Merge drm/drm-next into drm-misc-next-fixes

Backmerging to bring drm-misc-next-fixes up to the latest state for
the current release cycle.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>

1 files changed, 57 insertions, 96 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 3003ee1c9487..46cf48b3904a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -31,6 +31,8 @@
 #include <linux/dma-buf.h>
 #include "amdgpu_xgmi.h"
 #include <uapi/linux/kfd_ioctl.h>
+#include "amdgpu_ras.h"
+#include "amdgpu_umc.h"
 
 /* Total memory size in system memory and all GPU VRAM. Used to
  * estimate worst case amount of memory to reserve for page tables
@@ -70,8 +72,7 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev)
 	if (!kfd_initialized)
 		return;
 
-	adev->kfd.dev = kgd2kfd_probe((struct kgd_dev *)adev,
-				      adev->pdev, adev->asic_type, vf);
+	adev->kfd.dev = kgd2kfd_probe(adev, vf);
 
 	if (adev->kfd.dev)
 		amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size;
@@ -192,6 +193,16 @@ void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool run_pm)
 		kgd2kfd_suspend(adev->kfd.dev, run_pm);
 }
 
+int amdgpu_amdkfd_resume_iommu(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (adev->kfd.dev)
+		r = kgd2kfd_resume_iommu(adev->kfd.dev);
+
+	return r;
+}
+
 int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool run_pm)
 {
 	int r = 0;
@@ -222,19 +233,16 @@ int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev)
 	return r;
 }
 
-void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd)
+void amdgpu_amdkfd_gpu_reset(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	if (amdgpu_device_should_recover_gpu(adev))
 		amdgpu_device_gpu_recover(adev, NULL);
 }
 
-int amdgpu_amdkfd_alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
+int amdgpu_amdkfd_alloc_gtt_mem(struct amdgpu_device *adev, size_t size,
 				void **mem_obj, uint64_t *gpu_addr,
 				void **cpu_ptr, bool cp_mqd_gfx9)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct amdgpu_bo *bo = NULL;
 	struct amdgpu_bo_param bp;
 	int r;
@@ -303,7 +311,7 @@ allocate_mem_reserve_bo_failed:
 	return r;
 }
 
-void amdgpu_amdkfd_free_gtt_mem(struct kgd_dev *kgd, void *mem_obj)
+void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void *mem_obj)
 {
 	struct amdgpu_bo *bo = (struct amdgpu_bo *) mem_obj;
 
@@ -314,10 +322,9 @@ void amdgpu_amdkfd_free_gtt_mem(struct kgd_dev *kgd, void *mem_obj)
 	amdgpu_bo_unref(&(bo));
 }
 
-int amdgpu_amdkfd_alloc_gws(struct kgd_dev *kgd, size_t size,
+int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size,
 				void **mem_obj)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct amdgpu_bo *bo = NULL;
 	struct amdgpu_bo_user *ubo;
 	struct amdgpu_bo_param bp;
@@ -344,18 +351,16 @@ int amdgpu_amdkfd_alloc_gws(struct kgd_dev *kgd, size_t size,
 	return 0;
 }
 
-void amdgpu_amdkfd_free_gws(struct kgd_dev *kgd, void *mem_obj)
+void amdgpu_amdkfd_free_gws(struct amdgpu_device *adev, void *mem_obj)
 {
 	struct amdgpu_bo *bo = (struct amdgpu_bo *)mem_obj;
 
 	amdgpu_bo_unref(&bo);
 }
 
-uint32_t amdgpu_amdkfd_get_fw_version(struct kgd_dev *kgd,
+uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
 				      enum kgd_engine_type type)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	switch (type) {
 	case KGD_ENGINE_PFP:
 		return adev->gfx.pfp_fw_version;
@@ -388,11 +393,9 @@ uint32_t amdgpu_amdkfd_get_fw_version(struct kgd_dev *kgd,
 	return 0;
 }
 
-void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
+void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
 				      struct kfd_local_mem_info *mem_info)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	memset(mem_info, 0, sizeof(*mem_info));
 
 	mem_info->local_mem_size_public = adev->gmc.visible_vram_size;
@@ -417,19 +420,15 @@ void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
 		mem_info->mem_clk_max = 100;
 }
 
-uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd)
+uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	if (adev->gfx.funcs->get_gpu_clock_counter)
 		return adev->gfx.funcs->get_gpu_clock_counter(adev);
 	return 0;
 }
 
-uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
+uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	/* the sclk is in quantas of 10kHz */
 	if (amdgpu_sriov_vf(adev))
 		return adev->clock.default_sclk / 100;
@@ -439,9 +438,8 @@ uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
 		return 100;
 }
 
-void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info)
+void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev, struct kfd_cu_info *cu_info)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct amdgpu_cu_info acu_info = adev->gfx.cu_info;
 
 	memset(cu_info, 0, sizeof(*cu_info));
@@ -462,13 +460,12 @@ void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info)
 	cu_info->lds_size = acu_info.lds_size;
 }
 
-int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd,
-				  struct kgd_dev **dma_buf_kgd,
+int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
+				  struct amdgpu_device **dmabuf_adev,
 				  uint64_t *bo_size, void *metadata_buffer,
 				  size_t buffer_size, uint32_t *metadata_size,
 				  uint32_t *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct dma_buf *dma_buf;
 	struct drm_gem_object *obj;
 	struct amdgpu_bo *bo;
@@ -496,8 +493,8 @@ int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd,
 		goto out_put;
 
 	r = 0;
-	if (dma_buf_kgd)
-		*dma_buf_kgd = (struct kgd_dev *)adev;
+	if (dmabuf_adev)
+		*dmabuf_adev = adev;
 	if (bo_size)
 		*bo_size = amdgpu_bo_size(bo);
 	if (metadata_buffer)
@@ -517,32 +514,18 @@ out_put:
 	return r;
 }
 
-uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd)
+uint64_t amdgpu_amdkfd_get_vram_usage(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct ttm_resource_manager *vram_man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM);
 
 	return amdgpu_vram_mgr_usage(vram_man);
 }
 
-uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
-	return adev->gmc.xgmi.hive_id;
-}
-
-uint64_t amdgpu_amdkfd_get_unique_id(struct kgd_dev *kgd)
+uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
+					  struct amdgpu_device *src)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
-	return adev->unique_id;
-}
-
-uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *src)
-{
-	struct amdgpu_device *peer_adev = (struct amdgpu_device *)src;
-	struct amdgpu_device *adev = (struct amdgpu_device *)dst;
+	struct amdgpu_device *peer_adev = src;
+	struct amdgpu_device *adev = dst;
 	int ret = amdgpu_xgmi_get_hops_count(adev, peer_adev);
 
 	if (ret < 0) {
@@ -554,16 +537,18 @@ uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *s
 	return  (uint8_t)ret;
 }
 
-int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev *dst, struct kgd_dev *src, bool is_min)
+int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
+					    struct amdgpu_device *src,
+					    bool is_min)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)dst, *peer_adev;
+	struct amdgpu_device *adev = dst, *peer_adev;
 	int num_links;
 
 	if (adev->asic_type != CHIP_ALDEBARAN)
 		return 0;
 
 	if (src)
-		peer_adev = (struct amdgpu_device *)src;
+		peer_adev = src;
 
 	/* num links returns 0 for indirect peers since indirect route is unknown. */
 	num_links = is_min ? 1 : amdgpu_xgmi_get_num_links(adev, peer_adev);
@@ -578,9 +563,8 @@ int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev *dst, struct kgd_dev
 	return (num_links * 16 * 25000)/BITS_PER_BYTE;
 }
 
-int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool is_min)
+int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)dev;
 	int num_lanes_shift = (is_min ? ffs(adev->pm.pcie_mlw_mask) :
 							fls(adev->pm.pcie_mlw_mask)) - 1;
 	int gen_speed_shift = (is_min ? ffs(adev->pm.pcie_gen_mask &
@@ -636,39 +620,11 @@ int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool is_min)
 	return (num_lanes_factor * gen_speed_mbits_factor)/BITS_PER_BYTE;
 }
 
-uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct kgd_dev *kgd)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
-	return adev->rmmio_remap.bus_addr;
-}
-
-uint32_t amdgpu_amdkfd_get_num_gws(struct kgd_dev *kgd)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
-	return adev->gds.gws_size;
-}
-
-uint32_t amdgpu_amdkfd_get_asic_rev_id(struct kgd_dev *kgd)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
-	return adev->rev_id;
-}
-
-int amdgpu_amdkfd_get_noretry(struct kgd_dev *kgd)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
-	return adev->gmc.noretry;
-}
-
-int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine,
+int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
+				enum kgd_engine_type engine,
 				uint32_t vmid, uint64_t gpu_addr,
 				uint32_t *ib_cmd, uint32_t ib_len)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	struct amdgpu_ring *ring;
@@ -719,10 +675,8 @@ err:
 	return ret;
 }
 
-void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle)
+void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	amdgpu_dpm_switch_power_profile(adev,
 					PP_SMC_POWER_PROFILE_COMPUTE,
 					!idle);
@@ -736,10 +690,9 @@ bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid)
 	return false;
 }
 
-int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct kgd_dev *kgd, uint16_t vmid)
+int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
+				     uint16_t vmid)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	if (adev->family == AMDGPU_FAMILY_AI) {
 		int i;
 
@@ -752,10 +705,9 @@ int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct kgd_dev *kgd, uint16_t vmid)
 	return 0;
 }
 
-int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t pasid,
-				      enum TLB_FLUSH_TYPE flush_type)
+int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+				      uint16_t pasid, enum TLB_FLUSH_TYPE flush_type)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	bool all_hub = false;
 
 	if (adev->family == AMDGPU_FAMILY_AI)
@@ -764,9 +716,18 @@ int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t pasid,
 	return amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, flush_type, all_hub);
 }
 
-bool amdgpu_amdkfd_have_atomics_support(struct kgd_dev *kgd)
+bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	return adev->have_atomics_support;
 }
+
+void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev)
+{
+	struct ras_err_data err_data = {0, 0, 0, NULL};
+
+	/* CPU MCA will handle page retirement if connected_to_cpu is 1 */
+	if (!adev->gmc.xgmi.connected_to_cpu)
+		amdgpu_umc_process_ras_data_cb(adev, &err_data, NULL);
+	else
+		amdgpu_amdkfd_gpu_reset(adev);
+}