYPE_DMA1_INDEX] = &ci_dma_ring, [R600_RING_TYPE_UVD_INDEX] = &cayman_uvd_ring, + [TN_RING_TYPE_VCE1_INDEX] = &ci_vce_ring, + [TN_RING_TYPE_VCE2_INDEX] = &ci_vce_ring, }, .irq = { .set = &cik_irq_set, @@ -2061,6 +2076,7 @@ static struct radeon_asic ci_asic = { .set_pcie_lanes = NULL, .set_clock_gating = NULL, .set_uvd_clocks = &cik_set_uvd_clocks, + .set_vce_clocks = &cik_set_vce_clocks, .get_temperature = &ci_get_temp, }, .dpm = { @@ -2117,6 +2133,8 @@ static struct radeon_asic kv_asic = { [R600_RING_TYPE_DMA_INDEX] = &ci_dma_ring, [CAYMAN_RING_TYPE_DMA1_INDEX] = &ci_dma_ring, [R600_RING_TYPE_UVD_INDEX] = &cayman_uvd_ring, + [TN_RING_TYPE_VCE1_INDEX] = &ci_vce_ring, + [TN_RING_TYPE_VCE2_INDEX] = &ci_vce_ring, }, .irq = { .set = &cik_irq_set, @@ -2163,6 +2181,7 @@ static struct radeon_asic kv_asic = { .set_pcie_lanes = NULL, .set_clock_gating = NULL, .set_uvd_clocks = &cik_set_uvd_clocks, + .set_vce_clocks = &cik_set_vce_clocks, .get_temperature = &kv_get_temp, }, .dpm = { diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index ae637cfda783..3d55a3a39e82 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -717,6 +717,7 @@ u32 cik_get_xclk(struct radeon_device *rdev); uint32_t cik_pciep_rreg(struct radeon_device *rdev, uint32_t reg); void cik_pciep_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v); int cik_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk); +int cik_set_vce_clocks(struct radeon_device *rdev, u32 evclk, u32 ecclk); void cik_sdma_fence_ring_emit(struct radeon_device *rdev, struct radeon_fence *fence); bool cik_sdma_semaphore_ring_emit(struct radeon_device *rdev, @@ -863,4 +864,17 @@ bool uvd_v3_1_semaphore_emit(struct radeon_device *rdev, /* uvd v4.2 */ int uvd_v4_2_resume(struct radeon_device *rdev); +/* vce v1.0 */ +uint32_t vce_v1_0_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring); +uint32_t vce_v1_0_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); +void vce_v1_0_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); +int vce_v1_0_init(struct radeon_device *rdev); +int vce_v1_0_start(struct radeon_device *rdev); + +/* vce v2.0 */ +int vce_v2_0_resume(struct radeon_device *rdev); + #endif diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c index dfb5a1db87d4..f28a8d82fa19 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -147,6 +147,10 @@ static int radeon_cs_get_ring(struct radeon_cs_parser *p, u32 ring, s32 priority case RADEON_CS_RING_UVD: p->ring = R600_RING_TYPE_UVD_INDEX; break; + case RADEON_CS_RING_VCE: + /* TODO: only use the low priority ring for now */ + p->ring = TN_RING_TYPE_VCE1_INDEX; + break; } return 0; } @@ -343,6 +347,9 @@ static int radeon_cs_ib_chunk(struct radeon_device *rdev, if (parser->ring == R600_RING_TYPE_UVD_INDEX) radeon_uvd_note_usage(rdev); + else if ((parser->ring == TN_RING_TYPE_VCE1_INDEX) || + (parser->ring == TN_RING_TYPE_VCE2_INDEX)) + radeon_vce_note_usage(rdev); radeon_cs_sync_rings(parser); r = radeon_ib_schedule(rdev, &parser->ib, NULL); diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c index 114d1672d616..baff98be65b1 100644 --- a/drivers/gpu/drm/radeon/radeon_kms.c +++ b/drivers/gpu/drm/radeon/radeon_kms.c @@ -433,6 +433,9 @@ static int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file case RADEON_CS_RING_UVD: *value = rdev->ring[R600_RING_TYPE_UVD_INDEX].ready; break; + case RADEON_CS_RING_VCE: + *value = rdev->ring[TN_RING_TYPE_VCE1_INDEX].ready; + break; default: return -EINVAL; } @@ -477,6 +480,12 @@ static int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file else *value = rdev->pm.default_sclk * 10; break; + case RADEON_INFO_VCE_FW_VERSION: + *value = rdev->vce.fw_version; + break; + case RADEON_INFO_VCE_FB_VERSION: + *value = rdev->vce.fb_version; + break; default: DRM_DEBUG_KMS("Invalid request %d\n", info->request); return -EINVAL; @@ -610,6 +619,7 @@ void radeon_driver_preclose_kms(struct drm_device *dev, if (rdev->cmask_filp == file_priv) rdev->cmask_filp = NULL; radeon_uvd_free_handles(rdev, file_priv); + radeon_vce_free_handles(rdev, file_priv); } /* diff --git a/drivers/gpu/drm/radeon/radeon_pm.c b/drivers/gpu/drm/radeon/radeon_pm.c index 8e8153e471c2..4ad9af9fc517 100644 --- a/drivers/gpu/drm/radeon/radeon_pm.c +++ b/drivers/gpu/drm/radeon/radeon_pm.c @@ -826,6 +826,9 @@ static void radeon_dpm_change_power_state_locked(struct radeon_device *rdev) /* no need to reprogram if nothing changed unless we are on BTC+ */ if (rdev->pm.dpm.current_ps == rdev->pm.dpm.requested_ps) { + /* vce just modifies an existing state so force a change */ + if (ps->vce_active != rdev->pm.dpm.vce_active) + goto force; if ((rdev->family < CHIP_BARTS) || (rdev->flags & RADEON_IS_IGP)) { /* for pre-BTC and APUs if the num crtcs changed but state is the same, * all we need to do is update the display configuration. @@ -862,16 +865,21 @@ static void radeon_dpm_change_power_state_locked(struct radeon_device *rdev) } } +force: if (radeon_dpm == 1) { printk("switching from power state:\n"); radeon_dpm_print_power_state(rdev, rdev->pm.dpm.current_ps); printk("switching to power state:\n"); radeon_dpm_print_power_state(rdev, rdev->pm.dpm.requested_ps); } + mutex_lock(&rdev->ddev->struct_mutex); down_write(&rdev->pm.mclk_lock); mutex_lock(&rdev->ring_lock); + /* update whether vce is active */ + ps->vce_active = rdev->pm.dpm.vce_active; + ret = radeon_dpm_pre_set_power_state(rdev); if (ret) goto done; @@ -960,6 +968,23 @@ void radeon_dpm_enable_uvd(struct radeon_device *rdev, bool enable) } } +void radeon_dpm_enable_vce(struct radeon_device *rdev, bool enable) +{ + if (enable) { + mutex_lock(&rdev->pm.mutex); + rdev->pm.dpm.vce_active = true; + /* XXX select vce level based on ring/task */ + rdev->pm.dpm.vce_level = RADEON_VCE_LEVEL_AC_ALL; + mutex_unlock(&rdev->pm.mutex); + } else { + mutex_lock(&rdev->pm.mutex); + rdev->pm.dpm.vce_active = false; + mutex_unlock(&rdev->pm.mutex); + } + + radeon_pm_compute_clocks(rdev); +} + static void radeon_pm_suspend_old(struct radeon_device *rdev) { mutex_lock(&rdev->pm.mutex); diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index 15e44a7281ab..b14c86d57607 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -342,13 +342,17 @@ bool radeon_ring_supports_scratch_reg(struct radeon_device *rdev, */ void radeon_ring_free_size(struct radeon_device *rdev, struct radeon_ring *ring) { - ring->rptr = radeon_ring_get_rptr(rdev, ring); + uint32_t rptr = radeon_ring_get_rptr(rdev, ring); + /* This works because ring_size is a power of 2 */ - ring->ring_free_dw = (ring->rptr + (ring->ring_size / 4)); + ring->ring_free_dw = rptr + (ring->ring_size / 4); ring->ring_free_dw -= ring->wptr; ring->ring_free_dw &= ring->ptr_mask; if (!ring->ring_free_dw) { + /* this is an empty ring */ ring->ring_free_dw = ring->ring_size / 4; + /* update lockup info to avoid false positive */ + radeon_ring_lockup_update(rdev, ring); } } @@ -372,12 +376,6 @@ int radeon_ring_alloc(struct radeon_device *rdev, struct radeon_ring *ring, unsi /* Align requested size with padding so unlock_commit can * pad safely */ radeon_ring_free_size(rdev, ring); - if (ring->ring_free_dw == (ring->ring_size / 4)) { - /* This is an empty ring update lockup info to avoid - * false positive. - */ - radeon_ring_lockup_update(ring); - } ndw = (ndw + ring->align_mask) & ~ring->align_mask; while (ndw > (ring->ring_free_dw - 1)) { radeon_ring_free_size(rdev, ring); @@ -478,38 +476,16 @@ void radeon_ring_unlock_undo(struct radeon_device *rdev, struct radeon_ring *rin } /** - * radeon_ring_force_activity - add some nop packets to the ring - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * - * Add some nop packets to the ring to force activity (all asics). - * Used for lockup detection to see if the rptr is advancing. - */ -void radeon_ring_force_activity(struct radeon_device *rdev, struct radeon_ring *ring) -{ - int r; - - radeon_ring_free_size(rdev, ring); - if (ring->rptr == ring->wptr) { - r = radeon_ring_alloc(rdev, ring, 1); - if (!r) { - radeon_ring_write(ring, ring->nop); - radeon_ring_commit(rdev, ring); - } - } -} - -/** * radeon_ring_lockup_update - update lockup variables * * @ring: radeon_ring structure holding ring information * * Update the last rptr value and timestamp (all asics). */ -void radeon_ring_lockup_update(struct radeon_ring *ring) +void radeon_ring_lockup_update(struct radeon_device *rdev, + struct radeon_ring *ring) { - ring->last_rptr = ring->rptr; + ring->last_rptr = radeon_ring_get_rptr(rdev, ring); ring->last_activity = jiffies; } @@ -518,35 +494,21 @@ void radeon_ring_lockup_update(struct radeon_ring *ring) * @rdev: radeon device structure * @ring: radeon_ring structure holding ring information * - * We don't need to initialize the lockup tracking information as we will either - * have CP rptr to a different value of jiffies wrap around which will force - * initialization of the lockup tracking informations. - * - * A possible false positivie is if we get call after while and last_cp_rptr == - * the current CP rptr, even if it's unlikely it might happen. To avoid this - * if the elapsed time since last call is bigger than 2 second than we return - * false and update the tracking information. Due to this the caller must call - * radeon_ring_test_lockup several time in less than 2sec for lockup to be reported - * the fencing code should be cautious about that. - * - * Caller should write to the ring to force CP to do something so we don't get - * false positive when CP is just gived nothing to do. - * - **/ + */ bool radeon_ring_test_lockup(struct radeon_device *rdev, struct radeon_ring *ring) { + uint32_t rptr = radeon_ring_get_rptr(rdev, ring); unsigned long cjiffies, elapsed; cjiffies = jiffies; if (!time_after(cjiffies, ring->last_activity)) { /* likely a wrap around */ - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - ring->rptr = radeon_ring_get_rptr(rdev, ring); - if (ring->rptr != ring->last_rptr) { + if (rptr != ring->last_rptr) { /* CP is still working no lockup */ - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } elapsed = jiffies_to_msecs(cjiffies - ring->last_activity); @@ -709,7 +671,7 @@ int radeon_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsig if (radeon_debugfs_ring_init(rdev, ring)) { DRM_ERROR("Failed to register debugfs file for rings !\n"); } - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return 0; } @@ -780,8 +742,6 @@ static int radeon_debugfs_ring_info(struct seq_file *m, void *data) seq_printf(m, "driver's copy of the wptr: 0x%08x [%5d]\n", ring->wptr, ring->wptr); - seq_printf(m, "driver's copy of the rptr: 0x%08x [%5d]\n", - ring->rptr, ring->rptr); seq_printf(m, "last semaphore signal addr : 0x%016llx\n", ring->last_semaphore_signal_addr); seq_printf(m, "last semaphore wait addr : 0x%016llx\n", @@ -814,6 +774,8 @@ static int cayman_cp2_index = CAYMAN_RING_TYPE_CP2_INDEX; static int radeon_dma1_index = R600_RING_TYPE_DMA_INDEX; static int radeon_dma2_index = CAYMAN_RING_TYPE_DMA1_INDEX; static int r600_uvd_index = R600_RING_TYPE_UVD_INDEX; +static int si_vce1_index = TN_RING_TYPE_VCE1_INDEX; +static int si_vce2_index = TN_RING_TYPE_VCE2_INDEX; static struct drm_info_list radeon_debugfs_ring_info_list[] = { {"radeon_ring_gfx", radeon_debugfs_ring_info, 0, &radeon_gfx_index}, @@ -822,6 +784,8 @@ static struct drm_info_list radeon_debugfs_ring_info_list[] = { {"radeon_ring_dma1", radeon_debugfs_ring_info, 0, &radeon_dma1_index}, {"radeon_ring_dma2", radeon_debugfs_ring_info, 0, &radeon_dma2_index}, {"radeon_ring_uvd", radeon_debugfs_ring_info, 0, &r600_uvd_index}, + {"radeon_ring_vce1", radeon_debugfs_ring_info, 0, &si_vce1_index}, + {"radeon_ring_vce2", radeon_debugfs_ring_info, 0, &si_vce2_index}, }; static int radeon_debugfs_sa_info(struct seq_file *m, void *data) diff --git a/drivers/gpu/drm/radeon/radeon_test.c b/drivers/gpu/drm/radeon/radeon_test.c index 12e8099a0823..3a13e0d1055c 100644 --- a/drivers/gpu/drm/radeon/radeon_test.c +++ b/drivers/gpu/drm/radeon/radeon_test.c @@ -257,20 +257,36 @@ static int radeon_test_create_and_emit_fence(struct radeon_device *rdev, struct radeon_ring *ring, struct radeon_fence **fence) { + uint32_t handle = ring->idx ^ 0xdeafbeef; int r; if (ring->idx == R600_RING_TYPE_UVD_INDEX) { - r = radeon_uvd_get_create_msg(rdev, ring->idx, 1, NULL); + r = radeon_uvd_get_create_msg(rdev, ring->idx, handle, NULL); if (r) { DRM_ERROR("Failed to get dummy create msg\n"); return r; } - r = radeon_uvd_get_destroy_msg(rdev, ring->idx, 1, fence); + r = radeon_uvd_get_destroy_msg(rdev, ring->idx, handle, fence); if (r) { DRM_ERROR("Failed to get dummy destroy msg\n"); return r; } + + } else if (ring->idx == TN_RING_TYPE_VCE1_INDEX || + ring->idx == TN_RING_TYPE_VCE2_INDEX) { + r = radeon_vce_get_create_msg(rdev, ring->idx, handle, NULL); + if (r) { + DRM_ERROR("Failed to get dummy create msg\n"); + return r; + } + + r = radeon_vce_get_destroy_msg(rdev, ring->idx, handle, fence); + if (r) { + DRM_ERROR("Failed to get dummy destroy msg\n"); + return r; + } + } else { r = radeon_ring_lock(rdev, ring, 64); if (r) { @@ -486,6 +502,16 @@ out_cleanup: printk(KERN_WARNING "Error while testing ring sync (%d).\n", r); } +static bool radeon_test_sync_possible(struct radeon_ring *ringA, + struct radeon_ring *ringB) +{ + if (ringA->idx == TN_RING_TYPE_VCE2_INDEX && + ringB->idx == TN_RING_TYPE_VCE1_INDEX) + return false; + + return true; +} + void radeon_test_syncing(struct radeon_device *rdev) { int i, j, k; @@ -500,6 +526,9 @@ void radeon_test_syncing(struct radeon_device *rdev) if (!ringB->ready) continue; + if (!radeon_test_sync_possible(ringA, ringB)) + continue; + DRM_INFO("Testing syncing between rings %d and %d...\n", i, j); radeon_test_ring_sync(rdev, ringA, ringB); @@ -511,6 +540,12 @@ void radeon_test_syncing(struct radeon_device *rdev) if (!ringC->ready) continue; + if (!radeon_test_sync_possible(ringA, ringC)) + continue; + + if (!radeon_test_sync_possible(ringB, ringC)) + continue; + DRM_INFO("Testing syncing between rings %d, %d and %d...\n", i, j, k); radeon_test_ring_sync2(rdev, ringA, ringB, ringC); diff --git a/drivers/gpu/drm/radeon/radeon_vce.c b/drivers/gpu/drm/radeon/radeon_vce.c new file mode 100644 index 000000000000..d130432e313a --- /dev/null +++ b/drivers/gpu/drm/radeon/radeon_vce.c @@ -0,0 +1,694 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * 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, sub license, 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 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 NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * Authors: Christian König <christian.koenig@amd.com> + */ + +#include <linux/firmware.h> +#include <linux/module.h> +#include <drm/drmP.h> +#include <drm/drm.h> + +#include "radeon.h" +#include "radeon_asic.h" +#include "sid.h" + +/* 1 second timeout */ +#define VCE_IDLE_TIMEOUT_MS 1000 + +/* Firmware Names */ +#define FIRMWARE_BONAIRE "radeon/BONAIRE_vce.bin" + +MODULE_FIRMWARE(FIRMWARE_BONAIRE); + +static void radeon_vce_idle_work_handler(struct work_struct *work); + +/** + * radeon_vce_init - allocate memory, load vce firmware + * + * @rdev: radeon_device pointer + * + * First step to get VCE online, allocate memory and load the firmware + */ +int radeon_vce_init(struct radeon_device *rdev) +{ + static const char *fw_version = "[ATI LIB=VCEFW,"; + static const char *fb_version = "[ATI LIB=VCEFWSTATS,"; + unsigned long size; + const char *fw_name, *c; + uint8_t start, mid, end; + int i, r; + + INIT_DELAYED_WORK(&rdev->vce.idle_work, radeon_vce_idle_work_handler); + + switch (rdev->family) { + case CHIP_BONAIRE: + case CHIP_KAVERI: + case CHIP_KABINI: + fw_name = FIRMWARE_BONAIRE; + break; + + default: + return -EINVAL; + } + + r = request_firmware(&rdev->vce_fw, fw_name, rdev->dev); + if (r) { + dev_err(rdev->dev, "radeon_vce: Can't load firmware \"%s\"\n", + fw_name); + return r; + } + + /* search for firmware version */ + + size = rdev->vce_fw->size - strlen(fw_version) - 9; + c = rdev->vce_fw->data; + for (;size > 0; --size, ++c) + if (strncmp(c, fw_version, strlen(fw_version)) == 0) + break; + + if (size == 0) + return -EINVAL; + + c += strlen(fw_version); + if (sscanf(c, "%2hhd.%2hhd.%2hhd]", &start, &mid, &end) != 3) + return -EINVAL; + + /* search for feedback version */ + + size = rdev->vce_fw->size - strlen(fb_version) - 3; + c = rdev->vce_fw->data; + for (;size > 0; --size, ++c) + if (strncmp(c, fb_version, strlen(fb_version)) == 0) + break; + + if (size == 0) + return -EINVAL; + + c += strlen(fb_version); + if (sscanf(c, "%2u]", &rdev->vce.fb_version) != 1) + return -EINVAL; + + DRM_INFO("Found VCE firmware/feedback version %hhd.%hhd.%hhd / %d!\n", + start, mid, end, rdev->vce.fb_version); + + rdev->vce.fw_version = (start << 24) | (mid << 16) | (end << 8); + + /* we can only work with this fw version for now */ + if (rdev->vce.fw_version != ((40 << 24) | (2 << 16) | (2 << 8))) + return -EINVAL; + + /* load firmware into VRAM */ + + size = RADEON_GPU_PAGE_ALIGN(rdev->vce_fw->size) + + RADEON_VCE_STACK_SIZE + RADEON_VCE_HEAP_SIZE; + r = radeon_bo_create(rdev, size, PAGE_SIZE, true, + RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->vce.vcpu_bo); + if (r) { + dev_err(rdev->dev, "(%d) failed to allocate VCE bo\n", r); + return r; + } + + r = radeon_vce_resume(rdev); + if (r) + return r; + + memset(rdev->vce.cpu_addr, 0, size); + memcpy(rdev->vce.cpu_addr, rdev->vce_fw->data, rdev->vce_fw->size); + + r = radeon_vce_suspend(rdev); + if (r) + return r; + + for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) { + atomic_set(&rdev->vce.handles[i], 0); + rdev->vce.filp[i] = NULL; + } + + return 0; +} + +/** + * radeon_vce_fini - free memory + * + * @rdev: radeon_device pointer + * + * Last step on VCE teardown, free firmware memory + */ +void radeon_vce_fini(struct radeon_device *rdev) +{ + radeon_vce_suspend(rdev); + radeon_bo_unref(&rdev->vce.vcpu_bo); +} + +/** + * radeon_vce_suspend - unpin VCE fw memory + * + * @rdev: radeon_device pointer + * + * TODO: Test VCE suspend/resume + */ +int radeon_vce_suspend(struct radeon_device *rdev) +{ + int r; + + if (rdev->vce.vcpu_bo == NULL) + return 0; + + r = radeon_bo_reserve(rdev->vce.vcpu_bo, false); + if (!r) { + radeon_bo_kunmap(rdev->vce.vcpu_bo); + radeon_bo_unpin(rdev->vce.vcpu_bo); + radeon_bo_unreserve(rdev->vce.vcpu_bo); + } + return r; +} + +/** + * radeon_vce_resume - pin VCE fw memory + * + * @rdev: radeon_device pointer + * + * TODO: Test VCE suspend/resume + */ +int radeon_vce_resume(struct radeon_device *rdev) +{ + int r; + + if (rdev->vce.vcpu_bo == NULL) + return -EINVAL; + + r = radeon_bo_reserve(rdev->vce.vcpu_bo, false); + if (r) { + radeon_bo_unref(&rdev->vce.vcpu_bo); + dev_err(rdev->dev, "(%d) failed to reserve VCE bo\n", r); + return r; + } + + r = radeon_bo_pin(rdev->vce.vcpu_bo, RADEON_GEM_DOMAIN_VRAM, + &rdev->vce.gpu_addr); + if (r) { + radeon_bo_unreserve(rdev->vce.vcpu_bo); + radeon_bo_unref(&rdev->vce.vcpu_bo); + dev_err(rdev->dev, "(%d) VCE bo pin failed\n", r); + return r; + } + + r = radeon_bo_kmap(rdev->vce.vcpu_bo, &rdev->vce.cpu_addr); + if (r) { + dev_err(rdev->dev, "(%d) VCE map failed\n", r); + return r; + } + + radeon_bo_unreserve(rdev->vce.vcpu_bo); + + return 0; +} + +/** + * radeon_vce_idle_work_handler - power off VCE + * + * @work: pointer to work structure + * + * power of VCE when it's not used any more + */ +static void radeon_vce_idle_work_handler(struct work_struct *work) +{ + struct radeon_device *rdev = + container_of(work, struct radeon_device, vce.idle_work.work); + + if ((radeon_fence_count_emitted(rdev, TN_RING_TYPE_VCE1_INDEX) == 0) && + (radeon_fence_count_emitted(rdev, TN_RING_TYPE_VCE2_INDEX) == 0)) { + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + radeon_dpm_enable_vce(rdev, false); + } else { + radeon_set_vce_clocks(rdev, 0, 0); + } + } else { + schedule_delayed_work(&rdev->vce.idle_work, + msecs_to_jiffies(VCE_IDLE_TIMEOUT_MS)); + } +} + +/** + * radeon_vce_note_usage - power up VCE + * + * @rdev: radeon_device pointer + * + * Make sure VCE is powerd up when we want to use it + */ +void radeon_vce_note_usage(struct radeon_device *rdev) +{ + bool streams_changed = false; + bool set_clocks = !cancel_delayed_work_sync(&rdev->vce.idle_work); + set_clocks &= schedule_delayed_work(&rdev->vce.idle_work, + msecs_to_jiffies(VCE_IDLE_TIMEOUT_MS)); + + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + /* XXX figure out if the streams changed */ + streams_changed = false; + } + + if (set_clocks || streams_changed) { + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + radeon_dpm_enable_vce(rdev, true); + } else { + radeon_set_vce_clocks(rdev, 53300, 40000); + } + } +} + +/** + * radeon_vce_free_handles - free still open VCE handles + * + * @rdev: radeon_device pointer + * @filp: drm file pointer + * + * Close all VCE handles still open by this file pointer + */ +void radeon_vce_free_handles(struct radeon_device *rdev, struct drm_file *filp) +{ + int i, r; + for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) { + uint32_t handle = atomic_read(&rdev->vce.handles[i]); + if (!handle || rdev->vce.filp[i] != filp) + continue; + + radeon_vce_note_usage(rdev); + + r = radeon_vce_get_destroy_msg(rdev, TN_RING_TYPE_VCE1_INDEX, + handle, NULL); + if (r) + DRM_ERROR("Error destroying VCE handle (%d)!\n", r); + + rdev->vce.filp[i] = NULL; + atomic_set(&rdev->vce.handles[i], 0); + } +} + +/** + * radeon_vce_get_create_msg - generate a VCE create msg + * + * @rdev: radeon_device pointer + * @ring: ring we should submit the msg to + * @handle: VCE session handle to use + * @fence: optional fence to return + * + * Open up a stream for HW test + */ +int radeon_vce_get_create_msg(struct radeon_device *rdev, int ring, + uint32_t handle, struct radeon_fence **fence) +{ + const unsigned ib_size_dw = 1024; + struct radeon_ib ib; + uint64_t dummy; + int i, r; + + r = radeon_ib_get(rdev, ring, &ib, NULL, ib_size_dw * 4); + if (r) { + DRM_ERROR("radeon: failed to get ib (%d).\n", r); + return r; + } + + dummy = ib.gpu_addr + 1024; + + /* stitch together an VCE create msg */ + ib.length_dw = 0; + ib.ptr[ib.length_dw++] = 0x0000000c; /* len */ + ib.ptr[ib.length_dw++] = 0x00000001; /* session cmd */ + ib.ptr[ib.length_dw++] = handle; + + ib.ptr[ib.length_dw++] = 0x00000030; /* len */ + ib.ptr[ib.length_dw++] = 0x01000001; /* create cmd */ + ib.ptr[ib.length_dw++] = 0x00000000; + ib.ptr[ib.length_dw++] = 0x00000042; + ib.ptr[ib.length_dw++] = 0x0000000a; + ib.ptr[ib.length_dw++] = 0x00000001; + ib.ptr[ib.length_dw++] = 0x00000080; + ib.ptr[ib.length_dw++] = 0x00000060; + ib.ptr[ib.length_dw++] = 0x00000100; + ib.ptr[ib.length_dw++] = 0x00000100; + ib.ptr[ib.length_dw++] = 0x0000000c; + ib.ptr[ib.length_dw++] = 0x00000000; + + ib.ptr[ib.length_dw++] = 0x00000014; /* len */ + ib.ptr[ib.length_dw++] = 0x05000005; /* feedback buffer */ + ib.ptr[ib.length_dw++] = upper_32_bits(dummy); + ib.ptr[ib.length_dw++] = dummy; + ib.ptr[ib.length_dw++] = 0x00000001; + + for (i = ib.length_dw; i < ib_size_dw; ++i) + ib.ptr[i] = 0x0; + + r = radeon_ib_schedule(rdev, &ib, NULL); + if (r) { + DRM_ERROR("radeon: failed to schedule ib (%d).\n", r); + } + + if (fence) + *fence = radeon_fence_ref(ib.fence); + + radeon_ib_free(rdev, &ib); + + return r; +} + +/** + * radeon_vce_get_destroy_msg - generate a VCE destroy msg + * + * @rdev: radeon_device pointer + * @ring: ring we should submit the msg to + * @handle: VCE session handle to use + * @fence: optional fence to return + * + * Close up a stream for HW test or if userspace failed to do so + */ +int radeon_vce_get_destroy_msg(struct radeon_device *rdev, int ring, + uint32_t handle, struct radeon_fence **fence) +{ + const unsigned ib_size_dw = 1024; + struct radeon_ib ib; + uint64_t dummy; + int i, r; + + r = radeon_ib_get(rdev, ring, &ib, NULL, ib_size_dw * 4); + if (r) { + DRM_ERROR("radeon: failed to get ib (%d).\n", r); + return r; + } + + dummy = ib.gpu_addr + 1024; + + /* stitch together an VCE destroy msg */ + ib.length_dw = 0; + ib.ptr[ib.length_dw++] = 0x0000000c; /* len */ + ib.ptr[ib.length_dw++] = 0x00000001; /* session cmd */ + ib.ptr[ib.length_dw++] = handle; + + ib.ptr[ib.length_dw++] = 0x00000014; /* len */ + ib.ptr[ib.length_dw++] = 0x05000005; /* feedback buffer */ + ib.ptr[ib.length_dw++] = upper_32_bits(dummy); + ib.ptr[ib.length_dw++] = dummy; + ib.ptr[ib.length_dw++] = 0x00000001; + + ib.ptr[ib.length_dw++] = 0x00000008; /* len */ + ib.ptr[ib.length_dw++] = 0x02000001; /* destroy cmd */ + + for (i = ib.length_dw; i < ib_size_dw; ++i) + ib.ptr[i] = 0x0; + + r = radeon_ib_schedule(rdev, &ib, NULL); + if (r) { + DRM_ERROR("radeon: failed to schedule ib (%d).\n", r); + } + + if (fence) + *fence = radeon_fence_ref(ib.fence); + + radeon_ib_free(rdev, &ib); + + return r; +} + +/** + * radeon_vce_cs_reloc - command submission relocation + * + * @p: parser context + * @lo: address of lower dword + * @hi: address of higher dword + * + * Patch relocation inside command stream with real buffer address + */ +int radeon_vce_cs_reloc(struct radeon_cs_parser *p, int lo, int hi) +{ + struct radeon_cs_chunk *relocs_chunk; + uint64_t offset; + unsigned idx; + + relocs_chunk = &p->chunks[p->chunk_relocs_idx]; + offset = radeon_get_ib_value(p, lo); + idx = radeon_get_ib_value(p, hi); + + if (idx >= relocs_chunk->length_dw) { + DRM_ERROR("Relocs at %d after relocations chunk end %d !\n", + idx, relocs_chunk->length_dw); + return -EINVAL; + } + + offset += p->relocs_ptr[(idx / 4)]->lobj.gpu_offset; + + p->ib.ptr[lo] = offset & 0xFFFFFFFF; + p->ib.ptr[hi] = offset >> 32; + + return 0; +} + +/** + * radeon_vce_cs_parse - parse and validate the command stream + * + * @p: parser context + * + */ +int radeon_vce_cs_parse(struct radeon_cs_parser *p) +{ + uint32_t handle = 0; + bool destroy = false; + int i, r; + + while (p->idx < p->chunks[p->chunk_ib_idx].length_dw) { + uint32_t len = radeon_get_ib_value(p, p->idx); + uint32_t cmd = radeon_get_ib_value(p, p->idx + 1); + + if ((len < 8) || (len & 3)) { + DRM_ERROR("invalid VCE command length (%d)!\n", len); + return -EINVAL; + } + + switch (cmd) { + case 0x00000001: // session + handle = radeon_get_ib_value(p, p->idx + 2); + break; + + case 0x00000002: // task info + case 0x01000001: // create + case 0x04000001: // config extension + case 0x04000002: // pic control + case 0x04000005: // rate control + case 0x04000007: // motion estimation + case 0x04000008: // rdo + break; + + case 0x03000001: // encode + r = radeon_vce_cs_reloc(p, p->idx + 10, p->idx + 9); + if (r) + return r; + + r = radeon_vce_cs_reloc(p, p->idx + 12, p->idx + 11); + if (r) + return r; + break; + + case 0x02000001: // destroy + destroy = true; + break; + + case 0x05000001: // context buffer + case 0x05000004: // video bitstream buffer + case 0x05000005: // feedback buffer + r = radeon_vce_cs_reloc(p, p->idx + 3, p->idx + 2); + if (r) + return r; + break; + + default: + DRM_ERROR("invalid VCE command (0x%x)!\n", cmd); + return -EINVAL; + } + + p->idx += len / 4; + } + + if (destroy) { + /* IB contains a destroy msg, free the handle */ + for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) + atomic_cmpxchg(&p->rdev->vce.handles[i], handle, 0); + + return 0; + } + + /* create or encode, validate the handle */ + for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) { + if (atomic_read(&p->rdev->vce.handles[i]) == handle) + return 0; + } + + /* handle not found try to alloc a new one */ + for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) { + if (!atomic_cmpxchg(&p->rdev->vce.handles[i], 0, handle)) { + p->rdev->vce.filp[i] = p->filp; + return 0; + } + } + + DRM_ERROR("No more free VCE handles!\n"); + return -EINVAL; +} + +/** + * radeon_vce_semaphore_emit - emit a semaphore command + * + * @rdev: radeon_device pointer + * @ring: engine to use + * @semaphore: address of semaphore + * @emit_wait: true=emit wait, false=emit signal + * + */ +bool radeon_vce_semaphore_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait) +{ + uint64_t addr = semaphore->gpu_addr; + + radeon_ring_write(ring, VCE_CMD_SEMAPHORE); + radeon_ring_write(ring, (addr >> 3) & 0x000FFFFF); + radeon_ring_write(ring, (addr >> 23) & 0x000FFFFF); + radeon_ring_write(ring, 0x01003000 | (emit_wait ? 1 : 0)); + if (!emit_wait) + radeon_ring_write(ring, VCE_CMD_END); + + return true; +} + +/** + * radeon_vce_ib_execute - execute indirect buffer + * + * @rdev: radeon_device pointer + * @ib: the IB to execute + * + */ +void radeon_vce_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) +{ + struct radeon_ring *ring = &rdev->ring[ib->ring]; + radeon_ring_write(ring, VCE_CMD_IB); + radeon_ring_write(ring, ib->gpu_addr); + radeon_ring_write(ring, upper_32_bits(ib->gpu_addr)); + radeon_ring_write(ring, ib->length_dw); +} + +/** + * radeon_vce_fence_emit - add a fence command to the ring + * + * @rdev: radeon_device pointer + * @fence: the fence + * + */ +void radeon_vce_fence_emit(struct radeon_device *rdev, + struct radeon_fence *fence) +{ + struct radeon_ring *ring = &rdev->ring[fence->ring]; + uint32_t addr = rdev->fence_drv[fence->ring].gpu_addr; + + radeon_ring_write(ring, VCE_CMD_FENCE); + radeon_ring_write(ring, addr); + radeon_ring_write(ring, upper_32_bits(addr)); + radeon_ring_write(ring, fence->seq); + radeon_ring_write(ring, VCE_CMD_TRAP); + radeon_ring_write(ring, VCE_CMD_END); +} + +/** + * radeon_vce_ring_test - test if VCE ring is working + * + * @rdev: radeon_device pointer + * @ring: the engine to test on + * + */ +int radeon_vce_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) +{ + uint32_t rptr = vce_v1_0_get_rptr(rdev, ring); + unsigned i; + int r; + + r = radeon_ring_lock(rdev, ring, 16); + if (r) { + DRM_ERROR("radeon: vce failed to lock ring %d (%d).\n", + ring->idx, r); + return r; + } + radeon_ring_write(ring, VCE_CMD_END); + radeon_ring_unlock_commit(rdev, ring); + + for (i = 0; i < rdev->usec_timeout; i++) { + if (vce_v1_0_get_rptr(rdev, ring) != rptr) + break; + DRM_UDELAY(1); + } + + if (i < rdev->usec_timeout) { + DRM_INFO("ring test on %d succeeded in %d usecs\n", + ring->idx, i); + } else { + DRM_ERROR("radeon: ring %d test failed\n", + ring->idx); + r = -ETIMEDOUT; + } + + return r; +} + +/** + * radeon_vce_ib_test - test if VCE IBs are working + * + * @rdev: radeon_device pointer + * @ring: the engine to test on + * + */ +int radeon_vce_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) +{ + struct radeon_fence *fence = NULL; + int r; + + r = radeon_vce_get_create_msg(rdev, ring->idx, 1, NULL); + if (r) { + DRM_ERROR("radeon: failed to get create msg (%d).\n", r); + goto error; + } + + r = radeon_vce_get_destroy_msg(rdev, ring->idx, 1, &fence); + if (r) { + DRM_ERROR("radeon: failed to get destroy ib (%d).\n", r); + goto error; + } + + r = radeon_fence_wait(fence, false); + if (r) { + DRM_ERROR("radeon: fence wait failed (%d).\n", r); + } else { + DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + } +error: + radeon_fence_unref(&fence); + return r; +} diff --git a/drivers/gpu/drm/radeon/rs780_dpm.c b/drivers/gpu/drm/radeon/rs780_dpm.c index 8512085b0aef..02f7710de470 100644 --- a/drivers/gpu/drm/radeon/rs780_dpm.c +++ b/drivers/gpu/drm/radeon/rs780_dpm.c @@ -807,9 +807,6 @@ static int rs780_parse_power_table(struct radeon_device *rdev) power_info->pplib.ucNumStates, GFP_KERNEL); if (!rdev->pm.dpm.ps) return -ENOMEM; - rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); - rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); - rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); for (i = 0; i < power_info->pplib.ucNumStates; i++) { power_state = (union pplib_power_state *) @@ -859,6 +856,10 @@ int rs780_dpm_init(struct radeon_device *rdev) return -ENOMEM; rdev->pm.dpm.priv = pi; + ret = r600_get_platform_caps(rdev); + if (ret) + return ret; + ret = rs780_parse_power_table(rdev); if (ret) return ret; diff --git a/drivers/gpu/drm/radeon/rv6xx_dpm.c b/drivers/gpu/drm/radeon/rv6xx_dpm.c index bebf31c4d841..e7045b085715 100644 --- a/drivers/gpu/drm/radeon/rv6xx_dpm.c +++ b/drivers/gpu/drm/radeon/rv6xx_dpm.c @@ -1891,9 +1891,6 @@ static int rv6xx_parse_power_table(struct radeon_device *rdev) power_info->pplib.ucNumStates, GFP_KERNEL); if (!rdev->pm.dpm.ps) return -ENOMEM; - rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); - rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); - rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); for (i = 0; i < power_info->pplib.ucNumStates; i++) { power_state = (union pplib_power_state *) @@ -1943,6 +1940,10 @@ int rv6xx_dpm_init(struct radeon_device *rdev) return -ENOMEM; rdev->pm.dpm.priv = pi; + ret = r600_get_platform_caps(rdev); + if (ret) + return ret; + ret = rv6xx_parse_power_table(rdev); if (ret) return ret; diff --git a/drivers/gpu/drm/radeon/rv770_dpm.c b/drivers/gpu/drm/radeon/rv770_dpm.c index b5f63f5e22a3..da041a43d82e 100644 --- a/drivers/gpu/drm/radeon/rv770_dpm.c +++ b/drivers/gpu/drm/radeon/rv770_dpm.c @@ -2281,9 +2281,6 @@ int rv7xx_parse_power_table(struct radeon_device *rdev) power_info->pplib.ucNumStates, GFP_KERNEL); if (!rdev->pm.dpm.ps) return -ENOMEM; - rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); - rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); - rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); for (i = 0; i < power_info->pplib.ucNumStates; i++) { power_state = (union pplib_power_state *) @@ -2361,6 +2358,10 @@ int rv770_dpm_init(struct radeon_device *rdev) pi->min_vddc_in_table = 0; pi->max_vddc_in_table = 0; + ret = r600_get_platform_caps(rdev); + if (ret) + return ret; + ret = rv7xx_parse_power_table(rdev); if (ret) return ret; diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 83578324e5d1..8008cb8d5324 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -3434,8 +3434,6 @@ static int si_cp_resume(struct radeon_device *rdev) WREG32(CP_RB0_BASE, ring->gpu_addr >> 8); - ring->rptr = RREG32(CP_RB0_RPTR); - /* ring1 - compute only */ /* Set ring buffer size */ ring = &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX]; @@ -3460,8 +3458,6 @@ static int si_cp_resume(struct radeon_device *rdev) WREG32(CP_RB1_BASE, ring->gpu_addr >> 8); - ring->rptr = RREG32(CP_RB1_RPTR); - /* ring2 - compute only */ /* Set ring buffer size */ ring = &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX]; @@ -3486,8 +3482,6 @@ static int si_cp_resume(struct radeon_device *rdev) WREG32(CP_RB2_BASE, ring->gpu_addr >> 8); - ring->rptr = RREG32(CP_RB2_RPTR); - /* start the rings */ si_cp_start(rdev); rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = true; @@ -3872,11 +3866,9 @@ bool si_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) if (!(reset_mask & (RADEON_RESET_GFX | RADEON_RESET_COMPUTE | RADEON_RESET_CP))) { - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - /* force CP activities */ - radeon_ring_force_activity(rdev, ring); return radeon_ring_test_lockup(rdev, ring); } diff --git a/drivers/gpu/drm/radeon/si_dma.c b/drivers/gpu/drm/radeon/si_dma.c index 59be2cfcbb47..cf0fdad8c278 100644 --- a/drivers/gpu/drm/radeon/si_dma.c +++ b/drivers/gpu/drm/radeon/si_dma.c @@ -49,11 +49,9 @@ bool si_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) mask = RADEON_RESET_DMA1; if (!(reset_mask & mask)) { - radeon_ring_lockup_update(ring); + radeon_ring_lockup_update(rdev, ring); return false; } - /* force ring activities */ - radeon_ring_force_activity(rdev, ring); return radeon_ring_test_lockup(rdev, ring); } diff --git a/drivers/gpu/drm/radeon/si_dpm.c b/drivers/gpu/drm/radeon/si_dpm.c index 0a2f5b4bca43..9a3567bedaae 100644 --- a/drivers/gpu/drm/radeon/si_dpm.c +++ b/drivers/gpu/drm/radeon/si_dpm.c @@ -6271,9 +6271,6 @@ static int si_parse_power_table(struct radeon_device *rdev) if (!rdev->pm.dpm.ps) return -ENOMEM; power_state_offset = (u8 *)state_array->states; - rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); - rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); - rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); for (i = 0; i < state_array->ucNumEntries; i++) { u8 *idx; power_state = (union pplib_power_state *)power_state_offset; @@ -6350,6 +6347,10 @@ int si_dpm_init(struct radeon_device *rdev) pi->min_vddc_in_table = 0; pi->max_vddc_in_table = 0; + ret = r600_get_platform_caps(rdev); + if (ret) + return ret; + ret = si_parse_power_table(rdev); if (ret) return ret; diff --git a/drivers/gpu/drm/radeon/sid.h b/drivers/gpu/drm/radeon/sid.h index 9239a6d29128..683532f84931 100644 --- a/drivers/gpu/drm/radeon/sid.h +++ b/drivers/gpu/drm/radeon/sid.h @@ -1798,4 +1798,51 @@ #define DMA_PACKET_CONSTANT_FILL 0xd #define DMA_PACKET_NOP 0xf +#define VCE_STATUS 0x20004 +#define VCE_VCPU_CNTL 0x20014 +#define VCE_CLK_EN (1 << 0) +#define VCE_VCPU_CACHE_OFFSET0 0x20024 +#define VCE_VCPU_CACHE_SIZE0 0x20028 +#define VCE_VCPU_CACHE_OFFSET1 0x2002c +#define VCE_VCPU_CACHE_SIZE1 0x20030 +#define VCE_VCPU_CACHE_OFFSET2 0x20034 +#define VCE_VCPU_CACHE_SIZE2 0x20038 +#define VCE_SOFT_RESET 0x20120 +#define VCE_ECPU_SOFT_RESET (1 << 0) +#define VCE_FME_SOFT_RESET (1 << 2) +#define VCE_RB_BASE_LO2 0x2016c +#define VCE_RB_BASE_HI2 0x20170 +#define VCE_RB_SIZE2 0x20174 +#define VCE_RB_RPTR2 0x20178 +#define VCE_RB_WPTR2 0x2017c +#define VCE_RB_BASE_LO 0x20180 +#define VCE_RB_BASE_HI 0x20184 +#define VCE_RB_SIZE 0x20188 +#define VCE_RB_RPTR 0x2018c +#define VCE_RB_WPTR 0x20190 +#define VCE_CLOCK_GATING_A 0x202f8 +#define VCE_CLOCK_GATING_B 0x202fc +#define VCE_UENC_CLOCK_GATING 0x205bc +#define VCE_UENC_REG_CLOCK_GATING 0x205c0 +#define VCE_FW_REG_STATUS 0x20e10 +# define VCE_FW_REG_STATUS_BUSY (1 << 0) +# define VCE_FW_REG_STATUS_PASS (1 << 3) +# define VCE_FW_REG_STATUS_DONE (1 << 11) +#define VCE_LMI_FW_START_KEYSEL 0x20e18 +#define VCE_LMI_FW_PERIODIC_CTRL 0x20e20 +#define VCE_LMI_CTRL2 0x20e74 +#define VCE_LMI_CTRL 0x20e98 +#define VCE_LMI_VM_CTRL 0x20ea0 +#define VCE_LMI_SWAP_CNTL 0x20eb4 +#define VCE_LMI_SWAP_CNTL1 0x20eb8 +#define VCE_LMI_CACHE_CTRL 0x20ef4 + +#define VCE_CMD_NO_OP 0x00000000 +#define VCE_CMD_END 0x00000001 +#define VCE_CMD_IB 0x00000002 +#define VCE_CMD_FENCE 0x00000003 +#define VCE_CMD_TRAP 0x00000004 +#define VCE_CMD_IB_AUTO 0x00000005 +#define VCE_CMD_SEMAPHORE 0x00000006 + #endif diff --git a/drivers/gpu/drm/radeon/sumo_dpm.c b/drivers/gpu/drm/radeon/sumo_dpm.c index 8b47b3cd0357..3f0e8d7b8dbe 100644 --- a/drivers/gpu/drm/radeon/sumo_dpm.c +++ b/drivers/gpu/drm/radeon/sumo_dpm.c @@ -1484,9 +1484,6 @@ static int sumo_parse_power_table(struct radeon_device *rdev) if (!rdev->pm.dpm.ps) return -ENOMEM; power_state_offset = (u8 *)state_array->states; - rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); - rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); - rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); for (i = 0; i < state_array->ucNumEntries; i++) { u8 *idx; power_state = (union pplib_power_state *)power_state_offset; @@ -1772,6 +1769,10 @@ int sumo_dpm_init(struct radeon_device *rdev) sumo_construct_boot_and_acpi_state(rdev); + ret = r600_get_platform_caps(rdev); + if (ret) + return ret; + ret = sumo_parse_power_table(rdev); if (ret) return ret; diff --git a/drivers/gpu/drm/radeon/trinity_dpm.c b/drivers/gpu/drm/radeon/trinity_dpm.c index 2da0e17eb960..2a2822c03329 100644 --- a/drivers/gpu/drm/radeon/trinity_dpm.c +++ b/drivers/gpu/drm/radeon/trinity_dpm.c @@ -1694,9 +1694,6 @@ static int trinity_parse_power_table(struct radeon_device *rdev) if (!rdev->pm.dpm.ps) return -ENOMEM; power_state_offset = (u8 *)state_array->states; - rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); - rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); - rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); for (i = 0; i < state_array->ucNumEntries; i++) { u8 *idx; power_state = (union pplib_power_state *)power_state_offset; @@ -1895,6 +1892,10 @@ int trinity_dpm_init(struct radeon_device *rdev) trinity_construct_boot_state(rdev); + ret = r600_get_platform_caps(rdev); + if (ret) + return ret; + ret = trinity_parse_power_table(rdev); if (ret) return ret; diff --git a/drivers/gpu/drm/radeon/uvd_v1_0.c b/drivers/gpu/drm/radeon/uvd_v1_0.c index d4a68af1a279..0a243f0e5d68 100644 --- a/drivers/gpu/drm/radeon/uvd_v1_0.c +++ b/drivers/gpu/drm/radeon/uvd_v1_0.c @@ -262,7 +262,7 @@ int uvd_v1_0_start(struct radeon_device *rdev) /* Initialize the ring buffer's read and write pointers */ WREG32(UVD_RBC_RB_RPTR, 0x0); - ring->wptr = ring->rptr = RREG32(UVD_RBC_RB_RPTR); + ring->wptr = RREG32(UVD_RBC_RB_RPTR); WREG32(UVD_RBC_RB_WPTR, ring->wptr); /* set the ring address */ diff --git a/drivers/gpu/drm/radeon/vce_v1_0.c b/drivers/gpu/drm/radeon/vce_v1_0.c new file mode 100644 index 000000000000..b44d9c842f7b --- /dev/null +++ b/drivers/gpu/drm/radeon/vce_v1_0.c @@ -0,0 +1,187 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * 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, sub license, 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 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 NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * Authors: Christian König <christian.koenig@amd.com> + */ + +#include <linux/firmware.h> +#include <drm/drmP.h> +#include "radeon.h" +#include "radeon_asic.h" +#include "sid.h" + +/** + * vce_v1_0_get_rptr - get read pointer + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * + * Returns the current hardware read pointer + */ +uint32_t vce_v1_0_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + if (ring->idx == TN_RING_TYPE_VCE1_INDEX) + return RREG32(VCE_RB_RPTR); + else + return RREG32(VCE_RB_RPTR2); +} + +/** + * vce_v1_0_get_wptr - get write pointer + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * + * Returns the current hardware write pointer + */ +uint32_t vce_v1_0_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + if (ring->idx == TN_RING_TYPE_VCE1_INDEX) + return RREG32(VCE_RB_WPTR); + else + return RREG32(VCE_RB_WPTR2); +} + +/** + * vce_v1_0_set_wptr - set write pointer + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * + * Commits the write pointer to the hardware + */ +void vce_v1_0_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + if (ring->idx == TN_RING_TYPE_VCE1_INDEX) + WREG32(VCE_RB_WPTR, ring->wptr); + else + WREG32(VCE_RB_WPTR2, ring->wptr); +} + +/** + * vce_v1_0_start - start VCE block + * + * @rdev: radeon_device pointer + * + * Setup and start the VCE block + */ +int vce_v1_0_start(struct radeon_device *rdev) +{ + struct radeon_ring *ring; + int i, j, r; + + /* set BUSY flag */ + WREG32_P(VCE_STATUS, 1, ~1); + + ring = &rdev->ring[TN_RING_TYPE_VCE1_INDEX]; + WREG32(VCE_RB_RPTR, ring->wptr); + WREG32(VCE_RB_WPTR, ring->wptr); + WREG32(VCE_RB_BASE_LO, ring->gpu_addr); + WREG32(VCE_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); + WREG32(VCE_RB_SIZE, ring->ring_size / 4); + + ring = &rdev->ring[TN_RING_TYPE_VCE2_INDEX]; + WREG32(VCE_RB_RPTR2, ring->wptr); + WREG32(VCE_RB_WPTR2, ring->wptr); + WREG32(VCE_RB_BASE_LO2, ring->gpu_addr); + WREG32(VCE_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); + WREG32(VCE_RB_SIZE2, ring->ring_size / 4); + + WREG32_P(VCE_VCPU_CNTL, VCE_CLK_EN, ~VCE_CLK_EN); + + WREG32_P(VCE_SOFT_RESET, + VCE_ECPU_SOFT_RESET | + VCE_FME_SOFT_RESET, ~( + VCE_ECPU_SOFT_RESET | + VCE_FME_SOFT_RESET)); + + mdelay(100); + + WREG32_P(VCE_SOFT_RESET, 0, ~( + VCE_ECPU_SOFT_RESET | + VCE_FME_SOFT_RESET)); + + for (i = 0; i < 10; ++i) { + uint32_t status; + for (j = 0; j < 100; ++j) { + status = RREG32(VCE_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(VCE_SOFT_RESET, VCE_ECPU_SOFT_RESET, ~VCE_ECPU_SOFT_RESET); + mdelay(10); + WREG32_P(VCE_SOFT_RESET, 0, ~VCE_ECPU_SOFT_RESET); + mdelay(10); + r = -1; + } + + /* clear BUSY flag */ + WREG32_P(VCE_STATUS, 0, ~1); + + if (r) { + DRM_ERROR("VCE not responding, giving up!!!\n"); + return r; + } + + return 0; +} + +int vce_v1_0_init(struct radeon_device *rdev) +{ + struct radeon_ring *ring; + int r; + + r = vce_v1_0_start(rdev); + if (r) + return r; + + ring = &rdev->ring[TN_RING_TYPE_VCE1_INDEX]; + ring->ready = true; + r = radeon_ring_test(rdev, TN_RING_TYPE_VCE1_INDEX, ring); + if (r) { + ring->ready = false; + return r; + } + + ring = &rdev->ring[TN_RING_TYPE_VCE2_INDEX]; + ring->ready = true; + r = radeon_ring_test(rdev, TN_RING_TYPE_VCE2_INDEX, ring); + if (r) { + ring->ready = false; + return r; + } + + DRM_INFO("VCE initialized successfully.\n"); + + return 0; +} diff --git a/drivers/gpu/drm/radeon/vce_v2_0.c b/drivers/gpu/drm/radeon/vce_v2_0.c new file mode 100644 index 000000000000..1ac7bb825a1b --- /dev/null +++ b/drivers/gpu/drm/radeon/vce_v2_0.c @@ -0,0 +1,181 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * 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, sub license, 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 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 NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * Authors: Christian König <christian.koenig@amd.com> + */ + +#include <linux/firmware.h> +#include <drm/drmP.h> +#include "radeon.h" +#include "radeon_asic.h" +#include "cikd.h" + +static void vce_v2_0_set_sw_cg(struct radeon_device *rdev, bool gated) +{ + u32 tmp; + + if (gated) { + tmp = RREG32(VCE_CLOCK_GATING_B); + tmp |= 0xe70000; + WREG32(VCE_CLOCK_GATING_B, tmp); + + tmp = RREG32(VCE_UENC_CLOCK_GATING); + tmp |= 0xff000000; + WREG32(VCE_UENC_CLOCK_GATING, tmp); + + tmp = RREG32(VCE_UENC_REG_CLOCK_GATING); + tmp &= ~0x3fc; + WREG32(VCE_UENC_REG_CLOCK_GATING, tmp); + + WREG32(VCE_CGTT_CLK_OVERRIDE, 0); + } else { + tmp = RREG32(VCE_CLOCK_GATING_B); + tmp |= 0xe7; + tmp &= ~0xe70000; + WREG32(VCE_CLOCK_GATING_B, tmp); + + tmp = RREG32(VCE_UENC_CLOCK_GATING); + tmp |= 0x1fe000; + tmp &= ~0xff000000; + WREG32(VCE_UENC_CLOCK_GATING, tmp); + + tmp = RREG32(VCE_UENC_REG_CLOCK_GATING); + tmp |= 0x3fc; + WREG32(VCE_UENC_REG_CLOCK_GATING, tmp); + } +} + +static void vce_v2_0_set_dyn_cg(struct radeon_device *rdev, bool gated) +{ + u32 orig, tmp; + + tmp = RREG32(VCE_CLOCK_GATING_B); + tmp &= ~0x00060006; + if (gated) { + tmp |= 0xe10000; + } else { + tmp |= 0xe1; + tmp &= ~0xe10000; + } + WREG32(VCE_CLOCK_GATING_B, tmp); + + orig = tmp = RREG32(VCE_UENC_CLOCK_GATING); + tmp &= ~0x1fe000; + tmp &= ~0xff000000; + if (tmp != orig) + WREG32(VCE_UENC_CLOCK_GATING, tmp); + + orig = tmp = RREG32(VCE_UENC_REG_CLOCK_GATING); + tmp &= ~0x3fc; + if (tmp != orig) + WREG32(VCE_UENC_REG_CLOCK_GATING, tmp); + + if (gated) + WREG32(VCE_CGTT_CLK_OVERRIDE, 0); +} + +static void vce_v2_0_disable_cg(struct radeon_device *rdev) +{ + WREG32(VCE_CGTT_CLK_OVERRIDE, 7); +} + +void vce_v2_0_enable_mgcg(struct radeon_device *rdev, bool enable) +{ + bool sw_cg = false; + + if (enable && (rdev->cg_flags & RADEON_CG_SUPPORT_VCE_MGCG)) { + if (sw_cg) + vce_v2_0_set_sw_cg(rdev, true); + else + vce_v2_0_set_dyn_cg(rdev, true); + } else { + vce_v2_0_disable_cg(rdev); + + if (sw_cg) + vce_v2_0_set_sw_cg(rdev, false); + else + vce_v2_0_set_dyn_cg(rdev, false); + } +} + +static void vce_v2_0_init_cg(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32(VCE_CLOCK_GATING_A); + tmp &= ~(CGC_CLK_GATE_DLY_TIMER_MASK | CGC_CLK_GATER_OFF_DLY_TIMER_MASK); + tmp |= (CGC_CLK_GATE_DLY_TIMER(0) | CGC_CLK_GATER_OFF_DLY_TIMER(4)); + tmp |= CGC_UENC_WAIT_AWAKE; + WREG32(VCE_CLOCK_GATING_A, tmp); + + tmp = RREG32(VCE_UENC_CLOCK_GATING); + tmp &= ~(CLOCK_ON_DELAY_MASK | CLOCK_OFF_DELAY_MASK); + tmp |= (CLOCK_ON_DELAY(0) | CLOCK_OFF_DELAY(4)); + WREG32(VCE_UENC_CLOCK_GATING, tmp); + + tmp = RREG32(VCE_CLOCK_GATING_B); + tmp |= 0x10; + tmp &= ~0x100000; + WREG32(VCE_CLOCK_GATING_B, tmp); +} + +int vce_v2_0_resume(struct radeon_device *rdev) +{ + uint64_t addr = rdev->vce.gpu_addr; + uint32_t size; + + WREG32_P(VCE_CLOCK_GATING_A, 0, ~(1 << 16)); + WREG32_P(VCE_UENC_CLOCK_GATING, 0x1FF000, ~0xFF9FF000); + WREG32_P(VCE_UENC_REG_CLOCK_GATING, 0x3F, ~0x3F); + WREG32(VCE_CLOCK_GATING_B, 0xf7); + + WREG32(VCE_LMI_CTRL, 0x00398000); + WREG32_P(VCE_LMI_CACHE_CTRL, 0x0, ~0x1); + WREG32(VCE_LMI_SWAP_CNTL, 0); + WREG32(VCE_LMI_SWAP_CNTL1, 0); + WREG32(VCE_LMI_VM_CTRL, 0); + + size = RADEON_GPU_PAGE_ALIGN(rdev->vce_fw->size); + WREG32(VCE_VCPU_CACHE_OFFSET0, addr & 0x7fffffff); + WREG32(VCE_VCPU_CACHE_SIZE0, size); + + addr += size; + size = RADEON_VCE_STACK_SIZE; + WREG32(VCE_VCPU_CACHE_OFFSET1, addr & 0x7fffffff); + WREG32(VCE_VCPU_CACHE_SIZE1, size); + + addr += size; + size = RADEON_VCE_HEAP_SIZE; + WREG32(VCE_VCPU_CACHE_OFFSET2, addr & 0x7fffffff); + WREG32(VCE_VCPU_CACHE_SIZE2, size); + + WREG32_P(VCE_LMI_CTRL2, 0x0, ~0x100); + + WREG32_P(VCE_SYS_INT_EN, VCE_SYS_INT_TRAP_INTERRUPT_EN, + ~VCE_SYS_INT_TRAP_INTERRUPT_EN); + + vce_v2_0_init_cg(rdev); + + return 0; +} |