diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
83 files changed, 2282 insertions, 617 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index 2afecc550..260e32ef7 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -80,7 +80,7 @@ amdgpu-y += amdgpu_device.o amdgpu_doorbell_mgr.o amdgpu_kms.o \ amdgpu_umc.o smu_v11_0_i2c.o amdgpu_fru_eeprom.o amdgpu_rap.o \ amdgpu_fw_attestation.o amdgpu_securedisplay.o \ amdgpu_eeprom.o amdgpu_mca.o amdgpu_psp_ta.o amdgpu_lsdma.o \ - amdgpu_ring_mux.o amdgpu_xcp.o + amdgpu_ring_mux.o amdgpu_xcp.o amdgpu_seq64.o amdgpu-$(CONFIG_PROC_FS) += amdgpu_fdinfo.o diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 13c62a26a..79827a6dc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -109,6 +109,8 @@ #include "amdgpu_mca.h" #include "amdgpu_ras.h" #include "amdgpu_xcp.h" +#include "amdgpu_seq64.h" +#include "amdgpu_reg_state.h" #define MAX_GPU_INSTANCE 64 @@ -251,6 +253,8 @@ extern int amdgpu_seamless; extern int amdgpu_user_partt_mode; extern int amdgpu_agp; +extern int amdgpu_wbrf; + #define AMDGPU_VM_MAX_NUM_CTX 4096 #define AMDGPU_SG_THRESHOLD (256*1024*1024) #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS 3000 @@ -469,6 +473,7 @@ struct amdgpu_fpriv { struct amdgpu_vm vm; struct amdgpu_bo_va *prt_va; struct amdgpu_bo_va *csa_va; + struct amdgpu_bo_va *seq64_va; struct mutex bo_list_lock; struct idr bo_list_handles; struct amdgpu_ctx_mgr ctx_mgr; @@ -507,6 +512,31 @@ struct amdgpu_allowed_register_entry { bool grbm_indexed; }; +/** + * enum amd_reset_method - Methods for resetting AMD GPU devices + * + * @AMD_RESET_METHOD_NONE: The device will not be reset. + * @AMD_RESET_LEGACY: Method reserved for SI, CIK and VI ASICs. + * @AMD_RESET_MODE0: Reset the entire ASIC. Not currently available for the + * any device. + * @AMD_RESET_MODE1: Resets all IP blocks on the ASIC (SDMA, GFX, VCN, etc.) + * individually. Suitable only for some discrete GPU, not + * available for all ASICs. + * @AMD_RESET_MODE2: Resets a lesser level of IPs compared to MODE1. Which IPs + * are reset depends on the ASIC. Notably doesn't reset IPs + * shared with the CPU on APUs or the memory controllers (so + * VRAM is not lost). Not available on all ASICs. + * @AMD_RESET_BACO: BACO (Bus Alive, Chip Off) method powers off and on the card + * but without powering off the PCI bus. Suitable only for + * discrete GPUs. + * @AMD_RESET_PCI: Does a full bus reset using core Linux subsystem PCI reset + * and does a secondary bus reset or FLR, depending on what the + * underlying hardware supports. + * + * Methods available for AMD GPU driver for resetting the device. Not all + * methods are suitable for every device. User can override the method using + * module parameter `reset_method`. + */ enum amd_reset_method { AMD_RESET_METHOD_NONE = -1, AMD_RESET_METHOD_LEGACY = 0, @@ -586,6 +616,10 @@ struct amdgpu_asic_funcs { const struct amdgpu_video_codecs **codecs); /* encode "> 32bits" smn addressing */ u64 (*encode_ext_smn_addressing)(int ext_id); + + ssize_t (*get_reg_state)(struct amdgpu_device *adev, + enum amdgpu_reg_state reg_state, void *buf, + size_t max_size); }; /* @@ -988,6 +1022,9 @@ struct amdgpu_device { /* GDS */ struct amdgpu_gds gds; + /* for userq and VM fences */ + struct amdgpu_seq64 seq64; + /* KFD */ struct amdgpu_kfd_dev kfd; @@ -1110,6 +1147,7 @@ struct amdgpu_device { bool debug_vm; bool debug_largebar; bool debug_disable_soft_recovery; + bool debug_use_vram_fw_buf; }; static inline uint32_t amdgpu_ip_version(const struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 3d126f296..131983ed4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -138,6 +138,30 @@ static void amdgpu_amdkfd_reset_work(struct work_struct *work) amdgpu_device_gpu_recover(adev, NULL, &reset_context); } +static const struct drm_client_funcs kfd_client_funcs = { + .unregister = drm_client_release, +}; + +int amdgpu_amdkfd_drm_client_create(struct amdgpu_device *adev) +{ + int ret; + + if (!adev->kfd.init_complete || adev->kfd.client.dev) + return 0; + + ret = drm_client_init(&adev->ddev, &adev->kfd.client, "kfd", + &kfd_client_funcs); + if (ret) { + dev_err(adev->dev, "Failed to init DRM client: %d\n", + ret); + return ret; + } + + drm_client_register(&adev->kfd.client); + + return 0; +} + void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) { int i; @@ -708,35 +732,6 @@ bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid) return false; } -int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev, - uint16_t vmid) -{ - if (adev->family == AMDGPU_FAMILY_AI) { - int i; - - for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) - amdgpu_gmc_flush_gpu_tlb(adev, vmid, i, 0); - } else { - amdgpu_gmc_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB(0), 0); - } - - return 0; -} - -int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev, - uint16_t pasid, - enum TLB_FLUSH_TYPE flush_type, - uint32_t inst) -{ - bool all_hub = false; - - if (adev->family == AMDGPU_FAMILY_AI || - adev->family == AMDGPU_FAMILY_RV) - all_hub = true; - - return amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, flush_type, all_hub, inst); -} - bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev) { return adev->have_atomics_support; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index dac983da9..27c61c535 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -33,6 +33,7 @@ #include <linux/mmu_notifier.h> #include <linux/memremap.h> #include <kgd_kfd_interface.h> +#include <drm/drm_client.h> #include "amdgpu_sync.h" #include "amdgpu_vm.h" #include "amdgpu_xcp.h" @@ -83,6 +84,7 @@ struct kgd_mem { struct amdgpu_sync sync; + uint32_t gem_handle; bool aql_queue; bool is_imported; }; @@ -105,6 +107,9 @@ struct amdgpu_kfd_dev { /* HMM page migration MEMORY_DEVICE_PRIVATE mapping */ struct dev_pagemap pgmap; + + /* Client for KFD BO GEM handle allocations */ + struct drm_client_dev client; }; enum kgd_engine_type { @@ -162,11 +167,6 @@ int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev, uint32_t *ib_cmd, uint32_t ib_len); 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, - uint32_t inst); bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid); @@ -182,6 +182,8 @@ int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev, struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, struct mm_struct *mm, struct svm_range_bo *svm_bo); + +int amdgpu_amdkfd_drm_client_create(struct amdgpu_device *adev); #if defined(CONFIG_DEBUG_FS) int kfd_debugfs_kfd_mem_limits(struct seq_file *m, void *data); #endif @@ -301,7 +303,7 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv); int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv); -void amdgpu_amdkfd_gpuvm_dmaunmap_mem(struct kgd_mem *mem, void *drm_priv); +int amdgpu_amdkfd_gpuvm_dmaunmap_mem(struct kgd_mem *mem, void *drm_priv); int amdgpu_amdkfd_gpuvm_sync_memory( struct amdgpu_device *adev, struct kgd_mem *mem, bool intr); int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_mem *mem, @@ -311,14 +313,13 @@ void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem); int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_bo *bo); int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info, - struct dma_fence **ef); + struct dma_fence __rcu **ef); 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 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_gpuvm_import_dmabuf_fd(struct amdgpu_device *adev, int fd, + uint64_t va, void *drm_priv, + struct kgd_mem **mem, uint64_t *size, + uint64_t *mmap_offset); int amdgpu_amdkfd_gpuvm_export_dmabuf(struct kgd_mem *mem, struct dma_buf **dmabuf); void amdgpu_amdkfd_debug_mem_fence(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c index 625db444d..3a3f3ce09 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c @@ -200,7 +200,7 @@ int kgd_arcturus_hqd_sdma_dump(struct amdgpu_device *adev, #undef HQD_N_REGS #define HQD_N_REGS (19+6+7+10) - *dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL); + *dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL); if (*dump == NULL) return -ENOMEM; @@ -290,7 +290,7 @@ static int suspend_resume_compute_scheduler(struct amdgpu_device *adev, bool sus for (i = 0; i < adev->gfx.num_compute_rings; i++) { struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; - if (!(ring && ring->sched.thread)) + if (!amdgpu_ring_sched_ready(ring)) continue; /* stop secheduler and drain ring. */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c index f6598b9e4..a5c7259cf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c @@ -141,7 +141,7 @@ static int kgd_gfx_v9_4_3_hqd_sdma_dump(struct amdgpu_device *adev, (*dump)[i++][1] = RREG32(addr); \ } while (0) - *dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL); + *dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL); if (*dump == NULL) return -ENOMEM; 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 6bf448ab3..ca4a6b828 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -214,7 +214,7 @@ static int kgd_hqd_dump(struct amdgpu_device *adev, (*dump)[i++][1] = RREG32(addr); \ } while (0) - *dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL); + *dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL); if (*dump == NULL) return -ENOMEM; @@ -301,7 +301,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev, #undef HQD_N_REGS #define HQD_N_REGS (19+4) - *dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL); + *dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL); if (*dump == NULL) return -ENOMEM; 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 cd06e4a6d..0f3e2944e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -238,7 +238,7 @@ static int kgd_hqd_dump(struct amdgpu_device *adev, (*dump)[i++][1] = RREG32(addr); \ } while (0) - *dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL); + *dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL); if (*dump == NULL) return -ENOMEM; @@ -324,7 +324,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev, #undef HQD_N_REGS #define HQD_N_REGS (19+4+2+3+7) - *dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL); + *dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL); if (*dump == NULL) return -ENOMEM; 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 00fbc0f44..5a35a8ca8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -363,7 +363,7 @@ int kgd_gfx_v9_hqd_dump(struct amdgpu_device *adev, (*dump)[i++][1] = RREG32(addr); \ } while (0) - *dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL); + *dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL); if (*dump == NULL) return -ENOMEM; @@ -460,7 +460,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev, #undef HQD_N_REGS #define HQD_N_REGS (19+6+7+10) - *dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL); + *dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL); if (*dump == NULL) return -ENOMEM; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index 41fbc4fd0..daa66eb4f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -25,6 +25,7 @@ #include <linux/pagemap.h> #include <linux/sched/mm.h> #include <linux/sched/task.h> +#include <linux/fdtable.h> #include <drm/ttm/ttm_tt.h> #include <drm/drm_exec.h> @@ -806,13 +807,22 @@ kfd_mem_dmaunmap_attachment(struct kgd_mem *mem, static int kfd_mem_export_dmabuf(struct kgd_mem *mem) { if (!mem->dmabuf) { - struct dma_buf *ret = amdgpu_gem_prime_export( - &mem->bo->tbo.base, + struct amdgpu_device *bo_adev; + struct dma_buf *dmabuf; + int r, fd; + + bo_adev = amdgpu_ttm_adev(mem->bo->tbo.bdev); + r = drm_gem_prime_handle_to_fd(&bo_adev->ddev, bo_adev->kfd.client.file, + mem->gem_handle, mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ? - DRM_RDWR : 0); - if (IS_ERR(ret)) - return PTR_ERR(ret); - mem->dmabuf = ret; + DRM_RDWR : 0, &fd); + if (r) + return r; + dmabuf = dma_buf_get(fd); + close_fd(fd); + if (WARN_ON_ONCE(IS_ERR(dmabuf))) + return PTR_ERR(dmabuf); + mem->dmabuf = dmabuf; } return 0; @@ -1137,7 +1147,7 @@ static int reserve_bo_and_vm(struct kgd_mem *mem, ctx->n_vms = 1; ctx->sync = &mem->sync; - drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT); + drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0); drm_exec_until_all_locked(&ctx->exec) { ret = amdgpu_vm_lock_pd(vm, &ctx->exec, 2); drm_exec_retry_on_contention(&ctx->exec); @@ -1176,7 +1186,7 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem, int ret; ctx->sync = &mem->sync; - drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT); + drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0); drm_exec_until_all_locked(&ctx->exec) { ctx->n_vms = 0; list_for_each_entry(entry, &mem->attachments, list) { @@ -1384,7 +1394,6 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, amdgpu_amdkfd_restore_userptr_worker); *process_info = info; - *ef = dma_fence_get(&info->eviction_fence->base); } vm->process_info = *process_info; @@ -1415,6 +1424,8 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, list_add_tail(&vm->vm_list_node, &(vm->process_info->vm_list_head)); vm->process_info->n_vms++; + + *ef = dma_fence_get(&vm->process_info->eviction_fence->base); mutex_unlock(&vm->process_info->lock); return 0; @@ -1426,10 +1437,7 @@ validate_pd_fail: reserve_pd_fail: vm->process_info = NULL; if (info) { - /* Two fence references: one in info and one in *ef */ dma_fence_put(&info->eviction_fence->base); - dma_fence_put(*ef); - *ef = NULL; *process_info = NULL; put_pid(info->pid); create_evict_fence_fail: @@ -1623,7 +1631,8 @@ int amdgpu_amdkfd_criu_resume(void *p) goto out_unlock; } WRITE_ONCE(pinfo->block_mmu_notifications, false); - schedule_delayed_work(&pinfo->restore_userptr_work, 0); + queue_delayed_work(system_freezable_wq, + &pinfo->restore_userptr_work, 0); out_unlock: mutex_unlock(&pinfo->lock); @@ -1779,6 +1788,9 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( pr_debug("Failed to allow vma node access. ret %d\n", ret); goto err_node_allow; } + ret = drm_gem_handle_create(adev->kfd.client.file, gobj, &(*mem)->gem_handle); + if (ret) + goto err_gem_handle_create; bo = gem_to_amdgpu_bo(gobj); if (bo_type == ttm_bo_type_sg) { bo->tbo.sg = sg; @@ -1830,6 +1842,8 @@ allocate_init_user_pages_failed: err_pin_bo: err_validate_bo: remove_kgd_mem_from_kfd_bo_list(*mem, avm->process_info); + drm_gem_handle_delete(adev->kfd.client.file, (*mem)->gem_handle); +err_gem_handle_create: drm_vma_node_revoke(&gobj->vma_node, drm_priv); err_node_allow: /* Don't unreserve system mem limit twice */ @@ -1837,6 +1851,7 @@ err_node_allow: err_bo_create: amdgpu_amdkfd_unreserve_mem_limit(adev, aligned_size, flags, xcp_id); err_reserve_limit: + amdgpu_sync_free(&(*mem)->sync); mutex_destroy(&(*mem)->lock); if (gobj) drm_gem_object_put(gobj); @@ -1942,8 +1957,11 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( /* Free the BO*/ drm_vma_node_revoke(&mem->bo->tbo.base.vma_node, drm_priv); - if (mem->dmabuf) + drm_gem_handle_delete(adev->kfd.client.file, mem->gem_handle); + if (mem->dmabuf) { dma_buf_put(mem->dmabuf); + mem->dmabuf = NULL; + } mutex_destroy(&mem->lock); /* If this releases the last reference, it will end up calling @@ -2068,21 +2086,35 @@ out: return ret; } -void amdgpu_amdkfd_gpuvm_dmaunmap_mem(struct kgd_mem *mem, void *drm_priv) +int amdgpu_amdkfd_gpuvm_dmaunmap_mem(struct kgd_mem *mem, void *drm_priv) { struct kfd_mem_attachment *entry; struct amdgpu_vm *vm; + int ret; vm = drm_priv_to_vm(drm_priv); mutex_lock(&mem->lock); + ret = amdgpu_bo_reserve(mem->bo, true); + if (ret) + goto out; + list_for_each_entry(entry, &mem->attachments, list) { - if (entry->bo_va->base.vm == vm) - kfd_mem_dmaunmap_attachment(mem, entry); + if (entry->bo_va->base.vm != vm) + continue; + if (entry->bo_va->base.bo->tbo.ttm && + !entry->bo_va->base.bo->tbo.ttm->sg) + continue; + + kfd_mem_dmaunmap_attachment(mem, entry); } + amdgpu_bo_unreserve(mem->bo); +out: mutex_unlock(&mem->lock); + + return ret; } int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( @@ -2295,34 +2327,26 @@ int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev, return 0; } -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) +static int import_obj_create(struct amdgpu_device *adev, + struct dma_buf *dma_buf, + struct drm_gem_object *obj, + uint64_t va, void *drm_priv, + struct kgd_mem **mem, uint64_t *size, + uint64_t *mmap_offset) { struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); - struct drm_gem_object *obj; struct amdgpu_bo *bo; int ret; - obj = amdgpu_gem_prime_import(adev_to_drm(adev), dma_buf); - if (IS_ERR(obj)) - return PTR_ERR(obj); - bo = gem_to_amdgpu_bo(obj); if (!(bo->preferred_domains & (AMDGPU_GEM_DOMAIN_VRAM | - AMDGPU_GEM_DOMAIN_GTT))) { + AMDGPU_GEM_DOMAIN_GTT))) /* Only VRAM and GTT BOs are supported */ - ret = -EINVAL; - goto err_put_obj; - } + return -EINVAL; *mem = kzalloc(sizeof(struct kgd_mem), GFP_KERNEL); - if (!*mem) { - ret = -ENOMEM; - goto err_put_obj; - } + if (!*mem) + return -ENOMEM; ret = drm_vma_node_allow(&obj->vma_node, drm_priv); if (ret) @@ -2372,8 +2396,41 @@ err_remove_mem: drm_vma_node_revoke(&obj->vma_node, drm_priv); err_free_mem: kfree(*mem); + return ret; +} + +int amdgpu_amdkfd_gpuvm_import_dmabuf_fd(struct amdgpu_device *adev, int fd, + uint64_t va, void *drm_priv, + struct kgd_mem **mem, uint64_t *size, + uint64_t *mmap_offset) +{ + struct drm_gem_object *obj; + uint32_t handle; + int ret; + + ret = drm_gem_prime_fd_to_handle(&adev->ddev, adev->kfd.client.file, fd, + &handle); + if (ret) + return ret; + obj = drm_gem_object_lookup(adev->kfd.client.file, handle); + if (!obj) { + ret = -EINVAL; + goto err_release_handle; + } + + ret = import_obj_create(adev, obj->dma_buf, obj, va, drm_priv, mem, size, + mmap_offset); + if (ret) + goto err_put_obj; + + (*mem)->gem_handle = handle; + + return 0; + err_put_obj: drm_gem_object_put(obj); +err_release_handle: + drm_gem_handle_delete(adev->kfd.client.file, handle); return ret; } @@ -2426,7 +2483,8 @@ int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni, KFD_QUEUE_EVICTION_TRIGGER_USERPTR); if (r) pr_err("Failed to quiesce KFD\n"); - schedule_delayed_work(&process_info->restore_userptr_work, + queue_delayed_work(system_freezable_wq, + &process_info->restore_userptr_work, msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS)); } mutex_unlock(&process_info->notifier_lock); @@ -2552,7 +2610,7 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) amdgpu_sync_create(&sync); - drm_exec_init(&exec, 0); + drm_exec_init(&exec, 0, 0); /* Reserve all BOs and page tables for validation */ drm_exec_until_all_locked(&exec) { /* Reserve all the page directories */ @@ -2749,7 +2807,8 @@ unlock_out: /* If validation failed, reschedule another attempt */ if (evicted_bos) { - schedule_delayed_work(&process_info->restore_userptr_work, + queue_delayed_work(system_freezable_wq, + &process_info->restore_userptr_work, msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS)); kfd_smi_event_queue_restore_rescheduled(mm); @@ -2758,6 +2817,23 @@ unlock_out: put_task_struct(usertask); } +static void replace_eviction_fence(struct dma_fence __rcu **ef, + struct dma_fence *new_ef) +{ + struct dma_fence *old_ef = rcu_replace_pointer(*ef, new_ef, true + /* protected by process_info->lock */); + + /* If we're replacing an unsignaled eviction fence, that fence will + * never be signaled, and if anyone is still waiting on that fence, + * they will hang forever. This should never happen. We should only + * replace the fence in restore_work that only gets scheduled after + * eviction work signaled the fence. + */ + WARN_ONCE(!dma_fence_is_signaled(old_ef), + "Replacing unsignaled eviction fence"); + dma_fence_put(old_ef); +} + /** amdgpu_amdkfd_gpuvm_restore_process_bos - Restore all BOs for the given * KFD process identified by process_info * @@ -2776,12 +2852,11 @@ unlock_out: * 7. Add fence to all PD and PT BOs. * 8. Unreserve all BOs */ -int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) +int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu **ef) { struct amdkfd_process_info *process_info = info; struct amdgpu_vm *peer_vm; struct kgd_mem *mem; - struct amdgpu_amdkfd_fence *new_fence; struct list_head duplicate_save; struct amdgpu_sync sync_obj; unsigned long failed_size = 0; @@ -2793,7 +2868,7 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) mutex_lock(&process_info->lock); - drm_exec_init(&exec, 0); + drm_exec_init(&exec, 0, 0); drm_exec_until_all_locked(&exec) { list_for_each_entry(peer_vm, &process_info->vm_list_head, vm_list_node) { @@ -2825,12 +2900,6 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) if (ret) goto validate_map_fail; - ret = process_sync_pds_resv(process_info, &sync_obj); - if (ret) { - pr_debug("Memory eviction: Failed to sync to PD BO moving fence. Try again\n"); - goto validate_map_fail; - } - /* Validate BOs and map them to GPUVM (update VM page tables). */ list_for_each_entry(mem, &process_info->kfd_bo_list, validate_list) { @@ -2881,6 +2950,19 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) if (failed_size) pr_debug("0x%lx/0x%lx in system\n", failed_size, total_size); + /* Update mappings not managed by KFD */ + list_for_each_entry(peer_vm, &process_info->vm_list_head, + vm_list_node) { + struct amdgpu_device *adev = amdgpu_ttm_adev( + peer_vm->root.bo->tbo.bdev); + + ret = amdgpu_vm_handle_moved(adev, peer_vm, &exec.ticket); + if (ret) { + pr_debug("Memory eviction: handle moved failed. Try again\n"); + goto validate_map_fail; + } + } + /* Update page directories */ ret = process_update_pds(process_info, &sync_obj); if (ret) { @@ -2888,25 +2970,47 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) goto validate_map_fail; } + /* Sync with fences on all the page tables. They implicitly depend on any + * move fences from amdgpu_vm_handle_moved above. + */ + ret = process_sync_pds_resv(process_info, &sync_obj); + if (ret) { + pr_debug("Memory eviction: Failed to sync to PD BO moving fence. Try again\n"); + goto validate_map_fail; + } + /* Wait for validate and PT updates to finish */ amdgpu_sync_wait(&sync_obj, false); - /* Release old eviction fence and create new one, because fence only - * goes from unsignaled to signaled, fence cannot be reused. - * Use context and mm from the old fence. + /* The old eviction fence may be unsignaled if restore happens + * after a GPU reset or suspend/resume. Keep the old fence in that + * case. Otherwise release the old eviction fence and create new + * one, because fence only goes from unsignaled to signaled once + * and cannot be reused. Use context and mm from the old fence. + * + * If an old eviction fence signals after this check, that's OK. + * Anyone signaling an eviction fence must stop the queues first + * and schedule another restore worker. */ - new_fence = amdgpu_amdkfd_fence_create( + if (dma_fence_is_signaled(&process_info->eviction_fence->base)) { + struct amdgpu_amdkfd_fence *new_fence = + amdgpu_amdkfd_fence_create( process_info->eviction_fence->base.context, process_info->eviction_fence->mm, NULL); - if (!new_fence) { - pr_err("Failed to create eviction fence\n"); - ret = -ENOMEM; - goto validate_map_fail; + + if (!new_fence) { + pr_err("Failed to create eviction fence\n"); + ret = -ENOMEM; + goto validate_map_fail; + } + dma_fence_put(&process_info->eviction_fence->base); + process_info->eviction_fence = new_fence; + replace_eviction_fence(ef, dma_fence_get(&new_fence->base)); + } else { + WARN_ONCE(*ef != &process_info->eviction_fence->base, + "KFD eviction fence doesn't match KGD process_info"); } - dma_fence_put(&process_info->eviction_fence->base); - process_info->eviction_fence = new_fence; - *ef = dma_fence_get(&new_fence->base); /* Attach new eviction fence to all BOs except pinned ones */ list_for_each_entry(mem, &process_info->kfd_bo_list, validate_list) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index 7473a42f7..9caba1031 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -103,7 +103,7 @@ int amdgpu_connector_get_monitor_bpc(struct drm_connector *connector) struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector); struct amdgpu_connector_atom_dig *dig_connector; int bpc = 8; - unsigned mode_clock, max_tmds_clock; + unsigned int mode_clock, max_tmds_clock; switch (connector->connector_type) { case DRM_MODE_CONNECTOR_DVII: @@ -255,6 +255,7 @@ struct edid *amdgpu_connector_edid(struct drm_connector *connector) return amdgpu_connector->edid; } else if (edid_blob) { struct edid *edid = kmemdup(edid_blob->data, edid_blob->length, GFP_KERNEL); + if (edid) amdgpu_connector->edid = edid; } @@ -581,6 +582,7 @@ static int amdgpu_connector_set_property(struct drm_connector *connector, amdgpu_encoder = to_amdgpu_encoder(connector->encoder); } else { const struct drm_connector_helper_funcs *connector_funcs = connector->helper_private; + amdgpu_encoder = to_amdgpu_encoder(connector_funcs->best_encoder(connector)); } @@ -797,6 +799,7 @@ static int amdgpu_connector_set_lcd_property(struct drm_connector *connector, amdgpu_encoder = to_amdgpu_encoder(connector->encoder); else { const struct drm_connector_helper_funcs *connector_funcs = connector->helper_private; + amdgpu_encoder = to_amdgpu_encoder(connector_funcs->best_encoder(connector)); } @@ -979,6 +982,41 @@ amdgpu_connector_check_hpd_status_unchanged(struct drm_connector *connector) return false; } +static void amdgpu_connector_shared_ddc(enum drm_connector_status *status, + struct drm_connector *connector, + struct amdgpu_connector *amdgpu_connector) +{ + struct drm_connector *list_connector; + struct drm_connector_list_iter iter; + struct amdgpu_connector *list_amdgpu_connector; + struct drm_device *dev = connector->dev; + struct amdgpu_device *adev = drm_to_adev(dev); + + if (amdgpu_connector->shared_ddc && *status == connector_status_connected) { + drm_connector_list_iter_begin(dev, &iter); + drm_for_each_connector_iter(list_connector, + &iter) { + if (connector == list_connector) + continue; + list_amdgpu_connector = to_amdgpu_connector(list_connector); + if (list_amdgpu_connector->shared_ddc && + list_amdgpu_connector->ddc_bus->rec.i2c_id == + amdgpu_connector->ddc_bus->rec.i2c_id) { + /* cases where both connectors are digital */ + if (list_connector->connector_type != DRM_MODE_CONNECTOR_VGA) { + /* hpd is our only option in this case */ + if (!amdgpu_display_hpd_sense(adev, + amdgpu_connector->hpd.hpd)) { + amdgpu_connector_free_edid(connector); + *status = connector_status_disconnected; + } + } + } + } + drm_connector_list_iter_end(&iter); + } +} + /* * DVI is complicated * Do a DDC probe, if DDC probe passes, get the full EDID so @@ -1065,32 +1103,7 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force) * DDC line. The latter is more complex because with DVI<->HDMI adapters * you don't really know what's connected to which port as both are digital. */ - if (amdgpu_connector->shared_ddc && (ret == connector_status_connected)) { - struct drm_connector *list_connector; - struct drm_connector_list_iter iter; - struct amdgpu_connector *list_amdgpu_connector; - - drm_connector_list_iter_begin(dev, &iter); - drm_for_each_connector_iter(list_connector, - &iter) { - if (connector == list_connector) - continue; - list_amdgpu_connector = to_amdgpu_connector(list_connector); - if (list_amdgpu_connector->shared_ddc && - (list_amdgpu_connector->ddc_bus->rec.i2c_id == - amdgpu_connector->ddc_bus->rec.i2c_id)) { - /* cases where both connectors are digital */ - if (list_connector->connector_type != DRM_MODE_CONNECTOR_VGA) { - /* hpd is our only option in this case */ - if (!amdgpu_display_hpd_sense(adev, amdgpu_connector->hpd.hpd)) { - amdgpu_connector_free_edid(connector); - ret = connector_status_disconnected; - } - } - } - } - drm_connector_list_iter_end(&iter); - } + amdgpu_connector_shared_ddc(&ret, connector, amdgpu_connector); } } @@ -1192,6 +1205,7 @@ amdgpu_connector_dvi_encoder(struct drm_connector *connector) static void amdgpu_connector_dvi_force(struct drm_connector *connector) { struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector); + if (connector->force == DRM_FORCE_ON) amdgpu_connector->use_digital = false; if (connector->force == DRM_FORCE_ON_DIGITAL) @@ -1426,6 +1440,7 @@ amdgpu_connector_dp_detect(struct drm_connector *connector, bool force) ret = connector_status_connected; else if (amdgpu_connector->dac_load_detect) { /* try load detection */ const struct drm_encoder_helper_funcs *encoder_funcs = encoder->helper_private; + ret = encoder_funcs->detect(encoder, connector); } } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index e50be6500..116d3756c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -66,7 +66,7 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, amdgpu_sync_create(&p->sync); drm_exec_init(&p->exec, DRM_EXEC_INTERRUPTIBLE_WAIT | - DRM_EXEC_IGNORE_DUPLICATES); + DRM_EXEC_IGNORE_DUPLICATES, 0); return 0; } @@ -819,7 +819,7 @@ retry: p->bytes_moved += ctx.bytes_moved; if (!amdgpu_gmc_vram_full_visible(&adev->gmc) && - amdgpu_bo_in_cpu_visible_vram(bo)) + amdgpu_res_cpu_visible(adev, bo->tbo.resource)) p->bytes_moved_vis += ctx.bytes_moved; if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) { @@ -870,9 +870,9 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, struct amdgpu_bo *bo = e->bo; int i; - e->user_pages = kvmalloc_array(bo->tbo.ttm->num_pages, - sizeof(struct page *), - GFP_KERNEL | __GFP_ZERO); + e->user_pages = kvcalloc(bo->tbo.ttm->num_pages, + sizeof(struct page *), + GFP_KERNEL); if (!e->user_pages) { DRM_ERROR("kvmalloc_array failure\n"); r = -ENOMEM; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c index 720011019..796fa6f14 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c @@ -70,7 +70,7 @@ int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct drm_exec exec; int r; - drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT); + drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0); drm_exec_until_all_locked(&exec) { r = amdgpu_vm_lock_pd(vm, &exec, 0); if (likely(!r)) @@ -110,7 +110,7 @@ int amdgpu_unmap_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct drm_exec exec; int r; - drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT); + drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0); drm_exec_until_all_locked(&exec) { r = amdgpu_vm_lock_pd(vm, &exec, 0); if (likely(!r)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index e2ae9ba14..5cb33ac99 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -73,10 +73,10 @@ amdgpu_ctx_to_drm_sched_prio(int32_t ctx_prio) return DRM_SCHED_PRIORITY_NORMAL; case AMDGPU_CTX_PRIORITY_VERY_LOW: - return DRM_SCHED_PRIORITY_MIN; + return DRM_SCHED_PRIORITY_LOW; case AMDGPU_CTX_PRIORITY_LOW: - return DRM_SCHED_PRIORITY_MIN; + return DRM_SCHED_PRIORITY_LOW; case AMDGPU_CTX_PRIORITY_NORMAL: return DRM_SCHED_PRIORITY_NORMAL; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index 8d4a3ff65..1afbb2e93 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -540,7 +540,11 @@ static ssize_t amdgpu_debugfs_regs_pcie_read(struct file *f, char __user *buf, while (size) { uint32_t value; - value = RREG32_PCIE(*pos); + if (upper_32_bits(*pos)) + value = RREG32_PCIE_EXT(*pos); + else + value = RREG32_PCIE(*pos); + r = put_user(value, (uint32_t *)buf); if (r) goto out; @@ -600,7 +604,10 @@ static ssize_t amdgpu_debugfs_regs_pcie_write(struct file *f, const char __user if (r) goto out; - WREG32_PCIE(*pos, value); + if (upper_32_bits(*pos)) + WREG32_PCIE_EXT(*pos, value); + else + WREG32_PCIE(*pos, value); result += 4; buf += 4; @@ -1671,9 +1678,9 @@ static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused) for (i = 0; i < AMDGPU_MAX_RINGS; i++) { struct amdgpu_ring *ring = adev->rings[i]; - if (!ring || !ring->sched.thread) + if (!amdgpu_ring_sched_ready(ring)) continue; - kthread_park(ring->sched.thread); + drm_sched_wqueue_stop(&ring->sched); } seq_puts(m, "run ib test:\n"); @@ -1687,9 +1694,9 @@ static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused) for (i = 0; i < AMDGPU_MAX_RINGS; i++) { struct amdgpu_ring *ring = adev->rings[i]; - if (!ring || !ring->sched.thread) + if (!amdgpu_ring_sched_ready(ring)) continue; - kthread_unpark(ring->sched.thread); + drm_sched_wqueue_start(&ring->sched); } up_write(&adev->reset_domain->sem); @@ -1909,7 +1916,8 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val) ring = adev->rings[val]; - if (!ring || !ring->funcs->preempt_ib || !ring->sched.thread) + if (!amdgpu_ring_sched_ready(ring) || + !ring->funcs->preempt_ib) return -EINVAL; /* the last preemption failed */ @@ -1927,7 +1935,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val) goto pro_end; /* stop the scheduler */ - kthread_park(ring->sched.thread); + drm_sched_wqueue_stop(&ring->sched); /* preempt the IB */ r = amdgpu_ring_preempt_ib(ring); @@ -1961,7 +1969,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val) failure: /* restart the scheduler */ - kthread_unpark(ring->sched.thread); + drm_sched_wqueue_start(&ring->sched); up_read(&adev->reset_domain->sem); @@ -2146,6 +2154,8 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev) amdgpu_debugfs_firmware_init(adev); amdgpu_ta_if_debugfs_init(adev); + amdgpu_debugfs_mes_event_log_init(adev); + #if defined(CONFIG_DRM_AMD_DC) if (adev->dc_enabled) dtn_debugfs_init(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h index 371a6f0de..0425432d8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h @@ -32,3 +32,5 @@ void amdgpu_debugfs_fini(struct amdgpu_device *adev); void amdgpu_debugfs_fence_init(struct amdgpu_device *adev); void amdgpu_debugfs_firmware_init(struct amdgpu_device *adev); void amdgpu_debugfs_gem_init(struct amdgpu_device *adev); +void amdgpu_debugfs_mes_event_log_init(struct amdgpu_device *adev); + diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 7f48c7ec4..d0afb9ba3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -162,6 +162,65 @@ static ssize_t amdgpu_device_get_pcie_replay_count(struct device *dev, static DEVICE_ATTR(pcie_replay_count, 0444, amdgpu_device_get_pcie_replay_count, NULL); +static ssize_t amdgpu_sysfs_reg_state_get(struct file *f, struct kobject *kobj, + struct bin_attribute *attr, char *buf, + loff_t ppos, size_t count) +{ + struct device *dev = kobj_to_dev(kobj); + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(ddev); + ssize_t bytes_read; + + switch (ppos) { + case AMDGPU_SYS_REG_STATE_XGMI: + bytes_read = amdgpu_asic_get_reg_state( + adev, AMDGPU_REG_STATE_TYPE_XGMI, buf, count); + break; + case AMDGPU_SYS_REG_STATE_WAFL: + bytes_read = amdgpu_asic_get_reg_state( + adev, AMDGPU_REG_STATE_TYPE_WAFL, buf, count); + break; + case AMDGPU_SYS_REG_STATE_PCIE: + bytes_read = amdgpu_asic_get_reg_state( + adev, AMDGPU_REG_STATE_TYPE_PCIE, buf, count); + break; + case AMDGPU_SYS_REG_STATE_USR: + bytes_read = amdgpu_asic_get_reg_state( + adev, AMDGPU_REG_STATE_TYPE_USR, buf, count); + break; + case AMDGPU_SYS_REG_STATE_USR_1: + bytes_read = amdgpu_asic_get_reg_state( + adev, AMDGPU_REG_STATE_TYPE_USR_1, buf, count); + break; + default: + return -EINVAL; + } + + return bytes_read; +} + +BIN_ATTR(reg_state, 0444, amdgpu_sysfs_reg_state_get, NULL, + AMDGPU_SYS_REG_STATE_END); + +int amdgpu_reg_state_sysfs_init(struct amdgpu_device *adev) +{ + int ret; + + if (!amdgpu_asic_get_reg_state_supported(adev)) + return 0; + + ret = sysfs_create_bin_file(&adev->dev->kobj, &bin_attr_reg_state); + + return ret; +} + +void amdgpu_reg_state_sysfs_fini(struct amdgpu_device *adev) +{ + if (!amdgpu_asic_get_reg_state_supported(adev)) + return; + sysfs_remove_bin_file(&adev->dev->kobj, &bin_attr_reg_state); +} + /** * DOC: board_info * @@ -1541,7 +1600,7 @@ bool amdgpu_device_seamless_boot_supported(struct amdgpu_device *adev) if (adev->mman.keep_stolen_vga_memory) return false; - return adev->ip_versions[DCE_HWIP][0] >= IP_VERSION(3, 0, 0); + return amdgpu_ip_version(adev, DCE_HWIP, 0) >= IP_VERSION(3, 0, 0); } /* @@ -1552,11 +1611,15 @@ bool amdgpu_device_seamless_boot_supported(struct amdgpu_device *adev) * https://edc.intel.com/content/www/us/en/design/products/platforms/details/raptor-lake-s/13th-generation-core-processors-datasheet-volume-1-of-2/005/pci-express-support/ * https://gitlab.freedesktop.org/drm/amd/-/issues/2663 */ -static bool amdgpu_device_pcie_dynamic_switching_supported(void) +static bool amdgpu_device_pcie_dynamic_switching_supported(struct amdgpu_device *adev) { #if IS_ENABLED(CONFIG_X86) struct cpuinfo_x86 *c = &cpu_data(0); + /* eGPU change speeds based on USB4 fabric conditions */ + if (dev_is_removable(adev->dev)) + return true; + if (c->x86_vendor == X86_VENDOR_INTEL) return false; #endif @@ -2389,7 +2452,7 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) adev->pm.pp_feature &= ~PP_GFXOFF_MASK; if (amdgpu_sriov_vf(adev) && adev->asic_type == CHIP_SIENNA_CICHLID) adev->pm.pp_feature &= ~PP_OVERDRIVE_MASK; - if (!amdgpu_device_pcie_dynamic_switching_supported()) + if (!amdgpu_device_pcie_dynamic_switching_supported(adev)) adev->pm.pp_feature &= ~PP_PCIE_DPM_MASK; total = true; @@ -2567,7 +2630,7 @@ static int amdgpu_device_init_schedulers(struct amdgpu_device *adev) break; } - r = drm_sched_init(&ring->sched, &amdgpu_sched_ops, + r = drm_sched_init(&ring->sched, &amdgpu_sched_ops, NULL, DRM_SCHED_PRIORITY_COUNT, ring->num_hw_submission, 0, timeout, adev->reset_domain->wq, @@ -2670,6 +2733,12 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) goto init_failed; } } + + r = amdgpu_seq64_init(adev); + if (r) { + DRM_ERROR("allocate seq64 failed %d\n", r); + goto init_failed; + } } } @@ -3132,6 +3201,7 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) amdgpu_device_wb_fini(adev); amdgpu_device_mem_scratch_fini(adev); amdgpu_ib_pool_fini(adev); + amdgpu_seq64_fini(adev); } r = adev->ip_blocks[i].version->funcs->sw_fini((void *)adev); @@ -4202,6 +4272,7 @@ fence_driver_init: "Could not create amdgpu board attributes\n"); amdgpu_fru_sysfs_init(adev); + amdgpu_reg_state_sysfs_init(adev); if (IS_ENABLED(CONFIG_PERF_EVENTS)) r = amdgpu_pmu_init(adev); @@ -4324,6 +4395,8 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev) sysfs_remove_files(&adev->dev->kobj, amdgpu_dev_attributes); amdgpu_fru_sysfs_fini(adev); + amdgpu_reg_state_sysfs_fini(adev); + /* disable ras feature must before hw fini */ amdgpu_ras_pre_fini(adev); @@ -4451,6 +4524,8 @@ int amdgpu_device_prepare(struct drm_device *dev) if (r) goto unprepare; + flush_delayed_work(&adev->gfx.gfx_off_delay_work); + for (i = 0; i < adev->num_ip_blocks; i++) { if (!adev->ip_blocks[i].status.valid) continue; @@ -4954,7 +5029,7 @@ bool amdgpu_device_has_job_running(struct amdgpu_device *adev) for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = adev->rings[i]; - if (!ring || !ring->sched.thread) + if (!amdgpu_ring_sched_ready(ring)) continue; spin_lock(&ring->sched.job_list_lock); @@ -5093,7 +5168,7 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev, for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = adev->rings[i]; - if (!ring || !ring->sched.thread) + if (!amdgpu_ring_sched_ready(ring)) continue; /* Clear job fence from fence drv to avoid force_completion @@ -5560,7 +5635,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = tmp_adev->rings[i]; - if (!ring || !ring->sched.thread) + if (!amdgpu_ring_sched_ready(ring)) continue; drm_sched_stop(&ring->sched, job ? &job->base : NULL); @@ -5629,7 +5704,7 @@ skip_hw_reset: for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = tmp_adev->rings[i]; - if (!ring || !ring->sched.thread) + if (!amdgpu_ring_sched_ready(ring)) continue; drm_sched_start(&ring->sched, true); @@ -5691,6 +5766,39 @@ skip_sched_resume: } /** + * amdgpu_device_partner_bandwidth - find the bandwidth of appropriate partner + * + * @adev: amdgpu_device pointer + * @speed: pointer to the speed of the link + * @width: pointer to the width of the link + * + * Evaluate the hierarchy to find the speed and bandwidth capabilities of the + * first physical partner to an AMD dGPU. + * This will exclude any virtual switches and links. + */ +static void amdgpu_device_partner_bandwidth(struct amdgpu_device *adev, + enum pci_bus_speed *speed, + enum pcie_link_width *width) +{ + struct pci_dev *parent = adev->pdev; + + if (!speed || !width) + return; + + *speed = PCI_SPEED_UNKNOWN; + *width = PCIE_LNK_WIDTH_UNKNOWN; + + while ((parent = pci_upstream_bridge(parent))) { + /* skip upstream/downstream switches internal to dGPU*/ + if (parent->vendor == PCI_VENDOR_ID_ATI) + continue; + *speed = pcie_get_speed_cap(parent); + *width = pcie_get_width_cap(parent); + break; + } +} + +/** * amdgpu_device_get_pcie_info - fence pcie info about the PCIE slot * * @adev: amdgpu_device pointer @@ -5723,8 +5831,8 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev) if (adev->pm.pcie_gen_mask && adev->pm.pcie_mlw_mask) return; - pcie_bandwidth_available(adev->pdev, NULL, - &platform_speed_cap, &platform_link_width); + amdgpu_device_partner_bandwidth(adev, &platform_speed_cap, + &platform_link_width); if (adev->pm.pcie_gen_mask == 0) { /* asic caps */ @@ -5951,7 +6059,7 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = adev->rings[i]; - if (!ring || !ring->sched.thread) + if (!amdgpu_ring_sched_ready(ring)) continue; drm_sched_stop(&ring->sched, NULL); @@ -6001,6 +6109,20 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) struct amdgpu_reset_context reset_context; u32 memsize; struct list_head device_list; + struct amdgpu_hive_info *hive; + int hive_ras_recovery = 0; + struct amdgpu_ras *ras; + + /* PCI error slot reset should be skipped During RAS recovery */ + hive = amdgpu_get_xgmi_hive(adev); + if (hive) { + hive_ras_recovery = atomic_read(&hive->ras_recovery); + amdgpu_put_xgmi_hive(hive); + } + ras = amdgpu_ras_get_context(adev); + if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3)) && + ras && (atomic_read(&ras->in_recovery) || hive_ras_recovery)) + return PCI_ERS_RESULT_RECOVERED; DRM_INFO("PCI error: slot reset callback!!\n"); @@ -6079,7 +6201,7 @@ void amdgpu_pci_resume(struct pci_dev *pdev) for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = adev->rings[i]; - if (!ring || !ring->sched.thread) + if (!amdgpu_ring_sched_ready(ring)) continue; drm_sched_start(&ring->sched, true); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c index e7e87a3b2..decbbe3d4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c @@ -42,6 +42,7 @@ #include <linux/dma-fence-array.h> #include <linux/pci-p2pdma.h> #include <linux/pm_runtime.h> +#include "amdgpu_trace.h" /** * amdgpu_dma_buf_attach - &dma_buf_ops.attach implementation @@ -63,6 +64,7 @@ static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf, attach->peer2peer = false; r = pm_runtime_get_sync(adev_to_drm(adev)->dev); + trace_amdgpu_runpm_reference_dumps(1, __func__); if (r < 0) goto out; @@ -70,6 +72,7 @@ static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf, out: pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + trace_amdgpu_runpm_reference_dumps(0, __func__); return r; } @@ -90,6 +93,7 @@ static void amdgpu_dma_buf_detach(struct dma_buf *dmabuf, pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + trace_amdgpu_runpm_reference_dumps(0, __func__); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 855ab5963..64b1bb240 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -128,6 +128,7 @@ enum AMDGPU_DEBUG_MASK { AMDGPU_DEBUG_VM = BIT(0), AMDGPU_DEBUG_LARGEBAR = BIT(1), AMDGPU_DEBUG_DISABLE_GPU_SOFT_RECOVERY = BIT(2), + AMDGPU_DEBUG_USE_VRAM_FW_BUF = BIT(3), }; unsigned int amdgpu_vram_limit = UINT_MAX; @@ -209,6 +210,7 @@ int amdgpu_umsch_mm; int amdgpu_seamless = -1; /* auto */ uint amdgpu_debug_mask; int amdgpu_agp = -1; /* auto */ +int amdgpu_wbrf = -1; int amdgpu_damage_clips = -1; /* auto */ static void amdgpu_drv_delayed_reset_work_handler(struct work_struct *work); @@ -985,6 +987,22 @@ module_param_named(debug_mask, amdgpu_debug_mask, uint, 0444); MODULE_PARM_DESC(agp, "AGP (-1 = auto (default), 0 = disable, 1 = enable)"); module_param_named(agp, amdgpu_agp, int, 0444); +/** + * DOC: wbrf (int) + * Enable Wifi RFI interference mitigation feature. + * Due to electrical and mechanical constraints there may be likely interference of + * relatively high-powered harmonics of the (G-)DDR memory clocks with local radio + * module frequency bands used by Wifi 6/6e/7. To mitigate the possible RFI interference, + * with this feature enabled, PMFW will use either “shadowed P-State” or “P-State” based + * on active list of frequencies in-use (to be avoided) as part of initial setting or + * P-state transition. However, there may be potential performance impact with this + * feature enabled. + * (0 = disabled, 1 = enabled, -1 = auto (default setting, will be enabled if supported)) + */ +MODULE_PARM_DESC(wbrf, + "Enable Wifi RFI interference mitigation (0 = disabled, 1 = enabled, -1 = auto(default)"); +module_param_named(wbrf, amdgpu_wbrf, int, 0444); + /* These devices are not supported by amdgpu. * They are supported by the mach64, r128, radeon drivers */ @@ -2113,6 +2131,11 @@ static void amdgpu_init_debug_options(struct amdgpu_device *adev) pr_info("debug: soft reset for GPU recovery disabled\n"); adev->debug_disable_soft_recovery = true; } + + if (amdgpu_debug_mask & AMDGPU_DEBUG_USE_VRAM_FW_BUF) { + pr_info("debug: place fw in vram for frontdoor loading\n"); + adev->debug_use_vram_fw_buf = true; + } } static unsigned long amdgpu_fix_asic_type(struct pci_dev *pdev, unsigned long flags) @@ -2224,6 +2247,8 @@ static int amdgpu_pci_probe(struct pci_dev *pdev, pci_set_drvdata(pdev, ddev); + amdgpu_init_debug_options(adev); + ret = amdgpu_driver_load_kms(adev, flags); if (ret) goto err_pci; @@ -2243,6 +2268,10 @@ retry_init: if (ret) goto err_pci; + ret = amdgpu_amdkfd_drm_client_create(adev); + if (ret) + goto err_pci; + /* * 1. don't init fbdev on hw without DCE * 2. don't init fbdev if there are no connectors @@ -2304,8 +2333,6 @@ retry_init: amdgpu_get_secondary_funcs(adev); } - amdgpu_init_debug_options(adev); - return 0; err_pci: @@ -2424,8 +2451,11 @@ static void amdgpu_drv_delayed_reset_work_handler(struct work_struct *work) } for (i = 0; i < mgpu_info.num_dgpu; i++) { adev = mgpu_info.gpu_ins[i].adev; - if (!adev->kfd.init_complete) + if (!adev->kfd.init_complete) { + kgd2kfd_init_zone_device(adev); amdgpu_amdkfd_device_init(adev); + amdgpu_amdkfd_drm_client_create(adev); + } amdgpu_ttm_set_buffer_funcs_status(adev, true); } } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c index 5706b282a..c7df7fa34 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c @@ -97,6 +97,10 @@ void amdgpu_show_fdinfo(struct drm_printer *p, struct drm_file *file) stats.requested_visible_vram/1024UL); drm_printf(p, "amd-requested-gtt:\t%llu KiB\n", stats.requested_gtt/1024UL); + drm_printf(p, "drm-shared-vram:\t%llu KiB\n", stats.vram_shared/1024UL); + drm_printf(p, "drm-shared-gtt:\t%llu KiB\n", stats.gtt_shared/1024UL); + drm_printf(p, "drm-shared-cpu:\t%llu KiB\n", stats.cpu_shared/1024UL); + for (hw_ip = 0; hw_ip < AMDGPU_HW_IP_NUM; ++hw_ip) { if (!usage[hw_ip]) continue; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index dc2302127..70bff8cec 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -183,6 +183,7 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f, struct amd amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr, seq, flags | AMDGPU_FENCE_FLAG_INT); pm_runtime_get_noresume(adev_to_drm(adev)->dev); + trace_amdgpu_runpm_reference_dumps(1, __func__); ptr = &ring->fence_drv.fences[seq & ring->fence_drv.num_fences_mask]; if (unlikely(rcu_dereference_protected(*ptr, 1))) { struct dma_fence *old; @@ -310,6 +311,7 @@ bool amdgpu_fence_process(struct amdgpu_ring *ring) dma_fence_put(fence); pm_runtime_mark_last_busy(adev_to_drm(adev)->dev); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + trace_amdgpu_runpm_reference_dumps(0, __func__); } while (last_seq != seq); return true; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 84beeaa4d..49a5f1c73 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -203,7 +203,7 @@ static void amdgpu_gem_object_close(struct drm_gem_object *obj, struct drm_exec exec; long r; - drm_exec_init(&exec, DRM_EXEC_IGNORE_DUPLICATES); + drm_exec_init(&exec, DRM_EXEC_IGNORE_DUPLICATES, 0); drm_exec_until_all_locked(&exec) { r = drm_exec_prepare_obj(&exec, &bo->tbo.base, 1); drm_exec_retry_on_contention(&exec); @@ -739,7 +739,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, } drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT | - DRM_EXEC_IGNORE_DUPLICATES); + DRM_EXEC_IGNORE_DUPLICATES, 0); drm_exec_until_all_locked(&exec) { if (gobj) { r = drm_exec_lock_obj(&exec, gobj); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c index 57516a8c5..431ec7265 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c @@ -202,8 +202,8 @@ int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier, pr_debug("hmm range: start = 0x%lx, end = 0x%lx", hmm_range->start, hmm_range->end); - /* Assuming 128MB takes maximum 1 second to fault page address */ - timeout = max((hmm_range->end - hmm_range->start) >> 27, 1UL); + /* Assuming 64MB takes maximum 1 second to fault page address */ + timeout = max((hmm_range->end - hmm_range->start) >> 26, 1UL); timeout *= HMM_RANGE_DEFAULT_TIMEOUT; timeout = jiffies + msecs_to_jiffies(timeout); @@ -211,6 +211,7 @@ retry: hmm_range->notifier_seq = mmu_interval_read_begin(notifier); r = hmm_range_fault(hmm_range); if (unlikely(r)) { + schedule(); /* * FIXME: This timeout should encompass the retry from * mmu_interval_read_retry() as well. @@ -224,7 +225,6 @@ retry: break; hmm_range->hmm_pfns += MAX_WALK_BYTE >> PAGE_SHIFT; hmm_range->start = hmm_range->end; - schedule(); } while (hmm_range->end < end); hmm_range->start = start; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c index 82608df43..d79cb13e1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c @@ -175,7 +175,6 @@ struct amdgpu_i2c_chan *amdgpu_i2c_create(struct drm_device *dev, i2c->rec = *rec; i2c->adapter.owner = THIS_MODULE; - i2c->adapter.class = I2C_CLASS_DDC; i2c->adapter.dev.parent = dev->dev; i2c->dev = dev; i2c_set_adapdata(&i2c->adapter, i2c); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index 1f3571985..71a5cf37b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -115,7 +115,7 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm, if (!entity) return 0; - return drm_sched_job_init(&(*job)->base, entity, owner); + return drm_sched_job_init(&(*job)->base, entity, 1, owner); } int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, @@ -325,7 +325,7 @@ void amdgpu_job_stop_all_jobs_on_sched(struct drm_gpu_scheduler *sched) int i; /* Signal all jobs not yet scheduled */ - for (i = sched->num_rqs - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) { + for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs; i++) { struct drm_sched_rq *rq = sched->sched_rq[i]; spin_lock(&rq->lock); list_for_each_entry(s_entity, &rq->entities, list) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 598867919..bf4f48fe4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -1433,6 +1433,8 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, fpriv->csa_va = NULL; } + amdgpu_seq64_unmap(adev, fpriv); + pasid = fpriv->vm.pasid; pd = amdgpu_bo_ref(fpriv->vm.root.bo); if (!WARN_ON(amdgpu_bo_reserve(pd, true))) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c index 061d88f44..59fafb839 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c @@ -218,6 +218,7 @@ static void amdgpu_mca_smu_mca_bank_dump(struct amdgpu_device *adev, int idx, st int amdgpu_mca_smu_log_ras_error(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type, struct ras_err_data *err_data) { struct amdgpu_smuio_mcm_config_info mcm_info; + struct ras_err_addr err_addr = {0}; struct mca_bank_set mca_set; struct mca_bank_node *node; struct mca_bank_entry *entry; @@ -246,10 +247,18 @@ int amdgpu_mca_smu_log_ras_error(struct amdgpu_device *adev, enum amdgpu_ras_blo mcm_info.socket_id = entry->info.socket_id; mcm_info.die_id = entry->info.aid; + if (blk == AMDGPU_RAS_BLOCK__UMC) { + err_addr.err_status = entry->regs[MCA_REG_IDX_STATUS]; + err_addr.err_ipid = entry->regs[MCA_REG_IDX_IPID]; + err_addr.err_addr = entry->regs[MCA_REG_IDX_ADDR]; + } + if (type == AMDGPU_MCA_ERROR_TYPE_UE) - amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, (uint64_t)count); + amdgpu_ras_error_statistic_ue_count(err_data, + &mcm_info, &err_addr, (uint64_t)count); else - amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, (uint64_t)count); + amdgpu_ras_error_statistic_ce_count(err_data, + &mcm_info, &err_addr, (uint64_t)count); } out_mca_release: @@ -377,7 +386,7 @@ static int amdgpu_mca_smu_debug_mode_set(void *data, u64 val) struct amdgpu_device *adev = (struct amdgpu_device *)data; int ret; - ret = amdgpu_mca_smu_set_debug_mode(adev, val ? true : false); + ret = amdgpu_ras_set_mca_debug_mode(adev, val ? true : false); if (ret) return ret; @@ -485,7 +494,7 @@ DEFINE_DEBUGFS_ATTRIBUTE(mca_debug_mode_fops, NULL, amdgpu_mca_smu_debug_mode_se void amdgpu_mca_smu_debugfs_init(struct amdgpu_device *adev, struct dentry *root) { #if defined(CONFIG_DEBUG_FS) - if (!root || adev->ip_versions[MP1_HWIP][0] != IP_VERSION(13, 0, 6)) + if (!root || amdgpu_ip_version(adev, MP1_HWIP, 0) != IP_VERSION(13, 0, 6)) return; debugfs_create_file("mca_debug_mode", 0200, root, adev, &mca_debug_mode_fops); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h index e51e8918e..b399f1b62 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h @@ -46,6 +46,8 @@ #define MCA_REG__STATUS__ERRORCODEEXT(x) MCA_REG_FIELD(x, 21, 16) #define MCA_REG__STATUS__ERRORCODE(x) MCA_REG_FIELD(x, 15, 0) +#define MCA_REG__MISC0__ERRCNT(x) MCA_REG_FIELD(x, 43, 32) + #define MCA_REG__SYND__ERRORINFORMATION(x) MCA_REG_FIELD(x, 17, 0) enum amdgpu_mca_ip { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c index 30c010836..da48b6da0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c @@ -98,6 +98,26 @@ static int amdgpu_mes_doorbell_init(struct amdgpu_device *adev) return 0; } +static int amdgpu_mes_event_log_init(struct amdgpu_device *adev) +{ + int r; + + r = amdgpu_bo_create_kernel(adev, PAGE_SIZE, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &adev->mes.event_log_gpu_obj, + &adev->mes.event_log_gpu_addr, + &adev->mes.event_log_cpu_addr); + if (r) { + dev_warn(adev->dev, "failed to create MES event log buffer (%d)", r); + return r; + } + + memset(adev->mes.event_log_cpu_addr, 0, PAGE_SIZE); + + return 0; + +} + static void amdgpu_mes_doorbell_free(struct amdgpu_device *adev) { bitmap_free(adev->mes.doorbell_bitmap); @@ -182,8 +202,14 @@ int amdgpu_mes_init(struct amdgpu_device *adev) if (r) goto error; + r = amdgpu_mes_event_log_init(adev); + if (r) + goto error_doorbell; + return 0; +error_doorbell: + amdgpu_mes_doorbell_free(adev); error: amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); @@ -199,6 +225,10 @@ error_ids: void amdgpu_mes_fini(struct amdgpu_device *adev) { + amdgpu_bo_free_kernel(&adev->mes.event_log_gpu_obj, + &adev->mes.event_log_gpu_addr, + &adev->mes.event_log_cpu_addr); + amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); amdgpu_device_wb_free(adev, adev->mes.read_val_offs); @@ -1153,7 +1183,7 @@ int amdgpu_mes_ctx_map_meta_data(struct amdgpu_device *adev, amdgpu_sync_create(&sync); - drm_exec_init(&exec, 0); + drm_exec_init(&exec, 0, 0); drm_exec_until_all_locked(&exec) { r = drm_exec_lock_obj(&exec, &ctx_data->meta_data_obj->tbo.base); @@ -1224,7 +1254,7 @@ int amdgpu_mes_ctx_unmap_meta_data(struct amdgpu_device *adev, struct drm_exec exec; long r; - drm_exec_init(&exec, 0); + drm_exec_init(&exec, 0, 0); drm_exec_until_all_locked(&exec) { r = drm_exec_lock_obj(&exec, &ctx_data->meta_data_obj->tbo.base); @@ -1510,3 +1540,34 @@ out: amdgpu_ucode_release(&adev->mes.fw[pipe]); return r; } + +#if defined(CONFIG_DEBUG_FS) + +static int amdgpu_debugfs_mes_event_log_show(struct seq_file *m, void *unused) +{ + struct amdgpu_device *adev = m->private; + uint32_t *mem = (uint32_t *)(adev->mes.event_log_cpu_addr); + + seq_hex_dump(m, "", DUMP_PREFIX_OFFSET, 32, 4, + mem, PAGE_SIZE, false); + + return 0; +} + + +DEFINE_SHOW_ATTRIBUTE(amdgpu_debugfs_mes_event_log); + +#endif + +void amdgpu_debugfs_mes_event_log_init(struct amdgpu_device *adev) +{ + +#if defined(CONFIG_DEBUG_FS) + struct drm_minor *minor = adev_to_drm(adev)->primary; + struct dentry *root = minor->debugfs_root; + + debugfs_create_file("amdgpu_mes_event_log", 0444, root, + adev, &amdgpu_debugfs_mes_event_log_fops); + +#endif +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h index c2c88b772..7d4f93fea 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h @@ -133,6 +133,11 @@ struct amdgpu_mes { uint32_t num_mes_dbs; unsigned long *doorbell_bitmap; + /* MES event log buffer */ + struct amdgpu_bo *event_log_gpu_obj; + uint64_t event_log_gpu_addr; + void *event_log_cpu_addr; + /* ip specific functions */ const struct amdgpu_mes_funcs *funcs; }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index 32fe05c81..2e4911050 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -32,7 +32,6 @@ #include <drm/display/drm_dp_helper.h> #include <drm/drm_crtc.h> -#include <drm/drm_edid.h> #include <drm/drm_encoder.h> #include <drm/drm_fixed.h> #include <drm/drm_framebuffer.h> @@ -51,6 +50,7 @@ struct amdgpu_device; struct amdgpu_encoder; struct amdgpu_router; struct amdgpu_hpd; +struct edid; #define to_amdgpu_crtc(x) container_of(x, struct amdgpu_crtc, base) #define to_amdgpu_connector(x) container_of(x, struct amdgpu_connector, base) @@ -343,6 +343,97 @@ struct amdgpu_mode_info { int disp_priority; const struct amdgpu_display_funcs *funcs; const enum drm_plane_type *plane_type; + + /* Driver-private color mgmt props */ + + /* @plane_degamma_lut_property: Plane property to set a degamma LUT to + * convert encoded values to light linear values before sampling or + * blending. + */ + struct drm_property *plane_degamma_lut_property; + /* @plane_degamma_lut_size_property: Plane property to define the max + * size of degamma LUT as supported by the driver (read-only). + */ + struct drm_property *plane_degamma_lut_size_property; + /** + * @plane_degamma_tf_property: Plane pre-defined transfer function to + * to go from scanout/encoded values to linear values. + */ + struct drm_property *plane_degamma_tf_property; + /** + * @plane_hdr_mult_property: + */ + struct drm_property *plane_hdr_mult_property; + + struct drm_property *plane_ctm_property; + /** + * @shaper_lut_property: Plane property to set pre-blending shaper LUT + * that converts color content before 3D LUT. If + * plane_shaper_tf_property != Identity TF, AMD color module will + * combine the user LUT values with pre-defined TF into the LUT + * parameters to be programmed. + */ + struct drm_property *plane_shaper_lut_property; + /** + * @shaper_lut_size_property: Plane property for the size of + * pre-blending shaper LUT as supported by the driver (read-only). + */ + struct drm_property *plane_shaper_lut_size_property; + /** + * @plane_shaper_tf_property: Plane property to set a predefined + * transfer function for pre-blending shaper (before applying 3D LUT) + * with or without LUT. There is no shaper ROM, but we can use AMD + * color modules to program LUT parameters from predefined TF (or + * from a combination of pre-defined TF and the custom 1D LUT). + */ + struct drm_property *plane_shaper_tf_property; + /** + * @plane_lut3d_property: Plane property for color transformation using + * a 3D LUT (pre-blending), a three-dimensional array where each + * element is an RGB triplet. Each dimension has the size of + * lut3d_size. The array contains samples from the approximated + * function. On AMD, values between samples are estimated by + * tetrahedral interpolation. The array is accessed with three indices, + * one for each input dimension (color channel), blue being the + * outermost dimension, red the innermost. + */ + struct drm_property *plane_lut3d_property; + /** + * @plane_degamma_lut_size_property: Plane property to define the max + * size of 3D LUT as supported by the driver (read-only). The max size + * is the max size of one dimension and, therefore, the max number of + * entries for 3D LUT array is the 3D LUT size cubed; + */ + struct drm_property *plane_lut3d_size_property; + /** + * @plane_blend_lut_property: Plane property for output gamma before + * blending. Userspace set a blend LUT to convert colors after 3D LUT + * conversion. It works as a post-3DLUT 1D LUT. With shaper LUT, they + * are sandwiching 3D LUT with two 1D LUT. If plane_blend_tf_property + * != Identity TF, AMD color module will combine the user LUT values + * with pre-defined TF into the LUT parameters to be programmed. + */ + struct drm_property *plane_blend_lut_property; + /** + * @plane_blend_lut_size_property: Plane property to define the max + * size of blend LUT as supported by the driver (read-only). + */ + struct drm_property *plane_blend_lut_size_property; + /** + * @plane_blend_tf_property: Plane property to set a predefined + * transfer function for pre-blending blend/out_gamma (after applying + * 3D LUT) with or without LUT. There is no blend ROM, but we can use + * AMD color modules to program LUT parameters from predefined TF (or + * from a combination of pre-defined TF and the custom 1D LUT). + */ + struct drm_property *plane_blend_tf_property; + /* @regamma_tf_property: Transfer function for CRTC regamma + * (post-blending). Possible values are defined by `enum + * amdgpu_transfer_function`. There is no regamma ROM, but we can use + * AMD color modules to program LUT parameters from predefined TF (or + * from a combination of pre-defined TF and the custom 1D LUT). + */ + struct drm_property *regamma_tf_property; }; #define AMDGPU_MAX_BL_LEVEL 0xFF @@ -416,6 +507,10 @@ struct amdgpu_crtc { int otg_inst; struct drm_pending_vblank_event *event; + + bool wb_pending; + bool wb_enabled; + struct drm_writeback_connector *wb_conn; }; struct amdgpu_encoder_atom_dig { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 425cebcc5..866bfde1c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -620,8 +620,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev, return r; if (!amdgpu_gmc_vram_full_visible(&adev->gmc) && - bo->tbo.resource->mem_type == TTM_PL_VRAM && - amdgpu_bo_in_cpu_visible_vram(bo)) + amdgpu_res_cpu_visible(adev, bo->tbo.resource)) amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved, ctx.bytes_moved); else @@ -1275,26 +1274,39 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, bool evict) void amdgpu_bo_get_memory(struct amdgpu_bo *bo, struct amdgpu_mem_stats *stats) { + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + struct ttm_resource *res = bo->tbo.resource; uint64_t size = amdgpu_bo_size(bo); + struct drm_gem_object *obj; unsigned int domain; + bool shared; /* Abort if the BO doesn't currently have a backing store */ - if (!bo->tbo.resource) + if (!res) return; - domain = amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type); + obj = &bo->tbo.base; + shared = drm_gem_object_is_shared_for_memory_stats(obj); + + domain = amdgpu_mem_type_to_domain(res->mem_type); switch (domain) { case AMDGPU_GEM_DOMAIN_VRAM: stats->vram += size; - if (amdgpu_bo_in_cpu_visible_vram(bo)) + if (amdgpu_res_cpu_visible(adev, bo->tbo.resource)) stats->visible_vram += size; + if (shared) + stats->vram_shared += size; break; case AMDGPU_GEM_DOMAIN_GTT: stats->gtt += size; + if (shared) + stats->gtt_shared += size; break; case AMDGPU_GEM_DOMAIN_CPU: default: stats->cpu += size; + if (shared) + stats->cpu_shared += size; break; } @@ -1381,10 +1393,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) /* Remember that this BO was accessed by the CPU */ abo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; - if (bo->resource->mem_type != TTM_PL_VRAM) - return 0; - - if (amdgpu_bo_in_cpu_visible_vram(abo)) + if (amdgpu_res_cpu_visible(adev, bo->resource)) return 0; /* Can't move a pinned BO to visible VRAM */ @@ -1408,7 +1417,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) /* this should never happen */ if (bo->resource->mem_type == TTM_PL_VRAM && - !amdgpu_bo_in_cpu_visible_vram(abo)) + !amdgpu_res_cpu_visible(adev, bo->resource)) return VM_FAULT_SIGBUS; ttm_bo_move_to_lru_tail_unlocked(bo); @@ -1572,6 +1581,7 @@ uint32_t amdgpu_bo_get_preferred_domain(struct amdgpu_device *adev, */ u64 amdgpu_bo_print_info(int id, struct amdgpu_bo *bo, struct seq_file *m) { + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); struct dma_buf_attachment *attachment; struct dma_buf *dma_buf; const char *placement; @@ -1580,10 +1590,11 @@ u64 amdgpu_bo_print_info(int id, struct amdgpu_bo *bo, struct seq_file *m) if (dma_resv_trylock(bo->tbo.base.resv)) { unsigned int domain; + domain = amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type); switch (domain) { case AMDGPU_GEM_DOMAIN_VRAM: - if (amdgpu_bo_in_cpu_visible_vram(bo)) + if (amdgpu_res_cpu_visible(adev, bo->tbo.resource)) placement = "VRAM VISIBLE"; else placement = "VRAM"; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index a3ea8a82d..fa03d9e48 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -138,12 +138,18 @@ struct amdgpu_bo_vm { struct amdgpu_mem_stats { /* current VRAM usage, includes visible VRAM */ uint64_t vram; + /* current shared VRAM usage, includes visible VRAM */ + uint64_t vram_shared; /* current visible VRAM usage */ uint64_t visible_vram; /* current GTT usage */ uint64_t gtt; + /* current shared GTT usage */ + uint64_t gtt_shared; /* current system memory usage */ uint64_t cpu; + /* current shared system memory usage */ + uint64_t cpu_shared; /* sum of evicted buffers, includes visible VRAM */ uint64_t evicted_vram; /* sum of evicted buffers due to CPU access */ @@ -245,28 +251,6 @@ static inline u64 amdgpu_bo_mmap_offset(struct amdgpu_bo *bo) } /** - * amdgpu_bo_in_cpu_visible_vram - check if BO is (partly) in visible VRAM - */ -static inline bool amdgpu_bo_in_cpu_visible_vram(struct amdgpu_bo *bo) -{ - struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - struct amdgpu_res_cursor cursor; - - if (!bo->tbo.resource || bo->tbo.resource->mem_type != TTM_PL_VRAM) - return false; - - amdgpu_res_first(bo->tbo.resource, 0, amdgpu_bo_size(bo), &cursor); - while (cursor.remaining) { - if (cursor.start < adev->gmc.visible_vram_size) - return true; - - amdgpu_res_next(&cursor, cursor.size); - } - - return false; -} - -/** * amdgpu_bo_explicit_sync - return whether the bo is explicitly synced */ static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index a21045d01..032861647 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -466,7 +466,7 @@ static int psp_sw_init(void *handle) } ret = amdgpu_bo_create_kernel(adev, PSP_1_MEG, PSP_1_MEG, - amdgpu_sriov_vf(adev) ? + (amdgpu_sriov_vf(adev) || adev->debug_use_vram_fw_buf) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, &psp->fw_pri_bo, &psp->fw_pri_mc_addr, @@ -1433,8 +1433,8 @@ int psp_xgmi_get_topology_info(struct psp_context *psp, get_extended_data) || amdgpu_ip_version(psp->adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 6); - bool ta_port_num_support = psp->xgmi_context.xgmi_ta_caps & - EXTEND_PEER_LINK_INFO_CMD_FLAG; + bool ta_port_num_support = amdgpu_sriov_vf(psp->adev) ? 0 : + psp->xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG; /* popluate the shared output buffer rather than the cmd input buffer * with node_ids as the input for GET_PEER_LINKS command execution. diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 4a3726bb6..31823a30d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -305,11 +305,13 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f, return -EINVAL; data->head.block = block_id; - /* only ue and ce errors are supported */ + /* only ue, ce and poison errors are supported */ if (!memcmp("ue", err, 2)) data->head.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; else if (!memcmp("ce", err, 2)) data->head.type = AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE; + else if (!memcmp("poison", err, 6)) + data->head.type = AMDGPU_RAS_ERROR__POISON; else return -EINVAL; @@ -431,9 +433,10 @@ static void amdgpu_ras_instance_mask_check(struct amdgpu_device *adev, * The block is one of: umc, sdma, gfx, etc. * see ras_block_string[] for details * - * The error type is one of: ue, ce, where, + * The error type is one of: ue, ce and poison where, * ue is multi-uncorrectable * ce is single-correctable + * poison is poison * * The sub-block is a the sub-block index, pass 0 if there is no sub-block. * The address and value are hexadecimal numbers, leading 0x is optional. @@ -1067,8 +1070,7 @@ static void amdgpu_ras_error_print_error_data(struct amdgpu_device *adev, mcm_info = &err_info->mcm_info; if (err_info->ce_count) { dev_info(adev->dev, "socket: %d, die: %d, " - "%lld new correctable hardware errors detected in %s block, " - "no user action is needed\n", + "%lld new correctable hardware errors detected in %s block\n", mcm_info->socket_id, mcm_info->die_id, err_info->ce_count, @@ -1080,8 +1082,7 @@ static void amdgpu_ras_error_print_error_data(struct amdgpu_device *adev, err_info = &err_node->err_info; mcm_info = &err_info->mcm_info; dev_info(adev->dev, "socket: %d, die: %d, " - "%lld correctable hardware errors detected in total in %s block, " - "no user action is needed\n", + "%lld correctable hardware errors detected in total in %s block\n", mcm_info->socket_id, mcm_info->die_id, err_info->ce_count, blk_name); } } @@ -1108,16 +1109,14 @@ static void amdgpu_ras_error_generate_report(struct amdgpu_device *adev, adev->smuio.funcs->get_die_id) { dev_info(adev->dev, "socket: %d, die: %d " "%ld correctable hardware errors " - "detected in %s block, no user " - "action is needed.\n", + "detected in %s block\n", adev->smuio.funcs->get_socket_id(adev), adev->smuio.funcs->get_die_id(adev), ras_mgr->err_data.ce_count, blk_name); } else { dev_info(adev->dev, "%ld correctable hardware errors " - "detected in %s block, no user " - "action is needed.\n", + "detected in %s block\n", ras_mgr->err_data.ce_count, blk_name); } @@ -1156,8 +1155,10 @@ static void amdgpu_rasmgr_error_data_statistic_update(struct ras_manager *obj, s for_each_ras_error(err_node, err_data) { err_info = &err_node->err_info; - amdgpu_ras_error_statistic_ce_count(&obj->err_data, &err_info->mcm_info, err_info->ce_count); - amdgpu_ras_error_statistic_ue_count(&obj->err_data, &err_info->mcm_info, err_info->ue_count); + amdgpu_ras_error_statistic_ce_count(&obj->err_data, + &err_info->mcm_info, NULL, err_info->ce_count); + amdgpu_ras_error_statistic_ue_count(&obj->err_data, + &err_info->mcm_info, NULL, err_info->ue_count); } } else { /* for legacy asic path which doesn't has error source info */ @@ -1918,7 +1919,7 @@ static void amdgpu_ras_interrupt_poison_creation_handler(struct ras_manager *obj struct amdgpu_iv_entry *entry) { dev_info(obj->adev->dev, - "Poison is created, no user action is needed.\n"); + "Poison is created\n"); } static void amdgpu_ras_interrupt_umc_handler(struct ras_manager *obj, @@ -2541,7 +2542,7 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev) return 0; data = &con->eh_data; - *data = kmalloc(sizeof(**data), GFP_KERNEL | __GFP_ZERO); + *data = kzalloc(sizeof(**data), GFP_KERNEL); if (!*data) { ret = -ENOMEM; goto out; @@ -2828,10 +2829,10 @@ int amdgpu_ras_init(struct amdgpu_device *adev) if (con) return 0; - con = kmalloc(sizeof(struct amdgpu_ras) + + con = kzalloc(sizeof(*con) + sizeof(struct ras_manager) * AMDGPU_RAS_BLOCK_COUNT + sizeof(struct ras_manager) * AMDGPU_RAS_MCA_BLOCK_COUNT, - GFP_KERNEL|__GFP_ZERO); + GFP_KERNEL); if (!con) return -ENOMEM; @@ -2918,6 +2919,11 @@ int amdgpu_ras_init(struct amdgpu_device *adev) amdgpu_ras_query_poison_mode(adev); + /* Packed socket_id to ras feature mask bits[31:29] */ + if (adev->smuio.funcs && + adev->smuio.funcs->get_socket_id) + con->features |= ((adev->smuio.funcs->get_socket_id(adev)) << 29); + /* Get RAS schema for particular SOC */ con->schema = amdgpu_get_ras_schema(adev); @@ -3136,6 +3142,8 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev) if (amdgpu_sriov_vf(adev)) return 0; + amdgpu_ras_set_mca_debug_mode(adev, false); + list_for_each_entry_safe(node, tmp, &adev->ras_list, node) { if (!node->ras_obj) { dev_warn(adev->dev, "Warning: abnormal ras list node.\n"); @@ -3409,12 +3417,18 @@ int amdgpu_ras_reset_gpu(struct amdgpu_device *adev) return 0; } -void amdgpu_ras_set_mca_debug_mode(struct amdgpu_device *adev, bool enable) +int amdgpu_ras_set_mca_debug_mode(struct amdgpu_device *adev, bool enable) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); + int ret = 0; - if (con) - con->is_mca_debug_mode = enable; + if (con) { + ret = amdgpu_mca_smu_set_debug_mode(adev, enable); + if (!ret) + con->is_mca_debug_mode = enable; + } + + return ret; } bool amdgpu_ras_get_mca_debug_mode(struct amdgpu_device *adev) @@ -3685,7 +3699,8 @@ static int ras_err_info_cmp(void *priv, const struct list_head *a, const struct } static struct ras_err_info *amdgpu_ras_error_get_info(struct ras_err_data *err_data, - struct amdgpu_smuio_mcm_config_info *mcm_info) + struct amdgpu_smuio_mcm_config_info *mcm_info, + struct ras_err_addr *err_addr) { struct ras_err_node *err_node; @@ -3699,6 +3714,9 @@ static struct ras_err_info *amdgpu_ras_error_get_info(struct ras_err_data *err_d memcpy(&err_node->err_info.mcm_info, mcm_info, sizeof(*mcm_info)); + if (err_addr) + memcpy(&err_node->err_info.err_addr, err_addr, sizeof(*err_addr)); + err_data->err_list_count++; list_add_tail(&err_node->node, &err_data->err_node_list); list_sort(NULL, &err_data->err_node_list, ras_err_info_cmp); @@ -3707,7 +3725,8 @@ static struct ras_err_info *amdgpu_ras_error_get_info(struct ras_err_data *err_d } int amdgpu_ras_error_statistic_ue_count(struct ras_err_data *err_data, - struct amdgpu_smuio_mcm_config_info *mcm_info, u64 count) + struct amdgpu_smuio_mcm_config_info *mcm_info, + struct ras_err_addr *err_addr, u64 count) { struct ras_err_info *err_info; @@ -3717,7 +3736,7 @@ int amdgpu_ras_error_statistic_ue_count(struct ras_err_data *err_data, if (!count) return 0; - err_info = amdgpu_ras_error_get_info(err_data, mcm_info); + err_info = amdgpu_ras_error_get_info(err_data, mcm_info, err_addr); if (!err_info) return -EINVAL; @@ -3728,7 +3747,8 @@ int amdgpu_ras_error_statistic_ue_count(struct ras_err_data *err_data, } int amdgpu_ras_error_statistic_ce_count(struct ras_err_data *err_data, - struct amdgpu_smuio_mcm_config_info *mcm_info, u64 count) + struct amdgpu_smuio_mcm_config_info *mcm_info, + struct ras_err_addr *err_addr, u64 count) { struct ras_err_info *err_info; @@ -3738,7 +3758,7 @@ int amdgpu_ras_error_statistic_ce_count(struct ras_err_data *err_data, if (!count) return 0; - err_info = amdgpu_ras_error_get_info(err_data, mcm_info); + err_info = amdgpu_ras_error_get_info(err_data, mcm_info, err_addr); if (!err_info) return -EINVAL; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 19161916a..76fb85628 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -452,10 +452,17 @@ struct ras_fs_data { char debugfs_name[32]; }; +struct ras_err_addr { + uint64_t err_status; + uint64_t err_ipid; + uint64_t err_addr; +}; + struct ras_err_info { struct amdgpu_smuio_mcm_config_info mcm_info; u64 ce_count; u64 ue_count; + struct ras_err_addr err_addr; }; struct ras_err_node { @@ -773,7 +780,7 @@ struct amdgpu_ras* amdgpu_ras_get_context(struct amdgpu_device *adev); int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras *ras_con); -void amdgpu_ras_set_mca_debug_mode(struct amdgpu_device *adev, bool enable); +int amdgpu_ras_set_mca_debug_mode(struct amdgpu_device *adev, bool enable); bool amdgpu_ras_get_mca_debug_mode(struct amdgpu_device *adev); bool amdgpu_ras_get_error_query_mode(struct amdgpu_device *adev, unsigned int *mode); @@ -806,8 +813,10 @@ void amdgpu_ras_inst_reset_ras_error_count(struct amdgpu_device *adev, int amdgpu_ras_error_data_init(struct ras_err_data *err_data); void amdgpu_ras_error_data_fini(struct ras_err_data *err_data); int amdgpu_ras_error_statistic_ce_count(struct ras_err_data *err_data, - struct amdgpu_smuio_mcm_config_info *mcm_info, u64 count); + struct amdgpu_smuio_mcm_config_info *mcm_info, + struct ras_err_addr *err_addr, u64 count); int amdgpu_ras_error_statistic_ue_count(struct ras_err_data *err_data, - struct amdgpu_smuio_mcm_config_info *mcm_info, u64 count); + struct amdgpu_smuio_mcm_config_info *mcm_info, + struct ras_err_addr *err_addr, u64 count); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 4e526d773..06f0a6534 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -647,6 +647,7 @@ int amdgpu_ring_test_helper(struct amdgpu_ring *ring) ring->name); ring->sched.ready = !r; + return r; } @@ -729,3 +730,14 @@ void amdgpu_ring_ib_on_emit_de(struct amdgpu_ring *ring) if (ring->is_sw_ring) amdgpu_sw_ring_ib_mark_offset(ring, AMDGPU_MUX_OFFSET_TYPE_DE); } + +bool amdgpu_ring_sched_ready(struct amdgpu_ring *ring) +{ + if (!ring) + return false; + + if (ring->no_scheduler || !drm_sched_wqueue_ready(&ring->sched)) + return false; + + return true; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index bbb53720a..fe1a61eb6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -450,5 +450,5 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, int amdgpu_ib_pool_init(struct amdgpu_device *adev); void amdgpu_ib_pool_fini(struct amdgpu_device *adev); int amdgpu_ib_ring_tests(struct amdgpu_device *adev); - +bool amdgpu_ring_sched_ready(struct amdgpu_ring *ring); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c index 35e0ae9ac..2c3675d91 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c @@ -531,13 +531,12 @@ int amdgpu_gfx_rlc_init_microcode(struct amdgpu_device *adev, if (version_major == 2 && version_minor == 1) adev->gfx.rlc.is_rlc_v2_1 = true; - if (version_minor >= 0) { - err = amdgpu_gfx_rlc_init_microcode_v2_0(adev); - if (err) { - dev_err(adev->dev, "fail to init rlc v2_0 microcode\n"); - return err; - } + err = amdgpu_gfx_rlc_init_microcode_v2_0(adev); + if (err) { + dev_err(adev->dev, "fail to init rlc v2_0 microcode\n"); + return err; } + if (version_minor >= 1) amdgpu_gfx_rlc_init_microcode_v2_1(adev); if (version_minor >= 2) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c new file mode 100644 index 000000000..7a6a67275 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c @@ -0,0 +1,247 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright 2023 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "amdgpu_seq64.h" + +#include <drm/drm_exec.h> + +/** + * DOC: amdgpu_seq64 + * + * amdgpu_seq64 allocates a 64bit memory on each request in sequence order. + * seq64 driver is required for user queue fence memory allocation, TLB + * counters and VM updates. It has maximum count of 32768 64 bit slots. + */ + +/** + * amdgpu_seq64_map - Map the seq64 memory to VM + * + * @adev: amdgpu_device pointer + * @vm: vm pointer + * @bo_va: bo_va pointer + * @seq64_addr: seq64 vaddr start address + * @size: seq64 pool size + * + * Map the seq64 memory to the given VM. + * + * Returns: + * 0 on success or a negative error code on failure + */ +int amdgpu_seq64_map(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct amdgpu_bo_va **bo_va, u64 seq64_addr, + uint32_t size) +{ + struct amdgpu_bo *bo; + struct drm_exec exec; + int r; + + bo = adev->seq64.sbo; + if (!bo) + return -EINVAL; + + drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0); + drm_exec_until_all_locked(&exec) { + r = amdgpu_vm_lock_pd(vm, &exec, 0); + if (likely(!r)) + r = drm_exec_lock_obj(&exec, &bo->tbo.base); + drm_exec_retry_on_contention(&exec); + if (unlikely(r)) + goto error; + } + + *bo_va = amdgpu_vm_bo_add(adev, vm, bo); + if (!*bo_va) { + r = -ENOMEM; + goto error; + } + + r = amdgpu_vm_bo_map(adev, *bo_va, seq64_addr, 0, size, + AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE | + AMDGPU_PTE_EXECUTABLE); + if (r) { + DRM_ERROR("failed to do bo_map on userq sem, err=%d\n", r); + amdgpu_vm_bo_del(adev, *bo_va); + goto error; + } + + r = amdgpu_vm_bo_update(adev, *bo_va, false); + if (r) { + DRM_ERROR("failed to do vm_bo_update on userq sem\n"); + amdgpu_vm_bo_del(adev, *bo_va); + goto error; + } + +error: + drm_exec_fini(&exec); + return r; +} + +/** + * amdgpu_seq64_unmap - Unmap the seq64 memory + * + * @adev: amdgpu_device pointer + * @fpriv: DRM file private + * + * Unmap the seq64 memory from the given VM. + */ +void amdgpu_seq64_unmap(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv) +{ + struct amdgpu_vm *vm; + struct amdgpu_bo *bo; + struct drm_exec exec; + int r; + + if (!fpriv->seq64_va) + return; + + bo = adev->seq64.sbo; + if (!bo) + return; + + vm = &fpriv->vm; + + drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0); + drm_exec_until_all_locked(&exec) { + r = amdgpu_vm_lock_pd(vm, &exec, 0); + if (likely(!r)) + r = drm_exec_lock_obj(&exec, &bo->tbo.base); + drm_exec_retry_on_contention(&exec); + if (unlikely(r)) + goto error; + } + + amdgpu_vm_bo_del(adev, fpriv->seq64_va); + + fpriv->seq64_va = NULL; + +error: + drm_exec_fini(&exec); +} + +/** + * amdgpu_seq64_alloc - Allocate a 64 bit memory + * + * @adev: amdgpu_device pointer + * @gpu_addr: allocated gpu VA start address + * @cpu_addr: allocated cpu VA start address + * + * Alloc a 64 bit memory from seq64 pool. + * + * Returns: + * 0 on success or a negative error code on failure + */ +int amdgpu_seq64_alloc(struct amdgpu_device *adev, u64 *gpu_addr, + u64 **cpu_addr) +{ + unsigned long bit_pos; + u32 offset; + + bit_pos = find_first_zero_bit(adev->seq64.used, adev->seq64.num_sem); + + if (bit_pos < adev->seq64.num_sem) { + __set_bit(bit_pos, adev->seq64.used); + offset = bit_pos << 6; /* convert to qw offset */ + } else { + return -EINVAL; + } + + *gpu_addr = offset + AMDGPU_SEQ64_VADDR_START; + *cpu_addr = offset + adev->seq64.cpu_base_addr; + + return 0; +} + +/** + * amdgpu_seq64_free - Free the given 64 bit memory + * + * @adev: amdgpu_device pointer + * @gpu_addr: gpu start address to be freed + * + * Free the given 64 bit memory from seq64 pool. + * + */ +void amdgpu_seq64_free(struct amdgpu_device *adev, u64 gpu_addr) +{ + u32 offset; + + offset = gpu_addr - AMDGPU_SEQ64_VADDR_START; + + offset >>= 6; + if (offset < adev->seq64.num_sem) + __clear_bit(offset, adev->seq64.used); +} + +/** + * amdgpu_seq64_fini - Cleanup seq64 driver + * + * @adev: amdgpu_device pointer + * + * Free the memory space allocated for seq64. + * + */ +void amdgpu_seq64_fini(struct amdgpu_device *adev) +{ + amdgpu_bo_free_kernel(&adev->seq64.sbo, + NULL, + (void **)&adev->seq64.cpu_base_addr); +} + +/** + * amdgpu_seq64_init - Initialize seq64 driver + * + * @adev: amdgpu_device pointer + * + * Allocate the required memory space for seq64. + * + * Returns: + * 0 on success or a negative error code on failure + */ +int amdgpu_seq64_init(struct amdgpu_device *adev) +{ + int r; + + if (adev->seq64.sbo) + return 0; + + /* + * AMDGPU_MAX_SEQ64_SLOTS * sizeof(u64) * 8 = AMDGPU_MAX_SEQ64_SLOTS + * 64bit slots + */ + r = amdgpu_bo_create_kernel(adev, AMDGPU_SEQ64_SIZE, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + &adev->seq64.sbo, NULL, + (void **)&adev->seq64.cpu_base_addr); + if (r) { + dev_warn(adev->dev, "(%d) create seq64 failed\n", r); + return r; + } + + memset(adev->seq64.cpu_base_addr, 0, AMDGPU_SEQ64_SIZE); + + adev->seq64.num_sem = AMDGPU_MAX_SEQ64_SLOTS; + memset(&adev->seq64.used, 0, sizeof(adev->seq64.used)); + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h new file mode 100644 index 000000000..2196e72be --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h @@ -0,0 +1,49 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright 2023 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __AMDGPU_SEQ64_H__ +#define __AMDGPU_SEQ64_H__ + +#define AMDGPU_SEQ64_SIZE (2ULL << 20) +#define AMDGPU_MAX_SEQ64_SLOTS (AMDGPU_SEQ64_SIZE / (sizeof(u64) * 8)) +#define AMDGPU_SEQ64_VADDR_OFFSET 0x50000 +#define AMDGPU_SEQ64_VADDR_START (AMDGPU_VA_RESERVED_SIZE + AMDGPU_SEQ64_VADDR_OFFSET) + +struct amdgpu_seq64 { + struct amdgpu_bo *sbo; + u32 num_sem; + u64 *cpu_base_addr; + DECLARE_BITMAP(used, AMDGPU_MAX_SEQ64_SLOTS); +}; + +void amdgpu_seq64_fini(struct amdgpu_device *adev); +int amdgpu_seq64_init(struct amdgpu_device *adev); +int amdgpu_seq64_alloc(struct amdgpu_device *adev, u64 *gpu_addr, u64 **cpu_addr); +void amdgpu_seq64_free(struct amdgpu_device *adev, u64 gpu_addr); +int amdgpu_seq64_map(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct amdgpu_bo_va **bo_va, u64 seq64_addr, uint32_t size); +void amdgpu_seq64_unmap(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv); + +#endif + diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index 2fd1bfb35..f539b1d00 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -554,6 +554,21 @@ TRACE_EVENT(amdgpu_reset_reg_dumps, __entry->value) ); +TRACE_EVENT(amdgpu_runpm_reference_dumps, + TP_PROTO(uint32_t index, const char *func), + TP_ARGS(index, func), + TP_STRUCT__entry( + __field(uint32_t, index) + __string(func, func) + ), + TP_fast_assign( + __entry->index = index; + __assign_str(func, func); + ), + TP_printk("amdgpu runpm reference dump 0x%x: 0x%s\n", + __entry->index, + __get_str(func)) +); #undef AMDGPU_JOB_GET_TIMELINE_NAME #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index b0ed10f4d..851509c6e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -137,7 +137,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU); } else if (!amdgpu_gmc_vram_full_visible(&adev->gmc) && !(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) && - amdgpu_bo_in_cpu_visible_vram(abo)) { + amdgpu_res_cpu_visible(adev, bo->resource)) { /* Try evicting to the CPU inaccessible part of VRAM * first, but only set GTT as busy placement, so this @@ -408,40 +408,55 @@ error: return r; } -/* - * amdgpu_mem_visible - Check that memory can be accessed by ttm_bo_move_memcpy +/** + * amdgpu_res_cpu_visible - Check that resource can be accessed by CPU + * @adev: amdgpu device + * @res: the resource to check * - * Called by amdgpu_bo_move() + * Returns: true if the full resource is CPU visible, false otherwise. */ -static bool amdgpu_mem_visible(struct amdgpu_device *adev, - struct ttm_resource *mem) +bool amdgpu_res_cpu_visible(struct amdgpu_device *adev, + struct ttm_resource *res) { - u64 mem_size = (u64)mem->size; struct amdgpu_res_cursor cursor; - u64 end; - if (mem->mem_type == TTM_PL_SYSTEM || - mem->mem_type == TTM_PL_TT) + if (!res) + return false; + + if (res->mem_type == TTM_PL_SYSTEM || res->mem_type == TTM_PL_TT || + res->mem_type == AMDGPU_PL_PREEMPT) return true; - if (mem->mem_type != TTM_PL_VRAM) + + if (res->mem_type != TTM_PL_VRAM) return false; - amdgpu_res_first(mem, 0, mem_size, &cursor); - end = cursor.start + cursor.size; + amdgpu_res_first(res, 0, res->size, &cursor); while (cursor.remaining) { + if ((cursor.start + cursor.size) >= adev->gmc.visible_vram_size) + return false; amdgpu_res_next(&cursor, cursor.size); + } - if (!cursor.remaining) - break; + return true; +} - /* ttm_resource_ioremap only supports contiguous memory */ - if (end != cursor.start) - return false; +/* + * amdgpu_res_copyable - Check that memory can be accessed by ttm_bo_move_memcpy + * + * Called by amdgpu_bo_move() + */ +static bool amdgpu_res_copyable(struct amdgpu_device *adev, + struct ttm_resource *mem) +{ + if (!amdgpu_res_cpu_visible(adev, mem)) + return false; - end = cursor.start + cursor.size; - } + /* ttm_resource_ioremap only supports contiguous memory */ + if (mem->mem_type == TTM_PL_VRAM && + !(mem->placement & TTM_PL_FLAG_CONTIGUOUS)) + return false; - return end <= adev->gmc.visible_vram_size; + return true; } /* @@ -534,8 +549,8 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict, if (r) { /* Check that all memory is CPU accessible */ - if (!amdgpu_mem_visible(adev, old_mem) || - !amdgpu_mem_visible(adev, new_mem)) { + if (!amdgpu_res_copyable(adev, old_mem) || + !amdgpu_res_copyable(adev, new_mem)) { pr_err("Move buffer fallback to memcpy unavailable\n"); return r; } @@ -562,7 +577,6 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev, struct ttm_resource *mem) { struct amdgpu_device *adev = amdgpu_ttm_adev(bdev); - size_t bus_size = (size_t)mem->size; switch (mem->mem_type) { case TTM_PL_SYSTEM: @@ -573,9 +587,6 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev, break; case TTM_PL_VRAM: mem->bus.offset = mem->start << PAGE_SHIFT; - /* check if it's visible */ - if ((mem->bus.offset + bus_size) > adev->gmc.visible_vram_size) - return -EINVAL; if (adev->mman.aper_base_kaddr && mem->placement & TTM_PL_FLAG_CONTIGUOUS) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index 65ec82141..32cf6b6f6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -139,6 +139,9 @@ int amdgpu_vram_mgr_reserve_range(struct amdgpu_vram_mgr *mgr, int amdgpu_vram_mgr_query_page_status(struct amdgpu_vram_mgr *mgr, uint64_t start); +bool amdgpu_res_cpu_visible(struct amdgpu_device *adev, + struct ttm_resource *res); + int amdgpu_ttm_init(struct amdgpu_device *adev); void amdgpu_ttm_fini(struct amdgpu_device *adev); void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index 0efb2568c..3e12763e4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -1062,7 +1062,8 @@ int amdgpu_ucode_create_bo(struct amdgpu_device *adev) { if (adev->firmware.load_type != AMDGPU_FW_LOAD_DIRECT) { amdgpu_bo_create_kernel(adev, adev->firmware.fw_size, PAGE_SIZE, - amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, + (amdgpu_sriov_vf(adev) || adev->debug_use_vram_fw_buf) ? + AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, &adev->firmware.fw_buf, &adev->firmware.fw_buf_mc, &adev->firmware.fw_buf_ptr); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c index ca45ba8ac..1b5ef3210 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c @@ -86,7 +86,7 @@ static int map_ring_data(struct amdgpu_device *adev, struct amdgpu_vm *vm, amdgpu_sync_create(&sync); - drm_exec_init(&exec, 0); + drm_exec_init(&exec, 0, 0); drm_exec_until_all_locked(&exec) { r = drm_exec_lock_obj(&exec, &bo->tbo.base); drm_exec_retry_on_contention(&exec); @@ -149,7 +149,7 @@ static int unmap_ring_data(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct drm_exec exec; long r; - drm_exec_init(&exec, 0); + drm_exec_init(&exec, 0, 0); drm_exec_until_all_locked(&exec) { r = drm_exec_lock_obj(&exec, &bo->tbo.base); drm_exec_retry_on_contention(&exec); @@ -766,6 +766,9 @@ static int umsch_mm_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (amdgpu_in_reset(adev) || adev->in_s0ix || adev->in_suspend) + return 0; + return umsch_mm_test(adev); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c index 3a632c3b1..0dcff2889 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c @@ -1099,7 +1099,8 @@ bool amdgpu_sriov_xnack_support(struct amdgpu_device *adev) { bool xnack_mode = true; - if (amdgpu_sriov_vf(adev) && adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) + if (amdgpu_sriov_vf(adev) && + amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2)) xnack_mode = false; return xnack_mode; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c index db6fc0cb1..453a4b786 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c @@ -1,6 +1,7 @@ // SPDX-License-Identifier: GPL-2.0+ #include <drm/drm_atomic_helper.h> +#include <drm/drm_edid.h> #include <drm/drm_simple_kms_helper.h> #include <drm/drm_vblank.h> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 5baefb548..5b50f7359 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1439,6 +1439,51 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev, } /** + * amdgpu_vm_flush_compute_tlb - Flush TLB on compute VM + * + * @adev: amdgpu_device pointer + * @vm: requested vm + * @flush_type: flush type + * @xcc_mask: mask of XCCs that belong to the compute partition in need of a TLB flush. + * + * Flush TLB if needed for a compute VM. + * + * Returns: + * 0 for success. + */ +int amdgpu_vm_flush_compute_tlb(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + uint32_t flush_type, + uint32_t xcc_mask) +{ + uint64_t tlb_seq = amdgpu_vm_tlb_seq(vm); + bool all_hub = false; + int xcc = 0, r = 0; + + WARN_ON_ONCE(!vm->is_compute_context); + + /* + * It can be that we race and lose here, but that is extremely unlikely + * and the worst thing which could happen is that we flush the changes + * into the TLB once more which is harmless. + */ + if (atomic64_xchg(&vm->kfd_last_flushed_seq, tlb_seq) == tlb_seq) + return 0; + + if (adev->family == AMDGPU_FAMILY_AI || + adev->family == AMDGPU_FAMILY_RV) + all_hub = true; + + for_each_inst(xcc, xcc_mask) { + r = amdgpu_gmc_flush_gpu_tlb_pasid(adev, vm->pasid, flush_type, + all_hub, xcc); + if (r) + break; + } + return r; +} + +/** * amdgpu_vm_bo_add - add a bo to a specific vm * * @adev: amdgpu_device pointer @@ -1514,6 +1559,37 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev, trace_amdgpu_vm_bo_map(bo_va, mapping); } +/* Validate operation parameters to prevent potential abuse */ +static int amdgpu_vm_verify_parameters(struct amdgpu_device *adev, + struct amdgpu_bo *bo, + uint64_t saddr, + uint64_t offset, + uint64_t size) +{ + uint64_t tmp, lpfn; + + if (saddr & AMDGPU_GPU_PAGE_MASK + || offset & AMDGPU_GPU_PAGE_MASK + || size & AMDGPU_GPU_PAGE_MASK) + return -EINVAL; + + if (check_add_overflow(saddr, size, &tmp) + || check_add_overflow(offset, size, &tmp) + || size == 0 /* which also leads to end < begin */) + return -EINVAL; + + /* make sure object fit at this offset */ + if (bo && offset + size > amdgpu_bo_size(bo)) + return -EINVAL; + + /* Ensure last pfn not exceed max_pfn */ + lpfn = (saddr + size - 1) >> AMDGPU_GPU_PAGE_SHIFT; + if (lpfn >= adev->vm_manager.max_pfn) + return -EINVAL; + + return 0; +} + /** * amdgpu_vm_bo_map - map bo inside a vm * @@ -1540,21 +1616,14 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev, struct amdgpu_bo *bo = bo_va->base.bo; struct amdgpu_vm *vm = bo_va->base.vm; uint64_t eaddr; + int r; - /* validate the parameters */ - if (saddr & ~PAGE_MASK || offset & ~PAGE_MASK || size & ~PAGE_MASK) - return -EINVAL; - if (saddr + size <= saddr || offset + size <= offset) - return -EINVAL; - - /* make sure object fit at this offset */ - eaddr = saddr + size - 1; - if ((bo && offset + size > amdgpu_bo_size(bo)) || - (eaddr >= adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT)) - return -EINVAL; + r = amdgpu_vm_verify_parameters(adev, bo, saddr, offset, size); + if (r) + return r; saddr /= AMDGPU_GPU_PAGE_SIZE; - eaddr /= AMDGPU_GPU_PAGE_SIZE; + eaddr = saddr + (size - 1) / AMDGPU_GPU_PAGE_SIZE; tmp = amdgpu_vm_it_iter_first(&vm->va, saddr, eaddr); if (tmp) { @@ -1607,17 +1676,9 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev, uint64_t eaddr; int r; - /* validate the parameters */ - if (saddr & ~PAGE_MASK || offset & ~PAGE_MASK || size & ~PAGE_MASK) - return -EINVAL; - if (saddr + size <= saddr || offset + size <= offset) - return -EINVAL; - - /* make sure object fit at this offset */ - eaddr = saddr + size - 1; - if ((bo && offset + size > amdgpu_bo_size(bo)) || - (eaddr >= adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT)) - return -EINVAL; + r = amdgpu_vm_verify_parameters(adev, bo, saddr, offset, size); + if (r) + return r; /* Allocate all the needed memory */ mapping = kmalloc(sizeof(*mapping), GFP_KERNEL); @@ -1631,7 +1692,7 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev, } saddr /= AMDGPU_GPU_PAGE_SIZE; - eaddr /= AMDGPU_GPU_PAGE_SIZE; + eaddr = saddr + (size - 1) / AMDGPU_GPU_PAGE_SIZE; mapping->start = saddr; mapping->last = eaddr; @@ -1718,10 +1779,14 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, struct amdgpu_bo_va_mapping *before, *after, *tmp, *next; LIST_HEAD(removed); uint64_t eaddr; + int r; + + r = amdgpu_vm_verify_parameters(adev, NULL, saddr, 0, size); + if (r) + return r; - eaddr = saddr + size - 1; saddr /= AMDGPU_GPU_PAGE_SIZE; - eaddr /= AMDGPU_GPU_PAGE_SIZE; + eaddr = saddr + (size - 1) / AMDGPU_GPU_PAGE_SIZE; /* Allocate all the needed memory */ before = kzalloc(sizeof(*before), GFP_KERNEL); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 2cd86d2bf..4740dd65b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -116,7 +116,7 @@ struct amdgpu_mem_stats; #define AMDGPU_VM_FAULT_STOP_FIRST 1 #define AMDGPU_VM_FAULT_STOP_ALWAYS 2 -/* Reserve 4MB VRAM for page tables */ +/* How much VRAM be reserved for page tables */ #define AMDGPU_VM_RESERVED_VRAM (8ULL << 20) /* @@ -324,6 +324,7 @@ struct amdgpu_vm { /* Last finished delayed update */ atomic64_t tlb_seq; struct dma_fence *last_tlb_flush; + atomic64_t kfd_last_flushed_seq; /* How many times we had to re-generate the page tables */ uint64_t generation; @@ -445,6 +446,10 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, int amdgpu_vm_handle_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct ww_acquire_ctx *ticket); +int amdgpu_vm_flush_compute_tlb(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + uint32_t flush_type, + uint32_t xcc_mask); void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base, struct amdgpu_vm *vm, struct amdgpu_bo *bo); int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c index e81579708..ad44012cc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c @@ -26,6 +26,7 @@ #include "amdgpu.h" #include "amdgpu_ucode.h" #include "amdgpu_vpe.h" +#include "amdgpu_smu.h" #include "soc15_common.h" #include "vpe_v6_1.h" @@ -33,8 +34,180 @@ /* VPE CSA resides in the 4th page of CSA */ #define AMDGPU_CSA_VPE_OFFSET (4096 * 3) +/* 1 second timeout */ +#define VPE_IDLE_TIMEOUT msecs_to_jiffies(1000) + +#define VPE_MAX_DPM_LEVEL 4 +#define FIXED1_8_BITS_PER_FRACTIONAL_PART 8 +#define GET_PRATIO_INTEGER_PART(x) ((x) >> FIXED1_8_BITS_PER_FRACTIONAL_PART) + static void vpe_set_ring_funcs(struct amdgpu_device *adev); +static inline uint16_t div16_u16_rem(uint16_t dividend, uint16_t divisor, uint16_t *remainder) +{ + *remainder = dividend % divisor; + return dividend / divisor; +} + +static inline uint16_t complete_integer_division_u16( + uint16_t dividend, + uint16_t divisor, + uint16_t *remainder) +{ + return div16_u16_rem(dividend, divisor, (uint16_t *)remainder); +} + +static uint16_t vpe_u1_8_from_fraction(uint16_t numerator, uint16_t denominator) +{ + u16 arg1_value = numerator; + u16 arg2_value = denominator; + + uint16_t remainder; + + /* determine integer part */ + uint16_t res_value = complete_integer_division_u16( + arg1_value, arg2_value, &remainder); + + if (res_value > 127 /* CHAR_MAX */) + return 0; + + /* determine fractional part */ + { + unsigned int i = FIXED1_8_BITS_PER_FRACTIONAL_PART; + + do { + remainder <<= 1; + + res_value <<= 1; + + if (remainder >= arg2_value) { + res_value |= 1; + remainder -= arg2_value; + } + } while (--i != 0); + } + + /* round up LSB */ + { + uint16_t summand = (remainder << 1) >= arg2_value; + + if ((res_value + summand) > 32767 /* SHRT_MAX */) + return 0; + + res_value += summand; + } + + return res_value; +} + +static uint16_t vpe_internal_get_pratio(uint16_t from_frequency, uint16_t to_frequency) +{ + uint16_t pratio = vpe_u1_8_from_fraction(from_frequency, to_frequency); + + if (GET_PRATIO_INTEGER_PART(pratio) > 1) + pratio = 0; + + return pratio; +} + +/* + * VPE has 4 DPM levels from level 0 (lowerest) to 3 (highest), + * VPE FW will dynamically decide which level should be used according to current loading. + * + * Get VPE and SOC clocks from PM, and select the appropriate four clock values, + * calculate the ratios of adjusting from one clock to another. + * The VPE FW can then request the appropriate frequency from the PMFW. + */ +int amdgpu_vpe_configure_dpm(struct amdgpu_vpe *vpe) +{ + struct amdgpu_device *adev = vpe->ring.adev; + uint32_t dpm_ctl; + + if (adev->pm.dpm_enabled) { + struct dpm_clocks clock_table = { 0 }; + struct dpm_clock *VPEClks; + struct dpm_clock *SOCClks; + uint32_t idx; + uint32_t pratio_vmax_vnorm = 0, pratio_vnorm_vmid = 0, pratio_vmid_vmin = 0; + uint16_t pratio_vmin_freq = 0, pratio_vmid_freq = 0, pratio_vnorm_freq = 0, pratio_vmax_freq = 0; + + dpm_ctl = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable)); + dpm_ctl |= 1; /* DPM enablement */ + WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable), dpm_ctl); + + /* Get VPECLK and SOCCLK */ + if (amdgpu_dpm_get_dpm_clock_table(adev, &clock_table)) { + dev_dbg(adev->dev, "%s: get clock failed!\n", __func__); + goto disable_dpm; + } + + SOCClks = clock_table.SocClocks; + VPEClks = clock_table.VPEClocks; + + /* vpe dpm only cares 4 levels. */ + for (idx = 0; idx < VPE_MAX_DPM_LEVEL; idx++) { + uint32_t soc_dpm_level; + uint32_t min_freq; + + if (idx == 0) + soc_dpm_level = 0; + else + soc_dpm_level = (idx * 2) + 1; + + /* clamp the max level */ + if (soc_dpm_level > PP_SMU_NUM_VPECLK_DPM_LEVELS - 1) + soc_dpm_level = PP_SMU_NUM_VPECLK_DPM_LEVELS - 1; + + min_freq = (SOCClks[soc_dpm_level].Freq < VPEClks[soc_dpm_level].Freq) ? + SOCClks[soc_dpm_level].Freq : VPEClks[soc_dpm_level].Freq; + + switch (idx) { + case 0: + pratio_vmin_freq = min_freq; + break; + case 1: + pratio_vmid_freq = min_freq; + break; + case 2: + pratio_vnorm_freq = min_freq; + break; + case 3: + pratio_vmax_freq = min_freq; + break; + default: + break; + } + } + + if (pratio_vmin_freq && pratio_vmid_freq && pratio_vnorm_freq && pratio_vmax_freq) { + uint32_t pratio_ctl; + + pratio_vmax_vnorm = (uint32_t)vpe_internal_get_pratio(pratio_vmax_freq, pratio_vnorm_freq); + pratio_vnorm_vmid = (uint32_t)vpe_internal_get_pratio(pratio_vnorm_freq, pratio_vmid_freq); + pratio_vmid_vmin = (uint32_t)vpe_internal_get_pratio(pratio_vmid_freq, pratio_vmin_freq); + + pratio_ctl = pratio_vmax_vnorm | (pratio_vnorm_vmid << 9) | (pratio_vmid_vmin << 18); + WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_pratio), pratio_ctl); /* PRatio */ + WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_request_interval), 24000); /* 1ms, unit=1/24MHz */ + WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_decision_threshold), 1200000); /* 50ms */ + WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_busy_clamp_threshold), 1200000);/* 50ms */ + WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_idle_clamp_threshold), 1200000);/* 50ms */ + dev_dbg(adev->dev, "%s: configure vpe dpm pratio done!\n", __func__); + } else { + dev_dbg(adev->dev, "%s: invalid pratio parameters!\n", __func__); + goto disable_dpm; + } + } + return 0; + +disable_dpm: + dpm_ctl = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable)); + dpm_ctl &= 0xfffffffe; /* Disable DPM */ + WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable), dpm_ctl); + dev_dbg(adev->dev, "%s: disable vpe dpm\n", __func__); + return 0; +} + int amdgpu_vpe_psp_update_sram(struct amdgpu_device *adev) { struct amdgpu_firmware_info ucode = { @@ -134,6 +307,19 @@ static int vpe_early_init(void *handle) return 0; } +static void vpe_idle_work_handler(struct work_struct *work) +{ + struct amdgpu_device *adev = + container_of(work, struct amdgpu_device, vpe.idle_work.work); + unsigned int fences = 0; + + fences += amdgpu_fence_count_emitted(&adev->vpe.ring); + + if (fences == 0) + amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_GATE); + else + schedule_delayed_work(&adev->vpe.idle_work, VPE_IDLE_TIMEOUT); +} static int vpe_common_init(struct amdgpu_vpe *vpe) { @@ -150,6 +336,9 @@ static int vpe_common_init(struct amdgpu_vpe *vpe) return r; } + vpe->context_started = false; + INIT_DELAYED_WORK(&adev->vpe.idle_work, vpe_idle_work_handler); + return 0; } @@ -201,6 +390,12 @@ static int vpe_hw_init(void *handle) struct amdgpu_vpe *vpe = &adev->vpe; int ret; + /* Power on VPE */ + ret = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, + AMD_PG_STATE_UNGATE); + if (ret) + return ret; + ret = vpe_load_microcode(vpe); if (ret) return ret; @@ -219,6 +414,9 @@ static int vpe_hw_fini(void *handle) vpe_ring_stop(vpe); + /* Power off VPE */ + amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_GATE); + return 0; } @@ -226,6 +424,8 @@ static int vpe_suspend(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + cancel_delayed_work_sync(&adev->vpe.idle_work); + return vpe_hw_fini(adev); } @@ -430,6 +630,21 @@ static int vpe_set_clockgating_state(void *handle, static int vpe_set_powergating_state(void *handle, enum amd_powergating_state state) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_vpe *vpe = &adev->vpe; + + if (!adev->pm.dpm_enabled) + dev_err(adev->dev, "Without PM, cannot support powergating\n"); + + dev_dbg(adev->dev, "%s: %s!\n", __func__, (state == AMD_PG_STATE_GATE) ? "GATE":"UNGATE"); + + if (state == AMD_PG_STATE_GATE) { + amdgpu_dpm_enable_vpe(adev, false); + vpe->context_started = false; + } else { + amdgpu_dpm_enable_vpe(adev, true); + } + return 0; } @@ -595,6 +810,38 @@ err0: return ret; } +static void vpe_ring_begin_use(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_vpe *vpe = &adev->vpe; + + cancel_delayed_work_sync(&adev->vpe.idle_work); + + /* Power on VPE and notify VPE of new context */ + if (!vpe->context_started) { + uint32_t context_notify; + + /* Power on VPE */ + amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_UNGATE); + + /* Indicates that a job from a new context has been submitted. */ + context_notify = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.context_indicator)); + if ((context_notify & 0x1) == 0) + context_notify |= 0x1; + else + context_notify &= ~(0x1); + WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.context_indicator), context_notify); + vpe->context_started = true; + } +} + +static void vpe_ring_end_use(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + schedule_delayed_work(&adev->vpe.idle_work, VPE_IDLE_TIMEOUT); +} + static const struct amdgpu_ring_funcs vpe_ring_funcs = { .type = AMDGPU_RING_TYPE_VPE, .align_mask = 0xf, @@ -625,6 +872,8 @@ static const struct amdgpu_ring_funcs vpe_ring_funcs = { .init_cond_exec = vpe_ring_init_cond_exec, .patch_cond_exec = vpe_ring_patch_cond_exec, .preempt_ib = vpe_ring_preempt_ib, + .begin_use = vpe_ring_begin_use, + .end_use = vpe_ring_end_use, }; static void vpe_set_ring_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h index 29d56f7ae..1153ddaea 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h @@ -47,6 +47,15 @@ struct vpe_regs { uint32_t queue0_rb_wptr_lo; uint32_t queue0_rb_wptr_hi; uint32_t queue0_preempt; + + uint32_t dpm_enable; + uint32_t dpm_pratio; + uint32_t dpm_request_interval; + uint32_t dpm_decision_threshold; + uint32_t dpm_busy_clamp_threshold; + uint32_t dpm_idle_clamp_threshold; + uint32_t dpm_request_lv; + uint32_t context_indicator; }; struct amdgpu_vpe { @@ -63,12 +72,15 @@ struct amdgpu_vpe { struct amdgpu_bo *cmdbuf_obj; uint64_t cmdbuf_gpu_addr; uint32_t *cmdbuf_cpu_addr; + struct delayed_work idle_work; + bool context_started; }; int amdgpu_vpe_psp_update_sram(struct amdgpu_device *adev); int amdgpu_vpe_init_microcode(struct amdgpu_vpe *vpe); int amdgpu_vpe_ring_init(struct amdgpu_vpe *vpe); int amdgpu_vpe_ring_fini(struct amdgpu_vpe *vpe); +int amdgpu_vpe_configure_dpm(struct amdgpu_vpe *vpe); #define vpe_ring_init(vpe) ((vpe)->funcs->ring_init ? (vpe)->funcs->ring_init((vpe)) : 0) #define vpe_ring_start(vpe) ((vpe)->funcs->ring_start ? (vpe)->funcs->ring_start((vpe)) : 0) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index bd20cb3b9..a6c88f2fe 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -413,6 +413,38 @@ static ssize_t amdgpu_xgmi_show_num_links(struct device *dev, return sysfs_emit(buf, "%s\n", buf); } +static ssize_t amdgpu_xgmi_show_connected_port_num(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(ddev); + struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info; + int i, j, size = 0; + int current_node; + /* + * get the node id in the sysfs for the current socket and show + * it in the port num info output in the sysfs for easy reading. + * it is NOT the one retrieved from xgmi ta. + */ + for (i = 0; i < top->num_nodes; i++) { + if (top->nodes[i].node_id == adev->gmc.xgmi.node_id) { + current_node = i; + break; + } + } + + for (i = 0; i < top->num_nodes; i++) { + for (j = 0; j < top->nodes[i].num_links; j++) + /* node id in sysfs starts from 1 rather than 0 so +1 here */ + size += sysfs_emit_at(buf, size, "%02x:%02x -> %02x:%02x\n", current_node + 1, + top->nodes[i].port_num[j].src_xgmi_port_num, i + 1, + top->nodes[i].port_num[j].dst_xgmi_port_num); + } + + return size; +} + #define AMDGPU_XGMI_SET_FICAA(o) ((o) | 0x456801) static ssize_t amdgpu_xgmi_show_error(struct device *dev, struct device_attribute *attr, @@ -452,6 +484,7 @@ static DEVICE_ATTR(xgmi_physical_id, 0444, amdgpu_xgmi_show_physical_id, NULL); static DEVICE_ATTR(xgmi_error, S_IRUGO, amdgpu_xgmi_show_error, NULL); static DEVICE_ATTR(xgmi_num_hops, S_IRUGO, amdgpu_xgmi_show_num_hops, NULL); static DEVICE_ATTR(xgmi_num_links, S_IRUGO, amdgpu_xgmi_show_num_links, NULL); +static DEVICE_ATTR(xgmi_port_num, S_IRUGO, amdgpu_xgmi_show_connected_port_num, NULL); static int amdgpu_xgmi_sysfs_add_dev_info(struct amdgpu_device *adev, struct amdgpu_hive_info *hive) @@ -487,6 +520,13 @@ static int amdgpu_xgmi_sysfs_add_dev_info(struct amdgpu_device *adev, if (ret) pr_err("failed to create xgmi_num_links\n"); + /* Create xgmi port num file if supported */ + if (adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG) { + ret = device_create_file(adev->dev, &dev_attr_xgmi_port_num); + if (ret) + dev_err(adev->dev, "failed to create xgmi_port_num\n"); + } + /* Create sysfs link to hive info folder on the first device */ if (hive->kobj.parent != (&adev->dev->kobj)) { ret = sysfs_create_link(&adev->dev->kobj, &hive->kobj, @@ -517,6 +557,8 @@ remove_file: device_remove_file(adev->dev, &dev_attr_xgmi_error); device_remove_file(adev->dev, &dev_attr_xgmi_num_hops); device_remove_file(adev->dev, &dev_attr_xgmi_num_links); + if (adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG) + device_remove_file(adev->dev, &dev_attr_xgmi_port_num); success: return ret; @@ -533,6 +575,8 @@ static void amdgpu_xgmi_sysfs_rem_dev_info(struct amdgpu_device *adev, device_remove_file(adev->dev, &dev_attr_xgmi_error); device_remove_file(adev->dev, &dev_attr_xgmi_num_hops); device_remove_file(adev->dev, &dev_attr_xgmi_num_links); + if (adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG) + device_remove_file(adev->dev, &dev_attr_xgmi_port_num); if (hive->kobj.parent != (&adev->dev->kobj)) sysfs_remove_link(&adev->dev->kobj,"xgmi_hive_info"); @@ -779,6 +823,28 @@ static int amdgpu_xgmi_initialize_hive_get_data_partition(struct amdgpu_hive_inf return 0; } +static void amdgpu_xgmi_fill_topology_info(struct amdgpu_device *adev, + struct amdgpu_device *peer_adev) +{ + struct psp_xgmi_topology_info *top_info = &adev->psp.xgmi_context.top_info; + struct psp_xgmi_topology_info *peer_info = &peer_adev->psp.xgmi_context.top_info; + + for (int i = 0; i < peer_info->num_nodes; i++) { + if (peer_info->nodes[i].node_id == adev->gmc.xgmi.node_id) { + for (int j = 0; j < top_info->num_nodes; j++) { + if (top_info->nodes[j].node_id == peer_adev->gmc.xgmi.node_id) { + peer_info->nodes[i].num_hops = top_info->nodes[j].num_hops; + peer_info->nodes[i].is_sharing_enabled = + top_info->nodes[j].is_sharing_enabled; + peer_info->nodes[i].num_links = + top_info->nodes[j].num_links; + return; + } + } + } + } +} + int amdgpu_xgmi_add_device(struct amdgpu_device *adev) { struct psp_xgmi_topology_info *top_info; @@ -853,18 +919,38 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev) goto exit_unlock; } - /* get latest topology info for each device from psp */ - list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { - ret = psp_xgmi_get_topology_info(&tmp_adev->psp, count, - &tmp_adev->psp.xgmi_context.top_info, false); + if (amdgpu_sriov_vf(adev) && + adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG) { + /* only get topology for VF being init if it can support full duplex */ + ret = psp_xgmi_get_topology_info(&adev->psp, count, + &adev->psp.xgmi_context.top_info, false); if (ret) { - dev_err(tmp_adev->dev, + dev_err(adev->dev, "XGMI: Get topology failure on device %llx, hive %llx, ret %d", - tmp_adev->gmc.xgmi.node_id, - tmp_adev->gmc.xgmi.hive_id, ret); - /* To do : continue with some node failed or disable the whole hive */ + adev->gmc.xgmi.node_id, + adev->gmc.xgmi.hive_id, ret); + /* To do: continue with some node failed or disable the whole hive*/ goto exit_unlock; } + + /* fill the topology info for peers instead of getting from PSP */ + list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { + amdgpu_xgmi_fill_topology_info(adev, tmp_adev); + } + } else { + /* get latest topology info for each device from psp */ + list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { + ret = psp_xgmi_get_topology_info(&tmp_adev->psp, count, + &tmp_adev->psp.xgmi_context.top_info, false); + if (ret) { + dev_err(tmp_adev->dev, + "XGMI: Get topology failure on device %llx, hive %llx, ret %d", + tmp_adev->gmc.xgmi.node_id, + tmp_adev->gmc.xgmi.hive_id, ret); + /* To do : continue with some node failed or disable the whole hive */ + goto exit_unlock; + } + } } /* get topology again for hives that support extended data */ @@ -1227,10 +1313,10 @@ static void __xgmi_v6_4_0_query_error_count(struct amdgpu_device *adev, struct a switch (xgmi_v6_4_0_pcs_mca_get_error_type(adev, status)) { case AMDGPU_MCA_ERROR_TYPE_UE: - amdgpu_ras_error_statistic_ue_count(err_data, mcm_info, 1ULL); + amdgpu_ras_error_statistic_ue_count(err_data, mcm_info, NULL, 1ULL); break; case AMDGPU_MCA_ERROR_TYPE_CE: - amdgpu_ras_error_statistic_ce_count(err_data, mcm_info, 1ULL); + amdgpu_ras_error_statistic_ce_count(err_data, mcm_info, NULL, 1ULL); break; default: break; diff --git a/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c b/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c index 3f715e7fe..d6f808acf 100644 --- a/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c +++ b/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c @@ -24,6 +24,7 @@ #include "soc15.h" #include "soc15_common.h" +#include "amdgpu_reg_state.h" #include "amdgpu_xcp.h" #include "gfx_v9_4_3.h" #include "gfxhub_v1_2.h" @@ -656,3 +657,416 @@ int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev) return 0; } + +static void aqua_read_smn(struct amdgpu_device *adev, + struct amdgpu_smn_reg_data *regdata, + uint64_t smn_addr) +{ + regdata->addr = smn_addr; + regdata->value = RREG32_PCIE(smn_addr); +} + +struct aqua_reg_list { + uint64_t start_addr; + uint32_t num_regs; + uint32_t incrx; +}; + +#define DW_ADDR_INCR 4 + +static void aqua_read_smn_ext(struct amdgpu_device *adev, + struct amdgpu_smn_reg_data *regdata, + uint64_t smn_addr, int i) +{ + regdata->addr = + smn_addr + adev->asic_funcs->encode_ext_smn_addressing(i); + regdata->value = RREG32_PCIE_EXT(regdata->addr); +} + +#define smnreg_0x1A340218 0x1A340218 +#define smnreg_0x1A3402E4 0x1A3402E4 +#define smnreg_0x1A340294 0x1A340294 +#define smreg_0x1A380088 0x1A380088 + +#define NUM_PCIE_SMN_REGS 14 + +static struct aqua_reg_list pcie_reg_addrs[] = { + { smnreg_0x1A340218, 1, 0 }, + { smnreg_0x1A3402E4, 1, 0 }, + { smnreg_0x1A340294, 6, DW_ADDR_INCR }, + { smreg_0x1A380088, 6, DW_ADDR_INCR }, +}; + +static ssize_t aqua_vanjaram_read_pcie_state(struct amdgpu_device *adev, + void *buf, size_t max_size) +{ + struct amdgpu_reg_state_pcie_v1_0 *pcie_reg_state; + uint32_t start_addr, incrx, num_regs, szbuf; + struct amdgpu_regs_pcie_v1_0 *pcie_regs; + struct amdgpu_smn_reg_data *reg_data; + struct pci_dev *us_pdev, *ds_pdev; + int aer_cap, r, n; + + if (!buf || !max_size) + return -EINVAL; + + pcie_reg_state = (struct amdgpu_reg_state_pcie_v1_0 *)buf; + + szbuf = sizeof(*pcie_reg_state) + + amdgpu_reginst_size(1, sizeof(*pcie_regs), NUM_PCIE_SMN_REGS); + /* Only one instance of pcie regs */ + if (max_size < szbuf) + return -EOVERFLOW; + + pcie_regs = (struct amdgpu_regs_pcie_v1_0 *)((uint8_t *)buf + + sizeof(*pcie_reg_state)); + pcie_regs->inst_header.instance = 0; + pcie_regs->inst_header.state = AMDGPU_INST_S_OK; + pcie_regs->inst_header.num_smn_regs = NUM_PCIE_SMN_REGS; + + reg_data = pcie_regs->smn_reg_values; + + for (r = 0; r < ARRAY_SIZE(pcie_reg_addrs); r++) { + start_addr = pcie_reg_addrs[r].start_addr; + incrx = pcie_reg_addrs[r].incrx; + num_regs = pcie_reg_addrs[r].num_regs; + for (n = 0; n < num_regs; n++) { + aqua_read_smn(adev, reg_data, start_addr + n * incrx); + ++reg_data; + } + } + + ds_pdev = pci_upstream_bridge(adev->pdev); + us_pdev = pci_upstream_bridge(ds_pdev); + + pcie_capability_read_word(us_pdev, PCI_EXP_DEVSTA, + &pcie_regs->device_status); + pcie_capability_read_word(us_pdev, PCI_EXP_LNKSTA, + &pcie_regs->link_status); + + aer_cap = pci_find_ext_capability(us_pdev, PCI_EXT_CAP_ID_ERR); + if (aer_cap) { + pci_read_config_dword(us_pdev, aer_cap + PCI_ERR_COR_STATUS, + &pcie_regs->pcie_corr_err_status); + pci_read_config_dword(us_pdev, aer_cap + PCI_ERR_UNCOR_STATUS, + &pcie_regs->pcie_uncorr_err_status); + } + + pci_read_config_dword(us_pdev, PCI_PRIMARY_BUS, + &pcie_regs->sub_bus_number_latency); + + pcie_reg_state->common_header.structure_size = szbuf; + pcie_reg_state->common_header.format_revision = 1; + pcie_reg_state->common_header.content_revision = 0; + pcie_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_PCIE; + pcie_reg_state->common_header.num_instances = 1; + + return pcie_reg_state->common_header.structure_size; +} + +#define smnreg_0x11A00050 0x11A00050 +#define smnreg_0x11A00180 0x11A00180 +#define smnreg_0x11A00070 0x11A00070 +#define smnreg_0x11A00200 0x11A00200 +#define smnreg_0x11A0020C 0x11A0020C +#define smnreg_0x11A00210 0x11A00210 +#define smnreg_0x11A00108 0x11A00108 + +#define XGMI_LINK_REG(smnreg, l) ((smnreg) | (l << 20)) + +#define NUM_XGMI_SMN_REGS 25 + +static struct aqua_reg_list xgmi_reg_addrs[] = { + { smnreg_0x11A00050, 1, 0 }, + { smnreg_0x11A00180, 16, DW_ADDR_INCR }, + { smnreg_0x11A00070, 4, DW_ADDR_INCR }, + { smnreg_0x11A00200, 1, 0 }, + { smnreg_0x11A0020C, 1, 0 }, + { smnreg_0x11A00210, 1, 0 }, + { smnreg_0x11A00108, 1, 0 }, +}; + +static ssize_t aqua_vanjaram_read_xgmi_state(struct amdgpu_device *adev, + void *buf, size_t max_size) +{ + struct amdgpu_reg_state_xgmi_v1_0 *xgmi_reg_state; + uint32_t start_addr, incrx, num_regs, szbuf; + struct amdgpu_regs_xgmi_v1_0 *xgmi_regs; + struct amdgpu_smn_reg_data *reg_data; + const int max_xgmi_instances = 8; + int inst = 0, i, j, r, n; + const int xgmi_inst = 2; + void *p; + + if (!buf || !max_size) + return -EINVAL; + + xgmi_reg_state = (struct amdgpu_reg_state_xgmi_v1_0 *)buf; + + szbuf = sizeof(*xgmi_reg_state) + + amdgpu_reginst_size(max_xgmi_instances, sizeof(*xgmi_regs), + NUM_XGMI_SMN_REGS); + /* Only one instance of pcie regs */ + if (max_size < szbuf) + return -EOVERFLOW; + + p = &xgmi_reg_state->xgmi_state_regs[0]; + for_each_inst(i, adev->aid_mask) { + for (j = 0; j < xgmi_inst; ++j) { + xgmi_regs = (struct amdgpu_regs_xgmi_v1_0 *)p; + xgmi_regs->inst_header.instance = inst++; + + xgmi_regs->inst_header.state = AMDGPU_INST_S_OK; + xgmi_regs->inst_header.num_smn_regs = NUM_XGMI_SMN_REGS; + + reg_data = xgmi_regs->smn_reg_values; + + for (r = 0; r < ARRAY_SIZE(xgmi_reg_addrs); r++) { + start_addr = xgmi_reg_addrs[r].start_addr; + incrx = xgmi_reg_addrs[r].incrx; + num_regs = xgmi_reg_addrs[r].num_regs; + + for (n = 0; n < num_regs; n++) { + aqua_read_smn_ext( + adev, reg_data, + XGMI_LINK_REG(start_addr, j) + + n * incrx, + i); + ++reg_data; + } + } + p = reg_data; + } + } + + xgmi_reg_state->common_header.structure_size = szbuf; + xgmi_reg_state->common_header.format_revision = 1; + xgmi_reg_state->common_header.content_revision = 0; + xgmi_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_XGMI; + xgmi_reg_state->common_header.num_instances = max_xgmi_instances; + + return xgmi_reg_state->common_header.structure_size; +} + +#define smnreg_0x11C00070 0x11C00070 +#define smnreg_0x11C00210 0x11C00210 + +static struct aqua_reg_list wafl_reg_addrs[] = { + { smnreg_0x11C00070, 4, DW_ADDR_INCR }, + { smnreg_0x11C00210, 1, 0 }, +}; + +#define WAFL_LINK_REG(smnreg, l) ((smnreg) | (l << 20)) + +#define NUM_WAFL_SMN_REGS 5 + +static ssize_t aqua_vanjaram_read_wafl_state(struct amdgpu_device *adev, + void *buf, size_t max_size) +{ + struct amdgpu_reg_state_wafl_v1_0 *wafl_reg_state; + uint32_t start_addr, incrx, num_regs, szbuf; + struct amdgpu_regs_wafl_v1_0 *wafl_regs; + struct amdgpu_smn_reg_data *reg_data; + const int max_wafl_instances = 8; + int inst = 0, i, j, r, n; + const int wafl_inst = 2; + void *p; + + if (!buf || !max_size) + return -EINVAL; + + wafl_reg_state = (struct amdgpu_reg_state_wafl_v1_0 *)buf; + + szbuf = sizeof(*wafl_reg_state) + + amdgpu_reginst_size(max_wafl_instances, sizeof(*wafl_regs), + NUM_WAFL_SMN_REGS); + + if (max_size < szbuf) + return -EOVERFLOW; + + p = &wafl_reg_state->wafl_state_regs[0]; + for_each_inst(i, adev->aid_mask) { + for (j = 0; j < wafl_inst; ++j) { + wafl_regs = (struct amdgpu_regs_wafl_v1_0 *)p; + wafl_regs->inst_header.instance = inst++; + + wafl_regs->inst_header.state = AMDGPU_INST_S_OK; + wafl_regs->inst_header.num_smn_regs = NUM_WAFL_SMN_REGS; + + reg_data = wafl_regs->smn_reg_values; + + for (r = 0; r < ARRAY_SIZE(wafl_reg_addrs); r++) { + start_addr = wafl_reg_addrs[r].start_addr; + incrx = wafl_reg_addrs[r].incrx; + num_regs = wafl_reg_addrs[r].num_regs; + for (n = 0; n < num_regs; n++) { + aqua_read_smn_ext( + adev, reg_data, + WAFL_LINK_REG(start_addr, j) + + n * incrx, + i); + ++reg_data; + } + } + p = reg_data; + } + } + + wafl_reg_state->common_header.structure_size = szbuf; + wafl_reg_state->common_header.format_revision = 1; + wafl_reg_state->common_header.content_revision = 0; + wafl_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_WAFL; + wafl_reg_state->common_header.num_instances = max_wafl_instances; + + return wafl_reg_state->common_header.structure_size; +} + +#define smnreg_0x1B311060 0x1B311060 +#define smnreg_0x1B411060 0x1B411060 +#define smnreg_0x1B511060 0x1B511060 +#define smnreg_0x1B611060 0x1B611060 + +#define smnreg_0x1C307120 0x1C307120 +#define smnreg_0x1C317120 0x1C317120 + +#define smnreg_0x1C320830 0x1C320830 +#define smnreg_0x1C380830 0x1C380830 +#define smnreg_0x1C3D0830 0x1C3D0830 +#define smnreg_0x1C420830 0x1C420830 + +#define smnreg_0x1C320100 0x1C320100 +#define smnreg_0x1C380100 0x1C380100 +#define smnreg_0x1C3D0100 0x1C3D0100 +#define smnreg_0x1C420100 0x1C420100 + +#define smnreg_0x1B310500 0x1B310500 +#define smnreg_0x1C300400 0x1C300400 + +#define USR_CAKE_INCR 0x11000 +#define USR_LINK_INCR 0x100000 +#define USR_CP_INCR 0x10000 + +#define NUM_USR_SMN_REGS 20 + +struct aqua_reg_list usr_reg_addrs[] = { + { smnreg_0x1B311060, 4, DW_ADDR_INCR }, + { smnreg_0x1B411060, 4, DW_ADDR_INCR }, + { smnreg_0x1B511060, 4, DW_ADDR_INCR }, + { smnreg_0x1B611060, 4, DW_ADDR_INCR }, + { smnreg_0x1C307120, 2, DW_ADDR_INCR }, + { smnreg_0x1C317120, 2, DW_ADDR_INCR }, +}; + +#define NUM_USR1_SMN_REGS 46 +struct aqua_reg_list usr1_reg_addrs[] = { + { smnreg_0x1C320830, 6, USR_CAKE_INCR }, + { smnreg_0x1C380830, 5, USR_CAKE_INCR }, + { smnreg_0x1C3D0830, 5, USR_CAKE_INCR }, + { smnreg_0x1C420830, 4, USR_CAKE_INCR }, + { smnreg_0x1C320100, 6, USR_CAKE_INCR }, + { smnreg_0x1C380100, 5, USR_CAKE_INCR }, + { smnreg_0x1C3D0100, 5, USR_CAKE_INCR }, + { smnreg_0x1C420100, 4, USR_CAKE_INCR }, + { smnreg_0x1B310500, 4, USR_LINK_INCR }, + { smnreg_0x1C300400, 2, USR_CP_INCR }, +}; + +static ssize_t aqua_vanjaram_read_usr_state(struct amdgpu_device *adev, + void *buf, size_t max_size, + int reg_state) +{ + uint32_t start_addr, incrx, num_regs, szbuf, num_smn; + struct amdgpu_reg_state_usr_v1_0 *usr_reg_state; + struct amdgpu_regs_usr_v1_0 *usr_regs; + struct amdgpu_smn_reg_data *reg_data; + const int max_usr_instances = 4; + struct aqua_reg_list *reg_addrs; + int inst = 0, i, n, r, arr_size; + void *p; + + if (!buf || !max_size) + return -EINVAL; + + switch (reg_state) { + case AMDGPU_REG_STATE_TYPE_USR: + arr_size = ARRAY_SIZE(usr_reg_addrs); + reg_addrs = usr_reg_addrs; + num_smn = NUM_USR_SMN_REGS; + break; + case AMDGPU_REG_STATE_TYPE_USR_1: + arr_size = ARRAY_SIZE(usr1_reg_addrs); + reg_addrs = usr1_reg_addrs; + num_smn = NUM_USR1_SMN_REGS; + break; + default: + return -EINVAL; + } + + usr_reg_state = (struct amdgpu_reg_state_usr_v1_0 *)buf; + + szbuf = sizeof(*usr_reg_state) + amdgpu_reginst_size(max_usr_instances, + sizeof(*usr_regs), + num_smn); + if (max_size < szbuf) + return -EOVERFLOW; + + p = &usr_reg_state->usr_state_regs[0]; + for_each_inst(i, adev->aid_mask) { + usr_regs = (struct amdgpu_regs_usr_v1_0 *)p; + usr_regs->inst_header.instance = inst++; + usr_regs->inst_header.state = AMDGPU_INST_S_OK; + usr_regs->inst_header.num_smn_regs = num_smn; + reg_data = usr_regs->smn_reg_values; + + for (r = 0; r < arr_size; r++) { + start_addr = reg_addrs[r].start_addr; + incrx = reg_addrs[r].incrx; + num_regs = reg_addrs[r].num_regs; + for (n = 0; n < num_regs; n++) { + aqua_read_smn_ext(adev, reg_data, + start_addr + n * incrx, i); + reg_data++; + } + } + p = reg_data; + } + + usr_reg_state->common_header.structure_size = szbuf; + usr_reg_state->common_header.format_revision = 1; + usr_reg_state->common_header.content_revision = 0; + usr_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_USR; + usr_reg_state->common_header.num_instances = max_usr_instances; + + return usr_reg_state->common_header.structure_size; +} + +ssize_t aqua_vanjaram_get_reg_state(struct amdgpu_device *adev, + enum amdgpu_reg_state reg_state, void *buf, + size_t max_size) +{ + ssize_t size; + + switch (reg_state) { + case AMDGPU_REG_STATE_TYPE_PCIE: + size = aqua_vanjaram_read_pcie_state(adev, buf, max_size); + break; + case AMDGPU_REG_STATE_TYPE_XGMI: + size = aqua_vanjaram_read_xgmi_state(adev, buf, max_size); + break; + case AMDGPU_REG_STATE_TYPE_WAFL: + size = aqua_vanjaram_read_wafl_state(adev, buf, max_size); + break; + case AMDGPU_REG_STATE_TYPE_USR: + size = aqua_vanjaram_read_usr_state(adev, buf, max_size, + AMDGPU_REG_STATE_TYPE_USR); + break; + case AMDGPU_REG_STATE_TYPE_USR_1: + size = aqua_vanjaram_read_usr_state( + adev, buf, max_size, AMDGPU_REG_STATE_TYPE_USR_1); + break; + default: + return -EINVAL; + } + + return size; +} diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c index f0737fb3a..d1bba9c64 100644 --- a/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c @@ -30,6 +30,8 @@ #define regATHUB_MISC_CNTL_V3_0_1 0x00d7 #define regATHUB_MISC_CNTL_V3_0_1_BASE_IDX 0 +#define regATHUB_MISC_CNTL_V3_3_0 0x00d8 +#define regATHUB_MISC_CNTL_V3_3_0_BASE_IDX 0 static uint32_t athub_v3_0_get_cg_cntl(struct amdgpu_device *adev) @@ -40,6 +42,9 @@ static uint32_t athub_v3_0_get_cg_cntl(struct amdgpu_device *adev) case IP_VERSION(3, 0, 1): data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_0_1); break; + case IP_VERSION(3, 3, 0): + data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_3_0); + break; default: data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL); break; @@ -53,6 +58,9 @@ static void athub_v3_0_set_cg_cntl(struct amdgpu_device *adev, uint32_t data) case IP_VERSION(3, 0, 1): WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_0_1, data); break; + case IP_VERSION(3, 3, 0): + WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_3_0, data); + break; default: WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL, data); break; diff --git a/drivers/gpu/drm/amd/amdgpu/atom.c b/drivers/gpu/drm/amd/amdgpu/atom.c index 1b51be9b5..ac51d19b5 100644 --- a/drivers/gpu/drm/amd/amdgpu/atom.c +++ b/drivers/gpu/drm/amd/amdgpu/atom.c @@ -395,7 +395,6 @@ static void atom_skip_src_int(atom_exec_context *ctx, uint8_t attr, int *ptr) (*ptr)++; return; } - return; } } diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c index 3ee219aa2..7672abe6c 100644 --- a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c +++ b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c @@ -28,6 +28,7 @@ #include <acpi/video.h> +#include <drm/drm_edid.h> #include <drm/amdgpu_drm.h> #include "amdgpu.h" #include "amdgpu_connectors.h" diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c index bb666cb75..587ee632a 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c @@ -21,6 +21,7 @@ * */ +#include <drm/drm_edid.h> #include <drm/drm_fourcc.h> #include <drm/drm_modeset_helper.h> #include <drm/drm_modeset_helper_vtables.h> diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c index 7af277f61..f22ec2736 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c @@ -21,6 +21,7 @@ * */ +#include <drm/drm_edid.h> #include <drm/drm_fourcc.h> #include <drm/drm_modeset_helper.h> #include <drm/drm_modeset_helper_vtables.h> diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c index 143efc37a..4dbe9b325 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c @@ -23,6 +23,7 @@ #include <linux/pci.h> +#include <drm/drm_edid.h> #include <drm/drm_fourcc.h> #include <drm/drm_modeset_helper.h> #include <drm/drm_modeset_helper_vtables.h> diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index adeddfb7f..05bcce233 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -21,6 +21,7 @@ * */ +#include <drm/drm_edid.h> #include <drm/drm_fourcc.h> #include <drm/drm_modeset_helper.h> #include <drm/drm_modeset_helper_vtables.h> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c index 806a8cf90..0afe86bcc 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c @@ -67,6 +67,7 @@ MODULE_FIRMWARE("amdgpu/gc_11_0_0_pfp.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_0_me.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_0_mec.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc_1.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_0_toc.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_1_pfp.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_1_me.bin"); @@ -106,23 +107,6 @@ static const struct soc15_reg_golden golden_settings_gc_11_0_1[] = SOC15_REG_GOLDEN_VALUE(GC, 0, regTCP_CNTL2, 0xfcffffff, 0x0000000a) }; -static const struct soc15_reg_golden golden_settings_gc_11_5_0[] = { - SOC15_REG_GOLDEN_VALUE(GC, 0, regDB_DEBUG5, 0xffffffff, 0x00000800), - SOC15_REG_GOLDEN_VALUE(GC, 0, regGB_ADDR_CONFIG, 0x0c1807ff, 0x00000242), - SOC15_REG_GOLDEN_VALUE(GC, 0, regGCR_GENERAL_CNTL, 0x1ff1ffff, 0x00000500), - SOC15_REG_GOLDEN_VALUE(GC, 0, regGL2A_ADDR_MATCH_MASK, 0xffffffff, 0xfffffff3), - SOC15_REG_GOLDEN_VALUE(GC, 0, regGL2C_ADDR_MATCH_MASK, 0xffffffff, 0xfffffff3), - SOC15_REG_GOLDEN_VALUE(GC, 0, regGL2C_CTRL, 0xffffffff, 0xf37fff3f), - SOC15_REG_GOLDEN_VALUE(GC, 0, regGL2C_CTRL3, 0xfffffffb, 0x00f40188), - SOC15_REG_GOLDEN_VALUE(GC, 0, regGL2C_CTRL4, 0xf0ffffff, 0x80009007), - SOC15_REG_GOLDEN_VALUE(GC, 0, regPA_CL_ENHANCE, 0xf1ffffff, 0x00880007), - SOC15_REG_GOLDEN_VALUE(GC, 0, regPC_CONFIG_CNTL_1, 0xffffffff, 0x00010000), - SOC15_REG_GOLDEN_VALUE(GC, 0, regTA_CNTL_AUX, 0xf7f7ffff, 0x01030000), - SOC15_REG_GOLDEN_VALUE(GC, 0, regTA_CNTL2, 0x007f0000, 0x00000000), - SOC15_REG_GOLDEN_VALUE(GC, 0, regTCP_CNTL2, 0xffcfffff, 0x0000200a), - SOC15_REG_GOLDEN_VALUE(GC, 0, regUTCL1_CTRL_2, 0xffffffff, 0x0000048f) -}; - #define DEFAULT_SH_MEM_CONFIG \ ((SH_MEM_ADDRESS_MODE_64 << SH_MEM_CONFIG__ADDRESS_MODE__SHIFT) | \ (SH_MEM_ALIGNMENT_MODE_UNALIGNED << SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) | \ @@ -293,6 +277,9 @@ static void gfx_v11_0_set_kiq_pm4_funcs(struct amdgpu_device *adev) static void gfx_v11_0_init_golden_registers(struct amdgpu_device *adev) { + if (amdgpu_sriov_vf(adev)) + return; + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { case IP_VERSION(11, 0, 1): case IP_VERSION(11, 0, 4): @@ -300,11 +287,6 @@ static void gfx_v11_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_gc_11_0_1, (const u32)ARRAY_SIZE(golden_settings_gc_11_0_1)); break; - case IP_VERSION(11, 5, 0): - soc15_program_register_sequence(adev, - golden_settings_gc_11_5_0, - (const u32)ARRAY_SIZE(golden_settings_gc_11_5_0)); - break; default: break; } @@ -564,7 +546,11 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev) } if (!amdgpu_sriov_vf(adev)) { - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", ucode_prefix); + if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 0, 0) && + adev->pdev->revision == 0xCE) + snprintf(fw_name, sizeof(fw_name), "amdgpu/gc_11_0_0_rlc_1.bin"); + else + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", ucode_prefix); err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name); if (err) goto out; @@ -1644,7 +1630,7 @@ static void gfx_v11_0_setup_rb(struct amdgpu_device *adev) active_rb_bitmap |= (0x3 << (i * rb_bitmap_width_per_sa)); } - active_rb_bitmap |= global_active_rb_bitmap; + active_rb_bitmap &= global_active_rb_bitmap; adev->gfx.config.backend_enable_mask = active_rb_bitmap; adev->gfx.config.num_rbs = hweight32(active_rb_bitmap); } @@ -4466,11 +4452,43 @@ static int gfx_v11_0_wait_for_idle(void *handle) return -ETIMEDOUT; } +static int gfx_v11_0_request_gfx_index_mutex(struct amdgpu_device *adev, + int req) +{ + u32 i, tmp, val; + + for (i = 0; i < adev->usec_timeout; i++) { + /* Request with MeId=2, PipeId=0 */ + tmp = REG_SET_FIELD(0, CP_GFX_INDEX_MUTEX, REQUEST, req); + tmp = REG_SET_FIELD(tmp, CP_GFX_INDEX_MUTEX, CLIENTID, 4); + WREG32_SOC15(GC, 0, regCP_GFX_INDEX_MUTEX, tmp); + + val = RREG32_SOC15(GC, 0, regCP_GFX_INDEX_MUTEX); + if (req) { + if (val == tmp) + break; + } else { + tmp = REG_SET_FIELD(tmp, CP_GFX_INDEX_MUTEX, + REQUEST, 1); + + /* unlocked or locked by firmware */ + if (val != tmp) + break; + } + udelay(1); + } + + if (i >= adev->usec_timeout) + return -EINVAL; + + return 0; +} + static int gfx_v11_0_soft_reset(void *handle) { u32 grbm_soft_reset = 0; u32 tmp; - int i, j, k; + int r, i, j, k; struct amdgpu_device *adev = (struct amdgpu_device *)handle; tmp = RREG32_SOC15(GC, 0, regCP_INT_CNTL); @@ -4510,6 +4528,13 @@ static int gfx_v11_0_soft_reset(void *handle) } } + /* Try to acquire the gfx mutex before access to CP_VMID_RESET */ + r = gfx_v11_0_request_gfx_index_mutex(adev, 1); + if (r) { + DRM_ERROR("Failed to acquire the gfx mutex during soft reset\n"); + return r; + } + WREG32_SOC15(GC, 0, regCP_VMID_RESET, 0xfffffffe); // Read CP_VMID_RESET register three times. @@ -4518,6 +4543,13 @@ static int gfx_v11_0_soft_reset(void *handle) RREG32_SOC15(GC, 0, regCP_VMID_RESET); RREG32_SOC15(GC, 0, regCP_VMID_RESET); + /* release the gfx mutex */ + r = gfx_v11_0_request_gfx_index_mutex(adev, 0); + if (r) { + DRM_ERROR("Failed to release the gfx mutex during soft reset\n"); + return r; + } + for (i = 0; i < adev->usec_timeout; i++) { if (!RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) && !RREG32_SOC15(GC, 0, regCP_GFX_HQD_ACTIVE)) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c index 4a09cc0d8..131cddbdd 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c @@ -3828,8 +3828,8 @@ static void gfx_v9_4_3_inst_query_ras_err_count(struct amdgpu_device *adev, /* the caller should make sure initialize value of * err_data->ue_count and err_data->ce_count */ - amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, ue_count); - amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, ce_count); + amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, NULL, ue_count); + amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, NULL, ce_count); } static void gfx_v9_4_3_inst_reset_ras_err_count(struct amdgpu_device *adev, @@ -3882,150 +3882,6 @@ static void gfx_v9_4_3_inst_reset_ras_err_count(struct amdgpu_device *adev, mutex_unlock(&adev->grbm_idx_mutex); } -static void gfx_v9_4_3_inst_query_utc_err_status(struct amdgpu_device *adev, - int xcc_id) -{ - uint32_t data; - - data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regUTCL2_MEM_ECC_STATUS); - if (data) { - dev_warn(adev->dev, "GFX UTCL2 Mem Ecc Status: 0x%x!\n", data); - WREG32_SOC15(GC, GET_INST(GC, xcc_id), regUTCL2_MEM_ECC_STATUS, 0x3); - } - - data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regVML2_MEM_ECC_STATUS); - if (data) { - dev_warn(adev->dev, "GFX VML2 Mem Ecc Status: 0x%x!\n", data); - WREG32_SOC15(GC, GET_INST(GC, xcc_id), regVML2_MEM_ECC_STATUS, 0x3); - } - - data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), - regVML2_WALKER_MEM_ECC_STATUS); - if (data) { - dev_warn(adev->dev, "GFX VML2 Walker Mem Ecc Status: 0x%x!\n", data); - WREG32_SOC15(GC, GET_INST(GC, xcc_id), regVML2_WALKER_MEM_ECC_STATUS, - 0x3); - } -} - -static void gfx_v9_4_3_log_cu_timeout_status(struct amdgpu_device *adev, - uint32_t status, int xcc_id) -{ - struct amdgpu_cu_info *cu_info = &adev->gfx.cu_info; - uint32_t i, simd, wave; - uint32_t wave_status; - uint32_t wave_pc_lo, wave_pc_hi; - uint32_t wave_exec_lo, wave_exec_hi; - uint32_t wave_inst_dw0, wave_inst_dw1; - uint32_t wave_ib_sts; - - for (i = 0; i < 32; i++) { - if (!((i << 1) & status)) - continue; - - simd = i / cu_info->max_waves_per_simd; - wave = i % cu_info->max_waves_per_simd; - - wave_status = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_STATUS); - wave_pc_lo = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_PC_LO); - wave_pc_hi = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_PC_HI); - wave_exec_lo = - wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_EXEC_LO); - wave_exec_hi = - wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_EXEC_HI); - wave_inst_dw0 = - wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_INST_DW0); - wave_inst_dw1 = - wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_INST_DW1); - wave_ib_sts = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_IB_STS); - - dev_info( - adev->dev, - "\t SIMD %d, Wave %d: status 0x%x, pc 0x%llx, exec 0x%llx, inst 0x%llx, ib_sts 0x%x\n", - simd, wave, wave_status, - ((uint64_t)wave_pc_hi << 32 | wave_pc_lo), - ((uint64_t)wave_exec_hi << 32 | wave_exec_lo), - ((uint64_t)wave_inst_dw1 << 32 | wave_inst_dw0), - wave_ib_sts); - } -} - -static void gfx_v9_4_3_inst_query_sq_timeout_status(struct amdgpu_device *adev, - int xcc_id) -{ - uint32_t se_idx, sh_idx, cu_idx; - uint32_t status; - - mutex_lock(&adev->grbm_idx_mutex); - for (se_idx = 0; se_idx < adev->gfx.config.max_shader_engines; se_idx++) { - for (sh_idx = 0; sh_idx < adev->gfx.config.max_sh_per_se; sh_idx++) { - for (cu_idx = 0; cu_idx < adev->gfx.config.max_cu_per_sh; cu_idx++) { - gfx_v9_4_3_xcc_select_se_sh(adev, se_idx, sh_idx, - cu_idx, xcc_id); - status = RREG32_SOC15(GC, GET_INST(GC, xcc_id), - regSQ_TIMEOUT_STATUS); - if (status != 0) { - dev_info( - adev->dev, - "GFX Watchdog Timeout: SE %d, SH %d, CU %d\n", - se_idx, sh_idx, cu_idx); - gfx_v9_4_3_log_cu_timeout_status( - adev, status, xcc_id); - } - /* clear old status */ - WREG32_SOC15(GC, GET_INST(GC, xcc_id), - regSQ_TIMEOUT_STATUS, 0); - } - } - } - gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, - xcc_id); - mutex_unlock(&adev->grbm_idx_mutex); -} - -static void gfx_v9_4_3_inst_query_ras_err_status(struct amdgpu_device *adev, - void *ras_error_status, int xcc_id) -{ - gfx_v9_4_3_inst_query_utc_err_status(adev, xcc_id); - gfx_v9_4_3_inst_query_sq_timeout_status(adev, xcc_id); -} - -static void gfx_v9_4_3_inst_reset_utc_err_status(struct amdgpu_device *adev, - int xcc_id) -{ - WREG32_SOC15(GC, GET_INST(GC, xcc_id), regUTCL2_MEM_ECC_STATUS, 0x3); - WREG32_SOC15(GC, GET_INST(GC, xcc_id), regVML2_MEM_ECC_STATUS, 0x3); - WREG32_SOC15(GC, GET_INST(GC, xcc_id), regVML2_WALKER_MEM_ECC_STATUS, 0x3); -} - -static void gfx_v9_4_3_inst_reset_sq_timeout_status(struct amdgpu_device *adev, - int xcc_id) -{ - uint32_t se_idx, sh_idx, cu_idx; - - mutex_lock(&adev->grbm_idx_mutex); - for (se_idx = 0; se_idx < adev->gfx.config.max_shader_engines; se_idx++) { - for (sh_idx = 0; sh_idx < adev->gfx.config.max_sh_per_se; sh_idx++) { - for (cu_idx = 0; cu_idx < adev->gfx.config.max_cu_per_sh; cu_idx++) { - gfx_v9_4_3_xcc_select_se_sh(adev, se_idx, sh_idx, - cu_idx, xcc_id); - WREG32_SOC15(GC, GET_INST(GC, xcc_id), - regSQ_TIMEOUT_STATUS, 0); - } - } - } - gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, - xcc_id); - mutex_unlock(&adev->grbm_idx_mutex); -} - -static void gfx_v9_4_3_inst_reset_ras_err_status(struct amdgpu_device *adev, - void *ras_error_status, int xcc_id) -{ - gfx_v9_4_3_inst_reset_utc_err_status(adev, xcc_id); - gfx_v9_4_3_inst_reset_sq_timeout_status(adev, xcc_id); -} - static void gfx_v9_4_3_inst_enable_watchdog_timer(struct amdgpu_device *adev, void *ras_error_status, int xcc_id) { @@ -4067,16 +3923,6 @@ static void gfx_v9_4_3_reset_ras_error_count(struct amdgpu_device *adev) amdgpu_gfx_ras_error_func(adev, NULL, gfx_v9_4_3_inst_reset_ras_err_count); } -static void gfx_v9_4_3_query_ras_error_status(struct amdgpu_device *adev) -{ - amdgpu_gfx_ras_error_func(adev, NULL, gfx_v9_4_3_inst_query_ras_err_status); -} - -static void gfx_v9_4_3_reset_ras_error_status(struct amdgpu_device *adev) -{ - amdgpu_gfx_ras_error_func(adev, NULL, gfx_v9_4_3_inst_reset_ras_err_status); -} - static void gfx_v9_4_3_enable_watchdog_timer(struct amdgpu_device *adev) { amdgpu_gfx_ras_error_func(adev, NULL, gfx_v9_4_3_inst_enable_watchdog_timer); @@ -4394,8 +4240,6 @@ struct amdgpu_xcp_ip_funcs gfx_v9_4_3_xcp_funcs = { struct amdgpu_ras_block_hw_ops gfx_v9_4_3_ras_ops = { .query_ras_error_count = &gfx_v9_4_3_query_ras_error_count, .reset_ras_error_count = &gfx_v9_4_3_reset_ras_error_count, - .query_ras_error_status = &gfx_v9_4_3_query_ras_error_status, - .reset_ras_error_status = &gfx_v9_4_3_reset_ras_error_status, }; struct amdgpu_gfx_ras gfx_v9_4_3_ras = { diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c index 95d06da54..49aecdcee 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c @@ -456,10 +456,12 @@ static void gfxhub_v1_2_xcc_gart_disable(struct amdgpu_device *adev, WREG32_SOC15_RLC(GC, GET_INST(GC, j), regMC_VM_MX_L1_TLB_CNTL, tmp); /* Setup L2 cache */ - tmp = RREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL); - tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0); - WREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL, tmp); - WREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL3, 0); + if (!amdgpu_sriov_vf(adev)) { + tmp = RREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0); + WREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL, tmp); + WREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL3, 0); + } } } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index ced2e1bdc..e67a62db9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -883,7 +883,7 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, * GRBM interface. */ if ((vmhub == AMDGPU_GFXHUB(0)) && - (adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 4, 2))) + (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(9, 4, 2))) RREG32_NO_KIQ(req); for (j = 0; j < adev->usec_timeout; j++) { diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c index bc38b90f8..88ea58d5c 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c @@ -674,14 +674,6 @@ static int jpeg_v4_0_set_powergating_state(void *handle, return ret; } -static int jpeg_v4_0_set_interrupt_state(struct amdgpu_device *adev, - struct amdgpu_irq_src *source, - unsigned type, - enum amdgpu_interrupt_state state) -{ - return 0; -} - static int jpeg_v4_0_set_ras_interrupt_state(struct amdgpu_device *adev, struct amdgpu_irq_src *source, unsigned int type, @@ -765,7 +757,6 @@ static void jpeg_v4_0_set_dec_ring_funcs(struct amdgpu_device *adev) } static const struct amdgpu_irq_src_funcs jpeg_v4_0_irq_funcs = { - .set = jpeg_v4_0_set_interrupt_state, .process = jpeg_v4_0_process_interrupt, }; diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c index 6ede85b28..78b74daf4 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c @@ -181,7 +181,6 @@ static int jpeg_v4_0_5_hw_fini(void *handle) RREG32_SOC15(JPEG, 0, regUVD_JRBC_STATUS)) jpeg_v4_0_5_set_powergating_state(adev, AMD_PG_STATE_GATE); } - amdgpu_irq_put(adev, &adev->jpeg.inst->irq, 0); return 0; } @@ -516,14 +515,6 @@ static int jpeg_v4_0_5_set_powergating_state(void *handle, return ret; } -static int jpeg_v4_0_5_set_interrupt_state(struct amdgpu_device *adev, - struct amdgpu_irq_src *source, - unsigned type, - enum amdgpu_interrupt_state state) -{ - return 0; -} - static int jpeg_v4_0_5_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -603,7 +594,6 @@ static void jpeg_v4_0_5_set_dec_ring_funcs(struct amdgpu_device *adev) } static const struct amdgpu_irq_src_funcs jpeg_v4_0_5_irq_funcs = { - .set = jpeg_v4_0_5_set_interrupt_state, .process = jpeg_v4_0_5_process_interrupt, }; diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c index 4dfec56e1..26d71a223 100644 --- a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c @@ -408,6 +408,8 @@ static int mes_v11_0_set_hw_resources(struct amdgpu_mes *mes) mes_set_hw_res_pkt.enable_reg_active_poll = 1; mes_set_hw_res_pkt.enable_level_process_quantum_check = 1; mes_set_hw_res_pkt.oversubscription_timer = 50; + mes_set_hw_res_pkt.enable_mes_event_int_logging = 1; + mes_set_hw_res_pkt.event_intr_history_gpu_mc_ptr = mes->event_log_gpu_addr; return mes_v11_0_submit_pkt_and_poll_completion(mes, &mes_set_hw_res_pkt, sizeof(mes_set_hw_res_pkt), diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c index 9b0146732..fb53aacdc 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c @@ -652,8 +652,8 @@ static void mmhub_v1_8_inst_query_ras_error_count(struct amdgpu_device *adev, AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, &ue_count); - amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, ce_count); - amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, ue_count); + amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, NULL, ce_count); + amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, NULL, ue_count); } static void mmhub_v1_8_query_ras_error_count(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c index 6d24c8492..19986ff6a 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -401,8 +401,7 @@ static void nbio_v7_4_handle_ras_controller_intr_no_bifring(struct amdgpu_device if (err_data.ce_count) dev_info(adev->dev, "%ld correctable hardware " - "errors detected in %s block, " - "no user action is needed.\n", + "errors detected in %s block\n", obj->err_data.ce_count, get_ras_block_str(adev->nbio.ras_if)); diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c index db013f861..b4723d68e 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c @@ -603,8 +603,7 @@ static void nbio_v7_9_handle_ras_controller_intr_no_bifring(struct amdgpu_device if (err_data.ce_count) dev_info(adev->dev, "%ld correctable hardware " - "errors detected in %s block, " - "no user action is needed.\n", + "errors detected in %s block\n", obj->err_data.ce_count, get_ras_block_str(adev->nbio.ras_if)); diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c index 8eed9a73d..1caed0163 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c @@ -366,7 +366,8 @@ static void sdma_v4_4_2_ring_emit_hdp_flush(struct amdgpu_ring *ring) u32 ref_and_mask = 0; const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg; - ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 << ring->me; + ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 + << (ring->me % adev->sdma.num_inst_per_aid); sdma_v4_4_2_wait_reg_mem(ring, 0, 1, adev->nbio.funcs->get_hdp_flush_done_offset(adev), @@ -2135,7 +2136,7 @@ static void sdma_v4_4_2_inst_query_ras_error_count(struct amdgpu_device *adev, AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, &ue_count); - amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, ue_count); + amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, NULL, ue_count); } static void sdma_v4_4_2_query_ras_error_count(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c index 0058f3f7c..46f9f58d8 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c @@ -292,17 +292,21 @@ static void sdma_v5_2_ring_emit_hdp_flush(struct amdgpu_ring *ring) u32 ref_and_mask = 0; const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg; - ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 << ring->me; - - amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) | - SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(1) | - SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* == */ - amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_done_offset(adev)) << 2); - amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_req_offset(adev)) << 2); - amdgpu_ring_write(ring, ref_and_mask); /* reference */ - amdgpu_ring_write(ring, ref_and_mask); /* mask */ - amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | - SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */ + if (ring->me > 1) { + amdgpu_asic_flush_hdp(adev, ring); + } else { + ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 << ring->me; + + amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) | + SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(1) | + SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* == */ + amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_done_offset(adev)) << 2); + amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_req_offset(adev)) << 2); + amdgpu_ring_write(ring, ref_and_mask); /* reference */ + amdgpu_ring_write(ring, ref_and_mask); /* mask */ + amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | + SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */ + } } /** diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index f9ba18030..1c614451d 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -925,6 +925,7 @@ static const struct amdgpu_asic_funcs aqua_vanjaram_asic_funcs = .pre_asic_init = &soc15_pre_asic_init, .query_video_codecs = &soc15_query_video_codecs, .encode_ext_smn_addressing = &aqua_vanjaram_encode_ext_smn_addressing, + .get_reg_state = &aqua_vanjaram_get_reg_state, }; static int soc15_common_early_init(void *handle) diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.h b/drivers/gpu/drm/amd/amdgpu/soc15.h index eac54042c..1444b7765 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15.h @@ -27,6 +27,7 @@ #include "nbio_v6_1.h" #include "nbio_v7_0.h" #include "nbio_v7_4.h" +#include "amdgpu_reg_state.h" extern const struct amdgpu_ip_block_version vega10_common_ip_block; @@ -114,6 +115,9 @@ int aldebaran_reg_base_init(struct amdgpu_device *adev); void aqua_vanjaram_ip_map_init(struct amdgpu_device *adev); u64 aqua_vanjaram_encode_ext_smn_addressing(int ext_id); int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev); +ssize_t aqua_vanjaram_get_reg_state(struct amdgpu_device *adev, + enum amdgpu_reg_state reg_state, void *buf, + size_t max_size); void vega10_doorbell_index_init(struct amdgpu_device *adev); void vega20_doorbell_index_init(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/soc21.c b/drivers/gpu/drm/amd/amdgpu/soc21.c index 4d7188912..fd4505fa4 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc21.c +++ b/drivers/gpu/drm/amd/amdgpu/soc21.c @@ -450,10 +450,8 @@ static bool soc21_need_full_reset(struct amdgpu_device *adev) { switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { case IP_VERSION(11, 0, 0): - return amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC); case IP_VERSION(11, 0, 2): case IP_VERSION(11, 0, 3): - return false; default: return true; } @@ -714,7 +712,10 @@ static int soc21_common_early_init(void *handle) AMD_PG_SUPPORT_VCN | AMD_PG_SUPPORT_JPEG | AMD_PG_SUPPORT_GFX_PG; - adev->external_rev_id = adev->rev_id + 0x1; + if (adev->rev_id == 0) + adev->external_rev_id = 0x1; + else + adev->external_rev_id = adev->rev_id + 0x10; break; default: /* FIXME: not supported yet */ @@ -832,10 +833,35 @@ static int soc21_common_suspend(void *handle) return soc21_common_hw_fini(adev); } +static bool soc21_need_reset_on_resume(struct amdgpu_device *adev) +{ + u32 sol_reg1, sol_reg2; + + /* Will reset for the following suspend abort cases. + * 1) Only reset dGPU side. + * 2) S3 suspend got aborted and TOS is active. + */ + if (!(adev->flags & AMD_IS_APU) && adev->in_s3 && + !adev->suspend_complete) { + sol_reg1 = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_81); + msleep(100); + sol_reg2 = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_81); + + return (sol_reg1 != sol_reg2); + } + + return false; +} + static int soc21_common_resume(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (soc21_need_reset_on_resume(adev)) { + dev_info(adev->dev, "S3 suspend aborted, resetting..."); + soc21_asic_reset(adev); + } + return soc21_common_hw_init(adev); } diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c index e9c2ff74f..7458a218e 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c @@ -26,6 +26,7 @@ #include "amdgpu.h" #include "umc/umc_12_0_0_offset.h" #include "umc/umc_12_0_0_sh_mask.h" +#include "mp/mp_13_0_6_sh_mask.h" const uint32_t umc_v12_0_channel_idx_tbl[] @@ -88,16 +89,26 @@ static void umc_v12_0_reset_error_count(struct amdgpu_device *adev) umc_v12_0_reset_error_count_per_channel, NULL); } -bool umc_v12_0_is_uncorrectable_error(uint64_t mc_umc_status) +bool umc_v12_0_is_uncorrectable_error(struct amdgpu_device *adev, uint64_t mc_umc_status) { + if (amdgpu_ras_is_poison_mode_supported(adev) && + (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)) + return true; + return ((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 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)); } -bool umc_v12_0_is_correctable_error(uint64_t mc_umc_status) +bool umc_v12_0_is_correctable_error(struct amdgpu_device *adev, uint64_t mc_umc_status) { + if (amdgpu_ras_is_poison_mode_supported(adev) && + (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)) + return false; + return (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 || (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 && @@ -105,7 +116,7 @@ bool umc_v12_0_is_correctable_error(uint64_t mc_umc_status) /* Identify data parity error in replay mode */ ((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, ErrorCodeExt) == 0x5 || REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, ErrorCodeExt) == 0xb) && - !(umc_v12_0_is_uncorrectable_error(mc_umc_status))))); + !(umc_v12_0_is_uncorrectable_error(adev, mc_umc_status))))); } static void umc_v12_0_query_correctable_error_count(struct amdgpu_device *adev, @@ -124,7 +135,7 @@ static void umc_v12_0_query_correctable_error_count(struct amdgpu_device *adev, mc_umc_status = RREG64_PCIE_EXT((mc_umc_status_addr + umc_reg_offset) * 4); - if (umc_v12_0_is_correctable_error(mc_umc_status)) + if (umc_v12_0_is_correctable_error(adev, mc_umc_status)) *error_count += 1; } @@ -142,7 +153,7 @@ static void umc_v12_0_query_uncorrectable_error_count(struct amdgpu_device *adev mc_umc_status = RREG64_PCIE_EXT((mc_umc_status_addr + umc_reg_offset) * 4); - if (umc_v12_0_is_uncorrectable_error(mc_umc_status)) + if (umc_v12_0_is_uncorrectable_error(adev, mc_umc_status)) *error_count += 1; } @@ -166,8 +177,8 @@ static int umc_v12_0_query_error_count(struct amdgpu_device *adev, umc_v12_0_query_correctable_error_count(adev, umc_reg_offset, &ce_count); umc_v12_0_query_uncorrectable_error_count(adev, umc_reg_offset, &ue_count); - amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, ue_count); - amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, ce_count); + amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, NULL, ue_count); + amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, NULL, ce_count); return 0; } @@ -360,6 +371,59 @@ static int umc_v12_0_err_cnt_init_per_channel(struct amdgpu_device *adev, return 0; } +static void umc_v12_0_ecc_info_query_ras_error_count(struct amdgpu_device *adev, + void *ras_error_status) +{ + amdgpu_mca_smu_log_ras_error(adev, + AMDGPU_RAS_BLOCK__UMC, AMDGPU_MCA_ERROR_TYPE_CE, ras_error_status); + amdgpu_mca_smu_log_ras_error(adev, + AMDGPU_RAS_BLOCK__UMC, AMDGPU_MCA_ERROR_TYPE_UE, ras_error_status); +} + +static void umc_v12_0_ecc_info_query_ras_error_address(struct amdgpu_device *adev, + void *ras_error_status) +{ + struct ras_err_node *err_node; + uint64_t mc_umc_status; + struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; + + for_each_ras_error(err_node, err_data) { + mc_umc_status = err_node->err_info.err_addr.err_status; + if (!mc_umc_status) + continue; + + if (umc_v12_0_is_uncorrectable_error(adev, mc_umc_status)) { + uint64_t mca_addr, err_addr, mca_ipid; + uint32_t InstanceIdLo; + struct amdgpu_smuio_mcm_config_info *mcm_info; + + mcm_info = &err_node->err_info.mcm_info; + mca_addr = err_node->err_info.err_addr.err_addr; + mca_ipid = err_node->err_info.err_addr.err_ipid; + + err_addr = REG_GET_FIELD(mca_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr); + InstanceIdLo = REG_GET_FIELD(mca_ipid, MCMP1_IPIDT0, InstanceIdLo); + + dev_info(adev->dev, "UMC:IPID:0x%llx, aid:%d, inst:%d, ch:%d, err_addr:0x%llx\n", + mca_ipid, + mcm_info->die_id, + MCA_IPID_LO_2_UMC_INST(InstanceIdLo), + MCA_IPID_LO_2_UMC_CH(InstanceIdLo), + err_addr); + + umc_v12_0_convert_error_address(adev, + err_data, err_addr, + MCA_IPID_LO_2_UMC_CH(InstanceIdLo), + MCA_IPID_LO_2_UMC_INST(InstanceIdLo), + mcm_info->die_id); + + /* Clear umc error address content */ + memset(&err_node->err_info.err_addr, + 0, sizeof(err_node->err_info.err_addr)); + } + } +} + static void umc_v12_0_err_cnt_init(struct amdgpu_device *adev) { amdgpu_umc_loop_channels(adev, @@ -386,4 +450,6 @@ struct amdgpu_umc_ras umc_v12_0_ras = { }, .err_cnt_init = umc_v12_0_err_cnt_init, .query_ras_poison_mode = umc_v12_0_query_ras_poison_mode, + .ecc_info_query_ras_error_count = umc_v12_0_ecc_info_query_ras_error_count, + .ecc_info_query_ras_error_address = umc_v12_0_ecc_info_query_ras_error_address, }; diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h index b34b1e358..e8de3a922 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h +++ b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h @@ -117,8 +117,12 @@ (pa) |= (UMC_V12_0_CHANNEL_HASH_CH6(channel_idx, pa) << UMC_V12_0_PA_CH6_BIT); \ } while (0) -bool umc_v12_0_is_uncorrectable_error(uint64_t mc_umc_status); -bool umc_v12_0_is_correctable_error(uint64_t mc_umc_status); +#define MCA_IPID_LO_2_UMC_CH(_ipid_lo) (((((_ipid_lo) >> 20) & 0x1) * 4) + \ + (((_ipid_lo) >> 12) & 0xF)) +#define MCA_IPID_LO_2_UMC_INST(_ipid_lo) (((_ipid_lo) >> 21) & 0x7) + +bool umc_v12_0_is_uncorrectable_error(struct amdgpu_device *adev, uint64_t mc_umc_status); +bool umc_v12_0_is_correctable_error(struct amdgpu_device *adev, uint64_t mc_umc_status); extern const uint32_t umc_v12_0_channel_idx_tbl[] diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c index 530549314..a3ee3c4c6 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c @@ -64,7 +64,7 @@ static void umc_v6_7_query_error_status_helper(struct amdgpu_device *adev, uint64_t reg_value; if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1) - dev_info(adev->dev, "Deferred error, no user action is needed.\n"); + dev_info(adev->dev, "Deferred error\n"); if (mc_umc_status) dev_info(adev->dev, "MCA STATUS 0x%llx, umc_reg_offset 0x%x\n", mc_umc_status, umc_reg_offset); diff --git a/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c b/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c index 8e7b763cf..fd23cd348 100644 --- a/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c @@ -225,6 +225,8 @@ static int umsch_mm_v4_0_ring_start(struct amdgpu_umsch_mm *umsch) WREG32_SOC15(VCN, 0, regVCN_UMSCH_RB_SIZE, ring->ring_size); + ring->wptr = 0; + data = RREG32_SOC15(VCN, 0, regVCN_RB_ENABLE); data &= ~(VCN_RB_ENABLE__AUDIO_RB_EN_MASK); WREG32_SOC15(VCN, 0, regVCN_RB_ENABLE, data); diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c index 48bfcd0d5..8ab01ae91 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c @@ -100,6 +100,31 @@ static int vcn_v4_0_early_init(void *handle) return amdgpu_vcn_early_init(adev); } +static int vcn_v4_0_fw_shared_init(struct amdgpu_device *adev, int inst_idx) +{ + volatile struct amdgpu_vcn4_fw_shared *fw_shared; + + fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr; + fw_shared->present_flag_0 = cpu_to_le32(AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE); + fw_shared->sq.is_enabled = 1; + + fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_SMU_DPM_INTERFACE_FLAG); + fw_shared->smu_dpm_interface.smu_interface_type = (adev->flags & AMD_IS_APU) ? + AMDGPU_VCN_SMU_DPM_INTERFACE_APU : AMDGPU_VCN_SMU_DPM_INTERFACE_DGPU; + + if (amdgpu_ip_version(adev, VCN_HWIP, 0) == + IP_VERSION(4, 0, 2)) { + fw_shared->present_flag_0 |= AMDGPU_FW_SHARED_FLAG_0_DRM_KEY_INJECT; + fw_shared->drm_key_wa.method = + AMDGPU_DRM_KEY_INJECT_WORKAROUND_VCNFW_ASD_HANDSHAKING; + } + + if (amdgpu_vcnfw_log) + amdgpu_vcn_fwlog_init(&adev->vcn.inst[inst_idx]); + + return 0; +} + /** * vcn_v4_0_sw_init - sw init for VCN block * @@ -124,8 +149,6 @@ static int vcn_v4_0_sw_init(void *handle) return r; for (i = 0; i < adev->vcn.num_vcn_inst; i++) { - volatile struct amdgpu_vcn4_fw_shared *fw_shared; - if (adev->vcn.harvest_config & (1 << i)) continue; @@ -161,23 +184,7 @@ static int vcn_v4_0_sw_init(void *handle) if (r) return r; - fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr; - fw_shared->present_flag_0 = cpu_to_le32(AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE); - fw_shared->sq.is_enabled = 1; - - fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_SMU_DPM_INTERFACE_FLAG); - fw_shared->smu_dpm_interface.smu_interface_type = (adev->flags & AMD_IS_APU) ? - AMDGPU_VCN_SMU_DPM_INTERFACE_APU : AMDGPU_VCN_SMU_DPM_INTERFACE_DGPU; - - if (amdgpu_ip_version(adev, VCN_HWIP, 0) == - IP_VERSION(4, 0, 2)) { - fw_shared->present_flag_0 |= AMDGPU_FW_SHARED_FLAG_0_DRM_KEY_INJECT; - fw_shared->drm_key_wa.method = - AMDGPU_DRM_KEY_INJECT_WORKAROUND_VCNFW_ASD_HANDSHAKING; - } - - if (amdgpu_vcnfw_log) - amdgpu_vcn_fwlog_init(&adev->vcn.inst[i]); + vcn_v4_0_fw_shared_init(adev, i); } if (amdgpu_sriov_vf(adev)) { @@ -1273,6 +1280,9 @@ static int vcn_v4_0_start_sriov(struct amdgpu_device *adev) if (adev->vcn.harvest_config & (1 << i)) continue; + // Must re/init fw_shared at beginning + vcn_v4_0_fw_shared_init(adev, i); + table_size = 0; MMSCH_V4_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCN, i, @@ -2008,22 +2018,6 @@ static int vcn_v4_0_set_powergating_state(void *handle, enum amd_powergating_sta } /** - * vcn_v4_0_set_interrupt_state - set VCN block interrupt state - * - * @adev: amdgpu_device pointer - * @source: interrupt sources - * @type: interrupt types - * @state: interrupt states - * - * Set VCN block interrupt state - */ -static int vcn_v4_0_set_interrupt_state(struct amdgpu_device *adev, struct amdgpu_irq_src *source, - unsigned type, enum amdgpu_interrupt_state state) -{ - return 0; -} - -/** * vcn_v4_0_set_ras_interrupt_state - set VCN block RAS interrupt state * * @adev: amdgpu_device pointer @@ -2087,7 +2081,6 @@ static int vcn_v4_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_ } static const struct amdgpu_irq_src_funcs vcn_v4_0_irq_funcs = { - .set = vcn_v4_0_set_interrupt_state, .process = vcn_v4_0_process_interrupt, }; diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c index 2eda30e78..49e4c3c09 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c @@ -269,8 +269,6 @@ static int vcn_v4_0_5_hw_fini(void *handle) vcn_v4_0_5_set_powergating_state(adev, AMD_PG_STATE_GATE); } } - - amdgpu_irq_put(adev, &adev->vcn.inst[i].irq, 0); } return 0; @@ -1669,22 +1667,6 @@ static int vcn_v4_0_5_set_powergating_state(void *handle, enum amd_powergating_s } /** - * vcn_v4_0_5_set_interrupt_state - set VCN block interrupt state - * - * @adev: amdgpu_device pointer - * @source: interrupt sources - * @type: interrupt types - * @state: interrupt states - * - * Set VCN block interrupt state - */ -static int vcn_v4_0_5_set_interrupt_state(struct amdgpu_device *adev, struct amdgpu_irq_src *source, - unsigned type, enum amdgpu_interrupt_state state) -{ - return 0; -} - -/** * vcn_v4_0_5_process_interrupt - process VCN block interrupt * * @adev: amdgpu_device pointer @@ -1726,7 +1708,6 @@ static int vcn_v4_0_5_process_interrupt(struct amdgpu_device *adev, struct amdgp } static const struct amdgpu_irq_src_funcs vcn_v4_0_5_irq_funcs = { - .set = vcn_v4_0_5_set_interrupt_state, .process = vcn_v4_0_5_process_interrupt, }; diff --git a/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c b/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c index 174f13eff..d20060a51 100644 --- a/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c +++ b/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c @@ -96,6 +96,10 @@ static int vpe_v6_1_load_microcode(struct amdgpu_vpe *vpe) adev->vpe.cmdbuf_cpu_addr[1] = f32_cntl; amdgpu_vpe_psp_update_sram(adev); + + /* Config DPM */ + amdgpu_vpe_configure_dpm(vpe); + return 0; } @@ -128,6 +132,8 @@ static int vpe_v6_1_load_microcode(struct amdgpu_vpe *vpe) } vpe_v6_1_halt(vpe, false); + /* Config DPM */ + amdgpu_vpe_configure_dpm(vpe); return 0; } @@ -264,6 +270,15 @@ static int vpe_v6_1_set_regs(struct amdgpu_vpe *vpe) vpe->regs.queue0_rb_wptr_hi = regVPEC_QUEUE0_RB_WPTR_HI; vpe->regs.queue0_preempt = regVPEC_QUEUE0_PREEMPT; + vpe->regs.dpm_enable = regVPEC_PUB_DUMMY2; + vpe->regs.dpm_pratio = regVPEC_QUEUE6_DUMMY4; + vpe->regs.dpm_request_interval = regVPEC_QUEUE5_DUMMY3; + vpe->regs.dpm_decision_threshold = regVPEC_QUEUE5_DUMMY4; + vpe->regs.dpm_busy_clamp_threshold = regVPEC_QUEUE7_DUMMY2; + vpe->regs.dpm_idle_clamp_threshold = regVPEC_QUEUE7_DUMMY3; + vpe->regs.dpm_request_lv = regVPEC_QUEUE7_DUMMY1; + vpe->regs.context_indicator = regVPEC_QUEUE6_DUMMY3; + return 0; } |