diff --git a/MAINTAINERS b/MAINTAINERS index 53b859d10de6..ae021c31158d 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -966,6 +966,7 @@ F: drivers/gpu/drm/amd/include/kgd_kfd_interface.h F: drivers/gpu/drm/amd/include/v9_structs.h F: drivers/gpu/drm/amd/include/vi_structs.h F: include/uapi/linux/kfd_ioctl.h +F: include/uapi/linux/kfd_sysfs.h AMD SPI DRIVER M: Sanjay R Mehta diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index 653726588956..7fedbb725e17 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -45,7 +45,7 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \ amdgpu_atombios.o atombios_crtc.o amdgpu_connectors.o \ atom.o amdgpu_fence.o amdgpu_ttm.o amdgpu_object.o amdgpu_gart.o \ amdgpu_encoders.o amdgpu_display.o amdgpu_i2c.o \ - amdgpu_fb.o amdgpu_gem.o amdgpu_ring.o \ + amdgpu_gem.o amdgpu_ring.o \ amdgpu_cs.o amdgpu_bios.o amdgpu_benchmark.o amdgpu_test.o \ atombios_dp.o amdgpu_afmt.o amdgpu_trace_points.o \ atombios_encoders.o amdgpu_sa.o atombios_i2c.o \ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 7077f21f0021..46cf48b3904a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -72,7 +72,7 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) if (!kfd_initialized) return; - adev->kfd.dev = kgd2kfd_probe((struct kgd_dev *)adev, vf); + adev->kfd.dev = kgd2kfd_probe(adev, vf); if (adev->kfd.dev) amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; @@ -233,19 +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; @@ -314,7 +311,7 @@ int amdgpu_amdkfd_alloc_gtt_mem(struct kgd_dev *kgd, size_t size, 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; @@ -325,10 +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; @@ -355,18 +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; @@ -399,11 +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; @@ -428,19 +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; @@ -450,9 +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)); @@ -473,13 +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; @@ -507,8 +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) @@ -528,32 +514,18 @@ int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd, 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) +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->gmc.xgmi.hive_id; -} - -uint64_t amdgpu_amdkfd_get_unique_id(struct kgd_dev *kgd) -{ - 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) { @@ -565,16 +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); @@ -589,9 +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 & @@ -647,39 +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; @@ -730,10 +675,8 @@ int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine, 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); @@ -747,10 +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; @@ -763,10 +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) @@ -775,21 +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 kgd_dev *kgd) +void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev) { - struct amdgpu_device *adev = (struct amdgpu_device *)kgd; 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(kgd); + amdgpu_amdkfd_gpu_reset(adev); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index a15a4787c7ee..fcbc8a9c9e06 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -144,14 +144,16 @@ void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev, void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev); void amdgpu_amdkfd_device_init(struct amdgpu_device *adev); void amdgpu_amdkfd_device_fini_sw(struct amdgpu_device *adev); -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); -void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle); -bool amdgpu_amdkfd_have_atomics_support(struct kgd_dev *kgd); -int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct kgd_dev *kgd, uint16_t vmid); -int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t pasid, - enum TLB_FLUSH_TYPE flush_type); +void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle); +bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev); +int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev, + uint16_t vmid); +int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev, + uint16_t pasid, enum TLB_FLUSH_TYPE flush_type); bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid); @@ -159,7 +161,7 @@ int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev); int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev); -void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd); +void amdgpu_amdkfd_gpu_reset(struct amdgpu_device *adev); int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev, int queue_bit); @@ -198,37 +200,36 @@ int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm) } #endif /* Shared API */ -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 mqd_gfx9); -void amdgpu_amdkfd_free_gtt_mem(struct kgd_dev *kgd, void *mem_obj); -int amdgpu_amdkfd_alloc_gws(struct kgd_dev *kgd, size_t size, void **mem_obj); -void amdgpu_amdkfd_free_gws(struct kgd_dev *kgd, void *mem_obj); +void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void *mem_obj); +int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size, + void **mem_obj); +void amdgpu_amdkfd_free_gws(struct amdgpu_device *adev, void *mem_obj); int amdgpu_amdkfd_add_gws_to_process(void *info, void *gws, struct kgd_mem **mem); int amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem); -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); -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); -uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd); +uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev); -uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd); -void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info); -int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd, - struct kgd_dev **dmabuf_kgd, +uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev); +void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev, + struct kfd_cu_info *cu_info); +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); -uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd); -uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd); -uint64_t amdgpu_amdkfd_get_unique_id(struct kgd_dev *kgd); -uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct kgd_dev *kgd); -uint32_t amdgpu_amdkfd_get_num_gws(struct kgd_dev *kgd); -uint32_t amdgpu_amdkfd_get_asic_rev_id(struct kgd_dev *kgd); -int amdgpu_amdkfd_get_noretry(struct kgd_dev *kgd); -uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *src); -int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev *dst, struct kgd_dev *src, bool is_min); -int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool is_min); +uint64_t amdgpu_amdkfd_get_vram_usage(struct amdgpu_device *adev); +uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst, + struct amdgpu_device *src); +int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst, + struct amdgpu_device *src, + bool is_min); +int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min); /* Read user wptr from a specified user address space with page fault * disabled. The memory must be pinned and mapped to the hardware when @@ -258,45 +259,54 @@ int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool is_min); (&((struct amdgpu_fpriv *) \ ((struct drm_file *)(drm_priv))->driver_priv)->vm) -int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, +int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev, struct file *filp, u32 pasid, void **process_info, struct dma_fence **ef); -void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *drm_priv); +void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev, + void *drm_priv); uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv); int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( - struct kgd_dev *kgd, uint64_t va, uint64_t size, + struct amdgpu_device *adev, uint64_t va, uint64_t size, void *drm_priv, struct kgd_mem **mem, uint64_t *offset, uint32_t flags); int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv, + struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv, uint64_t *size); int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv, bool *table_freed); + struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv, + bool *table_freed); int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv); + struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv); int amdgpu_amdkfd_gpuvm_sync_memory( - struct kgd_dev *kgd, struct kgd_mem *mem, bool intr); -int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd, + struct amdgpu_device *adev, struct kgd_mem *mem, bool intr); +int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct amdgpu_device *adev, struct kgd_mem *mem, void **kptr, uint64_t *size); -void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_dev *kgd, struct kgd_mem *mem); +void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct amdgpu_device *adev, + struct kgd_mem *mem); int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info, struct dma_fence **ef); -int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, +int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev, struct kfd_vm_fault_info *info); -int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd, +int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev, struct dma_buf *dmabuf, uint64_t va, void *drm_priv, struct kgd_mem **mem, uint64_t *size, uint64_t *mmap_offset); -int amdgpu_amdkfd_get_tile_config(struct kgd_dev *kgd, +int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev, struct tile_config *config); -void amdgpu_amdkfd_ras_poison_consumption_handler(struct kgd_dev *kgd); +void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev); #if IS_ENABLED(CONFIG_HSA_AMD) void amdgpu_amdkfd_gpuvm_init_mem_limits(void); void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, struct amdgpu_vm *vm); + +/** + * @amdgpu_amdkfd_release_notify() - Notify KFD when GEM object is released + * + * Allows KFD to release its resources associated with the GEM object. + */ void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo); void amdgpu_amdkfd_reserve_system_mem(uint64_t size); #else @@ -324,7 +334,7 @@ int kgd2kfd_schedule_evict_and_restore_process(struct mm_struct *mm, #if IS_ENABLED(CONFIG_HSA_AMD) int kgd2kfd_init(void); void kgd2kfd_exit(void); -struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf); +struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf); bool kgd2kfd_device_init(struct kfd_dev *kfd, struct drm_device *ddev, const struct kgd2kfd_shared_resources *gpu_resources); @@ -348,7 +358,7 @@ static inline void kgd2kfd_exit(void) } static inline -struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf) +struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf) { return NULL; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c index 5a7f680bcb3f..abe93b3ff765 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c @@ -57,11 +57,6 @@ (*dump)[i++][1] = RREG32(addr); \ } while (0) -static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) -{ - return (struct amdgpu_device *)kgd; -} - static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd) { return (struct v9_sdma_mqd *)mqd; @@ -123,10 +118,9 @@ static uint32_t get_sdma_rlc_reg_offset(struct amdgpu_device *adev, return sdma_rlc_reg_offset; } -int kgd_arcturus_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, +int kgd_arcturus_hqd_sdma_load(struct amdgpu_device *adev, void *mqd, uint32_t __user *wptr, struct mm_struct *mm) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v9_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; unsigned long end_jiffies; @@ -193,11 +187,10 @@ int kgd_arcturus_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, return 0; } -int kgd_arcturus_hqd_sdma_dump(struct kgd_dev *kgd, +int kgd_arcturus_hqd_sdma_dump(struct amdgpu_device *adev, uint32_t engine_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, engine_id, queue_id); uint32_t i = 0, reg; @@ -225,9 +218,9 @@ int kgd_arcturus_hqd_sdma_dump(struct kgd_dev *kgd, return 0; } -bool kgd_arcturus_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) +bool kgd_arcturus_hqd_sdma_is_occupied(struct amdgpu_device *adev, + void *mqd) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v9_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; uint32_t sdma_rlc_rb_cntl; @@ -244,10 +237,9 @@ bool kgd_arcturus_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) return false; } -int kgd_arcturus_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, +int kgd_arcturus_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd, unsigned int utimeout) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v9_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; uint32_t temp; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.h index ce08131b7b5f..756c1a5679c0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.h @@ -20,11 +20,12 @@ * OTHER DEALINGS IN THE SOFTWARE. */ -int kgd_arcturus_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, +int kgd_arcturus_hqd_sdma_load(struct amdgpu_device *adev, void *mqd, uint32_t __user *wptr, struct mm_struct *mm); -int kgd_arcturus_hqd_sdma_dump(struct kgd_dev *kgd, +int kgd_arcturus_hqd_sdma_dump(struct amdgpu_device *adev, uint32_t engine_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs); -bool kgd_arcturus_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd); -int kgd_arcturus_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, +bool kgd_arcturus_hqd_sdma_is_occupied(struct amdgpu_device *adev, + void *mqd); +int kgd_arcturus_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd, unsigned int utimeout); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c index 960acf68150a..7b7f4b2764c1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c @@ -39,37 +39,26 @@ enum hqd_dequeue_request_type { SAVE_WAVES }; -static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) -{ - return (struct amdgpu_device *)kgd; -} - -static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe, +static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe, uint32_t queue, uint32_t vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - mutex_lock(&adev->srbm_mutex); nv_grbm_select(adev, mec, pipe, queue, vmid); } -static void unlock_srbm(struct kgd_dev *kgd) +static void unlock_srbm(struct amdgpu_device *adev) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - nv_grbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); } -static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id, +static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); - lock_srbm(kgd, mec, pipe, queue_id, 0); + lock_srbm(adev, mec, pipe, queue_id, 0); } static uint64_t get_queue_mask(struct amdgpu_device *adev, @@ -81,33 +70,29 @@ static uint64_t get_queue_mask(struct amdgpu_device *adev, return 1ull << bit; } -static void release_queue(struct kgd_dev *kgd) +static void release_queue(struct amdgpu_device *adev) { - unlock_srbm(kgd); + unlock_srbm(adev); } -static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, +static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid, uint32_t sh_mem_config, uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - - lock_srbm(kgd, 0, 0, 0, vmid); + lock_srbm(adev, 0, 0, 0, vmid); WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, sh_mem_config); WREG32_SOC15(GC, 0, mmSH_MEM_BASES, sh_mem_bases); /* APE1 no longer exists on GFX9 */ - unlock_srbm(kgd); + unlock_srbm(adev); } -static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid, +static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid, unsigned int vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - /* * We have to assume that there is no outstanding mapping. * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 because @@ -150,22 +135,21 @@ static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid, * but still works */ -static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id) +static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t mec; uint32_t pipe; mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); - lock_srbm(kgd, mec, pipe, 0, 0); + lock_srbm(adev, mec, pipe, 0, 0); WREG32_SOC15(GC, 0, mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK | CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK); - unlock_srbm(kgd); + unlock_srbm(adev); return 0; } @@ -218,12 +202,11 @@ static inline struct v10_sdma_mqd *get_sdma_mqd(void *mqd) return (struct v10_sdma_mqd *)mqd; } -static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, - uint32_t queue_id, uint32_t __user *wptr, - uint32_t wptr_shift, uint32_t wptr_mask, - struct mm_struct *mm) +static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t __user *wptr, uint32_t wptr_shift, + uint32_t wptr_mask, struct mm_struct *mm) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v10_compute_mqd *m; uint32_t *mqd_hqd; uint32_t reg, hqd_base, data; @@ -231,7 +214,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, m = get_mqd(mqd); pr_debug("Load hqd of pipe %d queue %d\n", pipe_id, queue_id); - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); /* HQD registers extend from CP_MQD_BASE_ADDR to CP_HQD_EOP_WPTR_MEM. */ mqd_hqd = &m->cp_mqd_base_addr_lo; @@ -296,16 +279,15 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1); WREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE, data); - release_queue(kgd); + release_queue(adev); return 0; } -static int kgd_hiq_mqd_load(struct kgd_dev *kgd, void *mqd, +static int kgd_hiq_mqd_load(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t doorbell_off) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; struct v10_compute_mqd *m; uint32_t mec, pipe; @@ -313,7 +295,7 @@ static int kgd_hiq_mqd_load(struct kgd_dev *kgd, void *mqd, m = get_mqd(mqd); - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); @@ -349,16 +331,15 @@ static int kgd_hiq_mqd_load(struct kgd_dev *kgd, void *mqd, out_unlock: spin_unlock(&adev->gfx.kiq.ring_lock); - release_queue(kgd); + release_queue(adev); return r; } -static int kgd_hqd_dump(struct kgd_dev *kgd, +static int kgd_hqd_dump(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t i = 0, reg; #define HQD_N_REGS 56 #define DUMP_REG(addr) do { \ @@ -372,13 +353,13 @@ static int kgd_hqd_dump(struct kgd_dev *kgd, if (*dump == NULL) return -ENOMEM; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); for (reg = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR); reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++) DUMP_REG(reg); - release_queue(kgd); + release_queue(adev); WARN_ON_ONCE(i != HQD_N_REGS); *n_regs = i; @@ -386,10 +367,9 @@ static int kgd_hqd_dump(struct kgd_dev *kgd, return 0; } -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, +static int kgd_hqd_sdma_load(struct amdgpu_device *adev, void *mqd, uint32_t __user *wptr, struct mm_struct *mm) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v10_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; unsigned long end_jiffies; @@ -456,11 +436,10 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, return 0; } -static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, +static int kgd_hqd_sdma_dump(struct amdgpu_device *adev, uint32_t engine_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, engine_id, queue_id); uint32_t i = 0, reg; @@ -488,15 +467,15 @@ static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, return 0; } -static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, - uint32_t pipe_id, uint32_t queue_id) +static bool kgd_hqd_is_occupied(struct amdgpu_device *adev, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t act; bool retval = false; uint32_t low, high; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); act = RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE); if (act) { low = lower_32_bits(queue_address >> 8); @@ -506,13 +485,12 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, high == RREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE_HI)) retval = true; } - release_queue(kgd); + release_queue(adev); return retval; } -static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) +static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v10_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; uint32_t sdma_rlc_rb_cntl; @@ -529,12 +507,11 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) return false; } -static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, +static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd, enum kfd_preempt_type reset_type, unsigned int utimeout, uint32_t pipe_id, uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); enum hqd_dequeue_request_type type; unsigned long end_jiffies; uint32_t temp; @@ -548,7 +525,7 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, int retry; #endif - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); if (m->cp_hqd_vmid == 0) WREG32_FIELD15(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0); @@ -633,20 +610,19 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, break; if (time_after(jiffies, end_jiffies)) { pr_err("cp queue preemption time out.\n"); - release_queue(kgd); + release_queue(adev); return -ETIME; } usleep_range(500, 1000); } - release_queue(kgd); + release_queue(adev); return 0; } -static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, +static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd, unsigned int utimeout) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v10_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; uint32_t temp; @@ -683,11 +659,10 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, return 0; } -static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd, +static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev, uint8_t vmid, uint16_t *p_pasid) { uint32_t value; - struct amdgpu_device *adev = (struct amdgpu_device *) kgd; value = RREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATC_VMID0_PASID_MAPPING) + vmid); @@ -696,12 +671,12 @@ static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd, return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK); } -static int kgd_address_watch_disable(struct kgd_dev *kgd) +static int kgd_address_watch_disable(struct amdgpu_device *adev) { return 0; } -static int kgd_address_watch_execute(struct kgd_dev *kgd, +static int kgd_address_watch_execute(struct amdgpu_device *adev, unsigned int watch_point_id, uint32_t cntl_val, uint32_t addr_hi, @@ -710,11 +685,10 @@ static int kgd_address_watch_execute(struct kgd_dev *kgd, return 0; } -static int kgd_wave_control_execute(struct kgd_dev *kgd, +static int kgd_wave_control_execute(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t data = 0; mutex_lock(&adev->grbm_idx_mutex); @@ -735,18 +709,16 @@ static int kgd_wave_control_execute(struct kgd_dev *kgd, return 0; } -static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd, +static uint32_t kgd_address_watch_get_offset(struct amdgpu_device *adev, unsigned int watch_point_id, unsigned int reg_offset) { return 0; } -static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, - uint64_t page_table_base) +static void set_vm_context_page_table_base(struct amdgpu_device *adev, + uint32_t vmid, uint64_t page_table_base) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) { pr_err("trying to set page table base for wrong VMID %u\n", vmid); @@ -757,12 +729,10 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, adev->gfxhub.funcs->setup_vm_pt_regs(adev, vmid, page_table_base); } -static void program_trap_handler_settings(struct kgd_dev *kgd, +static void program_trap_handler_settings(struct amdgpu_device *adev, uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - - lock_srbm(kgd, 0, 0, 0, vmid); + lock_srbm(adev, 0, 0, 0, vmid); /* * Program TBA registers @@ -781,7 +751,7 @@ static void program_trap_handler_settings(struct kgd_dev *kgd, WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_SHADER_TMA_HI), upper_32_bits(tma_addr >> 8)); - unlock_srbm(kgd); + unlock_srbm(adev); } const struct kfd2kgd_calls gfx_v10_kfd2kgd = { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c index dac0d751d5af..1f37d3574001 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c @@ -38,37 +38,26 @@ enum hqd_dequeue_request_type { SAVE_WAVES }; -static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) -{ - return (struct amdgpu_device *)kgd; -} - -static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe, +static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe, uint32_t queue, uint32_t vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - mutex_lock(&adev->srbm_mutex); nv_grbm_select(adev, mec, pipe, queue, vmid); } -static void unlock_srbm(struct kgd_dev *kgd) +static void unlock_srbm(struct amdgpu_device *adev) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - nv_grbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); } -static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id, +static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); - lock_srbm(kgd, mec, pipe, queue_id, 0); + lock_srbm(adev, mec, pipe, queue_id, 0); } static uint64_t get_queue_mask(struct amdgpu_device *adev, @@ -80,34 +69,30 @@ static uint64_t get_queue_mask(struct amdgpu_device *adev, return 1ull << bit; } -static void release_queue(struct kgd_dev *kgd) +static void release_queue(struct amdgpu_device *adev) { - unlock_srbm(kgd); + unlock_srbm(adev); } -static void program_sh_mem_settings_v10_3(struct kgd_dev *kgd, uint32_t vmid, +static void program_sh_mem_settings_v10_3(struct amdgpu_device *adev, uint32_t vmid, uint32_t sh_mem_config, uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - - lock_srbm(kgd, 0, 0, 0, vmid); + lock_srbm(adev, 0, 0, 0, vmid); WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, sh_mem_config); WREG32_SOC15(GC, 0, mmSH_MEM_BASES, sh_mem_bases); /* APE1 no longer exists on GFX9 */ - unlock_srbm(kgd); + unlock_srbm(adev); } /* ATC is defeatured on Sienna_Cichlid */ -static int set_pasid_vmid_mapping_v10_3(struct kgd_dev *kgd, unsigned int pasid, +static int set_pasid_vmid_mapping_v10_3(struct amdgpu_device *adev, unsigned int pasid, unsigned int vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint32_t value = pasid << IH_VMID_0_LUT__PASID__SHIFT; /* Mapping vmid to pasid also for IH block */ @@ -118,22 +103,21 @@ static int set_pasid_vmid_mapping_v10_3(struct kgd_dev *kgd, unsigned int pasid, return 0; } -static int init_interrupts_v10_3(struct kgd_dev *kgd, uint32_t pipe_id) +static int init_interrupts_v10_3(struct amdgpu_device *adev, uint32_t pipe_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t mec; uint32_t pipe; mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); - lock_srbm(kgd, mec, pipe, 0, 0); + lock_srbm(adev, mec, pipe, 0, 0); WREG32_SOC15(GC, 0, mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK | CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK); - unlock_srbm(kgd); + unlock_srbm(adev); return 0; } @@ -188,12 +172,11 @@ static inline struct v10_sdma_mqd *get_sdma_mqd(void *mqd) return (struct v10_sdma_mqd *)mqd; } -static int hqd_load_v10_3(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, - uint32_t queue_id, uint32_t __user *wptr, - uint32_t wptr_shift, uint32_t wptr_mask, - struct mm_struct *mm) +static int hqd_load_v10_3(struct amdgpu_device *adev, void *mqd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t __user *wptr, uint32_t wptr_shift, + uint32_t wptr_mask, struct mm_struct *mm) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v10_compute_mqd *m; uint32_t *mqd_hqd; uint32_t reg, hqd_base, data; @@ -201,7 +184,7 @@ static int hqd_load_v10_3(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, m = get_mqd(mqd); pr_debug("Load hqd of pipe %d queue %d\n", pipe_id, queue_id); - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); /* HIQ is set during driver init period with vmid set to 0*/ if (m->cp_hqd_vmid == 0) { @@ -281,16 +264,15 @@ static int hqd_load_v10_3(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1); WREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE, data); - release_queue(kgd); + release_queue(adev); return 0; } -static int hiq_mqd_load_v10_3(struct kgd_dev *kgd, void *mqd, +static int hiq_mqd_load_v10_3(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t doorbell_off) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; struct v10_compute_mqd *m; uint32_t mec, pipe; @@ -298,7 +280,7 @@ static int hiq_mqd_load_v10_3(struct kgd_dev *kgd, void *mqd, m = get_mqd(mqd); - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); @@ -334,16 +316,15 @@ static int hiq_mqd_load_v10_3(struct kgd_dev *kgd, void *mqd, out_unlock: spin_unlock(&adev->gfx.kiq.ring_lock); - release_queue(kgd); + release_queue(adev); return r; } -static int hqd_dump_v10_3(struct kgd_dev *kgd, +static int hqd_dump_v10_3(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t i = 0, reg; #define HQD_N_REGS 56 #define DUMP_REG(addr) do { \ @@ -357,13 +338,13 @@ static int hqd_dump_v10_3(struct kgd_dev *kgd, if (*dump == NULL) return -ENOMEM; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); for (reg = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR); reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++) DUMP_REG(reg); - release_queue(kgd); + release_queue(adev); WARN_ON_ONCE(i != HQD_N_REGS); *n_regs = i; @@ -371,10 +352,9 @@ static int hqd_dump_v10_3(struct kgd_dev *kgd, return 0; } -static int hqd_sdma_load_v10_3(struct kgd_dev *kgd, void *mqd, +static int hqd_sdma_load_v10_3(struct amdgpu_device *adev, void *mqd, uint32_t __user *wptr, struct mm_struct *mm) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v10_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; unsigned long end_jiffies; @@ -441,11 +421,10 @@ static int hqd_sdma_load_v10_3(struct kgd_dev *kgd, void *mqd, return 0; } -static int hqd_sdma_dump_v10_3(struct kgd_dev *kgd, +static int hqd_sdma_dump_v10_3(struct amdgpu_device *adev, uint32_t engine_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, engine_id, queue_id); uint32_t i = 0, reg; @@ -473,15 +452,15 @@ static int hqd_sdma_dump_v10_3(struct kgd_dev *kgd, return 0; } -static bool hqd_is_occupied_v10_3(struct kgd_dev *kgd, uint64_t queue_address, - uint32_t pipe_id, uint32_t queue_id) +static bool hqd_is_occupied_v10_3(struct amdgpu_device *adev, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t act; bool retval = false; uint32_t low, high; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); act = RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE); if (act) { low = lower_32_bits(queue_address >> 8); @@ -491,13 +470,13 @@ static bool hqd_is_occupied_v10_3(struct kgd_dev *kgd, uint64_t queue_address, high == RREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE_HI)) retval = true; } - release_queue(kgd); + release_queue(adev); return retval; } -static bool hqd_sdma_is_occupied_v10_3(struct kgd_dev *kgd, void *mqd) +static bool hqd_sdma_is_occupied_v10_3(struct amdgpu_device *adev, + void *mqd) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v10_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; uint32_t sdma_rlc_rb_cntl; @@ -514,18 +493,17 @@ static bool hqd_sdma_is_occupied_v10_3(struct kgd_dev *kgd, void *mqd) return false; } -static int hqd_destroy_v10_3(struct kgd_dev *kgd, void *mqd, +static int hqd_destroy_v10_3(struct amdgpu_device *adev, void *mqd, enum kfd_preempt_type reset_type, unsigned int utimeout, uint32_t pipe_id, uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); enum hqd_dequeue_request_type type; unsigned long end_jiffies; uint32_t temp; struct v10_compute_mqd *m = get_mqd(mqd); - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); if (m->cp_hqd_vmid == 0) WREG32_FIELD15(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0); @@ -555,20 +533,19 @@ static int hqd_destroy_v10_3(struct kgd_dev *kgd, void *mqd, if (time_after(jiffies, end_jiffies)) { pr_err("cp queue pipe %d queue %d preemption failed\n", pipe_id, queue_id); - release_queue(kgd); + release_queue(adev); return -ETIME; } usleep_range(500, 1000); } - release_queue(kgd); + release_queue(adev); return 0; } -static int hqd_sdma_destroy_v10_3(struct kgd_dev *kgd, void *mqd, +static int hqd_sdma_destroy_v10_3(struct amdgpu_device *adev, void *mqd, unsigned int utimeout) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v10_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; uint32_t temp; @@ -606,12 +583,12 @@ static int hqd_sdma_destroy_v10_3(struct kgd_dev *kgd, void *mqd, } -static int address_watch_disable_v10_3(struct kgd_dev *kgd) +static int address_watch_disable_v10_3(struct amdgpu_device *adev) { return 0; } -static int address_watch_execute_v10_3(struct kgd_dev *kgd, +static int address_watch_execute_v10_3(struct amdgpu_device *adev, unsigned int watch_point_id, uint32_t cntl_val, uint32_t addr_hi, @@ -620,11 +597,10 @@ static int address_watch_execute_v10_3(struct kgd_dev *kgd, return 0; } -static int wave_control_execute_v10_3(struct kgd_dev *kgd, +static int wave_control_execute_v10_3(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t data = 0; mutex_lock(&adev->grbm_idx_mutex); @@ -645,28 +621,24 @@ static int wave_control_execute_v10_3(struct kgd_dev *kgd, return 0; } -static uint32_t address_watch_get_offset_v10_3(struct kgd_dev *kgd, +static uint32_t address_watch_get_offset_v10_3(struct amdgpu_device *adev, unsigned int watch_point_id, unsigned int reg_offset) { return 0; } -static void set_vm_context_page_table_base_v10_3(struct kgd_dev *kgd, uint32_t vmid, - uint64_t page_table_base) +static void set_vm_context_page_table_base_v10_3(struct amdgpu_device *adev, + uint32_t vmid, uint64_t page_table_base) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - /* SDMA is on gfxhub as well for Navi1* series */ adev->gfxhub.funcs->setup_vm_pt_regs(adev, vmid, page_table_base); } -static void program_trap_handler_settings_v10_3(struct kgd_dev *kgd, +static void program_trap_handler_settings_v10_3(struct amdgpu_device *adev, uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - - lock_srbm(kgd, 0, 0, 0, vmid); + lock_srbm(adev, 0, 0, 0, vmid); /* * Program TBA registers @@ -685,15 +657,14 @@ static void program_trap_handler_settings_v10_3(struct kgd_dev *kgd, WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_SHADER_TMA_HI), upper_32_bits(tma_addr >> 8)); - unlock_srbm(kgd); + unlock_srbm(adev); } #if 0 -uint32_t enable_debug_trap_v10_3(struct kgd_dev *kgd, +uint32_t enable_debug_trap_v10_3(struct amdgpu_device *adev, uint32_t trap_debug_wave_launch_mode, uint32_t vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t data = 0; uint32_t orig_wave_cntl_value; uint32_t orig_stall_vmid; @@ -720,10 +691,8 @@ uint32_t enable_debug_trap_v10_3(struct kgd_dev *kgd, return 0; } -uint32_t disable_debug_trap_v10_3(struct kgd_dev *kgd) +uint32_t disable_debug_trap_v10_3(struct amdgpu_device *adev) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - mutex_lock(&adev->grbm_idx_mutex); WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0); @@ -733,11 +702,10 @@ uint32_t disable_debug_trap_v10_3(struct kgd_dev *kgd) return 0; } -uint32_t set_wave_launch_trap_override_v10_3(struct kgd_dev *kgd, +uint32_t set_wave_launch_trap_override_v10_3(struct amdgpu_device *adev, uint32_t trap_override, uint32_t trap_mask) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t data = 0; mutex_lock(&adev->grbm_idx_mutex); @@ -762,11 +730,10 @@ uint32_t set_wave_launch_trap_override_v10_3(struct kgd_dev *kgd, return 0; } -uint32_t set_wave_launch_mode_v10_3(struct kgd_dev *kgd, +uint32_t set_wave_launch_mode_v10_3(struct amdgpu_device *adev, uint8_t wave_launch_mode, uint32_t vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t data = 0; bool is_stall_mode; bool is_mode_set; @@ -805,16 +772,14 @@ uint32_t set_wave_launch_mode_v10_3(struct kgd_dev *kgd, * sem_rearm_wait_time -- Wait Count for Semaphore re-arm. * deq_retry_wait_time -- Wait Count for Global Wave Syncs. */ -void get_iq_wait_times_v10_3(struct kgd_dev *kgd, +void get_iq_wait_times_v10_3(struct amdgpu_device *adev, uint32_t *wait_times) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - *wait_times = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2)); } -void build_grace_period_packet_info_v10_3(struct kgd_dev *kgd, +void build_grace_period_packet_info_v10_3(struct amdgpu_device *adev, uint32_t wait_times, uint32_t grace_period, uint32_t *reg_offset, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index b91d27e39bad..36528dad7684 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -82,68 +82,54 @@ union TCP_WATCH_CNTL_BITS { float f32All; }; -static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) -{ - return (struct amdgpu_device *)kgd; -} - -static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe, +static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe, uint32_t queue, uint32_t vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) | QUEUEID(queue); mutex_lock(&adev->srbm_mutex); WREG32(mmSRBM_GFX_CNTL, value); } -static void unlock_srbm(struct kgd_dev *kgd) +static void unlock_srbm(struct amdgpu_device *adev) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - WREG32(mmSRBM_GFX_CNTL, 0); mutex_unlock(&adev->srbm_mutex); } -static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id, +static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); - lock_srbm(kgd, mec, pipe, queue_id, 0); + lock_srbm(adev, mec, pipe, queue_id, 0); } -static void release_queue(struct kgd_dev *kgd) +static void release_queue(struct amdgpu_device *adev) { - unlock_srbm(kgd); + unlock_srbm(adev); } -static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, +static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid, uint32_t sh_mem_config, uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - - lock_srbm(kgd, 0, 0, 0, vmid); + lock_srbm(adev, 0, 0, 0, vmid); WREG32(mmSH_MEM_CONFIG, sh_mem_config); WREG32(mmSH_MEM_APE1_BASE, sh_mem_ape1_base); WREG32(mmSH_MEM_APE1_LIMIT, sh_mem_ape1_limit); WREG32(mmSH_MEM_BASES, sh_mem_bases); - unlock_srbm(kgd); + unlock_srbm(adev); } -static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid, +static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid, unsigned int vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - /* * We have to assume that there is no outstanding mapping. * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 because @@ -165,21 +151,20 @@ static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid, return 0; } -static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id) +static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t mec; uint32_t pipe; mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); - lock_srbm(kgd, mec, pipe, 0, 0); + lock_srbm(adev, mec, pipe, 0, 0); WREG32(mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK | CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK); - unlock_srbm(kgd); + unlock_srbm(adev); return 0; } @@ -207,12 +192,11 @@ static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd) return (struct cik_sdma_rlc_registers *)mqd; } -static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, - uint32_t queue_id, uint32_t __user *wptr, - uint32_t wptr_shift, uint32_t wptr_mask, - struct mm_struct *mm) +static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t __user *wptr, uint32_t wptr_shift, + uint32_t wptr_mask, struct mm_struct *mm) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct cik_mqd *m; uint32_t *mqd_hqd; uint32_t reg, wptr_val, data; @@ -220,7 +204,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, m = get_mqd(mqd); - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); /* HQD registers extend from CP_MQD_BASE_ADDR to CP_MQD_CONTROL. */ mqd_hqd = &m->cp_mqd_base_addr_lo; @@ -239,25 +223,24 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, * release srbm_mutex to avoid circular dependency between * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex. */ - release_queue(kgd); + release_queue(adev); valid_wptr = read_user_wptr(mm, wptr, wptr_val); - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); if (valid_wptr) WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask); data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1); WREG32(mmCP_HQD_ACTIVE, data); - release_queue(kgd); + release_queue(adev); return 0; } -static int kgd_hqd_dump(struct kgd_dev *kgd, +static int kgd_hqd_dump(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t i = 0, reg; #define HQD_N_REGS (35+4) #define DUMP_REG(addr) do { \ @@ -271,7 +254,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd, if (*dump == NULL) return -ENOMEM; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0); DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1); @@ -281,7 +264,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd, for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_MQD_CONTROL; reg++) DUMP_REG(reg); - release_queue(kgd); + release_queue(adev); WARN_ON_ONCE(i != HQD_N_REGS); *n_regs = i; @@ -289,10 +272,9 @@ static int kgd_hqd_dump(struct kgd_dev *kgd, return 0; } -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, +static int kgd_hqd_sdma_load(struct amdgpu_device *adev, void *mqd, uint32_t __user *wptr, struct mm_struct *mm) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct cik_sdma_rlc_registers *m; unsigned long end_jiffies; uint32_t sdma_rlc_reg_offset; @@ -345,11 +327,10 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, return 0; } -static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, +static int kgd_hqd_sdma_dump(struct amdgpu_device *adev, uint32_t engine_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t sdma_offset = engine_id * SDMA1_REGISTER_OFFSET + queue_id * KFD_CIK_SDMA_QUEUE_OFFSET; uint32_t i = 0, reg; @@ -372,15 +353,15 @@ static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, return 0; } -static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, - uint32_t pipe_id, uint32_t queue_id) +static bool kgd_hqd_is_occupied(struct amdgpu_device *adev, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t act; bool retval = false; uint32_t low, high; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); act = RREG32(mmCP_HQD_ACTIVE); if (act) { low = lower_32_bits(queue_address >> 8); @@ -390,13 +371,12 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, high == RREG32(mmCP_HQD_PQ_BASE_HI)) retval = true; } - release_queue(kgd); + release_queue(adev); return retval; } -static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) +static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct cik_sdma_rlc_registers *m; uint32_t sdma_rlc_reg_offset; uint32_t sdma_rlc_rb_cntl; @@ -412,12 +392,11 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) return false; } -static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, +static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd, enum kfd_preempt_type reset_type, unsigned int utimeout, uint32_t pipe_id, uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t temp; enum hqd_dequeue_request_type type; unsigned long flags, end_jiffies; @@ -426,7 +405,7 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, if (amdgpu_in_reset(adev)) return -EIO; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, 0); switch (reset_type) { @@ -504,20 +483,19 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, break; if (time_after(jiffies, end_jiffies)) { pr_err("cp queue preemption time out\n"); - release_queue(kgd); + release_queue(adev); return -ETIME; } usleep_range(500, 1000); } - release_queue(kgd); + release_queue(adev); return 0; } -static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, +static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd, unsigned int utimeout) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct cik_sdma_rlc_registers *m; uint32_t sdma_rlc_reg_offset; uint32_t temp; @@ -551,9 +529,8 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, return 0; } -static int kgd_address_watch_disable(struct kgd_dev *kgd) +static int kgd_address_watch_disable(struct amdgpu_device *adev) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); union TCP_WATCH_CNTL_BITS cntl; unsigned int i; @@ -571,13 +548,12 @@ static int kgd_address_watch_disable(struct kgd_dev *kgd) return 0; } -static int kgd_address_watch_execute(struct kgd_dev *kgd, +static int kgd_address_watch_execute(struct amdgpu_device *adev, unsigned int watch_point_id, uint32_t cntl_val, uint32_t addr_hi, uint32_t addr_lo) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); union TCP_WATCH_CNTL_BITS cntl; cntl.u32All = cntl_val; @@ -602,11 +578,10 @@ static int kgd_address_watch_execute(struct kgd_dev *kgd, return 0; } -static int kgd_wave_control_execute(struct kgd_dev *kgd, +static int kgd_wave_control_execute(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t data; mutex_lock(&adev->grbm_idx_mutex); @@ -627,18 +602,17 @@ static int kgd_wave_control_execute(struct kgd_dev *kgd, return 0; } -static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd, +static uint32_t kgd_address_watch_get_offset(struct amdgpu_device *adev, unsigned int watch_point_id, unsigned int reg_offset) { return watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + reg_offset]; } -static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd, +static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev, uint8_t vmid, uint16_t *p_pasid) { uint32_t value; - struct amdgpu_device *adev = (struct amdgpu_device *) kgd; value = RREG32(mmATC_VMID0_PASID_MAPPING + vmid); *p_pasid = value & ATC_VMID0_PASID_MAPPING__PASID_MASK; @@ -646,21 +620,17 @@ static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd, return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK); } -static void set_scratch_backing_va(struct kgd_dev *kgd, +static void set_scratch_backing_va(struct amdgpu_device *adev, uint64_t va, uint32_t vmid) { - struct amdgpu_device *adev = (struct amdgpu_device *) kgd; - - lock_srbm(kgd, 0, 0, 0, vmid); + lock_srbm(adev, 0, 0, 0, vmid); WREG32(mmSH_HIDDEN_PRIVATE_BASE_VMID, va); - unlock_srbm(kgd); + unlock_srbm(adev); } -static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, - uint64_t page_table_base) +static void set_vm_context_page_table_base(struct amdgpu_device *adev, + uint32_t vmid, uint64_t page_table_base) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) { pr_err("trying to set page table base for wrong VMID\n"); return; @@ -676,10 +646,8 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, * @vmid: vmid pointer * read vmid from register (CIK). */ -static uint32_t read_vmid_from_vmfault_reg(struct kgd_dev *kgd) +static uint32_t read_vmid_from_vmfault_reg(struct amdgpu_device *adev) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint32_t status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS); return REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index 5ce0ce704a21..52832cd69a93 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -39,68 +39,54 @@ enum hqd_dequeue_request_type { RESET_WAVES }; -static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) -{ - return (struct amdgpu_device *)kgd; -} - -static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe, +static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe, uint32_t queue, uint32_t vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) | QUEUEID(queue); mutex_lock(&adev->srbm_mutex); WREG32(mmSRBM_GFX_CNTL, value); } -static void unlock_srbm(struct kgd_dev *kgd) +static void unlock_srbm(struct amdgpu_device *adev) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - WREG32(mmSRBM_GFX_CNTL, 0); mutex_unlock(&adev->srbm_mutex); } -static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id, +static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); - lock_srbm(kgd, mec, pipe, queue_id, 0); + lock_srbm(adev, mec, pipe, queue_id, 0); } -static void release_queue(struct kgd_dev *kgd) +static void release_queue(struct amdgpu_device *adev) { - unlock_srbm(kgd); + unlock_srbm(adev); } -static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, +static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid, uint32_t sh_mem_config, uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - - lock_srbm(kgd, 0, 0, 0, vmid); + lock_srbm(adev, 0, 0, 0, vmid); WREG32(mmSH_MEM_CONFIG, sh_mem_config); WREG32(mmSH_MEM_APE1_BASE, sh_mem_ape1_base); WREG32(mmSH_MEM_APE1_LIMIT, sh_mem_ape1_limit); WREG32(mmSH_MEM_BASES, sh_mem_bases); - unlock_srbm(kgd); + unlock_srbm(adev); } -static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid, +static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid, unsigned int vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - /* * We have to assume that there is no outstanding mapping. * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 because @@ -123,21 +109,20 @@ static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid, return 0; } -static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id) +static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t mec; uint32_t pipe; mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); - lock_srbm(kgd, mec, pipe, 0, 0); + lock_srbm(adev, mec, pipe, 0, 0); WREG32(mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK | CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK); - unlock_srbm(kgd); + unlock_srbm(adev); return 0; } @@ -165,12 +150,11 @@ static inline struct vi_sdma_mqd *get_sdma_mqd(void *mqd) return (struct vi_sdma_mqd *)mqd; } -static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, - uint32_t queue_id, uint32_t __user *wptr, - uint32_t wptr_shift, uint32_t wptr_mask, - struct mm_struct *mm) +static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t __user *wptr, uint32_t wptr_shift, + uint32_t wptr_mask, struct mm_struct *mm) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct vi_mqd *m; uint32_t *mqd_hqd; uint32_t reg, wptr_val, data; @@ -178,7 +162,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, m = get_mqd(mqd); - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); /* HIQ is set during driver init period with vmid set to 0*/ if (m->cp_hqd_vmid == 0) { @@ -206,7 +190,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, * on ASICs that do not support context-save. * EOP writes/reads can start anywhere in the ring. */ - if (get_amdgpu_device(kgd)->asic_type != CHIP_TONGA) { + if (adev->asic_type != CHIP_TONGA) { WREG32(mmCP_HQD_EOP_RPTR, m->cp_hqd_eop_rptr); WREG32(mmCP_HQD_EOP_WPTR, m->cp_hqd_eop_wptr); WREG32(mmCP_HQD_EOP_WPTR_MEM, m->cp_hqd_eop_wptr_mem); @@ -226,25 +210,24 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, * release srbm_mutex to avoid circular dependency between * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex. */ - release_queue(kgd); + release_queue(adev); valid_wptr = read_user_wptr(mm, wptr, wptr_val); - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); if (valid_wptr) WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask); data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1); WREG32(mmCP_HQD_ACTIVE, data); - release_queue(kgd); + release_queue(adev); return 0; } -static int kgd_hqd_dump(struct kgd_dev *kgd, +static int kgd_hqd_dump(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t i = 0, reg; #define HQD_N_REGS (54+4) #define DUMP_REG(addr) do { \ @@ -258,7 +241,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd, if (*dump == NULL) return -ENOMEM; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0); DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1); @@ -268,7 +251,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd, for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_HQD_EOP_DONES; reg++) DUMP_REG(reg); - release_queue(kgd); + release_queue(adev); WARN_ON_ONCE(i != HQD_N_REGS); *n_regs = i; @@ -276,10 +259,9 @@ static int kgd_hqd_dump(struct kgd_dev *kgd, return 0; } -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, +static int kgd_hqd_sdma_load(struct amdgpu_device *adev, void *mqd, uint32_t __user *wptr, struct mm_struct *mm) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct vi_sdma_mqd *m; unsigned long end_jiffies; uint32_t sdma_rlc_reg_offset; @@ -331,11 +313,10 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, return 0; } -static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, +static int kgd_hqd_sdma_dump(struct amdgpu_device *adev, uint32_t engine_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t sdma_offset = engine_id * SDMA1_REGISTER_OFFSET + queue_id * KFD_VI_SDMA_QUEUE_OFFSET; uint32_t i = 0, reg; @@ -367,15 +348,15 @@ static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, return 0; } -static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, - uint32_t pipe_id, uint32_t queue_id) +static bool kgd_hqd_is_occupied(struct amdgpu_device *adev, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t act; bool retval = false; uint32_t low, high; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); act = RREG32(mmCP_HQD_ACTIVE); if (act) { low = lower_32_bits(queue_address >> 8); @@ -385,13 +366,12 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, high == RREG32(mmCP_HQD_PQ_BASE_HI)) retval = true; } - release_queue(kgd); + release_queue(adev); return retval; } -static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) +static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct vi_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; uint32_t sdma_rlc_rb_cntl; @@ -407,12 +387,11 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) return false; } -static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, +static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd, enum kfd_preempt_type reset_type, unsigned int utimeout, uint32_t pipe_id, uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t temp; enum hqd_dequeue_request_type type; unsigned long flags, end_jiffies; @@ -422,7 +401,7 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, if (amdgpu_in_reset(adev)) return -EIO; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); if (m->cp_hqd_vmid == 0) WREG32_FIELD(RLC_CP_SCHEDULERS, scheduler1, 0); @@ -502,20 +481,19 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, break; if (time_after(jiffies, end_jiffies)) { pr_err("cp queue preemption time out.\n"); - release_queue(kgd); + release_queue(adev); return -ETIME; } usleep_range(500, 1000); } - release_queue(kgd); + release_queue(adev); return 0; } -static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, +static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd, unsigned int utimeout) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct vi_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; uint32_t temp; @@ -549,11 +527,10 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, return 0; } -static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd, +static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev, uint8_t vmid, uint16_t *p_pasid) { uint32_t value; - struct amdgpu_device *adev = (struct amdgpu_device *) kgd; value = RREG32(mmATC_VMID0_PASID_MAPPING + vmid); *p_pasid = value & ATC_VMID0_PASID_MAPPING__PASID_MASK; @@ -561,12 +538,12 @@ static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd, return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK); } -static int kgd_address_watch_disable(struct kgd_dev *kgd) +static int kgd_address_watch_disable(struct amdgpu_device *adev) { return 0; } -static int kgd_address_watch_execute(struct kgd_dev *kgd, +static int kgd_address_watch_execute(struct amdgpu_device *adev, unsigned int watch_point_id, uint32_t cntl_val, uint32_t addr_hi, @@ -575,11 +552,10 @@ static int kgd_address_watch_execute(struct kgd_dev *kgd, return 0; } -static int kgd_wave_control_execute(struct kgd_dev *kgd, +static int kgd_wave_control_execute(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t data = 0; mutex_lock(&adev->grbm_idx_mutex); @@ -600,28 +576,24 @@ static int kgd_wave_control_execute(struct kgd_dev *kgd, return 0; } -static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd, +static uint32_t kgd_address_watch_get_offset(struct amdgpu_device *adev, unsigned int watch_point_id, unsigned int reg_offset) { return 0; } -static void set_scratch_backing_va(struct kgd_dev *kgd, +static void set_scratch_backing_va(struct amdgpu_device *adev, uint64_t va, uint32_t vmid) { - struct amdgpu_device *adev = (struct amdgpu_device *) kgd; - - lock_srbm(kgd, 0, 0, 0, vmid); + lock_srbm(adev, 0, 0, 0, vmid); WREG32(mmSH_HIDDEN_PRIVATE_BASE_VMID, va); - unlock_srbm(kgd); + unlock_srbm(adev); } -static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, - uint64_t page_table_base) +static void set_vm_context_page_table_base(struct amdgpu_device *adev, + uint32_t vmid, uint64_t page_table_base) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) { pr_err("trying to set page table base for wrong VMID\n"); return; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c index bcc1cbeb8799..ddfe7aff919d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -46,37 +46,26 @@ enum hqd_dequeue_request_type { SAVE_WAVES }; -static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) -{ - return (struct amdgpu_device *)kgd; -} - -static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe, +static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe, uint32_t queue, uint32_t vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - mutex_lock(&adev->srbm_mutex); soc15_grbm_select(adev, mec, pipe, queue, vmid); } -static void unlock_srbm(struct kgd_dev *kgd) +static void unlock_srbm(struct amdgpu_device *adev) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - soc15_grbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); } -static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id, +static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); - lock_srbm(kgd, mec, pipe, queue_id, 0); + lock_srbm(adev, mec, pipe, queue_id, 0); } static uint64_t get_queue_mask(struct amdgpu_device *adev, @@ -88,33 +77,29 @@ static uint64_t get_queue_mask(struct amdgpu_device *adev, return 1ull << bit; } -static void release_queue(struct kgd_dev *kgd) +static void release_queue(struct amdgpu_device *adev) { - unlock_srbm(kgd); + unlock_srbm(adev); } -void kgd_gfx_v9_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, +void kgd_gfx_v9_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid, uint32_t sh_mem_config, uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - - lock_srbm(kgd, 0, 0, 0, vmid); + lock_srbm(adev, 0, 0, 0, vmid); WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), sh_mem_config); WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_BASES), sh_mem_bases); /* APE1 no longer exists on GFX9 */ - unlock_srbm(kgd); + unlock_srbm(adev); } -int kgd_gfx_v9_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid, +int kgd_gfx_v9_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid, unsigned int vmid) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - /* * We have to assume that there is no outstanding mapping. * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 because @@ -171,22 +156,21 @@ int kgd_gfx_v9_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid, * but still works */ -int kgd_gfx_v9_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id) +int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t mec; uint32_t pipe; mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); - lock_srbm(kgd, mec, pipe, 0, 0); + lock_srbm(adev, mec, pipe, 0, 0); WREG32(SOC15_REG_OFFSET(GC, 0, mmCPC_INT_CNTL), CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK | CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK); - unlock_srbm(kgd); + unlock_srbm(adev); return 0; } @@ -233,19 +217,18 @@ static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd) return (struct v9_sdma_mqd *)mqd; } -int kgd_gfx_v9_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, - uint32_t queue_id, uint32_t __user *wptr, - uint32_t wptr_shift, uint32_t wptr_mask, - struct mm_struct *mm) +int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t __user *wptr, uint32_t wptr_shift, + uint32_t wptr_mask, struct mm_struct *mm) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v9_mqd *m; uint32_t *mqd_hqd; uint32_t reg, hqd_base, data; m = get_mqd(mqd); - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); /* HQD registers extend from CP_MQD_BASE_ADDR to CP_HQD_EOP_WPTR_MEM. */ mqd_hqd = &m->cp_mqd_base_addr_lo; @@ -308,16 +291,15 @@ int kgd_gfx_v9_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1); WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE), data); - release_queue(kgd); + release_queue(adev); return 0; } -int kgd_gfx_v9_hiq_mqd_load(struct kgd_dev *kgd, void *mqd, +int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t doorbell_off) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; struct v9_mqd *m; uint32_t mec, pipe; @@ -325,7 +307,7 @@ int kgd_gfx_v9_hiq_mqd_load(struct kgd_dev *kgd, void *mqd, m = get_mqd(mqd); - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); @@ -361,16 +343,15 @@ int kgd_gfx_v9_hiq_mqd_load(struct kgd_dev *kgd, void *mqd, out_unlock: spin_unlock(&adev->gfx.kiq.ring_lock); - release_queue(kgd); + release_queue(adev); return r; } -int kgd_gfx_v9_hqd_dump(struct kgd_dev *kgd, +int kgd_gfx_v9_hqd_dump(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t i = 0, reg; #define HQD_N_REGS 56 #define DUMP_REG(addr) do { \ @@ -384,13 +365,13 @@ int kgd_gfx_v9_hqd_dump(struct kgd_dev *kgd, if (*dump == NULL) return -ENOMEM; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); for (reg = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR); reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++) DUMP_REG(reg); - release_queue(kgd); + release_queue(adev); WARN_ON_ONCE(i != HQD_N_REGS); *n_regs = i; @@ -398,10 +379,9 @@ int kgd_gfx_v9_hqd_dump(struct kgd_dev *kgd, return 0; } -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, +static int kgd_hqd_sdma_load(struct amdgpu_device *adev, void *mqd, uint32_t __user *wptr, struct mm_struct *mm) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v9_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; unsigned long end_jiffies; @@ -468,11 +448,10 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, return 0; } -static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, +static int kgd_hqd_sdma_dump(struct amdgpu_device *adev, uint32_t engine_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, engine_id, queue_id); uint32_t i = 0, reg; @@ -500,15 +479,15 @@ static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, return 0; } -bool kgd_gfx_v9_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, - uint32_t pipe_id, uint32_t queue_id) +bool kgd_gfx_v9_hqd_is_occupied(struct amdgpu_device *adev, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t act; bool retval = false; uint32_t low, high; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); act = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE)); if (act) { low = lower_32_bits(queue_address >> 8); @@ -518,13 +497,12 @@ bool kgd_gfx_v9_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, high == RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_BASE_HI))) retval = true; } - release_queue(kgd); + release_queue(adev); return retval; } -static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) +static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v9_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; uint32_t sdma_rlc_rb_cntl; @@ -541,12 +519,11 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) return false; } -int kgd_gfx_v9_hqd_destroy(struct kgd_dev *kgd, void *mqd, +int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd, enum kfd_preempt_type reset_type, unsigned int utimeout, uint32_t pipe_id, uint32_t queue_id) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); enum hqd_dequeue_request_type type; unsigned long end_jiffies; uint32_t temp; @@ -555,7 +532,7 @@ int kgd_gfx_v9_hqd_destroy(struct kgd_dev *kgd, void *mqd, if (amdgpu_in_reset(adev)) return -EIO; - acquire_queue(kgd, pipe_id, queue_id); + acquire_queue(adev, pipe_id, queue_id); if (m->cp_hqd_vmid == 0) WREG32_FIELD15_RLC(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0); @@ -584,20 +561,19 @@ int kgd_gfx_v9_hqd_destroy(struct kgd_dev *kgd, void *mqd, break; if (time_after(jiffies, end_jiffies)) { pr_err("cp queue preemption time out.\n"); - release_queue(kgd); + release_queue(adev); return -ETIME; } usleep_range(500, 1000); } - release_queue(kgd); + release_queue(adev); return 0; } -static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, +static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd, unsigned int utimeout) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct v9_sdma_mqd *m; uint32_t sdma_rlc_reg_offset; uint32_t temp; @@ -634,11 +610,10 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, return 0; } -bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd, +bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev, uint8_t vmid, uint16_t *p_pasid) { uint32_t value; - struct amdgpu_device *adev = (struct amdgpu_device *) kgd; value = RREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATC_VMID0_PASID_MAPPING) + vmid); @@ -647,12 +622,12 @@ bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd, return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK); } -int kgd_gfx_v9_address_watch_disable(struct kgd_dev *kgd) +int kgd_gfx_v9_address_watch_disable(struct amdgpu_device *adev) { return 0; } -int kgd_gfx_v9_address_watch_execute(struct kgd_dev *kgd, +int kgd_gfx_v9_address_watch_execute(struct amdgpu_device *adev, unsigned int watch_point_id, uint32_t cntl_val, uint32_t addr_hi, @@ -661,11 +636,10 @@ int kgd_gfx_v9_address_watch_execute(struct kgd_dev *kgd, return 0; } -int kgd_gfx_v9_wave_control_execute(struct kgd_dev *kgd, +int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t data = 0; mutex_lock(&adev->grbm_idx_mutex); @@ -686,18 +660,16 @@ int kgd_gfx_v9_wave_control_execute(struct kgd_dev *kgd, return 0; } -uint32_t kgd_gfx_v9_address_watch_get_offset(struct kgd_dev *kgd, +uint32_t kgd_gfx_v9_address_watch_get_offset(struct amdgpu_device *adev, unsigned int watch_point_id, unsigned int reg_offset) { return 0; } -void kgd_gfx_v9_set_vm_context_page_table_base(struct kgd_dev *kgd, +void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev, uint32_t vmid, uint64_t page_table_base) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) { pr_err("trying to set page table base for wrong VMID %u\n", vmid); @@ -804,7 +776,7 @@ static void get_wave_count(struct amdgpu_device *adev, int queue_idx, * * Reading registers referenced above involves programming GRBM appropriately */ -void kgd_gfx_v9_get_cu_occupancy(struct kgd_dev *kgd, int pasid, +void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid, int *pasid_wave_cnt, int *max_waves_per_cu) { int qidx; @@ -818,10 +790,8 @@ void kgd_gfx_v9_get_cu_occupancy(struct kgd_dev *kgd, int pasid, int pasid_tmp; int max_queue_cnt; int vmid_wave_cnt = 0; - struct amdgpu_device *adev; DECLARE_BITMAP(cp_queue_bitmap, KGD_MAX_QUEUES); - adev = get_amdgpu_device(kgd); lock_spi_csq_mutexes(adev); soc15_grbm_select(adev, 1, 0, 0, 0); @@ -882,12 +852,10 @@ void kgd_gfx_v9_get_cu_occupancy(struct kgd_dev *kgd, int pasid, adev->gfx.cu_info.max_waves_per_simd; } -void kgd_gfx_v9_program_trap_handler_settings(struct kgd_dev *kgd, +void kgd_gfx_v9_program_trap_handler_settings(struct amdgpu_device *adev, uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - - lock_srbm(kgd, 0, 0, 0, vmid); + lock_srbm(adev, 0, 0, 0, vmid); /* * Program TBA registers @@ -905,7 +873,7 @@ void kgd_gfx_v9_program_trap_handler_settings(struct kgd_dev *kgd, WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_SHADER_TMA_HI), upper_32_bits(tma_addr >> 8)); - unlock_srbm(kgd); + unlock_srbm(adev); } const struct kfd2kgd_calls gfx_v9_kfd2kgd = { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h index c63591106879..24be49df26fd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h @@ -22,48 +22,49 @@ -void kgd_gfx_v9_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, +void kgd_gfx_v9_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid, uint32_t sh_mem_config, uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases); -int kgd_gfx_v9_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid, +int kgd_gfx_v9_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid, unsigned int vmid); -int kgd_gfx_v9_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id); -int kgd_gfx_v9_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, +int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id); +int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t __user *wptr, uint32_t wptr_shift, uint32_t wptr_mask, struct mm_struct *mm); -int kgd_gfx_v9_hiq_mqd_load(struct kgd_dev *kgd, void *mqd, +int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t doorbell_off); -int kgd_gfx_v9_hqd_dump(struct kgd_dev *kgd, +int kgd_gfx_v9_hqd_dump(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs); -bool kgd_gfx_v9_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, - uint32_t pipe_id, uint32_t queue_id); -int kgd_gfx_v9_hqd_destroy(struct kgd_dev *kgd, void *mqd, +bool kgd_gfx_v9_hqd_is_occupied(struct amdgpu_device *adev, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id); +int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd, enum kfd_preempt_type reset_type, unsigned int utimeout, uint32_t pipe_id, uint32_t queue_id); -int kgd_gfx_v9_address_watch_disable(struct kgd_dev *kgd); -int kgd_gfx_v9_address_watch_execute(struct kgd_dev *kgd, +int kgd_gfx_v9_address_watch_disable(struct amdgpu_device *adev); +int kgd_gfx_v9_address_watch_execute(struct amdgpu_device *adev, unsigned int watch_point_id, uint32_t cntl_val, uint32_t addr_hi, uint32_t addr_lo); -int kgd_gfx_v9_wave_control_execute(struct kgd_dev *kgd, +int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd); -uint32_t kgd_gfx_v9_address_watch_get_offset(struct kgd_dev *kgd, +uint32_t kgd_gfx_v9_address_watch_get_offset(struct amdgpu_device *adev, unsigned int watch_point_id, unsigned int reg_offset); -bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd, +bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev, uint8_t vmid, uint16_t *p_pasid); -void kgd_gfx_v9_set_vm_context_page_table_base(struct kgd_dev *kgd, +void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev, uint32_t vmid, uint64_t page_table_base); -void kgd_gfx_v9_get_cu_occupancy(struct kgd_dev *kgd, int pasid, +void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid, int *pasid_wave_cnt, int *max_waves_per_cu); -void kgd_gfx_v9_program_trap_handler_settings(struct kgd_dev *kgd, +void kgd_gfx_v9_program_trap_handler_settings(struct amdgpu_device *adev, uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index 71acd577803e..c5362c23e687 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -60,12 +60,6 @@ static const char * const domain_bit_to_string[] = { static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work); - -static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) -{ - return (struct amdgpu_device *)kgd; -} - static bool kfd_mem_is_attached(struct amdgpu_vm *avm, struct kgd_mem *mem) { @@ -126,8 +120,19 @@ static size_t amdgpu_amdkfd_acc_size(uint64_t size) PAGE_ALIGN(size); } +/** + * @amdgpu_amdkfd_reserve_mem_limit() - Decrease available memory by size + * of buffer including any reserved for control structures + * + * @adev: Device to which allocated BO belongs to + * @size: Size of buffer, in bytes, encapsulated by B0. This should be + * equivalent to amdgpu_bo_size(BO) + * @alloc_flag: Flag used in allocating a BO as noted above + * + * Return: returns -ENOMEM in case of error, ZERO otherwise + */ static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, - uint64_t size, u32 domain, bool sg) + uint64_t size, u32 alloc_flag) { uint64_t reserved_for_pt = ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size); @@ -137,20 +142,24 @@ static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, acc_size = amdgpu_amdkfd_acc_size(size); vram_needed = 0; - if (domain == AMDGPU_GEM_DOMAIN_GTT) { - /* TTM GTT memory */ + if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_GTT) { system_mem_needed = acc_size + size; ttm_mem_needed = acc_size + size; - } else if (domain == AMDGPU_GEM_DOMAIN_CPU && !sg) { - /* Userptr */ - system_mem_needed = acc_size + size; - ttm_mem_needed = acc_size; - } else { - /* VRAM and SG */ + } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { system_mem_needed = acc_size; ttm_mem_needed = acc_size; - if (domain == AMDGPU_GEM_DOMAIN_VRAM) - vram_needed = size; + vram_needed = size; + } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) { + system_mem_needed = acc_size + size; + ttm_mem_needed = acc_size; + } else if (alloc_flag & + (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | + KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) { + system_mem_needed = acc_size; + ttm_mem_needed = acc_size; + } else { + pr_err("%s: Invalid BO type %#x\n", __func__, alloc_flag); + return -ENOMEM; } spin_lock(&kfd_mem_limit.mem_limit_lock); @@ -166,64 +175,72 @@ static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, (adev->kfd.vram_used + vram_needed > adev->gmc.real_vram_size - reserved_for_pt)) { ret = -ENOMEM; - } else { - kfd_mem_limit.system_mem_used += system_mem_needed; - kfd_mem_limit.ttm_mem_used += ttm_mem_needed; - adev->kfd.vram_used += vram_needed; + goto release; } + /* Update memory accounting by decreasing available system + * memory, TTM memory and GPU memory as computed above + */ + adev->kfd.vram_used += vram_needed; + kfd_mem_limit.system_mem_used += system_mem_needed; + kfd_mem_limit.ttm_mem_used += ttm_mem_needed; + +release: spin_unlock(&kfd_mem_limit.mem_limit_lock); return ret; } static void unreserve_mem_limit(struct amdgpu_device *adev, - uint64_t size, u32 domain, bool sg) + uint64_t size, u32 alloc_flag) { size_t acc_size; acc_size = amdgpu_amdkfd_acc_size(size); spin_lock(&kfd_mem_limit.mem_limit_lock); - if (domain == AMDGPU_GEM_DOMAIN_GTT) { + + if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_GTT) { kfd_mem_limit.system_mem_used -= (acc_size + size); kfd_mem_limit.ttm_mem_used -= (acc_size + size); - } else if (domain == AMDGPU_GEM_DOMAIN_CPU && !sg) { - kfd_mem_limit.system_mem_used -= (acc_size + size); - kfd_mem_limit.ttm_mem_used -= acc_size; - } else { + } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { kfd_mem_limit.system_mem_used -= acc_size; kfd_mem_limit.ttm_mem_used -= acc_size; - if (domain == AMDGPU_GEM_DOMAIN_VRAM) { - adev->kfd.vram_used -= size; - WARN_ONCE(adev->kfd.vram_used < 0, - "kfd VRAM memory accounting unbalanced"); - } + adev->kfd.vram_used -= size; + } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) { + kfd_mem_limit.system_mem_used -= (acc_size + size); + kfd_mem_limit.ttm_mem_used -= acc_size; + } else if (alloc_flag & + (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | + KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) { + kfd_mem_limit.system_mem_used -= acc_size; + kfd_mem_limit.ttm_mem_used -= acc_size; + } else { + pr_err("%s: Invalid BO type %#x\n", __func__, alloc_flag); + goto release; } - WARN_ONCE(kfd_mem_limit.system_mem_used < 0, - "kfd system memory accounting unbalanced"); - WARN_ONCE(kfd_mem_limit.ttm_mem_used < 0, - "kfd TTM memory accounting unbalanced"); + WARN_ONCE(adev->kfd.vram_used < 0, + "KFD VRAM memory accounting unbalanced"); + WARN_ONCE(kfd_mem_limit.ttm_mem_used < 0, + "KFD TTM memory accounting unbalanced"); + WARN_ONCE(kfd_mem_limit.system_mem_used < 0, + "KFD system memory accounting unbalanced"); + +release: spin_unlock(&kfd_mem_limit.mem_limit_lock); } void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - u32 domain = bo->preferred_domains; - bool sg = (bo->preferred_domains == AMDGPU_GEM_DOMAIN_CPU); + u32 alloc_flags = bo->kfd_bo->alloc_flags; + u64 size = amdgpu_bo_size(bo); - if (bo->flags & AMDGPU_AMDKFD_CREATE_USERPTR_BO) { - domain = AMDGPU_GEM_DOMAIN_CPU; - sg = false; - } - - unreserve_mem_limit(adev, amdgpu_bo_size(bo), domain, sg); + unreserve_mem_limit(adev, size, alloc_flags); kfree(bo->kfd_bo); } - /* amdgpu_amdkfd_remove_eviction_fence - Removes eviction fence from BO's * reservation object. * @@ -646,12 +663,6 @@ kfd_mem_attach_dmabuf(struct amdgpu_device *adev, struct kgd_mem *mem, if (IS_ERR(gobj)) return PTR_ERR(gobj); - /* Import takes an extra reference on the dmabuf. Drop it now to - * avoid leaking it. We only need the one reference in - * kgd_mem->dmabuf. - */ - dma_buf_put(mem->dmabuf); - *bo = gem_to_amdgpu_bo(gobj); (*bo)->flags |= AMDGPU_GEM_CREATE_PREEMPTIBLE; (*bo)->parent = amdgpu_bo_ref(mem->bo); @@ -1278,12 +1289,60 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, return ret; } -int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, +/** + * amdgpu_amdkfd_gpuvm_pin_bo() - Pins a BO using following criteria + * @bo: Handle of buffer object being pinned + * @domain: Domain into which BO should be pinned + * + * - USERPTR BOs are UNPINNABLE and will return error + * - All other BO types (GTT, VRAM, MMIO and DOORBELL) will have their + * PIN count incremented. It is valid to PIN a BO multiple times + * + * Return: ZERO if successful in pinning, Non-Zero in case of error. + */ +static int amdgpu_amdkfd_gpuvm_pin_bo(struct amdgpu_bo *bo, u32 domain) +{ + int ret = 0; + + ret = amdgpu_bo_reserve(bo, false); + if (unlikely(ret)) + return ret; + + ret = amdgpu_bo_pin_restricted(bo, domain, 0, 0); + if (ret) + pr_err("Error in Pinning BO to domain: %d\n", domain); + + amdgpu_bo_sync_wait(bo, AMDGPU_FENCE_OWNER_KFD, false); + amdgpu_bo_unreserve(bo); + + return ret; +} + +/** + * amdgpu_amdkfd_gpuvm_unpin_bo() - Unpins BO using following criteria + * @bo: Handle of buffer object being unpinned + * + * - Is a illegal request for USERPTR BOs and is ignored + * - All other BO types (GTT, VRAM, MMIO and DOORBELL) will have their + * PIN count decremented. Calls to UNPIN must balance calls to PIN + */ +static void amdgpu_amdkfd_gpuvm_unpin_bo(struct amdgpu_bo *bo) +{ + int ret = 0; + + ret = amdgpu_bo_reserve(bo, false); + if (unlikely(ret)) + return; + + amdgpu_bo_unpin(bo); + amdgpu_bo_unreserve(bo); +} + +int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev, struct file *filp, u32 pasid, void **process_info, struct dma_fence **ef) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct amdgpu_fpriv *drv_priv; struct amdgpu_vm *avm; int ret; @@ -1359,12 +1418,12 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, } } -void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *drm_priv) +void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev, + void *drm_priv) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct amdgpu_vm *avm; - if (WARN_ON(!kgd || !drm_priv)) + if (WARN_ON(!adev || !drm_priv)) return; avm = drm_priv_to_vm(drm_priv); @@ -1392,11 +1451,10 @@ uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv) } int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( - struct kgd_dev *kgd, uint64_t va, uint64_t size, + struct amdgpu_device *adev, uint64_t va, uint64_t size, void *drm_priv, struct kgd_mem **mem, uint64_t *offset, uint32_t flags) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); enum ttm_bo_type bo_type = ttm_bo_type_device; struct sg_table *sg = NULL; @@ -1460,7 +1518,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( amdgpu_sync_create(&(*mem)->sync); - ret = amdgpu_amdkfd_reserve_mem_limit(adev, size, alloc_domain, !!sg); + ret = amdgpu_amdkfd_reserve_mem_limit(adev, size, flags); if (ret) { pr_debug("Insufficient memory\n"); goto err_reserve_limit; @@ -1506,17 +1564,29 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( if (offset) *offset = amdgpu_bo_mmap_offset(bo); + if (flags & (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | + KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) { + ret = amdgpu_amdkfd_gpuvm_pin_bo(bo, AMDGPU_GEM_DOMAIN_GTT); + if (ret) { + pr_err("Pinning MMIO/DOORBELL BO during ALLOC FAILED\n"); + goto err_pin_bo; + } + bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT; + bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT; + } + return 0; allocate_init_user_pages_failed: remove_kgd_mem_from_kfd_bo_list(*mem, avm->process_info); +err_pin_bo: drm_vma_node_revoke(&gobj->vma_node, drm_priv); err_node_allow: drm_gem_object_put(gobj); /* Don't unreserve system mem limit twice */ goto err_reserve_limit; err_bo_create: - unreserve_mem_limit(adev, size, alloc_domain, !!sg); + unreserve_mem_limit(adev, size, flags); err_reserve_limit: mutex_destroy(&(*mem)->lock); kfree(*mem); @@ -1529,7 +1599,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( } int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv, + struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv, uint64_t *size) { struct amdkfd_process_info *process_info = mem->process_info; @@ -1542,6 +1612,14 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( bool is_imported = false; mutex_lock(&mem->lock); + + /* Unpin MMIO/DOORBELL BO's that were pinnned during allocation */ + if (mem->alloc_flags & + (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL | + KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) { + amdgpu_amdkfd_gpuvm_unpin_bo(mem->bo); + } + mapped_to_gpu_memory = mem->mapped_to_gpu_memory; is_imported = mem->is_imported; mutex_unlock(&mem->lock); @@ -1621,10 +1699,9 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( } int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, + struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv, bool *table_freed) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); int ret; struct amdgpu_bo *bo; @@ -1751,7 +1828,7 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( } int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv) + struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv) { struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); struct amdkfd_process_info *process_info = avm->process_info; @@ -1812,7 +1889,7 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( } int amdgpu_amdkfd_gpuvm_sync_memory( - struct kgd_dev *kgd, struct kgd_mem *mem, bool intr) + struct amdgpu_device *adev, struct kgd_mem *mem, bool intr) { struct amdgpu_sync sync; int ret; @@ -1828,7 +1905,7 @@ int amdgpu_amdkfd_gpuvm_sync_memory( return ret; } -int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd, +int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct amdgpu_device *adev, struct kgd_mem *mem, void **kptr, uint64_t *size) { int ret; @@ -1884,7 +1961,8 @@ int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd, return ret; } -void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_dev *kgd, struct kgd_mem *mem) +void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct amdgpu_device *adev, + struct kgd_mem *mem) { struct amdgpu_bo *bo = mem->bo; @@ -1894,12 +1972,9 @@ void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_dev *kgd, struct kg amdgpu_bo_unreserve(bo); } -int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, - struct kfd_vm_fault_info *mem) +int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev, + struct kfd_vm_fault_info *mem) { - struct amdgpu_device *adev; - - adev = (struct amdgpu_device *)kgd; if (atomic_read(&adev->gmc.vm_fault_info_updated) == 1) { *mem = *adev->gmc.vm_fault_info; mb(); @@ -1908,13 +1983,12 @@ int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, return 0; } -int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd, +int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev, struct dma_buf *dma_buf, uint64_t va, void *drm_priv, struct kgd_mem **mem, uint64_t *size, uint64_t *mmap_offset) { - struct amdgpu_device *adev = (struct amdgpu_device *)kgd; struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); struct drm_gem_object *obj; struct amdgpu_bo *bo; @@ -2541,11 +2615,9 @@ int amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem) } /* Returns GPU-specific tiling mode information */ -int amdgpu_amdkfd_get_tile_config(struct kgd_dev *kgd, +int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev, struct tile_config *config) { - struct amdgpu_device *adev = (struct amdgpu_device *)kgd; - config->gb_addr_config = adev->gfx.config.gb_addr_config; config->tile_config_ptr = adev->gfx.config.tile_mode_array; config->num_tile_configs = diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c index 96b7bb13a2dd..12a6b1c99c93 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c @@ -1569,6 +1569,18 @@ void amdgpu_atombios_scratch_regs_engine_hung(struct amdgpu_device *adev, WREG32(adev->bios_scratch_reg_offset + 3, tmp); } +void amdgpu_atombios_scratch_regs_set_backlight_level(struct amdgpu_device *adev, + u32 backlight_level) +{ + u32 tmp = RREG32(adev->bios_scratch_reg_offset + 2); + + tmp &= ~ATOM_S2_CURRENT_BL_LEVEL_MASK; + tmp |= (backlight_level << ATOM_S2_CURRENT_BL_LEVEL_SHIFT) & + ATOM_S2_CURRENT_BL_LEVEL_MASK; + + WREG32(adev->bios_scratch_reg_offset + 2, tmp); +} + bool amdgpu_atombios_scratch_need_asic_init(struct amdgpu_device *adev) { u32 tmp = RREG32(adev->bios_scratch_reg_offset + 7); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h index 8cc0222dba19..27e74b1fc260 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h @@ -185,6 +185,8 @@ bool amdgpu_atombios_has_gpu_virtualization_table(struct amdgpu_device *adev); void amdgpu_atombios_scratch_regs_lock(struct amdgpu_device *adev, bool lock); void amdgpu_atombios_scratch_regs_engine_hung(struct amdgpu_device *adev, bool hung); +void amdgpu_atombios_scratch_regs_set_backlight_level(struct amdgpu_device *adev, + u32 backlight_level); bool amdgpu_atombios_scratch_need_asic_init(struct amdgpu_device *adev); void amdgpu_atombios_copy_swap(u8 *dst, u8 *src, u8 num_bytes, bool to_le); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 188accb71249..6acdf7cfe0a4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3687,8 +3687,6 @@ int amdgpu_device_init(struct amdgpu_device *adev, /* Get a log2 for easy divisions. */ adev->mm_stats.log2_max_MBps = ilog2(max(1u, max_MBps)); - amdgpu_fbdev_init(adev); - r = amdgpu_pm_sysfs_init(adev); if (r) { adev->pm_sysfs_en = false; @@ -3846,8 +3844,6 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev) amdgpu_ucode_sysfs_fini(adev); sysfs_remove_files(&adev->dev->kobj, amdgpu_dev_attributes); - amdgpu_fbdev_fini(adev); - amdgpu_device_ip_fini_early(adev); amdgpu_irq_fini_hw(adev); @@ -3942,7 +3938,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool fbcon) drm_kms_helper_poll_disable(dev); if (fbcon) - amdgpu_fbdev_set_suspend(adev, 1); + drm_fb_helper_set_suspend_unlocked(adev_to_drm(adev)->fb_helper, true); cancel_delayed_work_sync(&adev->delayed_init_work); @@ -4019,7 +4015,7 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon) flush_delayed_work(&adev->delayed_init_work); if (fbcon) - amdgpu_fbdev_set_suspend(adev, 0); + drm_fb_helper_set_suspend_unlocked(adev_to_drm(adev)->fb_helper, false); drm_kms_helper_poll_enable(dev); @@ -4316,7 +4312,6 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, amdgpu_irq_gpu_reset_resume_helper(adev); r = amdgpu_ib_ring_tests(adev); - amdgpu_amdkfd_post_reset(adev); error: if (!r && adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST) { @@ -4649,7 +4644,7 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle, if (r) goto out; - amdgpu_fbdev_set_suspend(tmp_adev, 0); + drm_fb_helper_set_suspend_unlocked(adev_to_drm(tmp_adev)->fb_helper, false); /* * The GPU enters bad state once faulty pages @@ -5039,7 +5034,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, */ amdgpu_unregister_gpu_instance(tmp_adev); - amdgpu_fbdev_set_suspend(tmp_adev, 1); + drm_fb_helper_set_suspend_unlocked(adev_to_drm(adev)->fb_helper, true); /* disable ras on ALL IPs */ if (!need_emergency_restart && @@ -5089,7 +5084,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, tmp_vram_lost_counter = atomic_read(&((adev)->vram_lost_counter)); /* Actual ASIC resets if needed.*/ - /* TODO Implement XGMI hive reset logic for SRIOV */ + /* Host driver will handle XGMI hive reset for SRIOV */ if (amdgpu_sriov_vf(adev)) { r = amdgpu_device_reset_sriov(adev, job ? false : true); if (r) @@ -5149,8 +5144,8 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, skip_sched_resume: list_for_each_entry(tmp_adev, device_list_handle, reset_list) { - /* unlock kfd: SRIOV would do it separately */ - if (!need_emergency_restart && !amdgpu_sriov_vf(tmp_adev)) + /* unlock kfd */ + if (!need_emergency_restart) amdgpu_amdkfd_post_reset(tmp_adev); /* kfd_post_reset will do nothing if kfd device is not initialized, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 4e3669407518..503995c7ff6c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -248,8 +248,8 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) offset = offsetof(struct binary_header, binary_checksum) + sizeof(bhdr->binary_checksum); - size = bhdr->binary_size - offset; - checksum = bhdr->binary_checksum; + size = le16_to_cpu(bhdr->binary_size) - offset; + checksum = le16_to_cpu(bhdr->binary_checksum); if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, size, checksum)) { @@ -270,7 +270,7 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) } if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, - ihdr->size, checksum)) { + le16_to_cpu(ihdr->size), checksum)) { DRM_ERROR("invalid ip discovery data table checksum\n"); r = -EINVAL; goto out; @@ -282,7 +282,7 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) ghdr = (struct gpu_info_header *)(adev->mman.discovery_bin + offset); if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset, - ghdr->size, checksum)) { + le32_to_cpu(ghdr->size), checksum)) { DRM_ERROR("invalid gc data table checksum\n"); r = -EINVAL; goto out; @@ -489,10 +489,10 @@ void amdgpu_discovery_harvest_ip(struct amdgpu_device *adev) le16_to_cpu(bhdr->table_list[HARVEST_INFO].offset)); for (i = 0; i < 32; i++) { - if (le32_to_cpu(harvest_info->list[i].hw_id) == 0) + if (le16_to_cpu(harvest_info->list[i].hw_id) == 0) break; - switch (le32_to_cpu(harvest_info->list[i].hw_id)) { + switch (le16_to_cpu(harvest_info->list[i].hw_id)) { case VCN_HWID: vcn_harvest_count++; if (harvest_info->list[i].number_instance == 0) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index 68108f151dad..18cc7155e667 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -1599,13 +1599,10 @@ int amdgpu_display_suspend_helper(struct amdgpu_device *adev) continue; } robj = gem_to_amdgpu_bo(fb->obj[0]); - /* don't unpin kernel fb objects */ - if (!amdgpu_fbdev_robj_is_fb(adev, robj)) { - r = amdgpu_bo_reserve(robj, true); - if (r == 0) { - amdgpu_bo_unpin(robj); - amdgpu_bo_unreserve(robj); - } + r = amdgpu_bo_reserve(robj, true); + if (r == 0) { + amdgpu_bo_unpin(robj); + amdgpu_bo_unreserve(robj); } } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 438468b82eb6..3a6f125c6dc9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -2001,6 +2001,19 @@ static int amdgpu_pci_probe(struct pci_dev *pdev, goto err_pci; } + /* + * 1. don't init fbdev on hw without DCE + * 2. don't init fbdev if there are no connectors + */ + if (adev->mode_info.mode_config_initialized && + !list_empty(&adev_to_drm(adev)->mode_config.connector_list)) { + /* select 8 bpp console on low vram cards */ + if (adev->gmc.real_vram_size <= (32*1024*1024)) + drm_fbdev_generic_setup(adev_to_drm(adev), 8); + else + drm_fbdev_generic_setup(adev_to_drm(adev), 32); + } + ret = amdgpu_debugfs_init(adev); if (ret) DRM_ERROR("Creating debugfs files failed (%d).\n", ret); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c deleted file mode 100644 index cd0acbea75da..000000000000 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +++ /dev/null @@ -1,388 +0,0 @@ -/* - * Copyright © 2007 David Airlie - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS 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. - * - * Authors: - * David Airlie - */ - -#include -#include -#include -#include - -#include -#include -#include -#include -#include - -#include "amdgpu.h" -#include "cikd.h" -#include "amdgpu_gem.h" - -#include "amdgpu_display.h" - -/* object hierarchy - - this contains a helper + a amdgpu fb - the helper contains a pointer to amdgpu framebuffer baseclass. -*/ - -static int -amdgpufb_open(struct fb_info *info, int user) -{ - struct drm_fb_helper *fb_helper = info->par; - int ret = pm_runtime_get_sync(fb_helper->dev->dev); - if (ret < 0 && ret != -EACCES) { - pm_runtime_mark_last_busy(fb_helper->dev->dev); - pm_runtime_put_autosuspend(fb_helper->dev->dev); - return ret; - } - return 0; -} - -static int -amdgpufb_release(struct fb_info *info, int user) -{ - struct drm_fb_helper *fb_helper = info->par; - - pm_runtime_mark_last_busy(fb_helper->dev->dev); - pm_runtime_put_autosuspend(fb_helper->dev->dev); - return 0; -} - -static const struct fb_ops amdgpufb_ops = { - .owner = THIS_MODULE, - DRM_FB_HELPER_DEFAULT_OPS, - .fb_open = amdgpufb_open, - .fb_release = amdgpufb_release, - .fb_fillrect = drm_fb_helper_cfb_fillrect, - .fb_copyarea = drm_fb_helper_cfb_copyarea, - .fb_imageblit = drm_fb_helper_cfb_imageblit, -}; - - -int amdgpu_align_pitch(struct amdgpu_device *adev, int width, int cpp, bool tiled) -{ - int aligned = width; - int pitch_mask = 0; - - switch (cpp) { - case 1: - pitch_mask = 255; - break; - case 2: - pitch_mask = 127; - break; - case 3: - case 4: - pitch_mask = 63; - break; - } - - aligned += pitch_mask; - aligned &= ~pitch_mask; - return aligned * cpp; -} - -static void amdgpufb_destroy_pinned_object(struct drm_gem_object *gobj) -{ - struct amdgpu_bo *abo = gem_to_amdgpu_bo(gobj); - int ret; - - ret = amdgpu_bo_reserve(abo, true); - if (likely(ret == 0)) { - amdgpu_bo_kunmap(abo); - amdgpu_bo_unpin(abo); - amdgpu_bo_unreserve(abo); - } - drm_gem_object_put(gobj); -} - -static int amdgpufb_create_pinned_object(struct amdgpu_fbdev *rfbdev, - struct drm_mode_fb_cmd2 *mode_cmd, - struct drm_gem_object **gobj_p) -{ - const struct drm_format_info *info; - struct amdgpu_device *adev = rfbdev->adev; - struct drm_gem_object *gobj = NULL; - struct amdgpu_bo *abo = NULL; - bool fb_tiled = false; /* useful for testing */ - u32 tiling_flags = 0, domain; - int ret; - int aligned_size, size; - int height = mode_cmd->height; - u32 cpp; - u64 flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | - AMDGPU_GEM_CREATE_VRAM_CLEARED; - - info = drm_get_format_info(adev_to_drm(adev), mode_cmd); - cpp = info->cpp[0]; - - /* need to align pitch with crtc limits */ - mode_cmd->pitches[0] = amdgpu_align_pitch(adev, mode_cmd->width, cpp, - fb_tiled); - domain = amdgpu_display_supported_domains(adev, flags); - height = ALIGN(mode_cmd->height, 8); - size = mode_cmd->pitches[0] * height; - aligned_size = ALIGN(size, PAGE_SIZE); - ret = amdgpu_gem_object_create(adev, aligned_size, 0, domain, flags, - ttm_bo_type_device, NULL, &gobj); - if (ret) { - pr_err("failed to allocate framebuffer (%d)\n", aligned_size); - return -ENOMEM; - } - abo = gem_to_amdgpu_bo(gobj); - - if (fb_tiled) - tiling_flags = AMDGPU_TILING_SET(ARRAY_MODE, GRPH_ARRAY_2D_TILED_THIN1); - - ret = amdgpu_bo_reserve(abo, false); - if (unlikely(ret != 0)) - goto out_unref; - - if (tiling_flags) { - ret = amdgpu_bo_set_tiling_flags(abo, - tiling_flags); - if (ret) - dev_err(adev->dev, "FB failed to set tiling flags\n"); - } - - ret = amdgpu_bo_pin(abo, domain); - if (ret) { - amdgpu_bo_unreserve(abo); - goto out_unref; - } - - ret = amdgpu_ttm_alloc_gart(&abo->tbo); - if (ret) { - amdgpu_bo_unreserve(abo); - dev_err(adev->dev, "%p bind failed\n", abo); - goto out_unref; - } - - ret = amdgpu_bo_kmap(abo, NULL); - amdgpu_bo_unreserve(abo); - if (ret) { - goto out_unref; - } - - *gobj_p = gobj; - return 0; -out_unref: - amdgpufb_destroy_pinned_object(gobj); - *gobj_p = NULL; - return ret; -} - -static int amdgpufb_create(struct drm_fb_helper *helper, - struct drm_fb_helper_surface_size *sizes) -{ - struct amdgpu_fbdev *rfbdev = (struct amdgpu_fbdev *)helper; - struct amdgpu_device *adev = rfbdev->adev; - struct fb_info *info; - struct drm_framebuffer *fb = NULL; - struct drm_mode_fb_cmd2 mode_cmd; - struct drm_gem_object *gobj = NULL; - struct amdgpu_bo *abo = NULL; - int ret; - - memset(&mode_cmd, 0, sizeof(mode_cmd)); - mode_cmd.width = sizes->surface_width; - mode_cmd.height = sizes->surface_height; - - if (sizes->surface_bpp == 24) - sizes->surface_bpp = 32; - - mode_cmd.pixel_format = drm_mode_legacy_fb_format(sizes->surface_bpp, - sizes->surface_depth); - - ret = amdgpufb_create_pinned_object(rfbdev, &mode_cmd, &gobj); - if (ret) { - DRM_ERROR("failed to create fbcon object %d\n", ret); - return ret; - } - - abo = gem_to_amdgpu_bo(gobj); - - /* okay we have an object now allocate the framebuffer */ - info = drm_fb_helper_alloc_fbi(helper); - if (IS_ERR(info)) { - ret = PTR_ERR(info); - goto out; - } - - ret = amdgpu_display_gem_fb_init(adev_to_drm(adev), &rfbdev->rfb, - &mode_cmd, gobj); - if (ret) { - DRM_ERROR("failed to initialize framebuffer %d\n", ret); - goto out; - } - - fb = &rfbdev->rfb.base; - - /* setup helper */ - rfbdev->helper.fb = fb; - - info->fbops = &amdgpufb_ops; - - info->fix.smem_start = amdgpu_gmc_vram_cpu_pa(adev, abo); - info->fix.smem_len = amdgpu_bo_size(abo); - info->screen_base = amdgpu_bo_kptr(abo); - info->screen_size = amdgpu_bo_size(abo); - - drm_fb_helper_fill_info(info, &rfbdev->helper, sizes); - - /* setup aperture base/size for vesafb takeover */ - info->apertures->ranges[0].base = adev_to_drm(adev)->mode_config.fb_base; - info->apertures->ranges[0].size = adev->gmc.aper_size; - - /* Use default scratch pixmap (info->pixmap.flags = FB_PIXMAP_SYSTEM) */ - - if (info->screen_base == NULL) { - ret = -ENOSPC; - goto out; - } - - DRM_INFO("fb mappable at 0x%lX\n", info->fix.smem_start); - DRM_INFO("vram apper at 0x%lX\n", (unsigned long)adev->gmc.aper_base); - DRM_INFO("size %lu\n", (unsigned long)amdgpu_bo_size(abo)); - DRM_INFO("fb depth is %d\n", fb->format->depth); - DRM_INFO(" pitch is %d\n", fb->pitches[0]); - - vga_switcheroo_client_fb_set(adev->pdev, info); - return 0; - -out: - if (fb && ret) { - drm_gem_object_put(gobj); - drm_framebuffer_unregister_private(fb); - drm_framebuffer_cleanup(fb); - kfree(fb); - } - return ret; -} - -static int amdgpu_fbdev_destroy(struct drm_device *dev, struct amdgpu_fbdev *rfbdev) -{ - struct amdgpu_framebuffer *rfb = &rfbdev->rfb; - int i; - - drm_fb_helper_unregister_fbi(&rfbdev->helper); - - if (rfb->base.obj[0]) { - for (i = 0; i < rfb->base.format->num_planes; i++) - drm_gem_object_put(rfb->base.obj[0]); - amdgpufb_destroy_pinned_object(rfb->base.obj[0]); - rfb->base.obj[0] = NULL; - drm_framebuffer_unregister_private(&rfb->base); - drm_framebuffer_cleanup(&rfb->base); - } - drm_fb_helper_fini(&rfbdev->helper); - - return 0; -} - -static const struct drm_fb_helper_funcs amdgpu_fb_helper_funcs = { - .fb_probe = amdgpufb_create, -}; - -int amdgpu_fbdev_init(struct amdgpu_device *adev) -{ - struct amdgpu_fbdev *rfbdev; - int bpp_sel = 32; - int ret; - - /* don't init fbdev on hw without DCE */ - if (!adev->mode_info.mode_config_initialized) - return 0; - - /* don't init fbdev if there are no connectors */ - if (list_empty(&adev_to_drm(adev)->mode_config.connector_list)) - return 0; - - /* select 8 bpp console on low vram cards */ - if (adev->gmc.real_vram_size <= (32*1024*1024)) - bpp_sel = 8; - - rfbdev = kzalloc(sizeof(struct amdgpu_fbdev), GFP_KERNEL); - if (!rfbdev) - return -ENOMEM; - - rfbdev->adev = adev; - adev->mode_info.rfbdev = rfbdev; - - drm_fb_helper_prepare(adev_to_drm(adev), &rfbdev->helper, - &amdgpu_fb_helper_funcs); - - ret = drm_fb_helper_init(adev_to_drm(adev), &rfbdev->helper); - if (ret) { - kfree(rfbdev); - return ret; - } - - /* disable all the possible outputs/crtcs before entering KMS mode */ - if (!amdgpu_device_has_dc_support(adev) && !amdgpu_virtual_display) - drm_helper_disable_unused_functions(adev_to_drm(adev)); - - drm_fb_helper_initial_config(&rfbdev->helper, bpp_sel); - return 0; -} - -void amdgpu_fbdev_fini(struct amdgpu_device *adev) -{ - if (!adev->mode_info.rfbdev) - return; - - amdgpu_fbdev_destroy(adev_to_drm(adev), adev->mode_info.rfbdev); - kfree(adev->mode_info.rfbdev); - adev->mode_info.rfbdev = NULL; -} - -void amdgpu_fbdev_set_suspend(struct amdgpu_device *adev, int state) -{ - if (adev->mode_info.rfbdev) - drm_fb_helper_set_suspend_unlocked(&adev->mode_info.rfbdev->helper, - state); -} - -int amdgpu_fbdev_total_size(struct amdgpu_device *adev) -{ - struct amdgpu_bo *robj; - int size = 0; - - if (!adev->mode_info.rfbdev) - return 0; - - robj = gem_to_amdgpu_bo(adev->mode_info.rfbdev->rfb.base.obj[0]); - size += amdgpu_bo_size(robj); - return size; -} - -bool amdgpu_fbdev_robj_is_fb(struct amdgpu_device *adev, struct amdgpu_bo *robj) -{ - if (!adev->mode_info.rfbdev) - return false; - if (robj == gem_to_amdgpu_bo(adev->mode_info.rfbdev->rfb.base.obj[0])) - return true; - return false; -} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index a1e63ba4c54a..c0d8f40a5b45 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -877,6 +877,32 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, return r; } +static int amdgpu_gem_align_pitch(struct amdgpu_device *adev, + int width, + int cpp, + bool tiled) +{ + int aligned = width; + int pitch_mask = 0; + + switch (cpp) { + case 1: + pitch_mask = 255; + break; + case 2: + pitch_mask = 127; + break; + case 3: + case 4: + pitch_mask = 63; + break; + } + + aligned += pitch_mask; + aligned &= ~pitch_mask; + return aligned * cpp; +} + int amdgpu_mode_dumb_create(struct drm_file *file_priv, struct drm_device *dev, struct drm_mode_create_dumb *args) @@ -885,7 +911,8 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv, struct drm_gem_object *gobj; uint32_t handle; u64 flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_CPU_GTT_USWC; + AMDGPU_GEM_CREATE_CPU_GTT_USWC | + AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; u32 domain; int r; @@ -897,8 +924,8 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv, if (adev->mman.buffer_funcs_enabled) flags |= AMDGPU_GEM_CREATE_VRAM_CLEARED; - args->pitch = amdgpu_align_pitch(adev, args->width, - DIV_ROUND_UP(args->bpp, 8), 0); + args->pitch = amdgpu_gem_align_pitch(adev, args->width, + DIV_ROUND_UP(args->bpp, 8), 0); args->size = (u64)args->pitch * args->height; args->size = ALIGN(args->size, PAGE_SIZE); domain = amdgpu_bo_get_preferred_domain(adev, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c index f3d62e196901..0c7963dfacad 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c @@ -223,7 +223,7 @@ int amdgpu_ih_wait_on_checkpoint_process(struct amdgpu_device *adev, */ int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih) { - unsigned int count = AMDGPU_IH_MAX_NUM_IVS; + unsigned int count; u32 wptr; if (!ih->enabled || adev->shutdown) @@ -232,6 +232,7 @@ int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih) wptr = amdgpu_ih_get_wptr(adev, ih); restart_ih: + count = AMDGPU_IH_MAX_NUM_IVS; DRM_DEBUG("%s: rptr %d, wptr %d\n", __func__, ih->rptr, wptr); /* Order reading of wptr vs. reading of IH ring data */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c index cc2e0c9cfe0a..4f3c62adccbd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c @@ -333,7 +333,6 @@ int amdgpu_irq_init(struct amdgpu_device *adev) if (!amdgpu_device_has_dc_support(adev)) { if (!adev->enable_virtual_display) /* Disable vblank IRQs aggressively for power-saving */ - /* XXX: can this be enabled for DC? */ adev_to_drm(adev)->vblank_disable_immediate = true; r = drm_vblank_init(adev_to_drm(adev), adev->mode_info.num_crtc); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index 89fb372ed49c..6043bf6fd414 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -232,8 +232,6 @@ struct amdgpu_i2c_chan { struct mutex mutex; }; -struct amdgpu_fbdev; - struct amdgpu_afmt { bool enabled; int offset; @@ -309,13 +307,6 @@ struct amdgpu_framebuffer { uint64_t address; }; -struct amdgpu_fbdev { - struct drm_fb_helper helper; - struct amdgpu_framebuffer rfb; - struct list_head fbdev_list; - struct amdgpu_device *adev; -}; - struct amdgpu_mode_info { struct atom_context *atom_context; struct card_info *atom_card_info; @@ -341,8 +332,6 @@ struct amdgpu_mode_info { struct edid *bios_hardcoded_edid; int bios_hardcoded_edid_size; - /* pointer to fbdev info structure */ - struct amdgpu_fbdev *rfbdev; /* firmware flags */ u32 firmware_flags; /* pointer to backlight encoder */ @@ -631,15 +620,6 @@ bool amdgpu_crtc_get_scanout_position(struct drm_crtc *crtc, int *hpos, ktime_t *stime, ktime_t *etime, const struct drm_display_mode *mode); -/* fbdev layer */ -int amdgpu_fbdev_init(struct amdgpu_device *adev); -void amdgpu_fbdev_fini(struct amdgpu_device *adev); -void amdgpu_fbdev_set_suspend(struct amdgpu_device *adev, int state); -int amdgpu_fbdev_total_size(struct amdgpu_device *adev); -bool amdgpu_fbdev_robj_is_fb(struct amdgpu_device *adev, struct amdgpu_bo *robj); - -int amdgpu_align_pitch(struct amdgpu_device *adev, int width, int bpp, bool tiled); - /* amdgpu_display.c */ void amdgpu_display_print_display_setup(struct drm_device *dev); int amdgpu_display_modeset_create_props(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 4fcfc2313b8c..3a7b56e57cec 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -1032,9 +1032,14 @@ int amdgpu_bo_init(struct amdgpu_device *adev) /* On A+A platform, VRAM can be mapped as WB */ if (!adev->gmc.xgmi.connected_to_cpu) { /* reserve PAT memory space to WC for VRAM */ - arch_io_reserve_memtype_wc(adev->gmc.aper_base, + int r = arch_io_reserve_memtype_wc(adev->gmc.aper_base, adev->gmc.aper_size); + if (r) { + DRM_ERROR("Unable to set WC memtype for the aperture base\n"); + return r; + } + /* Add an MTRR for the VRAM */ adev->gmc.vram_mtrr = arch_phys_wc_add(adev->gmc.aper_base, adev->gmc.aper_size); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 08133de21fdd..46910e7b2927 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -892,6 +892,38 @@ void amdgpu_ras_mca_query_error_status(struct amdgpu_device *adev, } } +static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_data *err_data) +{ + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + int ret = 0; + + /* + * choosing right query method according to + * whether smu support query error information + */ + ret = smu_get_ecc_info(&adev->smu, (void *)&(ras->umc_ecc)); + if (ret == -EOPNOTSUPP) { + if (adev->umc.ras_funcs && + adev->umc.ras_funcs->query_ras_error_count) + adev->umc.ras_funcs->query_ras_error_count(adev, err_data); + + /* umc query_ras_error_address is also responsible for clearing + * error status + */ + if (adev->umc.ras_funcs && + adev->umc.ras_funcs->query_ras_error_address) + adev->umc.ras_funcs->query_ras_error_address(adev, err_data); + } else if (!ret) { + if (adev->umc.ras_funcs && + adev->umc.ras_funcs->ecc_info_query_ras_error_count) + adev->umc.ras_funcs->ecc_info_query_ras_error_count(adev, err_data); + + if (adev->umc.ras_funcs && + adev->umc.ras_funcs->ecc_info_query_ras_error_address) + adev->umc.ras_funcs->ecc_info_query_ras_error_address(adev, err_data); + } +} + /* query/inject/cure begin */ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, struct ras_query_if *info) @@ -905,15 +937,7 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, switch (info->head.block) { case AMDGPU_RAS_BLOCK__UMC: - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_count) - adev->umc.ras_funcs->query_ras_error_count(adev, &err_data); - /* umc query_ras_error_address is also responsible for clearing - * error status - */ - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_address) - adev->umc.ras_funcs->query_ras_error_address(adev, &err_data); + amdgpu_ras_get_ecc_info(adev, &err_data); break; case AMDGPU_RAS_BLOCK__SDMA: if (adev->sdma.funcs->query_ras_error_count) { @@ -1935,9 +1959,11 @@ int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev) if (!con || !con->eh_data) return 0; + mutex_lock(&con->recovery_lock); control = &con->eeprom_control; data = con->eh_data; save_count = data->count - control->ras_num_recs; + mutex_unlock(&con->recovery_lock); /* only new entries are saved */ if (save_count > 0) { if (amdgpu_ras_eeprom_append(control, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index e36f4de9fa55..1c708122d492 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -319,6 +319,19 @@ struct ras_common_if { char name[32]; }; +#define MAX_UMC_CHANNEL_NUM 32 + +struct ecc_info_per_ch { + uint16_t ce_count_lo_chip; + uint16_t ce_count_hi_chip; + uint64_t mca_umc_status; + uint64_t mca_umc_addr; +}; + +struct umc_ecc_info { + struct ecc_info_per_ch ecc[MAX_UMC_CHANNEL_NUM]; +}; + struct amdgpu_ras { /* ras infrastructure */ /* for ras itself. */ @@ -358,6 +371,9 @@ struct amdgpu_ras { struct delayed_work ras_counte_delay_work; atomic_t ras_ue_count; atomic_t ras_ce_count; + + /* record umc error info queried from smu */ + struct umc_ecc_info umc_ecc; }; struct ras_fs_data { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index c15687ce67c4..fb0d8bffdce2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -913,11 +913,6 @@ static int amdgpu_ttm_backend_bind(struct ttm_device *bdev, ttm->num_pages, bo_mem, ttm); } - if (bo_mem->mem_type == AMDGPU_PL_GDS || - bo_mem->mem_type == AMDGPU_PL_GWS || - bo_mem->mem_type == AMDGPU_PL_OA) - return -EINVAL; - if (bo_mem->mem_type != TTM_PL_TT || !amdgpu_gtt_mgr_has_gart_addr(bo_mem)) { gtt->offset = AMDGPU_BO_INVALID_OFFSET; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c index a90029ee9733..6e4bea012ea4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c @@ -94,30 +94,58 @@ int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev, { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; struct amdgpu_ras *con = amdgpu_ras_get_context(adev); + int ret = 0; kgd2kfd_set_sram_ecc_flag(adev->kfd.dev); - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_count) - adev->umc.ras_funcs->query_ras_error_count(adev, ras_error_status); + ret = smu_get_ecc_info(&adev->smu, (void *)&(con->umc_ecc)); + if (ret == -EOPNOTSUPP) { + if (adev->umc.ras_funcs && + adev->umc.ras_funcs->query_ras_error_count) + adev->umc.ras_funcs->query_ras_error_count(adev, ras_error_status); - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_address && - adev->umc.max_ras_err_cnt_per_query) { - err_data->err_addr = - kcalloc(adev->umc.max_ras_err_cnt_per_query, - sizeof(struct eeprom_table_record), GFP_KERNEL); + if (adev->umc.ras_funcs && + adev->umc.ras_funcs->query_ras_error_address && + adev->umc.max_ras_err_cnt_per_query) { + err_data->err_addr = + kcalloc(adev->umc.max_ras_err_cnt_per_query, + sizeof(struct eeprom_table_record), GFP_KERNEL); - /* still call query_ras_error_address to clear error status - * even NOMEM error is encountered - */ - if(!err_data->err_addr) - dev_warn(adev->dev, "Failed to alloc memory for " - "umc error address record!\n"); + /* still call query_ras_error_address to clear error status + * even NOMEM error is encountered + */ + if(!err_data->err_addr) + dev_warn(adev->dev, "Failed to alloc memory for " + "umc error address record!\n"); - /* umc query_ras_error_address is also responsible for clearing - * error status - */ - adev->umc.ras_funcs->query_ras_error_address(adev, ras_error_status); + /* umc query_ras_error_address is also responsible for clearing + * error status + */ + adev->umc.ras_funcs->query_ras_error_address(adev, ras_error_status); + } + } else if (!ret) { + if (adev->umc.ras_funcs && + adev->umc.ras_funcs->ecc_info_query_ras_error_count) + adev->umc.ras_funcs->ecc_info_query_ras_error_count(adev, ras_error_status); + + if (adev->umc.ras_funcs && + adev->umc.ras_funcs->ecc_info_query_ras_error_address && + adev->umc.max_ras_err_cnt_per_query) { + err_data->err_addr = + kcalloc(adev->umc.max_ras_err_cnt_per_query, + sizeof(struct eeprom_table_record), GFP_KERNEL); + + /* still call query_ras_error_address to clear error status + * even NOMEM error is encountered + */ + if(!err_data->err_addr) + dev_warn(adev->dev, "Failed to alloc memory for " + "umc error address record!\n"); + + /* umc query_ras_error_address is also responsible for clearing + * error status + */ + adev->umc.ras_funcs->ecc_info_query_ras_error_address(adev, ras_error_status); + } } /* only uncorrectable error needs gpu reset */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h index 1f5fe2315236..9e40bade0a68 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h @@ -49,6 +49,10 @@ struct amdgpu_umc_ras_funcs { void (*query_ras_error_address)(struct amdgpu_device *adev, void *ras_error_status); bool (*query_ras_poison_mode)(struct amdgpu_device *adev); + void (*ecc_info_query_ras_error_count)(struct amdgpu_device *adev, + void *ras_error_status); + void (*ecc_info_query_ras_error_address)(struct amdgpu_device *adev, + void *ras_error_status); }; struct amdgpu_umc_funcs { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c index 04cf9b207e62..3fc49823f527 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c @@ -283,17 +283,15 @@ static int amdgpu_virt_init_ras_err_handler_data(struct amdgpu_device *adev) *data = kmalloc(sizeof(struct amdgpu_virt_ras_err_handler_data), GFP_KERNEL); if (!*data) - return -ENOMEM; + goto data_failure; bps = kmalloc_array(align_space, sizeof((*data)->bps), GFP_KERNEL); - bps_bo = kmalloc_array(align_space, sizeof((*data)->bps_bo), GFP_KERNEL); + if (!bps) + goto bps_failure; - if (!bps || !bps_bo) { - kfree(bps); - kfree(bps_bo); - kfree(*data); - return -ENOMEM; - } + bps_bo = kmalloc_array(align_space, sizeof((*data)->bps_bo), GFP_KERNEL); + if (!bps_bo) + goto bps_bo_failure; (*data)->bps = bps; (*data)->bps_bo = bps_bo; @@ -303,6 +301,13 @@ static int amdgpu_virt_init_ras_err_handler_data(struct amdgpu_device *adev) virt->ras_init_done = true; return 0; + +bps_bo_failure: + kfree(bps); +bps_failure: + kfree(*data); +data_failure: + return -ENOMEM; } static void amdgpu_virt_ras_release_bp(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index b200b9e722d9..8318ee8339f1 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -2092,22 +2092,18 @@ static int dce_v8_0_pick_dig_encoder(struct drm_encoder *encoder) return 1; else return 0; - break; case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1: if (dig->linkb) return 3; else return 2; - break; case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2: if (dig->linkb) return 5; else return 4; - break; case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: return 6; - break; default: DRM_ERROR("invalid encoder_id: 0x%x\n", amdgpu_encoder->encoder_id); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c index e7dfeb466a0e..dbe7442fb25c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c @@ -7707,8 +7707,19 @@ static uint64_t gfx_v10_0_get_gpu_clock_counter(struct amdgpu_device *adev) switch (adev->ip_versions[GC_HWIP][0]) { case IP_VERSION(10, 3, 1): case IP_VERSION(10, 3, 3): - clock = (uint64_t)RREG32_SOC15(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Vangogh) | - ((uint64_t)RREG32_SOC15(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Vangogh) << 32ULL); + preempt_disable(); + clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Vangogh); + clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Vangogh); + hi_check = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Vangogh); + /* The SMUIO TSC clock frequency is 100MHz, which sets 32-bit carry over + * roughly every 42 seconds. + */ + if (hi_check != clock_hi) { + clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Vangogh); + clock_hi = hi_check; + } + preempt_enable(); + clock = clock_lo | (clock_hi << 32ULL); break; default: preempt_disable(); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index b4b80f27b894..34478bcc4d09 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -140,6 +140,11 @@ MODULE_FIRMWARE("amdgpu/aldebaran_rlc.bin"); #define mmTCP_CHAN_STEER_5_ARCT 0x0b0c #define mmTCP_CHAN_STEER_5_ARCT_BASE_IDX 0 +#define mmGOLDEN_TSC_COUNT_UPPER_Renoir 0x0025 +#define mmGOLDEN_TSC_COUNT_UPPER_Renoir_BASE_IDX 1 +#define mmGOLDEN_TSC_COUNT_LOWER_Renoir 0x0026 +#define mmGOLDEN_TSC_COUNT_LOWER_Renoir_BASE_IDX 1 + enum ta_ras_gfx_subblock { /*CPC*/ TA_RAS_BLOCK__GFX_CPC_INDEX_START = 0, @@ -4238,19 +4243,38 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev) { - uint64_t clock; + uint64_t clock, clock_lo, clock_hi, hi_check; - amdgpu_gfx_off_ctrl(adev, false); - mutex_lock(&adev->gfx.gpu_clock_mutex); - if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 0, 1) && amdgpu_sriov_runtime(adev)) { - clock = gfx_v9_0_kiq_read_clock(adev); - } else { - WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1); - clock = (uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_LSB) | - ((uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_MSB) << 32ULL); + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 3, 0): + preempt_disable(); + clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Renoir); + clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Renoir); + hi_check = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Renoir); + /* The SMUIO TSC clock frequency is 100MHz, which sets 32-bit carry over + * roughly every 42 seconds. + */ + if (hi_check != clock_hi) { + clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Renoir); + clock_hi = hi_check; + } + preempt_enable(); + clock = clock_lo | (clock_hi << 32ULL); + break; + default: + amdgpu_gfx_off_ctrl(adev, false); + mutex_lock(&adev->gfx.gpu_clock_mutex); + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 0, 1) && amdgpu_sriov_runtime(adev)) { + clock = gfx_v9_0_kiq_read_clock(adev); + } else { + WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1); + clock = (uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_LSB) | + ((uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_MSB) << 32ULL); + } + mutex_unlock(&adev->gfx.gpu_clock_mutex); + amdgpu_gfx_off_ctrl(adev, true); + break; } - mutex_unlock(&adev->gfx.gpu_clock_mutex); - amdgpu_gfx_off_ctrl(adev, true); return clock; } diff --git a/drivers/gpu/drm/amd/amdgpu/navi10_ih.c b/drivers/gpu/drm/amd/amdgpu/navi10_ih.c index 1d8414c3fadb..38241cf0e1f1 100644 --- a/drivers/gpu/drm/amd/amdgpu/navi10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/navi10_ih.c @@ -160,6 +160,7 @@ static int navi10_ih_toggle_ring_interrupts(struct amdgpu_device *adev, tmp = RREG32(ih_regs->ih_rb_cntl); tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0)); + tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_GPU_TS_ENABLE, 1); /* enable_intr field is only valid in ring0 */ if (ih == &adev->irq.ih) tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0)); @@ -275,10 +276,8 @@ static int navi10_ih_enable_ring(struct amdgpu_device *adev, tmp = navi10_ih_rb_cntl(ih, tmp); if (ih == &adev->irq.ih) tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RPTR_REARM, !!adev->irq.msi_enabled); - if (ih == &adev->irq.ih1) { - tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_ENABLE, 0); + if (ih == &adev->irq.ih1) tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_FULL_DRAIN_ENABLE, 1); - } if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) { if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp)) { @@ -319,7 +318,6 @@ static int navi10_ih_irq_init(struct amdgpu_device *adev) { struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1, &adev->irq.ih2}; u32 ih_chicken; - u32 tmp; int ret; int i; @@ -363,15 +361,6 @@ static int navi10_ih_irq_init(struct amdgpu_device *adev) adev->nbio.funcs->ih_doorbell_range(adev, ih[0]->use_doorbell, ih[0]->doorbell_index); - tmp = RREG32_SOC15(OSSSYS, 0, mmIH_STORM_CLIENT_LIST_CNTL); - tmp = REG_SET_FIELD(tmp, IH_STORM_CLIENT_LIST_CNTL, - CLIENT18_IS_STORM_CLIENT, 1); - WREG32_SOC15(OSSSYS, 0, mmIH_STORM_CLIENT_LIST_CNTL, tmp); - - tmp = RREG32_SOC15(OSSSYS, 0, mmIH_INT_FLOOD_CNTL); - tmp = REG_SET_FIELD(tmp, IH_INT_FLOOD_CNTL, FLOOD_CNTL_ENABLE, 1); - WREG32_SOC15(OSSSYS, 0, mmIH_INT_FLOOD_CNTL, tmp); - pci_set_master(adev->pdev); /* enable interrupts */ @@ -420,12 +409,19 @@ static u32 navi10_ih_get_wptr(struct amdgpu_device *adev, u32 wptr, tmp; struct amdgpu_ih_regs *ih_regs; - wptr = le32_to_cpu(*ih->wptr_cpu); + if (ih == &adev->irq.ih) { + /* Only ring0 supports writeback. On other rings fall back + * to register-based code with overflow checking below. + */ + wptr = le32_to_cpu(*ih->wptr_cpu); + + if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) + goto out; + } + ih_regs = &ih->ih_regs; - if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) - goto out; - + /* Double check that the overflow wasn't already cleared. */ wptr = RREG32_NO_KIQ(ih_regs->ih_rb_wptr); if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) goto out; @@ -513,15 +509,11 @@ static int navi10_ih_self_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - uint32_t wptr = cpu_to_le32(entry->src_data[0]); - switch (entry->ring_id) { case 1: - *adev->irq.ih1.wptr_cpu = wptr; schedule_work(&adev->irq.ih1_work); break; case 2: - *adev->irq.ih2.wptr_cpu = wptr; schedule_work(&adev->irq.ih2_work); break; default: break; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c b/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c index 4ecd2b5808ce..ee7cab37dfd5 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c @@ -359,6 +359,10 @@ static void nbio_v2_3_init_registers(struct amdgpu_device *adev) if (def != data) WREG32_PCIE(smnPCIE_CONFIG_CNTL, data); + + if (amdgpu_sriov_vf(adev)) + adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0, + mmBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2; } #define NAVI10_PCIE__LC_L0S_INACTIVITY_DEFAULT 0x00000000 // off by default, no gains over L1 diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c index 0d2d629e2d6a..4bbacf1be25a 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c @@ -276,6 +276,10 @@ static void nbio_v6_1_init_registers(struct amdgpu_device *adev) if (def != data) WREG32_PCIE(smnPCIE_CI_CNTL, data); + + if (amdgpu_sriov_vf(adev)) + adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0, + mmBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2; } static void nbio_v6_1_program_ltr(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c index 3c00666a13e1..37a4039fdfc5 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c @@ -273,7 +273,9 @@ const struct nbio_hdp_flush_reg nbio_v7_0_hdp_flush_reg = { static void nbio_v7_0_init_registers(struct amdgpu_device *adev) { - + if (amdgpu_sriov_vf(adev)) + adev->rmmio_remap.reg_offset = + SOC15_REG_OFFSET(NBIO, 0, mmHDP_MEM_COHERENCY_FLUSH_CNTL) << 2; } const struct amdgpu_nbio_funcs nbio_v7_0_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c index 8f2a315e7c73..3444332ea110 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c @@ -371,6 +371,10 @@ static void nbio_v7_2_init_registers(struct amdgpu_device *adev) if (def != data) WREG32_PCIE_PORT(SOC15_REG_OFFSET(NBIO, 0, regPCIE_CONFIG_CNTL), data); } + + if (amdgpu_sriov_vf(adev)) + adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0, + regBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2; } const struct amdgpu_nbio_funcs nbio_v7_2_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c index b8bd03d16dba..dc5e93756fea 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -362,7 +362,9 @@ const struct nbio_hdp_flush_reg nbio_v7_4_hdp_flush_reg_ald = { static void nbio_v7_4_init_registers(struct amdgpu_device *adev) { - + if (amdgpu_sriov_vf(adev)) + adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0, + mmBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2; } static void nbio_v7_4_handle_ras_controller_intr_no_bifring(struct amdgpu_device *adev) @@ -692,6 +694,9 @@ static void nbio_v7_4_program_aspm(struct amdgpu_device *adev) { uint32_t def, data; + if (adev->ip_versions[NBIO_HWIP][0] == IP_VERSION(7, 4, 4)) + return; + def = data = RREG32_PCIE(smnPCIE_LC_CNTL); data &= ~PCIE_LC_CNTL__LC_L1_INACTIVITY_MASK; data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK; diff --git a/drivers/gpu/drm/amd/amdgpu/nv.c b/drivers/gpu/drm/amd/amdgpu/nv.c index 59eafa31c626..a6659d9ecdd2 100644 --- a/drivers/gpu/drm/amd/amdgpu/nv.c +++ b/drivers/gpu/drm/amd/amdgpu/nv.c @@ -731,8 +731,10 @@ static int nv_common_early_init(void *handle) #define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE) struct amdgpu_device *adev = (struct amdgpu_device *)handle; - adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET; - adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET; + if (!amdgpu_sriov_vf(adev)) { + adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET; + adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET; + } adev->smc_rreg = NULL; adev->smc_wreg = NULL; adev->pcie_rreg = &nv_pcie_rreg; @@ -1032,7 +1034,7 @@ static int nv_common_hw_init(void *handle) * for the purpose of expose those registers * to process space */ - if (adev->nbio.funcs->remap_hdp_registers) + if (adev->nbio.funcs->remap_hdp_registers && !amdgpu_sriov_vf(adev)) adev->nbio.funcs->remap_hdp_registers(adev); /* enable the doorbell aperture */ nv_enable_doorbell_aperture(adev, true); diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 0c316a2d42ed..de9b55383e9f 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -971,8 +971,10 @@ static int soc15_common_early_init(void *handle) #define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE) struct amdgpu_device *adev = (struct amdgpu_device *)handle; - adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET; - adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET; + if (!amdgpu_sriov_vf(adev)) { + adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET; + adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET; + } adev->smc_rreg = NULL; adev->smc_wreg = NULL; adev->pcie_rreg = &soc15_pcie_rreg; @@ -1285,7 +1287,7 @@ static int soc15_common_hw_init(void *handle) * for the purpose of expose those registers * to process space */ - if (adev->nbio.funcs->remap_hdp_registers) + if (adev->nbio.funcs->remap_hdp_registers && !amdgpu_sriov_vf(adev)) adev->nbio.funcs->remap_hdp_registers(adev); /* enable the doorbell aperture */ diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c index f7ec3fe134e5..6dd1e19e8d43 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c @@ -50,6 +50,165 @@ static inline uint32_t get_umc_v6_7_reg_offset(struct amdgpu_device *adev, return adev->umc.channel_offs * ch_inst + UMC_V6_7_INST_DIST * umc_inst; } +static inline uint32_t get_umc_v6_7_channel_index(struct amdgpu_device *adev, + uint32_t umc_inst, + uint32_t ch_inst) +{ + return adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; +} + +static void umc_v6_7_ecc_info_query_correctable_error_count(struct amdgpu_device *adev, + uint32_t channel_index, + unsigned long *error_count) +{ + uint32_t ecc_err_cnt; + uint64_t mc_umc_status; + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + + /* + * select the lower chip and check the error count + * skip add error count, calc error counter only from mca_umc_status + */ + ecc_err_cnt = ras->umc_ecc.ecc[channel_index].ce_count_lo_chip; + + /* + * select the higher chip and check the err counter + * skip add error count, calc error counter only from mca_umc_status + */ + ecc_err_cnt = ras->umc_ecc.ecc[channel_index].ce_count_hi_chip; + + /* check for SRAM correctable error + MCUMC_STATUS is a 64 bit register */ + mc_umc_status = ras->umc_ecc.ecc[channel_index].mca_umc_status; + if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1) + *error_count += 1; +} + +static void umc_v6_7_ecc_info_querry_uncorrectable_error_count(struct amdgpu_device *adev, + uint32_t channel_index, + unsigned long *error_count) +{ + uint64_t mc_umc_status; + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + + /* check the MCUMC_STATUS */ + mc_umc_status = ras->umc_ecc.ecc[channel_index].mca_umc_status; + if ((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1) && + (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, PCC) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UC) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, TCC) == 1)) + *error_count += 1; +} + +static void umc_v6_7_ecc_info_query_ras_error_count(struct amdgpu_device *adev, + void *ras_error_status) +{ + struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; + + uint32_t umc_inst = 0; + uint32_t ch_inst = 0; + uint32_t umc_reg_offset = 0; + uint32_t channel_index = 0; + + /*TODO: driver needs to toggle DF Cstate to ensure + * safe access of UMC registers. Will add the protection */ + LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { + umc_reg_offset = get_umc_v6_7_reg_offset(adev, + umc_inst, + ch_inst); + channel_index = get_umc_v6_7_channel_index(adev, + umc_inst, + ch_inst); + umc_v6_7_ecc_info_query_correctable_error_count(adev, + channel_index, + &(err_data->ce_count)); + umc_v6_7_ecc_info_querry_uncorrectable_error_count(adev, + channel_index, + &(err_data->ue_count)); + } +} + +static void umc_v6_7_ecc_info_query_error_address(struct amdgpu_device *adev, + struct ras_err_data *err_data, + uint32_t umc_reg_offset, + uint32_t ch_inst, + uint32_t umc_inst) +{ + uint64_t mc_umc_status, err_addr, retired_page; + struct eeprom_table_record *err_rec; + uint32_t channel_index; + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + + channel_index = + adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; + + mc_umc_status = ras->umc_ecc.ecc[channel_index].mca_umc_status; + + if (mc_umc_status == 0) + return; + + if (!err_data->err_addr) + return; + + err_rec = &err_data->err_addr[err_data->err_addr_cnt]; + + /* calculate error address if ue/ce error is detected */ + if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && + (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || + REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1)) { + + err_addr = ras->umc_ecc.ecc[channel_index].mca_umc_addr; + err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr); + + /* translate umc channel address to soc pa, 3 parts are included */ + retired_page = ADDR_OF_8KB_BLOCK(err_addr) | + ADDR_OF_256B_BLOCK(channel_index) | + OFFSET_IN_256B_BLOCK(err_addr); + + /* we only save ue error information currently, ce is skipped */ + if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) + == 1) { + err_rec->address = err_addr; + /* page frame address is saved */ + err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; + err_rec->ts = (uint64_t)ktime_get_real_seconds(); + err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; + err_rec->cu = 0; + err_rec->mem_channel = channel_index; + err_rec->mcumc_id = umc_inst; + + err_data->err_addr_cnt++; + } + } +} + +static void umc_v6_7_ecc_info_query_ras_error_address(struct amdgpu_device *adev, + void *ras_error_status) +{ + struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; + + uint32_t umc_inst = 0; + uint32_t ch_inst = 0; + uint32_t umc_reg_offset = 0; + + /*TODO: driver needs to toggle DF Cstate to ensure + * safe access of UMC resgisters. Will add the protection + * when firmware interface is ready */ + LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { + umc_reg_offset = get_umc_v6_7_reg_offset(adev, + umc_inst, + ch_inst); + umc_v6_7_ecc_info_query_error_address(adev, + err_data, + umc_reg_offset, + ch_inst, + umc_inst); + } +} + static void umc_v6_7_query_correctable_error_count(struct amdgpu_device *adev, uint32_t umc_reg_offset, unsigned long *error_count) @@ -327,4 +486,6 @@ const struct amdgpu_umc_ras_funcs umc_v6_7_ras_funcs = { .query_ras_error_count = umc_v6_7_query_ras_error_count, .query_ras_error_address = umc_v6_7_query_ras_error_address, .query_ras_poison_mode = umc_v6_7_query_ras_poison_mode, + .ecc_info_query_ras_error_count = umc_v6_7_ecc_info_query_ras_error_count, + .ecc_info_query_ras_error_address = umc_v6_7_ecc_info_query_ras_error_address, }; diff --git a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c index f6233019f042..d60576ce10cd 100644 --- a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c +++ b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c @@ -43,15 +43,15 @@ static bool cik_event_interrupt_isr(struct kfd_dev *dev, */ if ((ihre->source_id == CIK_INTSRC_GFX_PAGE_INV_FAULT || ihre->source_id == CIK_INTSRC_GFX_MEM_PROT_FAULT) && - dev->device_info->asic_family == CHIP_HAWAII) { + dev->adev->asic_type == CHIP_HAWAII) { struct cik_ih_ring_entry *tmp_ihre = (struct cik_ih_ring_entry *)patched_ihre; *patched_flag = true; *tmp_ihre = *ihre; - vmid = f2g->read_vmid_from_vmfault_reg(dev->kgd); - ret = f2g->get_atc_vmid_pasid_mapping_info(dev->kgd, vmid, &pasid); + vmid = f2g->read_vmid_from_vmfault_reg(dev->adev); + ret = f2g->get_atc_vmid_pasid_mapping_info(dev->adev, vmid, &pasid); tmp_ihre->ring_id &= 0x000000ff; tmp_ihre->ring_id |= vmid << 8; @@ -113,7 +113,7 @@ static void cik_event_interrupt_wq(struct kfd_dev *dev, kfd_process_vm_fault(dev->dqm, pasid); memset(&info, 0, sizeof(info)); - amdgpu_amdkfd_gpuvm_get_vm_fault_info(dev->kgd, &info); + amdgpu_amdkfd_gpuvm_get_vm_fault_info(dev->adev, &info); if (!info.page_addr && !info.status) return; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 24ebd61395d8..4bfc0c8ab764 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -321,7 +321,7 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p, /* Return gpu_id as doorbell offset for mmap usage */ args->doorbell_offset = KFD_MMAP_TYPE_DOORBELL; args->doorbell_offset |= KFD_MMAP_GPU_ID(args->gpu_id); - if (KFD_IS_SOC15(dev->device_info->asic_family)) + if (KFD_IS_SOC15(dev)) /* On SOC15 ASICs, include the doorbell offset within the * process doorbell frame, which is 2 pages. */ @@ -580,7 +580,7 @@ static int kfd_ioctl_dbg_register(struct file *filep, if (!dev) return -EINVAL; - if (dev->device_info->asic_family == CHIP_CARRIZO) { + if (dev->adev->asic_type == CHIP_CARRIZO) { pr_debug("kfd_ioctl_dbg_register not supported on CZ\n"); return -EINVAL; } @@ -631,7 +631,7 @@ static int kfd_ioctl_dbg_unregister(struct file *filep, if (!dev || !dev->dbgmgr) return -EINVAL; - if (dev->device_info->asic_family == CHIP_CARRIZO) { + if (dev->adev->asic_type == CHIP_CARRIZO) { pr_debug("kfd_ioctl_dbg_unregister not supported on CZ\n"); return -EINVAL; } @@ -676,7 +676,7 @@ static int kfd_ioctl_dbg_address_watch(struct file *filep, if (!dev) return -EINVAL; - if (dev->device_info->asic_family == CHIP_CARRIZO) { + if (dev->adev->asic_type == CHIP_CARRIZO) { pr_debug("kfd_ioctl_dbg_wave_control not supported on CZ\n"); return -EINVAL; } @@ -784,7 +784,7 @@ static int kfd_ioctl_dbg_wave_control(struct file *filep, if (!dev) return -EINVAL; - if (dev->device_info->asic_family == CHIP_CARRIZO) { + if (dev->adev->asic_type == CHIP_CARRIZO) { pr_debug("kfd_ioctl_dbg_wave_control not supported on CZ\n"); return -EINVAL; } @@ -851,7 +851,7 @@ static int kfd_ioctl_get_clock_counters(struct file *filep, dev = kfd_device_by_id(args->gpu_id); if (dev) /* Reading GPU clock counter from KGD */ - args->gpu_clock_counter = amdgpu_amdkfd_get_gpu_clock_counter(dev->kgd); + args->gpu_clock_counter = amdgpu_amdkfd_get_gpu_clock_counter(dev->adev); else /* Node without GPU resource */ args->gpu_clock_counter = 0; @@ -1041,7 +1041,7 @@ static int kfd_ioctl_create_event(struct file *filp, struct kfd_process *p, goto out_unlock; } - err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kfd->kgd, + err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kfd->adev, mem, &kern_addr, &size); if (err) { pr_err("Failed to map event page to kernel\n"); @@ -1051,7 +1051,7 @@ static int kfd_ioctl_create_event(struct file *filp, struct kfd_process *p, err = kfd_event_page_set(p, kern_addr, size); if (err) { pr_err("Failed to set event page\n"); - amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(kfd->kgd, mem); + amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(kfd->adev, mem); goto out_unlock; } @@ -1137,7 +1137,7 @@ static int kfd_ioctl_set_scratch_backing_va(struct file *filep, if (dev->dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS && pdd->qpd.vmid != 0 && dev->kfd2kgd->set_scratch_backing_va) dev->kfd2kgd->set_scratch_backing_va( - dev->kgd, args->va_addr, pdd->qpd.vmid); + dev->adev, args->va_addr, pdd->qpd.vmid); return 0; @@ -1158,7 +1158,7 @@ static int kfd_ioctl_get_tile_config(struct file *filep, if (!dev) return -EINVAL; - amdgpu_amdkfd_get_tile_config(dev->kgd, &config); + amdgpu_amdkfd_get_tile_config(dev->adev, &config); args->gb_addr_config = config.gb_addr_config; args->num_banks = config.num_banks; @@ -1244,7 +1244,7 @@ bool kfd_dev_is_large_bar(struct kfd_dev *dev) if (dev->use_iommu_v2) return false; - amdgpu_amdkfd_get_local_mem_info(dev->kgd, &mem_info); + amdgpu_amdkfd_get_local_mem_info(dev->adev, &mem_info); if (mem_info.local_mem_size_private == 0 && mem_info.local_mem_size_public > 0) return true; @@ -1313,7 +1313,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep, err = -EINVAL; goto err_unlock; } - offset = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->kgd); + offset = dev->adev->rmmio_remap.bus_addr; if (!offset) { err = -ENOMEM; goto err_unlock; @@ -1321,7 +1321,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep, } err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( - dev->kgd, args->va_addr, args->size, + dev->adev, args->va_addr, args->size, pdd->drm_priv, (struct kgd_mem **) &mem, &offset, flags); @@ -1353,7 +1353,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep, return 0; err_free: - amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem, + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev, (struct kgd_mem *)mem, pdd->drm_priv, NULL); err_unlock: mutex_unlock(&p->mutex); @@ -1399,7 +1399,7 @@ static int kfd_ioctl_free_memory_of_gpu(struct file *filep, goto err_unlock; } - ret = amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, + ret = amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev, (struct kgd_mem *)mem, pdd->drm_priv, &size); /* If freeing the buffer failed, leave the handle in place for @@ -1484,7 +1484,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep, goto get_mem_obj_from_handle_failed; } err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu( - peer->kgd, (struct kgd_mem *)mem, + peer->adev, (struct kgd_mem *)mem, peer_pdd->drm_priv, &table_freed); if (err) { pr_err("Failed to map to gpu %d/%d\n", @@ -1496,7 +1496,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep, mutex_unlock(&p->mutex); - err = amdgpu_amdkfd_gpuvm_sync_memory(dev->kgd, (struct kgd_mem *) mem, true); + err = amdgpu_amdkfd_gpuvm_sync_memory(dev->adev, (struct kgd_mem *) mem, true); if (err) { pr_debug("Sync memory failed, wait interrupted by user signal\n"); goto sync_memory_failed; @@ -1593,7 +1593,7 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep, goto get_mem_obj_from_handle_failed; } err = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( - peer->kgd, (struct kgd_mem *)mem, peer_pdd->drm_priv); + peer->adev, (struct kgd_mem *)mem, peer_pdd->drm_priv); if (err) { pr_err("Failed to unmap from gpu %d/%d\n", i, args->n_devices); @@ -1603,8 +1603,8 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep, } mutex_unlock(&p->mutex); - if (dev->device_info->asic_family == CHIP_ALDEBARAN) { - err = amdgpu_amdkfd_gpuvm_sync_memory(dev->kgd, + if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 2)) { + err = amdgpu_amdkfd_gpuvm_sync_memory(dev->adev, (struct kgd_mem *) mem, true); if (err) { pr_debug("Sync memory failed, wait interrupted by user signal\n"); @@ -1680,7 +1680,7 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep, { struct kfd_ioctl_get_dmabuf_info_args *args = data; struct kfd_dev *dev = NULL; - struct kgd_dev *dma_buf_kgd; + struct amdgpu_device *dmabuf_adev; void *metadata_buffer = NULL; uint32_t flags; unsigned int i; @@ -1700,15 +1700,15 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep, } /* Get dmabuf info from KGD */ - r = amdgpu_amdkfd_get_dmabuf_info(dev->kgd, args->dmabuf_fd, - &dma_buf_kgd, &args->size, + r = amdgpu_amdkfd_get_dmabuf_info(dev->adev, args->dmabuf_fd, + &dmabuf_adev, &args->size, metadata_buffer, args->metadata_size, &args->metadata_size, &flags); if (r) goto exit; /* Reverse-lookup gpu_id from kgd pointer */ - dev = kfd_device_by_kgd(dma_buf_kgd); + dev = kfd_device_by_adev(dmabuf_adev); if (!dev) { r = -EINVAL; goto exit; @@ -1758,7 +1758,7 @@ static int kfd_ioctl_import_dmabuf(struct file *filep, goto err_unlock; } - r = amdgpu_amdkfd_gpuvm_import_dmabuf(dev->kgd, dmabuf, + r = amdgpu_amdkfd_gpuvm_import_dmabuf(dev->adev, dmabuf, args->va_addr, pdd->drm_priv, (struct kgd_mem **)&mem, &size, NULL); @@ -1779,7 +1779,7 @@ static int kfd_ioctl_import_dmabuf(struct file *filep, return 0; err_free: - amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem, + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev, (struct kgd_mem *)mem, pdd->drm_priv, NULL); err_unlock: mutex_unlock(&p->mutex); @@ -2066,7 +2066,7 @@ static int kfd_mmio_mmap(struct kfd_dev *dev, struct kfd_process *process, if (vma->vm_end - vma->vm_start != PAGE_SIZE) return -EINVAL; - address = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->kgd); + address = dev->adev->rmmio_remap.bus_addr; vma->vm_flags |= VM_IO | VM_DONTCOPY | VM_DONTEXPAND | VM_NORESERVE | VM_DONTDUMP | VM_PFNMAP; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c index cfedfb1e8596..f187596faf66 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c @@ -1340,7 +1340,7 @@ static int kfd_fill_gpu_cache_info(struct kfd_dev *kdev, int ret; unsigned int num_cu_shared; - switch (kdev->device_info->asic_family) { + switch (kdev->adev->asic_type) { case CHIP_KAVERI: pcache_info = kaveri_cache_info; num_of_cache_types = ARRAY_SIZE(kaveri_cache_info); @@ -1377,67 +1377,71 @@ static int kfd_fill_gpu_cache_info(struct kfd_dev *kdev, pcache_info = vegam_cache_info; num_of_cache_types = ARRAY_SIZE(vegam_cache_info); break; - case CHIP_VEGA10: - pcache_info = vega10_cache_info; - num_of_cache_types = ARRAY_SIZE(vega10_cache_info); - break; - case CHIP_VEGA12: - pcache_info = vega12_cache_info; - num_of_cache_types = ARRAY_SIZE(vega12_cache_info); - break; - case CHIP_VEGA20: - case CHIP_ARCTURUS: - pcache_info = vega20_cache_info; - num_of_cache_types = ARRAY_SIZE(vega20_cache_info); - break; - case CHIP_ALDEBARAN: - pcache_info = aldebaran_cache_info; - num_of_cache_types = ARRAY_SIZE(aldebaran_cache_info); - break; - case CHIP_RAVEN: - pcache_info = raven_cache_info; - num_of_cache_types = ARRAY_SIZE(raven_cache_info); - break; - case CHIP_RENOIR: - pcache_info = renoir_cache_info; - num_of_cache_types = ARRAY_SIZE(renoir_cache_info); - break; - case CHIP_NAVI10: - case CHIP_NAVI12: - case CHIP_CYAN_SKILLFISH: - pcache_info = navi10_cache_info; - num_of_cache_types = ARRAY_SIZE(navi10_cache_info); - break; - case CHIP_NAVI14: - pcache_info = navi14_cache_info; - num_of_cache_types = ARRAY_SIZE(navi14_cache_info); - break; - case CHIP_SIENNA_CICHLID: - pcache_info = sienna_cichlid_cache_info; - num_of_cache_types = ARRAY_SIZE(sienna_cichlid_cache_info); - break; - case CHIP_NAVY_FLOUNDER: - pcache_info = navy_flounder_cache_info; - num_of_cache_types = ARRAY_SIZE(navy_flounder_cache_info); - break; - case CHIP_DIMGREY_CAVEFISH: - pcache_info = dimgrey_cavefish_cache_info; - num_of_cache_types = ARRAY_SIZE(dimgrey_cavefish_cache_info); - break; - case CHIP_VANGOGH: - pcache_info = vangogh_cache_info; - num_of_cache_types = ARRAY_SIZE(vangogh_cache_info); - break; - case CHIP_BEIGE_GOBY: - pcache_info = beige_goby_cache_info; - num_of_cache_types = ARRAY_SIZE(beige_goby_cache_info); - break; - case CHIP_YELLOW_CARP: - pcache_info = yellow_carp_cache_info; - num_of_cache_types = ARRAY_SIZE(yellow_carp_cache_info); - break; default: - return -EINVAL; + switch(KFD_GC_VERSION(kdev)) { + case IP_VERSION(9, 0, 1): + pcache_info = vega10_cache_info; + num_of_cache_types = ARRAY_SIZE(vega10_cache_info); + break; + case IP_VERSION(9, 2, 1): + pcache_info = vega12_cache_info; + num_of_cache_types = ARRAY_SIZE(vega12_cache_info); + break; + case IP_VERSION(9, 4, 0): + case IP_VERSION(9, 4, 1): + pcache_info = vega20_cache_info; + num_of_cache_types = ARRAY_SIZE(vega20_cache_info); + break; + case IP_VERSION(9, 4, 2): + pcache_info = aldebaran_cache_info; + num_of_cache_types = ARRAY_SIZE(aldebaran_cache_info); + break; + case IP_VERSION(9, 1, 0): + case IP_VERSION(9, 2, 2): + pcache_info = raven_cache_info; + num_of_cache_types = ARRAY_SIZE(raven_cache_info); + break; + case IP_VERSION(9, 3, 0): + pcache_info = renoir_cache_info; + num_of_cache_types = ARRAY_SIZE(renoir_cache_info); + break; + case IP_VERSION(10, 1, 10): + case IP_VERSION(10, 1, 2): + case IP_VERSION(10, 1, 3): + pcache_info = navi10_cache_info; + num_of_cache_types = ARRAY_SIZE(navi10_cache_info); + break; + case IP_VERSION(10, 1, 1): + pcache_info = navi14_cache_info; + num_of_cache_types = ARRAY_SIZE(navi14_cache_info); + break; + case IP_VERSION(10, 3, 0): + pcache_info = sienna_cichlid_cache_info; + num_of_cache_types = ARRAY_SIZE(sienna_cichlid_cache_info); + break; + case IP_VERSION(10, 3, 2): + pcache_info = navy_flounder_cache_info; + num_of_cache_types = ARRAY_SIZE(navy_flounder_cache_info); + break; + case IP_VERSION(10, 3, 4): + pcache_info = dimgrey_cavefish_cache_info; + num_of_cache_types = ARRAY_SIZE(dimgrey_cavefish_cache_info); + break; + case IP_VERSION(10, 3, 1): + pcache_info = vangogh_cache_info; + num_of_cache_types = ARRAY_SIZE(vangogh_cache_info); + break; + case IP_VERSION(10, 3, 5): + pcache_info = beige_goby_cache_info; + num_of_cache_types = ARRAY_SIZE(beige_goby_cache_info); + break; + case IP_VERSION(10, 3, 3): + pcache_info = yellow_carp_cache_info; + num_of_cache_types = ARRAY_SIZE(yellow_carp_cache_info); + break; + default: + return -EINVAL; + } } *size_filled = 0; @@ -1963,8 +1967,6 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size, struct crat_subtype_iolink *sub_type_hdr, uint32_t proximity_domain) { - struct amdgpu_device *adev = (struct amdgpu_device *)kdev->kgd; - *avail_size -= sizeof(struct crat_subtype_iolink); if (*avail_size < 0) return -ENOMEM; @@ -1981,7 +1983,7 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size, /* Fill in IOLINK subtype. * TODO: Fill-in other fields of iolink subtype */ - if (adev->gmc.xgmi.connected_to_cpu) { + if (kdev->adev->gmc.xgmi.connected_to_cpu) { /* * with host gpu xgmi link, host can access gpu memory whether * or not pcie bar type is large, so always create bidirectional @@ -1990,19 +1992,19 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size, sub_type_hdr->flags |= CRAT_IOLINK_FLAGS_BI_DIRECTIONAL; sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_XGMI; sub_type_hdr->num_hops_xgmi = 1; - if (adev->asic_type == CHIP_ALDEBARAN) { + if (KFD_GC_VERSION(kdev) == IP_VERSION(9, 4, 2)) { sub_type_hdr->minimum_bandwidth_mbs = amdgpu_amdkfd_get_xgmi_bandwidth_mbytes( - kdev->kgd, NULL, true); + kdev->adev, NULL, true); sub_type_hdr->maximum_bandwidth_mbs = sub_type_hdr->minimum_bandwidth_mbs; } } else { sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_PCIEXPRESS; sub_type_hdr->minimum_bandwidth_mbs = - amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->kgd, true); + amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->adev, true); sub_type_hdr->maximum_bandwidth_mbs = - amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->kgd, false); + amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->adev, false); } sub_type_hdr->proximity_domain_from = proximity_domain; @@ -2044,11 +2046,11 @@ static int kfd_fill_gpu_xgmi_link_to_gpu(int *avail_size, sub_type_hdr->proximity_domain_from = proximity_domain_from; sub_type_hdr->proximity_domain_to = proximity_domain_to; sub_type_hdr->num_hops_xgmi = - amdgpu_amdkfd_get_xgmi_hops_count(kdev->kgd, peer_kdev->kgd); + amdgpu_amdkfd_get_xgmi_hops_count(kdev->adev, peer_kdev->adev); sub_type_hdr->maximum_bandwidth_mbs = - amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->kgd, peer_kdev->kgd, false); + amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->adev, peer_kdev->adev, false); sub_type_hdr->minimum_bandwidth_mbs = sub_type_hdr->maximum_bandwidth_mbs ? - amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->kgd, NULL, true) : 0; + amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->adev, NULL, true) : 0; return 0; } @@ -2114,7 +2116,7 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image, cu->flags |= CRAT_CU_FLAGS_GPU_PRESENT; cu->proximity_domain = proximity_domain; - amdgpu_amdkfd_get_cu_info(kdev->kgd, &cu_info); + amdgpu_amdkfd_get_cu_info(kdev->adev, &cu_info); cu->num_simd_per_cu = cu_info.simd_per_cu; cu->num_simd_cores = cu_info.simd_per_cu * cu_info.cu_active_number; cu->max_waves_simd = cu_info.max_waves_per_simd; @@ -2145,7 +2147,7 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image, * report the total FB size (public+private) as a single * private heap. */ - amdgpu_amdkfd_get_local_mem_info(kdev->kgd, &local_mem_info); + amdgpu_amdkfd_get_local_mem_info(kdev->adev, &local_mem_info); sub_type_hdr = (typeof(sub_type_hdr))((char *)sub_type_hdr + sub_type_hdr->length); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c b/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c index 159add0f5aaa..1e30717b5253 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c @@ -41,7 +41,7 @@ static void dbgdev_address_watch_disable_nodiq(struct kfd_dev *dev) { - dev->kfd2kgd->address_watch_disable(dev->kgd); + dev->kfd2kgd->address_watch_disable(dev->adev); } static int dbgdev_diq_submit_ib(struct kfd_dbgdev *dbgdev, @@ -322,7 +322,7 @@ static int dbgdev_address_watch_nodiq(struct kfd_dbgdev *dbgdev, pr_debug("\t\t%30s\n", "* * * * * * * * * * * * * * * * * *"); pdd->dev->kfd2kgd->address_watch_execute( - dbgdev->dev->kgd, + dbgdev->dev->adev, i, cntl.u32All, addrHi.u32All, @@ -420,7 +420,7 @@ static int dbgdev_address_watch_diq(struct kfd_dbgdev *dbgdev, aw_reg_add_dword = dbgdev->dev->kfd2kgd->address_watch_get_offset( - dbgdev->dev->kgd, + dbgdev->dev->adev, i, ADDRESS_WATCH_REG_CNTL); @@ -431,7 +431,7 @@ static int dbgdev_address_watch_diq(struct kfd_dbgdev *dbgdev, aw_reg_add_dword = dbgdev->dev->kfd2kgd->address_watch_get_offset( - dbgdev->dev->kgd, + dbgdev->dev->adev, i, ADDRESS_WATCH_REG_ADDR_HI); @@ -441,7 +441,7 @@ static int dbgdev_address_watch_diq(struct kfd_dbgdev *dbgdev, aw_reg_add_dword = dbgdev->dev->kfd2kgd->address_watch_get_offset( - dbgdev->dev->kgd, + dbgdev->dev->adev, i, ADDRESS_WATCH_REG_ADDR_LO); @@ -457,7 +457,7 @@ static int dbgdev_address_watch_diq(struct kfd_dbgdev *dbgdev, aw_reg_add_dword = dbgdev->dev->kfd2kgd->address_watch_get_offset( - dbgdev->dev->kgd, + dbgdev->dev->adev, i, ADDRESS_WATCH_REG_CNTL); @@ -752,7 +752,7 @@ static int dbgdev_wave_control_nodiq(struct kfd_dbgdev *dbgdev, pr_debug("\t\t %30s\n", "* * * * * * * * * * * * * * * * * *"); - return dbgdev->dev->kfd2kgd->wave_control_execute(dbgdev->dev->kgd, + return dbgdev->dev->kfd2kgd->wave_control_execute(dbgdev->dev->adev, reg_gfx_index.u32All, reg_sq_cmd.u32All); } @@ -784,7 +784,7 @@ int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process *p) for (vmid = first_vmid_to_scan; vmid <= last_vmid_to_scan; vmid++) { status = dev->kfd2kgd->get_atc_vmid_pasid_mapping_info - (dev->kgd, vmid, &queried_pasid); + (dev->adev, vmid, &queried_pasid); if (status && queried_pasid == p->pasid) { pr_debug("Killing wave fronts of vmid %d and pasid 0x%x\n", @@ -811,7 +811,7 @@ int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process *p) /* for non DIQ we need to patch the VMID: */ reg_sq_cmd.bits.vm_id = vmid; - dev->kfd2kgd->wave_control_execute(dev->kgd, + dev->kfd2kgd->wave_control_execute(dev->adev, reg_gfx_index.u32All, reg_sq_cmd.u32All); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index 3b119db16003..e1294fba0c26 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -55,7 +55,6 @@ extern const struct kfd2kgd_calls gfx_v10_3_kfd2kgd; #ifdef KFD_SUPPORT_IOMMU_V2 static const struct kfd_device_info kaveri_device_info = { - .asic_family = CHIP_KAVERI, .asic_name = "kaveri", .gfx_target_version = 70000, .max_pasid_bits = 16, @@ -69,13 +68,10 @@ static const struct kfd_device_info kaveri_device_info = { .supports_cwsr = false, .needs_iommu_device = true, .needs_pci_atomics = false, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info carrizo_device_info = { - .asic_family = CHIP_CARRIZO, .asic_name = "carrizo", .gfx_target_version = 80001, .max_pasid_bits = 16, @@ -89,13 +85,10 @@ static const struct kfd_device_info carrizo_device_info = { .supports_cwsr = true, .needs_iommu_device = true, .needs_pci_atomics = false, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info raven_device_info = { - .asic_family = CHIP_RAVEN, .asic_name = "raven", .gfx_target_version = 90002, .max_pasid_bits = 16, @@ -108,15 +101,12 @@ static const struct kfd_device_info raven_device_info = { .supports_cwsr = true, .needs_iommu_device = true, .needs_pci_atomics = true, - .num_sdma_engines = 1, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; #endif #ifdef CONFIG_DRM_AMDGPU_CIK static const struct kfd_device_info hawaii_device_info = { - .asic_family = CHIP_HAWAII, .asic_name = "hawaii", .gfx_target_version = 70001, .max_pasid_bits = 16, @@ -130,14 +120,11 @@ static const struct kfd_device_info hawaii_device_info = { .supports_cwsr = false, .needs_iommu_device = false, .needs_pci_atomics = false, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; #endif static const struct kfd_device_info tonga_device_info = { - .asic_family = CHIP_TONGA, .asic_name = "tonga", .gfx_target_version = 80002, .max_pasid_bits = 16, @@ -150,13 +137,10 @@ static const struct kfd_device_info tonga_device_info = { .supports_cwsr = false, .needs_iommu_device = false, .needs_pci_atomics = true, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info fiji_device_info = { - .asic_family = CHIP_FIJI, .asic_name = "fiji", .gfx_target_version = 80003, .max_pasid_bits = 16, @@ -169,13 +153,10 @@ static const struct kfd_device_info fiji_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = true, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info fiji_vf_device_info = { - .asic_family = CHIP_FIJI, .asic_name = "fiji", .gfx_target_version = 80003, .max_pasid_bits = 16, @@ -188,14 +169,11 @@ static const struct kfd_device_info fiji_vf_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = false, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info polaris10_device_info = { - .asic_family = CHIP_POLARIS10, .asic_name = "polaris10", .gfx_target_version = 80003, .max_pasid_bits = 16, @@ -208,13 +186,10 @@ static const struct kfd_device_info polaris10_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = true, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info polaris10_vf_device_info = { - .asic_family = CHIP_POLARIS10, .asic_name = "polaris10", .gfx_target_version = 80003, .max_pasid_bits = 16, @@ -227,13 +202,10 @@ static const struct kfd_device_info polaris10_vf_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = false, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info polaris11_device_info = { - .asic_family = CHIP_POLARIS11, .asic_name = "polaris11", .gfx_target_version = 80003, .max_pasid_bits = 16, @@ -246,13 +218,10 @@ static const struct kfd_device_info polaris11_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = true, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info polaris12_device_info = { - .asic_family = CHIP_POLARIS12, .asic_name = "polaris12", .gfx_target_version = 80003, .max_pasid_bits = 16, @@ -265,13 +234,10 @@ static const struct kfd_device_info polaris12_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = true, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info vegam_device_info = { - .asic_family = CHIP_VEGAM, .asic_name = "vegam", .gfx_target_version = 80003, .max_pasid_bits = 16, @@ -284,13 +250,10 @@ static const struct kfd_device_info vegam_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = true, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info vega10_device_info = { - .asic_family = CHIP_VEGA10, .asic_name = "vega10", .gfx_target_version = 90000, .max_pasid_bits = 16, @@ -303,13 +266,10 @@ static const struct kfd_device_info vega10_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = false, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info vega10_vf_device_info = { - .asic_family = CHIP_VEGA10, .asic_name = "vega10", .gfx_target_version = 90000, .max_pasid_bits = 16, @@ -322,13 +282,10 @@ static const struct kfd_device_info vega10_vf_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = false, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info vega12_device_info = { - .asic_family = CHIP_VEGA12, .asic_name = "vega12", .gfx_target_version = 90004, .max_pasid_bits = 16, @@ -341,13 +298,10 @@ static const struct kfd_device_info vega12_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = false, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info vega20_device_info = { - .asic_family = CHIP_VEGA20, .asic_name = "vega20", .gfx_target_version = 90006, .max_pasid_bits = 16, @@ -360,13 +314,10 @@ static const struct kfd_device_info vega20_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = false, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 8, }; static const struct kfd_device_info arcturus_device_info = { - .asic_family = CHIP_ARCTURUS, .asic_name = "arcturus", .gfx_target_version = 90008, .max_pasid_bits = 16, @@ -379,13 +330,10 @@ static const struct kfd_device_info arcturus_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = false, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 6, .num_sdma_queues_per_engine = 8, }; static const struct kfd_device_info aldebaran_device_info = { - .asic_family = CHIP_ALDEBARAN, .asic_name = "aldebaran", .gfx_target_version = 90010, .max_pasid_bits = 16, @@ -398,13 +346,10 @@ static const struct kfd_device_info aldebaran_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = false, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 3, .num_sdma_queues_per_engine = 8, }; static const struct kfd_device_info renoir_device_info = { - .asic_family = CHIP_RENOIR, .asic_name = "renoir", .gfx_target_version = 90012, .max_pasid_bits = 16, @@ -417,13 +362,10 @@ static const struct kfd_device_info renoir_device_info = { .supports_cwsr = true, .needs_iommu_device = false, .needs_pci_atomics = false, - .num_sdma_engines = 1, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info navi10_device_info = { - .asic_family = CHIP_NAVI10, .asic_name = "navi10", .gfx_target_version = 100100, .max_pasid_bits = 16, @@ -437,13 +379,10 @@ static const struct kfd_device_info navi10_device_info = { .supports_cwsr = true, .needs_pci_atomics = true, .no_atomic_fw_version = 145, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 8, }; static const struct kfd_device_info navi12_device_info = { - .asic_family = CHIP_NAVI12, .asic_name = "navi12", .gfx_target_version = 100101, .max_pasid_bits = 16, @@ -457,13 +396,10 @@ static const struct kfd_device_info navi12_device_info = { .supports_cwsr = true, .needs_pci_atomics = true, .no_atomic_fw_version = 145, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 8, }; static const struct kfd_device_info navi14_device_info = { - .asic_family = CHIP_NAVI14, .asic_name = "navi14", .gfx_target_version = 100102, .max_pasid_bits = 16, @@ -477,13 +413,10 @@ static const struct kfd_device_info navi14_device_info = { .supports_cwsr = true, .needs_pci_atomics = true, .no_atomic_fw_version = 145, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 8, }; static const struct kfd_device_info sienna_cichlid_device_info = { - .asic_family = CHIP_SIENNA_CICHLID, .asic_name = "sienna_cichlid", .gfx_target_version = 100300, .max_pasid_bits = 16, @@ -497,13 +430,10 @@ static const struct kfd_device_info sienna_cichlid_device_info = { .supports_cwsr = true, .needs_pci_atomics = true, .no_atomic_fw_version = 92, - .num_sdma_engines = 4, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 8, }; static const struct kfd_device_info navy_flounder_device_info = { - .asic_family = CHIP_NAVY_FLOUNDER, .asic_name = "navy_flounder", .gfx_target_version = 100301, .max_pasid_bits = 16, @@ -517,13 +447,10 @@ static const struct kfd_device_info navy_flounder_device_info = { .supports_cwsr = true, .needs_pci_atomics = true, .no_atomic_fw_version = 92, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 8, }; static const struct kfd_device_info vangogh_device_info = { - .asic_family = CHIP_VANGOGH, .asic_name = "vangogh", .gfx_target_version = 100303, .max_pasid_bits = 16, @@ -537,13 +464,10 @@ static const struct kfd_device_info vangogh_device_info = { .supports_cwsr = true, .needs_pci_atomics = true, .no_atomic_fw_version = 92, - .num_sdma_engines = 1, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info dimgrey_cavefish_device_info = { - .asic_family = CHIP_DIMGREY_CAVEFISH, .asic_name = "dimgrey_cavefish", .gfx_target_version = 100302, .max_pasid_bits = 16, @@ -557,13 +481,10 @@ static const struct kfd_device_info dimgrey_cavefish_device_info = { .supports_cwsr = true, .needs_pci_atomics = true, .no_atomic_fw_version = 92, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 8, }; static const struct kfd_device_info beige_goby_device_info = { - .asic_family = CHIP_BEIGE_GOBY, .asic_name = "beige_goby", .gfx_target_version = 100304, .max_pasid_bits = 16, @@ -577,13 +498,10 @@ static const struct kfd_device_info beige_goby_device_info = { .supports_cwsr = true, .needs_pci_atomics = true, .no_atomic_fw_version = 92, - .num_sdma_engines = 1, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 8, }; static const struct kfd_device_info yellow_carp_device_info = { - .asic_family = CHIP_YELLOW_CARP, .asic_name = "yellow_carp", .gfx_target_version = 100305, .max_pasid_bits = 16, @@ -597,13 +515,10 @@ static const struct kfd_device_info yellow_carp_device_info = { .supports_cwsr = true, .needs_pci_atomics = true, .no_atomic_fw_version = 92, - .num_sdma_engines = 1, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 2, }; static const struct kfd_device_info cyan_skillfish_device_info = { - .asic_family = CHIP_CYAN_SKILLFISH, .asic_name = "cyan_skillfish", .gfx_target_version = 100103, .max_pasid_bits = 16, @@ -616,8 +531,6 @@ static const struct kfd_device_info cyan_skillfish_device_info = { .needs_iommu_device = false, .supports_cwsr = true, .needs_pci_atomics = true, - .num_sdma_engines = 2, - .num_xgmi_sdma_engines = 0, .num_sdma_queues_per_engine = 8, }; @@ -627,12 +540,11 @@ static void kfd_gtt_sa_fini(struct kfd_dev *kfd); static int kfd_resume(struct kfd_dev *kfd); -struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf) +struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf) { struct kfd_dev *kfd; const struct kfd_device_info *device_info; const struct kfd2kgd_calls *f2g; - struct amdgpu_device *adev = (struct amdgpu_device *)kgd; struct pci_dev *pdev = adev->pdev; switch (adev->asic_type) { @@ -815,8 +727,12 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf) } if (!device_info || !f2g) { - dev_err(kfd_device, "%s %s not supported in kfd\n", - amdgpu_asic_name[adev->asic_type], vf ? "VF" : ""); + if (adev->ip_versions[GC_HWIP][0]) + dev_err(kfd_device, "GC IP %06x %s not supported in kfd\n", + adev->ip_versions[GC_HWIP][0], vf ? "VF" : ""); + else + dev_err(kfd_device, "%s %s not supported in kfd\n", + amdgpu_asic_name[adev->asic_type], vf ? "VF" : ""); return NULL; } @@ -824,7 +740,7 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf) if (!kfd) return NULL; - kfd->kgd = kgd; + kfd->adev = adev; kfd->device_info = device_info; kfd->pdev = pdev; kfd->init_complete = false; @@ -845,23 +761,23 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf) static void kfd_cwsr_init(struct kfd_dev *kfd) { if (cwsr_enable && kfd->device_info->supports_cwsr) { - if (kfd->device_info->asic_family < CHIP_VEGA10) { + if (KFD_GC_VERSION(kfd) < IP_VERSION(9, 0, 1)) { BUILD_BUG_ON(sizeof(cwsr_trap_gfx8_hex) > PAGE_SIZE); kfd->cwsr_isa = cwsr_trap_gfx8_hex; kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx8_hex); - } else if (kfd->device_info->asic_family == CHIP_ARCTURUS) { + } else if (KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 1)) { BUILD_BUG_ON(sizeof(cwsr_trap_arcturus_hex) > PAGE_SIZE); kfd->cwsr_isa = cwsr_trap_arcturus_hex; kfd->cwsr_isa_size = sizeof(cwsr_trap_arcturus_hex); - } else if (kfd->device_info->asic_family == CHIP_ALDEBARAN) { + } else if (KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 2)) { BUILD_BUG_ON(sizeof(cwsr_trap_aldebaran_hex) > PAGE_SIZE); kfd->cwsr_isa = cwsr_trap_aldebaran_hex; kfd->cwsr_isa_size = sizeof(cwsr_trap_aldebaran_hex); - } else if (kfd->device_info->asic_family < CHIP_NAVI10) { + } else if (KFD_GC_VERSION(kfd) < IP_VERSION(10, 1, 1)) { BUILD_BUG_ON(sizeof(cwsr_trap_gfx9_hex) > PAGE_SIZE); kfd->cwsr_isa = cwsr_trap_gfx9_hex; kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx9_hex); - } else if (kfd->device_info->asic_family < CHIP_SIENNA_CICHLID) { + } else if (KFD_GC_VERSION(kfd) < IP_VERSION(10, 3, 0)) { BUILD_BUG_ON(sizeof(cwsr_trap_nv1x_hex) > PAGE_SIZE); kfd->cwsr_isa = cwsr_trap_nv1x_hex; kfd->cwsr_isa_size = sizeof(cwsr_trap_nv1x_hex); @@ -882,18 +798,17 @@ static int kfd_gws_init(struct kfd_dev *kfd) if (kfd->dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) return 0; - if (hws_gws_support - || (kfd->device_info->asic_family == CHIP_VEGA10 - && kfd->mec2_fw_version >= 0x81b3) - || (kfd->device_info->asic_family >= CHIP_VEGA12 - && kfd->device_info->asic_family <= CHIP_RAVEN - && kfd->mec2_fw_version >= 0x1b3) - || (kfd->device_info->asic_family == CHIP_ARCTURUS - && kfd->mec2_fw_version >= 0x30) - || (kfd->device_info->asic_family == CHIP_ALDEBARAN - && kfd->mec2_fw_version >= 0x28)) - ret = amdgpu_amdkfd_alloc_gws(kfd->kgd, - amdgpu_amdkfd_get_num_gws(kfd->kgd), &kfd->gws); + if (hws_gws_support || (KFD_IS_SOC15(kfd) && + ((KFD_GC_VERSION(kfd) == IP_VERSION(9, 0, 1) + && kfd->mec2_fw_version >= 0x81b3) || + (KFD_GC_VERSION(kfd) <= IP_VERSION(9, 4, 0) + && kfd->mec2_fw_version >= 0x1b3) || + (KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 1) + && kfd->mec2_fw_version >= 0x30) || + (KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 2) + && kfd->mec2_fw_version >= 0x28)))) + ret = amdgpu_amdkfd_alloc_gws(kfd->adev, + kfd->adev->gds.gws_size, &kfd->gws); return ret; } @@ -910,11 +825,11 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, unsigned int size, map_process_packet_size; kfd->ddev = ddev; - kfd->mec_fw_version = amdgpu_amdkfd_get_fw_version(kfd->kgd, + kfd->mec_fw_version = amdgpu_amdkfd_get_fw_version(kfd->adev, KGD_ENGINE_MEC1); - kfd->mec2_fw_version = amdgpu_amdkfd_get_fw_version(kfd->kgd, + kfd->mec2_fw_version = amdgpu_amdkfd_get_fw_version(kfd->adev, KGD_ENGINE_MEC2); - kfd->sdma_fw_version = amdgpu_amdkfd_get_fw_version(kfd->kgd, + kfd->sdma_fw_version = amdgpu_amdkfd_get_fw_version(kfd->adev, KGD_ENGINE_SDMA1); kfd->shared_resources = *gpu_resources; @@ -927,7 +842,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, * 32 and 64-bit requests are possible and must be * supported. */ - kfd->pci_atomic_requested = amdgpu_amdkfd_have_atomics_support(kfd->kgd); + kfd->pci_atomic_requested = amdgpu_amdkfd_have_atomics_support(kfd->adev); if (!kfd->pci_atomic_requested && kfd->device_info->needs_pci_atomics && (!kfd->device_info->no_atomic_fw_version || @@ -959,10 +874,9 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, * calculate max size of runlist packet. * There can be only 2 packets at once */ - map_process_packet_size = - kfd->device_info->asic_family == CHIP_ALDEBARAN ? + map_process_packet_size = KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 2) ? sizeof(struct pm4_mes_map_process_aldebaran) : - sizeof(struct pm4_mes_map_process); + sizeof(struct pm4_mes_map_process); size += (KFD_MAX_NUM_OF_PROCESSES * map_process_packet_size + max_num_of_queues_per_device * sizeof(struct pm4_mes_map_queues) + sizeof(struct pm4_mes_runlist)) * 2; @@ -974,7 +888,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, size += 512 * 1024; if (amdgpu_amdkfd_alloc_gtt_mem( - kfd->kgd, size, &kfd->gtt_mem, + kfd->adev, size, &kfd->gtt_mem, &kfd->gtt_start_gpu_addr, &kfd->gtt_start_cpu_ptr, false)) { dev_err(kfd_device, "Could not allocate %d bytes\n", size); @@ -995,9 +909,9 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, goto kfd_doorbell_error; } - kfd->hive_id = amdgpu_amdkfd_get_hive_id(kfd->kgd); + kfd->hive_id = kfd->adev->gmc.xgmi.hive_id; - kfd->noretry = amdgpu_amdkfd_get_noretry(kfd->kgd); + kfd->noretry = kfd->adev->gmc.noretry; if (kfd_interrupt_init(kfd)) { dev_err(kfd_device, "Error initializing interrupts\n"); @@ -1015,7 +929,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, */ if (kfd_gws_init(kfd)) { dev_err(kfd_device, "Could not allocate %d gws\n", - amdgpu_amdkfd_get_num_gws(kfd->kgd)); + kfd->adev->gds.gws_size); goto gws_error; } @@ -1030,7 +944,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, kfd_cwsr_init(kfd); - svm_migrate_init((struct amdgpu_device *)kfd->kgd); + svm_migrate_init(kfd->adev); if(kgd2kfd_resume_iommu(kfd)) goto device_iommu_error; @@ -1068,10 +982,10 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, kfd_doorbell_error: kfd_gtt_sa_fini(kfd); kfd_gtt_sa_init_error: - amdgpu_amdkfd_free_gtt_mem(kfd->kgd, kfd->gtt_mem); + amdgpu_amdkfd_free_gtt_mem(kfd->adev, kfd->gtt_mem); alloc_gtt_mem_failure: if (kfd->gws) - amdgpu_amdkfd_free_gws(kfd->kgd, kfd->gws); + amdgpu_amdkfd_free_gws(kfd->adev, kfd->gws); dev_err(kfd_device, "device %x:%x NOT added due to errors\n", kfd->pdev->vendor, kfd->pdev->device); @@ -1088,9 +1002,9 @@ void kgd2kfd_device_exit(struct kfd_dev *kfd) kfd_doorbell_fini(kfd); ida_destroy(&kfd->doorbell_ida); kfd_gtt_sa_fini(kfd); - amdgpu_amdkfd_free_gtt_mem(kfd->kgd, kfd->gtt_mem); + amdgpu_amdkfd_free_gtt_mem(kfd->adev, kfd->gtt_mem); if (kfd->gws) - amdgpu_amdkfd_free_gws(kfd->kgd, kfd->gws); + amdgpu_amdkfd_free_gws(kfd->adev, kfd->gws); } kfree(kfd); @@ -1526,7 +1440,7 @@ void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd) void kfd_inc_compute_active(struct kfd_dev *kfd) { if (atomic_inc_return(&kfd->compute_profile) == 1) - amdgpu_amdkfd_set_compute_idle(kfd->kgd, false); + amdgpu_amdkfd_set_compute_idle(kfd->adev, false); } void kfd_dec_compute_active(struct kfd_dev *kfd) @@ -1534,7 +1448,7 @@ void kfd_dec_compute_active(struct kfd_dev *kfd) int count = atomic_dec_return(&kfd->compute_profile); if (count == 0) - amdgpu_amdkfd_set_compute_idle(kfd->kgd, true); + amdgpu_amdkfd_set_compute_idle(kfd->adev, true); WARN_ONCE(count < 0, "Compute profile ref. count error"); } @@ -1544,6 +1458,26 @@ void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask) kfd_smi_event_update_thermal_throttling(kfd, throttle_bitmask); } +/* kfd_get_num_sdma_engines returns the number of PCIe optimized SDMA and + * kfd_get_num_xgmi_sdma_engines returns the number of XGMI SDMA. + * When the device has more than two engines, we reserve two for PCIe to enable + * full-duplex and the rest are used as XGMI. + */ +unsigned int kfd_get_num_sdma_engines(struct kfd_dev *kdev) +{ + /* If XGMI is not supported, all SDMA engines are PCIe */ + if (!kdev->adev->gmc.xgmi.supported) + return kdev->adev->sdma.num_instances; + + return min(kdev->adev->sdma.num_instances, 2); +} + +unsigned int kfd_get_num_xgmi_sdma_engines(struct kfd_dev *kdev) +{ + /* After reserved for PCIe, the rest of engines are XGMI */ + return kdev->adev->sdma.num_instances - kfd_get_num_sdma_engines(kdev); +} + #if defined(CONFIG_DEBUG_FS) /* This function will send a package to HIQ to hang the HWS diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 93e33dd84dd4..2af2b3268171 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -99,38 +99,29 @@ unsigned int get_pipes_per_mec(struct device_queue_manager *dqm) return dqm->dev->shared_resources.num_pipe_per_mec; } -static unsigned int get_num_sdma_engines(struct device_queue_manager *dqm) -{ - return dqm->dev->device_info->num_sdma_engines; -} - -static unsigned int get_num_xgmi_sdma_engines(struct device_queue_manager *dqm) -{ - return dqm->dev->device_info->num_xgmi_sdma_engines; -} - static unsigned int get_num_all_sdma_engines(struct device_queue_manager *dqm) { - return get_num_sdma_engines(dqm) + get_num_xgmi_sdma_engines(dqm); + return kfd_get_num_sdma_engines(dqm->dev) + + kfd_get_num_xgmi_sdma_engines(dqm->dev); } unsigned int get_num_sdma_queues(struct device_queue_manager *dqm) { - return dqm->dev->device_info->num_sdma_engines - * dqm->dev->device_info->num_sdma_queues_per_engine; + return kfd_get_num_sdma_engines(dqm->dev) * + dqm->dev->device_info->num_sdma_queues_per_engine; } unsigned int get_num_xgmi_sdma_queues(struct device_queue_manager *dqm) { - return dqm->dev->device_info->num_xgmi_sdma_engines - * dqm->dev->device_info->num_sdma_queues_per_engine; + return kfd_get_num_xgmi_sdma_engines(dqm->dev) * + dqm->dev->device_info->num_sdma_queues_per_engine; } void program_sh_mem_settings(struct device_queue_manager *dqm, struct qcm_process_device *qpd) { return dqm->dev->kfd2kgd->program_sh_mem_settings( - dqm->dev->kgd, qpd->vmid, + dqm->dev->adev, qpd->vmid, qpd->sh_mem_config, qpd->sh_mem_ape1_base, qpd->sh_mem_ape1_limit, @@ -157,7 +148,7 @@ static int allocate_doorbell(struct qcm_process_device *qpd, struct queue *q) { struct kfd_dev *dev = qpd->dqm->dev; - if (!KFD_IS_SOC15(dev->device_info->asic_family)) { + if (!KFD_IS_SOC15(dev)) { /* On pre-SOC15 chips we need to use the queue ID to * preserve the user mode ABI. */ @@ -202,7 +193,7 @@ static void deallocate_doorbell(struct qcm_process_device *qpd, unsigned int old; struct kfd_dev *dev = qpd->dqm->dev; - if (!KFD_IS_SOC15(dev->device_info->asic_family) || + if (!KFD_IS_SOC15(dev) || q->properties.type == KFD_QUEUE_TYPE_SDMA || q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI) return; @@ -216,7 +207,7 @@ static void program_trap_handler_settings(struct device_queue_manager *dqm, { if (dqm->dev->kfd2kgd->program_trap_handler_settings) dqm->dev->kfd2kgd->program_trap_handler_settings( - dqm->dev->kgd, qpd->vmid, + dqm->dev->adev, qpd->vmid, qpd->tba_addr, qpd->tma_addr); } @@ -250,21 +241,20 @@ static int allocate_vmid(struct device_queue_manager *dqm, program_sh_mem_settings(dqm, qpd); - if (dqm->dev->device_info->asic_family >= CHIP_VEGA10 && - dqm->dev->cwsr_enabled) + if (KFD_IS_SOC15(dqm->dev) && dqm->dev->cwsr_enabled) program_trap_handler_settings(dqm, qpd); /* qpd->page_table_base is set earlier when register_process() * is called, i.e. when the first queue is created. */ - dqm->dev->kfd2kgd->set_vm_context_page_table_base(dqm->dev->kgd, + dqm->dev->kfd2kgd->set_vm_context_page_table_base(dqm->dev->adev, qpd->vmid, qpd->page_table_base); /* invalidate the VM context after pasid and vmid mapping is set up */ kfd_flush_tlb(qpd_to_pdd(qpd), TLB_FLUSH_LEGACY); if (dqm->dev->kfd2kgd->set_scratch_backing_va) - dqm->dev->kfd2kgd->set_scratch_backing_va(dqm->dev->kgd, + dqm->dev->kfd2kgd->set_scratch_backing_va(dqm->dev->adev, qpd->sh_hidden_private_base, qpd->vmid); return 0; @@ -283,7 +273,7 @@ static int flush_texture_cache_nocpsch(struct kfd_dev *kdev, if (ret) return ret; - return amdgpu_amdkfd_submit_ib(kdev->kgd, KGD_ENGINE_MEC1, qpd->vmid, + return amdgpu_amdkfd_submit_ib(kdev->adev, KGD_ENGINE_MEC1, qpd->vmid, qpd->ib_base, (uint32_t *)qpd->ib_kaddr, pmf->release_mem_size / sizeof(uint32_t)); } @@ -293,7 +283,7 @@ static void deallocate_vmid(struct device_queue_manager *dqm, struct queue *q) { /* On GFX v7, CP doesn't flush TC at dequeue */ - if (q->device->device_info->asic_family == CHIP_HAWAII) + if (q->device->adev->asic_type == CHIP_HAWAII) if (flush_texture_cache_nocpsch(q->device, qpd)) pr_err("Failed to flush TC\n"); @@ -776,7 +766,7 @@ static int restore_process_queues_nocpsch(struct device_queue_manager *dqm, if (!list_empty(&qpd->queues_list)) { dqm->dev->kfd2kgd->set_vm_context_page_table_base( - dqm->dev->kgd, + dqm->dev->adev, qpd->vmid, qpd->page_table_base); kfd_flush_tlb(pdd, TLB_FLUSH_LEGACY); @@ -954,7 +944,7 @@ set_pasid_vmid_mapping(struct device_queue_manager *dqm, u32 pasid, unsigned int vmid) { return dqm->dev->kfd2kgd->set_pasid_vmid_mapping( - dqm->dev->kgd, pasid, vmid); + dqm->dev->adev, pasid, vmid); } static void init_interrupts(struct device_queue_manager *dqm) @@ -963,7 +953,7 @@ static void init_interrupts(struct device_queue_manager *dqm) for (i = 0 ; i < get_pipes_per_mec(dqm) ; i++) if (is_pipe_enabled(dqm, 0, i)) - dqm->dev->kfd2kgd->init_interrupts(dqm->dev->kgd, i); + dqm->dev->kfd2kgd->init_interrupts(dqm->dev->adev, i); } static int initialize_nocpsch(struct device_queue_manager *dqm) @@ -1017,7 +1007,7 @@ static int start_nocpsch(struct device_queue_manager *dqm) pr_info("SW scheduler is used"); init_interrupts(dqm); - if (dqm->dev->device_info->asic_family == CHIP_HAWAII) + if (dqm->dev->adev->asic_type == CHIP_HAWAII) return pm_init(&dqm->packet_mgr, dqm); dqm->sched_running = true; @@ -1026,7 +1016,7 @@ static int start_nocpsch(struct device_queue_manager *dqm) static int stop_nocpsch(struct device_queue_manager *dqm) { - if (dqm->dev->device_info->asic_family == CHIP_HAWAII) + if (dqm->dev->adev->asic_type == CHIP_HAWAII) pm_uninit(&dqm->packet_mgr, false); dqm->sched_running = false; @@ -1055,9 +1045,9 @@ static int allocate_sdma_queue(struct device_queue_manager *dqm, dqm->sdma_bitmap &= ~(1ULL << bit); q->sdma_id = bit; q->properties.sdma_engine_id = q->sdma_id % - get_num_sdma_engines(dqm); + kfd_get_num_sdma_engines(dqm->dev); q->properties.sdma_queue_id = q->sdma_id / - get_num_sdma_engines(dqm); + kfd_get_num_sdma_engines(dqm->dev); } else if (q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI) { if (dqm->xgmi_sdma_bitmap == 0) { pr_err("No more XGMI SDMA queue to allocate\n"); @@ -1072,10 +1062,11 @@ static int allocate_sdma_queue(struct device_queue_manager *dqm, * assumes the first N engines are always * PCIe-optimized ones */ - q->properties.sdma_engine_id = get_num_sdma_engines(dqm) + - q->sdma_id % get_num_xgmi_sdma_engines(dqm); + q->properties.sdma_engine_id = + kfd_get_num_sdma_engines(dqm->dev) + + q->sdma_id % kfd_get_num_xgmi_sdma_engines(dqm->dev); q->properties.sdma_queue_id = q->sdma_id / - get_num_xgmi_sdma_engines(dqm); + kfd_get_num_xgmi_sdma_engines(dqm->dev); } pr_debug("SDMA engine id: %d\n", q->properties.sdma_engine_id); @@ -1132,7 +1123,7 @@ static int set_sched_resources(struct device_queue_manager *dqm) res.queue_mask |= 1ull << amdgpu_queue_mask_bit_to_set_resource_bit( - (struct amdgpu_device *)dqm->dev->kgd, i); + dqm->dev->adev, i); } res.gws_mask = ~0ull; res.oac_mask = res.gds_heap_base = res.gds_heap_size = 0; @@ -1850,7 +1841,7 @@ static int allocate_hiq_sdma_mqd(struct device_queue_manager *dqm) dev->device_info->num_sdma_queues_per_engine + dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ]->mqd_size; - retval = amdgpu_amdkfd_alloc_gtt_mem(dev->kgd, size, + retval = amdgpu_amdkfd_alloc_gtt_mem(dev->adev, size, &(mem_obj->gtt_mem), &(mem_obj->gpu_addr), (void *)&(mem_obj->cpu_ptr), false); @@ -1867,7 +1858,7 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev) if (!dqm) return NULL; - switch (dev->device_info->asic_family) { + switch (dev->adev->asic_type) { /* HWS is not available on Hawaii. */ case CHIP_HAWAII: /* HWS depends on CWSR for timely dequeue. CWSR is not @@ -1930,7 +1921,7 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev) goto out_free; } - switch (dev->device_info->asic_family) { + switch (dev->adev->asic_type) { case CHIP_CARRIZO: device_queue_manager_init_vi(&dqm->asic_ops); break; @@ -1952,31 +1943,16 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev) device_queue_manager_init_vi_tonga(&dqm->asic_ops); break; - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_RAVEN: - case CHIP_RENOIR: - case CHIP_ARCTURUS: - case CHIP_ALDEBARAN: - device_queue_manager_init_v9(&dqm->asic_ops); - break; - case CHIP_NAVI10: - case CHIP_NAVI12: - case CHIP_NAVI14: - case CHIP_SIENNA_CICHLID: - case CHIP_NAVY_FLOUNDER: - case CHIP_VANGOGH: - case CHIP_DIMGREY_CAVEFISH: - case CHIP_BEIGE_GOBY: - case CHIP_YELLOW_CARP: - case CHIP_CYAN_SKILLFISH: - device_queue_manager_init_v10_navi10(&dqm->asic_ops); - break; default: - WARN(1, "Unexpected ASIC family %u", - dev->device_info->asic_family); - goto out_free; + if (KFD_GC_VERSION(dev) >= IP_VERSION(10, 1, 1)) + device_queue_manager_init_v10_navi10(&dqm->asic_ops); + else if (KFD_GC_VERSION(dev) >= IP_VERSION(9, 0, 1)) + device_queue_manager_init_v9(&dqm->asic_ops); + else { + WARN(1, "Unexpected ASIC family %u", + dev->adev->asic_type); + goto out_free; + } } if (init_mqd_managers(dqm)) @@ -2000,7 +1976,7 @@ static void deallocate_hiq_sdma_mqd(struct kfd_dev *dev, { WARN(!mqd, "No hiq sdma mqd trunk to free"); - amdgpu_amdkfd_free_gtt_mem(dev->kgd, mqd->gtt_mem); + amdgpu_amdkfd_free_gtt_mem(dev->adev, mqd->gtt_mem); } void device_queue_manager_uninit(struct device_queue_manager *dqm) @@ -2031,7 +2007,7 @@ static void kfd_process_hw_exception(struct work_struct *work) { struct device_queue_manager *dqm = container_of(work, struct device_queue_manager, hw_exception_work); - amdgpu_amdkfd_gpu_reset(dqm->dev->kgd); + amdgpu_amdkfd_gpu_reset(dqm->dev->adev); } #if defined(CONFIG_DEBUG_FS) @@ -2070,7 +2046,7 @@ int dqm_debugfs_hqds(struct seq_file *m, void *data) return 0; } - r = dqm->dev->kfd2kgd->hqd_dump(dqm->dev->kgd, + r = dqm->dev->kfd2kgd->hqd_dump(dqm->dev->adev, KFD_CIK_HIQ_PIPE, KFD_CIK_HIQ_QUEUE, &dump, &n_regs); if (!r) { @@ -2092,7 +2068,7 @@ int dqm_debugfs_hqds(struct seq_file *m, void *data) continue; r = dqm->dev->kfd2kgd->hqd_dump( - dqm->dev->kgd, pipe, queue, &dump, &n_regs); + dqm->dev->adev, pipe, queue, &dump, &n_regs); if (r) break; @@ -2109,7 +2085,7 @@ int dqm_debugfs_hqds(struct seq_file *m, void *data) queue < dqm->dev->device_info->num_sdma_queues_per_engine; queue++) { r = dqm->dev->kfd2kgd->hqd_sdma_dump( - dqm->dev->kgd, pipe, queue, &dump, &n_regs); + dqm->dev->adev, pipe, queue, &dump, &n_regs); if (r) break; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c index b5c3d13643f1..f20434d9980e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c @@ -62,7 +62,7 @@ static int update_qpd_v9(struct device_queue_manager *dqm, SH_MEM_ALIGNMENT_MODE_UNALIGNED << SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT; - if (dqm->dev->device_info->asic_family == CHIP_ALDEBARAN) { + if (KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 4, 2)) { /* Aldebaran can safely support different XNACK modes * per process */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_events.c index 3eea4edee355..afe72dd11325 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_events.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.c @@ -935,8 +935,10 @@ void kfd_signal_iommu_event(struct kfd_dev *dev, u32 pasid, /* Workaround on Raven to not kill the process when memory is freed * before IOMMU is able to finish processing all the excessive PPRs */ - if (dev->device_info->asic_family != CHIP_RAVEN && - dev->device_info->asic_family != CHIP_RENOIR) { + + if (KFD_GC_VERSION(dev) != IP_VERSION(9, 1, 0) && + KFD_GC_VERSION(dev) != IP_VERSION(9, 2, 2) && + KFD_GC_VERSION(dev) != IP_VERSION(9, 3, 0)) { mutex_lock(&p->event_mutex); /* Lookup events by type and signal them */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c index d1388896f9c1..2e2b7ceb71db 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c @@ -394,7 +394,7 @@ int kfd_init_apertures(struct kfd_process *process) pdd->gpuvm_base = pdd->gpuvm_limit = 0; pdd->scratch_base = pdd->scratch_limit = 0; } else { - switch (dev->device_info->asic_family) { + switch (dev->adev->asic_type) { case CHIP_KAVERI: case CHIP_HAWAII: case CHIP_CARRIZO: @@ -406,29 +406,14 @@ int kfd_init_apertures(struct kfd_process *process) case CHIP_VEGAM: kfd_init_apertures_vi(pdd, id); break; - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_RAVEN: - case CHIP_RENOIR: - case CHIP_ARCTURUS: - case CHIP_ALDEBARAN: - case CHIP_NAVI10: - case CHIP_NAVI12: - case CHIP_NAVI14: - case CHIP_SIENNA_CICHLID: - case CHIP_NAVY_FLOUNDER: - case CHIP_VANGOGH: - case CHIP_DIMGREY_CAVEFISH: - case CHIP_BEIGE_GOBY: - case CHIP_YELLOW_CARP: - case CHIP_CYAN_SKILLFISH: - kfd_init_apertures_v9(pdd, id); - break; default: - WARN(1, "Unexpected ASIC family %u", - dev->device_info->asic_family); - return -EINVAL; + if (KFD_GC_VERSION(dev) >= IP_VERSION(9, 0, 1)) + kfd_init_apertures_v9(pdd, id); + else { + WARN(1, "Unexpected ASIC family %u", + dev->adev->asic_type); + return -EINVAL; + } } if (!dev->use_iommu_v2) { diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c index 543e7ea75593..20512a4e9a91 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c @@ -231,7 +231,7 @@ static void event_interrupt_wq_v9(struct kfd_dev *dev, if (sq_intr_err != SQ_INTERRUPT_ERROR_TYPE_ILLEGAL_INST && sq_intr_err != SQ_INTERRUPT_ERROR_TYPE_MEMVIOL) { kfd_signal_poison_consumed_event(dev, pasid); - amdgpu_amdkfd_ras_poison_consumption_handler(dev->kgd); + amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev); return; } break; @@ -253,7 +253,7 @@ static void event_interrupt_wq_v9(struct kfd_dev *dev, kfd_signal_event_interrupt(pasid, context_id0 & 0xfffffff, 28); } else if (source_id == SOC15_INTSRC_SDMA_ECC) { kfd_signal_poison_consumed_event(dev, pasid); - amdgpu_amdkfd_ras_poison_consumption_handler(dev->kgd); + amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev); return; } } else if (client_id == SOC15_IH_CLIENTID_VMC || diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c index 64b4ac339904..406479a369a9 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c @@ -91,7 +91,7 @@ static bool kq_initialize(struct kernel_queue *kq, struct kfd_dev *dev, kq->pq_gpu_addr = kq->pq->gpu_addr; /* For CIK family asics, kq->eop_mem is not needed */ - if (dev->device_info->asic_family > CHIP_MULLINS) { + if (dev->adev->asic_type > CHIP_MULLINS) { retval = kfd_gtt_sa_allocate(dev, PAGE_SIZE, &kq->eop_mem); if (retval != 0) goto err_eop_allocate_vidmem; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c index 9b9c2b9bf2ef..d84cec0022b1 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c @@ -938,7 +938,7 @@ int svm_migrate_init(struct amdgpu_device *adev) void *r; /* Page migration works on Vega10 or newer */ - if (kfddev->device_info->asic_family < CHIP_VEGA10) + if (!KFD_IS_SOC15(kfddev)) return -EINVAL; pgmap = &kfddev->pgmap; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c index c021519af810..7b4118915bf6 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c @@ -100,7 +100,7 @@ void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm, struct kfd_cu_info cu_info; uint32_t cu_per_sh[KFD_MAX_NUM_SE][KFD_MAX_NUM_SH_PER_SE] = {0}; int i, se, sh, cu; - amdgpu_amdkfd_get_cu_info(mm->dev->kgd, &cu_info); + amdgpu_amdkfd_get_cu_info(mm->dev->adev, &cu_info); if (cu_mask_count > cu_info.cu_active_number) cu_mask_count = cu_info.cu_active_number; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c index 8128f4d312f1..e9a8e21e144e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c @@ -171,7 +171,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd, uint32_t pipe_id, uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0); uint32_t wptr_mask = (uint32_t)((p->queue_size / 4) - 1); - return mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id, + return mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id, (uint32_t __user *)p->write_ptr, wptr_shift, wptr_mask, mms); } @@ -180,7 +180,7 @@ static int load_mqd_sdma(struct mqd_manager *mm, void *mqd, uint32_t pipe_id, uint32_t queue_id, struct queue_properties *p, struct mm_struct *mms) { - return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd, + return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->adev, mqd, (uint32_t __user *)p->write_ptr, mms); } @@ -276,7 +276,7 @@ static int destroy_mqd(struct mqd_manager *mm, void *mqd, unsigned int timeout, uint32_t pipe_id, uint32_t queue_id) { - return mm->dev->kfd2kgd->hqd_destroy(mm->dev->kgd, mqd, type, timeout, + return mm->dev->kfd2kgd->hqd_destroy(mm->dev->adev, mqd, type, timeout, pipe_id, queue_id); } @@ -289,7 +289,7 @@ static int destroy_mqd_sdma(struct mqd_manager *mm, void *mqd, unsigned int timeout, uint32_t pipe_id, uint32_t queue_id) { - return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->kgd, mqd, timeout); + return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->adev, mqd, timeout); } static bool is_occupied(struct mqd_manager *mm, void *mqd, @@ -297,7 +297,7 @@ static bool is_occupied(struct mqd_manager *mm, void *mqd, uint32_t queue_id) { - return mm->dev->kfd2kgd->hqd_is_occupied(mm->dev->kgd, queue_address, + return mm->dev->kfd2kgd->hqd_is_occupied(mm->dev->adev, queue_address, pipe_id, queue_id); } @@ -306,7 +306,7 @@ static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id) { - return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd); + return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->adev, mqd); } /* diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c index 270160fc401b..d74d8a6ac27a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c @@ -148,7 +148,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd, /* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */ uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0); - r = mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id, + r = mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id, (uint32_t __user *)p->write_ptr, wptr_shift, 0, mms); return r; @@ -158,7 +158,7 @@ static int hiq_load_mqd_kiq(struct mqd_manager *mm, void *mqd, uint32_t pipe_id, uint32_t queue_id, struct queue_properties *p, struct mm_struct *mms) { - return mm->dev->kfd2kgd->hiq_mqd_load(mm->dev->kgd, mqd, pipe_id, + return mm->dev->kfd2kgd->hiq_mqd_load(mm->dev->adev, mqd, pipe_id, queue_id, p->doorbell_off); } @@ -239,7 +239,7 @@ static int destroy_mqd(struct mqd_manager *mm, void *mqd, uint32_t queue_id) { return mm->dev->kfd2kgd->hqd_destroy - (mm->dev->kgd, mqd, type, timeout, + (mm->dev->adev, mqd, type, timeout, pipe_id, queue_id); } @@ -254,7 +254,7 @@ static bool is_occupied(struct mqd_manager *mm, void *mqd, uint32_t queue_id) { return mm->dev->kfd2kgd->hqd_is_occupied( - mm->dev->kgd, queue_address, + mm->dev->adev, queue_address, pipe_id, queue_id); } @@ -320,7 +320,7 @@ static int load_mqd_sdma(struct mqd_manager *mm, void *mqd, uint32_t pipe_id, uint32_t queue_id, struct queue_properties *p, struct mm_struct *mms) { - return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd, + return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->adev, mqd, (uint32_t __user *)p->write_ptr, mms); } @@ -363,14 +363,14 @@ static int destroy_mqd_sdma(struct mqd_manager *mm, void *mqd, unsigned int timeout, uint32_t pipe_id, uint32_t queue_id) { - return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->kgd, mqd, timeout); + return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->adev, mqd, timeout); } static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id) { - return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd); + return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->adev, mqd); } #if defined(CONFIG_DEBUG_FS) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c index 4e5932f54b5a..326eb2285029 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c @@ -108,7 +108,7 @@ static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd, mqd_mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL); if (!mqd_mem_obj) return NULL; - retval = amdgpu_amdkfd_alloc_gtt_mem(kfd->kgd, + retval = amdgpu_amdkfd_alloc_gtt_mem(kfd->adev, ALIGN(q->ctl_stack_size, PAGE_SIZE) + ALIGN(sizeof(struct v9_mqd), PAGE_SIZE), &(mqd_mem_obj->gtt_mem), @@ -199,7 +199,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd, /* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */ uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0); - return mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id, + return mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id, (uint32_t __user *)p->write_ptr, wptr_shift, 0, mms); } @@ -208,7 +208,7 @@ static int hiq_load_mqd_kiq(struct mqd_manager *mm, void *mqd, uint32_t pipe_id, uint32_t queue_id, struct queue_properties *p, struct mm_struct *mms) { - return mm->dev->kfd2kgd->hiq_mqd_load(mm->dev->kgd, mqd, pipe_id, + return mm->dev->kfd2kgd->hiq_mqd_load(mm->dev->adev, mqd, pipe_id, queue_id, p->doorbell_off); } @@ -291,7 +291,7 @@ static int destroy_mqd(struct mqd_manager *mm, void *mqd, uint32_t queue_id) { return mm->dev->kfd2kgd->hqd_destroy - (mm->dev->kgd, mqd, type, timeout, + (mm->dev->adev, mqd, type, timeout, pipe_id, queue_id); } @@ -301,7 +301,7 @@ static void free_mqd(struct mqd_manager *mm, void *mqd, struct kfd_dev *kfd = mm->dev; if (mqd_mem_obj->gtt_mem) { - amdgpu_amdkfd_free_gtt_mem(kfd->kgd, mqd_mem_obj->gtt_mem); + amdgpu_amdkfd_free_gtt_mem(kfd->adev, mqd_mem_obj->gtt_mem); kfree(mqd_mem_obj); } else { kfd_gtt_sa_free(mm->dev, mqd_mem_obj); @@ -313,7 +313,7 @@ static bool is_occupied(struct mqd_manager *mm, void *mqd, uint32_t queue_id) { return mm->dev->kfd2kgd->hqd_is_occupied( - mm->dev->kgd, queue_address, + mm->dev->adev, queue_address, pipe_id, queue_id); } @@ -375,7 +375,7 @@ static int load_mqd_sdma(struct mqd_manager *mm, void *mqd, uint32_t pipe_id, uint32_t queue_id, struct queue_properties *p, struct mm_struct *mms) { - return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd, + return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->adev, mqd, (uint32_t __user *)p->write_ptr, mms); } @@ -418,14 +418,14 @@ static int destroy_mqd_sdma(struct mqd_manager *mm, void *mqd, unsigned int timeout, uint32_t pipe_id, uint32_t queue_id) { - return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->kgd, mqd, timeout); + return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->adev, mqd, timeout); } static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id) { - return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd); + return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->adev, mqd); } #if defined(CONFIG_DEBUG_FS) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c index cd9220eb8a7a..d456e950ce1d 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c @@ -162,7 +162,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd, uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0); uint32_t wptr_mask = (uint32_t)((p->queue_size / 4) - 1); - return mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id, + return mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id, (uint32_t __user *)p->write_ptr, wptr_shift, wptr_mask, mms); } @@ -265,7 +265,7 @@ static int destroy_mqd(struct mqd_manager *mm, void *mqd, uint32_t queue_id) { return mm->dev->kfd2kgd->hqd_destroy - (mm->dev->kgd, mqd, type, timeout, + (mm->dev->adev, mqd, type, timeout, pipe_id, queue_id); } @@ -280,7 +280,7 @@ static bool is_occupied(struct mqd_manager *mm, void *mqd, uint32_t queue_id) { return mm->dev->kfd2kgd->hqd_is_occupied( - mm->dev->kgd, queue_address, + mm->dev->adev, queue_address, pipe_id, queue_id); } @@ -347,7 +347,7 @@ static int load_mqd_sdma(struct mqd_manager *mm, void *mqd, uint32_t pipe_id, uint32_t queue_id, struct queue_properties *p, struct mm_struct *mms) { - return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd, + return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->adev, mqd, (uint32_t __user *)p->write_ptr, mms); } @@ -389,14 +389,14 @@ static int destroy_mqd_sdma(struct mqd_manager *mm, void *mqd, unsigned int timeout, uint32_t pipe_id, uint32_t queue_id) { - return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->kgd, mqd, timeout); + return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->adev, mqd, timeout); } static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id) { - return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd); + return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->adev, mqd); } #if defined(CONFIG_DEBUG_FS) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c index e547f1f8c49f..1439420925a0 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c @@ -223,7 +223,7 @@ static int pm_create_runlist_ib(struct packet_manager *pm, int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm) { - switch (dqm->dev->device_info->asic_family) { + switch (dqm->dev->adev->asic_type) { case CHIP_KAVERI: case CHIP_HAWAII: /* PM4 packet structures on CIK are the same as on VI */ @@ -236,31 +236,16 @@ int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm) case CHIP_VEGAM: pm->pmf = &kfd_vi_pm_funcs; break; - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_RAVEN: - case CHIP_RENOIR: - case CHIP_ARCTURUS: - case CHIP_NAVI10: - case CHIP_NAVI12: - case CHIP_NAVI14: - case CHIP_SIENNA_CICHLID: - case CHIP_NAVY_FLOUNDER: - case CHIP_VANGOGH: - case CHIP_DIMGREY_CAVEFISH: - case CHIP_BEIGE_GOBY: - case CHIP_YELLOW_CARP: - case CHIP_CYAN_SKILLFISH: - pm->pmf = &kfd_v9_pm_funcs; - break; - case CHIP_ALDEBARAN: - pm->pmf = &kfd_aldebaran_pm_funcs; - break; default: - WARN(1, "Unexpected ASIC family %u", - dqm->dev->device_info->asic_family); - return -EINVAL; + if (KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 4, 2)) + pm->pmf = &kfd_aldebaran_pm_funcs; + else if (KFD_GC_VERSION(dqm->dev) >= IP_VERSION(9, 0, 1)) + pm->pmf = &kfd_v9_pm_funcs; + else { + WARN(1, "Unexpected ASIC family %u", + dqm->dev->adev->asic_type); + return -EINVAL; + } } pm->dqm = dqm; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 94e92c0812db..7ea528941951 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -183,7 +183,8 @@ enum cache_policy { cache_policy_noncoherent }; -#define KFD_IS_SOC15(chip) ((chip) >= CHIP_VEGA10) +#define KFD_GC_VERSION(dev) ((dev)->adev->ip_versions[GC_HWIP][0]) +#define KFD_IS_SOC15(dev) ((KFD_GC_VERSION(dev)) >= (IP_VERSION(9, 0, 1))) struct kfd_event_interrupt_class { bool (*interrupt_isr)(struct kfd_dev *dev, @@ -194,7 +195,6 @@ struct kfd_event_interrupt_class { }; struct kfd_device_info { - enum amd_asic_type asic_family; const char *asic_name; uint32_t gfx_target_version; const struct kfd_event_interrupt_class *event_interrupt_class; @@ -208,11 +208,12 @@ struct kfd_device_info { bool needs_iommu_device; bool needs_pci_atomics; uint32_t no_atomic_fw_version; - unsigned int num_sdma_engines; - unsigned int num_xgmi_sdma_engines; unsigned int num_sdma_queues_per_engine; }; +unsigned int kfd_get_num_sdma_engines(struct kfd_dev *kdev); +unsigned int kfd_get_num_xgmi_sdma_engines(struct kfd_dev *kdev); + struct kfd_mem_obj { uint32_t range_start; uint32_t range_end; @@ -228,7 +229,7 @@ struct kfd_vmid_info { }; struct kfd_dev { - struct kgd_dev *kgd; + struct amdgpu_device *adev; const struct kfd_device_info *device_info; struct pci_dev *pdev; @@ -766,7 +767,7 @@ struct svm_range_list { struct list_head deferred_range_list; spinlock_t deferred_list_lock; atomic_t evicted_ranges; - bool drain_pagefaults; + atomic_t drain_pagefaults; struct delayed_work restore_work; DECLARE_BITMAP(bitmap_supported, MAX_GPU_INSTANCE); struct task_struct *faulting_task; @@ -891,7 +892,7 @@ struct kfd_process *kfd_lookup_process_by_pasid(u32 pasid); struct kfd_process *kfd_lookup_process_by_mm(const struct mm_struct *mm); int kfd_process_gpuidx_from_gpuid(struct kfd_process *p, uint32_t gpu_id); -int kfd_process_gpuid_from_kgd(struct kfd_process *p, +int kfd_process_gpuid_from_adev(struct kfd_process *p, struct amdgpu_device *adev, uint32_t *gpuid, uint32_t *gpuidx); static inline int kfd_process_gpuid_from_gpuidx(struct kfd_process *p, @@ -984,7 +985,7 @@ struct kfd_topology_device *kfd_topology_device_by_proximity_domain( struct kfd_topology_device *kfd_topology_device_by_id(uint32_t gpu_id); struct kfd_dev *kfd_device_by_id(uint32_t gpu_id); struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev); -struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd); +struct kfd_dev *kfd_device_by_adev(const struct amdgpu_device *adev); int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_dev **kdev); int kfd_numa_node_to_apic_id(int numa_node_id); void kfd_double_confirm_iommu_support(struct kfd_dev *gpu); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index b993011cfa64..d4c8a6948a9f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -288,7 +288,7 @@ static int kfd_get_cu_occupancy(struct attribute *attr, char *buffer) /* Collect wave count from device if it supports */ wave_cnt = 0; max_waves_per_cu = 0; - dev->kfd2kgd->get_cu_occupancy(dev->kgd, proc->pasid, &wave_cnt, + dev->kfd2kgd->get_cu_occupancy(dev->adev, proc->pasid, &wave_cnt, &max_waves_per_cu); /* Translate wave count to number of compute units */ @@ -692,12 +692,12 @@ static void kfd_process_free_gpuvm(struct kgd_mem *mem, struct kfd_dev *dev = pdd->dev; if (kptr) { - amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(dev->kgd, mem); + amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(dev->adev, mem); kptr = NULL; } - amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(dev->kgd, mem, pdd->drm_priv); - amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, mem, pdd->drm_priv, + amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(dev->adev, mem, pdd->drm_priv); + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev, mem, pdd->drm_priv, NULL); } @@ -714,24 +714,24 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd, struct kfd_dev *kdev = pdd->dev; int err; - err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(kdev->kgd, gpu_va, size, + err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(kdev->adev, gpu_va, size, pdd->drm_priv, mem, NULL, flags); if (err) goto err_alloc_mem; - err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->kgd, *mem, + err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->adev, *mem, pdd->drm_priv, NULL); if (err) goto err_map_mem; - err = amdgpu_amdkfd_gpuvm_sync_memory(kdev->kgd, *mem, true); + err = amdgpu_amdkfd_gpuvm_sync_memory(kdev->adev, *mem, true); if (err) { pr_debug("Sync memory failed, wait interrupted by user signal\n"); goto sync_memory_failed; } if (kptr) { - err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kdev->kgd, + err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kdev->adev, (struct kgd_mem *)*mem, kptr, NULL); if (err) { pr_debug("Map GTT BO to kernel failed\n"); @@ -742,10 +742,10 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd, return err; sync_memory_failed: - amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(kdev->kgd, *mem, pdd->drm_priv); + amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(kdev->adev, *mem, pdd->drm_priv); err_map_mem: - amdgpu_amdkfd_gpuvm_free_memory_of_gpu(kdev->kgd, *mem, pdd->drm_priv, + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(kdev->adev, *mem, pdd->drm_priv, NULL); err_alloc_mem: *mem = NULL; @@ -940,10 +940,10 @@ static void kfd_process_device_free_bos(struct kfd_process_device *pdd) if (!peer_pdd->drm_priv) continue; amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( - peer_pdd->dev->kgd, mem, peer_pdd->drm_priv); + peer_pdd->dev->adev, mem, peer_pdd->drm_priv); } - amdgpu_amdkfd_gpuvm_free_memory_of_gpu(pdd->dev->kgd, mem, + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(pdd->dev->adev, mem, pdd->drm_priv, NULL); kfd_process_device_remove_obj_handle(pdd, id); } @@ -974,7 +974,7 @@ static void kfd_process_kunmap_signal_bo(struct kfd_process *p) if (!mem) goto out; - amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(kdev->kgd, mem); + amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(kdev->adev, mem); out: mutex_unlock(&p->mutex); @@ -1003,7 +1003,7 @@ static void kfd_process_destroy_pdds(struct kfd_process *p) if (pdd->drm_file) { amdgpu_amdkfd_gpuvm_release_process_vm( - pdd->dev->kgd, pdd->drm_priv); + pdd->dev->adev, pdd->drm_priv); fput(pdd->drm_file); } @@ -1317,14 +1317,13 @@ bool kfd_process_xnack_mode(struct kfd_process *p, bool supported) * support the SVM APIs and don't need to be considered * for the XNACK mode selection. */ - if (dev->device_info->asic_family < CHIP_VEGA10) + if (!KFD_IS_SOC15(dev)) continue; /* Aldebaran can always support XNACK because it can support * per-process XNACK mode selection. But let the dev->noretry * setting still influence the default XNACK mode. */ - if (supported && - dev->device_info->asic_family == CHIP_ALDEBARAN) + if (supported && KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 2)) continue; /* GFXv10 and later GPUs do not support shader preemption @@ -1332,7 +1331,7 @@ bool kfd_process_xnack_mode(struct kfd_process *p, bool supported) * management and memory-manager-related preemptions or * even deadlocks. */ - if (dev->device_info->asic_family >= CHIP_NAVI10) + if (KFD_GC_VERSION(dev) >= IP_VERSION(10, 1, 1)) return false; if (dev->noretry) @@ -1431,7 +1430,7 @@ static int init_doorbell_bitmap(struct qcm_process_device *qpd, int range_start = dev->shared_resources.non_cp_doorbells_start; int range_end = dev->shared_resources.non_cp_doorbells_end; - if (!KFD_IS_SOC15(dev->device_info->asic_family)) + if (!KFD_IS_SOC15(dev)) return 0; qpd->doorbell_bitmap = @@ -1547,7 +1546,7 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd, dev = pdd->dev; ret = amdgpu_amdkfd_gpuvm_acquire_process_vm( - dev->kgd, drm_file, p->pasid, + dev->adev, drm_file, p->pasid, &p->kgd_process_info, &p->ef); if (ret) { pr_err("Failed to create process VM object\n"); @@ -1779,14 +1778,13 @@ int kfd_process_gpuidx_from_gpuid(struct kfd_process *p, uint32_t gpu_id) } int -kfd_process_gpuid_from_kgd(struct kfd_process *p, struct amdgpu_device *adev, +kfd_process_gpuid_from_adev(struct kfd_process *p, struct amdgpu_device *adev, uint32_t *gpuid, uint32_t *gpuidx) { - struct kgd_dev *kgd = (struct kgd_dev *)adev; int i; for (i = 0; i < p->n_pdds; i++) - if (p->pdds[i] && p->pdds[i]->dev->kgd == kgd) { + if (p->pdds[i] && p->pdds[i]->dev->adev == adev) { *gpuid = p->pdds[i]->dev->id; *gpuidx = i; return 0; @@ -1951,10 +1949,10 @@ void kfd_flush_tlb(struct kfd_process_device *pdd, enum TLB_FLUSH_TYPE type) * only happens when the first queue is created. */ if (pdd->qpd.vmid) - amdgpu_amdkfd_flush_gpu_tlb_vmid(dev->kgd, + amdgpu_amdkfd_flush_gpu_tlb_vmid(dev->adev, pdd->qpd.vmid); } else { - amdgpu_amdkfd_flush_gpu_tlb_pasid(dev->kgd, + amdgpu_amdkfd_flush_gpu_tlb_pasid(dev->adev, pdd->process->pasid, type); } } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index 3627e7ac161b..4f8464658daf 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -118,7 +118,7 @@ int pqm_set_gws(struct process_queue_manager *pqm, unsigned int qid, return ret; pqn->q->gws = mem; - pdd->qpd.num_gws = gws ? amdgpu_amdkfd_get_num_gws(dev->kgd) : 0; + pdd->qpd.num_gws = gws ? dev->adev->gds.gws_size : 0; return pqn->q->device->dqm->ops.update_queue(pqn->q->device->dqm, pqn->q, NULL); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c index ed4bc5f844ce..deae12dc777d 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c @@ -207,7 +207,6 @@ void kfd_smi_event_update_gpu_reset(struct kfd_dev *dev, bool post_reset) void kfd_smi_event_update_thermal_throttling(struct kfd_dev *dev, uint64_t throttle_bitmask) { - struct amdgpu_device *adev = (struct amdgpu_device *)dev->kgd; /* * ThermalThrottle msg = throttle_bitmask(8): * thermal_interrupt_count(16): @@ -223,14 +222,13 @@ void kfd_smi_event_update_thermal_throttling(struct kfd_dev *dev, len = snprintf(fifo_in, sizeof(fifo_in), "%x %llx:%llx\n", KFD_SMI_EVENT_THERMAL_THROTTLE, throttle_bitmask, - atomic64_read(&adev->smu.throttle_int_counter)); + atomic64_read(&dev->adev->smu.throttle_int_counter)); add_event_to_kfifo(dev, KFD_SMI_EVENT_THERMAL_THROTTLE, fifo_in, len); } void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid) { - struct amdgpu_device *adev = (struct amdgpu_device *)dev->kgd; struct amdgpu_task_info task_info; /* VmFault msg = (hex)uint32_pid(8) + :(1) + task name(16) = 25 */ /* 1 byte event + 1 byte space + 25 bytes msg + 1 byte \n + @@ -243,7 +241,7 @@ void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid) return; memset(&task_info, 0, sizeof(struct amdgpu_task_info)); - amdgpu_vm_get_task_info(adev, pasid, &task_info); + amdgpu_vm_get_task_info(dev->adev, pasid, &task_info); /* Report VM faults from user applications, not retry from kernel */ if (!task_info.pid) return; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c index 16137c4247bb..10868d5b549f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c @@ -193,7 +193,6 @@ svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap, for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) { struct kfd_process_device *pdd; - struct amdgpu_device *adev; pr_debug("mapping to gpu idx 0x%x\n", gpuidx); pdd = kfd_process_device_from_gpuidx(p, gpuidx); @@ -201,9 +200,8 @@ svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap, pr_debug("failed to find device idx %d\n", gpuidx); return -EINVAL; } - adev = (struct amdgpu_device *)pdd->dev->kgd; - r = svm_range_dma_map_dev(adev, prange, offset, npages, + r = svm_range_dma_map_dev(pdd->dev->adev, prange, offset, npages, hmm_pfns, gpuidx); if (r) break; @@ -581,7 +579,7 @@ svm_range_get_adev_by_id(struct svm_range *prange, uint32_t gpu_id) return NULL; } - return (struct amdgpu_device *)pdd->dev->kgd; + return pdd->dev->adev; } struct kfd_process_device * @@ -593,7 +591,7 @@ svm_range_get_pdd_by_adev(struct svm_range *prange, struct amdgpu_device *adev) p = container_of(prange->svms, struct kfd_process, svms); - r = kfd_process_gpuid_from_kgd(p, adev, &gpuid, &gpu_idx); + r = kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpu_idx); if (r) { pr_debug("failed to get device id by adev %p\n", adev); return NULL; @@ -1053,8 +1051,8 @@ svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange, if (domain == SVM_RANGE_VRAM_DOMAIN) bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev); - switch (adev->asic_type) { - case CHIP_ARCTURUS: + switch (KFD_GC_VERSION(adev->kfd.dev)) { + case IP_VERSION(9, 4, 1): if (domain == SVM_RANGE_VRAM_DOMAIN) { if (bo_adev == adev) { mapping_flags |= coherent ? @@ -1070,7 +1068,7 @@ svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange, AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; } break; - case CHIP_ALDEBARAN: + case IP_VERSION(9, 4, 2): if (domain == SVM_RANGE_VRAM_DOMAIN) { if (bo_adev == adev) { mapping_flags |= coherent ? @@ -1129,7 +1127,6 @@ svm_range_unmap_from_gpus(struct svm_range *prange, unsigned long start, DECLARE_BITMAP(bitmap, MAX_GPU_INSTANCE); struct kfd_process_device *pdd; struct dma_fence *fence = NULL; - struct amdgpu_device *adev; struct kfd_process *p; uint32_t gpuidx; int r = 0; @@ -1145,9 +1142,9 @@ svm_range_unmap_from_gpus(struct svm_range *prange, unsigned long start, pr_debug("failed to find device idx %d\n", gpuidx); return -EINVAL; } - adev = (struct amdgpu_device *)pdd->dev->kgd; - r = svm_range_unmap_from_gpu(adev, drm_priv_to_vm(pdd->drm_priv), + r = svm_range_unmap_from_gpu(pdd->dev->adev, + drm_priv_to_vm(pdd->drm_priv), start, last, &fence); if (r) break; @@ -1159,7 +1156,7 @@ svm_range_unmap_from_gpus(struct svm_range *prange, unsigned long start, if (r) break; } - amdgpu_amdkfd_flush_gpu_tlb_pasid((struct kgd_dev *)adev, + amdgpu_amdkfd_flush_gpu_tlb_pasid(pdd->dev->adev, p->pasid, TLB_FLUSH_HEAVYWEIGHT); } @@ -1243,8 +1240,7 @@ svm_range_map_to_gpu(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct kfd_process *p; p = container_of(prange->svms, struct kfd_process, svms); - amdgpu_amdkfd_flush_gpu_tlb_pasid((struct kgd_dev *)adev, - p->pasid, TLB_FLUSH_LEGACY); + amdgpu_amdkfd_flush_gpu_tlb_pasid(adev, p->pasid, TLB_FLUSH_LEGACY); } out: return r; @@ -1257,7 +1253,6 @@ svm_range_map_to_gpus(struct svm_range *prange, unsigned long offset, { struct kfd_process_device *pdd; struct amdgpu_device *bo_adev; - struct amdgpu_device *adev; struct kfd_process *p; struct dma_fence *fence = NULL; uint32_t gpuidx; @@ -1276,19 +1271,18 @@ svm_range_map_to_gpus(struct svm_range *prange, unsigned long offset, pr_debug("failed to find device idx %d\n", gpuidx); return -EINVAL; } - adev = (struct amdgpu_device *)pdd->dev->kgd; pdd = kfd_bind_process_to_device(pdd->dev, p); if (IS_ERR(pdd)) return -EINVAL; - if (bo_adev && adev != bo_adev && - !amdgpu_xgmi_same_hive(adev, bo_adev)) { + if (bo_adev && pdd->dev->adev != bo_adev && + !amdgpu_xgmi_same_hive(pdd->dev->adev, bo_adev)) { pr_debug("cannot map to device idx %d\n", gpuidx); continue; } - r = svm_range_map_to_gpu(adev, drm_priv_to_vm(pdd->drm_priv), + r = svm_range_map_to_gpu(pdd->dev->adev, drm_priv_to_vm(pdd->drm_priv), prange, offset, npages, readonly, prange->dma_addr[gpuidx], bo_adev, wait ? &fence : NULL); @@ -1322,7 +1316,6 @@ struct svm_validate_context { static int svm_range_reserve_bos(struct svm_validate_context *ctx) { struct kfd_process_device *pdd; - struct amdgpu_device *adev; struct amdgpu_vm *vm; uint32_t gpuidx; int r; @@ -1334,7 +1327,6 @@ static int svm_range_reserve_bos(struct svm_validate_context *ctx) pr_debug("failed to find device idx %d\n", gpuidx); return -EINVAL; } - adev = (struct amdgpu_device *)pdd->dev->kgd; vm = drm_priv_to_vm(pdd->drm_priv); ctx->tv[gpuidx].bo = &vm->root.bo->tbo; @@ -1356,9 +1348,9 @@ static int svm_range_reserve_bos(struct svm_validate_context *ctx) r = -EINVAL; goto unreserve_out; } - adev = (struct amdgpu_device *)pdd->dev->kgd; - r = amdgpu_vm_validate_pt_bos(adev, drm_priv_to_vm(pdd->drm_priv), + r = amdgpu_vm_validate_pt_bos(pdd->dev->adev, + drm_priv_to_vm(pdd->drm_priv), svm_range_bo_validate, NULL); if (r) { pr_debug("failed %d validate pt bos\n", r); @@ -1381,12 +1373,10 @@ static void svm_range_unreserve_bos(struct svm_validate_context *ctx) static void *kfd_svm_page_owner(struct kfd_process *p, int32_t gpuidx) { struct kfd_process_device *pdd; - struct amdgpu_device *adev; pdd = kfd_process_device_from_gpuidx(p, gpuidx); - adev = (struct amdgpu_device *)pdd->dev->kgd; - return SVM_ADEV_PGMAP_OWNER(adev); + return SVM_ADEV_PGMAP_OWNER(pdd->dev->adev); } /* @@ -1966,23 +1956,30 @@ svm_range_handle_list_op(struct svm_range_list *svms, struct svm_range *prange) static void svm_range_drain_retry_fault(struct svm_range_list *svms) { struct kfd_process_device *pdd; - struct amdgpu_device *adev; struct kfd_process *p; + int drain; uint32_t i; p = container_of(svms, struct kfd_process, svms); +restart: + drain = atomic_read(&svms->drain_pagefaults); + if (!drain) + return; + for_each_set_bit(i, svms->bitmap_supported, p->n_pdds) { pdd = p->pdds[i]; if (!pdd) continue; pr_debug("drain retry fault gpu %d svms %p\n", i, svms); - adev = (struct amdgpu_device *)pdd->dev->kgd; - amdgpu_ih_wait_on_checkpoint_process(adev, &adev->irq.ih1); + amdgpu_ih_wait_on_checkpoint_process(pdd->dev->adev, + &pdd->dev->adev->irq.ih1); pr_debug("drain retry fault gpu %d svms 0x%p done\n", i, svms); } + if (atomic_cmpxchg(&svms->drain_pagefaults, drain, 0) != drain) + goto restart; } static void svm_range_deferred_list_work(struct work_struct *work) @@ -1990,43 +1987,41 @@ static void svm_range_deferred_list_work(struct work_struct *work) struct svm_range_list *svms; struct svm_range *prange; struct mm_struct *mm; + struct kfd_process *p; svms = container_of(work, struct svm_range_list, deferred_list_work); pr_debug("enter svms 0x%p\n", svms); + p = container_of(svms, struct kfd_process, svms); + /* Avoid mm is gone when inserting mmu notifier */ + mm = get_task_mm(p->lead_thread); + if (!mm) { + pr_debug("svms 0x%p process mm gone\n", svms); + return; + } +retry: + mmap_write_lock(mm); + + /* Checking for the need to drain retry faults must be inside + * mmap write lock to serialize with munmap notifiers. + */ + if (unlikely(atomic_read(&svms->drain_pagefaults))) { + mmap_write_unlock(mm); + svm_range_drain_retry_fault(svms); + goto retry; + } + spin_lock(&svms->deferred_list_lock); while (!list_empty(&svms->deferred_range_list)) { prange = list_first_entry(&svms->deferred_range_list, struct svm_range, deferred_list); - spin_unlock(&svms->deferred_list_lock); - pr_debug("prange 0x%p [0x%lx 0x%lx] op %d\n", prange, - prange->start, prange->last, prange->work_item.op); - - mm = prange->work_item.mm; -retry: - mmap_write_lock(mm); - mutex_lock(&svms->lock); - - /* Checking for the need to drain retry faults must be in - * mmap write lock to serialize with munmap notifiers. - * - * Remove from deferred_list must be inside mmap write lock, - * otherwise, svm_range_list_lock_and_flush_work may hold mmap - * write lock, and continue because deferred_list is empty, then - * deferred_list handle is blocked by mmap write lock. - */ - spin_lock(&svms->deferred_list_lock); - if (unlikely(svms->drain_pagefaults)) { - svms->drain_pagefaults = false; - spin_unlock(&svms->deferred_list_lock); - mutex_unlock(&svms->lock); - mmap_write_unlock(mm); - svm_range_drain_retry_fault(svms); - goto retry; - } list_del_init(&prange->deferred_list); spin_unlock(&svms->deferred_list_lock); + pr_debug("prange 0x%p [0x%lx 0x%lx] op %d\n", prange, + prange->start, prange->last, prange->work_item.op); + + mutex_lock(&svms->lock); mutex_lock(&prange->migrate_mutex); while (!list_empty(&prange->child_list)) { struct svm_range *pchild; @@ -2042,12 +2037,13 @@ static void svm_range_deferred_list_work(struct work_struct *work) svm_range_handle_list_op(svms, prange); mutex_unlock(&svms->lock); - mmap_write_unlock(mm); spin_lock(&svms->deferred_list_lock); } spin_unlock(&svms->deferred_list_lock); + mmap_write_unlock(mm); + mmput(mm); pr_debug("exit svms 0x%p\n", svms); } @@ -2056,12 +2052,6 @@ svm_range_add_list_work(struct svm_range_list *svms, struct svm_range *prange, struct mm_struct *mm, enum svm_work_list_ops op) { spin_lock(&svms->deferred_list_lock); - /* Make sure pending page faults are drained in the deferred worker - * before the range is freed to avoid straggler interrupts on - * unmapped memory causing "phantom faults". - */ - if (op == SVM_OP_UNMAP_RANGE) - svms->drain_pagefaults = true; /* if prange is on the deferred list */ if (!list_empty(&prange->deferred_list)) { pr_debug("update exist prange 0x%p work op %d\n", prange, op); @@ -2140,6 +2130,12 @@ svm_range_unmap_from_cpu(struct mm_struct *mm, struct svm_range *prange, pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx] [0x%lx 0x%lx]\n", svms, prange, prange->start, prange->last, start, last); + /* Make sure pending page faults are drained in the deferred worker + * before the range is freed to avoid straggler interrupts on + * unmapped memory causing "phantom faults". + */ + atomic_inc(&svms->drain_pagefaults); + unmap_parent = start <= prange->start && last >= prange->last; list_for_each_entry(pchild, &prange->child_list, child_list) { @@ -2301,7 +2297,7 @@ svm_range_best_restore_location(struct svm_range *prange, p = container_of(prange->svms, struct kfd_process, svms); - r = kfd_process_gpuid_from_kgd(p, adev, &gpuid, gpuidx); + r = kfd_process_gpuid_from_adev(p, adev, &gpuid, gpuidx); if (r < 0) { pr_debug("failed to get gpuid from kgd\n"); return -1; @@ -2478,7 +2474,7 @@ svm_range *svm_range_create_unregistered_range(struct amdgpu_device *adev, pr_debug("Failed to create prange in address [0x%llx]\n", addr); return NULL; } - if (kfd_process_gpuid_from_kgd(p, adev, &gpuid, &gpuidx)) { + if (kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpuidx)) { pr_debug("failed to get gpuid from kgd\n"); svm_range_free(prange); return NULL; @@ -2545,7 +2541,7 @@ svm_range_count_fault(struct amdgpu_device *adev, struct kfd_process *p, uint32_t gpuid; int r; - r = kfd_process_gpuid_from_kgd(p, adev, &gpuid, &gpuidx); + r = kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpuidx); if (r < 0) return; } @@ -2559,20 +2555,13 @@ svm_range_count_fault(struct amdgpu_device *adev, struct kfd_process *p, } static bool -svm_fault_allowed(struct mm_struct *mm, uint64_t addr, bool write_fault) +svm_fault_allowed(struct vm_area_struct *vma, bool write_fault) { unsigned long requested = VM_READ; - struct vm_area_struct *vma; if (write_fault) requested |= VM_WRITE; - vma = find_vma(mm, addr << PAGE_SHIFT); - if (!vma || (addr << PAGE_SHIFT) < vma->vm_start) { - pr_debug("address 0x%llx VMA is removed\n", addr); - return true; - } - pr_debug("requested 0x%lx, vma permission flags 0x%lx\n", requested, vma->vm_flags); return (vma->vm_flags & requested) == requested; @@ -2590,6 +2579,7 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid, int32_t best_loc; int32_t gpuidx = MAX_GPU_INSTANCE; bool write_locked = false; + struct vm_area_struct *vma; int r = 0; if (!KFD_IS_SVM_API_SUPPORTED(adev->kfd.dev)) { @@ -2600,7 +2590,7 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid, p = kfd_lookup_process_by_pasid(pasid); if (!p) { pr_debug("kfd process not founded pasid 0x%x\n", pasid); - return -ESRCH; + return 0; } if (!p->xnack_enabled) { pr_debug("XNACK not enabled for pasid 0x%x\n", pasid); @@ -2611,10 +2601,17 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid, pr_debug("restoring svms 0x%p fault address 0x%llx\n", svms, addr); + if (atomic_read(&svms->drain_pagefaults)) { + pr_debug("draining retry fault, drop fault 0x%llx\n", addr); + goto out; + } + + /* p->lead_thread is available as kfd_process_wq_release flush the work + * before releasing task ref. + */ mm = get_task_mm(p->lead_thread); if (!mm) { pr_debug("svms 0x%p failed to get mm\n", svms); - r = -ESRCH; goto out; } @@ -2663,7 +2660,17 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid, goto out_unlock_range; } - if (!svm_fault_allowed(mm, addr, write_fault)) { + /* __do_munmap removed VMA, return success as we are handling stale + * retry fault. + */ + vma = find_vma(mm, addr << PAGE_SHIFT); + if (!vma || (addr << PAGE_SHIFT) < vma->vm_start) { + pr_debug("address 0x%llx VMA is removed\n", addr); + r = 0; + goto out_unlock_range; + } + + if (!svm_fault_allowed(vma, write_fault)) { pr_debug("fault addr 0x%llx no %s permission\n", addr, write_fault ? "write" : "read"); r = -EPERM; @@ -2741,6 +2748,14 @@ void svm_range_list_fini(struct kfd_process *p) /* Ensure list work is finished before process is destroyed */ flush_work(&p->svms.deferred_list_work); + /* + * Ensure no retry fault comes in afterwards, as page fault handler will + * not find kfd process and take mm lock to recover fault. + */ + atomic_inc(&p->svms.drain_pagefaults); + svm_range_drain_retry_fault(&p->svms); + + list_for_each_entry_safe(prange, next, &p->svms.list, list) { svm_range_unlink(prange); svm_range_remove_notifier(prange); @@ -2761,6 +2776,7 @@ int svm_range_list_init(struct kfd_process *p) mutex_init(&svms->lock); INIT_LIST_HEAD(&svms->list); atomic_set(&svms->evicted_ranges, 0); + atomic_set(&svms->drain_pagefaults, 0); INIT_DELAYED_WORK(&svms->restore_work, svm_range_restore_work); INIT_WORK(&svms->deferred_list_work, svm_range_deferred_list_work); INIT_LIST_HEAD(&svms->deferred_range_list); @@ -2953,7 +2969,6 @@ svm_range_best_prefetch_location(struct svm_range *prange) uint32_t best_loc = prange->prefetch_loc; struct kfd_process_device *pdd; struct amdgpu_device *bo_adev; - struct amdgpu_device *adev; struct kfd_process *p; uint32_t gpuidx; @@ -2981,12 +2996,11 @@ svm_range_best_prefetch_location(struct svm_range *prange) pr_debug("failed to get device by idx 0x%x\n", gpuidx); continue; } - adev = (struct amdgpu_device *)pdd->dev->kgd; - if (adev == bo_adev) + if (pdd->dev->adev == bo_adev) continue; - if (!amdgpu_xgmi_same_hive(adev, bo_adev)) { + if (!amdgpu_xgmi_same_hive(pdd->dev->adev, bo_adev)) { best_loc = 0; break; } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c index dd593ad0614a..2d44b26b6657 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c @@ -113,7 +113,7 @@ struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev) return device; } -struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd) +struct kfd_dev *kfd_device_by_adev(const struct amdgpu_device *adev) { struct kfd_topology_device *top_dev; struct kfd_dev *device = NULL; @@ -121,7 +121,7 @@ struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd) down_read(&topology_lock); list_for_each_entry(top_dev, &topology_device_list, list) - if (top_dev->gpu && top_dev->gpu->kgd == kgd) { + if (top_dev->gpu && top_dev->gpu->adev == adev) { device = top_dev->gpu; break; } @@ -515,7 +515,7 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr, HSA_CAP_WATCH_POINTS_TOTALBITS_MASK); } - if (dev->gpu->device_info->asic_family == CHIP_TONGA) + if (dev->gpu->adev->asic_type == CHIP_TONGA) dev->node_props.capability |= HSA_CAP_AQL_QUEUE_DOUBLE_MAP; @@ -531,7 +531,7 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr, sysfs_show_32bit_prop(buffer, offs, "sdma_fw_version", dev->gpu->sdma_fw_version); sysfs_show_64bit_prop(buffer, offs, "unique_id", - amdgpu_amdkfd_get_unique_id(dev->gpu->kgd)); + dev->gpu->adev->unique_id); } @@ -1106,7 +1106,7 @@ static uint32_t kfd_generate_gpu_id(struct kfd_dev *gpu) if (!gpu) return 0; - amdgpu_amdkfd_get_local_mem_info(gpu->kgd, &local_mem_info); + amdgpu_amdkfd_get_local_mem_info(gpu->adev, &local_mem_info); local_mem_size = local_mem_info.local_mem_size_private + local_mem_info.local_mem_size_public; @@ -1189,7 +1189,7 @@ static void kfd_fill_mem_clk_max_info(struct kfd_topology_device *dev) * for APUs - If CRAT from ACPI reports more than one bank, then * all the banks will report the same mem_clk_max information */ - amdgpu_amdkfd_get_local_mem_info(dev->gpu->kgd, &local_mem_info); + amdgpu_amdkfd_get_local_mem_info(dev->gpu->adev, &local_mem_info); list_for_each_entry(mem, &dev->mem_props, list) mem->mem_clk_max = local_mem_info.mem_clk_max; @@ -1217,8 +1217,7 @@ static void kfd_set_iolink_no_atomics(struct kfd_topology_device *dev, /* set gpu (dev) flags. */ } else { if (!dev->gpu->pci_atomic_requested || - dev->gpu->device_info->asic_family == - CHIP_HAWAII) + dev->gpu->adev->asic_type == CHIP_HAWAII) link->flags |= CRAT_IOLINK_FLAGS_NO_ATOMICS_32_BIT | CRAT_IOLINK_FLAGS_NO_ATOMICS_64_BIT; } @@ -1239,7 +1238,7 @@ static void kfd_set_iolink_non_coherent(struct kfd_topology_device *to_dev, */ if (inbound_link->iolink_type == CRAT_IOLINK_TYPE_PCIEXPRESS || (inbound_link->iolink_type == CRAT_IOLINK_TYPE_XGMI && - to_dev->gpu->device_info->asic_family == CHIP_VEGA20)) { + KFD_GC_VERSION(to_dev->gpu) == IP_VERSION(9, 4, 0))) { outbound_link->flags |= CRAT_IOLINK_FLAGS_NON_COHERENT; inbound_link->flags |= CRAT_IOLINK_FLAGS_NON_COHERENT; } @@ -1286,7 +1285,6 @@ int kfd_topology_add_device(struct kfd_dev *gpu) void *crat_image = NULL; size_t image_size = 0; int proximity_domain; - struct amdgpu_device *adev; INIT_LIST_HEAD(&temp_topology_device_list); @@ -1296,10 +1294,8 @@ int kfd_topology_add_device(struct kfd_dev *gpu) proximity_domain = atomic_inc_return(&topology_crat_proximity_domain); - adev = (struct amdgpu_device *)(gpu->kgd); - /* Include the CPU in xGMI hive if xGMI connected by assigning it the hive ID. */ - if (gpu->hive_id && adev->gmc.xgmi.connected_to_cpu) { + if (gpu->hive_id && gpu->adev->gmc.xgmi.connected_to_cpu) { struct kfd_topology_device *top_dev; down_read(&topology_lock); @@ -1372,7 +1368,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu) * needed for the topology */ - amdgpu_amdkfd_get_cu_info(dev->gpu->kgd, &cu_info); + amdgpu_amdkfd_get_cu_info(dev->gpu->adev, &cu_info); strncpy(dev->node_props.name, gpu->device_info->asic_name, KFD_TOPOLOGY_PUBLIC_NAME_SIZE); @@ -1384,33 +1380,32 @@ int kfd_topology_add_device(struct kfd_dev *gpu) dev->node_props.vendor_id = gpu->pdev->vendor; dev->node_props.device_id = gpu->pdev->device; dev->node_props.capability |= - ((amdgpu_amdkfd_get_asic_rev_id(dev->gpu->kgd) << - HSA_CAP_ASIC_REVISION_SHIFT) & + ((dev->gpu->adev->rev_id << HSA_CAP_ASIC_REVISION_SHIFT) & HSA_CAP_ASIC_REVISION_MASK); dev->node_props.location_id = pci_dev_id(gpu->pdev); dev->node_props.domain = pci_domain_nr(gpu->pdev->bus); dev->node_props.max_engine_clk_fcompute = - amdgpu_amdkfd_get_max_engine_clock_in_mhz(dev->gpu->kgd); + amdgpu_amdkfd_get_max_engine_clock_in_mhz(dev->gpu->adev); dev->node_props.max_engine_clk_ccompute = cpufreq_quick_get_max(0) / 1000; dev->node_props.drm_render_minor = gpu->shared_resources.drm_render_minor; dev->node_props.hive_id = gpu->hive_id; - dev->node_props.num_sdma_engines = gpu->device_info->num_sdma_engines; + dev->node_props.num_sdma_engines = kfd_get_num_sdma_engines(gpu); dev->node_props.num_sdma_xgmi_engines = - gpu->device_info->num_xgmi_sdma_engines; + kfd_get_num_xgmi_sdma_engines(gpu); dev->node_props.num_sdma_queues_per_engine = gpu->device_info->num_sdma_queues_per_engine; dev->node_props.num_gws = (dev->gpu->gws && dev->gpu->dqm->sched_policy != KFD_SCHED_POLICY_NO_HWS) ? - amdgpu_amdkfd_get_num_gws(dev->gpu->kgd) : 0; + dev->gpu->adev->gds.gws_size : 0; dev->node_props.num_cp_queues = get_cp_queues_num(dev->gpu->dqm); kfd_fill_mem_clk_max_info(dev); kfd_fill_iolink_non_crat_info(dev); - switch (dev->gpu->device_info->asic_family) { + switch (dev->gpu->adev->asic_type) { case CHIP_KAVERI: case CHIP_HAWAII: case CHIP_TONGA: @@ -1429,30 +1424,14 @@ int kfd_topology_add_device(struct kfd_dev *gpu) HSA_CAP_DOORBELL_TYPE_TOTALBITS_SHIFT) & HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK); break; - case CHIP_VEGA10: - case CHIP_VEGA12: - case CHIP_VEGA20: - case CHIP_RAVEN: - case CHIP_RENOIR: - case CHIP_ARCTURUS: - case CHIP_ALDEBARAN: - case CHIP_NAVI10: - case CHIP_NAVI12: - case CHIP_NAVI14: - case CHIP_SIENNA_CICHLID: - case CHIP_NAVY_FLOUNDER: - case CHIP_VANGOGH: - case CHIP_DIMGREY_CAVEFISH: - case CHIP_BEIGE_GOBY: - case CHIP_YELLOW_CARP: - case CHIP_CYAN_SKILLFISH: - dev->node_props.capability |= ((HSA_CAP_DOORBELL_TYPE_2_0 << - HSA_CAP_DOORBELL_TYPE_TOTALBITS_SHIFT) & - HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK); - break; default: - WARN(1, "Unexpected ASIC family %u", - dev->gpu->device_info->asic_family); + if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(9, 0, 1)) + dev->node_props.capability |= ((HSA_CAP_DOORBELL_TYPE_2_0 << + HSA_CAP_DOORBELL_TYPE_TOTALBITS_SHIFT) & + HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK); + else + WARN(1, "Unexpected ASIC family %u", + dev->gpu->adev->asic_type); } /* @@ -1469,7 +1448,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu) * because it doesn't consider masked out CUs * max_waves_per_simd: Carrizo reports wrong max_waves_per_simd */ - if (dev->gpu->device_info->asic_family == CHIP_CARRIZO) { + if (dev->gpu->adev->asic_type == CHIP_CARRIZO) { dev->node_props.simd_count = cu_info.simd_per_cu * cu_info.cu_active_number; dev->node_props.max_waves_per_simd = 10; @@ -1477,16 +1456,17 @@ int kfd_topology_add_device(struct kfd_dev *gpu) /* kfd only concerns sram ecc on GFX and HBM ecc on UMC */ dev->node_props.capability |= - ((adev->ras_enabled & BIT(AMDGPU_RAS_BLOCK__GFX)) != 0) ? + ((dev->gpu->adev->ras_enabled & BIT(AMDGPU_RAS_BLOCK__GFX)) != 0) ? HSA_CAP_SRAM_EDCSUPPORTED : 0; - dev->node_props.capability |= ((adev->ras_enabled & BIT(AMDGPU_RAS_BLOCK__UMC)) != 0) ? + dev->node_props.capability |= + ((dev->gpu->adev->ras_enabled & BIT(AMDGPU_RAS_BLOCK__UMC)) != 0) ? HSA_CAP_MEM_EDCSUPPORTED : 0; - if (adev->asic_type != CHIP_VEGA10) - dev->node_props.capability |= (adev->ras_enabled != 0) ? + if (KFD_GC_VERSION(dev->gpu) != IP_VERSION(9, 0, 1)) + dev->node_props.capability |= (dev->gpu->adev->ras_enabled != 0) ? HSA_CAP_RASEVENTNOTIFY : 0; - if (KFD_IS_SVM_API_SUPPORTED(adev->kfd.dev)) + if (KFD_IS_SVM_API_SUPPORTED(dev->gpu->adev->kfd.dev)) dev->node_props.capability |= HSA_CAP_SVMAPI_SUPPORTED; kfd_debug_print_topology(); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h index a8db017c9b8e..f0cc59d2fd5d 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h @@ -25,38 +25,11 @@ #include #include +#include #include "kfd_crat.h" #define KFD_TOPOLOGY_PUBLIC_NAME_SIZE 32 -#define HSA_CAP_HOT_PLUGGABLE 0x00000001 -#define HSA_CAP_ATS_PRESENT 0x00000002 -#define HSA_CAP_SHARED_WITH_GRAPHICS 0x00000004 -#define HSA_CAP_QUEUE_SIZE_POW2 0x00000008 -#define HSA_CAP_QUEUE_SIZE_32BIT 0x00000010 -#define HSA_CAP_QUEUE_IDLE_EVENT 0x00000020 -#define HSA_CAP_VA_LIMIT 0x00000040 -#define HSA_CAP_WATCH_POINTS_SUPPORTED 0x00000080 -#define HSA_CAP_WATCH_POINTS_TOTALBITS_MASK 0x00000f00 -#define HSA_CAP_WATCH_POINTS_TOTALBITS_SHIFT 8 -#define HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK 0x00003000 -#define HSA_CAP_DOORBELL_TYPE_TOTALBITS_SHIFT 12 - -#define HSA_CAP_DOORBELL_TYPE_PRE_1_0 0x0 -#define HSA_CAP_DOORBELL_TYPE_1_0 0x1 -#define HSA_CAP_DOORBELL_TYPE_2_0 0x2 -#define HSA_CAP_AQL_QUEUE_DOUBLE_MAP 0x00004000 - -#define HSA_CAP_RESERVED_WAS_SRAM_EDCSUPPORTED 0x00080000 /* Old buggy user mode depends on this being 0 */ -#define HSA_CAP_MEM_EDCSUPPORTED 0x00100000 -#define HSA_CAP_RASEVENTNOTIFY 0x00200000 -#define HSA_CAP_ASIC_REVISION_MASK 0x03c00000 -#define HSA_CAP_ASIC_REVISION_SHIFT 22 -#define HSA_CAP_SRAM_EDCSUPPORTED 0x04000000 -#define HSA_CAP_SVMAPI_SUPPORTED 0x08000000 -#define HSA_CAP_FLAGS_COHERENTHOSTACCESS 0x10000000 -#define HSA_CAP_RESERVED 0xe00f8000 - struct kfd_node_properties { uint64_t hive_id; uint32_t cpu_cores_count; @@ -93,17 +66,6 @@ struct kfd_node_properties { char name[KFD_TOPOLOGY_PUBLIC_NAME_SIZE]; }; -#define HSA_MEM_HEAP_TYPE_SYSTEM 0 -#define HSA_MEM_HEAP_TYPE_FB_PUBLIC 1 -#define HSA_MEM_HEAP_TYPE_FB_PRIVATE 2 -#define HSA_MEM_HEAP_TYPE_GPU_GDS 3 -#define HSA_MEM_HEAP_TYPE_GPU_LDS 4 -#define HSA_MEM_HEAP_TYPE_GPU_SCRATCH 5 - -#define HSA_MEM_FLAGS_HOT_PLUGGABLE 0x00000001 -#define HSA_MEM_FLAGS_NON_VOLATILE 0x00000002 -#define HSA_MEM_FLAGS_RESERVED 0xfffffffc - struct kfd_mem_properties { struct list_head list; uint32_t heap_type; @@ -116,12 +78,6 @@ struct kfd_mem_properties { struct attribute attr; }; -#define HSA_CACHE_TYPE_DATA 0x00000001 -#define HSA_CACHE_TYPE_INSTRUCTION 0x00000002 -#define HSA_CACHE_TYPE_CPU 0x00000004 -#define HSA_CACHE_TYPE_HSACU 0x00000008 -#define HSA_CACHE_TYPE_RESERVED 0xfffffff0 - struct kfd_cache_properties { struct list_head list; uint32_t processor_id_low; diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index 4595c59f2bf0..3fcf911c8b61 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -51,6 +51,7 @@ #include #endif #include "amdgpu_pm.h" +#include "amdgpu_atombios.h" #include "amd_shared.h" #include "amdgpu_dm_irq.h" @@ -789,8 +790,7 @@ static void dm_dmub_outbox1_low_irq(void *interrupt_params) plink = adev->dm.dc->links[notify.link_index]; if (plink) { plink->hpd_status = - notify.hpd_status == - DP_HPD_PLUG ? true : false; + notify.hpd_status == DP_HPD_PLUG; } } queue_work(adev->dm.delayed_hpd_wq, &dmub_hpd_wrk->handle_hpd_work); @@ -1455,6 +1455,12 @@ static int amdgpu_dm_init(struct amdgpu_device *adev) init_data.flags.power_down_display_on_boot = true; + if (check_seamless_boot_capability(adev)) { + init_data.flags.power_down_display_on_boot = false; + init_data.flags.allow_seamless_boot_optimization = true; + DRM_INFO("Seamless boot condition check passed\n"); + } + INIT_LIST_HEAD(&adev->dm.da_list); /* Display Core create. */ adev->dm.dc = dc_create(&init_data); @@ -1479,8 +1485,10 @@ static int amdgpu_dm_init(struct amdgpu_device *adev) if (amdgpu_dc_debug_mask & DC_DISABLE_STUTTER) adev->dm.dc->debug.disable_stutter = true; - if (amdgpu_dc_debug_mask & DC_DISABLE_DSC) + if (amdgpu_dc_debug_mask & DC_DISABLE_DSC) { adev->dm.dc->debug.disable_dsc = true; + adev->dm.dc->debug.disable_dsc_edp = true; + } if (amdgpu_dc_debug_mask & DC_DISABLE_CLOCK_GATING) adev->dm.dc->debug.disable_clock_gate = true; @@ -2303,14 +2311,6 @@ static enum dc_status amdgpu_dm_commit_zero_streams(struct dc *dc) goto fail; } - - res = dc_validate_global_state(dc, context, false); - - if (res != DC_OK) { - DRM_ERROR("%s:resource validation failed, dc_status:%d\n", __func__, res); - goto fail; - } - res = dc_commit_state(dc, context); fail: @@ -2561,6 +2561,22 @@ static int dm_resume(void *handle) if (amdgpu_in_reset(adev)) { dc_state = dm->cached_dc_state; + /* + * The dc->current_state is backed up into dm->cached_dc_state + * before we commit 0 streams. + * + * DC will clear link encoder assignments on the real state + * but the changes won't propagate over to the copy we made + * before the 0 streams commit. + * + * DC expects that link encoder assignments are *not* valid + * when committing a state, so as a workaround it needs to be + * cleared here. + */ + link_enc_cfg_init(dm->dc, dc_state); + + amdgpu_dm_outbox_init(adev); + r = dm_dmub_hw_init(adev); if (r) DRM_ERROR("DMUB interface failed to initialize: status=%d\n", r); @@ -2572,20 +2588,11 @@ static int dm_resume(void *handle) for (i = 0; i < dc_state->stream_count; i++) { dc_state->streams[i]->mode_changed = true; - for (j = 0; j < dc_state->stream_status->plane_count; j++) { - dc_state->stream_status->plane_states[j]->update_flags.raw + for (j = 0; j < dc_state->stream_status[i].plane_count; j++) { + dc_state->stream_status[i].plane_states[j]->update_flags.raw = 0xffffffff; } } -#if defined(CONFIG_DRM_AMD_DC_DCN) - /* - * Resource allocation happens for link encoders for newer ASIC in - * dc_validate_global_state, so we need to revalidate it. - * - * This shouldn't fail (it passed once before), so warn if it does. - */ - WARN_ON(dc_validate_global_state(dm->dc, dc_state, false) != DC_OK); -#endif WARN_ON(!dc_commit_state(dm->dc, dc_state)); @@ -3909,6 +3916,9 @@ static int amdgpu_dm_backlight_set_level(struct amdgpu_display_manager *dm, caps = dm->backlight_caps[bl_idx]; dm->brightness[bl_idx] = user_brightness; + /* update scratch register */ + if (bl_idx == 0) + amdgpu_atombios_scratch_regs_set_backlight_level(dm->adev, dm->brightness[bl_idx]); brightness = convert_brightness_from_user(&caps, dm->brightness[bl_idx]); link = (struct dc_link *)dm->backlight_link[bl_idx]; @@ -4251,6 +4261,14 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev) } + /* + * Disable vblank IRQs aggressively for power-saving. + * + * TODO: Fix vblank control helpers to delay PSR entry to allow this when PSR + * is also supported. + */ + adev_to_drm(adev)->vblank_disable_immediate = !psr_feature_enabled; + /* Software is initialized. Now we can register interrupt handlers. */ switch (adev->asic_type) { #if defined(CONFIG_DRM_AMD_DC_SI) @@ -6036,7 +6054,8 @@ static void update_dsc_caps(struct amdgpu_dm_connector *aconnector, { stream->timing.flags.DSC = 0; - if (aconnector->dc_link && sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT) { + if (aconnector->dc_link && (sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT || + sink->sink_signal == SIGNAL_TYPE_EDP)) { dc_dsc_parse_dsc_dpcd(aconnector->dc_link->ctx->dc, aconnector->dc_link->dpcd_caps.dsc_caps.dsc_basic_caps.raw, aconnector->dc_link->dpcd_caps.dsc_caps.dsc_branch_decoder_caps.raw, @@ -6044,6 +6063,64 @@ static void update_dsc_caps(struct amdgpu_dm_connector *aconnector, } } +static void apply_dsc_policy_for_edp(struct amdgpu_dm_connector *aconnector, + struct dc_sink *sink, struct dc_stream_state *stream, + struct dsc_dec_dpcd_caps *dsc_caps, + uint32_t max_dsc_target_bpp_limit_override) +{ + const struct dc_link_settings *verified_link_cap = NULL; + uint32_t link_bw_in_kbps; + uint32_t edp_min_bpp_x16, edp_max_bpp_x16; + struct dc *dc = sink->ctx->dc; + struct dc_dsc_bw_range bw_range = {0}; + struct dc_dsc_config dsc_cfg = {0}; + + verified_link_cap = dc_link_get_link_cap(stream->link); + link_bw_in_kbps = dc_link_bandwidth_kbps(stream->link, verified_link_cap); + edp_min_bpp_x16 = 8 * 16; + edp_max_bpp_x16 = 8 * 16; + + if (edp_max_bpp_x16 > dsc_caps->edp_max_bits_per_pixel) + edp_max_bpp_x16 = dsc_caps->edp_max_bits_per_pixel; + + if (edp_max_bpp_x16 < edp_min_bpp_x16) + edp_min_bpp_x16 = edp_max_bpp_x16; + + if (dc_dsc_compute_bandwidth_range(dc->res_pool->dscs[0], + dc->debug.dsc_min_slice_height_override, + edp_min_bpp_x16, edp_max_bpp_x16, + dsc_caps, + &stream->timing, + &bw_range)) { + + if (bw_range.max_kbps < link_bw_in_kbps) { + if (dc_dsc_compute_config(dc->res_pool->dscs[0], + dsc_caps, + dc->debug.dsc_min_slice_height_override, + max_dsc_target_bpp_limit_override, + 0, + &stream->timing, + &dsc_cfg)) { + stream->timing.dsc_cfg = dsc_cfg; + stream->timing.flags.DSC = 1; + stream->timing.dsc_cfg.bits_per_pixel = edp_max_bpp_x16; + } + return; + } + } + + if (dc_dsc_compute_config(dc->res_pool->dscs[0], + dsc_caps, + dc->debug.dsc_min_slice_height_override, + max_dsc_target_bpp_limit_override, + link_bw_in_kbps, + &stream->timing, + &dsc_cfg)) { + stream->timing.dsc_cfg = dsc_cfg; + stream->timing.flags.DSC = 1; + } +} + static void apply_dsc_policy_for_stream(struct amdgpu_dm_connector *aconnector, struct dc_sink *sink, struct dc_stream_state *stream, struct dsc_dec_dpcd_caps *dsc_caps) @@ -6051,6 +6128,7 @@ static void apply_dsc_policy_for_stream(struct amdgpu_dm_connector *aconnector, struct drm_connector *drm_connector = &aconnector->base; uint32_t link_bandwidth_kbps; uint32_t max_dsc_target_bpp_limit_override = 0; + struct dc *dc = sink->ctx->dc; link_bandwidth_kbps = dc_link_bandwidth_kbps(aconnector->dc_link, dc_link_get_link_cap(aconnector->dc_link)); @@ -6063,7 +6141,12 @@ static void apply_dsc_policy_for_stream(struct amdgpu_dm_connector *aconnector, dc_dsc_policy_set_enable_dsc_when_not_needed( aconnector->dsc_settings.dsc_force_enable == DSC_CLK_FORCE_ENABLE); - if (aconnector->dc_link && sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT) { + if (aconnector->dc_link && sink->sink_signal == SIGNAL_TYPE_EDP && !dc->debug.disable_dsc_edp && + dc->caps.edp_dsc_support && aconnector->dsc_settings.dsc_force_enable != DSC_CLK_FORCE_DISABLE) { + + apply_dsc_policy_for_edp(aconnector, sink, stream, dsc_caps, max_dsc_target_bpp_limit_override); + + } else if (aconnector->dc_link && sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT) { if (dc_dsc_compute_config(aconnector->dc_link->ctx->dc->res_pool->dscs[0], dsc_caps, @@ -10759,8 +10842,10 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, trace_amdgpu_dm_atomic_check_begin(state); ret = drm_atomic_helper_check_modeset(dev, state); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("drm_atomic_helper_check_modeset() failed\n"); goto fail; + } /* Check connector changes */ for_each_oldnew_connector_in_state(state, connector, old_con_state, new_con_state, i) { @@ -10776,6 +10861,7 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, new_crtc_state = drm_atomic_get_crtc_state(state, new_con_state->crtc); if (IS_ERR(new_crtc_state)) { + DRM_DEBUG_DRIVER("drm_atomic_get_crtc_state() failed\n"); ret = PTR_ERR(new_crtc_state); goto fail; } @@ -10790,8 +10876,10 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { if (drm_atomic_crtc_needs_modeset(new_crtc_state)) { ret = add_affected_mst_dsc_crtcs(state, crtc); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("add_affected_mst_dsc_crtcs() failed\n"); goto fail; + } } } } @@ -10806,19 +10894,25 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, continue; ret = amdgpu_dm_verify_lut_sizes(new_crtc_state); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("amdgpu_dm_verify_lut_sizes() failed\n"); goto fail; + } if (!new_crtc_state->enable) continue; ret = drm_atomic_add_affected_connectors(state, crtc); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("drm_atomic_add_affected_connectors() failed\n"); goto fail; + } ret = drm_atomic_add_affected_planes(state, crtc); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("drm_atomic_add_affected_planes() failed\n"); goto fail; + } if (dm_old_crtc_state->dsc_force_changed) new_crtc_state->mode_changed = true; @@ -10855,6 +10949,7 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, if (IS_ERR(new_plane_state)) { ret = PTR_ERR(new_plane_state); + DRM_DEBUG_DRIVER("new_plane_state is BAD\n"); goto fail; } } @@ -10867,8 +10962,10 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, new_plane_state, false, &lock_and_validation_needed); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("dm_update_plane_state() failed\n"); goto fail; + } } /* Disable all crtcs which require disable */ @@ -10878,8 +10975,10 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, new_crtc_state, false, &lock_and_validation_needed); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("DISABLE: dm_update_crtc_state() failed\n"); goto fail; + } } /* Enable all crtcs which require enable */ @@ -10889,8 +10988,10 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, new_crtc_state, true, &lock_and_validation_needed); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("ENABLE: dm_update_crtc_state() failed\n"); goto fail; + } } /* Add new/modified planes */ @@ -10900,20 +11001,26 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, new_plane_state, true, &lock_and_validation_needed); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("dm_update_plane_state() failed\n"); goto fail; + } } /* Run this here since we want to validate the streams we created */ ret = drm_atomic_helper_check_planes(dev, state); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("drm_atomic_helper_check_planes() failed\n"); goto fail; + } /* Check cursor planes scaling */ for_each_new_crtc_in_state(state, crtc, new_crtc_state, i) { ret = dm_check_crtc_cursor(state, crtc, new_crtc_state); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("dm_check_crtc_cursor() failed\n"); goto fail; + } } if (state->legacy_cursor_update) { @@ -11000,20 +11107,28 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, */ if (lock_and_validation_needed) { ret = dm_atomic_get_state(state, &dm_state); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("dm_atomic_get_state() failed\n"); goto fail; + } ret = do_aquire_global_lock(dev, state); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("do_aquire_global_lock() failed\n"); goto fail; + } #if defined(CONFIG_DRM_AMD_DC_DCN) - if (!compute_mst_dsc_configs_for_state(state, dm_state->context, vars)) + if (!compute_mst_dsc_configs_for_state(state, dm_state->context, vars)) { + DRM_DEBUG_DRIVER("compute_mst_dsc_configs_for_state() failed\n"); goto fail; + } ret = dm_update_mst_vcpi_slots_for_dsc(state, dm_state->context, vars); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("dm_update_mst_vcpi_slots_for_dsc() failed\n"); goto fail; + } #endif /* @@ -11023,12 +11138,13 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, * to get stuck in an infinite loop and hang eventually. */ ret = drm_dp_mst_atomic_check(state); - if (ret) + if (ret) { + DRM_DEBUG_DRIVER("drm_dp_mst_atomic_check() failed\n"); goto fail; - status = dc_validate_global_state(dc, dm_state->context, false); + } + status = dc_validate_global_state(dc, dm_state->context, true); if (status != DC_OK) { - drm_dbg_atomic(dev, - "DC global validation failure: %s (%d)", + DRM_DEBUG_DRIVER("DC global validation failure: %s (%d)", dc_status_to_str(status), status); ret = -EINVAL; goto fail; @@ -11529,3 +11645,24 @@ int amdgpu_dm_process_dmub_aux_transfer_sync(bool is_cmd_aux, struct dc_context ctx, DMUB_ASYNC_TO_SYNC_ACCESS_SUCCESS, (uint32_t *)operation_result); } + +/* + * Check whether seamless boot is supported. + * + * So far we only support seamless boot on CHIP_VANGOGH. + * If everything goes well, we may consider expanding + * seamless boot to other ASICs. + */ +bool check_seamless_boot_capability(struct amdgpu_device *adev) +{ + switch (adev->asic_type) { + case CHIP_VANGOGH: + if (!adev->mman.keep_stolen_vga_memory) + return true; + break; + default: + break; + } + + return false; +} diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h index 37e61a88d49e..bb65f41d1a59 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h @@ -731,4 +731,7 @@ extern const struct drm_encoder_helper_funcs amdgpu_dm_encoder_helper_funcs; int amdgpu_dm_process_dmub_aux_transfer_sync(bool is_cmd_aux, struct dc_context *ctx, unsigned int link_index, void *payload, void *operation_result); + +bool check_seamless_boot_capability(struct amdgpu_device *adev); + #endif /* __AMDGPU_DM_H__ */ diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c index 8cbeeb7c986d..72a2e84645df 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c @@ -584,7 +584,7 @@ bool dm_helpers_dp_write_dsc_enable( ret = drm_dp_dpcd_write(aconnector->dsc_aux, DP_DSC_ENABLE, &enable_dsc, 1); } - if (stream->signal == SIGNAL_TYPE_DISPLAY_PORT) { + if (stream->signal == SIGNAL_TYPE_DISPLAY_PORT || stream->signal == SIGNAL_TYPE_EDP) { ret = dm_helpers_dp_write_dpcd(ctx, stream->link, DP_DSC_ENABLE, &enable_dsc, 1); DC_LOG_DC("Send DSC %s to sst display\n", enable_dsc ? "enable" : "disable"); } diff --git a/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c b/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c index a4bef4364afd..1e385d55e7fb 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c +++ b/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c @@ -2995,7 +2995,7 @@ static bool bios_parser2_construct( &bp->object_info_tbl.revision); if (bp->object_info_tbl.revision.major == 1 - && bp->object_info_tbl.revision.minor >= 4) { + && bp->object_info_tbl.revision.minor == 4) { struct display_object_info_table_v1_4 *tbl_v1_4; tbl_v1_4 = GET_IMAGE(struct display_object_info_table_v1_4, @@ -3004,8 +3004,10 @@ static bool bios_parser2_construct( return false; bp->object_info_tbl.v1_4 = tbl_v1_4; - } else + } else { + ASSERT(0); return false; + } dal_firmware_parser_init_cmd_tbl(bp); dal_bios_parser_init_cmd_tbl_helper2(&bp->cmd_helper, dce_version); diff --git a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c index 6b248cd2a461..c8b0a2f05b4d 100644 --- a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c +++ b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c @@ -763,7 +763,7 @@ unsigned int get_highest_allowed_voltage_level(uint32_t chip_family, uint32_t hw return 4; } -bool dcn_validate_bandwidth( +bool dcn10_validate_bandwidth( struct dc *dc, struct dc_state *context, bool fast_validate) diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c index 26f96ee32472..9200c8ce02ba 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c @@ -308,8 +308,7 @@ void dc_destroy_clk_mgr(struct clk_mgr *clk_mgr_base) case FAMILY_NV: if (ASICREV_IS_SIENNA_CICHLID_P(clk_mgr_base->ctx->asic_id.hw_internal_rev)) { dcn3_clk_mgr_destroy(clk_mgr); - } - if (ASICREV_IS_DIMGREY_CAVEFISH_P(clk_mgr_base->ctx->asic_id.hw_internal_rev)) { + } else if (ASICREV_IS_DIMGREY_CAVEFISH_P(clk_mgr_base->ctx->asic_id.hw_internal_rev)) { dcn3_clk_mgr_destroy(clk_mgr); } if (ASICREV_IS_BEIGE_GOBY_P(clk_mgr_base->ctx->asic_id.hw_internal_rev)) { diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.c index f4c9a458ace8..a13ff1783b9b 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.c @@ -66,7 +66,7 @@ #define TO_CLK_MGR_DCN31(clk_mgr)\ container_of(clk_mgr, struct clk_mgr_dcn31, base) -int dcn31_get_active_display_cnt_wa( +static int dcn31_get_active_display_cnt_wa( struct dc *dc, struct dc_state *context) { @@ -118,7 +118,7 @@ static void dcn31_disable_otg_wa(struct clk_mgr *clk_mgr_base, bool disable) } } -static void dcn31_update_clocks(struct clk_mgr *clk_mgr_base, +void dcn31_update_clocks(struct clk_mgr *clk_mgr_base, struct dc_state *context, bool safe_to_lower) { @@ -284,7 +284,7 @@ static void dcn31_enable_pme_wa(struct clk_mgr *clk_mgr_base) dcn31_smu_enable_pme_wa(clk_mgr); } -static void dcn31_init_clocks(struct clk_mgr *clk_mgr) +void dcn31_init_clocks(struct clk_mgr *clk_mgr) { memset(&(clk_mgr->clks), 0, sizeof(struct dc_clocks)); // Assumption is that boot state always supports pstate @@ -294,7 +294,7 @@ static void dcn31_init_clocks(struct clk_mgr *clk_mgr) clk_mgr->clks.zstate_support = DCN_ZSTATE_SUPPORT_UNKNOWN; } -static bool dcn31_are_clock_states_equal(struct dc_clocks *a, +bool dcn31_are_clock_states_equal(struct dc_clocks *a, struct dc_clocks *b) { if (a->dispclk_khz != b->dispclk_khz) diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.h index f8f100535526..961b10a49486 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.h +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.h @@ -39,6 +39,13 @@ struct clk_mgr_dcn31 { struct dcn31_smu_watermark_set smu_wm_set; }; +bool dcn31_are_clock_states_equal(struct dc_clocks *a, + struct dc_clocks *b); +void dcn31_init_clocks(struct clk_mgr *clk_mgr); +void dcn31_update_clocks(struct clk_mgr *clk_mgr_base, + struct dc_state *context, + bool safe_to_lower); + void dcn31_clk_mgr_construct(struct dc_context *ctx, struct clk_mgr_dcn31 *clk_mgr, struct pp_smu_funcs *pp_smu, diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 0ded4decee05..17b7408d84b7 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -221,9 +221,9 @@ static bool create_links( link = link_create(&link_init_params); if (link) { - dc->links[dc->link_count] = link; - link->dc = dc; - ++dc->link_count; + dc->links[dc->link_count] = link; + link->dc = dc; + ++dc->link_count; } } @@ -808,6 +808,10 @@ void dc_stream_set_static_screen_params(struct dc *dc, static void dc_destruct(struct dc *dc) { + // reset link encoder assignment table on destruct + if (dc->res_pool->funcs->link_encs_assign) + link_enc_cfg_init(dc, dc->current_state); + if (dc->current_state) { dc_release_state(dc->current_state); dc->current_state = NULL; @@ -1016,8 +1020,6 @@ static bool dc_construct(struct dc *dc, goto fail; } - dc_resource_state_construct(dc, dc->current_state); - if (!create_links(dc, init_params->num_virtual_links)) goto fail; @@ -1027,8 +1029,7 @@ static bool dc_construct(struct dc *dc, if (!create_link_encoders(dc)) goto fail; - /* Initialise DIG link encoder resource tracking variables. */ - link_enc_cfg_init(dc, dc->current_state); + dc_resource_state_construct(dc, dc->current_state); return true; @@ -1830,6 +1831,19 @@ bool dc_commit_state(struct dc *dc, struct dc_state *context) dc_stream_log(dc, stream); } + /* + * Previous validation was perfomred with fast_validation = true and + * the full DML state required for hardware programming was skipped. + * + * Re-validate here to calculate these parameters / watermarks. + */ + result = dc_validate_global_state(dc, context, false); + if (result != DC_OK) { + DC_LOG_ERROR("DC commit global validation failure: %s (%d)", + dc_status_to_str(result), result); + return result; + } + result = dc_commit_state_no_check(dc, context); return (result == DC_OK); @@ -2870,7 +2884,8 @@ static void commit_planes_for_stream(struct dc *dc, #endif if ((update_type != UPDATE_TYPE_FAST) && stream->update_flags.bits.dsc_changed) - if (top_pipe_to_program->stream_res.tg->funcs->lock_doublebuffer_enable) { + if (top_pipe_to_program && + top_pipe_to_program->stream_res.tg->funcs->lock_doublebuffer_enable) { if (should_use_dmub_lock(stream->link)) { union dmub_hw_lock_flags hw_locks = { 0 }; struct dmub_hw_lock_inst_flags inst_flags = { 0 }; diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link.c b/drivers/gpu/drm/amd/display/dc/core/dc_link.c index 60544788e911..3d08f8eba402 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link.c @@ -270,10 +270,10 @@ bool dc_link_detect_sink(struct dc_link *link, enum dc_connection_type *type) /* Link may not have physical HPD pin. */ if (link->ep_type != DISPLAY_ENDPOINT_PHY) { - if (link->hpd_status) - *type = dc_connection_single; - else + if (link->is_hpd_pending || !link->hpd_status) *type = dc_connection_none; + else + *type = dc_connection_single; return true; } @@ -1999,6 +1999,57 @@ static enum dc_status enable_link_dp_mst( return enable_link_dp(state, pipe_ctx); } +void dc_link_blank_all_dp_displays(struct dc *dc) +{ + unsigned int i; + uint8_t dpcd_power_state = '\0'; + enum dc_status status = DC_ERROR_UNEXPECTED; + + for (i = 0; i < dc->link_count; i++) { + if ((dc->links[i]->connector_signal != SIGNAL_TYPE_DISPLAY_PORT) || + (dc->links[i]->priv == NULL) || (dc->links[i]->local_sink == NULL)) + continue; + + /* DP 2.0 spec requires that we read LTTPR caps first */ + dp_retrieve_lttpr_cap(dc->links[i]); + /* if any of the displays are lit up turn them off */ + status = core_link_read_dpcd(dc->links[i], DP_SET_POWER, + &dpcd_power_state, sizeof(dpcd_power_state)); + + if (status == DC_OK && dpcd_power_state == DP_POWER_STATE_D0) + dc_link_blank_dp_stream(dc->links[i], true); + } + +} + +void dc_link_blank_dp_stream(struct dc_link *link, bool hw_init) +{ + unsigned int j; + struct dc *dc = link->ctx->dc; + enum signal_type signal = link->connector_signal; + + if ((signal == SIGNAL_TYPE_EDP) || + (signal == SIGNAL_TYPE_DISPLAY_PORT)) { + if (link->ep_type == DISPLAY_ENDPOINT_PHY && + link->link_enc->funcs->get_dig_frontend && + link->link_enc->funcs->is_dig_enabled(link->link_enc)) { + unsigned int fe = link->link_enc->funcs->get_dig_frontend(link->link_enc); + + if (fe != ENGINE_ID_UNKNOWN) + for (j = 0; j < dc->res_pool->stream_enc_count; j++) { + if (fe == dc->res_pool->stream_enc[j]->id) { + dc->res_pool->stream_enc[j]->funcs->dp_blank(link, + dc->res_pool->stream_enc[j]); + break; + } + } + } + + if ((!link->wa_flags.dp_keep_receiver_powered) || hw_init) + dp_receiver_power_ctrl(link, false); + } +} + static bool get_ext_hdmi_settings(struct pipe_ctx *pipe_ctx, enum engine_id eng_id, struct ext_hdmi_settings *settings) @@ -2946,7 +2997,7 @@ bool dc_link_set_psr_allow_active(struct dc_link *link, const bool *allow_active link->psr_settings.psr_power_opt = *power_opts; if (psr != NULL && link->psr_settings.psr_feature_enabled && psr->funcs->psr_set_power_opt) - psr->funcs->psr_set_power_opt(psr, link->psr_settings.psr_power_opt); + psr->funcs->psr_set_power_opt(psr, link->psr_settings.psr_power_opt, panel_inst); } /* Enable or Disable PSR */ @@ -3913,9 +3964,6 @@ static void update_psp_stream_config(struct pipe_ctx *pipe_ctx, bool dpms_off) struct cp_psp *cp_psp = &pipe_ctx->stream->ctx->cp_psp; #if defined(CONFIG_DRM_AMD_DC_DCN) struct link_encoder *link_enc = NULL; - struct dc_state *state = pipe_ctx->stream->ctx->dc->current_state; - struct link_enc_assignment link_enc_assign; - int i; #endif if (cp_psp && cp_psp->funcs.update_stream_config) { @@ -3943,18 +3991,15 @@ static void update_psp_stream_config(struct pipe_ctx *pipe_ctx, bool dpms_off) pipe_ctx->stream->ctx->dc, pipe_ctx->stream); } + ASSERT(link_enc); + // Initialize PHY ID with ABCDE - 01234 mapping except when it is B0 config.phy_idx = link_enc->transmitter - TRANSMITTER_UNIPHY_A; - //look up the link_enc_assignment for the current pipe_ctx - for (i = 0; i < state->stream_count; i++) { - if (pipe_ctx->stream == state->streams[i]) { - link_enc_assign = state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i]; - } - } // Add flag to guard new A0 DIG mapping - if (pipe_ctx->stream->ctx->dc->enable_c20_dtm_b0 == true) { - config.dig_be = link_enc_assign.eng_id; + if (pipe_ctx->stream->ctx->dc->enable_c20_dtm_b0 == true && + pipe_ctx->stream->link->dc->ctx->dce_version == DCN_VERSION_3_1) { + config.dig_be = link_enc->preferred_engine; config.dio_output_type = pipe_ctx->stream->link->ep_type; config.dio_output_idx = link_enc->transmitter - TRANSMITTER_UNIPHY_A; } else { @@ -3966,10 +4011,8 @@ static void update_psp_stream_config(struct pipe_ctx *pipe_ctx, bool dpms_off) if (pipe_ctx->stream->ctx->dc->enable_c20_dtm_b0 == true && link_enc->ctx->asic_id.hw_internal_rev == YELLOW_CARP_B0) { if (pipe_ctx->stream->link->ep_type == DISPLAY_ENDPOINT_USB4_DPIA) { - link_enc = link_enc_assign.stream->link_enc; - // enum ID 1-4 maps to DPIA PHY ID 0-3 - config.phy_idx = link_enc_assign.ep_id.link_id.enum_id - ENUM_ID_1; + config.phy_idx = pipe_ctx->stream->link->link_id.enum_id - ENUM_ID_1; } else { // for non DPIA mode over B0, ABCDE maps to 01564 switch (link_enc->transmitter) { @@ -4242,7 +4285,8 @@ void core_link_enable_stream( /* eDP lit up by bios already, no need to enable again. */ if (pipe_ctx->stream->signal == SIGNAL_TYPE_EDP && apply_edp_fast_boot_optimization && - !pipe_ctx->stream->timing.flags.DSC) { + !pipe_ctx->stream->timing.flags.DSC && + !pipe_ctx->next_odm_pipe) { pipe_ctx->stream->dpms_off = false; #if defined(CONFIG_DRM_AMD_DC_HDCP) update_psp_stream_config(pipe_ctx, false); @@ -4749,6 +4793,8 @@ bool dc_link_should_enable_fec(const struct dc_link *link) link->local_sink && link->local_sink->edid_caps.panel_patch.disable_fec) || (link->connector_signal == SIGNAL_TYPE_EDP + // enable FEC for EDP if DSC is supported + && link->dpcd_caps.dsc_caps.dsc_basic_caps.fields.dsc_support.DSC_SUPPORT == false )) is_fec_disable = true; diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c index 60539b1f2a80..24dc662ec3e4 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c @@ -626,7 +626,7 @@ bool dal_ddc_submit_aux_command(struct ddc_service *ddc, do { struct aux_payload current_payload; bool is_end_of_payload = (retrieved + DEFAULT_AUX_MAX_DATA_SIZE) >= - payload->length ? true : false; + payload->length; uint32_t payload_length = is_end_of_payload ? payload->length - retrieved : DEFAULT_AUX_MAX_DATA_SIZE; diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c index cb7bf9148904..84f3545c3032 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c @@ -430,7 +430,7 @@ enum dc_status dpcd_set_link_settings( status = core_link_write_dpcd(link, DP_LANE_COUNT_SET, &lane_count_set.raw, 1); - if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_14 && + if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_13 && lt_settings->link_settings.use_link_rate_set == true) { rate = 0; /* WA for some MUX chips that will power down with eDP and lose supported @@ -3346,6 +3346,148 @@ bool decide_edp_link_settings(struct dc_link *link, struct dc_link_settings *lin return false; } +static bool decide_edp_link_settings_with_dsc(struct dc_link *link, + struct dc_link_settings *link_setting, + uint32_t req_bw, + enum dc_link_rate max_link_rate) +{ + struct dc_link_settings initial_link_setting; + struct dc_link_settings current_link_setting; + uint32_t link_bw; + + unsigned int policy = 0; + + policy = link->ctx->dc->debug.force_dsc_edp_policy; + if (max_link_rate == LINK_RATE_UNKNOWN) + max_link_rate = link->verified_link_cap.link_rate; + /* + * edp_supported_link_rates_count is only valid for eDP v1.4 or higher. + * Per VESA eDP spec, "The DPCD revision for eDP v1.4 is 13h" + */ + if ((link->dpcd_caps.dpcd_rev.raw < DPCD_REV_13 || + link->dpcd_caps.edp_supported_link_rates_count == 0)) { + /* for DSC enabled case, we search for minimum lane count */ + memset(&initial_link_setting, 0, sizeof(initial_link_setting)); + initial_link_setting.lane_count = LANE_COUNT_ONE; + initial_link_setting.link_rate = LINK_RATE_LOW; + initial_link_setting.link_spread = LINK_SPREAD_DISABLED; + initial_link_setting.use_link_rate_set = false; + initial_link_setting.link_rate_set = 0; + current_link_setting = initial_link_setting; + if (req_bw > dc_link_bandwidth_kbps(link, &link->verified_link_cap)) + return false; + + /* search for the minimum link setting that: + * 1. is supported according to the link training result + * 2. could support the b/w requested by the timing + */ + while (current_link_setting.link_rate <= + max_link_rate) { + link_bw = dc_link_bandwidth_kbps( + link, + ¤t_link_setting); + if (req_bw <= link_bw) { + *link_setting = current_link_setting; + return true; + } + if (policy) { + /* minimize lane */ + if (current_link_setting.link_rate < max_link_rate) { + current_link_setting.link_rate = + increase_link_rate( + current_link_setting.link_rate); + } else { + if (current_link_setting.lane_count < + link->verified_link_cap.lane_count) { + current_link_setting.lane_count = + increase_lane_count( + current_link_setting.lane_count); + current_link_setting.link_rate = initial_link_setting.link_rate; + } else + break; + } + } else { + /* minimize link rate */ + if (current_link_setting.lane_count < + link->verified_link_cap.lane_count) { + current_link_setting.lane_count = + increase_lane_count( + current_link_setting.lane_count); + } else { + current_link_setting.link_rate = + increase_link_rate( + current_link_setting.link_rate); + current_link_setting.lane_count = + initial_link_setting.lane_count; + } + } + } + return false; + } + + /* if optimize edp link is supported */ + memset(&initial_link_setting, 0, sizeof(initial_link_setting)); + initial_link_setting.lane_count = LANE_COUNT_ONE; + initial_link_setting.link_rate = link->dpcd_caps.edp_supported_link_rates[0]; + initial_link_setting.link_spread = LINK_SPREAD_DISABLED; + initial_link_setting.use_link_rate_set = true; + initial_link_setting.link_rate_set = 0; + current_link_setting = initial_link_setting; + + /* search for the minimum link setting that: + * 1. is supported according to the link training result + * 2. could support the b/w requested by the timing + */ + while (current_link_setting.link_rate <= + max_link_rate) { + link_bw = dc_link_bandwidth_kbps( + link, + ¤t_link_setting); + if (req_bw <= link_bw) { + *link_setting = current_link_setting; + return true; + } + if (policy) { + /* minimize lane */ + if (current_link_setting.link_rate_set < + link->dpcd_caps.edp_supported_link_rates_count + && current_link_setting.link_rate < max_link_rate) { + current_link_setting.link_rate_set++; + current_link_setting.link_rate = + link->dpcd_caps.edp_supported_link_rates[current_link_setting.link_rate_set]; + } else { + if (current_link_setting.lane_count < link->verified_link_cap.lane_count) { + current_link_setting.lane_count = + increase_lane_count( + current_link_setting.lane_count); + current_link_setting.link_rate_set = initial_link_setting.link_rate_set; + current_link_setting.link_rate = + link->dpcd_caps.edp_supported_link_rates[current_link_setting.link_rate_set]; + } else + break; + } + } else { + /* minimize link rate */ + if (current_link_setting.lane_count < + link->verified_link_cap.lane_count) { + current_link_setting.lane_count = + increase_lane_count( + current_link_setting.lane_count); + } else { + if (current_link_setting.link_rate_set < link->dpcd_caps.edp_supported_link_rates_count) { + current_link_setting.link_rate_set++; + current_link_setting.link_rate = + link->dpcd_caps.edp_supported_link_rates[current_link_setting.link_rate_set]; + current_link_setting.lane_count = + initial_link_setting.lane_count; + } else + break; + } + } + } + return false; +} + static bool decide_mst_link_settings(const struct dc_link *link, struct dc_link_settings *link_setting) { *link_setting = link->verified_link_cap; @@ -3380,7 +3522,25 @@ void decide_link_settings(struct dc_stream_state *stream, if (decide_mst_link_settings(link, link_setting)) return; } else if (link->connector_signal == SIGNAL_TYPE_EDP) { - if (decide_edp_link_settings(link, link_setting, req_bw)) + /* enable edp link optimization for DSC eDP case */ + if (stream->timing.flags.DSC) { + enum dc_link_rate max_link_rate = LINK_RATE_UNKNOWN; + + if (link->ctx->dc->debug.force_dsc_edp_policy) { + /* calculate link max link rate cap*/ + struct dc_link_settings tmp_link_setting; + struct dc_crtc_timing tmp_timing = stream->timing; + uint32_t orig_req_bw; + + tmp_link_setting.link_rate = LINK_RATE_UNKNOWN; + tmp_timing.flags.DSC = 0; + orig_req_bw = dc_bandwidth_in_kbps_from_timing(&tmp_timing); + decide_edp_link_settings(link, &tmp_link_setting, orig_req_bw); + max_link_rate = tmp_link_setting.link_rate; + } + if (decide_edp_link_settings_with_dsc(link, link_setting, req_bw, max_link_rate)) + return; + } else if (decide_edp_link_settings(link, link_setting, req_bw)) return; } else if (decide_dp_link_settings(link, link_setting, req_bw)) return; @@ -4454,7 +4614,7 @@ bool dp_retrieve_lttpr_cap(struct dc_link *link) lttpr_dpcd_data, sizeof(lttpr_dpcd_data)); if (status != DC_OK) { - dm_error("%s: Read LTTPR caps data failed.\n", __func__); + DC_LOG_DP2("%s: Read LTTPR caps data failed.\n", __func__); return false; } @@ -5885,7 +6045,10 @@ bool is_edp_ilr_optimization_required(struct dc_link *link, struct dc_crtc_timin req_bw = dc_bandwidth_in_kbps_from_timing(crtc_timing); - decide_edp_link_settings(link, &link_setting, req_bw); + if (!crtc_timing->flags.DSC) + decide_edp_link_settings(link, &link_setting, req_bw); + else + decide_edp_link_settings_with_dsc(link, &link_setting, req_bw, LINK_RATE_UNKNOWN); if (link->dpcd_caps.edp_supported_link_rates[link_rate_set] != link_setting.link_rate || lane_count_set.bits.LANE_COUNT_SET != link_setting.lane_count) { diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_dpia.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_dpia.c index b1c9f77d6bf4..d72122593959 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_dpia.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_dpia.c @@ -94,17 +94,17 @@ static enum link_training_result dpia_configure_link(struct dc_link *link, lt_settings); status = dpcd_configure_channel_coding(link, lt_settings); - if (status != DC_OK && !link->hpd_status) + if (status != DC_OK && link->is_hpd_pending) return LINK_TRAINING_ABORT; /* Configure lttpr mode */ status = dpcd_configure_lttpr_mode(link, lt_settings); - if (status != DC_OK && !link->hpd_status) + if (status != DC_OK && link->is_hpd_pending) return LINK_TRAINING_ABORT; /* Set link rate, lane count and spread. */ status = dpcd_set_link_settings(link, lt_settings); - if (status != DC_OK && !link->hpd_status) + if (status != DC_OK && link->is_hpd_pending) return LINK_TRAINING_ABORT; if (link->preferred_training_settings.fec_enable) @@ -112,7 +112,7 @@ static enum link_training_result dpia_configure_link(struct dc_link *link, else fec_enable = true; status = dp_set_fec_ready(link, fec_enable); - if (status != DC_OK && !link->hpd_status) + if (status != DC_OK && link->is_hpd_pending) return LINK_TRAINING_ABORT; return LINK_TRAINING_SUCCESS; @@ -388,7 +388,7 @@ static enum link_training_result dpia_training_cr_non_transparent(struct dc_link } /* Abort link training if clock recovery failed due to HPD unplug. */ - if (!link->hpd_status) + if (link->is_hpd_pending) result = LINK_TRAINING_ABORT; DC_LOG_HW_LINK_TRAINING("%s\n DPIA(%d) clock recovery\n" @@ -490,7 +490,7 @@ static enum link_training_result dpia_training_cr_transparent(struct dc_link *li } /* Abort link training if clock recovery failed due to HPD unplug. */ - if (!link->hpd_status) + if (link->is_hpd_pending) result = LINK_TRAINING_ABORT; DC_LOG_HW_LINK_TRAINING("%s\n DPIA(%d) clock recovery\n" @@ -675,7 +675,7 @@ static enum link_training_result dpia_training_eq_non_transparent(struct dc_link } /* Abort link training if equalization failed due to HPD unplug. */ - if (!link->hpd_status) + if (link->is_hpd_pending) result = LINK_TRAINING_ABORT; DC_LOG_HW_LINK_TRAINING("%s\n DPIA(%d) equalization\n" @@ -758,7 +758,7 @@ static enum link_training_result dpia_training_eq_transparent(struct dc_link *li } /* Abort link training if equalization failed due to HPD unplug. */ - if (!link->hpd_status) + if (link->is_hpd_pending) result = LINK_TRAINING_ABORT; DC_LOG_HW_LINK_TRAINING("%s\n DPIA(%d) equalization\n" @@ -892,10 +892,10 @@ static void dpia_training_abort(struct dc_link *link, uint32_t hop) __func__, link->link_id.enum_id - ENUM_ID_1, link->lttpr_mode, - link->hpd_status); + link->is_hpd_pending); /* Abandon clean-up if sink unplugged. */ - if (!link->hpd_status) + if (link->is_hpd_pending) return; if (hop != DPRX) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_enc_cfg.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_enc_cfg.c index 25e48a8cbb78..a55944da8d53 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_enc_cfg.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_enc_cfg.c @@ -118,7 +118,10 @@ static void remove_link_enc_assignment( */ if (get_stream_using_link_enc(state, eng_id) == NULL) state->res_ctx.link_enc_cfg_ctx.link_enc_avail[eng_idx] = eng_id; + stream->link_enc = NULL; + state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i].eng_id = ENGINE_ID_UNKNOWN; + state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i].stream = NULL; break; } } @@ -148,6 +151,7 @@ static void add_link_enc_assignment( .ep_type = stream->link->ep_type}, .eng_id = eng_id, .stream = stream}; + dc_stream_retain(stream); state->res_ctx.link_enc_cfg_ctx.link_enc_avail[eng_idx] = ENGINE_ID_UNKNOWN; stream->link_enc = stream->ctx->dc->res_pool->link_encoders[eng_idx]; break; @@ -227,7 +231,7 @@ static struct link_encoder *get_link_enc_used_by_link( .link_id = link->link_id, .ep_type = link->ep_type}; - for (i = 0; i < state->stream_count; i++) { + for (i = 0; i < MAX_PIPES; i++) { struct link_enc_assignment assignment = state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i]; if (assignment.valid == true && are_ep_ids_equal(&assignment.ep_id, &ep_id)) @@ -237,28 +241,18 @@ static struct link_encoder *get_link_enc_used_by_link( return link_enc; } /* Clear all link encoder assignments. */ -static void clear_enc_assignments(struct dc_state *state) +static void clear_enc_assignments(const struct dc *dc, struct dc_state *state) { int i; - enum engine_id eng_id; - struct dc_stream_state *stream; for (i = 0; i < MAX_PIPES; i++) { state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i].valid = false; - eng_id = state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i].eng_id; - stream = state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i].stream; - if (eng_id != ENGINE_ID_UNKNOWN) - state->res_ctx.link_enc_cfg_ctx.link_enc_avail[eng_id - ENGINE_ID_DIGA] = eng_id; - if (stream) - stream->link_enc = NULL; + state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i].eng_id = ENGINE_ID_UNKNOWN; + if (state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i].stream != NULL) { + dc_stream_release(state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i].stream); + state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i].stream = NULL; + } } -} - -void link_enc_cfg_init( - struct dc *dc, - struct dc_state *state) -{ - int i; for (i = 0; i < dc->res_pool->res_cap->num_dig_link_enc; i++) { if (dc->res_pool->link_encoders[i]) @@ -266,8 +260,13 @@ void link_enc_cfg_init( else state->res_ctx.link_enc_cfg_ctx.link_enc_avail[i] = ENGINE_ID_UNKNOWN; } +} - clear_enc_assignments(state); +void link_enc_cfg_init( + const struct dc *dc, + struct dc_state *state) +{ + clear_enc_assignments(dc, state); state->res_ctx.link_enc_cfg_ctx.mode = LINK_ENC_CFG_STEADY; } @@ -284,12 +283,9 @@ void link_enc_cfg_link_encs_assign( ASSERT(state->stream_count == stream_count); - if (stream_count == 0) - clear_enc_assignments(state); - /* Release DIG link encoder resources before running assignment algorithm. */ - for (i = 0; i < stream_count; i++) - dc->res_pool->funcs->link_enc_unassign(state, streams[i]); + for (i = 0; i < dc->current_state->stream_count; i++) + dc->res_pool->funcs->link_enc_unassign(state, dc->current_state->streams[i]); for (i = 0; i < MAX_PIPES; i++) ASSERT(state->res_ctx.link_enc_cfg_ctx.link_enc_assignments[i].valid == false); @@ -544,6 +540,7 @@ bool link_enc_cfg_validate(struct dc *dc, struct dc_state *state) uint8_t dig_stream_count = 0; int matching_stream_ptrs = 0; int eng_ids_per_ep_id[MAX_PIPES] = {0}; + int valid_bitmap = 0; /* (1) No. valid entries same as stream count. */ for (i = 0; i < MAX_PIPES; i++) { @@ -625,5 +622,15 @@ bool link_enc_cfg_validate(struct dc *dc, struct dc_state *state) is_valid = valid_entries && valid_stream_ptrs && valid_uniqueness && valid_avail && valid_streams; ASSERT(is_valid); + if (is_valid == false) { + valid_bitmap = + (valid_entries & 0x1) | + ((valid_stream_ptrs & 0x1) << 1) | + ((valid_uniqueness & 0x1) << 2) | + ((valid_avail & 0x1) << 3) | + ((valid_streams & 0x1) << 4); + dm_error("Invalid link encoder assignments: 0x%x\n", valid_bitmap); + } + return is_valid; } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c index c32fdccd4d92..1da91f250afa 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c @@ -2078,7 +2078,6 @@ static void mark_seamless_boot_stream( { struct dc_bios *dcb = dc->ctx->dc_bios; - /* TODO: Check Linux */ if (dc->config.allow_seamless_boot_optimization && !dcb->funcs->is_accelerated_mode(dcb)) { if (dc_validate_seamless_boot_timing(dc, stream->sink, &stream->timing)) @@ -2224,6 +2223,9 @@ void dc_resource_state_construct( struct dc_state *dst_ctx) { dst_ctx->clk_mgr = dc->clk_mgr; + + /* Initialise DIG link encoder resource tracking variables. */ + link_enc_cfg_init(dc, dst_ctx); } @@ -2506,17 +2508,7 @@ static void set_avi_info_frame( /* TODO : We should handle YCC quantization */ /* but we do not have matrix calculation */ - if (stream->qy_bit == 1) { - if (color_space == COLOR_SPACE_SRGB || - color_space == COLOR_SPACE_2020_RGB_FULLRANGE) - hdmi_info.bits.YQ0_YQ1 = YYC_QUANTIZATION_LIMITED_RANGE; - else if (color_space == COLOR_SPACE_SRGB_LIMITED || - color_space == COLOR_SPACE_2020_RGB_LIMITEDRANGE) - hdmi_info.bits.YQ0_YQ1 = YYC_QUANTIZATION_LIMITED_RANGE; - else - hdmi_info.bits.YQ0_YQ1 = YYC_QUANTIZATION_LIMITED_RANGE; - } else - hdmi_info.bits.YQ0_YQ1 = YYC_QUANTIZATION_LIMITED_RANGE; + hdmi_info.bits.YQ0_YQ1 = YYC_QUANTIZATION_LIMITED_RANGE; ///VIC format = stream->timing.timing_3d_format; diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_sink.c b/drivers/gpu/drm/amd/display/dc/core/dc_sink.c index a249a0e5edd0..4b5e4d8e7735 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_sink.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_sink.c @@ -33,14 +33,6 @@ * Private functions ******************************************************************************/ -static void dc_sink_destruct(struct dc_sink *sink) -{ - if (sink->dc_container_id) { - kfree(sink->dc_container_id); - sink->dc_container_id = NULL; - } -} - static bool dc_sink_construct(struct dc_sink *sink, const struct dc_sink_init_data *init_params) { @@ -75,7 +67,7 @@ void dc_sink_retain(struct dc_sink *sink) static void dc_sink_free(struct kref *kref) { struct dc_sink *sink = container_of(kref, struct dc_sink, refcount); - dc_sink_destruct(sink); + kfree(sink->dc_container_id); kfree(sink); } diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 3aac3f4a2852..c14e7db3f69d 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -47,7 +47,7 @@ struct aux_payload; struct set_config_cmd_payload; struct dmub_notification; -#define DC_VER "3.2.160" +#define DC_VER "3.2.163" #define MAX_SURFACES 3 #define MAX_PLANES 6 @@ -188,6 +188,7 @@ struct dc_caps { #if defined(CONFIG_DRM_AMD_DC_DCN) bool dp_hpo; #endif + bool edp_dsc_support; bool vbios_lttpr_aware; bool vbios_lttpr_enable; }; @@ -573,6 +574,8 @@ struct dc_debug_options { bool native422_support; bool disable_dsc; enum visual_confirm visual_confirm; + int visual_confirm_rect_height; + bool sanity_checks; bool max_disp_clk; bool surface_trace; @@ -667,6 +670,8 @@ struct dc_debug_options { bool validate_dml_output; bool enable_dmcub_surface_flip; bool usbc_combo_phy_reset_wa; + bool disable_dsc_edp; + unsigned int force_dsc_edp_policy; bool enable_dram_clock_change_one_display_vactive; #if defined(CONFIG_DRM_AMD_DC_DCN) /* TODO - remove once tested */ diff --git a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c index 360f3199ea6f..541376fabbef 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c +++ b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c @@ -115,13 +115,44 @@ void dc_dmub_srv_wait_idle(struct dc_dmub_srv *dc_dmub_srv) } } +void dc_dmub_srv_clear_inbox0_ack(struct dc_dmub_srv *dmub_srv) +{ + struct dmub_srv *dmub = dmub_srv->dmub; + struct dc_context *dc_ctx = dmub_srv->ctx; + enum dmub_status status = DMUB_STATUS_OK; + + status = dmub_srv_clear_inbox0_ack(dmub); + if (status != DMUB_STATUS_OK) { + DC_ERROR("Error clearing INBOX0 ack: status=%d\n", status); + dc_dmub_srv_log_diagnostic_data(dmub_srv); + } +} + +void dc_dmub_srv_wait_for_inbox0_ack(struct dc_dmub_srv *dmub_srv) +{ + struct dmub_srv *dmub = dmub_srv->dmub; + struct dc_context *dc_ctx = dmub_srv->ctx; + enum dmub_status status = DMUB_STATUS_OK; + + status = dmub_srv_wait_for_inbox0_ack(dmub, 100000); + if (status != DMUB_STATUS_OK) { + DC_ERROR("Error waiting for INBOX0 HW Lock Ack\n"); + dc_dmub_srv_log_diagnostic_data(dmub_srv); + } +} + void dc_dmub_srv_send_inbox0_cmd(struct dc_dmub_srv *dmub_srv, union dmub_inbox0_data_register data) { struct dmub_srv *dmub = dmub_srv->dmub; - if (dmub->hw_funcs.send_inbox0_cmd) - dmub->hw_funcs.send_inbox0_cmd(dmub, data); - // TODO: Add wait command -- poll register for ACK + struct dc_context *dc_ctx = dmub_srv->ctx; + enum dmub_status status = DMUB_STATUS_OK; + + status = dmub_srv_send_inbox0_cmd(dmub, data); + if (status != DMUB_STATUS_OK) { + DC_ERROR("Error sending INBOX0 cmd\n"); + dc_dmub_srv_log_diagnostic_data(dmub_srv); + } } bool dc_dmub_srv_cmd_with_reply_data(struct dc_dmub_srv *dc_dmub_srv, union dmub_rb_cmd *cmd) diff --git a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.h b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.h index 3e35eee7188c..7e4e2ec5915d 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.h +++ b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.h @@ -68,6 +68,8 @@ bool dc_dmub_srv_get_dmub_outbox0_msg(const struct dc *dc, struct dmcub_trace_bu void dc_dmub_trace_event_control(struct dc *dc, bool enable); +void dc_dmub_srv_clear_inbox0_ack(struct dc_dmub_srv *dmub_srv); +void dc_dmub_srv_wait_for_inbox0_ack(struct dc_dmub_srv *dmub_srv); void dc_dmub_srv_send_inbox0_cmd(struct dc_dmub_srv *dmub_srv, union dmub_inbox0_data_register data); bool dc_dmub_srv_get_diagnostic_data(struct dc_dmub_srv *dc_dmub_srv, struct dmub_diagnostic_data *dmub_oca); diff --git a/drivers/gpu/drm/amd/display/dc/dc_link.h b/drivers/gpu/drm/amd/display/dc/dc_link.h index 180ecd860296..d449e72a4e2a 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_link.h +++ b/drivers/gpu/drm/amd/display/dc/dc_link.h @@ -113,6 +113,7 @@ struct dc_link { * DIG encoder. */ bool is_dig_mapping_flexible; bool hpd_status; /* HPD status of link without physical HPD pin. */ + bool is_hpd_pending; /* Indicates a new received hpd */ bool edp_sink_present; @@ -287,6 +288,10 @@ bool dc_link_setup_psr(struct dc_link *dc_link, void dc_link_get_psr_residency(const struct dc_link *link, uint32_t *residency); +void dc_link_blank_all_dp_displays(struct dc *dc); + +void dc_link_blank_dp_stream(struct dc_link *link, bool hw_init); + /* Request DC to detect if there is a Panel connected. * boot - If this call is during initial boot. * Return false for any type of detection failure or MST detection @@ -298,7 +303,7 @@ enum dc_detect_reason { DETECT_REASON_HPD, DETECT_REASON_HPDRX, DETECT_REASON_FALLBACK, - DETECT_REASON_RETRAIN + DETECT_REASON_RETRAIN, }; bool dc_link_detect(struct dc_link *dc_link, enum dc_detect_reason reason); diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_audio.c b/drivers/gpu/drm/amd/display/dc/dce/dce_audio.c index 27218ede150a..70eaac017624 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_audio.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_audio.c @@ -67,9 +67,6 @@ static void write_indirect_azalia_reg(struct audio *audio, /* AZALIA_F0_CODEC_ENDPOINT_DATA endpoint data */ REG_SET(AZALIA_F0_CODEC_ENDPOINT_DATA, 0, AZALIA_ENDPOINT_REG_DATA, reg_data); - - DC_LOG_HW_AUDIO("AUDIO:write_indirect_azalia_reg: index: %u data: %u\n", - reg_index, reg_data); } static uint32_t read_indirect_azalia_reg(struct audio *audio, uint32_t reg_index) @@ -85,9 +82,6 @@ static uint32_t read_indirect_azalia_reg(struct audio *audio, uint32_t reg_index /* AZALIA_F0_CODEC_ENDPOINT_DATA endpoint data */ value = REG_READ(AZALIA_F0_CODEC_ENDPOINT_DATA); - DC_LOG_HW_AUDIO("AUDIO:read_indirect_azalia_reg: index: %u data: %u\n", - reg_index, value); - return value; } diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_audio.h b/drivers/gpu/drm/amd/display/dc/dce/dce_audio.h index 5622d5e32d81..dbd2cfed0603 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_audio.h +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_audio.h @@ -113,6 +113,7 @@ struct dce_audio_shift { uint8_t DCCG_AUDIO_DTO2_USE_512FBR_DTO; uint32_t DCCG_AUDIO_DTO0_USE_512FBR_DTO; uint32_t DCCG_AUDIO_DTO1_USE_512FBR_DTO; + uint32_t CLOCK_GATING_DISABLE; }; struct dce_audio_mask { @@ -132,6 +133,7 @@ struct dce_audio_mask { uint32_t DCCG_AUDIO_DTO2_USE_512FBR_DTO; uint32_t DCCG_AUDIO_DTO0_USE_512FBR_DTO; uint32_t DCCG_AUDIO_DTO1_USE_512FBR_DTO; + uint32_t CLOCK_GATING_DISABLE; }; diff --git a/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.c b/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.c index 9baf8ca0a920..b1b2e3c6f379 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.c @@ -56,8 +56,11 @@ void dmub_hw_lock_mgr_inbox0_cmd(struct dc_dmub_srv *dmub_srv, union dmub_inbox0_cmd_lock_hw hw_lock_cmd) { union dmub_inbox0_data_register data = { 0 }; + data.inbox0_cmd_lock_hw = hw_lock_cmd; + dc_dmub_srv_clear_inbox0_ack(dmub_srv); dc_dmub_srv_send_inbox0_cmd(dmub_srv, data); + dc_dmub_srv_wait_for_inbox0_ack(dmub_srv); } bool should_use_dmub_lock(struct dc_link *link) diff --git a/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.c b/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.c index 90eb8eedacf2..87ed48d5530d 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.c @@ -230,7 +230,7 @@ static void dmub_psr_set_level(struct dmub_psr *dmub, uint16_t psr_level, uint8_ /** * Set PSR power optimization flags. */ -static void dmub_psr_set_power_opt(struct dmub_psr *dmub, unsigned int power_opt) +static void dmub_psr_set_power_opt(struct dmub_psr *dmub, unsigned int power_opt, uint8_t panel_inst) { union dmub_rb_cmd cmd; struct dc_context *dc = dmub->ctx; @@ -239,7 +239,9 @@ static void dmub_psr_set_power_opt(struct dmub_psr *dmub, unsigned int power_opt cmd.psr_set_power_opt.header.type = DMUB_CMD__PSR; cmd.psr_set_power_opt.header.sub_type = DMUB_CMD__SET_PSR_POWER_OPT; cmd.psr_set_power_opt.header.payload_bytes = sizeof(struct dmub_cmd_psr_set_power_opt_data); + cmd.psr_set_power_opt.psr_set_power_opt_data.cmd_version = DMUB_CMD_PSR_CONTROL_VERSION_1; cmd.psr_set_power_opt.psr_set_power_opt_data.power_opt = power_opt; + cmd.psr_set_power_opt.psr_set_power_opt_data.panel_inst = panel_inst; dc_dmub_srv_cmd_queue(dc->dmub_srv, &cmd); dc_dmub_srv_cmd_execute(dc->dmub_srv); @@ -327,6 +329,16 @@ static bool dmub_psr_copy_settings(struct dmub_psr *dmub, copy_settings_data->fec_enable_delay_in100us = link->dc->debug.fec_enable_delay_in100us; copy_settings_data->cmd_version = DMUB_CMD_PSR_CONTROL_VERSION_1; copy_settings_data->panel_inst = panel_inst; + copy_settings_data->dsc_enable_status = (pipe_ctx->stream->timing.flags.DSC == 1); + + if (link->fec_state == dc_link_fec_enabled && + (!memcmp(link->dpcd_caps.sink_dev_id_str, DP_SINK_DEVICE_STR_ID_1, + sizeof(link->dpcd_caps.sink_dev_id_str)) || + !memcmp(link->dpcd_caps.sink_dev_id_str, DP_SINK_DEVICE_STR_ID_2, + sizeof(link->dpcd_caps.sink_dev_id_str)))) + copy_settings_data->debug.bitfields.force_wakeup_by_tps3 = 1; + else + copy_settings_data->debug.bitfields.force_wakeup_by_tps3 = 0; dc_dmub_srv_cmd_queue(dc->dmub_srv, &cmd); dc_dmub_srv_cmd_execute(dc->dmub_srv); diff --git a/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.h b/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.h index 5dbd479660f1..01acc01cc191 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.h +++ b/drivers/gpu/drm/amd/display/dc/dce/dmub_psr.h @@ -46,7 +46,7 @@ struct dmub_psr_funcs { void (*psr_force_static)(struct dmub_psr *dmub, uint8_t panel_inst); void (*psr_get_residency)(struct dmub_psr *dmub, uint32_t *residency, uint8_t panel_inst); - void (*psr_set_power_opt)(struct dmub_psr *dmub, unsigned int power_opt); + void (*psr_set_power_opt)(struct dmub_psr *dmub, unsigned int power_opt, uint8_t panel_inst); }; struct dmub_psr *dmub_psr_create(struct dc_context *ctx); diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c index 24e47df526f6..3d421583e9ca 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c @@ -1602,6 +1602,11 @@ static enum dc_status apply_single_controller_ctx_to_hw( pipe_ctx->stream_res.stream_enc, pipe_ctx->stream_res.tg->inst); + if (dc_is_dp_signal(pipe_ctx->stream->signal) && + pipe_ctx->stream_res.stream_enc->funcs->reset_fifo) + pipe_ctx->stream_res.stream_enc->funcs->reset_fifo( + pipe_ctx->stream_res.stream_enc); + if (dc_is_dp_signal(pipe_ctx->stream->signal)) dp_source_sequence_trace(link, DPCD_SOURCE_SEQ_AFTER_CONNECT_DIG_FE_OTG); @@ -1655,30 +1660,12 @@ static enum dc_status apply_single_controller_ctx_to_hw( static void power_down_encoders(struct dc *dc) { - int i, j; + int i; for (i = 0; i < dc->link_count; i++) { enum signal_type signal = dc->links[i]->connector_signal; - if ((signal == SIGNAL_TYPE_EDP) || - (signal == SIGNAL_TYPE_DISPLAY_PORT)) { - if (dc->links[i]->link_enc->funcs->get_dig_frontend && - dc->links[i]->link_enc->funcs->is_dig_enabled(dc->links[i]->link_enc)) { - unsigned int fe = dc->links[i]->link_enc->funcs->get_dig_frontend( - dc->links[i]->link_enc); - - for (j = 0; j < dc->res_pool->stream_enc_count; j++) { - if (fe == dc->res_pool->stream_enc[j]->id) { - dc->res_pool->stream_enc[j]->funcs->dp_blank(dc->links[i], - dc->res_pool->stream_enc[j]); - break; - } - } - } - - if (!dc->links[i]->wa_flags.dp_keep_receiver_powered) - dp_receiver_power_ctrl(dc->links[i], false); - } + dc_link_blank_dp_stream(dc->links[i], false); if (signal != SIGNAL_TYPE_EDP) signal = SIGNAL_TYPE_NONE; @@ -1846,7 +1833,7 @@ void dce110_enable_accelerated_mode(struct dc *dc, struct dc_state *context) } } // We are trying to enable eDP, don't power down VDD - if (edp_stream_num) + if (edp_stream_num && can_apply_edp_fast_boot) keep_edp_vdd_on = true; } diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c index 44293d66b46b..e31a6f1516bb 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c @@ -39,6 +39,10 @@ #define BLACK_OFFSET_RGB_Y 0x0 #define BLACK_OFFSET_CBCR 0x8000 +#define VISUAL_CONFIRM_RECT_HEIGHT_DEFAULT 3 +#define VISUAL_CONFIRM_RECT_HEIGHT_MIN 1 +#define VISUAL_CONFIRM_RECT_HEIGHT_MAX 10 + #define REG(reg)\ dpp->tf_regs->reg @@ -685,9 +689,17 @@ static void dpp1_dscl_set_recout(struct dcn10_dpp *dpp, const struct rect *recout) { int visual_confirm_on = 0; + unsigned short visual_confirm_rect_height = VISUAL_CONFIRM_RECT_HEIGHT_DEFAULT; + if (dpp->base.ctx->dc->debug.visual_confirm != VISUAL_CONFIRM_DISABLE) visual_confirm_on = 1; + /* Check bounds to ensure the VC bar height was set to a sane value */ + if ((dpp->base.ctx->dc->debug.visual_confirm_rect_height >= VISUAL_CONFIRM_RECT_HEIGHT_MIN) && + (dpp->base.ctx->dc->debug.visual_confirm_rect_height <= VISUAL_CONFIRM_RECT_HEIGHT_MAX)) { + visual_confirm_rect_height = dpp->base.ctx->dc->debug.visual_confirm_rect_height; + } + REG_SET_2(RECOUT_START, 0, /* First pixel of RECOUT in the active OTG area */ RECOUT_START_X, recout->x, @@ -699,7 +711,7 @@ static void dpp1_dscl_set_recout(struct dcn10_dpp *dpp, RECOUT_WIDTH, recout->width, /* Number of RECOUT vertical lines */ RECOUT_HEIGHT, recout->height - - visual_confirm_on * 2 * (dpp->base.inst + 1)); + - visual_confirm_on * 2 * (dpp->base.inst + visual_confirm_rect_height)); } /** diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c index 0b788d794fb3..a2b925cc4132 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c @@ -1362,11 +1362,48 @@ void dcn10_init_pipes(struct dc *dc, struct dc_state *context) tg->funcs->tg_init(tg); } + + /* Power gate DSCs */ + if (hws->funcs.dsc_pg_control != NULL) { + uint32_t num_opps = 0; + uint32_t opp_id_src0 = OPP_ID_INVALID; + uint32_t opp_id_src1 = OPP_ID_INVALID; + + // Step 1: To find out which OPTC is running & OPTC DSC is ON + for (i = 0; i < dc->res_pool->res_cap->num_timing_generator; i++) { + uint32_t optc_dsc_state = 0; + struct timing_generator *tg = dc->res_pool->timing_generators[i]; + + if (tg->funcs->is_tg_enabled(tg)) { + if (tg->funcs->get_dsc_status) + tg->funcs->get_dsc_status(tg, &optc_dsc_state); + // Only one OPTC with DSC is ON, so if we got one result, we would exit this block. + // non-zero value is DSC enabled + if (optc_dsc_state != 0) { + tg->funcs->get_optc_source(tg, &num_opps, &opp_id_src0, &opp_id_src1); + break; + } + } + } + + // Step 2: To power down DSC but skip DSC of running OPTC + for (i = 0; i < dc->res_pool->res_cap->num_dsc; i++) { + struct dcn_dsc_state s = {0}; + + dc->res_pool->dscs[i]->funcs->dsc_read_state(dc->res_pool->dscs[i], &s); + + if ((s.dsc_opp_source == opp_id_src0 || s.dsc_opp_source == opp_id_src1) && + s.dsc_clock_en && s.dsc_fw_en) + continue; + + hws->funcs.dsc_pg_control(hws, dc->res_pool->dscs[i]->inst, false); + } + } } void dcn10_init_hw(struct dc *dc) { - int i, j; + int i; struct abm *abm = dc->res_pool->abm; struct dmcu *dmcu = dc->res_pool->dmcu; struct dce_hwseq *hws = dc->hwseq; @@ -1468,43 +1505,8 @@ void dcn10_init_hw(struct dc *dc) dmub_enable_outbox_notification(dc); /* we want to turn off all dp displays before doing detection */ - if (dc->config.power_down_display_on_boot) { - uint8_t dpcd_power_state = '\0'; - enum dc_status status = DC_ERROR_UNEXPECTED; - - for (i = 0; i < dc->link_count; i++) { - if (dc->links[i]->connector_signal != SIGNAL_TYPE_DISPLAY_PORT) - continue; - - /* DP 2.0 requires that LTTPR Caps be read first */ - dp_retrieve_lttpr_cap(dc->links[i]); - - /* - * If any of the displays are lit up turn them off. - * The reason is that some MST hubs cannot be turned off - * completely until we tell them to do so. - * If not turned off, then displays connected to MST hub - * won't light up. - */ - status = core_link_read_dpcd(dc->links[i], DP_SET_POWER, - &dpcd_power_state, sizeof(dpcd_power_state)); - if (status == DC_OK && dpcd_power_state == DP_POWER_STATE_D0) { - /* blank dp stream before power off receiver*/ - if (dc->links[i]->link_enc->funcs->get_dig_frontend) { - unsigned int fe = dc->links[i]->link_enc->funcs->get_dig_frontend(dc->links[i]->link_enc); - - for (j = 0; j < dc->res_pool->stream_enc_count; j++) { - if (fe == dc->res_pool->stream_enc[j]->id) { - dc->res_pool->stream_enc[j]->funcs->dp_blank(dc->links[i], - dc->res_pool->stream_enc[j]); - break; - } - } - } - dp_receiver_power_ctrl(dc->links[i], false); - } - } - } + if (dc->config.power_down_display_on_boot) + dc_link_blank_all_dp_displays(dc); /* If taking control over from VBIOS, we may want to optimize our first * mode set, so we need to skip powering down pipes until we know which @@ -1637,7 +1639,7 @@ void dcn10_reset_hw_ctx_wrap( dcn10_reset_back_end_for_pipe(dc, pipe_ctx_old, dc->current_state); if (hws->funcs.enable_stream_gating) - hws->funcs.enable_stream_gating(dc, pipe_ctx); + hws->funcs.enable_stream_gating(dc, pipe_ctx_old); if (old_clk) old_clk->funcs->cs_power_down(old_clk); } @@ -2624,7 +2626,7 @@ static void dcn10_update_dchubp_dpp( /* new calculated dispclk, dppclk are stored in * context->bw_ctx.bw.dcn.clk.dispclk_khz / dppclk_khz. current * dispclk, dppclk are from dc->clk_mgr->clks.dispclk_khz. - * dcn_validate_bandwidth compute new dispclk, dppclk. + * dcn10_validate_bandwidth compute new dispclk, dppclk. * dispclk will put in use after optimize_bandwidth when * ramp_up_dispclk_with_dpp is called. * there are two places for dppclk be put in use. One location @@ -2638,7 +2640,7 @@ static void dcn10_update_dchubp_dpp( * for example, eDP + external dp, change resolution of DP from * 1920x1080x144hz to 1280x960x60hz. * before change: dispclk = 337889 dppclk = 337889 - * change mode, dcn_validate_bandwidth calculate + * change mode, dcn10_validate_bandwidth calculate * dispclk = 143122 dppclk = 143122 * update_dchubp_dpp be executed before dispclk be updated, * dispclk = 337889, but dppclk use new value dispclk /2 = diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c index f37551e00023..19a2dd619ec7 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c @@ -978,10 +978,8 @@ static void dcn10_resource_destruct(struct dcn10_resource_pool *pool) pool->base.mpc = NULL; } - if (pool->base.hubbub != NULL) { - kfree(pool->base.hubbub); - pool->base.hubbub = NULL; - } + kfree(pool->base.hubbub); + pool->base.hubbub = NULL; for (i = 0; i < pool->base.pipe_count; i++) { if (pool->base.opps[i] != NULL) @@ -1011,14 +1009,10 @@ static void dcn10_resource_destruct(struct dcn10_resource_pool *pool) for (i = 0; i < pool->base.res_cap->num_ddc; i++) { if (pool->base.engines[i] != NULL) dce110_engine_destroy(&pool->base.engines[i]); - if (pool->base.hw_i2cs[i] != NULL) { - kfree(pool->base.hw_i2cs[i]); - pool->base.hw_i2cs[i] = NULL; - } - if (pool->base.sw_i2cs[i] != NULL) { - kfree(pool->base.sw_i2cs[i]); - pool->base.sw_i2cs[i] = NULL; - } + kfree(pool->base.hw_i2cs[i]); + pool->base.hw_i2cs[i] = NULL; + kfree(pool->base.sw_i2cs[i]); + pool->base.sw_i2cs[i] = NULL; } for (i = 0; i < pool->base.audio_count; i++) { @@ -1320,7 +1314,7 @@ static const struct resource_funcs dcn10_res_pool_funcs = { .destroy = dcn10_destroy_resource_pool, .link_enc_create = dcn10_link_encoder_create, .panel_cntl_create = dcn10_panel_cntl_create, - .validate_bandwidth = dcn_validate_bandwidth, + .validate_bandwidth = dcn10_validate_bandwidth, .acquire_idle_pipe_for_layer = dcn10_acquire_idle_pipe_for_layer, .validate_plane = dcn10_validate_plane, .validate_global = dcn10_validate_global, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.c index b0c08ee6bc2c..bf4436d7aaab 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.c @@ -902,6 +902,19 @@ void enc1_stream_encoder_stop_dp_info_packets( } +void enc1_stream_encoder_reset_fifo( + struct stream_encoder *enc) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + /* set DIG_START to 0x1 to reset FIFO */ + REG_UPDATE(DIG_FE_CNTL, DIG_START, 1); + udelay(100); + + /* write 0 to take the FIFO out of reset */ + REG_UPDATE(DIG_FE_CNTL, DIG_START, 0); +} + void enc1_stream_encoder_dp_blank( struct dc_link *link, struct stream_encoder *enc) @@ -1587,6 +1600,8 @@ static const struct stream_encoder_funcs dcn10_str_enc_funcs = { enc1_stream_encoder_send_immediate_sdp_message, .stop_dp_info_packets = enc1_stream_encoder_stop_dp_info_packets, + .reset_fifo = + enc1_stream_encoder_reset_fifo, .dp_blank = enc1_stream_encoder_dp_blank, .dp_unblank = diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.h index 687d7e4bf7ca..a146a41f68e9 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.h @@ -626,6 +626,9 @@ void enc1_stream_encoder_send_immediate_sdp_message( void enc1_stream_encoder_stop_dp_info_packets( struct stream_encoder *enc); +void enc1_stream_encoder_reset_fifo( + struct stream_encoder *enc); + void enc1_stream_encoder_dp_blank( struct dc_link *link, struct stream_encoder *enc); diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_dsc.c b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_dsc.c index 79b640e202eb..ef5c4c0f4d6c 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_dsc.c +++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_dsc.c @@ -162,6 +162,8 @@ static void dsc2_read_state(struct display_stream_compressor *dsc, struct dcn_ds REG_GET(DSCC_PPS_CONFIG2, PIC_WIDTH, &s->dsc_pic_width); REG_GET(DSCC_PPS_CONFIG2, PIC_HEIGHT, &s->dsc_pic_height); REG_GET(DSCC_PPS_CONFIG7, SLICE_BPG_OFFSET, &s->dsc_slice_bpg_offset); + REG_GET_2(DSCRM_DSC_FORWARD_CONFIG, DSCRM_DSC_FORWARD_EN, &s->dsc_fw_en, + DSCRM_DSC_OPP_PIPE_SOURCE, &s->dsc_opp_source); } diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c index 4f88376a118f..e6af99ae3d9f 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c @@ -2270,7 +2270,7 @@ void dcn20_reset_hw_ctx_wrap( dcn20_reset_back_end_for_pipe(dc, pipe_ctx_old, dc->current_state); if (hws->funcs.enable_stream_gating) - hws->funcs.enable_stream_gating(dc, pipe_ctx); + hws->funcs.enable_stream_gating(dc, pipe_ctx_old); if (old_clk) old_clk->funcs->cs_power_down(old_clk); } diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_optc.c b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_optc.c index c90b8516dcc1..8c34751b0e58 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_optc.c +++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_optc.c @@ -190,6 +190,19 @@ void optc2_set_dsc_config(struct timing_generator *optc, OPTC_DSC_SLICE_WIDTH, dsc_slice_width); } +/* Get DSC-related configuration. + * dsc_mode: 0 disables DSC, other values enable DSC in specified format + */ +void optc2_get_dsc_status(struct timing_generator *optc, + uint32_t *dsc_mode) +{ + struct optc *optc1 = DCN10TG_FROM_TG(optc); + + REG_GET(OPTC_DATA_FORMAT_CONTROL, + OPTC_DSC_MODE, dsc_mode); +} + + /*TEMP: Need to figure out inheritance model here.*/ bool optc2_is_two_pixels_per_containter(const struct dc_crtc_timing *timing) { @@ -579,6 +592,7 @@ static struct timing_generator_funcs dcn20_tg_funcs = { .get_crc = optc1_get_crc, .configure_crc = optc2_configure_crc, .set_dsc_config = optc2_set_dsc_config, + .get_dsc_status = optc2_get_dsc_status, .set_dwb_source = optc2_set_dwb_source, .set_odm_bypass = optc2_set_odm_bypass, .set_odm_combine = optc2_set_odm_combine, diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_optc.h b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_optc.h index be19a6885fbf..f7968b9ca16e 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_optc.h +++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_optc.h @@ -98,6 +98,9 @@ void optc2_set_dsc_config(struct timing_generator *optc, uint32_t dsc_bytes_per_pixel, uint32_t dsc_slice_width); +void optc2_get_dsc_status(struct timing_generator *optc, + uint32_t *dsc_mode); + void optc2_set_odm_bypass(struct timing_generator *optc, const struct dc_crtc_timing *dc_crtc_timing); diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_stream_encoder.c b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_stream_encoder.c index aab25ca8343a..8a70f92795c2 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_stream_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_stream_encoder.c @@ -593,6 +593,8 @@ static const struct stream_encoder_funcs dcn20_str_enc_funcs = { enc1_stream_encoder_send_immediate_sdp_message, .stop_dp_info_packets = enc1_stream_encoder_stop_dp_info_packets, + .reset_fifo = + enc1_stream_encoder_reset_fifo, .dp_blank = enc1_stream_encoder_dp_blank, .dp_unblank = diff --git a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_dio_stream_encoder.c b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_dio_stream_encoder.c index ebd9c35c914f..7aa9aaf5db4c 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_dio_stream_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_dio_stream_encoder.c @@ -805,6 +805,8 @@ static const struct stream_encoder_funcs dcn30_str_enc_funcs = { enc3_stream_encoder_update_dp_info_packets, .stop_dp_info_packets = enc1_stream_encoder_stop_dp_info_packets, + .reset_fifo = + enc1_stream_encoder_reset_fifo, .dp_blank = enc1_stream_encoder_dp_blank, .dp_unblank = diff --git a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_hwseq.c b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_hwseq.c index df2717116604..3e99bb9c70ab 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_hwseq.c @@ -437,7 +437,7 @@ void dcn30_init_hw(struct dc *dc) struct dce_hwseq *hws = dc->hwseq; struct dc_bios *dcb = dc->ctx->dc_bios; struct resource_pool *res_pool = dc->res_pool; - int i, j; + int i; int edp_num; uint32_t backlight = MAX_BACKLIGHT_LEVEL; @@ -534,41 +534,8 @@ void dcn30_init_hw(struct dc *dc) hws->funcs.dsc_pg_control(hws, res_pool->dscs[i]->inst, false); /* we want to turn off all dp displays before doing detection */ - if (dc->config.power_down_display_on_boot) { - uint8_t dpcd_power_state = '\0'; - enum dc_status status = DC_ERROR_UNEXPECTED; - - for (i = 0; i < dc->link_count; i++) { - if (dc->links[i]->connector_signal != SIGNAL_TYPE_DISPLAY_PORT) - continue; - /* DP 2.0 states that LTTPR regs must be read first */ - dp_retrieve_lttpr_cap(dc->links[i]); - - /* if any of the displays are lit up turn them off */ - status = core_link_read_dpcd(dc->links[i], DP_SET_POWER, - &dpcd_power_state, sizeof(dpcd_power_state)); - if (status == DC_OK && dpcd_power_state == DP_POWER_STATE_D0) { - /* blank dp stream before power off receiver*/ - if (dc->links[i]->link_enc->funcs->get_dig_frontend) { - unsigned int fe; - - fe = dc->links[i]->link_enc->funcs->get_dig_frontend( - dc->links[i]->link_enc); - if (fe == ENGINE_ID_UNKNOWN) - continue; - - for (j = 0; j < dc->res_pool->stream_enc_count; j++) { - if (fe == dc->res_pool->stream_enc[j]->id) { - dc->res_pool->stream_enc[j]->funcs->dp_blank(dc->links[i], - dc->res_pool->stream_enc[j]); - break; - } - } - } - dp_receiver_power_ctrl(dc->links[i], false); - } - } - } + if (dc->config.power_down_display_on_boot) + dc_link_blank_all_dp_displays(dc); /* If taking control over from VBIOS, we may want to optimize our first * mode set, so we need to skip powering down pipes until we know which diff --git a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_optc.c b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_optc.c index 5d9e6413d67a..f5e8916601d3 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_optc.c +++ b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_optc.c @@ -332,6 +332,7 @@ static struct timing_generator_funcs dcn30_tg_funcs = { .get_crc = optc1_get_crc, .configure_crc = optc2_configure_crc, .set_dsc_config = optc3_set_dsc_config, + .get_dsc_status = optc2_get_dsc_status, .set_dwb_source = NULL, .set_odm_bypass = optc3_set_odm_bypass, .set_odm_combine = optc3_set_odm_combine, diff --git a/drivers/gpu/drm/amd/display/dc/dcn301/dcn301_resource.c b/drivers/gpu/drm/amd/display/dc/dcn301/dcn301_resource.c index fbaa03f26d8b..7abc36a4ff76 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn301/dcn301_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn301/dcn301_resource.c @@ -686,7 +686,7 @@ static const struct dc_debug_options debug_defaults_drv = { .disable_clock_gate = true, .disable_pplib_clock_request = true, .disable_pplib_wm_range = true, - .pipe_split_policy = MPC_SPLIT_AVOID_MULT_DISP, + .pipe_split_policy = MPC_SPLIT_AVOID, .force_single_disp_pipe_split = false, .disable_dcc = DCC_ENABLE, .vsr_support = true, @@ -1449,9 +1449,7 @@ static bool dcn301_resource_construct( dc->caps.post_blend_color_processing = true; dc->caps.force_dp_tps4_for_cp2520 = true; dc->caps.extended_aux_timeout_support = true; -#ifdef CONFIG_DRM_AMD_DC_DMUB dc->caps.dmcub_support = true; -#endif /* Color pipeline capabilities */ dc->caps.color.dpp.dcn_arch = 1; @@ -1487,6 +1485,23 @@ static bool dcn301_resource_construct( dc->caps.color.mpc.ogam_rom_caps.hlg = 0; dc->caps.color.mpc.ocsc = 1; + /* read VBIOS LTTPR caps */ + if (ctx->dc_bios->funcs->get_lttpr_caps) { + enum bp_result bp_query_result; + uint8_t is_vbios_lttpr_enable = 0; + + bp_query_result = ctx->dc_bios->funcs->get_lttpr_caps(ctx->dc_bios, &is_vbios_lttpr_enable); + dc->caps.vbios_lttpr_enable = (bp_query_result == BP_RESULT_OK) && !!is_vbios_lttpr_enable; + } + + if (ctx->dc_bios->funcs->get_lttpr_interop) { + enum bp_result bp_query_result; + uint8_t is_vbios_interop_enabled = 0; + + bp_query_result = ctx->dc_bios->funcs->get_lttpr_interop(ctx->dc_bios, &is_vbios_interop_enabled); + dc->caps.vbios_lttpr_aware = (bp_query_result == BP_RESULT_OK) && !!is_vbios_interop_enabled; + } + if (dc->ctx->dce_environment == DCE_ENV_PRODUCTION_DRV) dc->debug = debug_defaults_drv; else if (dc->ctx->dce_environment == DCE_ENV_FPGA_MAXIMUS) { diff --git a/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.c b/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.c index fcf96cf08c76..058f5d71e037 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.c @@ -1557,6 +1557,24 @@ static bool dcn302_resource_construct( dc->caps.color.mpc.ogam_rom_caps.hlg = 0; dc->caps.color.mpc.ocsc = 1; + /* read VBIOS LTTPR caps */ + if (ctx->dc_bios->funcs->get_lttpr_caps) { + enum bp_result bp_query_result; + uint8_t is_vbios_lttpr_enable = 0; + + bp_query_result = ctx->dc_bios->funcs->get_lttpr_caps(ctx->dc_bios, &is_vbios_lttpr_enable); + dc->caps.vbios_lttpr_enable = (bp_query_result == BP_RESULT_OK) && !!is_vbios_lttpr_enable; + } + + if (ctx->dc_bios->funcs->get_lttpr_interop) { + enum bp_result bp_query_result; + uint8_t is_vbios_interop_enabled = 0; + + bp_query_result = ctx->dc_bios->funcs->get_lttpr_interop(ctx->dc_bios, + &is_vbios_interop_enabled); + dc->caps.vbios_lttpr_aware = (bp_query_result == BP_RESULT_OK) && !!is_vbios_interop_enabled; + } + if (dc->ctx->dce_environment == DCE_ENV_PRODUCTION_DRV) dc->debug = debug_defaults_drv; else diff --git a/drivers/gpu/drm/amd/display/dc/dcn303/dcn303_resource.c b/drivers/gpu/drm/amd/display/dc/dcn303/dcn303_resource.c index 4a9b64023675..7024aeb0884c 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn303/dcn303_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn303/dcn303_resource.c @@ -1500,6 +1500,23 @@ static bool dcn303_resource_construct( dc->caps.color.mpc.ogam_rom_caps.hlg = 0; dc->caps.color.mpc.ocsc = 1; + /* read VBIOS LTTPR caps */ + if (ctx->dc_bios->funcs->get_lttpr_caps) { + enum bp_result bp_query_result; + uint8_t is_vbios_lttpr_enable = 0; + + bp_query_result = ctx->dc_bios->funcs->get_lttpr_caps(ctx->dc_bios, &is_vbios_lttpr_enable); + dc->caps.vbios_lttpr_enable = (bp_query_result == BP_RESULT_OK) && !!is_vbios_lttpr_enable; + } + + if (ctx->dc_bios->funcs->get_lttpr_interop) { + enum bp_result bp_query_result; + uint8_t is_vbios_interop_enabled = 0; + + bp_query_result = ctx->dc_bios->funcs->get_lttpr_interop(ctx->dc_bios, &is_vbios_interop_enabled); + dc->caps.vbios_lttpr_aware = (bp_query_result == BP_RESULT_OK) && !!is_vbios_interop_enabled; + } + if (dc->ctx->dce_environment == DCE_ENV_PRODUCTION_DRV) dc->debug = debug_defaults_drv; else diff --git a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_hpo_dp_stream_encoder.c b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_hpo_dp_stream_encoder.c index 565f12dd179a..5065904c7833 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_hpo_dp_stream_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_hpo_dp_stream_encoder.c @@ -358,8 +358,8 @@ static void dcn31_hpo_dp_stream_enc_set_stream_attribute( h_width = hw_crtc_timing.h_border_left + hw_crtc_timing.h_addressable + hw_crtc_timing.h_border_right; v_height = hw_crtc_timing.v_border_top + hw_crtc_timing.v_addressable + hw_crtc_timing.v_border_bottom; - hsp = hw_crtc_timing.flags.HSYNC_POSITIVE_POLARITY ? 0x80 : 0; - vsp = hw_crtc_timing.flags.VSYNC_POSITIVE_POLARITY ? 0x80 : 0; + hsp = hw_crtc_timing.flags.HSYNC_POSITIVE_POLARITY ? 0 : 0x80; + vsp = hw_crtc_timing.flags.VSYNC_POSITIVE_POLARITY ? 0 : 0x80; v_freq = hw_crtc_timing.pix_clk_100hz * 100; /* MSA Packet Mapping to 32-bit Link Symbols - DP2 spec, section 2.7.4.1 diff --git a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_hwseq.c b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_hwseq.c index 5dd1ce9ddb53..4206ce5bf9a9 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_hwseq.c @@ -112,7 +112,7 @@ void dcn31_init_hw(struct dc *dc) struct dc_bios *dcb = dc->ctx->dc_bios; struct resource_pool *res_pool = dc->res_pool; uint32_t backlight = MAX_BACKLIGHT_LEVEL; - int i, j; + int i; if (dc->clk_mgr && dc->clk_mgr->funcs->init_clocks) dc->clk_mgr->funcs->init_clocks(dc->clk_mgr); @@ -192,50 +192,13 @@ void dcn31_init_hw(struct dc *dc) link->link_status.link_active = true; } - /* Power gate DSCs */ - for (i = 0; i < res_pool->res_cap->num_dsc; i++) - if (hws->funcs.dsc_pg_control != NULL) - hws->funcs.dsc_pg_control(hws, res_pool->dscs[i]->inst, false); - /* Enables outbox notifications for usb4 dpia */ if (dc->res_pool->usb4_dpia_count) dmub_enable_outbox_notification(dc); /* we want to turn off all dp displays before doing detection */ - if (dc->config.power_down_display_on_boot) { - uint8_t dpcd_power_state = '\0'; - enum dc_status status = DC_ERROR_UNEXPECTED; - - for (i = 0; i < dc->link_count; i++) { - if (dc->links[i]->connector_signal != SIGNAL_TYPE_DISPLAY_PORT) - continue; - - /* if any of the displays are lit up turn them off */ - status = core_link_read_dpcd(dc->links[i], DP_SET_POWER, - &dpcd_power_state, sizeof(dpcd_power_state)); - if (status == DC_OK && dpcd_power_state == DP_POWER_STATE_D0) { - /* blank dp stream before power off receiver*/ - if (dc->links[i]->ep_type == DISPLAY_ENDPOINT_PHY && - dc->links[i]->link_enc->funcs->get_dig_frontend) { - unsigned int fe; - - fe = dc->links[i]->link_enc->funcs->get_dig_frontend( - dc->links[i]->link_enc); - if (fe == ENGINE_ID_UNKNOWN) - continue; - - for (j = 0; j < dc->res_pool->stream_enc_count; j++) { - if (fe == dc->res_pool->stream_enc[j]->id) { - dc->res_pool->stream_enc[j]->funcs->dp_blank(dc->links[i], - dc->res_pool->stream_enc[j]); - break; - } - } - } - dp_receiver_power_ctrl(dc->links[i], false); - } - } - } + if (dc->config.power_down_display_on_boot) + dc_link_blank_all_dp_displays(dc); /* If taking control over from VBIOS, we may want to optimize our first * mode set, so we need to skip powering down pipes until we know which @@ -602,7 +565,7 @@ void dcn31_reset_hw_ctx_wrap( dcn31_reset_back_end_for_pipe(dc, pipe_ctx_old, dc->current_state); if (hws->funcs.enable_stream_gating) - hws->funcs.enable_stream_gating(dc, pipe_ctx); + hws->funcs.enable_stream_gating(dc, pipe_ctx_old); if (old_clk) old_clk->funcs->cs_power_down(old_clk); } diff --git a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_init.c b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_init.c index 05335a8c3c2d..e175b6cc0125 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_init.c +++ b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_init.c @@ -149,4 +149,9 @@ void dcn31_hw_sequencer_construct(struct dc *dc) dc->hwss.init_hw = dcn20_fpga_init_hw; dc->hwseq->funcs.init_pipes = NULL; } + if (dc->debug.disable_z10) { + /*hw not support z10 or sw disable it*/ + dc->hwss.z10_restore = NULL; + dc->hwss.z10_save_init = NULL; + } } diff --git a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_optc.c b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_optc.c index a4b1d98f0007..e8562fa11366 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_optc.c +++ b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_optc.c @@ -256,6 +256,7 @@ static struct timing_generator_funcs dcn31_tg_funcs = { .get_crc = optc1_get_crc, .configure_crc = optc2_configure_crc, .set_dsc_config = optc3_set_dsc_config, + .get_dsc_status = optc2_get_dsc_status, .set_dwb_source = NULL, .set_odm_bypass = optc3_set_odm_bypass, .set_odm_combine = optc31_set_odm_combine, diff --git a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_resource.c b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_resource.c index 18896294ae12..88e040687940 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_resource.c @@ -2199,6 +2199,7 @@ static bool dcn31_resource_construct( dc->caps.post_blend_color_processing = true; dc->caps.force_dp_tps4_for_cp2520 = true; dc->caps.dp_hpo = true; + dc->caps.edp_dsc_support = true; dc->caps.extended_aux_timeout_support = true; dc->caps.dmcub_support = true; dc->caps.is_apu = true; diff --git a/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.h b/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.h index 6905ef1e75a6..d76251fd1566 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.h +++ b/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.h @@ -73,6 +73,7 @@ struct display_mode_lib { struct vba_vars_st vba; struct dal_logger *logger; struct dml_funcs funcs; + struct _vcs_dpi_display_e2e_pipe_params_st dml_pipe_state[6]; }; void dml_init_instance(struct display_mode_lib *lib, diff --git a/drivers/gpu/drm/amd/display/dc/dml/dml_wrapper.c b/drivers/gpu/drm/amd/display/dc/dml/dml_wrapper.c new file mode 100644 index 000000000000..ece34b0b8a46 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dml/dml_wrapper.c @@ -0,0 +1,1889 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: AMD + * + */ + +#include "dml_wrapper.h" +#include "resource.h" +#include "core_types.h" +#include "dsc.h" +#include "clk_mgr.h" + +#ifndef DC_LOGGER_INIT +#define DC_LOGGER_INIT +#undef DC_LOG_WARNING +#define DC_LOG_WARNING +#endif + +#define DML_WRAPPER_TRANSLATION_ +#include "dml_wrapper_translation.c" +#undef DML_WRAPPER_TRANSLATION_ + +static bool is_dual_plane(enum surface_pixel_format format) +{ + return format >= SURFACE_PIXEL_FORMAT_VIDEO_BEGIN || format == SURFACE_PIXEL_FORMAT_GRPH_RGBE_ALPHA; +} + +static void build_clamping_params(struct dc_stream_state *stream) +{ + stream->clamping.clamping_level = CLAMPING_FULL_RANGE; + stream->clamping.c_depth = stream->timing.display_color_depth; + stream->clamping.pixel_encoding = stream->timing.pixel_encoding; +} + +static void get_pixel_clock_parameters( + const struct pipe_ctx *pipe_ctx, + struct pixel_clk_params *pixel_clk_params) +{ + const struct dc_stream_state *stream = pipe_ctx->stream; + + /*TODO: is this halved for YCbCr 420? in that case we might want to move + * the pixel clock normalization for hdmi up to here instead of doing it + * in pll_adjust_pix_clk + */ + pixel_clk_params->requested_pix_clk_100hz = stream->timing.pix_clk_100hz; + pixel_clk_params->encoder_object_id = stream->link->link_enc->id; + pixel_clk_params->signal_type = pipe_ctx->stream->signal; + pixel_clk_params->controller_id = pipe_ctx->stream_res.tg->inst + 1; + /* TODO: un-hardcode*/ + pixel_clk_params->requested_sym_clk = LINK_RATE_LOW * + LINK_RATE_REF_FREQ_IN_KHZ; + pixel_clk_params->flags.ENABLE_SS = 0; + pixel_clk_params->color_depth = + stream->timing.display_color_depth; + pixel_clk_params->flags.DISPLAY_BLANKED = 1; + pixel_clk_params->flags.SUPPORT_YCBCR420 = (stream->timing.pixel_encoding == + PIXEL_ENCODING_YCBCR420); + pixel_clk_params->pixel_encoding = stream->timing.pixel_encoding; + if (stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR422) { + pixel_clk_params->color_depth = COLOR_DEPTH_888; + } + if (stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR420) { + pixel_clk_params->requested_pix_clk_100hz = pixel_clk_params->requested_pix_clk_100hz / 2; + } + if (stream->timing.timing_3d_format == TIMING_3D_FORMAT_HW_FRAME_PACKING) + pixel_clk_params->requested_pix_clk_100hz *= 2; + +} + +static enum dc_status build_pipe_hw_param(struct pipe_ctx *pipe_ctx) +{ + get_pixel_clock_parameters(pipe_ctx, &pipe_ctx->stream_res.pix_clk_params); + + if (pipe_ctx->clock_source) + pipe_ctx->clock_source->funcs->get_pix_clk_dividers( + pipe_ctx->clock_source, + &pipe_ctx->stream_res.pix_clk_params, + &pipe_ctx->pll_settings); + + pipe_ctx->stream->clamping.pixel_encoding = pipe_ctx->stream->timing.pixel_encoding; + + resource_build_bit_depth_reduction_params(pipe_ctx->stream, + &pipe_ctx->stream->bit_depth_params); + build_clamping_params(pipe_ctx->stream); + + return DC_OK; +} + +static void resource_build_bit_depth_reduction_params(struct dc_stream_state *stream, + struct bit_depth_reduction_params *fmt_bit_depth) +{ + enum dc_dither_option option = stream->dither_option; + enum dc_pixel_encoding pixel_encoding = + stream->timing.pixel_encoding; + + memset(fmt_bit_depth, 0, sizeof(*fmt_bit_depth)); + + if (option == DITHER_OPTION_DEFAULT) { + switch (stream->timing.display_color_depth) { + case COLOR_DEPTH_666: + option = DITHER_OPTION_SPATIAL6; + break; + case COLOR_DEPTH_888: + option = DITHER_OPTION_SPATIAL8; + break; + case COLOR_DEPTH_101010: + option = DITHER_OPTION_SPATIAL10; + break; + default: + option = DITHER_OPTION_DISABLE; + } + } + + if (option == DITHER_OPTION_DISABLE) + return; + + if (option == DITHER_OPTION_TRUN6) { + fmt_bit_depth->flags.TRUNCATE_ENABLED = 1; + fmt_bit_depth->flags.TRUNCATE_DEPTH = 0; + } else if (option == DITHER_OPTION_TRUN8 || + option == DITHER_OPTION_TRUN8_SPATIAL6 || + option == DITHER_OPTION_TRUN8_FM6) { + fmt_bit_depth->flags.TRUNCATE_ENABLED = 1; + fmt_bit_depth->flags.TRUNCATE_DEPTH = 1; + } else if (option == DITHER_OPTION_TRUN10 || + option == DITHER_OPTION_TRUN10_SPATIAL6 || + option == DITHER_OPTION_TRUN10_SPATIAL8 || + option == DITHER_OPTION_TRUN10_FM8 || + option == DITHER_OPTION_TRUN10_FM6 || + option == DITHER_OPTION_TRUN10_SPATIAL8_FM6) { + fmt_bit_depth->flags.TRUNCATE_ENABLED = 1; + fmt_bit_depth->flags.TRUNCATE_DEPTH = 2; + } + + /* special case - Formatter can only reduce by 4 bits at most. + * When reducing from 12 to 6 bits, + * HW recommends we use trunc with round mode + * (if we did nothing, trunc to 10 bits would be used) + * note that any 12->10 bit reduction is ignored prior to DCE8, + * as the input was 10 bits. + */ + if (option == DITHER_OPTION_SPATIAL6_FRAME_RANDOM || + option == DITHER_OPTION_SPATIAL6 || + option == DITHER_OPTION_FM6) { + fmt_bit_depth->flags.TRUNCATE_ENABLED = 1; + fmt_bit_depth->flags.TRUNCATE_DEPTH = 2; + fmt_bit_depth->flags.TRUNCATE_MODE = 1; + } + + /* spatial dither + * note that spatial modes 1-3 are never used + */ + if (option == DITHER_OPTION_SPATIAL6_FRAME_RANDOM || + option == DITHER_OPTION_SPATIAL6 || + option == DITHER_OPTION_TRUN10_SPATIAL6 || + option == DITHER_OPTION_TRUN8_SPATIAL6) { + fmt_bit_depth->flags.SPATIAL_DITHER_ENABLED = 1; + fmt_bit_depth->flags.SPATIAL_DITHER_DEPTH = 0; + fmt_bit_depth->flags.HIGHPASS_RANDOM = 1; + fmt_bit_depth->flags.RGB_RANDOM = + (pixel_encoding == PIXEL_ENCODING_RGB) ? 1 : 0; + } else if (option == DITHER_OPTION_SPATIAL8_FRAME_RANDOM || + option == DITHER_OPTION_SPATIAL8 || + option == DITHER_OPTION_SPATIAL8_FM6 || + option == DITHER_OPTION_TRUN10_SPATIAL8 || + option == DITHER_OPTION_TRUN10_SPATIAL8_FM6) { + fmt_bit_depth->flags.SPATIAL_DITHER_ENABLED = 1; + fmt_bit_depth->flags.SPATIAL_DITHER_DEPTH = 1; + fmt_bit_depth->flags.HIGHPASS_RANDOM = 1; + fmt_bit_depth->flags.RGB_RANDOM = + (pixel_encoding == PIXEL_ENCODING_RGB) ? 1 : 0; + } else if (option == DITHER_OPTION_SPATIAL10_FRAME_RANDOM || + option == DITHER_OPTION_SPATIAL10 || + option == DITHER_OPTION_SPATIAL10_FM8 || + option == DITHER_OPTION_SPATIAL10_FM6) { + fmt_bit_depth->flags.SPATIAL_DITHER_ENABLED = 1; + fmt_bit_depth->flags.SPATIAL_DITHER_DEPTH = 2; + fmt_bit_depth->flags.HIGHPASS_RANDOM = 1; + fmt_bit_depth->flags.RGB_RANDOM = + (pixel_encoding == PIXEL_ENCODING_RGB) ? 1 : 0; + } + + if (option == DITHER_OPTION_SPATIAL6 || + option == DITHER_OPTION_SPATIAL8 || + option == DITHER_OPTION_SPATIAL10) { + fmt_bit_depth->flags.FRAME_RANDOM = 0; + } else { + fmt_bit_depth->flags.FRAME_RANDOM = 1; + } + + ////////////////////// + //// temporal dither + ////////////////////// + if (option == DITHER_OPTION_FM6 || + option == DITHER_OPTION_SPATIAL8_FM6 || + option == DITHER_OPTION_SPATIAL10_FM6 || + option == DITHER_OPTION_TRUN10_FM6 || + option == DITHER_OPTION_TRUN8_FM6 || + option == DITHER_OPTION_TRUN10_SPATIAL8_FM6) { + fmt_bit_depth->flags.FRAME_MODULATION_ENABLED = 1; + fmt_bit_depth->flags.FRAME_MODULATION_DEPTH = 0; + } else if (option == DITHER_OPTION_FM8 || + option == DITHER_OPTION_SPATIAL10_FM8 || + option == DITHER_OPTION_TRUN10_FM8) { + fmt_bit_depth->flags.FRAME_MODULATION_ENABLED = 1; + fmt_bit_depth->flags.FRAME_MODULATION_DEPTH = 1; + } else if (option == DITHER_OPTION_FM10) { + fmt_bit_depth->flags.FRAME_MODULATION_ENABLED = 1; + fmt_bit_depth->flags.FRAME_MODULATION_DEPTH = 2; + } + + fmt_bit_depth->pixel_encoding = pixel_encoding; +} + +bool dml_validate_dsc(struct dc *dc, struct dc_state *new_ctx) +{ + int i; + + /* Validate DSC config, dsc count validation is already done */ + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe_ctx = &new_ctx->res_ctx.pipe_ctx[i]; + struct dc_stream_state *stream = pipe_ctx->stream; + struct dsc_config dsc_cfg; + struct pipe_ctx *odm_pipe; + int opp_cnt = 1; + + for (odm_pipe = pipe_ctx->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe) + opp_cnt++; + + /* Only need to validate top pipe */ + if (pipe_ctx->top_pipe || pipe_ctx->prev_odm_pipe || !stream || !stream->timing.flags.DSC) + continue; + + dsc_cfg.pic_width = (stream->timing.h_addressable + stream->timing.h_border_left + + stream->timing.h_border_right) / opp_cnt; + dsc_cfg.pic_height = stream->timing.v_addressable + stream->timing.v_border_top + + stream->timing.v_border_bottom; + dsc_cfg.pixel_encoding = stream->timing.pixel_encoding; + dsc_cfg.color_depth = stream->timing.display_color_depth; + dsc_cfg.is_odm = pipe_ctx->next_odm_pipe ? true : false; + dsc_cfg.dc_dsc_cfg = stream->timing.dsc_cfg; + dsc_cfg.dc_dsc_cfg.num_slices_h /= opp_cnt; + + if (pipe_ctx->stream_res.dsc && !pipe_ctx->stream_res.dsc->funcs->dsc_validate_stream(pipe_ctx->stream_res.dsc, &dsc_cfg)) + return false; + } + return true; +} + +enum dc_status dml_build_mapped_resource(const struct dc *dc, struct dc_state *context, struct dc_stream_state *stream) +{ + enum dc_status status = DC_OK; + struct pipe_ctx *pipe_ctx = resource_get_head_pipe_for_stream(&context->res_ctx, stream); + + if (!pipe_ctx) + return DC_ERROR_UNEXPECTED; + + + status = build_pipe_hw_param(pipe_ctx); + + return status; +} + +void dml_acquire_dsc(const struct dc *dc, + struct resource_context *res_ctx, + struct display_stream_compressor **dsc, + int pipe_idx) +{ + int i; + const struct resource_pool *pool = dc->res_pool; + struct display_stream_compressor *dsc_old = dc->current_state->res_ctx.pipe_ctx[pipe_idx].stream_res.dsc; + + ASSERT(*dsc == NULL); /* If this ASSERT fails, dsc was not released properly */ + *dsc = NULL; + + /* Always do 1-to-1 mapping when number of DSCs is same as number of pipes */ + if (pool->res_cap->num_dsc == pool->res_cap->num_opp) { + *dsc = pool->dscs[pipe_idx]; + res_ctx->is_dsc_acquired[pipe_idx] = true; + return; + } + + /* Return old DSC to avoid the need for redo it */ + if (dsc_old && !res_ctx->is_dsc_acquired[dsc_old->inst]) { + *dsc = dsc_old; + res_ctx->is_dsc_acquired[dsc_old->inst] = true; + return ; + } + + /* Find first free DSC */ + for (i = 0; i < pool->res_cap->num_dsc; i++) + if (!res_ctx->is_dsc_acquired[i]) { + *dsc = pool->dscs[i]; + res_ctx->is_dsc_acquired[i] = true; + break; + } +} + +static bool dml_split_stream_for_mpc_or_odm( + const struct dc *dc, + struct resource_context *res_ctx, + struct pipe_ctx *pri_pipe, + struct pipe_ctx *sec_pipe, + bool odm) +{ + int pipe_idx = sec_pipe->pipe_idx; + const struct resource_pool *pool = dc->res_pool; + + *sec_pipe = *pri_pipe; + + sec_pipe->pipe_idx = pipe_idx; + sec_pipe->plane_res.mi = pool->mis[pipe_idx]; + sec_pipe->plane_res.hubp = pool->hubps[pipe_idx]; + sec_pipe->plane_res.ipp = pool->ipps[pipe_idx]; + sec_pipe->plane_res.xfm = pool->transforms[pipe_idx]; + sec_pipe->plane_res.dpp = pool->dpps[pipe_idx]; + sec_pipe->plane_res.mpcc_inst = pool->dpps[pipe_idx]->inst; + sec_pipe->stream_res.dsc = NULL; + if (odm) { + if (pri_pipe->next_odm_pipe) { + ASSERT(pri_pipe->next_odm_pipe != sec_pipe); + sec_pipe->next_odm_pipe = pri_pipe->next_odm_pipe; + sec_pipe->next_odm_pipe->prev_odm_pipe = sec_pipe; + } + if (pri_pipe->top_pipe && pri_pipe->top_pipe->next_odm_pipe) { + pri_pipe->top_pipe->next_odm_pipe->bottom_pipe = sec_pipe; + sec_pipe->top_pipe = pri_pipe->top_pipe->next_odm_pipe; + } + if (pri_pipe->bottom_pipe && pri_pipe->bottom_pipe->next_odm_pipe) { + pri_pipe->bottom_pipe->next_odm_pipe->top_pipe = sec_pipe; + sec_pipe->bottom_pipe = pri_pipe->bottom_pipe->next_odm_pipe; + } + pri_pipe->next_odm_pipe = sec_pipe; + sec_pipe->prev_odm_pipe = pri_pipe; + ASSERT(sec_pipe->top_pipe == NULL); + + if (!sec_pipe->top_pipe) + sec_pipe->stream_res.opp = pool->opps[pipe_idx]; + else + sec_pipe->stream_res.opp = sec_pipe->top_pipe->stream_res.opp; + if (sec_pipe->stream->timing.flags.DSC == 1) { + dml_acquire_dsc(dc, res_ctx, &sec_pipe->stream_res.dsc, pipe_idx); + ASSERT(sec_pipe->stream_res.dsc); + if (sec_pipe->stream_res.dsc == NULL) + return false; + } + } else { + if (pri_pipe->bottom_pipe) { + ASSERT(pri_pipe->bottom_pipe != sec_pipe); + sec_pipe->bottom_pipe = pri_pipe->bottom_pipe; + sec_pipe->bottom_pipe->top_pipe = sec_pipe; + } + pri_pipe->bottom_pipe = sec_pipe; + sec_pipe->top_pipe = pri_pipe; + + ASSERT(pri_pipe->plane_state); + } + + return true; +} + +static struct pipe_ctx *dml_find_split_pipe( + struct dc *dc, + struct dc_state *context, + int old_index) +{ + struct pipe_ctx *pipe = NULL; + int i; + + if (old_index >= 0 && context->res_ctx.pipe_ctx[old_index].stream == NULL) { + pipe = &context->res_ctx.pipe_ctx[old_index]; + pipe->pipe_idx = old_index; + } + + if (!pipe) + for (i = dc->res_pool->pipe_count - 1; i >= 0; i--) { + if (dc->current_state->res_ctx.pipe_ctx[i].top_pipe == NULL + && dc->current_state->res_ctx.pipe_ctx[i].prev_odm_pipe == NULL) { + if (context->res_ctx.pipe_ctx[i].stream == NULL) { + pipe = &context->res_ctx.pipe_ctx[i]; + pipe->pipe_idx = i; + break; + } + } + } + + /* + * May need to fix pipes getting tossed from 1 opp to another on flip + * Add for debugging transient underflow during topology updates: + * ASSERT(pipe); + */ + if (!pipe) + for (i = dc->res_pool->pipe_count - 1; i >= 0; i--) { + if (context->res_ctx.pipe_ctx[i].stream == NULL) { + pipe = &context->res_ctx.pipe_ctx[i]; + pipe->pipe_idx = i; + break; + } + } + + return pipe; +} + +static void dml_release_dsc(struct resource_context *res_ctx, + const struct resource_pool *pool, + struct display_stream_compressor **dsc) +{ + int i; + + for (i = 0; i < pool->res_cap->num_dsc; i++) + if (pool->dscs[i] == *dsc) { + res_ctx->is_dsc_acquired[i] = false; + *dsc = NULL; + break; + } +} + +static int dml_get_num_mpc_splits(struct pipe_ctx *pipe) +{ + int mpc_split_count = 0; + struct pipe_ctx *other_pipe = pipe->bottom_pipe; + + while (other_pipe && other_pipe->plane_state == pipe->plane_state) { + mpc_split_count++; + other_pipe = other_pipe->bottom_pipe; + } + other_pipe = pipe->top_pipe; + while (other_pipe && other_pipe->plane_state == pipe->plane_state) { + mpc_split_count++; + other_pipe = other_pipe->top_pipe; + } + + return mpc_split_count; +} + +static bool dml_enough_pipes_for_subvp(struct dc *dc, + struct dc_state *context) +{ + int i = 0; + int num_pipes = 0; + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + + if (pipe->stream && pipe->plane_state) + num_pipes++; + } + + // Sub-VP only possible if the number of "real" pipes is + // less than or equal to half the number of available pipes + if (num_pipes * 2 > dc->res_pool->pipe_count) + return false; + + return true; +} + +static int dml_validate_apply_pipe_split_flags( + struct dc *dc, + struct dc_state *context, + int vlevel, + int *split, + bool *merge) +{ + int i, pipe_idx, vlevel_split; + int plane_count = 0; + bool force_split = false; + bool avoid_split = dc->debug.pipe_split_policy == MPC_SPLIT_AVOID; + struct vba_vars_st *v = &context->bw_ctx.dml.vba; + int max_mpc_comb = v->maxMpcComb; + + if (context->stream_count > 1) { + if (dc->debug.pipe_split_policy == MPC_SPLIT_AVOID_MULT_DISP) + avoid_split = true; + } else if (dc->debug.force_single_disp_pipe_split) + force_split = true; + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + + /** + * Workaround for avoiding pipe-split in cases where we'd split + * planes that are too small, resulting in splits that aren't + * valid for the scaler. + */ + if (pipe->plane_state && + (pipe->plane_state->dst_rect.width <= 16 || + pipe->plane_state->dst_rect.height <= 16 || + pipe->plane_state->src_rect.width <= 16 || + pipe->plane_state->src_rect.height <= 16)) + avoid_split = true; + + /* TODO: fix dc bugs and remove this split threshold thing */ + if (pipe->stream && !pipe->prev_odm_pipe && + (!pipe->top_pipe || pipe->top_pipe->plane_state != pipe->plane_state)) + ++plane_count; + } + if (plane_count > dc->res_pool->pipe_count / 2) + avoid_split = true; + + /* W/A: Mode timing with borders may not work well with pipe split, avoid for this corner case */ + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + struct dc_crtc_timing timing; + + if (!pipe->stream) + continue; + else { + timing = pipe->stream->timing; + if (timing.h_border_left + timing.h_border_right + + timing.v_border_top + timing.v_border_bottom > 0) { + avoid_split = true; + break; + } + } + } + + /* Avoid split loop looks for lowest voltage level that allows most unsplit pipes possible */ + if (avoid_split) { + for (i = 0, pipe_idx = 0; i < dc->res_pool->pipe_count; i++) { + if (!context->res_ctx.pipe_ctx[i].stream) + continue; + + for (vlevel_split = vlevel; vlevel <= context->bw_ctx.dml.soc.num_states; vlevel++) + if (v->NoOfDPP[vlevel][0][pipe_idx] == 1 && + v->ModeSupport[vlevel][0]) + break; + /* Impossible to not split this pipe */ + if (vlevel > context->bw_ctx.dml.soc.num_states) + vlevel = vlevel_split; + else + max_mpc_comb = 0; + pipe_idx++; + } + v->maxMpcComb = max_mpc_comb; + } + + /* Split loop sets which pipe should be split based on dml outputs and dc flags */ + for (i = 0, pipe_idx = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + int pipe_plane = v->pipe_plane[pipe_idx]; + bool split4mpc = context->stream_count == 1 && plane_count == 1 + && dc->config.enable_4to1MPC && dc->res_pool->pipe_count >= 4; + + if (!context->res_ctx.pipe_ctx[i].stream) + continue; + + if (split4mpc || v->NoOfDPP[vlevel][max_mpc_comb][pipe_plane] == 4) + split[i] = 4; + else if (force_split || v->NoOfDPP[vlevel][max_mpc_comb][pipe_plane] == 2) + split[i] = 2; + + if ((pipe->stream->view_format == + VIEW_3D_FORMAT_SIDE_BY_SIDE || + pipe->stream->view_format == + VIEW_3D_FORMAT_TOP_AND_BOTTOM) && + (pipe->stream->timing.timing_3d_format == + TIMING_3D_FORMAT_TOP_AND_BOTTOM || + pipe->stream->timing.timing_3d_format == + TIMING_3D_FORMAT_SIDE_BY_SIDE)) + split[i] = 2; + if (dc->debug.force_odm_combine & (1 << pipe->stream_res.tg->inst)) { + split[i] = 2; + v->ODMCombineEnablePerState[vlevel][pipe_plane] = dm_odm_combine_mode_2to1; + } + if (dc->debug.force_odm_combine_4to1 & (1 << pipe->stream_res.tg->inst)) { + split[i] = 4; + v->ODMCombineEnablePerState[vlevel][pipe_plane] = dm_odm_combine_mode_4to1; + } + /*420 format workaround*/ + if (pipe->stream->timing.h_addressable > 7680 && + pipe->stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR420) { + split[i] = 4; + } + + v->ODMCombineEnabled[pipe_plane] = + v->ODMCombineEnablePerState[vlevel][pipe_plane]; + + if (v->ODMCombineEnabled[pipe_plane] == dm_odm_combine_mode_disabled) { + if (dml_get_num_mpc_splits(pipe) == 1) { + /*If need split for mpc but 2 way split already*/ + if (split[i] == 4) + split[i] = 2; /* 2 -> 4 MPC */ + else if (split[i] == 2) + split[i] = 0; /* 2 -> 2 MPC */ + else if (pipe->top_pipe && pipe->top_pipe->plane_state == pipe->plane_state) + merge[i] = true; /* 2 -> 1 MPC */ + } else if (dml_get_num_mpc_splits(pipe) == 3) { + /*If need split for mpc but 4 way split already*/ + if (split[i] == 2 && ((pipe->top_pipe && !pipe->top_pipe->top_pipe) + || !pipe->bottom_pipe)) { + merge[i] = true; /* 4 -> 2 MPC */ + } else if (split[i] == 0 && pipe->top_pipe && + pipe->top_pipe->plane_state == pipe->plane_state) + merge[i] = true; /* 4 -> 1 MPC */ + split[i] = 0; + } else if (dml_get_num_mpc_splits(pipe)) { + /* ODM -> MPC transition */ + if (pipe->prev_odm_pipe) { + split[i] = 0; + merge[i] = true; + } + } + } else { + if (dml_get_num_mpc_splits(pipe) == 1) { + /*If need split for odm but 2 way split already*/ + if (split[i] == 4) + split[i] = 2; /* 2 -> 4 ODM */ + else if (split[i] == 2) + split[i] = 0; /* 2 -> 2 ODM */ + else if (pipe->prev_odm_pipe) { + ASSERT(0); /* NOT expected yet */ + merge[i] = true; /* exit ODM */ + } + } else if (dml_get_num_mpc_splits(pipe) == 3) { + /*If need split for odm but 4 way split already*/ + if (split[i] == 2 && ((pipe->prev_odm_pipe && !pipe->prev_odm_pipe->prev_odm_pipe) + || !pipe->next_odm_pipe)) { + ASSERT(0); /* NOT expected yet */ + merge[i] = true; /* 4 -> 2 ODM */ + } else if (split[i] == 0 && pipe->prev_odm_pipe) { + ASSERT(0); /* NOT expected yet */ + merge[i] = true; /* exit ODM */ + } + split[i] = 0; + } else if (dml_get_num_mpc_splits(pipe)) { + /* MPC -> ODM transition */ + ASSERT(0); /* NOT expected yet */ + if (pipe->top_pipe && pipe->top_pipe->plane_state == pipe->plane_state) { + split[i] = 0; + merge[i] = true; + } + } + } + + /* Adjust dppclk when split is forced, do not bother with dispclk */ + if (split[i] != 0 && v->NoOfDPP[vlevel][max_mpc_comb][pipe_idx] == 1) + v->RequiredDPPCLK[vlevel][max_mpc_comb][pipe_idx] /= 2; + pipe_idx++; + } + + return vlevel; +} + +static void dml_set_phantom_stream_timing(struct dc *dc, + struct dc_state *context, + struct pipe_ctx *ref_pipe, + struct dc_stream_state *phantom_stream) +{ + // phantom_vactive = blackout (latency + margin) + fw_processing_delays + pstate allow width + uint32_t phantom_vactive_us = context->bw_ctx.dml.soc.dram_clock_change_latency_us + 60 + + dc->caps.subvp_fw_processing_delay_us + + dc->caps.subvp_pstate_allow_width_us; + uint32_t phantom_vactive = ((double)phantom_vactive_us/1000000) * + (ref_pipe->stream->timing.pix_clk_100hz * 100) / + (double)ref_pipe->stream->timing.h_total; + uint32_t phantom_bp = ref_pipe->pipe_dlg_param.vstartup_start; + + phantom_stream->dst.y = 0; + phantom_stream->dst.height = phantom_vactive; + phantom_stream->src.y = 0; + phantom_stream->src.height = phantom_vactive; + + phantom_stream->timing.v_addressable = phantom_vactive; + phantom_stream->timing.v_front_porch = 1; + phantom_stream->timing.v_total = phantom_stream->timing.v_addressable + + phantom_stream->timing.v_front_porch + + phantom_stream->timing.v_sync_width + + phantom_bp; +} + +static struct dc_stream_state *dml_enable_phantom_stream(struct dc *dc, + struct dc_state *context, + struct pipe_ctx *ref_pipe) +{ + struct dc_stream_state *phantom_stream = NULL; + + phantom_stream = dc_create_stream_for_sink(ref_pipe->stream->sink); + phantom_stream->signal = SIGNAL_TYPE_VIRTUAL; + phantom_stream->dpms_off = true; + phantom_stream->mall_stream_config.type = SUBVP_PHANTOM; + phantom_stream->mall_stream_config.paired_stream = ref_pipe->stream; + ref_pipe->stream->mall_stream_config.type = SUBVP_MAIN; + ref_pipe->stream->mall_stream_config.paired_stream = phantom_stream; + + /* stream has limited viewport and small timing */ + memcpy(&phantom_stream->timing, &ref_pipe->stream->timing, sizeof(phantom_stream->timing)); + memcpy(&phantom_stream->src, &ref_pipe->stream->src, sizeof(phantom_stream->src)); + memcpy(&phantom_stream->dst, &ref_pipe->stream->dst, sizeof(phantom_stream->dst)); + dml_set_phantom_stream_timing(dc, context, ref_pipe, phantom_stream); + + dc_add_stream_to_ctx(dc, context, phantom_stream); + dc->hwss.apply_ctx_to_hw(dc, context); + return phantom_stream; +} + +static void dml_enable_phantom_plane(struct dc *dc, + struct dc_state *context, + struct dc_stream_state *phantom_stream, + struct pipe_ctx *main_pipe) +{ + struct dc_plane_state *phantom_plane = NULL; + struct dc_plane_state *prev_phantom_plane = NULL; + struct pipe_ctx *curr_pipe = main_pipe; + + while (curr_pipe) { + if (curr_pipe->top_pipe && curr_pipe->top_pipe->plane_state == curr_pipe->plane_state) + phantom_plane = prev_phantom_plane; + else + phantom_plane = dc_create_plane_state(dc); + + memcpy(&phantom_plane->address, &curr_pipe->plane_state->address, sizeof(phantom_plane->address)); + memcpy(&phantom_plane->scaling_quality, &curr_pipe->plane_state->scaling_quality, + sizeof(phantom_plane->scaling_quality)); + memcpy(&phantom_plane->src_rect, &curr_pipe->plane_state->src_rect, sizeof(phantom_plane->src_rect)); + memcpy(&phantom_plane->dst_rect, &curr_pipe->plane_state->dst_rect, sizeof(phantom_plane->dst_rect)); + memcpy(&phantom_plane->clip_rect, &curr_pipe->plane_state->clip_rect, sizeof(phantom_plane->clip_rect)); + memcpy(&phantom_plane->plane_size, &curr_pipe->plane_state->plane_size, + sizeof(phantom_plane->plane_size)); + memcpy(&phantom_plane->tiling_info, &curr_pipe->plane_state->tiling_info, + sizeof(phantom_plane->tiling_info)); + memcpy(&phantom_plane->dcc, &curr_pipe->plane_state->dcc, sizeof(phantom_plane->dcc)); + /* Currently compat_level is undefined in dc_state + * phantom_plane->compat_level = curr_pipe->plane_state->compat_level; + */ + phantom_plane->format = curr_pipe->plane_state->format; + phantom_plane->rotation = curr_pipe->plane_state->rotation; + phantom_plane->visible = curr_pipe->plane_state->visible; + + /* Shadow pipe has small viewport. */ + phantom_plane->clip_rect.y = 0; + phantom_plane->clip_rect.height = phantom_stream->timing.v_addressable; + + dc_add_plane_to_context(dc, phantom_stream, phantom_plane, context); + + curr_pipe = curr_pipe->bottom_pipe; + prev_phantom_plane = phantom_plane; + } +} + +static void dml_add_phantom_pipes(struct dc *dc, struct dc_state *context) +{ + int i = 0; + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + struct dc_stream_state *ref_stream = pipe->stream; + // Only construct phantom stream for top pipes that have plane enabled + if (!pipe->top_pipe && pipe->plane_state && pipe->stream && + pipe->stream->mall_stream_config.type == SUBVP_NONE) { + struct dc_stream_state *phantom_stream = NULL; + + phantom_stream = dml_enable_phantom_stream(dc, context, pipe); + dml_enable_phantom_plane(dc, context, phantom_stream, pipe); + } + } + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + + if (pipe->plane_state && pipe->stream && + pipe->stream->mall_stream_config.type == SUBVP_PHANTOM) { + pipe->stream->use_dynamic_meta = false; + pipe->plane_state->flip_immediate = false; + if (!resource_build_scaling_params(pipe)) { + // Log / remove phantom pipes since failed to build scaling params + } + } + } +} + +static void dml_remove_phantom_pipes(struct dc *dc, struct dc_state *context) +{ + int i; + bool removed_pipe = false; + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + // build scaling params for phantom pipes + if (pipe->plane_state && pipe->stream && pipe->stream->mall_stream_config.type == SUBVP_PHANTOM) { + dc_rem_all_planes_for_stream(dc, pipe->stream, context); + dc_remove_stream_from_ctx(dc, context, pipe->stream); + removed_pipe = true; + } + + // Clear all phantom stream info + if (pipe->stream) { + pipe->stream->mall_stream_config.type = SUBVP_NONE; + pipe->stream->mall_stream_config.paired_stream = NULL; + } + } + if (removed_pipe) + dc->hwss.apply_ctx_to_hw(dc, context); +} + +/* + * If the input state contains no upstream planes for a particular pipe (i.e. only timing) + * we need to populate some "conservative" plane information as DML cannot handle "no planes" + */ +static void populate_default_plane_from_timing(const struct dc_crtc_timing *timing, struct _vcs_dpi_display_pipe_params_st *pipe) +{ + pipe->src.is_hsplit = pipe->dest.odm_combine != dm_odm_combine_mode_disabled; + pipe->src.source_scan = dm_horz; + pipe->src.sw_mode = dm_sw_4kb_s; + pipe->src.macro_tile_size = dm_64k_tile; + pipe->src.viewport_width = timing->h_addressable; + if (pipe->src.viewport_width > 1920) + pipe->src.viewport_width = 1920; + pipe->src.viewport_height = timing->v_addressable; + if (pipe->src.viewport_height > 1080) + pipe->src.viewport_height = 1080; + pipe->src.surface_height_y = pipe->src.viewport_height; + pipe->src.surface_width_y = pipe->src.viewport_width; + pipe->src.surface_height_c = pipe->src.viewport_height; + pipe->src.surface_width_c = pipe->src.viewport_width; + pipe->src.data_pitch = ((pipe->src.viewport_width + 255) / 256) * 256; + pipe->src.source_format = dm_444_32; + pipe->dest.recout_width = pipe->src.viewport_width; + pipe->dest.recout_height = pipe->src.viewport_height; + pipe->dest.full_recout_width = pipe->dest.recout_width; + pipe->dest.full_recout_height = pipe->dest.recout_height; + pipe->scale_ratio_depth.lb_depth = dm_lb_16; + pipe->scale_ratio_depth.hscl_ratio = 1.0; + pipe->scale_ratio_depth.vscl_ratio = 1.0; + pipe->scale_ratio_depth.scl_enable = 0; + pipe->scale_taps.htaps = 1; + pipe->scale_taps.vtaps = 1; + pipe->dest.vtotal_min = timing->v_total; + pipe->dest.vtotal_max = timing->v_total; + + if (pipe->dest.odm_combine == dm_odm_combine_mode_2to1) { + pipe->src.viewport_width /= 2; + pipe->dest.recout_width /= 2; + } else if (pipe->dest.odm_combine == dm_odm_combine_mode_4to1) { + pipe->src.viewport_width /= 4; + pipe->dest.recout_width /= 4; + } + + pipe->src.dcc = false; + pipe->src.dcc_rate = 1; +} + +/* + * If the pipe is not blending (i.e. pipe_ctx->top pipe == null) then its + * hsplit group is equal to its own pipe ID + * Otherwise, all pipes part of the same blending tree have the same hsplit group + * ID as the top most pipe + * + * If the pipe ctx is ODM combined, then similar logic follows + */ +static void populate_hsplit_group_from_dc_pipe_ctx (const struct pipe_ctx *dc_pipe_ctx, struct _vcs_dpi_display_e2e_pipe_params_st *e2e_pipe) +{ + e2e_pipe->pipe.src.hsplit_grp = dc_pipe_ctx->pipe_idx; + + if (dc_pipe_ctx->top_pipe && dc_pipe_ctx->top_pipe->plane_state + == dc_pipe_ctx->plane_state) { + struct pipe_ctx *first_pipe = dc_pipe_ctx->top_pipe; + int split_idx = 0; + + while (first_pipe->top_pipe && first_pipe->top_pipe->plane_state + == dc_pipe_ctx->plane_state) { + first_pipe = first_pipe->top_pipe; + split_idx++; + } + + /* Treat 4to1 mpc combine as an mpo of 2 2-to-1 combines */ + if (split_idx == 0) + e2e_pipe->pipe.src.hsplit_grp = first_pipe->pipe_idx; + else if (split_idx == 1) + e2e_pipe->pipe.src.hsplit_grp = dc_pipe_ctx->pipe_idx; + else if (split_idx == 2) + e2e_pipe->pipe.src.hsplit_grp = dc_pipe_ctx->top_pipe->pipe_idx; + + } else if (dc_pipe_ctx->prev_odm_pipe) { + struct pipe_ctx *first_pipe = dc_pipe_ctx->prev_odm_pipe; + + while (first_pipe->prev_odm_pipe) + first_pipe = first_pipe->prev_odm_pipe; + e2e_pipe->pipe.src.hsplit_grp = first_pipe->pipe_idx; + } +} + +static void populate_dml_from_dc_pipe_ctx (const struct pipe_ctx *dc_pipe_ctx, struct _vcs_dpi_display_e2e_pipe_params_st *e2e_pipe, int always_scale) +{ + const struct dc_plane_state *pln = dc_pipe_ctx->plane_state; + const struct scaler_data *scl = &dc_pipe_ctx->plane_res.scl_data; + + e2e_pipe->pipe.src.immediate_flip = pln->flip_immediate; + e2e_pipe->pipe.src.is_hsplit = (dc_pipe_ctx->bottom_pipe && dc_pipe_ctx->bottom_pipe->plane_state == pln) + || (dc_pipe_ctx->top_pipe && dc_pipe_ctx->top_pipe->plane_state == pln) + || e2e_pipe->pipe.dest.odm_combine != dm_odm_combine_mode_disabled; + + /* stereo is not split */ + if (pln->stereo_format == PLANE_STEREO_FORMAT_SIDE_BY_SIDE || + pln->stereo_format == PLANE_STEREO_FORMAT_TOP_AND_BOTTOM) { + e2e_pipe->pipe.src.is_hsplit = false; + e2e_pipe->pipe.src.hsplit_grp = dc_pipe_ctx->pipe_idx; + } + + e2e_pipe->pipe.src.source_scan = pln->rotation == ROTATION_ANGLE_90 + || pln->rotation == ROTATION_ANGLE_270 ? dm_vert : dm_horz; + e2e_pipe->pipe.src.viewport_y_y = scl->viewport.y; + e2e_pipe->pipe.src.viewport_y_c = scl->viewport_c.y; + e2e_pipe->pipe.src.viewport_width = scl->viewport.width; + e2e_pipe->pipe.src.viewport_width_c = scl->viewport_c.width; + e2e_pipe->pipe.src.viewport_height = scl->viewport.height; + e2e_pipe->pipe.src.viewport_height_c = scl->viewport_c.height; + e2e_pipe->pipe.src.viewport_width_max = pln->src_rect.width; + e2e_pipe->pipe.src.viewport_height_max = pln->src_rect.height; + e2e_pipe->pipe.src.surface_width_y = pln->plane_size.surface_size.width; + e2e_pipe->pipe.src.surface_height_y = pln->plane_size.surface_size.height; + e2e_pipe->pipe.src.surface_width_c = pln->plane_size.chroma_size.width; + e2e_pipe->pipe.src.surface_height_c = pln->plane_size.chroma_size.height; + + if (pln->format == SURFACE_PIXEL_FORMAT_GRPH_RGBE_ALPHA + || pln->format >= SURFACE_PIXEL_FORMAT_VIDEO_BEGIN) { + e2e_pipe->pipe.src.data_pitch = pln->plane_size.surface_pitch; + e2e_pipe->pipe.src.data_pitch_c = pln->plane_size.chroma_pitch; + e2e_pipe->pipe.src.meta_pitch = pln->dcc.meta_pitch; + e2e_pipe->pipe.src.meta_pitch_c = pln->dcc.meta_pitch_c; + } else { + e2e_pipe->pipe.src.data_pitch = pln->plane_size.surface_pitch; + e2e_pipe->pipe.src.meta_pitch = pln->dcc.meta_pitch; + } + e2e_pipe->pipe.src.dcc = pln->dcc.enable; + e2e_pipe->pipe.src.dcc_rate = 1; + e2e_pipe->pipe.dest.recout_width = scl->recout.width; + e2e_pipe->pipe.dest.recout_height = scl->recout.height; + e2e_pipe->pipe.dest.full_recout_height = scl->recout.height; + e2e_pipe->pipe.dest.full_recout_width = scl->recout.width; + if (e2e_pipe->pipe.dest.odm_combine == dm_odm_combine_mode_2to1) + e2e_pipe->pipe.dest.full_recout_width *= 2; + else if (e2e_pipe->pipe.dest.odm_combine == dm_odm_combine_mode_4to1) + e2e_pipe->pipe.dest.full_recout_width *= 4; + else { + struct pipe_ctx *split_pipe = dc_pipe_ctx->bottom_pipe; + + while (split_pipe && split_pipe->plane_state == pln) { + e2e_pipe->pipe.dest.full_recout_width += split_pipe->plane_res.scl_data.recout.width; + split_pipe = split_pipe->bottom_pipe; + } + split_pipe = dc_pipe_ctx->top_pipe; + while (split_pipe && split_pipe->plane_state == pln) { + e2e_pipe->pipe.dest.full_recout_width += split_pipe->plane_res.scl_data.recout.width; + split_pipe = split_pipe->top_pipe; + } + } + + e2e_pipe->pipe.scale_ratio_depth.lb_depth = dm_lb_16; + e2e_pipe->pipe.scale_ratio_depth.hscl_ratio = (double) scl->ratios.horz.value / (1ULL<<32); + e2e_pipe->pipe.scale_ratio_depth.hscl_ratio_c = (double) scl->ratios.horz_c.value / (1ULL<<32); + e2e_pipe->pipe.scale_ratio_depth.vscl_ratio = (double) scl->ratios.vert.value / (1ULL<<32); + e2e_pipe->pipe.scale_ratio_depth.vscl_ratio_c = (double) scl->ratios.vert_c.value / (1ULL<<32); + e2e_pipe->pipe.scale_ratio_depth.scl_enable = + scl->ratios.vert.value != dc_fixpt_one.value + || scl->ratios.horz.value != dc_fixpt_one.value + || scl->ratios.vert_c.value != dc_fixpt_one.value + || scl->ratios.horz_c.value != dc_fixpt_one.value /*Lb only or Full scl*/ + || always_scale; /*support always scale*/ + e2e_pipe->pipe.scale_taps.htaps = scl->taps.h_taps; + e2e_pipe->pipe.scale_taps.htaps_c = scl->taps.h_taps_c; + e2e_pipe->pipe.scale_taps.vtaps = scl->taps.v_taps; + e2e_pipe->pipe.scale_taps.vtaps_c = scl->taps.v_taps_c; + + /* Currently compat_level is not defined. Commenting it until further resolution + * if (pln->compat_level == DC_LEGACY_TILING_ADDR_GEN_TWO) { + swizzle_to_dml_params(pln->tiling_info.gfx9.swizzle, + &e2e_pipe->pipe.src.sw_mode); + e2e_pipe->pipe.src.macro_tile_size = + swizzle_mode_to_macro_tile_size(pln->tiling_info.gfx9.swizzle); + } else { + gfx10array_mode_to_dml_params(pln->tiling_info.gfx10compatible.array_mode, + pln->compat_level, + &e2e_pipe->pipe.src.sw_mode); + e2e_pipe->pipe.src.macro_tile_size = dm_4k_tile; + }*/ + + e2e_pipe->pipe.src.source_format = dc_source_format_to_dml_source_format(pln->format); +} + +static void populate_dml_cursor_parameters_from_dc_pipe_ctx (const struct pipe_ctx *dc_pipe_ctx, struct _vcs_dpi_display_e2e_pipe_params_st *e2e_pipe) +{ + /* + * For graphic plane, cursor number is 1, nv12 is 0 + * bw calculations due to cursor on/off + */ + if (dc_pipe_ctx->plane_state && + (dc_pipe_ctx->plane_state->address.type == PLN_ADDR_TYPE_VIDEO_PROGRESSIVE || + dc_pipe_ctx->stream->mall_stream_config.type == SUBVP_PHANTOM)) + e2e_pipe->pipe.src.num_cursors = 0; + else + e2e_pipe->pipe.src.num_cursors = 1; + + e2e_pipe->pipe.src.cur0_src_width = 256; + e2e_pipe->pipe.src.cur0_bpp = dm_cur_32bit; +} + +static int populate_dml_pipes_from_context_base( + struct dc *dc, + struct dc_state *context, + display_e2e_pipe_params_st *pipes, + bool fast_validate) +{ + int pipe_cnt, i; + bool synchronized_vblank = true; + struct resource_context *res_ctx = &context->res_ctx; + + for (i = 0, pipe_cnt = -1; i < dc->res_pool->pipe_count; i++) { + if (!res_ctx->pipe_ctx[i].stream) + continue; + + if (pipe_cnt < 0) { + pipe_cnt = i; + continue; + } + + if (res_ctx->pipe_ctx[pipe_cnt].stream == res_ctx->pipe_ctx[i].stream) + continue; + + if (dc->debug.disable_timing_sync || + (!resource_are_streams_timing_synchronizable( + res_ctx->pipe_ctx[pipe_cnt].stream, + res_ctx->pipe_ctx[i].stream) && + !resource_are_vblanks_synchronizable( + res_ctx->pipe_ctx[pipe_cnt].stream, + res_ctx->pipe_ctx[i].stream))) { + synchronized_vblank = false; + break; + } + } + + for (i = 0, pipe_cnt = 0; i < dc->res_pool->pipe_count; i++) { + struct dc_crtc_timing *timing = &res_ctx->pipe_ctx[i].stream->timing; + + struct audio_check aud_check = {0}; + if (!res_ctx->pipe_ctx[i].stream) + continue; + + /* todo: + pipes[pipe_cnt].pipe.src.dynamic_metadata_enable = 0; + pipes[pipe_cnt].pipe.src.dcc = 0; + pipes[pipe_cnt].pipe.src.vm = 0;*/ + + pipes[pipe_cnt].clks_cfg.refclk_mhz = dc->res_pool->ref_clocks.dchub_ref_clock_inKhz / 1000.0; + + pipes[pipe_cnt].dout.dsc_enable = res_ctx->pipe_ctx[i].stream->timing.flags.DSC; + /* todo: rotation?*/ + pipes[pipe_cnt].dout.dsc_slices = res_ctx->pipe_ctx[i].stream->timing.dsc_cfg.num_slices_h; + if (res_ctx->pipe_ctx[i].stream->use_dynamic_meta) { + pipes[pipe_cnt].pipe.src.dynamic_metadata_enable = true; + /* 1/2 vblank */ + pipes[pipe_cnt].pipe.src.dynamic_metadata_lines_before_active = + (timing->v_total - timing->v_addressable + - timing->v_border_top - timing->v_border_bottom) / 2; + /* 36 bytes dp, 32 hdmi */ + pipes[pipe_cnt].pipe.src.dynamic_metadata_xmit_bytes = + dc_is_dp_signal(res_ctx->pipe_ctx[i].stream->signal) ? 36 : 32; + } + pipes[pipe_cnt].pipe.dest.synchronized_vblank_all_planes = synchronized_vblank; + + dc_timing_to_dml_timing(timing, &pipes[pipe_cnt].pipe.dest); + pipes[pipe_cnt].pipe.dest.vtotal_min = res_ctx->pipe_ctx[i].stream->adjust.v_total_min; + pipes[pipe_cnt].pipe.dest.vtotal_max = res_ctx->pipe_ctx[i].stream->adjust.v_total_max; + + pipes[pipe_cnt].pipe.dest.otg_inst = res_ctx->pipe_ctx[i].stream_res.tg->inst; + + pipes[pipe_cnt].pipe.dest.odm_combine = get_dml_odm_combine(&res_ctx->pipe_ctx[i]); + + populate_hsplit_group_from_dc_pipe_ctx(&res_ctx->pipe_ctx[i], &pipes[pipe_cnt]); + + pipes[pipe_cnt].dout.dp_lanes = 4; + pipes[pipe_cnt].dout.is_virtual = 0; + pipes[pipe_cnt].dout.output_type = get_dml_output_type(res_ctx->pipe_ctx[i].stream->signal); + if (pipes[pipe_cnt].dout.output_type < 0) { + pipes[pipe_cnt].dout.output_type = dm_dp; + pipes[pipe_cnt].dout.is_virtual = 1; + } + + populate_color_depth_and_encoding_from_timing(&res_ctx->pipe_ctx[i].stream->timing, &pipes[pipe_cnt].dout); + + if (res_ctx->pipe_ctx[i].stream->timing.flags.DSC) + pipes[pipe_cnt].dout.output_bpp = res_ctx->pipe_ctx[i].stream->timing.dsc_cfg.bits_per_pixel / 16.0; + + /* todo: default max for now, until there is logic reflecting this in dc*/ + pipes[pipe_cnt].dout.dsc_input_bpc = 12; + /*fill up the audio sample rate (unit in kHz)*/ + get_audio_check(&res_ctx->pipe_ctx[i].stream->audio_info, &aud_check); + pipes[pipe_cnt].dout.max_audio_sample_rate = aud_check.max_audiosample_rate / 1000; + + populate_dml_cursor_parameters_from_dc_pipe_ctx(&res_ctx->pipe_ctx[i], &pipes[pipe_cnt]); + + if (!res_ctx->pipe_ctx[i].plane_state) { + populate_default_plane_from_timing(timing, &pipes[pipe_cnt].pipe); + } else { + populate_dml_from_dc_pipe_ctx(&res_ctx->pipe_ctx[i], &pipes[pipe_cnt], dc->debug.always_scale); + } + + pipe_cnt++; + } + + /* populate writeback information */ + if (dc->res_pool) + dc->res_pool->funcs->populate_dml_writeback_from_context(dc, res_ctx, pipes); + + return pipe_cnt; +} + +static int dml_populate_dml_pipes_from_context( + struct dc *dc, struct dc_state *context, + display_e2e_pipe_params_st *pipes, + bool fast_validate) +{ + int i, pipe_cnt; + struct resource_context *res_ctx = &context->res_ctx; + struct pipe_ctx *pipe; + + populate_dml_pipes_from_context_base(dc, context, pipes, fast_validate); + + for (i = 0, pipe_cnt = 0; i < dc->res_pool->pipe_count; i++) { + struct dc_crtc_timing *timing; + + if (!res_ctx->pipe_ctx[i].stream) + continue; + pipe = &res_ctx->pipe_ctx[i]; + timing = &pipe->stream->timing; + + pipes[pipe_cnt].pipe.src.gpuvm = true; + pipes[pipe_cnt].pipe.src.dcc_fraction_of_zs_req_luma = 0; + pipes[pipe_cnt].pipe.src.dcc_fraction_of_zs_req_chroma = 0; + pipes[pipe_cnt].pipe.dest.vfront_porch = timing->v_front_porch; + + pipes[pipe_cnt].dout.dsc_input_bpc = 0; + if (pipes[pipe_cnt].dout.dsc_enable) { + switch (timing->display_color_depth) { + case COLOR_DEPTH_888: + pipes[pipe_cnt].dout.dsc_input_bpc = 8; + break; + case COLOR_DEPTH_101010: + pipes[pipe_cnt].dout.dsc_input_bpc = 10; + break; + case COLOR_DEPTH_121212: + pipes[pipe_cnt].dout.dsc_input_bpc = 12; + break; + default: + ASSERT(0); + break; + } + } + pipe_cnt++; + } + dc->config.enable_4to1MPC = false; + if (pipe_cnt == 1 && pipe->plane_state && !dc->debug.disable_z9_mpc) { + if (is_dual_plane(pipe->plane_state->format) + && pipe->plane_state->src_rect.width <= 1920 && pipe->plane_state->src_rect.height <= 1080) { + dc->config.enable_4to1MPC = true; + } else if (!is_dual_plane(pipe->plane_state->format)) { + context->bw_ctx.dml.ip.det_buffer_size_kbytes = 192; + pipes[0].pipe.src.unbounded_req_mode = true; + } + } + + return pipe_cnt; +} + +static void dml_full_validate_bw_helper(struct dc *dc, + struct dc_state *context, + display_e2e_pipe_params_st *pipes, + int *vlevel, + int *split, + bool *merge, + int *pipe_cnt) +{ + struct vba_vars_st *vba = &context->bw_ctx.dml.vba; + + /* + * DML favors voltage over p-state, but we're more interested in + * supporting p-state over voltage. We can't support p-state in + * prefetch mode > 0 so try capping the prefetch mode to start. + */ + context->bw_ctx.dml.soc.allow_dram_self_refresh_or_dram_clock_change_in_vblank = + dm_allow_self_refresh_and_mclk_switch; + *vlevel = dml_get_voltage_level(&context->bw_ctx.dml, pipes, *pipe_cnt); + /* This may adjust vlevel and maxMpcComb */ + if (*vlevel < context->bw_ctx.dml.soc.num_states) + *vlevel = dml_validate_apply_pipe_split_flags(dc, context, *vlevel, split, merge); + + /* Conditions for setting up phantom pipes for SubVP: + * 1. Not force disable SubVP + * 2. Full update (i.e. !fast_validate) + * 3. Enough pipes are available to support SubVP (TODO: Which pipes will use VACTIVE / VBLANK / SUBVP?) + * 4. Display configuration passes validation + * 5. (Config doesn't support MCLK in VACTIVE/VBLANK || dc->debug.force_subvp_mclk_switch) + */ + if (!dc->debug.force_disable_subvp && + dml_enough_pipes_for_subvp(dc, context) && + *vlevel < context->bw_ctx.dml.soc.num_states && + (vba->DRAMClockChangeSupport[*vlevel][vba->maxMpcComb] == dm_dram_clock_change_unsupported || + dc->debug.force_subvp_mclk_switch)) { + + dml_add_phantom_pipes(dc, context); + + /* Create input to DML based on new context which includes phantom pipes + * TODO: Input to DML should mark which pipes are phantom + */ + *pipe_cnt = dml_populate_dml_pipes_from_context(dc, context, pipes, false); + *vlevel = dml_get_voltage_level(&context->bw_ctx.dml, pipes, *pipe_cnt); + if (*vlevel < context->bw_ctx.dml.soc.num_states) { + memset(split, 0, sizeof(split)); + memset(merge, 0, sizeof(merge)); + *vlevel = dml_validate_apply_pipe_split_flags(dc, context, *vlevel, split, merge); + } + + // If SubVP pipe config is unsupported (or cannot be used for UCLK switching) + // remove phantom pipes and repopulate dml pipes + if (*vlevel == context->bw_ctx.dml.soc.num_states || + vba->DRAMClockChangeSupport[*vlevel][vba->maxMpcComb] == dm_dram_clock_change_unsupported) { + dml_remove_phantom_pipes(dc, context); + *pipe_cnt = dml_populate_dml_pipes_from_context(dc, context, pipes, false); + } + } +} + +static void dcn20_adjust_adaptive_sync_v_startup( + const struct dc_crtc_timing *dc_crtc_timing, int *vstartup_start) +{ + struct dc_crtc_timing patched_crtc_timing; + uint32_t asic_blank_end = 0; + uint32_t asic_blank_start = 0; + uint32_t newVstartup = 0; + + patched_crtc_timing = *dc_crtc_timing; + + if (patched_crtc_timing.flags.INTERLACE == 1) { + if (patched_crtc_timing.v_front_porch < 2) + patched_crtc_timing.v_front_porch = 2; + } else { + if (patched_crtc_timing.v_front_porch < 1) + patched_crtc_timing.v_front_porch = 1; + } + + /* blank_start = frame end - front porch */ + asic_blank_start = patched_crtc_timing.v_total - + patched_crtc_timing.v_front_porch; + + /* blank_end = blank_start - active */ + asic_blank_end = asic_blank_start - + patched_crtc_timing.v_border_bottom - + patched_crtc_timing.v_addressable - + patched_crtc_timing.v_border_top; + + newVstartup = asic_blank_end + (patched_crtc_timing.v_total - asic_blank_start); + + *vstartup_start = ((newVstartup > *vstartup_start) ? newVstartup : *vstartup_start); +} + +static bool is_dp_128b_132b_signal(struct pipe_ctx *pipe_ctx) +{ + return (pipe_ctx->stream_res.hpo_dp_stream_enc && + pipe_ctx->stream->link->hpo_dp_link_enc && + dc_is_dp_signal(pipe_ctx->stream->signal)); +} + +static bool is_dtbclk_required(struct dc *dc, struct dc_state *context) +{ + int i; + for (i = 0; i < dc->res_pool->pipe_count; i++) { + if (!context->res_ctx.pipe_ctx[i].stream) + continue; +#if defined (CONFIG_DRM_AMD_DC_DP2_0) + if (is_dp_128b_132b_signal(&context->res_ctx.pipe_ctx[i])) + return true; +#endif + } + return false; +} + +static void dml_update_soc_for_wm_a(struct dc *dc, struct dc_state *context) +{ + if (dc->clk_mgr->bw_params->wm_table.nv_entries[WM_A].valid) { + context->bw_ctx.dml.soc.dram_clock_change_latency_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_A].dml_input.pstate_latency_us; + context->bw_ctx.dml.soc.sr_enter_plus_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_A].dml_input.sr_enter_plus_exit_time_us; + context->bw_ctx.dml.soc.sr_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_A].dml_input.sr_exit_time_us; + } +} + +static bool dml_internal_validate( + struct dc *dc, + struct dc_state *context, + display_e2e_pipe_params_st *pipes, + int *pipe_cnt_out, + int *vlevel_out, + bool fast_validate) +{ + bool out = false; + bool repopulate_pipes = false; + int split[MAX_PIPES] = { 0 }; + bool merge[MAX_PIPES] = { false }; + bool newly_split[MAX_PIPES] = { false }; + int pipe_cnt, i, pipe_idx, vlevel; + struct vba_vars_st *vba = &context->bw_ctx.dml.vba; + + ASSERT(pipes); + if (!pipes) + return false; + + // For each full update, remove all existing phantom pipes first + dml_remove_phantom_pipes(dc, context); + + dml_update_soc_for_wm_a(dc, context); + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + + if (pipe->plane_state) { + // On initial pass through DML, we intend to use MALL for SS on all + // (non-PSR) surfaces with none using MALL for P-State + // 'mall_plane_config': is not a member of 'dc_plane_state' - commenting it out till mall_plane_config gets supported in dc_plant_state + //if (pipe->stream && pipe->stream->link->psr_settings.psr_version == DC_PSR_VERSION_UNSUPPORTED) + // pipe->plane_state->mall_plane_config.use_mall_for_ss = true; + } + } + pipe_cnt = dml_populate_dml_pipes_from_context(dc, context, pipes, fast_validate); + + if (!pipe_cnt) { + out = true; + goto validate_out; + } + + dml_log_pipe_params(&context->bw_ctx.dml, pipes, pipe_cnt); + + if (!fast_validate) { + dml_full_validate_bw_helper(dc, context, pipes, &vlevel, split, merge, &pipe_cnt); + } + + if (fast_validate || vlevel == context->bw_ctx.dml.soc.num_states || + vba->DRAMClockChangeSupport[vlevel][vba->maxMpcComb] == dm_dram_clock_change_unsupported) { + /* + * If mode is unsupported or there's still no p-state support then + * fall back to favoring voltage. + * + * We don't actually support prefetch mode 2, so require that we + * at least support prefetch mode 1. + */ + context->bw_ctx.dml.soc.allow_dram_self_refresh_or_dram_clock_change_in_vblank = + dm_allow_self_refresh; + + vlevel = dml_get_voltage_level(&context->bw_ctx.dml, pipes, pipe_cnt); + if (vlevel < context->bw_ctx.dml.soc.num_states) { + memset(split, 0, sizeof(split)); + memset(merge, 0, sizeof(merge)); + vlevel = dml_validate_apply_pipe_split_flags(dc, context, vlevel, split, merge); + } + } + + dml_log_mode_support_params(&context->bw_ctx.dml); + + if (vlevel == context->bw_ctx.dml.soc.num_states) + goto validate_fail; + + for (i = 0, pipe_idx = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + struct pipe_ctx *mpo_pipe = pipe->bottom_pipe; + + if (!pipe->stream) + continue; + + /* We only support full screen mpo with ODM */ + if (vba->ODMCombineEnabled[vba->pipe_plane[pipe_idx]] != dm_odm_combine_mode_disabled + && pipe->plane_state && mpo_pipe + && memcmp(&mpo_pipe->plane_res.scl_data.recout, + &pipe->plane_res.scl_data.recout, + sizeof(struct rect)) != 0) { + ASSERT(mpo_pipe->plane_state != pipe->plane_state); + goto validate_fail; + } + pipe_idx++; + } + + /* merge pipes if necessary */ + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + + /*skip pipes that don't need merging*/ + if (!merge[i]) + continue; + + /* if ODM merge we ignore mpc tree, mpo pipes will have their own flags */ + if (pipe->prev_odm_pipe) { + /*split off odm pipe*/ + pipe->prev_odm_pipe->next_odm_pipe = pipe->next_odm_pipe; + if (pipe->next_odm_pipe) + pipe->next_odm_pipe->prev_odm_pipe = pipe->prev_odm_pipe; + + pipe->bottom_pipe = NULL; + pipe->next_odm_pipe = NULL; + pipe->plane_state = NULL; + pipe->stream = NULL; + pipe->top_pipe = NULL; + pipe->prev_odm_pipe = NULL; + if (pipe->stream_res.dsc) + dml_release_dsc(&context->res_ctx, dc->res_pool, &pipe->stream_res.dsc); + memset(&pipe->plane_res, 0, sizeof(pipe->plane_res)); + memset(&pipe->stream_res, 0, sizeof(pipe->stream_res)); + repopulate_pipes = true; + } else if (pipe->top_pipe && pipe->top_pipe->plane_state == pipe->plane_state) { + struct pipe_ctx *top_pipe = pipe->top_pipe; + struct pipe_ctx *bottom_pipe = pipe->bottom_pipe; + + top_pipe->bottom_pipe = bottom_pipe; + if (bottom_pipe) + bottom_pipe->top_pipe = top_pipe; + + pipe->top_pipe = NULL; + pipe->bottom_pipe = NULL; + pipe->plane_state = NULL; + pipe->stream = NULL; + memset(&pipe->plane_res, 0, sizeof(pipe->plane_res)); + memset(&pipe->stream_res, 0, sizeof(pipe->stream_res)); + repopulate_pipes = true; + } else + ASSERT(0); /* Should never try to merge master pipe */ + + } + + for (i = 0, pipe_idx = -1; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + struct pipe_ctx *old_pipe = &dc->current_state->res_ctx.pipe_ctx[i]; + struct pipe_ctx *hsplit_pipe = NULL; + bool odm; + int old_index = -1; + + if (!pipe->stream || newly_split[i]) + continue; + + pipe_idx++; + odm = vba->ODMCombineEnabled[vba->pipe_plane[pipe_idx]] != dm_odm_combine_mode_disabled; + + if (!pipe->plane_state && !odm) + continue; + + if (split[i]) { + if (odm) { + if (split[i] == 4 && old_pipe->next_odm_pipe && old_pipe->next_odm_pipe->next_odm_pipe) + old_index = old_pipe->next_odm_pipe->next_odm_pipe->pipe_idx; + else if (old_pipe->next_odm_pipe) + old_index = old_pipe->next_odm_pipe->pipe_idx; + } else { + if (split[i] == 4 && old_pipe->bottom_pipe && old_pipe->bottom_pipe->bottom_pipe && + old_pipe->bottom_pipe->bottom_pipe->plane_state == old_pipe->plane_state) + old_index = old_pipe->bottom_pipe->bottom_pipe->pipe_idx; + else if (old_pipe->bottom_pipe && + old_pipe->bottom_pipe->plane_state == old_pipe->plane_state) + old_index = old_pipe->bottom_pipe->pipe_idx; + } + hsplit_pipe = dml_find_split_pipe(dc, context, old_index); + ASSERT(hsplit_pipe); + if (!hsplit_pipe) + goto validate_fail; + + if (!dml_split_stream_for_mpc_or_odm( + dc, &context->res_ctx, + pipe, hsplit_pipe, odm)) + goto validate_fail; + + newly_split[hsplit_pipe->pipe_idx] = true; + repopulate_pipes = true; + } + if (split[i] == 4) { + struct pipe_ctx *pipe_4to1; + + if (odm && old_pipe->next_odm_pipe) + old_index = old_pipe->next_odm_pipe->pipe_idx; + else if (!odm && old_pipe->bottom_pipe && + old_pipe->bottom_pipe->plane_state == old_pipe->plane_state) + old_index = old_pipe->bottom_pipe->pipe_idx; + else + old_index = -1; + pipe_4to1 = dml_find_split_pipe(dc, context, old_index); + ASSERT(pipe_4to1); + if (!pipe_4to1) + goto validate_fail; + if (!dml_split_stream_for_mpc_or_odm( + dc, &context->res_ctx, + pipe, pipe_4to1, odm)) + goto validate_fail; + newly_split[pipe_4to1->pipe_idx] = true; + + if (odm && old_pipe->next_odm_pipe && old_pipe->next_odm_pipe->next_odm_pipe + && old_pipe->next_odm_pipe->next_odm_pipe->next_odm_pipe) + old_index = old_pipe->next_odm_pipe->next_odm_pipe->next_odm_pipe->pipe_idx; + else if (!odm && old_pipe->bottom_pipe && old_pipe->bottom_pipe->bottom_pipe && + old_pipe->bottom_pipe->bottom_pipe->bottom_pipe && + old_pipe->bottom_pipe->bottom_pipe->bottom_pipe->plane_state == old_pipe->plane_state) + old_index = old_pipe->bottom_pipe->bottom_pipe->bottom_pipe->pipe_idx; + else + old_index = -1; + pipe_4to1 = dml_find_split_pipe(dc, context, old_index); + ASSERT(pipe_4to1); + if (!pipe_4to1) + goto validate_fail; + if (!dml_split_stream_for_mpc_or_odm( + dc, &context->res_ctx, + hsplit_pipe, pipe_4to1, odm)) + goto validate_fail; + newly_split[pipe_4to1->pipe_idx] = true; + } + if (odm) + dml_build_mapped_resource(dc, context, pipe->stream); + } + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; + + if (pipe->plane_state) { + if (!resource_build_scaling_params(pipe)) + goto validate_fail; + } + } + + /* Actual dsc count per stream dsc validation*/ + if (!dml_validate_dsc(dc, context)) { + vba->ValidationStatus[vba->soc.num_states] = DML_FAIL_DSC_VALIDATION_FAILURE; + goto validate_fail; + } + + if (repopulate_pipes) + pipe_cnt = dml_populate_dml_pipes_from_context(dc, context, pipes, fast_validate); + *vlevel_out = vlevel; + *pipe_cnt_out = pipe_cnt; + + out = true; + goto validate_out; + +validate_fail: + out = false; + +validate_out: + return out; +} + +static void dml_calculate_dlg_params( + struct dc *dc, struct dc_state *context, + display_e2e_pipe_params_st *pipes, + int pipe_cnt, + int vlevel) +{ + int i, pipe_idx; + int plane_count; + + /* Writeback MCIF_WB arbitration parameters */ + if (dc->res_pool) + dc->res_pool->funcs->set_mcif_arb_params(dc, context, pipes, pipe_cnt); + + context->bw_ctx.bw.dcn.clk.dispclk_khz = context->bw_ctx.dml.vba.DISPCLK * 1000; + context->bw_ctx.bw.dcn.clk.dcfclk_khz = context->bw_ctx.dml.vba.DCFCLK * 1000; + context->bw_ctx.bw.dcn.clk.socclk_khz = context->bw_ctx.dml.vba.SOCCLK * 1000; + context->bw_ctx.bw.dcn.clk.dramclk_khz = context->bw_ctx.dml.vba.DRAMSpeed * 1000 / 16; + context->bw_ctx.bw.dcn.clk.dcfclk_deep_sleep_khz = context->bw_ctx.dml.vba.DCFCLKDeepSleep * 1000; + context->bw_ctx.bw.dcn.clk.fclk_khz = context->bw_ctx.dml.vba.FabricClock * 1000; + context->bw_ctx.bw.dcn.clk.p_state_change_support = + context->bw_ctx.dml.vba.DRAMClockChangeSupport[vlevel][context->bw_ctx.dml.vba.maxMpcComb] + != dm_dram_clock_change_unsupported; + + context->bw_ctx.bw.dcn.clk.dppclk_khz = 0; + /* 'z9_support': is not a member of 'dc_clocks' - Commenting out till we have this support in dc_clocks + * context->bw_ctx.bw.dcn.clk.z9_support = (context->bw_ctx.dml.vba.StutterPeriod > 5000.0) ? + DCN_Z9_SUPPORT_ALLOW : DCN_Z9_SUPPORT_DISALLOW; + */ + plane_count = 0; + for (i = 0; i < dc->res_pool->pipe_count; i++) { + if (context->res_ctx.pipe_ctx[i].plane_state) + plane_count++; + } + + /* Commented out as per above error for now. + if (plane_count == 0) + context->bw_ctx.bw.dcn.clk.z9_support = DCN_Z9_SUPPORT_ALLOW; + */ + context->bw_ctx.bw.dcn.clk.dtbclk_en = is_dtbclk_required(dc, context); + /* TODO : Uncomment the below line and make changes + * as per DML nomenclature once it is available. + * context->bw_ctx.bw.dcn.clk.fclk_p_state_change_support = context->bw_ctx.dml.vba.fclk_pstate_support; + */ + + if (context->bw_ctx.bw.dcn.clk.dispclk_khz < dc->debug.min_disp_clk_khz) + context->bw_ctx.bw.dcn.clk.dispclk_khz = dc->debug.min_disp_clk_khz; + + for (i = 0, pipe_idx = 0; i < dc->res_pool->pipe_count; i++) { + if (!context->res_ctx.pipe_ctx[i].stream) + continue; + pipes[pipe_idx].pipe.dest.vstartup_start = get_vstartup(&context->bw_ctx.dml, pipes, pipe_cnt, pipe_idx); + pipes[pipe_idx].pipe.dest.vupdate_offset = get_vupdate_offset(&context->bw_ctx.dml, pipes, pipe_cnt, pipe_idx); + pipes[pipe_idx].pipe.dest.vupdate_width = get_vupdate_width(&context->bw_ctx.dml, pipes, pipe_cnt, pipe_idx); + pipes[pipe_idx].pipe.dest.vready_offset = get_vready_offset(&context->bw_ctx.dml, pipes, pipe_cnt, pipe_idx); + if (context->res_ctx.pipe_ctx[i].stream->mall_stream_config.type == SUBVP_PHANTOM) { + // Phantom pipe requires that DET_SIZE = 0 and no unbounded requests + context->res_ctx.pipe_ctx[i].det_buffer_size_kb = 0; + context->res_ctx.pipe_ctx[i].unbounded_req = false; + } else { + context->res_ctx.pipe_ctx[i].det_buffer_size_kb = context->bw_ctx.dml.ip.det_buffer_size_kbytes; + context->res_ctx.pipe_ctx[i].unbounded_req = pipes[pipe_idx].pipe.src.unbounded_req_mode; + } + + if (context->bw_ctx.bw.dcn.clk.dppclk_khz < pipes[pipe_idx].clks_cfg.dppclk_mhz * 1000) + context->bw_ctx.bw.dcn.clk.dppclk_khz = pipes[pipe_idx].clks_cfg.dppclk_mhz * 1000; + context->res_ctx.pipe_ctx[i].plane_res.bw.dppclk_khz = + pipes[pipe_idx].clks_cfg.dppclk_mhz * 1000; + context->res_ctx.pipe_ctx[i].pipe_dlg_param = pipes[pipe_idx].pipe.dest; + pipe_idx++; + } + /*save a original dppclock copy*/ + context->bw_ctx.bw.dcn.clk.bw_dppclk_khz = context->bw_ctx.bw.dcn.clk.dppclk_khz; + context->bw_ctx.bw.dcn.clk.bw_dispclk_khz = context->bw_ctx.bw.dcn.clk.dispclk_khz; + context->bw_ctx.bw.dcn.clk.max_supported_dppclk_khz = context->bw_ctx.dml.soc.clock_limits[vlevel].dppclk_mhz * 1000; + context->bw_ctx.bw.dcn.clk.max_supported_dispclk_khz = context->bw_ctx.dml.soc.clock_limits[vlevel].dispclk_mhz * 1000; + context->bw_ctx.bw.dcn.compbuf_size_kb = context->bw_ctx.dml.ip.config_return_buffer_size_in_kbytes + - context->bw_ctx.dml.ip.det_buffer_size_kbytes * pipe_idx; + + for (i = 0, pipe_idx = 0; i < dc->res_pool->pipe_count; i++) { + bool cstate_en = context->bw_ctx.dml.vba.PrefetchMode[vlevel][context->bw_ctx.dml.vba.maxMpcComb] != 2; + + if (!context->res_ctx.pipe_ctx[i].stream) + continue; + + context->bw_ctx.dml.funcs.rq_dlg_get_dlg_reg(&context->bw_ctx.dml, + &context->res_ctx.pipe_ctx[i].dlg_regs, + &context->res_ctx.pipe_ctx[i].ttu_regs, + pipes, + pipe_cnt, + pipe_idx, + cstate_en, + context->bw_ctx.bw.dcn.clk.p_state_change_support, + false, false, true); + + context->bw_ctx.dml.funcs.rq_dlg_get_rq_reg(&context->bw_ctx.dml, + &context->res_ctx.pipe_ctx[i].rq_regs, + &pipes[pipe_idx].pipe); + pipe_idx++; + } +} + +static void dml_calculate_wm_and_dlg( + struct dc *dc, struct dc_state *context, + display_e2e_pipe_params_st *pipes, + int pipe_cnt, + int vlevel) +{ + int i, pipe_idx, vlevel_temp = 0; + + double dcfclk = context->bw_ctx.dml.soc.clock_limits[0].dcfclk_mhz; + double dcfclk_from_validation = context->bw_ctx.dml.vba.DCFCLKState[vlevel][context->bw_ctx.dml.vba.maxMpcComb]; + unsigned int min_dram_speed_mts = context->bw_ctx.dml.vba.DRAMSpeed; + bool pstate_en = context->bw_ctx.dml.vba.DRAMClockChangeSupport[vlevel][context->bw_ctx.dml.vba.maxMpcComb] != + dm_dram_clock_change_unsupported; + + /* Set B: + * For Set B calculations use clocks from clock_limits[2] when available i.e. when SMU is present, + * otherwise use arbitrary low value from spreadsheet for DCFCLK as lower is safer for watermark + * calculations to cover bootup clocks. + * DCFCLK: soc.clock_limits[2] when available + * UCLK: soc.clock_limits[2] when available + */ + if (context->bw_ctx.dml.soc.num_states > 2) { + vlevel_temp = 2; + dcfclk = context->bw_ctx.dml.soc.clock_limits[2].dcfclk_mhz; + } else + dcfclk = 615; //DCFCLK Vmin_lv + + pipes[0].clks_cfg.voltage = vlevel_temp; + pipes[0].clks_cfg.dcfclk_mhz = dcfclk; + pipes[0].clks_cfg.socclk_mhz = context->bw_ctx.dml.soc.clock_limits[vlevel_temp].socclk_mhz; + + if (dc->clk_mgr->bw_params->wm_table.nv_entries[WM_B].valid) { + context->bw_ctx.dml.soc.dram_clock_change_latency_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_B].dml_input.pstate_latency_us; + context->bw_ctx.dml.soc.sr_enter_plus_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_B].dml_input.sr_enter_plus_exit_time_us; + context->bw_ctx.dml.soc.sr_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_B].dml_input.sr_exit_time_us; + } + context->bw_ctx.bw.dcn.watermarks.b.urgent_ns = get_wm_urgent(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.cstate_enter_plus_exit_ns = get_wm_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.cstate_exit_ns = get_wm_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.pstate_change_ns = get_wm_dram_clock_change(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.b.pte_meta_urgent_ns = get_wm_memory_trip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.b.frac_urg_bw_nom = get_fraction_of_urgent_bandwidth(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.b.frac_urg_bw_flip = get_fraction_of_urgent_bandwidth_imm_flip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.b.urgent_latency_ns = get_urgent_latency(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + //context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.fclk_pstate_change_ns = get_wm_fclk_pstate(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + //context->bw_ctx.bw.dcn.watermarks.b.usr_retraining_ns = get_wm_usr_retraining(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + + /* Temporary, to have some fclk_pstate_change_ns and usr_retraining_ns wm values until DML is implemented */ + context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.fclk_pstate_change_ns = context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.pstate_change_ns / 4; + context->bw_ctx.bw.dcn.watermarks.b.usr_retraining_ns = context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.pstate_change_ns / 8; + + /* Set D: + * All clocks min. + * DCFCLK: Min, as reported by PM FW when available + * UCLK : Min, as reported by PM FW when available + * sr_enter_exit/sr_exit should be lower than used for DRAM (TBD after bringup or later, use as decided in Clk Mgr) + */ + + if (context->bw_ctx.dml.soc.num_states > 2) { + vlevel_temp = 0; + dcfclk = dc->clk_mgr->bw_params->clk_table.entries[0].dcfclk_mhz; + } else + dcfclk = 615; //DCFCLK Vmin_lv + + pipes[0].clks_cfg.voltage = vlevel_temp; + pipes[0].clks_cfg.dcfclk_mhz = dcfclk; + pipes[0].clks_cfg.socclk_mhz = context->bw_ctx.dml.soc.clock_limits[vlevel_temp].socclk_mhz; + + if (dc->clk_mgr->bw_params->wm_table.nv_entries[WM_D].valid) { + context->bw_ctx.dml.soc.dram_clock_change_latency_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_D].dml_input.pstate_latency_us; + context->bw_ctx.dml.soc.sr_enter_plus_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_D].dml_input.sr_enter_plus_exit_time_us; + context->bw_ctx.dml.soc.sr_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_D].dml_input.sr_exit_time_us; + } + context->bw_ctx.bw.dcn.watermarks.d.urgent_ns = get_wm_urgent(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.cstate_enter_plus_exit_ns = get_wm_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.cstate_exit_ns = get_wm_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.pstate_change_ns = get_wm_dram_clock_change(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.d.pte_meta_urgent_ns = get_wm_memory_trip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.d.frac_urg_bw_nom = get_fraction_of_urgent_bandwidth(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.d.frac_urg_bw_flip = get_fraction_of_urgent_bandwidth_imm_flip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.d.urgent_latency_ns = get_urgent_latency(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + //context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.fclk_pstate_change_ns = get_wm_fclk_pstate(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + //context->bw_ctx.bw.dcn.watermarks.d.usr_retraining_ns = get_wm_usr_retraining(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + + /* Temporary, to have some fclk_pstate_change_ns and usr_retraining_ns wm values until DML is implemented */ + context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.fclk_pstate_change_ns = context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.pstate_change_ns / 4; + context->bw_ctx.bw.dcn.watermarks.d.usr_retraining_ns = context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.pstate_change_ns / 8; + + /* Set C, for Dummy P-State: + * All clocks min. + * DCFCLK: Min, as reported by PM FW, when available + * UCLK : Min, as reported by PM FW, when available + * pstate latency as per UCLK state dummy pstate latency + */ + if (dc->clk_mgr->bw_params->wm_table.nv_entries[WM_C].valid) { + unsigned int min_dram_speed_mts_margin = 160; + + if ((!pstate_en)) + min_dram_speed_mts = dc->clk_mgr->bw_params->clk_table.entries[dc->clk_mgr->bw_params->clk_table.num_entries - 1].memclk_mhz * 16; + + /* find largest table entry that is lower than dram speed, but lower than DPM0 still uses DPM0 */ + for (i = 3; i > 0; i--) + if (min_dram_speed_mts + min_dram_speed_mts_margin > dc->clk_mgr->bw_params->dummy_pstate_table[i].dram_speed_mts) + break; + + context->bw_ctx.dml.soc.dram_clock_change_latency_us = dc->clk_mgr->bw_params->dummy_pstate_table[i].dummy_pstate_latency_us; + context->bw_ctx.dml.soc.dummy_pstate_latency_us = dc->clk_mgr->bw_params->dummy_pstate_table[i].dummy_pstate_latency_us; + context->bw_ctx.dml.soc.sr_enter_plus_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_C].dml_input.sr_enter_plus_exit_time_us; + context->bw_ctx.dml.soc.sr_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_C].dml_input.sr_exit_time_us; + } + context->bw_ctx.bw.dcn.watermarks.c.urgent_ns = get_wm_urgent(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.cstate_enter_plus_exit_ns = get_wm_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.cstate_exit_ns = get_wm_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.pstate_change_ns = get_wm_dram_clock_change(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.c.pte_meta_urgent_ns = get_wm_memory_trip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.c.frac_urg_bw_nom = get_fraction_of_urgent_bandwidth(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.c.frac_urg_bw_flip = get_fraction_of_urgent_bandwidth_imm_flip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.c.urgent_latency_ns = get_urgent_latency(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + //context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.fclk_pstate_change_ns = get_wm_fclk_pstate(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + //context->bw_ctx.bw.dcn.watermarks.c.usr_retraining_ns = get_wm_usr_retraining(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + + /* Temporary, to have some fclk_pstate_change_ns and usr_retraining_ns wm values until DML is implemented */ + context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.fclk_pstate_change_ns = context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.pstate_change_ns / 4; + context->bw_ctx.bw.dcn.watermarks.c.usr_retraining_ns = context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.pstate_change_ns / 8; + + if ((!pstate_en) && (dc->clk_mgr->bw_params->wm_table.nv_entries[WM_C].valid)) { + /* The only difference between A and C is p-state latency, if p-state is not supported + * with full p-state latency we want to calculate DLG based on dummy p-state latency, + * Set A p-state watermark set to 0 previously, when p-state unsupported, for now keep as previous implementation. + */ + context->bw_ctx.bw.dcn.watermarks.a = context->bw_ctx.bw.dcn.watermarks.c; + context->bw_ctx.bw.dcn.watermarks.a.cstate_pstate.pstate_change_ns = 0; + } else { + /* Set A: + * All clocks min. + * DCFCLK: Min, as reported by PM FW, when available + * UCLK: Min, as reported by PM FW, when available + */ + dml_update_soc_for_wm_a(dc, context); + context->bw_ctx.bw.dcn.watermarks.a.urgent_ns = get_wm_urgent(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.a.cstate_pstate.cstate_enter_plus_exit_ns = get_wm_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.a.cstate_pstate.cstate_exit_ns = get_wm_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.a.cstate_pstate.pstate_change_ns = get_wm_dram_clock_change(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.a.pte_meta_urgent_ns = get_wm_memory_trip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.a.frac_urg_bw_nom = get_fraction_of_urgent_bandwidth(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.a.frac_urg_bw_flip = get_fraction_of_urgent_bandwidth_imm_flip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + context->bw_ctx.bw.dcn.watermarks.a.urgent_latency_ns = get_urgent_latency(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000; + } + + pipes[0].clks_cfg.voltage = vlevel; + pipes[0].clks_cfg.dcfclk_mhz = dcfclk_from_validation; + pipes[0].clks_cfg.socclk_mhz = context->bw_ctx.dml.soc.clock_limits[vlevel].socclk_mhz; + + for (i = 0, pipe_idx = 0; i < dc->res_pool->pipe_count; i++) { + if (!context->res_ctx.pipe_ctx[i].stream) + continue; + + pipes[pipe_idx].clks_cfg.dispclk_mhz = get_dispclk_calculated(&context->bw_ctx.dml, pipes, pipe_cnt); + pipes[pipe_idx].clks_cfg.dppclk_mhz = get_dppclk_calculated(&context->bw_ctx.dml, pipes, pipe_cnt, pipe_idx); + + if (dc->config.forced_clocks) { + pipes[pipe_idx].clks_cfg.dispclk_mhz = context->bw_ctx.dml.soc.clock_limits[0].dispclk_mhz; + pipes[pipe_idx].clks_cfg.dppclk_mhz = context->bw_ctx.dml.soc.clock_limits[0].dppclk_mhz; + } + if (dc->debug.min_disp_clk_khz > pipes[pipe_idx].clks_cfg.dispclk_mhz * 1000) + pipes[pipe_idx].clks_cfg.dispclk_mhz = dc->debug.min_disp_clk_khz / 1000.0; + if (dc->debug.min_dpp_clk_khz > pipes[pipe_idx].clks_cfg.dppclk_mhz * 1000) + pipes[pipe_idx].clks_cfg.dppclk_mhz = dc->debug.min_dpp_clk_khz / 1000.0; + + pipe_idx++; + } + + context->perf_params.stutter_period_us = context->bw_ctx.dml.vba.StutterPeriod; + + dml_calculate_dlg_params(dc, context, pipes, pipe_cnt, vlevel); + + if (!pstate_en) + /* Restore full p-state latency */ + context->bw_ctx.dml.soc.dram_clock_change_latency_us = + dc->clk_mgr->bw_params->wm_table.nv_entries[WM_A].dml_input.pstate_latency_us; +} + +bool dml_validate(struct dc *dc, + struct dc_state *context, + bool fast_validate) +{ + bool out = false; + + BW_VAL_TRACE_SETUP(); + + int vlevel = 0; + int pipe_cnt = 0; + display_e2e_pipe_params_st *pipes = context->bw_ctx.dml.dml_pipe_state; + DC_LOGGER_INIT(dc->ctx->logger); + + BW_VAL_TRACE_COUNT(); + + out = dml_internal_validate(dc, context, pipes, &pipe_cnt, &vlevel, fast_validate); + + if (pipe_cnt == 0) + goto validate_out; + + if (!out) + goto validate_fail; + + BW_VAL_TRACE_END_VOLTAGE_LEVEL(); + + if (fast_validate) { + BW_VAL_TRACE_SKIP(fast); + goto validate_out; + } + + dml_calculate_wm_and_dlg(dc, context, pipes, pipe_cnt, vlevel); + + BW_VAL_TRACE_END_WATERMARKS(); + + goto validate_out; + +validate_fail: + DC_LOG_WARNING("Mode Validation Warning: %s failed validation.\n", + dml_get_status_message(context->bw_ctx.dml.vba.ValidationStatus[context->bw_ctx.dml.vba.soc.num_states])); + + BW_VAL_TRACE_SKIP(fail); + out = false; + +validate_out: + BW_VAL_TRACE_FINISH(); + + return out; +} diff --git a/drivers/gpu/drm/amd/display/dc/dml/dml_wrapper_translation.c b/drivers/gpu/drm/amd/display/dc/dml/dml_wrapper_translation.c new file mode 100644 index 000000000000..4ec5310a2962 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dml/dml_wrapper_translation.c @@ -0,0 +1,284 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: AMD + * + */ + +#ifdef DML_WRAPPER_TRANSLATION_ + +static void gfx10array_mode_to_dml_params( + enum array_mode_values array_mode, + enum legacy_tiling_compat_level compat_level, + unsigned int *sw_mode) +{ + switch (array_mode) { + case DC_ARRAY_LINEAR_ALLIGNED: + case DC_ARRAY_LINEAR_GENERAL: + *sw_mode = dm_sw_linear; + break; + case DC_ARRAY_2D_TILED_THIN1: +// DC_LEGACY_TILING_ADDR_GEN_ZERO - undefined as per current code hence removed +#if 0 + if (compat_level == DC_LEGACY_TILING_ADDR_GEN_ZERO) + *sw_mode = dm_sw_gfx7_2d_thin_l_vp; + else + *sw_mode = dm_sw_gfx7_2d_thin_gl; +#endif + break; + default: + ASSERT(0); /* Not supported */ + break; + } +} + +static void swizzle_to_dml_params( + enum swizzle_mode_values swizzle, + unsigned int *sw_mode) +{ + switch (swizzle) { + case DC_SW_LINEAR: + *sw_mode = dm_sw_linear; + break; + case DC_SW_4KB_S: + *sw_mode = dm_sw_4kb_s; + break; + case DC_SW_4KB_S_X: + *sw_mode = dm_sw_4kb_s_x; + break; + case DC_SW_4KB_D: + *sw_mode = dm_sw_4kb_d; + break; + case DC_SW_4KB_D_X: + *sw_mode = dm_sw_4kb_d_x; + break; + case DC_SW_64KB_S: + *sw_mode = dm_sw_64kb_s; + break; + case DC_SW_64KB_S_X: + *sw_mode = dm_sw_64kb_s_x; + break; + case DC_SW_64KB_S_T: + *sw_mode = dm_sw_64kb_s_t; + break; + case DC_SW_64KB_D: + *sw_mode = dm_sw_64kb_d; + break; + case DC_SW_64KB_D_X: + *sw_mode = dm_sw_64kb_d_x; + break; + case DC_SW_64KB_D_T: + *sw_mode = dm_sw_64kb_d_t; + break; + case DC_SW_64KB_R_X: + *sw_mode = dm_sw_64kb_r_x; + break; + case DC_SW_VAR_S: + *sw_mode = dm_sw_var_s; + break; + case DC_SW_VAR_S_X: + *sw_mode = dm_sw_var_s_x; + break; + case DC_SW_VAR_D: + *sw_mode = dm_sw_var_d; + break; + case DC_SW_VAR_D_X: + *sw_mode = dm_sw_var_d_x; + break; + + default: + ASSERT(0); /* Not supported */ + break; + } +} + +static void dc_timing_to_dml_timing(const struct dc_crtc_timing *timing, struct _vcs_dpi_display_pipe_dest_params_st *dest) +{ + dest->hblank_start = timing->h_total - timing->h_front_porch; + dest->hblank_end = dest->hblank_start + - timing->h_addressable + - timing->h_border_left + - timing->h_border_right; + dest->vblank_start = timing->v_total - timing->v_front_porch; + dest->vblank_end = dest->vblank_start + - timing->v_addressable + - timing->v_border_top + - timing->v_border_bottom; + dest->htotal = timing->h_total; + dest->vtotal = timing->v_total; + dest->hactive = timing->h_addressable; + dest->vactive = timing->v_addressable; + dest->interlaced = timing->flags.INTERLACE; + dest->pixel_rate_mhz = timing->pix_clk_100hz/10000.0; + if (timing->timing_3d_format == TIMING_3D_FORMAT_HW_FRAME_PACKING) + dest->pixel_rate_mhz *= 2; +} + +static enum odm_combine_mode get_dml_odm_combine(const struct pipe_ctx *pipe) +{ + int odm_split_count = 0; + enum odm_combine_mode combine_mode = dm_odm_combine_mode_disabled; + struct pipe_ctx *next_pipe = pipe->next_odm_pipe; + + // Traverse pipe tree to determine odm split count + while (next_pipe) { + odm_split_count++; + next_pipe = next_pipe->next_odm_pipe; + } + pipe = pipe->prev_odm_pipe; + while (pipe) { + odm_split_count++; + pipe = pipe->prev_odm_pipe; + } + + // Translate split to DML odm combine factor + switch (odm_split_count) { + case 1: + combine_mode = dm_odm_combine_mode_2to1; + break; + case 3: + combine_mode = dm_odm_combine_mode_4to1; + break; + default: + combine_mode = dm_odm_combine_mode_disabled; + } + + return combine_mode; +} + +static int get_dml_output_type(enum signal_type dc_signal) +{ + int dml_output_type = -1; + + switch (dc_signal) { + case SIGNAL_TYPE_DISPLAY_PORT_MST: + case SIGNAL_TYPE_DISPLAY_PORT: + dml_output_type = dm_dp; + break; + case SIGNAL_TYPE_EDP: + dml_output_type = dm_edp; + break; + case SIGNAL_TYPE_HDMI_TYPE_A: + case SIGNAL_TYPE_DVI_SINGLE_LINK: + case SIGNAL_TYPE_DVI_DUAL_LINK: + dml_output_type = dm_hdmi; + break; + default: + break; + } + + return dml_output_type; +} + +static void populate_color_depth_and_encoding_from_timing(const struct dc_crtc_timing *timing, struct _vcs_dpi_display_output_params_st *dout) +{ + int output_bpc = 0; + + switch (timing->display_color_depth) { + case COLOR_DEPTH_666: + output_bpc = 6; + break; + case COLOR_DEPTH_888: + output_bpc = 8; + break; + case COLOR_DEPTH_101010: + output_bpc = 10; + break; + case COLOR_DEPTH_121212: + output_bpc = 12; + break; + case COLOR_DEPTH_141414: + output_bpc = 14; + break; + case COLOR_DEPTH_161616: + output_bpc = 16; + break; + case COLOR_DEPTH_999: + output_bpc = 9; + break; + case COLOR_DEPTH_111111: + output_bpc = 11; + break; + default: + output_bpc = 8; + break; + } + + switch (timing->pixel_encoding) { + case PIXEL_ENCODING_RGB: + case PIXEL_ENCODING_YCBCR444: + dout->output_format = dm_444; + dout->output_bpp = output_bpc * 3; + break; + case PIXEL_ENCODING_YCBCR420: + dout->output_format = dm_420; + dout->output_bpp = (output_bpc * 3.0) / 2; + break; + case PIXEL_ENCODING_YCBCR422: + if (timing->flags.DSC && !timing->dsc_cfg.ycbcr422_simple) + dout->output_format = dm_n422; + else + dout->output_format = dm_s422; + dout->output_bpp = output_bpc * 2; + break; + default: + dout->output_format = dm_444; + dout->output_bpp = output_bpc * 3; + } +} + +static enum source_format_class dc_source_format_to_dml_source_format(enum surface_pixel_format dc_format) +{ + enum source_format_class dml_format = dm_444_32; + + switch (dc_format) { + case SURFACE_PIXEL_FORMAT_VIDEO_420_YCbCr: + case SURFACE_PIXEL_FORMAT_VIDEO_420_YCrCb: + dml_format = dm_420_8; + break; + case SURFACE_PIXEL_FORMAT_VIDEO_420_10bpc_YCbCr: + case SURFACE_PIXEL_FORMAT_VIDEO_420_10bpc_YCrCb: + dml_format = dm_420_10; + break; + case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616: + case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616F: + case SURFACE_PIXEL_FORMAT_GRPH_ABGR16161616F: + dml_format = dm_444_64; + break; + case SURFACE_PIXEL_FORMAT_GRPH_ARGB1555: + case SURFACE_PIXEL_FORMAT_GRPH_RGB565: + dml_format = dm_444_16; + break; + case SURFACE_PIXEL_FORMAT_GRPH_PALETA_256_COLORS: + dml_format = dm_444_8; + break; + case SURFACE_PIXEL_FORMAT_GRPH_RGBE_ALPHA: + dml_format = dm_rgbe_alpha; + break; + default: + dml_format = dm_444_32; + break; + } + + return dml_format; +} + +#endif diff --git a/drivers/gpu/drm/amd/display/dc/dml/dsc/rc_calc_fpu.c b/drivers/gpu/drm/amd/display/dc/dml/dsc/rc_calc_fpu.c index 3ee858f311d1..122ba291a7ef 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/dsc/rc_calc_fpu.c +++ b/drivers/gpu/drm/amd/display/dc/dml/dsc/rc_calc_fpu.c @@ -61,16 +61,6 @@ static double dsc_roundf(double num) return (int)(num); } -static double dsc_ceil(double num) -{ - double retval = (int)num; - - if (retval != num && num > 0) - retval = num + 1; - - return (int)retval; -} - static void get_qp_set(qp_set qps, enum colour_mode cm, enum bits_per_comp bpc, enum max_min max_min, float bpp) { @@ -268,24 +258,3 @@ void _do_calc_rc_params(struct rc_params *rc, rc->rc_buf_thresh[13] = 8064; } -u32 _do_bytes_per_pixel_calc(int slice_width, - u16 drm_bpp, - bool is_navite_422_or_420) -{ - float bpp; - u32 bytes_per_pixel; - double d_bytes_per_pixel; - - dc_assert_fp_enabled(); - - bpp = ((float)drm_bpp / 16.0); - d_bytes_per_pixel = dsc_ceil(bpp * slice_width / 8.0) / slice_width; - // TODO: Make sure the formula for calculating this is precise (ceiling - // vs. floor, and at what point they should be applied) - if (is_navite_422_or_420) - d_bytes_per_pixel /= 2; - - bytes_per_pixel = (u32)dsc_ceil(d_bytes_per_pixel * 0x10000000); - - return bytes_per_pixel; -} diff --git a/drivers/gpu/drm/amd/display/dc/dml/dsc/rc_calc_fpu.h b/drivers/gpu/drm/amd/display/dc/dml/dsc/rc_calc_fpu.h index b93b95409fbe..cad244c023cd 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/dsc/rc_calc_fpu.h +++ b/drivers/gpu/drm/amd/display/dc/dml/dsc/rc_calc_fpu.h @@ -78,10 +78,6 @@ struct qp_entry { typedef struct qp_entry qp_table[]; -u32 _do_bytes_per_pixel_calc(int slice_width, - u16 drm_bpp, - bool is_navite_422_or_420); - void _do_calc_rc_params(struct rc_params *rc, enum colour_mode cm, enum bits_per_comp bpc, diff --git a/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c b/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c index 0321b4446e05..9c74564cbd8d 100644 --- a/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c +++ b/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c @@ -455,6 +455,7 @@ static bool intersect_dsc_caps( if (pixel_encoding == PIXEL_ENCODING_YCBCR422 || pixel_encoding == PIXEL_ENCODING_YCBCR420) dsc_common_caps->bpp_increment_div = min(dsc_common_caps->bpp_increment_div, (uint32_t)8); + dsc_common_caps->edp_sink_max_bits_per_pixel = dsc_sink_caps->edp_max_bits_per_pixel; dsc_common_caps->is_dp = dsc_sink_caps->is_dp; return true; } @@ -513,6 +514,13 @@ static bool decide_dsc_bandwidth_range( range->min_target_bpp_x16 = preferred_bpp_x16; } } + /* TODO - make this value generic to all signal types */ + else if (dsc_caps->edp_sink_max_bits_per_pixel) { + /* apply max bpp limitation from edp sink */ + range->max_target_bpp_x16 = MIN(dsc_caps->edp_sink_max_bits_per_pixel, + max_bpp_x16); + range->min_target_bpp_x16 = min_bpp_x16; + } else { range->max_target_bpp_x16 = max_bpp_x16; range->min_target_bpp_x16 = min_bpp_x16; @@ -574,7 +582,7 @@ static bool decide_dsc_target_bpp_x16( return *target_bpp_x16 != 0; } -#define MIN_AVAILABLE_SLICES_SIZE 4 +#define MIN_AVAILABLE_SLICES_SIZE 6 static int get_available_dsc_slices(union dsc_enc_slice_caps slice_caps, int *available_slices) { @@ -860,6 +868,10 @@ static bool setup_dsc_config( min_slices_h = 0; // DSC TODO: Maybe try increasing the number of slices first? is_dsc_possible = (min_slices_h <= max_slices_h); + + if (min_slices_h == 0 && max_slices_h == 0) + is_dsc_possible = false; + if (!is_dsc_possible) goto done; diff --git a/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.c b/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.c index b19d3aeb5962..e97cf09be9d5 100644 --- a/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.c +++ b/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.c @@ -60,31 +60,3 @@ void calc_rc_params(struct rc_params *rc, const struct drm_dsc_config *pps) pps->dsc_version_minor); DC_FP_END(); } - -/** - * calc_dsc_bytes_per_pixel - calculate bytes per pixel - * @pps: DRM struct with all required DSC values - * - * Based on the information inside drm_dsc_config, this function calculates the - * total of bytes per pixel. - * - * @note This calculation requires float point operation, most of it executes - * under kernel_fpu_{begin,end}. - * - * Return: - * Return the number of bytes per pixel - */ -u32 calc_dsc_bytes_per_pixel(const struct drm_dsc_config *pps) - -{ - u32 ret; - u16 drm_bpp = pps->bits_per_pixel; - int slice_width = pps->slice_width; - bool is_navite_422_or_420 = pps->native_422 || pps->native_420; - - DC_FP_START(); - ret = _do_bytes_per_pixel_calc(slice_width, drm_bpp, - is_navite_422_or_420); - DC_FP_END(); - return ret; -} diff --git a/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.h b/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.h index c2340e001b57..80921c1c0d53 100644 --- a/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.h +++ b/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.h @@ -30,7 +30,6 @@ #include "dml/dsc/rc_calc_fpu.h" void calc_rc_params(struct rc_params *rc, const struct drm_dsc_config *pps); -u32 calc_dsc_bytes_per_pixel(const struct drm_dsc_config *pps); #endif diff --git a/drivers/gpu/drm/amd/display/dc/dsc/rc_calc_dpi.c b/drivers/gpu/drm/amd/display/dc/dsc/rc_calc_dpi.c index 1e19dd674e5a..7e306aa3e2b9 100644 --- a/drivers/gpu/drm/amd/display/dc/dsc/rc_calc_dpi.c +++ b/drivers/gpu/drm/amd/display/dc/dsc/rc_calc_dpi.c @@ -100,8 +100,7 @@ int dscc_compute_dsc_parameters(const struct drm_dsc_config *pps, struct dsc_par int ret; struct rc_params rc; struct drm_dsc_config dsc_cfg; - - dsc_params->bytes_per_pixel = calc_dsc_bytes_per_pixel(pps); + unsigned long long tmp; calc_rc_params(&rc, pps); dsc_params->pps = *pps; @@ -113,6 +112,9 @@ int dscc_compute_dsc_parameters(const struct drm_dsc_config *pps, struct dsc_par dsc_cfg.mux_word_size = dsc_params->pps.bits_per_component <= 10 ? 48 : 64; ret = drm_dsc_compute_rc_parameters(&dsc_cfg); + tmp = (unsigned long long)dsc_cfg.slice_chunk_size * 0x10000000 + (dsc_cfg.slice_width - 1); + do_div(tmp, (uint32_t)dsc_cfg.slice_width); //ROUND-UP + dsc_params->bytes_per_pixel = (uint32_t)tmp; copy_pps_fields(&dsc_params->pps, &dsc_cfg); dsc_params->rc_buffer_model_size = dsc_cfg.rc_bits; diff --git a/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h b/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h index 806f3041db14..337c0161e72d 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h +++ b/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h @@ -619,7 +619,7 @@ struct dcn_ip_params { }; extern const struct dcn_ip_params dcn10_ip_defaults; -bool dcn_validate_bandwidth( +bool dcn10_validate_bandwidth( struct dc *dc, struct dc_state *context, bool fast_validate); diff --git a/drivers/gpu/drm/amd/display/dc/inc/dml_wrapper.h b/drivers/gpu/drm/amd/display/dc/inc/dml_wrapper.h new file mode 100644 index 000000000000..5dcfbd8e2697 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/inc/dml_wrapper.h @@ -0,0 +1,34 @@ +/* + * Copyright 2015 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: AMD + * + */ + +#ifndef DML_WRAPPER_H_ +#define DML_WRAPPER_H_ + +#include "dc.h" +#include "dml/display_mode_vba.h" + +bool dml_validate(struct dc *dc, struct dc_state *context, bool fast_validate); + +#endif diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/dsc.h b/drivers/gpu/drm/amd/display/dc/inc/hw/dsc.h index f94135c6e3c2..346f0ba73e86 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/dsc.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/dsc.h @@ -61,6 +61,8 @@ struct dcn_dsc_state { uint32_t dsc_pic_height; uint32_t dsc_slice_bpg_offset; uint32_t dsc_chunk_size; + uint32_t dsc_fw_en; + uint32_t dsc_opp_source; }; @@ -88,6 +90,7 @@ struct dsc_enc_caps { int32_t max_total_throughput_mps; /* Maximum total throughput with all the slices combined */ int32_t max_slice_width; uint32_t bpp_increment_div; /* bpp increment divisor, e.g. if 16, it's 1/16th of a bit */ + uint32_t edp_sink_max_bits_per_pixel; bool is_dp; }; diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h b/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h index c88e113b94d1..073f8b667eff 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h @@ -164,6 +164,10 @@ struct stream_encoder_funcs { void (*stop_dp_info_packets)( struct stream_encoder *enc); + void (*reset_fifo)( + struct stream_encoder *enc + ); + void (*dp_blank)( struct dc_link *link, struct stream_encoder *enc); diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h b/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h index 7390baf916b5..c29320b3855d 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h @@ -290,6 +290,8 @@ struct timing_generator_funcs { enum optc_dsc_mode dsc_mode, uint32_t dsc_bytes_per_pixel, uint32_t dsc_slice_width); + void (*get_dsc_status)(struct timing_generator *optc, + uint32_t *dsc_mode); void (*set_odm_bypass)(struct timing_generator *optc, const struct dc_crtc_timing *dc_crtc_timing); void (*set_odm_combine)(struct timing_generator *optc, int *opp_id, int opp_cnt, struct dc_crtc_timing *timing); diff --git a/drivers/gpu/drm/amd/display/dc/inc/link_enc_cfg.h b/drivers/gpu/drm/amd/display/dc/inc/link_enc_cfg.h index 10dcf6a5e9b1..a4e43b4826e0 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/link_enc_cfg.h +++ b/drivers/gpu/drm/amd/display/dc/inc/link_enc_cfg.h @@ -36,7 +36,7 @@ * Initialise link encoder resource tracking. */ void link_enc_cfg_init( - struct dc *dc, + const struct dc *dc, struct dc_state *state); /* diff --git a/drivers/gpu/drm/amd/display/dmub/dmub_srv.h b/drivers/gpu/drm/amd/display/dmub/dmub_srv.h index cd204eef073b..83855b8a32e9 100644 --- a/drivers/gpu/drm/amd/display/dmub/dmub_srv.h +++ b/drivers/gpu/drm/amd/display/dmub/dmub_srv.h @@ -360,6 +360,8 @@ struct dmub_srv_hw_funcs { uint32_t (*get_gpint_dataout)(struct dmub_srv *dmub); + void (*clear_inbox0_ack_register)(struct dmub_srv *dmub); + uint32_t (*read_inbox0_ack_register)(struct dmub_srv *dmub); void (*send_inbox0_cmd)(struct dmub_srv *dmub, union dmub_inbox0_data_register data); uint32_t (*get_current_time)(struct dmub_srv *dmub); @@ -409,6 +411,7 @@ struct dmub_srv { struct dmub_srv_base_funcs funcs; struct dmub_srv_hw_funcs hw_funcs; struct dmub_rb inbox1_rb; + uint32_t inbox1_last_wptr; /** * outbox1_rb is accessed without locks (dal & dc) * and to be used only in dmub_srv_stat_get_notification() @@ -735,6 +738,45 @@ bool dmub_srv_get_diagnostic_data(struct dmub_srv *dmub, struct dmub_diagnostic_ bool dmub_srv_should_detect(struct dmub_srv *dmub); +/** + * dmub_srv_send_inbox0_cmd() - Send command to DMUB using INBOX0 + * @dmub: the dmub service + * @data: the data to be sent in the INBOX0 command + * + * Send command by writing directly to INBOX0 WPTR + * + * Return: + * DMUB_STATUS_OK - success + * DMUB_STATUS_INVALID - hw_init false or hw function does not exist + */ +enum dmub_status dmub_srv_send_inbox0_cmd(struct dmub_srv *dmub, union dmub_inbox0_data_register data); + +/** + * dmub_srv_wait_for_inbox0_ack() - wait for DMUB to ACK INBOX0 command + * @dmub: the dmub service + * @timeout_us: the maximum number of microseconds to wait + * + * Wait for DMUB to ACK the INBOX0 message + * + * Return: + * DMUB_STATUS_OK - success + * DMUB_STATUS_INVALID - hw_init false or hw function does not exist + * DMUB_STATUS_TIMEOUT - wait for ack timed out + */ +enum dmub_status dmub_srv_wait_for_inbox0_ack(struct dmub_srv *dmub, uint32_t timeout_us); + +/** + * dmub_srv_wait_for_inbox0_ack() - clear ACK register for INBOX0 + * @dmub: the dmub service + * + * Clear ACK register for INBOX0 + * + * Return: + * DMUB_STATUS_OK - success + * DMUB_STATUS_INVALID - hw_init false or hw function does not exist + */ +enum dmub_status dmub_srv_clear_inbox0_ack(struct dmub_srv *dmub); + #if defined(__cplusplus) } #endif diff --git a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h index c29a67ccef17..7eec65090862 100644 --- a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h +++ b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h @@ -46,10 +46,10 @@ /* Firmware versioning. */ #ifdef DMUB_EXPOSE_VERSION -#define DMUB_FW_VERSION_GIT_HASH 0x1d82d23e +#define DMUB_FW_VERSION_GIT_HASH 0x465e619a #define DMUB_FW_VERSION_MAJOR 0 #define DMUB_FW_VERSION_MINOR 0 -#define DMUB_FW_VERSION_REVISION 91 +#define DMUB_FW_VERSION_REVISION 94 #define DMUB_FW_VERSION_TEST 0 #define DMUB_FW_VERSION_VBIOS 0 #define DMUB_FW_VERSION_HOTFIX 0 @@ -172,13 +172,6 @@ extern "C" { #define dmub_udelay(microseconds) udelay(microseconds) #endif -/** - * Number of nanoseconds per DMUB tick. - * DMCUB_TIMER_CURRENT increments in DMUB ticks, which are 10ns by default. - * If DMCUB_TIMER_WINDOW is non-zero this will no longer be true. - */ -#define NS_PER_DMUB_TICK 10 - /** * union dmub_addr - DMUB physical/virtual 64-bit address. */ @@ -208,10 +201,9 @@ union dmub_psr_debug_flags { uint32_t use_hw_lock_mgr : 1; /** - * Unused. - * TODO: Remove. + * Use TPS3 signal when restore main link. */ - uint32_t log_line_nums : 1; + uint32_t force_wakeup_by_tps3 : 1; } bitfields; /** @@ -1550,10 +1542,14 @@ struct dmub_cmd_psr_copy_settings_data { * Currently the support is only for 0 or 1 */ uint8_t panel_inst; - /** - * Explicit padding to 4 byte boundary. + /* + * DSC enable status in driver */ - uint8_t pad3[4]; + uint8_t dsc_enable_status; + /** + * Explicit padding to 3 byte boundary. + */ + uint8_t pad3[3]; }; /** @@ -2722,7 +2718,7 @@ static inline bool dmub_rb_full(struct dmub_rb *rb) static inline bool dmub_rb_push_front(struct dmub_rb *rb, const union dmub_rb_cmd *cmd) { - uint64_t volatile *dst = (uint64_t volatile *)(rb->base_address) + rb->wrpt / sizeof(uint64_t); + uint64_t volatile *dst = (uint64_t volatile *)((uint8_t *)(rb->base_address) + rb->wrpt); const uint64_t *src = (const uint64_t *)cmd; uint8_t i; @@ -2840,7 +2836,7 @@ static inline bool dmub_rb_peek_offset(struct dmub_rb *rb, static inline bool dmub_rb_out_front(struct dmub_rb *rb, union dmub_rb_out_cmd *cmd) { - const uint64_t volatile *src = (const uint64_t volatile *)(rb->base_address) + rb->rptr / sizeof(uint64_t); + const uint64_t volatile *src = (const uint64_t volatile *)((uint8_t *)(rb->base_address) + rb->rptr); uint64_t *dst = (uint64_t *)cmd; uint8_t i; @@ -2888,7 +2884,7 @@ static inline void dmub_rb_flush_pending(const struct dmub_rb *rb) uint32_t wptr = rb->wrpt; while (rptr != wptr) { - uint64_t volatile *data = (uint64_t volatile *)rb->base_address + rptr / sizeof(uint64_t); + uint64_t volatile *data = (uint64_t volatile *)((uint8_t *)(rb->base_address) + rptr); //uint64_t volatile *p = (uint64_t volatile *)data; uint64_t temp; uint8_t i; diff --git a/drivers/gpu/drm/amd/display/dmub/src/dmub_srv.c b/drivers/gpu/drm/amd/display/dmub/src/dmub_srv.c index 56d400ffa7ac..f673a1c1777a 100644 --- a/drivers/gpu/drm/amd/display/dmub/src/dmub_srv.c +++ b/drivers/gpu/drm/amd/display/dmub/src/dmub_srv.c @@ -100,24 +100,9 @@ void dmub_flush_buffer_mem(const struct dmub_fb *fb) } static const struct dmub_fw_meta_info * -dmub_get_fw_meta_info(const struct dmub_srv_region_params *params) +dmub_get_fw_meta_info_from_blob(const uint8_t *blob, uint32_t blob_size, uint32_t meta_offset) { const union dmub_fw_meta *meta; - const uint8_t *blob = NULL; - uint32_t blob_size = 0; - uint32_t meta_offset = 0; - - if (params->fw_bss_data && params->bss_data_size) { - /* Legacy metadata region. */ - blob = params->fw_bss_data; - blob_size = params->bss_data_size; - meta_offset = DMUB_FW_META_OFFSET; - } else if (params->fw_inst_const && params->inst_const_size) { - /* Combined metadata region. */ - blob = params->fw_inst_const; - blob_size = params->inst_const_size; - meta_offset = 0; - } if (!blob || !blob_size) return NULL; @@ -134,6 +119,32 @@ dmub_get_fw_meta_info(const struct dmub_srv_region_params *params) return &meta->info; } +static const struct dmub_fw_meta_info * +dmub_get_fw_meta_info(const struct dmub_srv_region_params *params) +{ + const struct dmub_fw_meta_info *info = NULL; + + if (params->fw_bss_data && params->bss_data_size) { + /* Legacy metadata region. */ + info = dmub_get_fw_meta_info_from_blob(params->fw_bss_data, + params->bss_data_size, + DMUB_FW_META_OFFSET); + } else if (params->fw_inst_const && params->inst_const_size) { + /* Combined metadata region - can be aligned to 16-bytes. */ + uint32_t i; + + for (i = 0; i < 16; ++i) { + info = dmub_get_fw_meta_info_from_blob( + params->fw_inst_const, params->inst_const_size, i); + + if (info) + break; + } + } + + return info; +} + static bool dmub_srv_hw_setup(struct dmub_srv *dmub, enum dmub_asic asic) { struct dmub_srv_hw_funcs *funcs = &dmub->hw_funcs; @@ -598,6 +609,8 @@ enum dmub_status dmub_srv_cmd_queue(struct dmub_srv *dmub, enum dmub_status dmub_srv_cmd_execute(struct dmub_srv *dmub) { + struct dmub_rb flush_rb; + if (!dmub->hw_init) return DMUB_STATUS_INVALID; @@ -606,9 +619,14 @@ enum dmub_status dmub_srv_cmd_execute(struct dmub_srv *dmub) * been flushed to framebuffer memory. Otherwise DMCUB might * read back stale, fully invalid or partially invalid data. */ - dmub_rb_flush_pending(&dmub->inbox1_rb); + flush_rb = dmub->inbox1_rb; + flush_rb.rptr = dmub->inbox1_last_wptr; + dmub_rb_flush_pending(&flush_rb); + + dmub->hw_funcs.set_inbox1_wptr(dmub, dmub->inbox1_rb.wrpt); + + dmub->inbox1_last_wptr = dmub->inbox1_rb.wrpt; - dmub->hw_funcs.set_inbox1_wptr(dmub, dmub->inbox1_rb.wrpt); return DMUB_STATUS_OK; } @@ -831,3 +849,38 @@ bool dmub_srv_should_detect(struct dmub_srv *dmub) return dmub->hw_funcs.should_detect(dmub); } + +enum dmub_status dmub_srv_clear_inbox0_ack(struct dmub_srv *dmub) +{ + if (!dmub->hw_init || dmub->hw_funcs.clear_inbox0_ack_register) + return DMUB_STATUS_INVALID; + + dmub->hw_funcs.clear_inbox0_ack_register(dmub); + return DMUB_STATUS_OK; +} + +enum dmub_status dmub_srv_wait_for_inbox0_ack(struct dmub_srv *dmub, uint32_t timeout_us) +{ + uint32_t i = 0; + uint32_t ack = 0; + + if (!dmub->hw_init || !dmub->hw_funcs.read_inbox0_ack_register) + return DMUB_STATUS_INVALID; + + for (i = 0; i <= timeout_us; i++) { + ack = dmub->hw_funcs.read_inbox0_ack_register(dmub); + if (ack) + return DMUB_STATUS_OK; + } + return DMUB_STATUS_TIMEOUT; +} + +enum dmub_status dmub_srv_send_inbox0_cmd(struct dmub_srv *dmub, + union dmub_inbox0_data_register data) +{ + if (!dmub->hw_init || dmub->hw_funcs.send_inbox0_cmd) + return DMUB_STATUS_INVALID; + + dmub->hw_funcs.send_inbox0_cmd(dmub, data); + return DMUB_STATUS_OK; +} diff --git a/drivers/gpu/drm/amd/display/include/ddc_service_types.h b/drivers/gpu/drm/amd/display/include/ddc_service_types.h index 4de59b66bb1a..a2b80514d83e 100644 --- a/drivers/gpu/drm/amd/display/include/ddc_service_types.h +++ b/drivers/gpu/drm/amd/display/include/ddc_service_types.h @@ -35,6 +35,7 @@ #define DP_BRANCH_DEVICE_ID_00E04C 0x00E04C #define DP_BRANCH_DEVICE_ID_006037 0x006037 +#define DP_DEVICE_ID_38EC11 0x38EC11 enum ddc_result { DDC_RESULT_UNKNOWN = 0, DDC_RESULT_SUCESSFULL, @@ -117,4 +118,7 @@ struct av_sync_data { uint8_t aud_del_ins3;/* DPCD 0002Dh */ }; +static const uint8_t DP_SINK_DEVICE_STR_ID_1[] = {7, 1, 8, 7, 3, 0}; +static const uint8_t DP_SINK_DEVICE_STR_ID_2[] = {7, 1, 8, 7, 5, 0}; + #endif /* __DAL_DDC_SERVICE_TYPES_H__ */ diff --git a/drivers/gpu/drm/amd/display/include/logger_types.h b/drivers/gpu/drm/amd/display/include/logger_types.h index 370fad883e33..f093b49c5e6e 100644 --- a/drivers/gpu/drm/amd/display/include/logger_types.h +++ b/drivers/gpu/drm/amd/display/include/logger_types.h @@ -72,9 +72,7 @@ #define DC_LOG_DSC(...) DRM_DEBUG_KMS(__VA_ARGS__) #define DC_LOG_SMU(...) pr_debug("[SMU_MSG]:"__VA_ARGS__) #define DC_LOG_DWB(...) DRM_DEBUG_KMS(__VA_ARGS__) -#if defined(CONFIG_DRM_AMD_DC_DCN) #define DC_LOG_DP2(...) DRM_DEBUG_KMS(__VA_ARGS__) -#endif struct dal_logger; @@ -126,9 +124,7 @@ enum dc_log_type { LOG_MAX_HW_POINTS, LOG_ALL_TF_CHANNELS, LOG_SAMPLE_1DLUT, -#if defined(CONFIG_DRM_AMD_DC_DCN) LOG_DP2, -#endif LOG_SECTION_TOTAL_COUNT }; diff --git a/drivers/gpu/drm/amd/include/asic_reg/mp/mp_11_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/mp/mp_11_0_offset.h index 6d0052ce6bed..da6d380c948b 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/mp/mp_11_0_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/mp/mp_11_0_offset.h @@ -354,5 +354,12 @@ #define mmMP1_SMN_EXT_SCRATCH7 0x03c7 #define mmMP1_SMN_EXT_SCRATCH7_BASE_IDX 0 +/* + * addressBlock: mp_SmuMp1Pub_MmuDec + * base address: 0x0 + */ +#define smnMP1_PMI_3_START 0x3030204 +#define smnMP1_PMI_3_FIFO 0x3030208 +#define smnMP1_PMI_3 0x3030600 #endif diff --git a/drivers/gpu/drm/amd/include/asic_reg/mp/mp_11_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/mp/mp_11_0_sh_mask.h index 136fb5de6a4c..a5ae2a801254 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/mp/mp_11_0_sh_mask.h +++ b/drivers/gpu/drm/amd/include/asic_reg/mp/mp_11_0_sh_mask.h @@ -959,5 +959,17 @@ #define MP1_SMN_EXT_SCRATCH7__DATA__SHIFT 0x0 #define MP1_SMN_EXT_SCRATCH7__DATA_MASK 0xFFFFFFFFL +// MP1_PMI_3_START +#define MP1_PMI_3_START__ENABLE_MASK 0x80000000L +// MP1_PMI_3_FIFO +#define MP1_PMI_3_FIFO__DEPTH_MASK 0x00000fffL + +// MP1_PMI_3_START +#define MP1_PMI_3_START__ENABLE__SHIFT 0x0000001f +// MP1_PMI_3_FIFO +#define MP1_PMI_3_FIFO__DEPTH__SHIFT 0x00000000 + + + #endif diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h index c84bd7b2cf59..ac941f62cbed 100644 --- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h @@ -33,12 +33,11 @@ #include struct pci_dev; +struct amdgpu_device; #define KGD_MAX_QUEUES 128 struct kfd_dev; -struct kgd_dev; - struct kgd_mem; enum kfd_preempt_type { @@ -228,61 +227,61 @@ struct tile_config { */ struct kfd2kgd_calls { /* Register access functions */ - void (*program_sh_mem_settings)(struct kgd_dev *kgd, uint32_t vmid, + void (*program_sh_mem_settings)(struct amdgpu_device *adev, uint32_t vmid, uint32_t sh_mem_config, uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases); - int (*set_pasid_vmid_mapping)(struct kgd_dev *kgd, u32 pasid, + int (*set_pasid_vmid_mapping)(struct amdgpu_device *adev, u32 pasid, unsigned int vmid); - int (*init_interrupts)(struct kgd_dev *kgd, uint32_t pipe_id); + int (*init_interrupts)(struct amdgpu_device *adev, uint32_t pipe_id); - int (*hqd_load)(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, + int (*hqd_load)(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t __user *wptr, uint32_t wptr_shift, uint32_t wptr_mask, struct mm_struct *mm); - int (*hiq_mqd_load)(struct kgd_dev *kgd, void *mqd, + int (*hiq_mqd_load)(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t doorbell_off); - int (*hqd_sdma_load)(struct kgd_dev *kgd, void *mqd, + int (*hqd_sdma_load)(struct amdgpu_device *adev, void *mqd, uint32_t __user *wptr, struct mm_struct *mm); - int (*hqd_dump)(struct kgd_dev *kgd, + int (*hqd_dump)(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs); - int (*hqd_sdma_dump)(struct kgd_dev *kgd, + int (*hqd_sdma_dump)(struct amdgpu_device *adev, uint32_t engine_id, uint32_t queue_id, uint32_t (**dump)[2], uint32_t *n_regs); - bool (*hqd_is_occupied)(struct kgd_dev *kgd, uint64_t queue_address, - uint32_t pipe_id, uint32_t queue_id); - - int (*hqd_destroy)(struct kgd_dev *kgd, void *mqd, uint32_t reset_type, - unsigned int timeout, uint32_t pipe_id, + bool (*hqd_is_occupied)(struct amdgpu_device *adev, + uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id); - bool (*hqd_sdma_is_occupied)(struct kgd_dev *kgd, void *mqd); + int (*hqd_destroy)(struct amdgpu_device *adev, void *mqd, + uint32_t reset_type, unsigned int timeout, + uint32_t pipe_id, uint32_t queue_id); - int (*hqd_sdma_destroy)(struct kgd_dev *kgd, void *mqd, + bool (*hqd_sdma_is_occupied)(struct amdgpu_device *adev, void *mqd); + + int (*hqd_sdma_destroy)(struct amdgpu_device *adev, void *mqd, unsigned int timeout); - int (*address_watch_disable)(struct kgd_dev *kgd); - int (*address_watch_execute)(struct kgd_dev *kgd, + int (*address_watch_disable)(struct amdgpu_device *adev); + int (*address_watch_execute)(struct amdgpu_device *adev, unsigned int watch_point_id, uint32_t cntl_val, uint32_t addr_hi, uint32_t addr_lo); - int (*wave_control_execute)(struct kgd_dev *kgd, + int (*wave_control_execute)(struct amdgpu_device *adev, uint32_t gfx_index_val, uint32_t sq_cmd); - uint32_t (*address_watch_get_offset)(struct kgd_dev *kgd, + uint32_t (*address_watch_get_offset)(struct amdgpu_device *adev, unsigned int watch_point_id, unsigned int reg_offset); - bool (*get_atc_vmid_pasid_mapping_info)( - struct kgd_dev *kgd, + bool (*get_atc_vmid_pasid_mapping_info)(struct amdgpu_device *adev, uint8_t vmid, uint16_t *p_pasid); @@ -290,16 +289,16 @@ struct kfd2kgd_calls { * passed to the shader by the CP. It's the user mode driver's * responsibility. */ - void (*set_scratch_backing_va)(struct kgd_dev *kgd, + void (*set_scratch_backing_va)(struct amdgpu_device *adev, uint64_t va, uint32_t vmid); - void (*set_vm_context_page_table_base)(struct kgd_dev *kgd, + void (*set_vm_context_page_table_base)(struct amdgpu_device *adev, uint32_t vmid, uint64_t page_table_base); - uint32_t (*read_vmid_from_vmfault_reg)(struct kgd_dev *kgd); + uint32_t (*read_vmid_from_vmfault_reg)(struct amdgpu_device *adev); - void (*get_cu_occupancy)(struct kgd_dev *kgd, int pasid, int *wave_cnt, - int *max_waves_per_cu); - void (*program_trap_handler_settings)(struct kgd_dev *kgd, + void (*get_cu_occupancy)(struct amdgpu_device *adev, int pasid, + int *wave_cnt, int *max_waves_per_cu); + void (*program_trap_handler_settings)(struct amdgpu_device *adev, uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr); }; diff --git a/drivers/gpu/drm/amd/pm/amdgpu_pm.c b/drivers/gpu/drm/amd/pm/amdgpu_pm.c index 41472ed99253..49df4c20f09e 100644 --- a/drivers/gpu/drm/amd/pm/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/pm/amdgpu_pm.c @@ -3759,5 +3759,7 @@ void amdgpu_debugfs_pm_init(struct amdgpu_device *adev) adev, &amdgpu_debugfs_pm_prv_buffer_fops, adev->pm.smu_prv_buffer_size); + + amdgpu_smu_stb_debug_fs_init(adev); #endif } diff --git a/drivers/gpu/drm/amd/pm/inc/amdgpu_smu.h b/drivers/gpu/drm/amd/pm/inc/amdgpu_smu.h index 3557f4e7fc30..f738f7dc20c9 100644 --- a/drivers/gpu/drm/amd/pm/inc/amdgpu_smu.h +++ b/drivers/gpu/drm/amd/pm/inc/amdgpu_smu.h @@ -324,6 +324,7 @@ enum smu_table_id SMU_TABLE_OVERDRIVE, SMU_TABLE_I2C_COMMANDS, SMU_TABLE_PACE, + SMU_TABLE_ECCINFO, SMU_TABLE_COUNT, }; @@ -340,6 +341,7 @@ struct smu_table_context void *max_sustainable_clocks; struct smu_bios_boot_up_values boot_values; void *driver_pptable; + void *ecc_table; struct smu_table tables[SMU_TABLE_COUNT]; /* * The driver table is just a staging buffer for @@ -472,6 +474,12 @@ struct cmn2asic_mapping { int map_to; }; +struct stb_context { + uint32_t stb_buf_size; + bool enabled; + spinlock_t lock; +}; + #define WORKLOAD_POLICY_MAX 7 struct smu_context { @@ -559,6 +567,8 @@ struct smu_context uint16_t cpu_core_num; struct smu_user_dpm_profile user_dpm_profile; + + struct stb_context stb_context; }; struct i2c_adapter; @@ -1261,6 +1271,17 @@ struct pptable_funcs { * of SMUBUS table. */ int (*send_hbm_bad_pages_num)(struct smu_context *smu, uint32_t size); + + /** + * @get_ecc_table: message SMU to get ECC INFO table. + */ + ssize_t (*get_ecc_info)(struct smu_context *smu, void *table); + + + /** + * @stb_collect_info: Collects Smart Trace Buffers data. + */ + int (*stb_collect_info)(struct smu_context *smu, void *buf, uint32_t size); }; typedef enum { @@ -1397,6 +1418,9 @@ int smu_set_light_sbr(struct smu_context *smu, bool enable); int smu_wait_for_event(struct amdgpu_device *adev, enum smu_event_type event, uint64_t event_arg); +int smu_get_ecc_info(struct smu_context *smu, void *umc_ecc); +int smu_stb_collect_info(struct smu_context *smu, void *buff, uint32_t size); +void amdgpu_smu_stb_debug_fs_init(struct amdgpu_device *adev); #endif #endif diff --git a/drivers/gpu/drm/amd/pm/inc/smu13_driver_if_aldebaran.h b/drivers/gpu/drm/amd/pm/inc/smu13_driver_if_aldebaran.h index a017983ff1fa..0f67c56c2863 100644 --- a/drivers/gpu/drm/amd/pm/inc/smu13_driver_if_aldebaran.h +++ b/drivers/gpu/drm/amd/pm/inc/smu13_driver_if_aldebaran.h @@ -140,6 +140,8 @@ #define MAX_SW_I2C_COMMANDS 24 +#define ALDEBARAN_UMC_CHANNEL_NUM 32 + typedef enum { I2C_CONTROLLER_PORT_0, //CKSVII2C0 I2C_CONTROLLER_PORT_1, //CKSVII2C1 @@ -507,6 +509,19 @@ typedef struct { uint32_t MmHubPadding[8]; // SMU internal use } AvfsDebugTable_t; +typedef struct { + uint64_t mca_umc_status; + uint64_t mca_umc_addr; + uint16_t ce_count_lo_chip; + uint16_t ce_count_hi_chip; + + uint32_t eccPadding; +} EccInfo_t; + +typedef struct { + EccInfo_t EccInfo[ALDEBARAN_UMC_CHANNEL_NUM]; +} EccInfoTable_t; + // These defines are used with the following messages: // SMC_MSG_TransferTableDram2Smu // SMC_MSG_TransferTableSmu2Dram @@ -517,6 +532,7 @@ typedef struct { #define TABLE_SMU_METRICS 4 #define TABLE_DRIVER_SMU_CONFIG 5 #define TABLE_I2C_COMMANDS 6 -#define TABLE_COUNT 7 +#define TABLE_ECCINFO 7 +#define TABLE_COUNT 8 #endif diff --git a/drivers/gpu/drm/amd/pm/inc/smu_v13_0.h b/drivers/gpu/drm/amd/pm/inc/smu_v13_0.h index e5d3b0d1a032..44af23ae059e 100644 --- a/drivers/gpu/drm/amd/pm/inc/smu_v13_0.h +++ b/drivers/gpu/drm/amd/pm/inc/smu_v13_0.h @@ -27,7 +27,9 @@ #define SMU13_DRIVER_IF_VERSION_INV 0xFFFFFFFF #define SMU13_DRIVER_IF_VERSION_YELLOW_CARP 0x04 -#define SMU13_DRIVER_IF_VERSION_ALDE 0x07 +#define SMU13_DRIVER_IF_VERSION_ALDE 0x08 + +#define SMU13_MODE1_RESET_WAIT_TIME_IN_MS 500 //500ms /* MP Apertures */ #define MP0_Public 0x03800000 @@ -216,7 +218,6 @@ int smu_v13_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state) int smu_v13_0_baco_enter(struct smu_context *smu); int smu_v13_0_baco_exit(struct smu_context *smu); -int smu_v13_0_mode1_reset(struct smu_context *smu); int smu_v13_0_mode2_reset(struct smu_context *smu); int smu_v13_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type, diff --git a/drivers/gpu/drm/amd/pm/powerplay/amd_powerplay.c b/drivers/gpu/drm/amd/pm/powerplay/amd_powerplay.c index 8d796ed3b7d1..20cb234d5061 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/amd_powerplay.c +++ b/drivers/gpu/drm/amd/pm/powerplay/amd_powerplay.c @@ -1551,7 +1551,7 @@ static int pp_set_ppfeature_status(void *handle, uint64_t ppfeature_masks) static int pp_asic_reset_mode_2(void *handle) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; + int ret = 0; if (!hwmgr || !hwmgr->pm_en) return -EINVAL; diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.c index 258c573acc97..1f406f21b452 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu10_hwmgr.c @@ -1024,8 +1024,6 @@ static int smu10_print_clock_levels(struct pp_hwmgr *hwmgr, uint32_t min_freq, max_freq = 0; uint32_t ret = 0; - phm_get_sysfs_buf(&buf, &size); - switch (type) { case PP_SCLK: smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetGfxclkFrequency, &now); @@ -1038,13 +1036,13 @@ static int smu10_print_clock_levels(struct pp_hwmgr *hwmgr, else i = 1; - size += sysfs_emit_at(buf, size, "0: %uMhz %s\n", + size += sprintf(buf + size, "0: %uMhz %s\n", data->gfx_min_freq_limit/100, i == 0 ? "*" : ""); - size += sysfs_emit_at(buf, size, "1: %uMhz %s\n", + size += sprintf(buf + size, "1: %uMhz %s\n", i == 1 ? now : SMU10_UMD_PSTATE_GFXCLK, i == 1 ? "*" : ""); - size += sysfs_emit_at(buf, size, "2: %uMhz %s\n", + size += sprintf(buf + size, "2: %uMhz %s\n", data->gfx_max_freq_limit/100, i == 2 ? "*" : ""); break; @@ -1052,7 +1050,7 @@ static int smu10_print_clock_levels(struct pp_hwmgr *hwmgr, smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetFclkFrequency, &now); for (i = 0; i < mclk_table->count; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, mclk_table->entries[i].clk / 100, ((mclk_table->entries[i].clk / 100) @@ -1067,10 +1065,10 @@ static int smu10_print_clock_levels(struct pp_hwmgr *hwmgr, if (ret) return ret; - size += sysfs_emit_at(buf, size, "%s:\n", "OD_SCLK"); - size += sysfs_emit_at(buf, size, "0: %10uMhz\n", + size += sprintf(buf + size, "%s:\n", "OD_SCLK"); + size += sprintf(buf + size, "0: %10uMhz\n", (data->gfx_actual_soft_min_freq > 0) ? data->gfx_actual_soft_min_freq : min_freq); - size += sysfs_emit_at(buf, size, "1: %10uMhz\n", + size += sprintf(buf + size, "1: %10uMhz\n", (data->gfx_actual_soft_max_freq > 0) ? data->gfx_actual_soft_max_freq : max_freq); } break; @@ -1083,8 +1081,8 @@ static int smu10_print_clock_levels(struct pp_hwmgr *hwmgr, if (ret) return ret; - size += sysfs_emit_at(buf, size, "%s:\n", "OD_RANGE"); - size += sysfs_emit_at(buf, size, "SCLK: %7uMHz %10uMHz\n", + size += sprintf(buf + size, "%s:\n", "OD_RANGE"); + size += sprintf(buf + size, "SCLK: %7uMHz %10uMHz\n", min_freq, max_freq); } break; diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu7_hwmgr.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu7_hwmgr.c index aceebf584225..611969bf4520 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu7_hwmgr.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu7_hwmgr.c @@ -4914,8 +4914,6 @@ static int smu7_print_clock_levels(struct pp_hwmgr *hwmgr, int size = 0; uint32_t i, now, clock, pcie_speed; - phm_get_sysfs_buf(&buf, &size); - switch (type) { case PP_SCLK: smum_send_msg_to_smc(hwmgr, PPSMC_MSG_API_GetSclkFrequency, &clock); @@ -4928,7 +4926,7 @@ static int smu7_print_clock_levels(struct pp_hwmgr *hwmgr, now = i; for (i = 0; i < sclk_table->count; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, sclk_table->dpm_levels[i].value / 100, (i == now) ? "*" : ""); break; @@ -4943,7 +4941,7 @@ static int smu7_print_clock_levels(struct pp_hwmgr *hwmgr, now = i; for (i = 0; i < mclk_table->count; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, mclk_table->dpm_levels[i].value / 100, (i == now) ? "*" : ""); break; @@ -4957,7 +4955,7 @@ static int smu7_print_clock_levels(struct pp_hwmgr *hwmgr, now = i; for (i = 0; i < pcie_table->count; i++) - size += sysfs_emit_at(buf, size, "%d: %s %s\n", i, + size += sprintf(buf + size, "%d: %s %s\n", i, (pcie_table->dpm_levels[i].value == 0) ? "2.5GT/s, x8" : (pcie_table->dpm_levels[i].value == 1) ? "5.0GT/s, x16" : (pcie_table->dpm_levels[i].value == 2) ? "8.0GT/s, x16" : "", @@ -4965,32 +4963,32 @@ static int smu7_print_clock_levels(struct pp_hwmgr *hwmgr, break; case OD_SCLK: if (hwmgr->od_enabled) { - size += sysfs_emit_at(buf, size, "%s:\n", "OD_SCLK"); + size += sprintf(buf + size, "%s:\n", "OD_SCLK"); for (i = 0; i < odn_sclk_table->num_of_pl; i++) - size += sysfs_emit_at(buf, size, "%d: %10uMHz %10umV\n", + size += sprintf(buf + size, "%d: %10uMHz %10umV\n", i, odn_sclk_table->entries[i].clock/100, odn_sclk_table->entries[i].vddc); } break; case OD_MCLK: if (hwmgr->od_enabled) { - size += sysfs_emit_at(buf, size, "%s:\n", "OD_MCLK"); + size += sprintf(buf + size, "%s:\n", "OD_MCLK"); for (i = 0; i < odn_mclk_table->num_of_pl; i++) - size += sysfs_emit_at(buf, size, "%d: %10uMHz %10umV\n", + size += sprintf(buf + size, "%d: %10uMHz %10umV\n", i, odn_mclk_table->entries[i].clock/100, odn_mclk_table->entries[i].vddc); } break; case OD_RANGE: if (hwmgr->od_enabled) { - size += sysfs_emit_at(buf, size, "%s:\n", "OD_RANGE"); - size += sysfs_emit_at(buf, size, "SCLK: %7uMHz %10uMHz\n", + size += sprintf(buf + size, "%s:\n", "OD_RANGE"); + size += sprintf(buf + size, "SCLK: %7uMHz %10uMHz\n", data->golden_dpm_table.sclk_table.dpm_levels[0].value/100, hwmgr->platform_descriptor.overdriveLimit.engineClock/100); - size += sysfs_emit_at(buf, size, "MCLK: %7uMHz %10uMHz\n", + size += sprintf(buf + size, "MCLK: %7uMHz %10uMHz\n", data->golden_dpm_table.mclk_table.dpm_levels[0].value/100, hwmgr->platform_descriptor.overdriveLimit.memoryClock/100); - size += sysfs_emit_at(buf, size, "VDDC: %7umV %11umV\n", + size += sprintf(buf + size, "VDDC: %7umV %11umV\n", data->odn_dpm_table.min_vddc, data->odn_dpm_table.max_vddc); } diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu8_hwmgr.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu8_hwmgr.c index 8e28a8eecefc..03bf8f069222 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu8_hwmgr.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu8_hwmgr.c @@ -1550,8 +1550,6 @@ static int smu8_print_clock_levels(struct pp_hwmgr *hwmgr, uint32_t i, now; int size = 0; - phm_get_sysfs_buf(&buf, &size); - switch (type) { case PP_SCLK: now = PHM_GET_FIELD(cgs_read_ind_register(hwmgr->device, @@ -1561,7 +1559,7 @@ static int smu8_print_clock_levels(struct pp_hwmgr *hwmgr, CURR_SCLK_INDEX); for (i = 0; i < sclk_table->count; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, sclk_table->entries[i].clk / 100, (i == now) ? "*" : ""); break; @@ -1573,7 +1571,7 @@ static int smu8_print_clock_levels(struct pp_hwmgr *hwmgr, CURR_MCLK_INDEX); for (i = SMU8_NUM_NBPMEMORYCLOCK; i > 0; i--) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", SMU8_NUM_NBPMEMORYCLOCK-i, data->sys_info.nbp_memory_clock[i-1] / 100, (SMU8_NUM_NBPMEMORYCLOCK-i == now) ? "*" : ""); break; diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega10_hwmgr.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega10_hwmgr.c index c981fc2882f0..e6336654c565 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega10_hwmgr.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega10_hwmgr.c @@ -4639,8 +4639,6 @@ static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, int i, now, size = 0, count = 0; - phm_get_sysfs_buf(&buf, &size); - switch (type) { case PP_SCLK: if (data->registry_data.sclk_dpm_key_disabled) @@ -4654,7 +4652,7 @@ static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, else count = sclk_table->count; for (i = 0; i < count; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, sclk_table->dpm_levels[i].value / 100, (i == now) ? "*" : ""); break; @@ -4665,7 +4663,7 @@ static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetCurrentUclkIndex, &now); for (i = 0; i < mclk_table->count; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, mclk_table->dpm_levels[i].value / 100, (i == now) ? "*" : ""); break; @@ -4676,7 +4674,7 @@ static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetCurrentSocclkIndex, &now); for (i = 0; i < soc_table->count; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, soc_table->dpm_levels[i].value / 100, (i == now) ? "*" : ""); break; @@ -4688,7 +4686,7 @@ static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, PPSMC_MSG_GetClockFreqMHz, CLK_DCEFCLK, &now); for (i = 0; i < dcef_table->count; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, dcef_table->dpm_levels[i].value / 100, (dcef_table->dpm_levels[i].value / 100 == now) ? "*" : ""); @@ -4702,7 +4700,7 @@ static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, gen_speed = pptable->PcieGenSpeed[i]; lane_width = pptable->PcieLaneCount[i]; - size += sysfs_emit_at(buf, size, "%d: %s %s %s\n", i, + size += sprintf(buf + size, "%d: %s %s %s\n", i, (gen_speed == 0) ? "2.5GT/s," : (gen_speed == 1) ? "5.0GT/s," : (gen_speed == 2) ? "8.0GT/s," : @@ -4721,34 +4719,34 @@ static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, case OD_SCLK: if (hwmgr->od_enabled) { - size += sysfs_emit_at(buf, size, "%s:\n", "OD_SCLK"); + size += sprintf(buf + size, "%s:\n", "OD_SCLK"); podn_vdd_dep = &data->odn_dpm_table.vdd_dep_on_sclk; for (i = 0; i < podn_vdd_dep->count; i++) - size += sysfs_emit_at(buf, size, "%d: %10uMhz %10umV\n", + size += sprintf(buf + size, "%d: %10uMhz %10umV\n", i, podn_vdd_dep->entries[i].clk / 100, podn_vdd_dep->entries[i].vddc); } break; case OD_MCLK: if (hwmgr->od_enabled) { - size += sysfs_emit_at(buf, size, "%s:\n", "OD_MCLK"); + size += sprintf(buf + size, "%s:\n", "OD_MCLK"); podn_vdd_dep = &data->odn_dpm_table.vdd_dep_on_mclk; for (i = 0; i < podn_vdd_dep->count; i++) - size += sysfs_emit_at(buf, size, "%d: %10uMhz %10umV\n", + size += sprintf(buf + size, "%d: %10uMhz %10umV\n", i, podn_vdd_dep->entries[i].clk/100, podn_vdd_dep->entries[i].vddc); } break; case OD_RANGE: if (hwmgr->od_enabled) { - size += sysfs_emit_at(buf, size, "%s:\n", "OD_RANGE"); - size += sysfs_emit_at(buf, size, "SCLK: %7uMHz %10uMHz\n", + size += sprintf(buf + size, "%s:\n", "OD_RANGE"); + size += sprintf(buf + size, "SCLK: %7uMHz %10uMHz\n", data->golden_dpm_table.gfx_table.dpm_levels[0].value/100, hwmgr->platform_descriptor.overdriveLimit.engineClock/100); - size += sysfs_emit_at(buf, size, "MCLK: %7uMHz %10uMHz\n", + size += sprintf(buf + size, "MCLK: %7uMHz %10uMHz\n", data->golden_dpm_table.mem_table.dpm_levels[0].value/100, hwmgr->platform_descriptor.overdriveLimit.memoryClock/100); - size += sysfs_emit_at(buf, size, "VDDC: %7umV %11umV\n", + size += sprintf(buf + size, "VDDC: %7umV %11umV\n", data->odn_dpm_table.min_vddc, data->odn_dpm_table.max_vddc); } diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega12_hwmgr.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega12_hwmgr.c index f7e783e1c888..a2f4d6773d45 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega12_hwmgr.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega12_hwmgr.c @@ -2246,8 +2246,6 @@ static int vega12_print_clock_levels(struct pp_hwmgr *hwmgr, int i, now, size = 0; struct pp_clock_levels_with_latency clocks; - phm_get_sysfs_buf(&buf, &size); - switch (type) { case PP_SCLK: PP_ASSERT_WITH_CODE( @@ -2260,7 +2258,7 @@ static int vega12_print_clock_levels(struct pp_hwmgr *hwmgr, "Attempt to get gfx clk levels Failed!", return -1); for (i = 0; i < clocks.num_levels; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, clocks.data[i].clocks_in_khz / 1000, (clocks.data[i].clocks_in_khz / 1000 == now / 100) ? "*" : ""); break; @@ -2276,7 +2274,7 @@ static int vega12_print_clock_levels(struct pp_hwmgr *hwmgr, "Attempt to get memory clk levels Failed!", return -1); for (i = 0; i < clocks.num_levels; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, clocks.data[i].clocks_in_khz / 1000, (clocks.data[i].clocks_in_khz / 1000 == now / 100) ? "*" : ""); break; @@ -2294,7 +2292,7 @@ static int vega12_print_clock_levels(struct pp_hwmgr *hwmgr, "Attempt to get soc clk levels Failed!", return -1); for (i = 0; i < clocks.num_levels; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, clocks.data[i].clocks_in_khz / 1000, (clocks.data[i].clocks_in_khz / 1000 == now) ? "*" : ""); break; @@ -2312,7 +2310,7 @@ static int vega12_print_clock_levels(struct pp_hwmgr *hwmgr, "Attempt to get dcef clk levels Failed!", return -1); for (i = 0; i < clocks.num_levels; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, clocks.data[i].clocks_in_khz / 1000, (clocks.data[i].clocks_in_khz / 1000 == now) ? "*" : ""); break; diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_hwmgr.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_hwmgr.c index 03e63be4ee27..85d55ab4e369 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_hwmgr.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_hwmgr.c @@ -3366,8 +3366,6 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, int ret = 0; uint32_t gen_speed, lane_width, current_gen_speed, current_lane_width; - phm_get_sysfs_buf(&buf, &size); - switch (type) { case PP_SCLK: ret = vega20_get_current_clk_freq(hwmgr, PPCLK_GFXCLK, &now); @@ -3376,13 +3374,13 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, return ret); if (vega20_get_sclks(hwmgr, &clocks)) { - size += sysfs_emit_at(buf, size, "0: %uMhz * (DPM disabled)\n", + size += sprintf(buf + size, "0: %uMhz * (DPM disabled)\n", now / 100); break; } for (i = 0; i < clocks.num_levels; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, clocks.data[i].clocks_in_khz / 1000, (clocks.data[i].clocks_in_khz == now * 10) ? "*" : ""); break; @@ -3394,13 +3392,13 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, return ret); if (vega20_get_memclocks(hwmgr, &clocks)) { - size += sysfs_emit_at(buf, size, "0: %uMhz * (DPM disabled)\n", + size += sprintf(buf + size, "0: %uMhz * (DPM disabled)\n", now / 100); break; } for (i = 0; i < clocks.num_levels; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, clocks.data[i].clocks_in_khz / 1000, (clocks.data[i].clocks_in_khz == now * 10) ? "*" : ""); break; @@ -3412,13 +3410,13 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, return ret); if (vega20_get_socclocks(hwmgr, &clocks)) { - size += sysfs_emit_at(buf, size, "0: %uMhz * (DPM disabled)\n", + size += sprintf(buf + size, "0: %uMhz * (DPM disabled)\n", now / 100); break; } for (i = 0; i < clocks.num_levels; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, clocks.data[i].clocks_in_khz / 1000, (clocks.data[i].clocks_in_khz == now * 10) ? "*" : ""); break; @@ -3430,7 +3428,7 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, return ret); for (i = 0; i < fclk_dpm_table->count; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, fclk_dpm_table->dpm_levels[i].value, fclk_dpm_table->dpm_levels[i].value == (now / 100) ? "*" : ""); break; @@ -3442,13 +3440,13 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, return ret); if (vega20_get_dcefclocks(hwmgr, &clocks)) { - size += sysfs_emit_at(buf, size, "0: %uMhz * (DPM disabled)\n", + size += sprintf(buf + size, "0: %uMhz * (DPM disabled)\n", now / 100); break; } for (i = 0; i < clocks.num_levels; i++) - size += sysfs_emit_at(buf, size, "%d: %uMhz %s\n", + size += sprintf(buf + size, "%d: %uMhz %s\n", i, clocks.data[i].clocks_in_khz / 1000, (clocks.data[i].clocks_in_khz == now * 10) ? "*" : ""); break; @@ -3462,7 +3460,7 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, gen_speed = pptable->PcieGenSpeed[i]; lane_width = pptable->PcieLaneCount[i]; - size += sysfs_emit_at(buf, size, "%d: %s %s %dMhz %s\n", i, + size += sprintf(buf + size, "%d: %s %s %dMhz %s\n", i, (gen_speed == 0) ? "2.5GT/s," : (gen_speed == 1) ? "5.0GT/s," : (gen_speed == 2) ? "8.0GT/s," : @@ -3483,18 +3481,18 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, case OD_SCLK: if (od8_settings[OD8_SETTING_GFXCLK_FMIN].feature_id && od8_settings[OD8_SETTING_GFXCLK_FMAX].feature_id) { - size += sysfs_emit_at(buf, size, "%s:\n", "OD_SCLK"); - size += sysfs_emit_at(buf, size, "0: %10uMhz\n", + size += sprintf(buf + size, "%s:\n", "OD_SCLK"); + size += sprintf(buf + size, "0: %10uMhz\n", od_table->GfxclkFmin); - size += sysfs_emit_at(buf, size, "1: %10uMhz\n", + size += sprintf(buf + size, "1: %10uMhz\n", od_table->GfxclkFmax); } break; case OD_MCLK: if (od8_settings[OD8_SETTING_UCLK_FMAX].feature_id) { - size += sysfs_emit_at(buf, size, "%s:\n", "OD_MCLK"); - size += sysfs_emit_at(buf, size, "1: %10uMhz\n", + size += sprintf(buf + size, "%s:\n", "OD_MCLK"); + size += sprintf(buf + size, "1: %10uMhz\n", od_table->UclkFmax); } @@ -3507,14 +3505,14 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, od8_settings[OD8_SETTING_GFXCLK_VOLTAGE1].feature_id && od8_settings[OD8_SETTING_GFXCLK_VOLTAGE2].feature_id && od8_settings[OD8_SETTING_GFXCLK_VOLTAGE3].feature_id) { - size += sysfs_emit_at(buf, size, "%s:\n", "OD_VDDC_CURVE"); - size += sysfs_emit_at(buf, size, "0: %10uMhz %10dmV\n", + size += sprintf(buf + size, "%s:\n", "OD_VDDC_CURVE"); + size += sprintf(buf + size, "0: %10uMhz %10dmV\n", od_table->GfxclkFreq1, od_table->GfxclkVolt1 / VOLTAGE_SCALE); - size += sysfs_emit_at(buf, size, "1: %10uMhz %10dmV\n", + size += sprintf(buf + size, "1: %10uMhz %10dmV\n", od_table->GfxclkFreq2, od_table->GfxclkVolt2 / VOLTAGE_SCALE); - size += sysfs_emit_at(buf, size, "2: %10uMhz %10dmV\n", + size += sprintf(buf + size, "2: %10uMhz %10dmV\n", od_table->GfxclkFreq3, od_table->GfxclkVolt3 / VOLTAGE_SCALE); } @@ -3522,17 +3520,17 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, break; case OD_RANGE: - size += sysfs_emit_at(buf, size, "%s:\n", "OD_RANGE"); + size += sprintf(buf + size, "%s:\n", "OD_RANGE"); if (od8_settings[OD8_SETTING_GFXCLK_FMIN].feature_id && od8_settings[OD8_SETTING_GFXCLK_FMAX].feature_id) { - size += sysfs_emit_at(buf, size, "SCLK: %7uMhz %10uMhz\n", + size += sprintf(buf + size, "SCLK: %7uMhz %10uMhz\n", od8_settings[OD8_SETTING_GFXCLK_FMIN].min_value, od8_settings[OD8_SETTING_GFXCLK_FMAX].max_value); } if (od8_settings[OD8_SETTING_UCLK_FMAX].feature_id) { - size += sysfs_emit_at(buf, size, "MCLK: %7uMhz %10uMhz\n", + size += sprintf(buf + size, "MCLK: %7uMhz %10uMhz\n", od8_settings[OD8_SETTING_UCLK_FMAX].min_value, od8_settings[OD8_SETTING_UCLK_FMAX].max_value); } @@ -3543,22 +3541,22 @@ static int vega20_print_clock_levels(struct pp_hwmgr *hwmgr, od8_settings[OD8_SETTING_GFXCLK_VOLTAGE1].feature_id && od8_settings[OD8_SETTING_GFXCLK_VOLTAGE2].feature_id && od8_settings[OD8_SETTING_GFXCLK_VOLTAGE3].feature_id) { - size += sysfs_emit_at(buf, size, "VDDC_CURVE_SCLK[0]: %7uMhz %10uMhz\n", + size += sprintf(buf + size, "VDDC_CURVE_SCLK[0]: %7uMhz %10uMhz\n", od8_settings[OD8_SETTING_GFXCLK_FREQ1].min_value, od8_settings[OD8_SETTING_GFXCLK_FREQ1].max_value); - size += sysfs_emit_at(buf, size, "VDDC_CURVE_VOLT[0]: %7dmV %11dmV\n", + size += sprintf(buf + size, "VDDC_CURVE_VOLT[0]: %7dmV %11dmV\n", od8_settings[OD8_SETTING_GFXCLK_VOLTAGE1].min_value, od8_settings[OD8_SETTING_GFXCLK_VOLTAGE1].max_value); - size += sysfs_emit_at(buf, size, "VDDC_CURVE_SCLK[1]: %7uMhz %10uMhz\n", + size += sprintf(buf + size, "VDDC_CURVE_SCLK[1]: %7uMhz %10uMhz\n", od8_settings[OD8_SETTING_GFXCLK_FREQ2].min_value, od8_settings[OD8_SETTING_GFXCLK_FREQ2].max_value); - size += sysfs_emit_at(buf, size, "VDDC_CURVE_VOLT[1]: %7dmV %11dmV\n", + size += sprintf(buf + size, "VDDC_CURVE_VOLT[1]: %7dmV %11dmV\n", od8_settings[OD8_SETTING_GFXCLK_VOLTAGE2].min_value, od8_settings[OD8_SETTING_GFXCLK_VOLTAGE2].max_value); - size += sysfs_emit_at(buf, size, "VDDC_CURVE_SCLK[2]: %7uMhz %10uMhz\n", + size += sprintf(buf + size, "VDDC_CURVE_SCLK[2]: %7uMhz %10uMhz\n", od8_settings[OD8_SETTING_GFXCLK_FREQ3].min_value, od8_settings[OD8_SETTING_GFXCLK_FREQ3].max_value); - size += sysfs_emit_at(buf, size, "VDDC_CURVE_VOLT[2]: %7dmV %11dmV\n", + size += sprintf(buf + size, "VDDC_CURVE_VOLT[2]: %7dmV %11dmV\n", od8_settings[OD8_SETTING_GFXCLK_VOLTAGE3].min_value, od8_settings[OD8_SETTING_GFXCLK_VOLTAGE3].max_value); } diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index 01168b8955bf..e156add7b560 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -1153,6 +1153,8 @@ static int smu_smc_hw_setup(struct smu_context *smu) case IP_VERSION(11, 5, 0): case IP_VERSION(11, 0, 12): ret = smu_system_features_control(smu, true); + if (ret) + dev_err(adev->dev, "Failed system features control!\n"); break; default: break; @@ -1277,8 +1279,10 @@ static int smu_smc_hw_setup(struct smu_context *smu) } ret = smu_notify_display_change(smu); - if (ret) + if (ret) { + dev_err(adev->dev, "Failed to notify display change!\n"); return ret; + } /* * Set min deep sleep dce fclk with bootup value from vbios via @@ -1286,8 +1290,6 @@ static int smu_smc_hw_setup(struct smu_context *smu) */ ret = smu_set_min_dcef_deep_sleep(smu, smu->smu_table.boot_values.dcefclk / 100); - if (ret) - return ret; return ret; } @@ -3072,6 +3074,20 @@ int smu_set_light_sbr(struct smu_context *smu, bool enable) return ret; } +int smu_get_ecc_info(struct smu_context *smu, void *umc_ecc) +{ + int ret = -EOPNOTSUPP; + + mutex_lock(&smu->mutex); + if (smu->ppt_funcs && + smu->ppt_funcs->get_ecc_info) + ret = smu->ppt_funcs->get_ecc_info(smu, umc_ecc); + mutex_unlock(&smu->mutex); + + return ret; + +} + static int smu_get_prv_buffer_details(void *handle, void **addr, size_t *size) { struct smu_context *smu = handle; @@ -3161,3 +3177,107 @@ int smu_wait_for_event(struct amdgpu_device *adev, enum smu_event_type event, return ret; } + +int smu_stb_collect_info(struct smu_context *smu, void *buf, uint32_t size) +{ + + if (!smu->ppt_funcs->stb_collect_info || !smu->stb_context.enabled) + return -EOPNOTSUPP; + + /* Confirm the buffer allocated is of correct size */ + if (size != smu->stb_context.stb_buf_size) + return -EINVAL; + + /* + * No need to lock smu mutex as we access STB directly through MMIO + * and not going through SMU messaging route (for now at least). + * For registers access rely on implementation internal locking. + */ + return smu->ppt_funcs->stb_collect_info(smu, buf, size); +} + +#if defined(CONFIG_DEBUG_FS) + +static int smu_stb_debugfs_open(struct inode *inode, struct file *filp) +{ + struct amdgpu_device *adev = filp->f_inode->i_private; + struct smu_context *smu = &adev->smu; + unsigned char *buf; + int r; + + buf = kvmalloc_array(smu->stb_context.stb_buf_size, sizeof(*buf), GFP_KERNEL); + if (!buf) + return -ENOMEM; + + r = smu_stb_collect_info(smu, buf, smu->stb_context.stb_buf_size); + if (r) + goto out; + + filp->private_data = buf; + + return 0; + +out: + kvfree(buf); + return r; +} + +static ssize_t smu_stb_debugfs_read(struct file *filp, char __user *buf, size_t size, + loff_t *pos) +{ + struct amdgpu_device *adev = filp->f_inode->i_private; + struct smu_context *smu = &adev->smu; + + + if (!filp->private_data) + return -EINVAL; + + return simple_read_from_buffer(buf, + size, + pos, filp->private_data, + smu->stb_context.stb_buf_size); +} + +static int smu_stb_debugfs_release(struct inode *inode, struct file *filp) +{ + kvfree(filp->private_data); + filp->private_data = NULL; + + return 0; +} + +/* + * We have to define not only read method but also + * open and release because .read takes up to PAGE_SIZE + * data each time so and so is invoked multiple times. + * We allocate the STB buffer in .open and release it + * in .release + */ +static const struct file_operations smu_stb_debugfs_fops = { + .owner = THIS_MODULE, + .open = smu_stb_debugfs_open, + .read = smu_stb_debugfs_read, + .release = smu_stb_debugfs_release, + .llseek = default_llseek, +}; + +#endif + +void amdgpu_smu_stb_debug_fs_init(struct amdgpu_device *adev) +{ +#if defined(CONFIG_DEBUG_FS) + + struct smu_context *smu = &adev->smu; + + if (!smu->stb_context.stb_buf_size) + return; + + debugfs_create_file_size("amdgpu_smu_stb_dump", + S_IRUSR, + adev_to_drm(adev)->primary->debugfs_root, + adev, + &smu_stb_debugfs_fops, + smu->stb_context.stb_buf_size); +#endif + +} diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c index a4108025fe29..a673e05853fe 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c @@ -80,6 +80,9 @@ (*member) = (smu->smu_table.driver_pptable + offsetof(PPTable_t, field));\ } while(0) +/* STB FIFO depth is in 64bit units */ +#define SIENNA_CICHLID_STB_DEPTH_UNIT_BYTES 8 + static int get_table_size(struct smu_context *smu) { if (smu->adev->ip_versions[MP1_HWIP][0] == IP_VERSION(11, 0, 13)) @@ -650,6 +653,8 @@ static int sienna_cichlid_allocate_dpm_context(struct smu_context *smu) return 0; } +static void sienna_cichlid_stb_init(struct smu_context *smu); + static int sienna_cichlid_init_smc_tables(struct smu_context *smu) { int ret = 0; @@ -662,6 +667,8 @@ static int sienna_cichlid_init_smc_tables(struct smu_context *smu) if (ret) return ret; + sienna_cichlid_stb_init(smu); + return smu_v11_0_init_smc_tables(smu); } @@ -1171,7 +1178,7 @@ static int sienna_cichlid_force_clk_levels(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t mask) { struct amdgpu_device *adev = smu->adev; - int ret = 0, size = 0; + int ret = 0; uint32_t soft_min_level = 0, soft_max_level = 0, min_freq = 0, max_freq = 0; soft_min_level = mask ? (ffs(mask) - 1) : 0; @@ -1216,7 +1223,7 @@ static int sienna_cichlid_force_clk_levels(struct smu_context *smu, if ((clk_type == SMU_GFXCLK) || (clk_type == SMU_SCLK)) amdgpu_gfx_off_ctrl(adev, true); - return size; + return 0; } static int sienna_cichlid_populate_umd_state_clk(struct smu_context *smu) @@ -2135,7 +2142,13 @@ static int sienna_cichlid_od_edit_dpm_table(struct smu_context *smu, static int sienna_cichlid_run_btc(struct smu_context *smu) { - return smu_cmn_send_smc_msg(smu, SMU_MSG_RunDcBtc, NULL); + int res; + + res = smu_cmn_send_smc_msg(smu, SMU_MSG_RunDcBtc, NULL); + if (res) + dev_err(smu->adev->dev, "RunDcBtc failed!\n"); + + return res; } static int sienna_cichlid_baco_enter(struct smu_context *smu) @@ -3619,6 +3632,16 @@ static ssize_t sienna_cichlid_get_gpu_metrics(struct smu_context *smu, gpu_metrics->energy_accumulator = use_metrics_v2 ? metrics_v2->EnergyAccumulator : metrics->EnergyAccumulator; + if (metrics->CurrGfxVoltageOffset) + gpu_metrics->voltage_gfx = + (155000 - 625 * metrics->CurrGfxVoltageOffset) / 100; + if (metrics->CurrMemVidOffset) + gpu_metrics->voltage_mem = + (155000 - 625 * metrics->CurrMemVidOffset) / 100; + if (metrics->CurrSocVoltageOffset) + gpu_metrics->voltage_soc = + (155000 - 625 * metrics->CurrSocVoltageOffset) / 100; + average_gfx_activity = use_metrics_v2 ? metrics_v2->AverageGfxActivity : metrics->AverageGfxActivity; if (average_gfx_activity <= SMU_11_0_7_GFX_BUSY_THRESHOLD) gpu_metrics->average_gfxclk_frequency = @@ -3793,6 +3816,53 @@ static int sienna_cichlid_set_mp1_state(struct smu_context *smu, return ret; } +static void sienna_cichlid_stb_init(struct smu_context *smu) +{ + struct amdgpu_device *adev = smu->adev; + uint32_t reg; + + reg = RREG32_PCIE(MP1_Public | smnMP1_PMI_3_START); + smu->stb_context.enabled = REG_GET_FIELD(reg, MP1_PMI_3_START, ENABLE); + + /* STB is disabled */ + if (!smu->stb_context.enabled) + return; + + spin_lock_init(&smu->stb_context.lock); + + /* STB buffer size in bytes as function of FIFO depth */ + reg = RREG32_PCIE(MP1_Public | smnMP1_PMI_3_FIFO); + smu->stb_context.stb_buf_size = 1 << REG_GET_FIELD(reg, MP1_PMI_3_FIFO, DEPTH); + smu->stb_context.stb_buf_size *= SIENNA_CICHLID_STB_DEPTH_UNIT_BYTES; + + dev_info(smu->adev->dev, "STB initialized to %d entries", + smu->stb_context.stb_buf_size / SIENNA_CICHLID_STB_DEPTH_UNIT_BYTES); + +} + +int sienna_cichlid_stb_get_data_direct(struct smu_context *smu, + void *buf, + uint32_t size) +{ + uint32_t *p = buf; + struct amdgpu_device *adev = smu->adev; + + /* No need to disable interrupts for now as we don't lock it yet from ISR */ + spin_lock(&smu->stb_context.lock); + + /* + * Read the STB FIFO in units of 32bit since this is the accessor window + * (register width) we have. + */ + buf = ((char *) buf) + size; + while ((void *)p < buf) + *p++ = cpu_to_le32(RREG32_PCIE(MP1_Public | smnMP1_PMI_3)); + + spin_unlock(&smu->stb_context.lock); + + return 0; +} + static const struct pptable_funcs sienna_cichlid_ppt_funcs = { .get_allowed_feature_mask = sienna_cichlid_get_allowed_feature_mask, .set_default_dpm_table = sienna_cichlid_set_default_dpm_table, @@ -3882,6 +3952,7 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = { .interrupt_work = smu_v11_0_interrupt_work, .gpo_control = sienna_cichlid_gpo_control, .set_mp1_state = sienna_cichlid_set_mp1_state, + .stb_collect_info = sienna_cichlid_stb_get_data_direct, }; void sienna_cichlid_set_ppt_funcs(struct smu_context *smu) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c index 59a7d276541d..6e781cee8bb6 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c @@ -78,6 +78,12 @@ #define smnPCIE_ESM_CTRL 0x111003D0 +/* + * SMU support ECCTABLE since version 68.42.0, + * use this to check ECCTALE feature whether support + */ +#define SUPPORT_ECCTABLE_SMU_VERSION 0x00442a00 + static const struct smu_temperature_range smu13_thermal_policy[] = { {-273150, 99000, 99000, -273150, 99000, 99000, -273150, 99000, 99000}, @@ -190,6 +196,7 @@ static const struct cmn2asic_mapping aldebaran_table_map[SMU_TABLE_COUNT] = { TAB_MAP(SMU_METRICS), TAB_MAP(DRIVER_SMU_CONFIG), TAB_MAP(I2C_COMMANDS), + TAB_MAP(ECCINFO), }; static const uint8_t aldebaran_throttler_map[] = { @@ -223,6 +230,9 @@ static int aldebaran_tables_init(struct smu_context *smu) SMU_TABLE_INIT(tables, SMU_TABLE_I2C_COMMANDS, sizeof(SwI2cRequest_t), PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); + SMU_TABLE_INIT(tables, SMU_TABLE_ECCINFO, sizeof(EccInfoTable_t), + PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); + smu_table->metrics_table = kzalloc(sizeof(SmuMetrics_t), GFP_KERNEL); if (!smu_table->metrics_table) return -ENOMEM; @@ -235,6 +245,10 @@ static int aldebaran_tables_init(struct smu_context *smu) return -ENOMEM; } + smu_table->ecc_table = kzalloc(tables[SMU_TABLE_ECCINFO].size, GFP_KERNEL); + if (!smu_table->ecc_table) + return -ENOMEM; + return 0; } @@ -1765,6 +1779,98 @@ static ssize_t aldebaran_get_gpu_metrics(struct smu_context *smu, return sizeof(struct gpu_metrics_v1_3); } +static int aldebaran_check_ecc_table_support(struct smu_context *smu) +{ + uint32_t if_version = 0xff, smu_version = 0xff; + int ret = 0; + + ret = smu_cmn_get_smc_version(smu, &if_version, &smu_version); + if (ret) { + /* return not support if failed get smu_version */ + ret = -EOPNOTSUPP; + } + + if (smu_version < SUPPORT_ECCTABLE_SMU_VERSION) + ret = -EOPNOTSUPP; + + return ret; +} + +static ssize_t aldebaran_get_ecc_info(struct smu_context *smu, + void *table) +{ + struct smu_table_context *smu_table = &smu->smu_table; + EccInfoTable_t *ecc_table = NULL; + struct ecc_info_per_ch *ecc_info_per_channel = NULL; + int i, ret = 0; + struct umc_ecc_info *eccinfo = (struct umc_ecc_info *)table; + + ret = aldebaran_check_ecc_table_support(smu); + if (ret) + return ret; + + ret = smu_cmn_update_table(smu, + SMU_TABLE_ECCINFO, + 0, + smu_table->ecc_table, + false); + if (ret) { + dev_info(smu->adev->dev, "Failed to export SMU ecc table!\n"); + return ret; + } + + ecc_table = (EccInfoTable_t *)smu_table->ecc_table; + + for (i = 0; i < ALDEBARAN_UMC_CHANNEL_NUM; i++) { + ecc_info_per_channel = &(eccinfo->ecc[i]); + ecc_info_per_channel->ce_count_lo_chip = + ecc_table->EccInfo[i].ce_count_lo_chip; + ecc_info_per_channel->ce_count_hi_chip = + ecc_table->EccInfo[i].ce_count_hi_chip; + ecc_info_per_channel->mca_umc_status = + ecc_table->EccInfo[i].mca_umc_status; + ecc_info_per_channel->mca_umc_addr = + ecc_table->EccInfo[i].mca_umc_addr; + } + + return ret; +} + +static int aldebaran_mode1_reset(struct smu_context *smu) +{ + u32 smu_version, fatal_err, param; + int ret = 0; + struct amdgpu_device *adev = smu->adev; + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + + fatal_err = 0; + param = SMU_RESET_MODE_1; + + /* + * PM FW support SMU_MSG_GfxDeviceDriverReset from 68.07 + */ + smu_cmn_get_smc_version(smu, NULL, &smu_version); + if (smu_version < 0x00440700) { + ret = smu_cmn_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL); + } + else { + /* fatal error triggered by ras, PMFW supports the flag + from 68.44.0 */ + if ((smu_version >= 0x00442c00) && ras && + atomic_read(&ras->in_recovery)) + fatal_err = 1; + + param |= (fatal_err << 16); + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_GfxDeviceDriverReset, param, NULL); + } + + if (!ret) + msleep(SMU13_MODE1_RESET_WAIT_TIME_IN_MS); + + return ret; +} + static int aldebaran_mode2_reset(struct smu_context *smu) { u32 smu_version; @@ -1925,13 +2031,14 @@ static const struct pptable_funcs aldebaran_ppt_funcs = { .get_gpu_metrics = aldebaran_get_gpu_metrics, .mode1_reset_is_support = aldebaran_is_mode1_reset_supported, .mode2_reset_is_support = aldebaran_is_mode2_reset_supported, - .mode1_reset = smu_v13_0_mode1_reset, + .mode1_reset = aldebaran_mode1_reset, .set_mp1_state = aldebaran_set_mp1_state, .mode2_reset = aldebaran_mode2_reset, .wait_for_event = smu_v13_0_wait_for_event, .i2c_init = aldebaran_i2c_control_init, .i2c_fini = aldebaran_i2c_control_fini, .send_hbm_bad_pages_num = aldebaran_smu_send_hbm_bad_page_num, + .get_ecc_info = aldebaran_get_ecc_info, }; void aldebaran_set_ppt_funcs(struct smu_context *smu) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c index 35145db6eedf..55421ea622fb 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c @@ -60,8 +60,6 @@ MODULE_FIRMWARE("amdgpu/aldebaran_smc.bin"); #define SMU13_VOLTAGE_SCALE 4 -#define SMU13_MODE1_RESET_WAIT_TIME_IN_MS 500 //500ms - #define LINK_WIDTH_MAX 6 #define LINK_SPEED_MAX 3 @@ -430,8 +428,10 @@ int smu_v13_0_fini_smc_tables(struct smu_context *smu) kfree(smu_table->hardcode_pptable); smu_table->hardcode_pptable = NULL; + kfree(smu_table->ecc_table); kfree(smu_table->metrics_table); kfree(smu_table->watermarks_table); + smu_table->ecc_table = NULL; smu_table->metrics_table = NULL; smu_table->watermarks_table = NULL; smu_table->metrics_time = 0; @@ -1424,25 +1424,6 @@ int smu_v13_0_set_azalia_d3_pme(struct smu_context *smu) return ret; } -int smu_v13_0_mode1_reset(struct smu_context *smu) -{ - u32 smu_version; - int ret = 0; - /* - * PM FW support SMU_MSG_GfxDeviceDriverReset from 68.07 - */ - smu_cmn_get_smc_version(smu, NULL, &smu_version); - if (smu_version < 0x00440700) - ret = smu_cmn_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL); - else - ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_GfxDeviceDriverReset, SMU_RESET_MODE_1, NULL); - - if (!ret) - msleep(SMU13_MODE1_RESET_WAIT_TIME_IN_MS); - - return ret; -} - static int smu_v13_0_wait_for_reset_complete(struct smu_context *smu, uint64_t event_arg) { diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c index ea6f50c08c5f..500af6f8adcb 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c @@ -97,7 +97,7 @@ static void smu_cmn_read_arg(struct smu_context *smu, * smu: a pointer to SMU context * * Returns the status of the SMU, which could be, - * 0, the SMU is busy with your previous command; + * 0, the SMU is busy with your command; * 1, execution status: success, execution result: success; * 0xFF, execution status: success, execution result: failure; * 0xFE, unknown command; @@ -352,7 +352,7 @@ int smu_cmn_send_smc_msg_with_param(struct smu_context *smu, __smu_cmn_send_msg(smu, (uint16_t) index, param); reg = __smu_cmn_poll_stat(smu); res = __smu_cmn_reg2errno(smu, reg); - if (res == -EREMOTEIO) + if (res != 0) __smu_cmn_reg_print_error(smu, reg, index, param, msg); if (read_arg) smu_cmn_read_arg(smu, read_arg); diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c index 482fb0ae6cb5..7afe28408085 100644 --- a/drivers/gpu/drm/radeon/radeon_kms.c +++ b/drivers/gpu/drm/radeon/radeon_kms.c @@ -168,7 +168,7 @@ int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags) if (!r) { acpi_status = radeon_acpi_init(rdev); if (acpi_status) - dev_dbg(dev->dev, "Error during ACPI methods call\n"); + dev_dbg(dev->dev, "Error during ACPI methods call\n"); } if (radeon_is_px(dev)) { diff --git a/include/uapi/linux/kfd_sysfs.h b/include/uapi/linux/kfd_sysfs.h new file mode 100644 index 000000000000..e1fb78b4bf09 --- /dev/null +++ b/include/uapi/linux/kfd_sysfs.h @@ -0,0 +1,108 @@ +/* SPDX-License-Identifier: GPL-2.0 OR MIT WITH Linux-syscall-note */ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef KFD_SYSFS_H_INCLUDED +#define KFD_SYSFS_H_INCLUDED + +/* Capability bits in node properties */ +#define HSA_CAP_HOT_PLUGGABLE 0x00000001 +#define HSA_CAP_ATS_PRESENT 0x00000002 +#define HSA_CAP_SHARED_WITH_GRAPHICS 0x00000004 +#define HSA_CAP_QUEUE_SIZE_POW2 0x00000008 +#define HSA_CAP_QUEUE_SIZE_32BIT 0x00000010 +#define HSA_CAP_QUEUE_IDLE_EVENT 0x00000020 +#define HSA_CAP_VA_LIMIT 0x00000040 +#define HSA_CAP_WATCH_POINTS_SUPPORTED 0x00000080 +#define HSA_CAP_WATCH_POINTS_TOTALBITS_MASK 0x00000f00 +#define HSA_CAP_WATCH_POINTS_TOTALBITS_SHIFT 8 +#define HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK 0x00003000 +#define HSA_CAP_DOORBELL_TYPE_TOTALBITS_SHIFT 12 + +#define HSA_CAP_DOORBELL_TYPE_PRE_1_0 0x0 +#define HSA_CAP_DOORBELL_TYPE_1_0 0x1 +#define HSA_CAP_DOORBELL_TYPE_2_0 0x2 +#define HSA_CAP_AQL_QUEUE_DOUBLE_MAP 0x00004000 + +/* Old buggy user mode depends on this being 0 */ +#define HSA_CAP_RESERVED_WAS_SRAM_EDCSUPPORTED 0x00080000 + +#define HSA_CAP_MEM_EDCSUPPORTED 0x00100000 +#define HSA_CAP_RASEVENTNOTIFY 0x00200000 +#define HSA_CAP_ASIC_REVISION_MASK 0x03c00000 +#define HSA_CAP_ASIC_REVISION_SHIFT 22 +#define HSA_CAP_SRAM_EDCSUPPORTED 0x04000000 +#define HSA_CAP_SVMAPI_SUPPORTED 0x08000000 +#define HSA_CAP_FLAGS_COHERENTHOSTACCESS 0x10000000 +#define HSA_CAP_RESERVED 0xe00f8000 + +/* Heap types in memory properties */ +#define HSA_MEM_HEAP_TYPE_SYSTEM 0 +#define HSA_MEM_HEAP_TYPE_FB_PUBLIC 1 +#define HSA_MEM_HEAP_TYPE_FB_PRIVATE 2 +#define HSA_MEM_HEAP_TYPE_GPU_GDS 3 +#define HSA_MEM_HEAP_TYPE_GPU_LDS 4 +#define HSA_MEM_HEAP_TYPE_GPU_SCRATCH 5 + +/* Flag bits in memory properties */ +#define HSA_MEM_FLAGS_HOT_PLUGGABLE 0x00000001 +#define HSA_MEM_FLAGS_NON_VOLATILE 0x00000002 +#define HSA_MEM_FLAGS_RESERVED 0xfffffffc + +/* Cache types in cache properties */ +#define HSA_CACHE_TYPE_DATA 0x00000001 +#define HSA_CACHE_TYPE_INSTRUCTION 0x00000002 +#define HSA_CACHE_TYPE_CPU 0x00000004 +#define HSA_CACHE_TYPE_HSACU 0x00000008 +#define HSA_CACHE_TYPE_RESERVED 0xfffffff0 + +/* Link types in IO link properties (matches CRAT link types) */ +#define HSA_IOLINK_TYPE_UNDEFINED 0 +#define HSA_IOLINK_TYPE_HYPERTRANSPORT 1 +#define HSA_IOLINK_TYPE_PCIEXPRESS 2 +#define HSA_IOLINK_TYPE_AMBA 3 +#define HSA_IOLINK_TYPE_MIPI 4 +#define HSA_IOLINK_TYPE_QPI_1_1 5 +#define HSA_IOLINK_TYPE_RESERVED1 6 +#define HSA_IOLINK_TYPE_RESERVED2 7 +#define HSA_IOLINK_TYPE_RAPID_IO 8 +#define HSA_IOLINK_TYPE_INFINIBAND 9 +#define HSA_IOLINK_TYPE_RESERVED3 10 +#define HSA_IOLINK_TYPE_XGMI 11 +#define HSA_IOLINK_TYPE_XGOP 12 +#define HSA_IOLINK_TYPE_GZ 13 +#define HSA_IOLINK_TYPE_ETHERNET_RDMA 14 +#define HSA_IOLINK_TYPE_RDMA_OTHER 15 +#define HSA_IOLINK_TYPE_OTHER 16 + +/* Flag bits in IO link properties (matches CRAT flags, excluding the + * bi-directional flag, which is not offially part of the CRAT spec, and + * only used internally in KFD) + */ +#define HSA_IOLINK_FLAGS_ENABLED (1 << 0) +#define HSA_IOLINK_FLAGS_NON_COHERENT (1 << 1) +#define HSA_IOLINK_FLAGS_NO_ATOMICS_32_BIT (1 << 2) +#define HSA_IOLINK_FLAGS_NO_ATOMICS_64_BIT (1 << 3) +#define HSA_IOLINK_FLAGS_NO_PEER_TO_PEER_DMA (1 << 4) +#define HSA_IOLINK_FLAGS_RESERVED 0xffffffe0 + +#endif