diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
91 files changed, 3290 insertions, 1198 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index ee85e8aba636..6331a11299d0 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -56,7 +56,9 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \ amdgpu_gmc.o amdgpu_mmhub.o amdgpu_xgmi.o amdgpu_csa.o amdgpu_ras.o amdgpu_vm_cpu.o \ amdgpu_vm_sdma.o amdgpu_discovery.o amdgpu_ras_eeprom.o amdgpu_nbio.o \ amdgpu_umc.o smu_v11_0_i2c.o amdgpu_fru_eeprom.o amdgpu_rap.o \ - amdgpu_fw_attestation.o amdgpu_securedisplay.o + amdgpu_fw_attestation.o amdgpu_securedisplay.o amdgpu_hdp.o + +amdgpu-$(CONFIG_PROC_FS) += amdgpu_fdinfo.o amdgpu-$(CONFIG_PERF_EVENTS) += amdgpu_pmu.o @@ -71,7 +73,8 @@ amdgpu-y += \ vi.o mxgpu_vi.o nbio_v6_1.o soc15.o emu_soc.o mxgpu_ai.o nbio_v7_0.o vega10_reg_init.o \ vega20_reg_init.o nbio_v7_4.o nbio_v2_3.o nv.o navi10_reg_init.o navi14_reg_init.o \ arct_reg_init.o navi12_reg_init.o mxgpu_nv.o sienna_cichlid_reg_init.o vangogh_reg_init.o \ - nbio_v7_2.o dimgrey_cavefish_reg_init.o hdp_v4_0.o hdp_v5_0.o aldebaran_reg_init.o aldebaran.o + nbio_v7_2.o dimgrey_cavefish_reg_init.o hdp_v4_0.o hdp_v5_0.o aldebaran_reg_init.o aldebaran.o \ + beige_goby_reg_init.o # add DF block amdgpu-y += \ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index dc3a69296321..ad5f508924b8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -107,6 +107,7 @@ #include "amdgpu_gfxhub.h" #include "amdgpu_df.h" #include "amdgpu_smuio.h" +#include "amdgpu_fdinfo.h" #define MAX_GPU_INSTANCE 16 @@ -1006,6 +1007,7 @@ struct amdgpu_device { struct amdgpu_df df; struct amdgpu_ip_block ip_blocks[AMDGPU_MAX_IP_NUM]; + uint32_t harvest_ip_mask; int num_ip_blocks; struct mutex mn_lock; DECLARE_HASHTABLE(mn_hash, 7); @@ -1073,7 +1075,8 @@ struct amdgpu_device { atomic_t throttling_logging_enabled; struct ratelimit_state throttling_logging_rs; - uint32_t ras_features; + uint32_t ras_hw_enabled; + uint32_t ras_enabled; bool in_pci_err_recovery; struct pci_saved_state *pci_state; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c index 2e9b16fb3fcd..bf2939b6eb43 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c @@ -76,7 +76,7 @@ struct amdgpu_atif { /** * amdgpu_atif_call - call an ATIF method * - * @handle: acpi handle + * @atif: acpi handle * @function: the ATIF function to execute * @params: ATIF function params * @@ -166,7 +166,6 @@ static void amdgpu_atif_parse_functions(struct amdgpu_atif_functions *f, u32 mas /** * amdgpu_atif_verify_interface - verify ATIF * - * @handle: acpi handle * @atif: amdgpu atif struct * * Execute the ATIF_FUNCTION_VERIFY_INTERFACE ATIF function @@ -240,8 +239,7 @@ out: /** * amdgpu_atif_get_notification_params - determine notify configuration * - * @handle: acpi handle - * @n: atif notification configuration struct + * @atif: acpi handle * * Execute the ATIF_FUNCTION_GET_SYSTEM_PARAMETERS ATIF function * to determine if a notifier is used and if so which one @@ -304,7 +302,7 @@ out: /** * amdgpu_atif_query_backlight_caps - get min and max backlight input signal * - * @handle: acpi handle + * @atif: acpi handle * * Execute the QUERY_BRIGHTNESS_TRANSFER_CHARACTERISTICS ATIF function * to determine the acceptable range of backlight values @@ -363,7 +361,7 @@ out: /** * amdgpu_atif_get_sbios_requests - get requested sbios event * - * @handle: acpi handle + * @atif: acpi handle * @req: atif sbios request struct * * Execute the ATIF_FUNCTION_GET_SYSTEM_BIOS_REQUESTS ATIF function @@ -899,6 +897,8 @@ void amdgpu_acpi_fini(struct amdgpu_device *adev) /** * amdgpu_acpi_is_s0ix_supported * + * @adev: amdgpu_device_pointer + * * returns true if supported, false if not. */ bool amdgpu_acpi_is_s0ix_supported(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 5f6696a3c778..bfab2f9fdd17 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -670,10 +670,10 @@ int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct kgd_dev *kgd, uint16_t vmid) return 0; } -int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t pasid) +int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t pasid, + enum TLB_FLUSH_TYPE flush_type) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; - const uint32_t flush_type = 0; bool all_hub = false; if (adev->family == AMDGPU_FAMILY_AI) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index 5ffb07b02810..81264517d532 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -36,13 +36,26 @@ extern uint64_t amdgpu_amdkfd_total_mem_size; +enum TLB_FLUSH_TYPE { + TLB_FLUSH_LEGACY = 0, + TLB_FLUSH_LIGHTWEIGHT, + TLB_FLUSH_HEAVYWEIGHT +}; + struct amdgpu_device; -struct kfd_bo_va_list { - struct list_head bo_list; - struct amdgpu_bo_va *bo_va; - void *kgd_dev; +enum kfd_mem_attachment_type { + KFD_MEM_ATT_SHARED, /* Share kgd_mem->bo or another attachment's */ + KFD_MEM_ATT_USERPTR, /* SG bo to DMA map pages from a userptr bo */ + KFD_MEM_ATT_DMABUF, /* DMAbuf to DMA map TTM BOs */ +}; + +struct kfd_mem_attachment { + struct list_head list; + enum kfd_mem_attachment_type type; bool is_mapped; + struct amdgpu_bo_va *bo_va; + struct amdgpu_device *adev; uint64_t va; uint64_t pte_flags; }; @@ -50,7 +63,8 @@ struct kfd_bo_va_list { struct kgd_mem { struct mutex lock; struct amdgpu_bo *bo; - struct list_head bo_va_list; + struct dma_buf *dmabuf; + struct list_head attachments; /* protected by amdkfd_process_info.lock */ struct ttm_validate_buffer validate_list; struct ttm_validate_buffer resv_list; @@ -75,6 +89,7 @@ struct amdgpu_amdkfd_fence { struct mm_struct *mm; spinlock_t lock; char timeline_name[TASK_COMM_LEN]; + struct svm_range_bo *svm_bo; }; struct amdgpu_kfd_dev { @@ -134,7 +149,8 @@ int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine, void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle); bool amdgpu_amdkfd_have_atomics_support(struct kgd_dev *kgd); int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct kgd_dev *kgd, uint16_t vmid); -int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t pasid); +int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t pasid, + enum TLB_FLUSH_TYPE flush_type); bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid); @@ -148,7 +164,8 @@ int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev, int queue_bit); struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, - struct mm_struct *mm); + struct mm_struct *mm, + struct svm_range_bo *svm_bo); #if IS_ENABLED(CONFIG_HSA_AMD) bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm); struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f); @@ -234,22 +251,27 @@ uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *s }) /* GPUVM API */ +#define drm_priv_to_vm(drm_priv) \ + (&((struct amdgpu_fpriv *) \ + ((struct drm_file *)(drm_priv))->driver_priv)->vm) + int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, struct file *filp, u32 pasid, - void **vm, void **process_info, + void **process_info, struct dma_fence **ef); -void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm); -uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm); +void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *drm_priv); +uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv); int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( struct kgd_dev *kgd, uint64_t va, uint64_t size, - void *vm, struct kgd_mem **mem, + void *drm_priv, struct kgd_mem **mem, uint64_t *offset, uint32_t flags); int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, uint64_t *size); + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv, + uint64_t *size); int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *vm); + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv); int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *vm); + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv); int amdgpu_amdkfd_gpuvm_sync_memory( struct kgd_dev *kgd, struct kgd_mem *mem, bool intr); int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd, @@ -260,7 +282,7 @@ int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, struct kfd_vm_fault_info *info); int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd, struct dma_buf *dmabuf, - uint64_t va, void *vm, + uint64_t va, void *drm_priv, struct kgd_mem **mem, uint64_t *size, uint64_t *mmap_offset); int amdgpu_amdkfd_get_tile_config(struct kgd_dev *kgd, @@ -270,6 +292,7 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void); void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_amdkfd_unreserve_memory_limit(struct amdgpu_bo *bo); +void amdgpu_amdkfd_reserve_system_mem(uint64_t size); #else static inline void amdgpu_amdkfd_gpuvm_init_mem_limits(void) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c index 9ef9f3ddad48..6409d6b1b2df 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c @@ -25,6 +25,7 @@ #include <linux/firmware.h> #include "amdgpu.h" #include "amdgpu_amdkfd.h" +#include "amdgpu_amdkfd_arcturus.h" #include "sdma0/sdma0_4_2_2_offset.h" #include "sdma0/sdma0_4_2_2_sh_mask.h" #include "sdma1/sdma1_4_2_2_offset.h" diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c index 5af464933976..1d0dbff87d3f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c @@ -28,6 +28,7 @@ #include <linux/slab.h> #include <linux/sched/mm.h> #include "amdgpu_amdkfd.h" +#include "kfd_svm.h" static const struct dma_fence_ops amdkfd_fence_ops; static atomic_t fence_seq = ATOMIC_INIT(0); @@ -60,7 +61,8 @@ static atomic_t fence_seq = ATOMIC_INIT(0); */ struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, - struct mm_struct *mm) + struct mm_struct *mm, + struct svm_range_bo *svm_bo) { struct amdgpu_amdkfd_fence *fence; @@ -73,7 +75,7 @@ struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, fence->mm = mm; get_task_comm(fence->timeline_name, current); spin_lock_init(&fence->lock); - + fence->svm_bo = svm_bo; dma_fence_init(&fence->base, &amdkfd_fence_ops, &fence->lock, context, atomic_inc_return(&fence_seq)); @@ -111,6 +113,8 @@ static const char *amdkfd_fence_get_timeline_name(struct dma_fence *f) * a KFD BO and schedules a job to move the BO. * If fence is already signaled return true. * If fence is not signaled schedule a evict KFD process work item. + * + * @f: dma_fence */ static bool amdkfd_fence_enable_signaling(struct dma_fence *f) { @@ -122,16 +126,20 @@ static bool amdkfd_fence_enable_signaling(struct dma_fence *f) if (dma_fence_is_signaled(f)) return true; - if (!kgd2kfd_schedule_evict_and_restore_process(fence->mm, f)) - return true; - + if (!fence->svm_bo) { + if (!kgd2kfd_schedule_evict_and_restore_process(fence->mm, f)) + return true; + } else { + if (!svm_range_schedule_evict_svm_bo(fence)) + return true; + } return false; } /** * amdkfd_fence_release - callback that fence can be freed * - * @fence: fence + * @f: dma_fence * * This function is called when the reference count becomes zero. * Drops the mm_struct reference and RCU schedules freeing up the fence. 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 b43e68fc1378..ed3014fbb563 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -719,7 +719,7 @@ static void unlock_spi_csq_mutexes(struct amdgpu_device *adev) } /** - * @get_wave_count: Read device registers to get number of waves in flight for + * get_wave_count: Read device registers to get number of waves in flight for * a particular queue. The method also returns the VMID associated with the * queue. * @@ -755,19 +755,19 @@ static void get_wave_count(struct amdgpu_device *adev, int queue_idx, } /** - * @kgd_gfx_v9_get_cu_occupancy: Reads relevant registers associated with each + * kgd_gfx_v9_get_cu_occupancy: Reads relevant registers associated with each * shader engine and aggregates the number of waves that are in flight for the * process whose pasid is provided as a parameter. The process could have ZERO * or more queues running and submitting waves to compute units. * * @kgd: Handle of device from which to get number of waves in flight * @pasid: Identifies the process for which this query call is invoked - * @wave_cnt: Output parameter updated with number of waves in flight that + * @pasid_wave_cnt: Output parameter updated with number of waves in flight that * belong to process with given pasid * @max_waves_per_cu: Output parameter updated with maximum number of waves * possible per Compute Unit * - * @note: It's possible that the device has too many queues (oversubscription) + * Note: It's possible that the device has too many queues (oversubscription) * in which case a VMID could be remapped to a different PASID. This could lead * to an iaccurate wave count. Following is a high-level sequence: * Time T1: vmid = getVmid(); vmid is associated with Pasid P1 diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index 7d4118c8128a..928e8d57cd08 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -33,9 +33,6 @@ #include <uapi/linux/kfd_ioctl.h> #include "amdgpu_xgmi.h" -/* BO flag to indicate a KFD userptr BO */ -#define AMDGPU_AMDKFD_USERPTR_BO (1ULL << 63) - /* Userptr restore delay, just long enough to allow consecutive VM * changes to accumulate */ @@ -75,16 +72,16 @@ static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) return (struct amdgpu_device *)kgd; } -static bool check_if_add_bo_to_vm(struct amdgpu_vm *avm, +static bool kfd_mem_is_attached(struct amdgpu_vm *avm, struct kgd_mem *mem) { - struct kfd_bo_va_list *entry; + struct kfd_mem_attachment *entry; - list_for_each_entry(entry, &mem->bo_va_list, bo_list) + list_for_each_entry(entry, &mem->attachments, list) if (entry->bo_va->base.vm == avm) - return false; + return true; - return true; + return false; } /* Set memory usage limits. Current, limits are @@ -108,6 +105,11 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void) (kfd_mem_limit.max_ttm_mem_limit >> 20)); } +void amdgpu_amdkfd_reserve_system_mem(uint64_t size) +{ + kfd_mem_limit.system_mem_used += size; +} + /* Estimate page table size needed to represent a given memory size * * With 4KB pages, we need one 8 byte PTE for each 4KB of memory @@ -217,7 +219,7 @@ void amdgpu_amdkfd_unreserve_memory_limit(struct amdgpu_bo *bo) u32 domain = bo->preferred_domains; bool sg = (bo->preferred_domains == AMDGPU_GEM_DOMAIN_CPU); - if (bo->flags & AMDGPU_AMDKFD_USERPTR_BO) { + if (bo->flags & AMDGPU_AMDKFD_CREATE_USERPTR_BO) { domain = AMDGPU_GEM_DOMAIN_CPU; sg = false; } @@ -431,7 +433,8 @@ static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem) mapping_flags |= coherent ? AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW; else - mapping_flags |= AMDGPU_VM_MTYPE_UC; + mapping_flags |= coherent ? + AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; } else { mapping_flags |= coherent ? AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; @@ -450,7 +453,8 @@ static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem) if (adev->gmc.xgmi.connected_to_cpu) snoop = true; } else { - mapping_flags |= AMDGPU_VM_MTYPE_UC; + mapping_flags |= coherent ? + AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; if (amdgpu_xgmi_same_hive(adev, bo_adev)) snoop = true; } @@ -471,87 +475,318 @@ static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem) return pte_flags; } -/* add_bo_to_vm - Add a BO to a VM +static int +kfd_mem_dmamap_userptr(struct kgd_mem *mem, + struct kfd_mem_attachment *attachment) +{ + enum dma_data_direction direction = + mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ? + DMA_BIDIRECTIONAL : DMA_TO_DEVICE; + struct ttm_operation_ctx ctx = {.interruptible = true}; + struct amdgpu_bo *bo = attachment->bo_va->base.bo; + struct amdgpu_device *adev = attachment->adev; + struct ttm_tt *src_ttm = mem->bo->tbo.ttm; + struct ttm_tt *ttm = bo->tbo.ttm; + int ret; + + ttm->sg = kmalloc(sizeof(*ttm->sg), GFP_KERNEL); + if (unlikely(!ttm->sg)) + return -ENOMEM; + + if (WARN_ON(ttm->num_pages != src_ttm->num_pages)) + return -EINVAL; + + /* Same sequence as in amdgpu_ttm_tt_pin_userptr */ + ret = sg_alloc_table_from_pages(ttm->sg, src_ttm->pages, + ttm->num_pages, 0, + (u64)ttm->num_pages << PAGE_SHIFT, + GFP_KERNEL); + if (unlikely(ret)) + goto free_sg; + + ret = dma_map_sgtable(adev->dev, ttm->sg, direction, 0); + if (unlikely(ret)) + goto release_sg; + + drm_prime_sg_to_dma_addr_array(ttm->sg, ttm->dma_address, + ttm->num_pages); + + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT); + ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); + if (ret) + goto unmap_sg; + + return 0; + +unmap_sg: + dma_unmap_sgtable(adev->dev, ttm->sg, direction, 0); +release_sg: + pr_err("DMA map userptr failed: %d\n", ret); + sg_free_table(ttm->sg); +free_sg: + kfree(ttm->sg); + ttm->sg = NULL; + return ret; +} + +static int +kfd_mem_dmamap_dmabuf(struct kfd_mem_attachment *attachment) +{ + struct ttm_operation_ctx ctx = {.interruptible = true}; + struct amdgpu_bo *bo = attachment->bo_va->base.bo; + + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT); + return ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); +} + +static int +kfd_mem_dmamap_attachment(struct kgd_mem *mem, + struct kfd_mem_attachment *attachment) +{ + switch (attachment->type) { + case KFD_MEM_ATT_SHARED: + return 0; + case KFD_MEM_ATT_USERPTR: + return kfd_mem_dmamap_userptr(mem, attachment); + case KFD_MEM_ATT_DMABUF: + return kfd_mem_dmamap_dmabuf(attachment); + default: + WARN_ON_ONCE(1); + } + return -EINVAL; +} + +static void +kfd_mem_dmaunmap_userptr(struct kgd_mem *mem, + struct kfd_mem_attachment *attachment) +{ + enum dma_data_direction direction = + mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ? + DMA_BIDIRECTIONAL : DMA_TO_DEVICE; + struct ttm_operation_ctx ctx = {.interruptible = false}; + struct amdgpu_bo *bo = attachment->bo_va->base.bo; + struct amdgpu_device *adev = attachment->adev; + struct ttm_tt *ttm = bo->tbo.ttm; + + if (unlikely(!ttm->sg)) + return; + + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); + ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); + + dma_unmap_sgtable(adev->dev, ttm->sg, direction, 0); + sg_free_table(ttm->sg); + ttm->sg = NULL; +} + +static void +kfd_mem_dmaunmap_dmabuf(struct kfd_mem_attachment *attachment) +{ + struct ttm_operation_ctx ctx = {.interruptible = true}; + struct amdgpu_bo *bo = attachment->bo_va->base.bo; + + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); + ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); +} + +static void +kfd_mem_dmaunmap_attachment(struct kgd_mem *mem, + struct kfd_mem_attachment *attachment) +{ + switch (attachment->type) { + case KFD_MEM_ATT_SHARED: + break; + case KFD_MEM_ATT_USERPTR: + kfd_mem_dmaunmap_userptr(mem, attachment); + break; + case KFD_MEM_ATT_DMABUF: + kfd_mem_dmaunmap_dmabuf(attachment); + break; + default: + WARN_ON_ONCE(1); + } +} + +static int +kfd_mem_attach_userptr(struct amdgpu_device *adev, struct kgd_mem *mem, + struct amdgpu_bo **bo) +{ + unsigned long bo_size = mem->bo->tbo.base.size; + struct drm_gem_object *gobj; + int ret; + + ret = amdgpu_bo_reserve(mem->bo, false); + if (ret) + return ret; + + ret = amdgpu_gem_object_create(adev, bo_size, 1, + AMDGPU_GEM_DOMAIN_CPU, + 0, ttm_bo_type_sg, + mem->bo->tbo.base.resv, + &gobj); + if (ret) + return ret; + + amdgpu_bo_unreserve(mem->bo); + + *bo = gem_to_amdgpu_bo(gobj); + (*bo)->parent = amdgpu_bo_ref(mem->bo); + + return 0; +} + +static int +kfd_mem_attach_dmabuf(struct amdgpu_device *adev, struct kgd_mem *mem, + struct amdgpu_bo **bo) +{ + struct drm_gem_object *gobj; + + if (!mem->dmabuf) { + mem->dmabuf = amdgpu_gem_prime_export(&mem->bo->tbo.base, + mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ? + DRM_RDWR : 0); + if (IS_ERR(mem->dmabuf)) { + mem->dmabuf = NULL; + return PTR_ERR(mem->dmabuf); + } + } + + gobj = amdgpu_gem_prime_import(&adev->ddev, mem->dmabuf); + if (IS_ERR(gobj)) + return PTR_ERR(gobj); + + /* Import takes an extra reference on the dmabuf. Drop it now to + * avoid leaking it. We only need the one reference in + * kgd_mem->dmabuf. + */ + dma_buf_put(mem->dmabuf); + + *bo = gem_to_amdgpu_bo(gobj); + (*bo)->parent = amdgpu_bo_ref(mem->bo); + + return 0; +} + +/* kfd_mem_attach - Add a BO to a VM * * Everything that needs to bo done only once when a BO is first added * to a VM. It can later be mapped and unmapped many times without * repeating these steps. * + * 0. Create BO for DMA mapping, if needed * 1. Allocate and initialize BO VA entry data structure * 2. Add BO to the VM * 3. Determine ASIC-specific PTE flags * 4. Alloc page tables and directories if needed * 4a. Validate new page tables and directories */ -static int add_bo_to_vm(struct amdgpu_device *adev, struct kgd_mem *mem, - struct amdgpu_vm *vm, bool is_aql, - struct kfd_bo_va_list **p_bo_va_entry) +static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem, + struct amdgpu_vm *vm, bool is_aql) { - int ret; - struct kfd_bo_va_list *bo_va_entry; - struct amdgpu_bo *bo = mem->bo; + struct amdgpu_device *bo_adev = amdgpu_ttm_adev(mem->bo->tbo.bdev); + unsigned long bo_size = mem->bo->tbo.base.size; uint64_t va = mem->va; - struct list_head *list_bo_va = &mem->bo_va_list; - unsigned long bo_size = bo->tbo.base.size; + struct kfd_mem_attachment *attachment[2] = {NULL, NULL}; + struct amdgpu_bo *bo[2] = {NULL, NULL}; + int i, ret; if (!va) { pr_err("Invalid VA when adding BO to VM\n"); return -EINVAL; } - if (is_aql) - va += bo_size; - - bo_va_entry = kzalloc(sizeof(*bo_va_entry), GFP_KERNEL); - if (!bo_va_entry) - return -ENOMEM; + for (i = 0; i <= is_aql; i++) { + attachment[i] = kzalloc(sizeof(*attachment[i]), GFP_KERNEL); + if (unlikely(!attachment[i])) { + ret = -ENOMEM; + goto unwind; + } - pr_debug("\t add VA 0x%llx - 0x%llx to vm %p\n", va, - va + bo_size, vm); + pr_debug("\t add VA 0x%llx - 0x%llx to vm %p\n", va, + va + bo_size, vm); - /* Add BO to VM internal data structures*/ - bo_va_entry->bo_va = amdgpu_vm_bo_add(adev, vm, bo); - if (!bo_va_entry->bo_va) { - ret = -EINVAL; - pr_err("Failed to add BO object to VM. ret == %d\n", - ret); - goto err_vmadd; - } + if (adev == bo_adev || (mem->domain == AMDGPU_GEM_DOMAIN_VRAM && + amdgpu_xgmi_same_hive(adev, bo_adev))) { + /* Mappings on the local GPU and VRAM mappings in the + * local hive share the original BO + */ + attachment[i]->type = KFD_MEM_ATT_SHARED; + bo[i] = mem->bo; + drm_gem_object_get(&bo[i]->tbo.base); + } else if (i > 0) { + /* Multiple mappings on the same GPU share the BO */ + attachment[i]->type = KFD_MEM_ATT_SHARED; + bo[i] = bo[0]; + drm_gem_object_get(&bo[i]->tbo.base); + } else if (amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm)) { + /* Create an SG BO to DMA-map userptrs on other GPUs */ + attachment[i]->type = KFD_MEM_ATT_USERPTR; + ret = kfd_mem_attach_userptr(adev, mem, &bo[i]); + if (ret) + goto unwind; + } else if (mem->domain == AMDGPU_GEM_DOMAIN_GTT && + mem->bo->tbo.type != ttm_bo_type_sg) { + /* GTT BOs use DMA-mapping ability of dynamic-attach + * DMA bufs. TODO: The same should work for VRAM on + * large-BAR GPUs. + */ + attachment[i]->type = KFD_MEM_ATT_DMABUF; + ret = kfd_mem_attach_dmabuf(adev, mem, &bo[i]); + if (ret) + goto unwind; + } else { + /* FIXME: Need to DMA-map other BO types: + * large-BAR VRAM, doorbells, MMIO remap + */ + attachment[i]->type = KFD_MEM_ATT_SHARED; + bo[i] = mem->bo; + drm_gem_object_get(&bo[i]->tbo.base); + } - bo_va_entry->va = va; - bo_va_entry->pte_flags = get_pte_flags(adev, mem); - bo_va_entry->kgd_dev = (void *)adev; - list_add(&bo_va_entry->bo_list, list_bo_va); + /* Add BO to VM internal data structures */ + attachment[i]->bo_va = amdgpu_vm_bo_add(adev, vm, bo[i]); + if (unlikely(!attachment[i]->bo_va)) { + ret = -ENOMEM; + pr_err("Failed to add BO object to VM. ret == %d\n", + ret); + goto unwind; + } - if (p_bo_va_entry) - *p_bo_va_entry = bo_va_entry; + attachment[i]->va = va; + attachment[i]->pte_flags = get_pte_flags(adev, mem); + attachment[i]->adev = adev; + list_add(&attachment[i]->list, &mem->attachments); - /* Allocate validate page tables if needed */ - ret = vm_validate_pt_pd_bos(vm); - if (ret) { - pr_err("validate_pt_pd_bos() failed\n"); - goto err_alloc_pts; + va += bo_size; } return 0; -err_alloc_pts: - amdgpu_vm_bo_rmv(adev, bo_va_entry->bo_va); - list_del(&bo_va_entry->bo_list); -err_vmadd: - kfree(bo_va_entry); +unwind: + for (; i >= 0; i--) { + if (!attachment[i]) + continue; + if (attachment[i]->bo_va) { + amdgpu_vm_bo_rmv(adev, attachment[i]->bo_va); + list_del(&attachment[i]->list); + } + if (bo[i]) + drm_gem_object_put(&bo[i]->tbo.base); + kfree(attachment[i]); + } return ret; } -static void remove_bo_from_vm(struct amdgpu_device *adev, - struct kfd_bo_va_list *entry, unsigned long size) +static void kfd_mem_detach(struct kfd_mem_attachment *attachment) { - pr_debug("\t remove VA 0x%llx - 0x%llx in entry %p\n", - entry->va, - entry->va + size, entry); - amdgpu_vm_bo_rmv(adev, entry->bo_va); - list_del(&entry->bo_list); - kfree(entry); + struct amdgpu_bo *bo = attachment->bo_va->base.bo; + + pr_debug("\t remove VA 0x%llx in entry %p\n", + attachment->va, attachment); + amdgpu_vm_bo_rmv(attachment->adev, attachment->bo_va); + drm_gem_object_put(&bo->tbo.base); + list_del(&attachment->list); + kfree(attachment); } static void add_kgd_mem_to_kfd_bo_list(struct kgd_mem *mem, @@ -726,7 +961,7 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem, struct bo_vm_reservation_context *ctx) { struct amdgpu_bo *bo = mem->bo; - struct kfd_bo_va_list *entry; + struct kfd_mem_attachment *entry; unsigned int i; int ret; @@ -738,7 +973,7 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem, INIT_LIST_HEAD(&ctx->list); INIT_LIST_HEAD(&ctx->duplicates); - list_for_each_entry(entry, &mem->bo_va_list, bo_list) { + list_for_each_entry(entry, &mem->attachments, list) { if ((vm && vm != entry->bo_va->base.vm) || (entry->is_mapped != map_type && map_type != BO_VM_ALL)) @@ -760,7 +995,7 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem, list_add(&ctx->kfd_bo.tv.head, &ctx->list); i = 0; - list_for_each_entry(entry, &mem->bo_va_list, bo_list) { + list_for_each_entry(entry, &mem->attachments, list) { if ((vm && vm != entry->bo_va->base.vm) || (entry->is_mapped != map_type && map_type != BO_VM_ALL)) @@ -814,11 +1049,12 @@ static int unreserve_bo_and_vms(struct bo_vm_reservation_context *ctx, return ret; } -static int unmap_bo_from_gpuvm(struct amdgpu_device *adev, - struct kfd_bo_va_list *entry, +static void unmap_bo_from_gpuvm(struct kgd_mem *mem, + struct kfd_mem_attachment *entry, struct amdgpu_sync *sync) { struct amdgpu_bo_va *bo_va = entry->bo_va; + struct amdgpu_device *adev = entry->adev; struct amdgpu_vm *vm = bo_va->base.vm; amdgpu_vm_bo_unmap(adev, bo_va, entry->va); @@ -827,15 +1063,20 @@ static int unmap_bo_from_gpuvm(struct amdgpu_device *adev, amdgpu_sync_fence(sync, bo_va->last_pt_update); - return 0; + kfd_mem_dmaunmap_attachment(mem, entry); } -static int update_gpuvm_pte(struct amdgpu_device *adev, - struct kfd_bo_va_list *entry, - struct amdgpu_sync *sync) +static int update_gpuvm_pte(struct kgd_mem *mem, + struct kfd_mem_attachment *entry, + struct amdgpu_sync *sync) { - int ret; struct amdgpu_bo_va *bo_va = entry->bo_va; + struct amdgpu_device *adev = entry->adev; + int ret; + + ret = kfd_mem_dmamap_attachment(mem, entry); + if (ret) + return ret; /* Update the page tables */ ret = amdgpu_vm_bo_update(adev, bo_va, false); @@ -847,14 +1088,15 @@ static int update_gpuvm_pte(struct amdgpu_device *adev, return amdgpu_sync_fence(sync, bo_va->last_pt_update); } -static int map_bo_to_gpuvm(struct amdgpu_device *adev, - struct kfd_bo_va_list *entry, struct amdgpu_sync *sync, - bool no_update_pte) +static int map_bo_to_gpuvm(struct kgd_mem *mem, + struct kfd_mem_attachment *entry, + struct amdgpu_sync *sync, + bool no_update_pte) { int ret; /* Set virtual address for the allocation */ - ret = amdgpu_vm_bo_map(adev, entry->bo_va, entry->va, 0, + ret = amdgpu_vm_bo_map(entry->adev, entry->bo_va, entry->va, 0, amdgpu_bo_size(entry->bo_va->base.bo), entry->pte_flags); if (ret) { @@ -866,7 +1108,7 @@ static int map_bo_to_gpuvm(struct amdgpu_device *adev, if (no_update_pte) return 0; - ret = update_gpuvm_pte(adev, entry, sync); + ret = update_gpuvm_pte(mem, entry, sync); if (ret) { pr_err("update_gpuvm_pte() failed\n"); goto update_gpuvm_pte_failed; @@ -875,7 +1117,7 @@ static int map_bo_to_gpuvm(struct amdgpu_device *adev, return 0; update_gpuvm_pte_failed: - unmap_bo_from_gpuvm(adev, entry, sync); + unmap_bo_from_gpuvm(mem, entry, sync); return ret; } @@ -967,7 +1209,8 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, info->eviction_fence = amdgpu_amdkfd_fence_create(dma_fence_context_alloc(1), - current->mm); + current->mm, + NULL); if (!info->eviction_fence) { pr_err("Failed to create eviction fence\n"); ret = -ENOMEM; @@ -1036,15 +1279,19 @@ create_evict_fence_fail: int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, struct file *filp, u32 pasid, - void **vm, void **process_info, + void **process_info, struct dma_fence **ef) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct drm_file *drm_priv = filp->private_data; - struct amdgpu_fpriv *drv_priv = drm_priv->driver_priv; - struct amdgpu_vm *avm = &drv_priv->vm; + struct amdgpu_fpriv *drv_priv; + struct amdgpu_vm *avm; int ret; + ret = amdgpu_file_to_fpriv(filp, &drv_priv); + if (ret) + return ret; + avm = &drv_priv->vm; + /* Already a compute VM? */ if (avm->process_info) return -EINVAL; @@ -1059,7 +1306,7 @@ int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, if (ret) return ret; - *vm = (void *)avm; + amdgpu_vm_set_task_info(avm); return 0; } @@ -1100,15 +1347,17 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, } } -void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm) +void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *drm_priv) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + struct amdgpu_vm *avm; - if (WARN_ON(!kgd || !vm)) + if (WARN_ON(!kgd || !drm_priv)) return; - pr_debug("Releasing process vm %p\n", vm); + avm = drm_priv_to_vm(drm_priv); + + pr_debug("Releasing process vm %p\n", avm); /* The original pasid of amdgpu vm has already been * released during making a amdgpu vm to a compute vm @@ -1119,9 +1368,9 @@ void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm) amdgpu_vm_release_compute(adev, avm); } -uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm) +uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv) { - struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); struct amdgpu_bo *pd = avm->root.base.bo; struct amdgpu_device *adev = amdgpu_ttm_adev(pd->tbo.bdev); @@ -1132,11 +1381,11 @@ uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm) int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( struct kgd_dev *kgd, uint64_t va, uint64_t size, - void *vm, struct kgd_mem **mem, + void *drm_priv, struct kgd_mem **mem, uint64_t *offset, uint32_t flags) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); enum ttm_bo_type bo_type = ttm_bo_type_device; struct sg_table *sg = NULL; uint64_t user_addr = 0; @@ -1185,7 +1434,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( ret = -ENOMEM; goto err; } - INIT_LIST_HEAD(&(*mem)->bo_va_list); + INIT_LIST_HEAD(&(*mem)->attachments); mutex_init(&(*mem)->lock); (*mem)->aql_queue = !!(flags & KFD_IOC_ALLOC_MEM_FLAGS_AQL_QUEUE_MEM); @@ -1216,6 +1465,11 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( domain_string(alloc_domain), ret); goto err_bo_create; } + ret = drm_vma_node_allow(&gobj->vma_node, drm_priv); + if (ret) { + pr_debug("Failed to allow vma node access. ret %d\n", ret); + goto err_node_allow; + } bo = gem_to_amdgpu_bo(gobj); if (bo_type == ttm_bo_type_sg) { bo->tbo.sg = sg; @@ -1224,7 +1478,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( bo->kfd_bo = *mem; (*mem)->bo = bo; if (user_addr) - bo->flags |= AMDGPU_AMDKFD_USERPTR_BO; + bo->flags |= AMDGPU_AMDKFD_CREATE_USERPTR_BO; (*mem)->va = va; (*mem)->domain = domain; @@ -1245,6 +1499,8 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( allocate_init_user_pages_failed: remove_kgd_mem_from_kfd_bo_list(*mem, avm->process_info); + drm_vma_node_revoke(&gobj->vma_node, drm_priv); +err_node_allow: amdgpu_bo_unref(&bo); /* Don't unreserve system mem limit twice */ goto err_reserve_limit; @@ -1262,11 +1518,12 @@ err: } int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, uint64_t *size) + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv, + uint64_t *size) { struct amdkfd_process_info *process_info = mem->process_info; unsigned long bo_size = mem->bo->tbo.base.size; - struct kfd_bo_va_list *entry, *tmp; + struct kfd_mem_attachment *entry, *tmp; struct bo_vm_reservation_context ctx; struct ttm_validate_buffer *bo_list_entry; unsigned int mapped_to_gpu_memory; @@ -1309,13 +1566,12 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( pr_debug("Release VA 0x%llx - 0x%llx\n", mem->va, mem->va + bo_size * (1 + mem->aql_queue)); - /* Remove from VM internal data structures */ - list_for_each_entry_safe(entry, tmp, &mem->bo_va_list, bo_list) - remove_bo_from_vm((struct amdgpu_device *)entry->kgd_dev, - entry, bo_size); - ret = unreserve_bo_and_vms(&ctx, false, false); + /* Remove from VM internal data structures */ + list_for_each_entry_safe(entry, tmp, &mem->attachments, list) + kfd_mem_detach(entry); + /* Free the sync object */ amdgpu_sync_free(&mem->sync); @@ -1339,6 +1595,9 @@ 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) + dma_buf_put(mem->dmabuf); drm_gem_object_put(&mem->bo->tbo.base); mutex_destroy(&mem->lock); kfree(mem); @@ -1347,17 +1606,15 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( } int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *vm) + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); int ret; struct amdgpu_bo *bo; uint32_t domain; - struct kfd_bo_va_list *entry; + struct kfd_mem_attachment *entry; struct bo_vm_reservation_context ctx; - struct kfd_bo_va_list *bo_va_entry = NULL; - struct kfd_bo_va_list *bo_va_entry_aql = NULL; unsigned long bo_size; bool is_invalid_userptr = false; @@ -1391,9 +1648,15 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( pr_debug("Map VA 0x%llx - 0x%llx to vm %p domain %s\n", mem->va, mem->va + bo_size * (1 + mem->aql_queue), - vm, domain_string(domain)); + avm, domain_string(domain)); - ret = reserve_bo_and_vm(mem, vm, &ctx); + if (!kfd_mem_is_attached(avm, mem)) { + ret = kfd_mem_attach(adev, mem, avm, mem->aql_queue); + if (ret) + goto out; + } + + ret = reserve_bo_and_vm(mem, avm, &ctx); if (unlikely(ret)) goto out; @@ -1406,22 +1669,9 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( bo->tbo.mem.mem_type == TTM_PL_SYSTEM) is_invalid_userptr = true; - if (check_if_add_bo_to_vm(avm, mem)) { - ret = add_bo_to_vm(adev, mem, avm, false, - &bo_va_entry); - if (ret) - goto add_bo_to_vm_failed; - if (mem->aql_queue) { - ret = add_bo_to_vm(adev, mem, avm, - true, &bo_va_entry_aql); - if (ret) - goto add_bo_to_vm_failed_aql; - } - } else { - ret = vm_validate_pt_pd_bos(avm); - if (unlikely(ret)) - goto add_bo_to_vm_failed; - } + ret = vm_validate_pt_pd_bos(avm); + if (unlikely(ret)) + goto out_unreserve; if (mem->mapped_to_gpu_memory == 0 && !amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) { @@ -1432,34 +1682,34 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( ret = amdgpu_amdkfd_bo_validate(bo, domain, true); if (ret) { pr_debug("Validate failed\n"); - goto map_bo_to_gpuvm_failed; + goto out_unreserve; } } - list_for_each_entry(entry, &mem->bo_va_list, bo_list) { - if (entry->bo_va->base.vm == vm && !entry->is_mapped) { - pr_debug("\t map VA 0x%llx - 0x%llx in entry %p\n", - entry->va, entry->va + bo_size, - entry); + list_for_each_entry(entry, &mem->attachments, list) { + if (entry->bo_va->base.vm != avm || entry->is_mapped) + continue; - ret = map_bo_to_gpuvm(adev, entry, ctx.sync, - is_invalid_userptr); - if (ret) { - pr_err("Failed to map bo to gpuvm\n"); - goto map_bo_to_gpuvm_failed; - } + pr_debug("\t map VA 0x%llx - 0x%llx in entry %p\n", + entry->va, entry->va + bo_size, entry); - ret = vm_update_pds(vm, ctx.sync); - if (ret) { - pr_err("Failed to update page directories\n"); - goto map_bo_to_gpuvm_failed; - } + ret = map_bo_to_gpuvm(mem, entry, ctx.sync, + is_invalid_userptr); + if (ret) { + pr_err("Failed to map bo to gpuvm\n"); + goto out_unreserve; + } - entry->is_mapped = true; - mem->mapped_to_gpu_memory++; - pr_debug("\t INC mapping count %d\n", - mem->mapped_to_gpu_memory); + ret = vm_update_pds(avm, ctx.sync); + if (ret) { + pr_err("Failed to update page directories\n"); + goto out_unreserve; } + + entry->is_mapped = true; + mem->mapped_to_gpu_memory++; + pr_debug("\t INC mapping count %d\n", + mem->mapped_to_gpu_memory); } if (!amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) && !bo->tbo.pin_count) @@ -1470,13 +1720,7 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( goto out; -map_bo_to_gpuvm_failed: - if (bo_va_entry_aql) - remove_bo_from_vm(adev, bo_va_entry_aql, bo_size); -add_bo_to_vm_failed_aql: - if (bo_va_entry) - remove_bo_from_vm(adev, bo_va_entry, bo_size); -add_bo_to_vm_failed: +out_unreserve: unreserve_bo_and_vms(&ctx, false, false); out: mutex_unlock(&mem->process_info->lock); @@ -1485,19 +1729,18 @@ out: } int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *vm) + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv) { - struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct amdkfd_process_info *process_info = - ((struct amdgpu_vm *)vm)->process_info; + struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); + struct amdkfd_process_info *process_info = avm->process_info; unsigned long bo_size = mem->bo->tbo.base.size; - struct kfd_bo_va_list *entry; + struct kfd_mem_attachment *entry; struct bo_vm_reservation_context ctx; int ret; mutex_lock(&mem->lock); - ret = reserve_bo_and_cond_vms(mem, vm, BO_VM_MAPPED, &ctx); + ret = reserve_bo_and_cond_vms(mem, avm, BO_VM_MAPPED, &ctx); if (unlikely(ret)) goto out; /* If no VMs were reserved, it means the BO wasn't actually mapped */ @@ -1506,35 +1749,28 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( goto unreserve_out; } - ret = vm_validate_pt_pd_bos((struct amdgpu_vm *)vm); + ret = vm_validate_pt_pd_bos(avm); if (unlikely(ret)) goto unreserve_out; pr_debug("Unmap VA 0x%llx - 0x%llx from vm %p\n", mem->va, mem->va + bo_size * (1 + mem->aql_queue), - vm); - - list_for_each_entry(entry, &mem->bo_va_list, bo_list) { - if (entry->bo_va->base.vm == vm && entry->is_mapped) { - pr_debug("\t unmap VA 0x%llx - 0x%llx from entry %p\n", - entry->va, - entry->va + bo_size, - entry); - - ret = unmap_bo_from_gpuvm(adev, entry, ctx.sync); - if (ret == 0) { - entry->is_mapped = false; - } else { - pr_err("failed to unmap VA 0x%llx\n", - mem->va); - goto unreserve_out; - } + avm); - mem->mapped_to_gpu_memory--; - pr_debug("\t DEC mapping count %d\n", - mem->mapped_to_gpu_memory); - } + list_for_each_entry(entry, &mem->attachments, list) { + if (entry->bo_va->base.vm != avm || !entry->is_mapped) + continue; + + pr_debug("\t unmap VA 0x%llx - 0x%llx from entry %p\n", + entry->va, entry->va + bo_size, entry); + + unmap_bo_from_gpuvm(mem, entry, ctx.sync); + entry->is_mapped = false; + + mem->mapped_to_gpu_memory--; + pr_debug("\t DEC mapping count %d\n", + mem->mapped_to_gpu_memory); } /* If BO is unmapped from all VMs, unfence it. It can be evicted if @@ -1642,14 +1878,15 @@ int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd, struct dma_buf *dma_buf, - uint64_t va, void *vm, + uint64_t va, void *drm_priv, struct kgd_mem **mem, uint64_t *size, uint64_t *mmap_offset) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); struct drm_gem_object *obj; struct amdgpu_bo *bo; - struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + int ret; if (dma_buf->ops != &amdgpu_dmabuf_ops) /* Can't handle non-graphics buffers */ @@ -1670,13 +1907,19 @@ int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd, if (!*mem) return -ENOMEM; + ret = drm_vma_node_allow(&obj->vma_node, drm_priv); + if (ret) { + kfree(mem); + return ret; + } + if (size) *size = amdgpu_bo_size(bo); if (mmap_offset) *mmap_offset = amdgpu_bo_mmap_offset(bo); - INIT_LIST_HEAD(&(*mem)->bo_va_list); + INIT_LIST_HEAD(&(*mem)->attachments); mutex_init(&(*mem)->lock); (*mem)->alloc_flags = @@ -1873,7 +2116,7 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) list_for_each_entry_safe(mem, tmp_mem, &process_info->userptr_inval_list, validate_list.head) { - struct kfd_bo_va_list *bo_va_entry; + struct kfd_mem_attachment *attachment; bo = mem->bo; @@ -1896,13 +2139,12 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) * VM faults if the GPU tries to access the invalid * memory. */ - list_for_each_entry(bo_va_entry, &mem->bo_va_list, bo_list) { - if (!bo_va_entry->is_mapped) + list_for_each_entry(attachment, &mem->attachments, list) { + if (!attachment->is_mapped) continue; - ret = update_gpuvm_pte((struct amdgpu_device *) - bo_va_entry->kgd_dev, - bo_va_entry, &sync); + kfd_mem_dmaunmap_attachment(mem, attachment); + ret = update_gpuvm_pte(mem, attachment, &sync); if (ret) { pr_err("%s: update PTE failed\n", __func__); /* make sure this gets validated again */ @@ -2083,7 +2325,7 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) struct amdgpu_bo *bo = mem->bo; uint32_t domain = mem->domain; - struct kfd_bo_va_list *bo_va_entry; + struct kfd_mem_attachment *attachment; total_size += amdgpu_bo_size(bo); @@ -2103,12 +2345,12 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) pr_debug("Memory eviction: Sync BO fence failed. Try again\n"); goto validate_map_fail; } - list_for_each_entry(bo_va_entry, &mem->bo_va_list, - bo_list) { - ret = update_gpuvm_pte((struct amdgpu_device *) - bo_va_entry->kgd_dev, - bo_va_entry, - &sync_obj); + list_for_each_entry(attachment, &mem->attachments, list) { + if (!attachment->is_mapped) + continue; + + kfd_mem_dmaunmap_attachment(mem, attachment); + ret = update_gpuvm_pte(mem, attachment, &sync_obj); if (ret) { pr_debug("Memory eviction: update PTE failed. Try again\n"); goto validate_map_fail; @@ -2135,7 +2377,8 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) */ new_fence = amdgpu_amdkfd_fence_create( process_info->eviction_fence->base.context, - process_info->eviction_fence->mm); + process_info->eviction_fence->mm, + NULL); if (!new_fence) { pr_err("Failed to create eviction fence\n"); ret = -ENOMEM; @@ -2182,7 +2425,7 @@ int amdgpu_amdkfd_add_gws_to_process(void *info, void *gws, struct kgd_mem **mem return -ENOMEM; mutex_init(&(*mem)->lock); - INIT_LIST_HEAD(&(*mem)->bo_va_list); + INIT_LIST_HEAD(&(*mem)->attachments); (*mem)->bo = amdgpu_bo_ref(gws_bo); (*mem)->domain = AMDGPU_GEM_DOMAIN_GWS; (*mem)->process_info = process_info; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c index 494b2e1717d5..18bd1b49ced6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c @@ -1828,6 +1828,9 @@ int amdgpu_atombios_init(struct amdgpu_device *adev) if (adev->is_atom_fw) { amdgpu_atomfirmware_scratch_regs_init(adev); amdgpu_atomfirmware_allocate_fb_scratch(adev); + /* cached firmware_flags for further usage */ + adev->mode_info.firmware_flags = + amdgpu_atomfirmware_query_firmware_capability(adev); } else { amdgpu_atombios_scratch_regs_init(adev); amdgpu_atombios_allocate_fb_scratch(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c index 60716b35444b..3b5d13189073 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c @@ -29,23 +29,59 @@ #include "atombios.h" #include "soc15_hw_ip.h" -bool amdgpu_atomfirmware_gpu_supports_virtualization(struct amdgpu_device *adev) +union firmware_info { + struct atom_firmware_info_v3_1 v31; + struct atom_firmware_info_v3_2 v32; + struct atom_firmware_info_v3_3 v33; + struct atom_firmware_info_v3_4 v34; +}; + +/* + * Helper function to query firmware capability + * + * @adev: amdgpu_device pointer + * + * Return firmware_capability in firmwareinfo table on success or 0 if not + */ +uint32_t amdgpu_atomfirmware_query_firmware_capability(struct amdgpu_device *adev) { - int index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, - firmwareinfo); - uint16_t data_offset; + struct amdgpu_mode_info *mode_info = &adev->mode_info; + int index; + u16 data_offset, size; + union firmware_info *firmware_info; + u8 frev, crev; + u32 fw_cap = 0; - if (amdgpu_atom_parse_data_header(adev->mode_info.atom_context, index, NULL, - NULL, NULL, &data_offset)) { - struct atom_firmware_info_v3_1 *firmware_info = - (struct atom_firmware_info_v3_1 *)(adev->mode_info.atom_context->bios + - data_offset); + index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, + firmwareinfo); - if (le32_to_cpu(firmware_info->firmware_capability) & - ATOM_FIRMWARE_CAP_GPU_VIRTUALIZATION) - return true; + if (amdgpu_atom_parse_data_header(adev->mode_info.atom_context, + index, &size, &frev, &crev, &data_offset)) { + /* support firmware_info 3.1 + */ + if ((frev == 3 && crev >=1) || (frev > 3)) { + firmware_info = (union firmware_info *) + (mode_info->atom_context->bios + data_offset); + fw_cap = le32_to_cpu(firmware_info->v31.firmware_capability); + } } - return false; + + return fw_cap; +} + +/* + * Helper function to query gpu virtualizaiton capability + * + * @adev: amdgpu_device pointer + * + * Return true if gpu virtualization is supported or false if not + */ +bool amdgpu_atomfirmware_gpu_virtualization_supported(struct amdgpu_device *adev) +{ + u32 fw_cap; + + fw_cap = adev->mode_info.firmware_flags; + + return (fw_cap & ATOM_FIRMWARE_CAP_GPU_VIRTUALIZATION) ? true : false; } void amdgpu_atomfirmware_scratch_regs_init(struct amdgpu_device *adev) @@ -400,41 +436,36 @@ bool amdgpu_atomfirmware_mem_ecc_supported(struct amdgpu_device *adev) return ecc_default_enabled; } -union firmware_info { - struct atom_firmware_info_v3_1 v31; - struct atom_firmware_info_v3_2 v32; - struct atom_firmware_info_v3_3 v33; - struct atom_firmware_info_v3_4 v34; -}; - /* + * Helper function to query sram ecc capablity + * + * @adev: amdgpu_device pointer + * * Return true if vbios supports sram ecc or false if not */ bool amdgpu_atomfirmware_sram_ecc_supported(struct amdgpu_device *adev) { - struct amdgpu_mode_info *mode_info = &adev->mode_info; - int index; - u16 data_offset, size; - union firmware_info *firmware_info; - u8 frev, crev; - bool sram_ecc_supported = false; + u32 fw_cap; - index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, - firmwareinfo); + fw_cap = adev->mode_info.firmware_flags; - if (amdgpu_atom_parse_data_header(adev->mode_info.atom_context, - index, &size, &frev, &crev, &data_offset)) { - /* support firmware_info 3.1 + */ - if ((frev == 3 && crev >=1) || (frev > 3)) { - firmware_info = (union firmware_info *) - (mode_info->atom_context->bios + data_offset); - sram_ecc_supported = - (le32_to_cpu(firmware_info->v31.firmware_capability) & - ATOM_FIRMWARE_CAP_SRAM_ECC) ? true : false; - } - } + return (fw_cap & ATOM_FIRMWARE_CAP_SRAM_ECC) ? true : false; +} - return sram_ecc_supported; +/* + * Helper function to query dynamic boot config capability + * + * @adev: amdgpu_device pointer + * + * Return true if vbios supports dynamic boot config or false if not + */ +bool amdgpu_atomfirmware_dynamic_boot_config_supported(struct amdgpu_device *adev) +{ + u32 fw_cap; + + fw_cap = adev->mode_info.firmware_flags; + + return (fw_cap & ATOM_FIRMWARE_CAP_DYNAMIC_BOOT_CFG_ENABLE) ? true : false; } union smu_info { @@ -466,10 +497,6 @@ int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev) adev->pm.current_sclk = adev->clock.default_sclk; adev->pm.current_mclk = adev->clock.default_mclk; - /* not technically a clock, but... */ - adev->mode_info.firmware_flags = - le32_to_cpu(firmware_info->v31.firmware_capability); - ret = 0; } @@ -519,6 +546,21 @@ int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev) ret = 0; } + /* if asic is Navi+, the rlc reference clock is used for system clock + * from vbios gfx_info table */ + if (adev->asic_type >= CHIP_NAVI10) { + index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, + gfx_info); + if (amdgpu_atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) { + struct atom_gfx_info_v2_2 *gfx_info = (struct atom_gfx_info_v2_2*) + (mode_info->atom_context->bios + data_offset); + if ((frev == 2) && (crev >= 2)) + spll->reference_freq = le32_to_cpu(gfx_info->rlc_gpu_timer_refclk); + ret = 0; + } + } + return ret; } @@ -584,67 +626,19 @@ int amdgpu_atomfirmware_get_gfx_info(struct amdgpu_device *adev) } /* - * Check if VBIOS supports GDDR6 training data save/restore + * Helper function to query two stage mem training capability + * + * @adev: amdgpu_device pointer + * + * Return true if two stage mem training is supported or false if not */ -static bool gddr6_mem_train_vbios_support(struct amdgpu_device *adev) +bool amdgpu_atomfirmware_mem_training_supported(struct amdgpu_device *adev) { - uint16_t data_offset; - int index; - - index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, - firmwareinfo); - if (amdgpu_atom_parse_data_header(adev->mode_info.atom_context, index, NULL, - NULL, NULL, &data_offset)) { - struct atom_firmware_info_v3_1 *firmware_info = - (struct atom_firmware_info_v3_1 *)(adev->mode_info.atom_context->bios + - data_offset); - - DRM_DEBUG("atom firmware capability:0x%08x.\n", - le32_to_cpu(firmware_info->firmware_capability)); + u32 fw_cap; - if (le32_to_cpu(firmware_info->firmware_capability) & - ATOM_FIRMWARE_CAP_ENABLE_2STAGE_BIST_TRAINING) - return true; - } + fw_cap = adev->mode_info.firmware_flags; - return false; -} - -int amdgpu_mem_train_support(struct amdgpu_device *adev) -{ - int ret; - uint32_t major, minor, revision, hw_v; - - if (gddr6_mem_train_vbios_support(adev)) { - amdgpu_discovery_get_ip_version(adev, MP0_HWID, &major, &minor, &revision); - hw_v = HW_REV(major, minor, revision); - /* - * treat 0 revision as a special case since register for MP0 and MMHUB is missing - * for some Navi10 A0, preventing driver from discovering the hwip information since - * none of the functions will be initialized, it should not cause any problems - */ - switch (hw_v) { - case HW_REV(11, 0, 0): - case HW_REV(11, 0, 5): - case HW_REV(11, 0, 7): - case HW_REV(11, 0, 11): - case HW_REV(11, 0, 12): - ret = 1; - break; - default: - DRM_ERROR("memory training vbios supports but psp hw(%08x)" - " doesn't support!\n", hw_v); - ret = -1; - break; - } - } else { - ret = 0; - hw_v = -1; - } - - - DRM_DEBUG("mp0 hw_v %08x, ret:%d.\n", hw_v, ret); - return ret; + return (fw_cap & ATOM_FIRMWARE_CAP_ENABLE_2STAGE_BIST_TRAINING) ? true : false; } int amdgpu_atomfirmware_get_fw_reserved_fb_size(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h index 9f0d4356e8df..1bbbb195015d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h @@ -26,7 +26,8 @@ #define get_index_into_master_table(master_table, table_name) (offsetof(struct master_table, table_name) / sizeof(uint16_t)) -bool amdgpu_atomfirmware_gpu_supports_virtualization(struct amdgpu_device *adev); +uint32_t amdgpu_atomfirmware_query_firmware_capability(struct amdgpu_device *adev); +bool amdgpu_atomfirmware_gpu_virtualization_supported(struct amdgpu_device *adev); void amdgpu_atomfirmware_scratch_regs_init(struct amdgpu_device *adev); int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev); int amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev, @@ -35,7 +36,8 @@ int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev); int amdgpu_atomfirmware_get_gfx_info(struct amdgpu_device *adev); bool amdgpu_atomfirmware_mem_ecc_supported(struct amdgpu_device *adev); bool amdgpu_atomfirmware_sram_ecc_supported(struct amdgpu_device *adev); +bool amdgpu_atomfirmware_mem_training_supported(struct amdgpu_device *adev); +bool amdgpu_atomfirmware_dynamic_boot_config_supported(struct amdgpu_device *adev); int amdgpu_atomfirmware_get_fw_reserved_fb_size(struct amdgpu_device *adev); -int amdgpu_mem_train_support(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index b5c766998045..90136f9dedd6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -672,7 +672,7 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p) } /** - * cs_parser_fini() - clean parser states + * amdgpu_cs_parser_fini() - clean parser states * @parser: parser structure holding parsing context. * @error: error number * @backoff: indicator to backoff the reservation @@ -1488,7 +1488,7 @@ int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data, } /** - * amdgpu_cs_wait_all_fence - wait on all fences to signal + * amdgpu_cs_wait_all_fences - wait on all fences to signal * * @adev: amdgpu device * @filp: file private @@ -1639,7 +1639,7 @@ err_free_fences: } /** - * amdgpu_cs_find_bo_va - find bo_va for VM address + * amdgpu_cs_find_mapping - find bo_va for VM address * * @parser: command submission parser context * @addr: VM address diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index 0350205c4897..fc83445fbc40 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -651,3 +651,81 @@ void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr) idr_destroy(&mgr->ctx_handles); mutex_destroy(&mgr->lock); } + +static void amdgpu_ctx_fence_time(struct amdgpu_ctx *ctx, + struct amdgpu_ctx_entity *centity, ktime_t *total, ktime_t *max) +{ + ktime_t now, t1; + uint32_t i; + + *total = *max = 0; + + now = ktime_get(); + for (i = 0; i < amdgpu_sched_jobs; i++) { + struct dma_fence *fence; + struct drm_sched_fence *s_fence; + + spin_lock(&ctx->ring_lock); + fence = dma_fence_get(centity->fences[i]); + spin_unlock(&ctx->ring_lock); + if (!fence) + continue; + s_fence = to_drm_sched_fence(fence); + if (!dma_fence_is_signaled(&s_fence->scheduled)) { + dma_fence_put(fence); + continue; + } + t1 = s_fence->scheduled.timestamp; + if (!ktime_before(t1, now)) { + dma_fence_put(fence); + continue; + } + if (dma_fence_is_signaled(&s_fence->finished) && + s_fence->finished.timestamp < now) + *total += ktime_sub(s_fence->finished.timestamp, t1); + else + *total += ktime_sub(now, t1); + t1 = ktime_sub(now, t1); + dma_fence_put(fence); + *max = max(t1, *max); + } +} + +ktime_t amdgpu_ctx_mgr_fence_usage(struct amdgpu_ctx_mgr *mgr, uint32_t hwip, + uint32_t idx, uint64_t *elapsed) +{ + struct idr *idp; + struct amdgpu_ctx *ctx; + uint32_t id; + struct amdgpu_ctx_entity *centity; + ktime_t total = 0, max = 0; + + if (idx >= AMDGPU_MAX_ENTITY_NUM) + return 0; + idp = &mgr->ctx_handles; + mutex_lock(&mgr->lock); + idr_for_each_entry(idp, ctx, id) { + ktime_t ttotal, tmax; + + if (!ctx->entities[hwip][idx]) + continue; + + centity = ctx->entities[hwip][idx]; + amdgpu_ctx_fence_time(ctx, centity, &ttotal, &tmax); + + /* Harmonic mean approximation diverges for very small + * values. If ratio < 0.01% ignore + */ + if (AMDGPU_CTX_FENCE_USAGE_MIN_RATIO(tmax, ttotal)) + continue; + + total = ktime_add(total, ttotal); + max = ktime_after(tmax, max) ? tmax : max; + } + + mutex_unlock(&mgr->lock); + if (elapsed) + *elapsed = max; + + return total; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h index f54e10314661..14db16bc3322 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h @@ -30,6 +30,7 @@ struct drm_file; struct amdgpu_fpriv; #define AMDGPU_MAX_ENTITY_NUM 4 +#define AMDGPU_CTX_FENCE_USAGE_MIN_RATIO(max, total) ((max) > 16384ULL*(total)) struct amdgpu_ctx_entity { uint64_t sequence; @@ -87,5 +88,6 @@ void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr); void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr); long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout); void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr); - +ktime_t amdgpu_ctx_mgr_fence_usage(struct amdgpu_ctx_mgr *mgr, uint32_t hwip, + uint32_t idx, uint64_t *elapsed); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 7d3b54615147..b1eb005fe83e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -119,6 +119,7 @@ const char *amdgpu_asic_name[] = { "NAVY_FLOUNDER", "VANGOGH", "DIMGREY_CAVEFISH", + "BEIGE_GOBY", "LAST", }; @@ -1683,6 +1684,19 @@ int amdgpu_device_ip_block_add(struct amdgpu_device *adev, if (!ip_block_version) return -EINVAL; + switch (ip_block_version->type) { + case AMD_IP_BLOCK_TYPE_VCN: + if (adev->harvest_ip_mask & AMD_HARVEST_IP_VCN_MASK) + return 0; + break; + case AMD_IP_BLOCK_TYPE_JPEG: + if (adev->harvest_ip_mask & AMD_HARVEST_IP_JPEG_MASK) + return 0; + break; + default: + break; + } + DRM_INFO("add ip block number %d <%s>\n", adev->num_ip_blocks, ip_block_version->funcs->name); @@ -1807,6 +1821,7 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev) case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: default: return 0; case CHIP_VEGA10: @@ -2020,6 +2035,7 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: case CHIP_VANGOGH: if (adev->asic_type == CHIP_VANGOGH) adev->family = AMDGPU_FAMILY_VGH; @@ -2843,7 +2859,7 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev) AMD_IP_BLOCK_TYPE_IH, }; - for (i = 0; i < ARRAY_SIZE(ip_order); i++) { + for (i = 0; i < adev->num_ip_blocks; i++) { int j; struct amdgpu_ip_block *block; @@ -3021,7 +3037,7 @@ static void amdgpu_device_detect_sriov_bios(struct amdgpu_device *adev) { if (amdgpu_sriov_vf(adev)) { if (adev->is_atom_fw) { - if (amdgpu_atomfirmware_gpu_supports_virtualization(adev)) + if (amdgpu_atomfirmware_gpu_virtualization_supported(adev)) adev->virt.caps |= AMDGPU_SRIOV_CAPS_SRIOV_VBIOS; } else { if (amdgpu_atombios_has_gpu_virtualization_table(adev)) @@ -3084,6 +3100,7 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type) case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: case CHIP_VANGOGH: #endif return amdgpu_dc != 0; @@ -3111,7 +3128,6 @@ bool amdgpu_device_has_dc_support(struct amdgpu_device *adev) return amdgpu_device_asic_has_dc_support(adev->asic_type); } - static void amdgpu_device_xgmi_reset_func(struct work_struct *__work) { struct amdgpu_device *adev = @@ -3167,8 +3183,8 @@ static int amdgpu_device_get_job_timeout_settings(struct amdgpu_device *adev) int ret = 0; /* - * By default timeout for non compute jobs is 10000. - * And there is no timeout enforced on compute jobs. + * By default timeout for non compute jobs is 10000 + * and 60000 for compute jobs. * In SR-IOV or passthrough mode, timeout for compute * jobs are 60000 by default. */ @@ -3177,10 +3193,8 @@ static int amdgpu_device_get_job_timeout_settings(struct amdgpu_device *adev) if (amdgpu_sriov_vf(adev)) adev->compute_timeout = amdgpu_sriov_is_pp_one_vf(adev) ? msecs_to_jiffies(60000) : msecs_to_jiffies(10000); - else if (amdgpu_passthrough(adev)) - adev->compute_timeout = msecs_to_jiffies(60000); else - adev->compute_timeout = MAX_SCHEDULE_TIMEOUT; + adev->compute_timeout = msecs_to_jiffies(60000); if (strnlen(input, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) { while ((timeout_setting = strsep(&input, ",")) && @@ -3276,6 +3290,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->vm_manager.vm_pte_funcs = NULL; adev->vm_manager.vm_pte_num_scheds = 0; adev->gmc.gmc_funcs = NULL; + adev->harvest_ip_mask = 0x0; adev->fence_context = dma_fence_context_alloc(AMDGPU_MAX_RINGS); bitmap_zero(adev->gfx.pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); @@ -3728,7 +3743,6 @@ void amdgpu_device_fini(struct amdgpu_device *adev) int amdgpu_device_suspend(struct drm_device *dev, bool fbcon) { struct amdgpu_device *adev = drm_to_adev(dev); - int r; if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; @@ -3743,7 +3757,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool fbcon) amdgpu_ras_suspend(adev); - r = amdgpu_device_ip_suspend_phase1(adev); + amdgpu_device_ip_suspend_phase1(adev); if (!adev->in_s0ix) amdgpu_amdkfd_suspend(adev, adev->in_runpm); @@ -3753,7 +3767,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool fbcon) amdgpu_fence_driver_suspend(adev); - r = amdgpu_device_ip_suspend_phase2(adev); + amdgpu_device_ip_suspend_phase2(adev); /* evict remaining vram memory * This second call to evict vram is to evict the gart page table * using the CPU. @@ -4466,7 +4480,6 @@ out: r = amdgpu_ib_ring_tests(tmp_adev); if (r) { dev_err(tmp_adev->dev, "ib ring test failed (%d).\n", r); - r = amdgpu_device_ip_suspend(tmp_adev); need_full_reset = true; r = -EAGAIN; goto end; @@ -5111,7 +5124,8 @@ int amdgpu_device_baco_enter(struct drm_device *dev) if (!amdgpu_device_supports_baco(adev_to_drm(adev))) return -ENOTSUPP; - if (ras && ras->supported && adev->nbio.funcs->enable_doorbell_interrupt) + if (ras && adev->ras_enabled && + adev->nbio.funcs->enable_doorbell_interrupt) adev->nbio.funcs->enable_doorbell_interrupt(adev, false); return amdgpu_dpm_baco_enter(adev); @@ -5130,7 +5144,8 @@ int amdgpu_device_baco_exit(struct drm_device *dev) if (ret) return ret; - if (ras && ras->supported && adev->nbio.funcs->enable_doorbell_interrupt) + if (ras && adev->ras_enabled && + adev->nbio.funcs->enable_doorbell_interrupt) adev->nbio.funcs->enable_doorbell_interrupt(adev, true); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index b2dbcb4df020..e1b6f5891759 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -373,6 +373,34 @@ int amdgpu_discovery_get_ip_version(struct amdgpu_device *adev, int hw_id, return -EINVAL; } +void amdgpu_discovery_harvest_ip(struct amdgpu_device *adev) +{ + struct binary_header *bhdr; + struct harvest_table *harvest_info; + int i; + + bhdr = (struct binary_header *)adev->mman.discovery_bin; + harvest_info = (struct harvest_table *)(adev->mman.discovery_bin + + le16_to_cpu(bhdr->table_list[HARVEST_INFO].offset)); + + for (i = 0; i < 32; i++) { + if (le32_to_cpu(harvest_info->list[i].hw_id) == 0) + break; + + switch (le32_to_cpu(harvest_info->list[i].hw_id)) { + case VCN_HWID: + adev->harvest_ip_mask |= AMD_HARVEST_IP_VCN_MASK; + adev->harvest_ip_mask |= AMD_HARVEST_IP_JPEG_MASK; + break; + case DMU_HWID: + adev->harvest_ip_mask |= AMD_HARVEST_IP_DMU_MASK; + break; + default: + break; + } + } +} + int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev) { struct binary_header *bhdr; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h index 8f6183801cb3..1b1ae21b1037 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h @@ -29,6 +29,7 @@ void amdgpu_discovery_fini(struct amdgpu_device *adev); int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev); +void amdgpu_discovery_harvest_ip(struct amdgpu_device *adev); int amdgpu_discovery_get_ip_version(struct amdgpu_device *adev, int hw_id, int *major, int *minor, int *revision); int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index f93883db2b46..1ed9748b9bc7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -23,6 +23,7 @@ */ #include <drm/amdgpu_drm.h> +#include <drm/drm_aperture.h> #include <drm/drm_drv.h> #include <drm/drm_gem.h> #include <drm/drm_vblank.h> @@ -42,7 +43,7 @@ #include "amdgpu_irq.h" #include "amdgpu_dma_buf.h" #include "amdgpu_sched.h" - +#include "amdgpu_fdinfo.h" #include "amdgpu_amdkfd.h" #include "amdgpu_ras.h" @@ -287,9 +288,9 @@ module_param_named(msi, amdgpu_msi, int, 0444); * for SDMA and Video. * * By default(with no lockup_timeout settings), the timeout for all non-compute(GFX, SDMA and Video) - * jobs is 10000. And there is no timeout enforced on compute jobs. + * jobs is 10000. The timeout for compute is 60000. */ -MODULE_PARM_DESC(lockup_timeout, "GPU lockup timeout in ms (default: for bare metal 10000 for non-compute jobs and infinity timeout for compute jobs; " +MODULE_PARM_DESC(lockup_timeout, "GPU lockup timeout in ms (default: for bare metal 10000 for non-compute jobs and 60000 for compute jobs; " "for passthrough or sriov, 10000 for all jobs." " 0: keep default value. negative: infinity timeout), " "format: for bare metal [Non-Compute] or [GFX,Compute,SDMA,Video]; " @@ -640,7 +641,8 @@ module_param_named(mes, amdgpu_mes, int, 0444); /** * DOC: noretry (int) - * Disable retry faults in the GPU memory controller. + * Disable XNACK retry in the SQ by default on GFXv9 hardware. On ASICs that + * do not support per-process XNACK this also disables retry page faults. * (0 = retry enabled, 1 = retry disabled, -1 auto (default)) */ MODULE_PARM_DESC(noretry, @@ -1185,6 +1187,7 @@ static const struct pci_device_id pciidlist[] = { {0x1002, 0x7408, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, {0x1002, 0x740C, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, {0x1002, 0x740F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, + {0x1002, 0x7410, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, {0, 0, 0} }; @@ -1258,7 +1261,7 @@ static int amdgpu_pci_probe(struct pci_dev *pdev, #endif /* Get rid of things like offb */ - ret = drm_fb_helper_remove_conflicting_pci_framebuffers(pdev, "amdgpudrmfb"); + ret = drm_aperture_remove_conflicting_pci_framebuffers(pdev, "amdgpudrmfb"); if (ret) return ret; @@ -1597,17 +1600,15 @@ static int amdgpu_pmops_runtime_idle(struct device *dev) if (amdgpu_device_has_dc_support(adev)) { struct drm_crtc *crtc; - drm_modeset_lock_all(drm_dev); - drm_for_each_crtc(crtc, drm_dev) { - if (crtc->state->active) { + drm_modeset_lock(&crtc->mutex, NULL); + if (crtc->state->active) ret = -EBUSY; + drm_modeset_unlock(&crtc->mutex); + if (ret < 0) break; - } } - drm_modeset_unlock_all(drm_dev); - } else { struct drm_connector *list_connector; struct drm_connector_list_iter iter; @@ -1694,6 +1695,9 @@ static const struct file_operations amdgpu_driver_kms_fops = { #ifdef CONFIG_COMPAT .compat_ioctl = amdgpu_kms_compat_ioctl, #endif +#ifdef CONFIG_PROC_FS + .show_fdinfo = amdgpu_show_fdinfo +#endif }; int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c index 4f10c4529840..09b048647523 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c @@ -288,10 +288,13 @@ out: static int amdgpu_fbdev_destroy(struct drm_device *dev, struct amdgpu_fbdev *rfbdev) { struct amdgpu_framebuffer *rfb = &rfbdev->rfb; + int i; drm_fb_helper_unregister_fbi(&rfbdev->helper); if (rfb->base.obj[0]) { + for (i = 0; i < rfb->base.format->num_planes; i++) + drm_gem_object_put(rfb->base.obj[0]); amdgpufb_destroy_pinned_object(rfb->base.obj[0]); rfb->base.obj[0] = NULL; drm_framebuffer_unregister_private(&rfb->base); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c new file mode 100644 index 000000000000..dbebbe16e3b3 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c @@ -0,0 +1,104 @@ +// SPDX-License-Identifier: MIT +/* Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: David Nieto + * Roy Sun + */ + +#include <linux/debugfs.h> +#include <linux/list.h> +#include <linux/module.h> +#include <linux/uaccess.h> +#include <linux/reboot.h> +#include <linux/syscalls.h> + +#include <drm/amdgpu_drm.h> +#include <drm/drm_debugfs.h> + +#include "amdgpu.h" +#include "amdgpu_vm.h" +#include "amdgpu_gem.h" +#include "amdgpu_ctx.h" +#include "amdgpu_fdinfo.h" + + +static const char *amdgpu_ip_name[AMDGPU_HW_IP_NUM] = { + [AMDGPU_HW_IP_GFX] = "gfx", + [AMDGPU_HW_IP_COMPUTE] = "compute", + [AMDGPU_HW_IP_DMA] = "dma", + [AMDGPU_HW_IP_UVD] = "dec", + [AMDGPU_HW_IP_VCE] = "enc", + [AMDGPU_HW_IP_UVD_ENC] = "enc_1", + [AMDGPU_HW_IP_VCN_DEC] = "dec", + [AMDGPU_HW_IP_VCN_ENC] = "enc", + [AMDGPU_HW_IP_VCN_JPEG] = "jpeg", +}; + +void amdgpu_show_fdinfo(struct seq_file *m, struct file *f) +{ + struct amdgpu_fpriv *fpriv; + uint32_t bus, dev, fn, i, domain; + uint64_t vram_mem = 0, gtt_mem = 0, cpu_mem = 0; + struct drm_file *file = f->private_data; + struct amdgpu_device *adev = drm_to_adev(file->minor->dev); + int ret; + + ret = amdgpu_file_to_fpriv(f, &fpriv); + if (ret) + return; + bus = adev->pdev->bus->number; + domain = pci_domain_nr(adev->pdev->bus); + dev = PCI_SLOT(adev->pdev->devfn); + fn = PCI_FUNC(adev->pdev->devfn); + + ret = amdgpu_bo_reserve(fpriv->vm.root.base.bo, false); + if (ret) { + DRM_ERROR("Fail to reserve bo\n"); + return; + } + amdgpu_vm_get_memory(&fpriv->vm, &vram_mem, >t_mem, &cpu_mem); + amdgpu_bo_unreserve(fpriv->vm.root.base.bo); + seq_printf(m, "pdev:\t%04x:%02x:%02x.%d\npasid:\t%u\n", domain, bus, + dev, fn, fpriv->vm.pasid); + seq_printf(m, "vram mem:\t%llu kB\n", vram_mem/1024UL); + seq_printf(m, "gtt mem:\t%llu kB\n", gtt_mem/1024UL); + seq_printf(m, "cpu mem:\t%llu kB\n", cpu_mem/1024UL); + for (i = 0; i < AMDGPU_HW_IP_NUM; i++) { + uint32_t count = amdgpu_ctx_num_entities[i]; + int idx = 0; + uint64_t total = 0, min = 0; + uint32_t perc, frac; + + for (idx = 0; idx < count; idx++) { + total = amdgpu_ctx_mgr_fence_usage(&fpriv->ctx_mgr, + i, idx, &min); + if ((total == 0) || (min == 0)) + continue; + + perc = div64_u64(10000 * total, min); + frac = perc % 100; + + seq_printf(m, "%s%d:\t%d.%d%%\n", + amdgpu_ip_name[i], + idx, perc/100, frac); + } + } +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h new file mode 100644 index 000000000000..41a4c7056729 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h @@ -0,0 +1,43 @@ +/* SPDX-License-Identifier: MIT + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: David Nieto + * Roy Sun + */ +#ifndef __AMDGPU_SMI_H__ +#define __AMDGPU_SMI_H__ + +#include <linux/idr.h> +#include <linux/kfifo.h> +#include <linux/rbtree.h> +#include <drm/gpu_scheduler.h> +#include <drm/drm_file.h> +#include <drm/ttm/ttm_bo_driver.h> +#include <linux/sched/mm.h> + +#include "amdgpu_sync.h" +#include "amdgpu_ring.h" +#include "amdgpu_ids.h" + +uint32_t amdgpu_get_ip_count(struct amdgpu_device *adev, int id); +void amdgpu_show_fdinfo(struct seq_file *m, struct file *f); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index 47ea46859618..30772608eac6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -434,6 +434,7 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring, * * @ring: ring to init the fence driver on * @num_hw_submission: number of entries on the hardware queue + * @sched_score: optional score atomic shared with other schedulers * * Init the fence driver for the requested ring (all asics). * Helper function for amdgpu_fence_driver_init(). diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index c5a9a4fb10d2..5562b5c90c03 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -60,7 +60,7 @@ */ /** - * amdgpu_dummy_page_init - init dummy page used by the driver + * amdgpu_gart_dummy_page_init - init dummy page used by the driver * * @adev: amdgpu_device pointer * @@ -86,7 +86,7 @@ static int amdgpu_gart_dummy_page_init(struct amdgpu_device *adev) } /** - * amdgpu_dummy_page_fini - free dummy page used by the driver + * amdgpu_gart_dummy_page_fini - free dummy page used by the driver * * @adev: amdgpu_device pointer * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 311bcdc59eda..18974bd081f0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -766,7 +766,7 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, void __user *out = u64_to_user_ptr(args->value); info.bo_size = robj->tbo.base.size; - info.alignment = robj->tbo.mem.page_alignment << PAGE_SHIFT; + info.alignment = robj->tbo.page_alignment << PAGE_SHIFT; info.domains = robj->preferred_domains; info.domain_flags = robj->flags; amdgpu_bo_unreserve(robj); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 95d4f43a03df..a0be0772c8b3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -607,7 +607,6 @@ int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev) struct ras_ih_if ih_info = { .cb = amdgpu_gfx_process_ras_data_cb, }; - struct ras_query_if info = { 0 }; if (!adev->gfx.ras_if) { adev->gfx.ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); @@ -625,12 +624,8 @@ int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev) goto free; if (amdgpu_ras_is_supported(adev, adev->gfx.ras_if->block)) { - if (adev->gmc.xgmi.connected_to_cpu) { - info.head = *adev->gfx.ras_if; - amdgpu_ras_query_error_status(adev, &info); - } else { + if (!amdgpu_persistent_edc_harvesting_supported(adev)) amdgpu_ras_reset_error_status(adev, AMDGPU_RAS_BLOCK__GFX); - } r = amdgpu_irq_get(adev, &adev->gfx.cp_ecc_error_irq, 0); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index c39ed9eb0987..a129ecc73869 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -333,6 +333,17 @@ void amdgpu_gmc_agp_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc) } /** + * amdgpu_gmc_fault_key - get hask key from vm fault address and pasid + * + * @addr: 48 bit physical address, page aligned (36 significant bits) + * @pasid: 16 bit process address space identifier + */ +static inline uint64_t amdgpu_gmc_fault_key(uint64_t addr, uint16_t pasid) +{ + return addr << 4 | pasid; +} + +/** * amdgpu_gmc_filter_faults - filter VM faults * * @adev: amdgpu device structure @@ -348,8 +359,7 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr, uint16_t pasid, uint64_t timestamp) { struct amdgpu_gmc *gmc = &adev->gmc; - - uint64_t stamp, key = addr << 4 | pasid; + uint64_t stamp, key = amdgpu_gmc_fault_key(addr, pasid); struct amdgpu_gmc_fault *fault; uint32_t hash; @@ -365,7 +375,7 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr, while (fault->timestamp >= stamp) { uint64_t tmp; - if (fault->key == key) + if (atomic64_read(&fault->key) == key) return true; tmp = fault->timestamp; @@ -378,7 +388,7 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr, /* Add the fault to the ring */ fault = &gmc->fault_ring[gmc->last_fault]; - fault->key = key; + atomic64_set(&fault->key, key); fault->timestamp = timestamp; /* And update the hash */ @@ -387,6 +397,36 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr, return false; } +/** + * amdgpu_gmc_filter_faults_remove - remove address from VM faults filter + * + * @adev: amdgpu device structure + * @addr: address of the VM fault + * @pasid: PASID of the process causing the fault + * + * Remove the address from fault filter, then future vm fault on this address + * will pass to retry fault handler to recover. + */ +void amdgpu_gmc_filter_faults_remove(struct amdgpu_device *adev, uint64_t addr, + uint16_t pasid) +{ + struct amdgpu_gmc *gmc = &adev->gmc; + uint64_t key = amdgpu_gmc_fault_key(addr, pasid); + struct amdgpu_gmc_fault *fault; + uint32_t hash; + uint64_t tmp; + + hash = hash_64(key, AMDGPU_GMC_FAULT_HASH_ORDER); + fault = &gmc->fault_ring[gmc->fault_hash[hash].idx]; + do { + if (atomic64_cmpxchg(&fault->key, key, 0) == key) + break; + + tmp = fault->timestamp; + fault = &gmc->fault_ring[fault->next]; + } while (fault->timestamp < tmp); +} + int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) { int r; @@ -415,6 +455,13 @@ int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) return r; } + if (adev->hdp.ras_funcs && + adev->hdp.ras_funcs->ras_late_init) { + r = adev->hdp.ras_funcs->ras_late_init(adev); + if (r) + return r; + } + return 0; } @@ -426,11 +473,15 @@ void amdgpu_gmc_ras_fini(struct amdgpu_device *adev) if (adev->mmhub.ras_funcs && adev->mmhub.ras_funcs->ras_fini) - amdgpu_mmhub_ras_fini(adev); + adev->mmhub.ras_funcs->ras_fini(adev); if (adev->gmc.xgmi.ras_funcs && adev->gmc.xgmi.ras_funcs->ras_fini) adev->gmc.xgmi.ras_funcs->ras_fini(adev); + + if (adev->hdp.ras_funcs && + adev->hdp.ras_funcs->ras_fini) + adev->hdp.ras_funcs->ras_fini(adev); } /* diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index 9d11c02a3938..6aa1d52d3aee 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -66,9 +66,9 @@ struct firmware; * GMC page fault information */ struct amdgpu_gmc_fault { - uint64_t timestamp; + uint64_t timestamp:48; uint64_t next:AMDGPU_GMC_FAULT_RING_ORDER; - uint64_t key:52; + atomic64_t key; }; /* @@ -318,6 +318,8 @@ void amdgpu_gmc_agp_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc); bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr, uint16_t pasid, uint64_t timestamp); +void amdgpu_gmc_filter_faults_remove(struct amdgpu_device *adev, uint64_t addr, + uint16_t pasid); int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev); void amdgpu_gmc_ras_fini(struct amdgpu_device *adev); int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c index 540c01052b21..c026972ca9a1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c @@ -24,7 +24,8 @@ #include "amdgpu.h" -static inline struct amdgpu_gtt_mgr *to_gtt_mgr(struct ttm_resource_manager *man) +static inline struct amdgpu_gtt_mgr * +to_gtt_mgr(struct ttm_resource_manager *man) { return container_of(man, struct amdgpu_gtt_mgr, manager); } @@ -43,12 +44,14 @@ struct amdgpu_gtt_node { * the GTT block, in bytes */ static ssize_t amdgpu_mem_info_gtt_total_show(struct device *dev, - struct device_attribute *attr, char *buf) + struct device_attribute *attr, + char *buf) { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); + struct ttm_resource_manager *man; + man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); return sysfs_emit(buf, "%llu\n", man->size * PAGE_SIZE); } @@ -61,12 +64,14 @@ static ssize_t amdgpu_mem_info_gtt_total_show(struct device *dev, * size of the GTT block, in bytes */ static ssize_t amdgpu_mem_info_gtt_used_show(struct device *dev, - struct device_attribute *attr, char *buf) + struct device_attribute *attr, + char *buf) { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); + struct ttm_resource_manager *man; + man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); return sysfs_emit(buf, "%llu\n", amdgpu_gtt_mgr_usage(man)); } @@ -75,80 +80,6 @@ static DEVICE_ATTR(mem_info_gtt_total, S_IRUGO, static DEVICE_ATTR(mem_info_gtt_used, S_IRUGO, amdgpu_mem_info_gtt_used_show, NULL); -static const struct ttm_resource_manager_func amdgpu_gtt_mgr_func; -/** - * amdgpu_gtt_mgr_init - init GTT manager and DRM MM - * - * @adev: amdgpu_device pointer - * @gtt_size: maximum size of GTT - * - * Allocate and initialize the GTT manager. - */ -int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size) -{ - struct amdgpu_gtt_mgr *mgr = &adev->mman.gtt_mgr; - struct ttm_resource_manager *man = &mgr->manager; - uint64_t start, size; - int ret; - - man->use_tt = true; - man->func = &amdgpu_gtt_mgr_func; - - ttm_resource_manager_init(man, gtt_size >> PAGE_SHIFT); - - start = AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GTT_NUM_TRANSFER_WINDOWS; - size = (adev->gmc.gart_size >> PAGE_SHIFT) - start; - drm_mm_init(&mgr->mm, start, size); - spin_lock_init(&mgr->lock); - atomic64_set(&mgr->available, gtt_size >> PAGE_SHIFT); - - ret = device_create_file(adev->dev, &dev_attr_mem_info_gtt_total); - if (ret) { - DRM_ERROR("Failed to create device file mem_info_gtt_total\n"); - return ret; - } - ret = device_create_file(adev->dev, &dev_attr_mem_info_gtt_used); - if (ret) { - DRM_ERROR("Failed to create device file mem_info_gtt_used\n"); - return ret; - } - - ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_TT, &mgr->manager); - ttm_resource_manager_set_used(man, true); - return 0; -} - -/** - * amdgpu_gtt_mgr_fini - free and destroy GTT manager - * - * @adev: amdgpu_device pointer - * - * Destroy and free the GTT manager, returns -EBUSY if ranges are still - * allocated inside it. - */ -void amdgpu_gtt_mgr_fini(struct amdgpu_device *adev) -{ - struct amdgpu_gtt_mgr *mgr = &adev->mman.gtt_mgr; - struct ttm_resource_manager *man = &mgr->manager; - int ret; - - ttm_resource_manager_set_used(man, false); - - ret = ttm_resource_manager_evict_all(&adev->mman.bdev, man); - if (ret) - return; - - spin_lock(&mgr->lock); - drm_mm_takedown(&mgr->mm); - spin_unlock(&mgr->lock); - - device_remove_file(adev->dev, &dev_attr_mem_info_gtt_total); - device_remove_file(adev->dev, &dev_attr_mem_info_gtt_used); - - ttm_resource_manager_cleanup(man); - ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_TT, NULL); -} - /** * amdgpu_gtt_mgr_has_gart_addr - Check if mem has address space * @@ -205,7 +136,7 @@ static int amdgpu_gtt_mgr_new(struct ttm_resource_manager *man, spin_lock(&mgr->lock); r = drm_mm_insert_node_in_range(&mgr->mm, &node->node, mem->num_pages, - mem->page_alignment, 0, place->fpfn, + tbo->page_alignment, 0, place->fpfn, place->lpfn, DRM_MM_INSERT_BEST); spin_unlock(&mgr->lock); @@ -265,6 +196,13 @@ uint64_t amdgpu_gtt_mgr_usage(struct ttm_resource_manager *man) return (result > 0 ? result : 0) * PAGE_SIZE; } +/** + * amdgpu_gtt_mgr_recover - re-init gart + * + * @man: TTM memory type manager + * + * Re-init the gart for each known BO in the GTT. + */ int amdgpu_gtt_mgr_recover(struct ttm_resource_manager *man) { struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man); @@ -311,3 +249,76 @@ static const struct ttm_resource_manager_func amdgpu_gtt_mgr_func = { .free = amdgpu_gtt_mgr_del, .debug = amdgpu_gtt_mgr_debug }; + +/** + * amdgpu_gtt_mgr_init - init GTT manager and DRM MM + * + * @adev: amdgpu_device pointer + * @gtt_size: maximum size of GTT + * + * Allocate and initialize the GTT manager. + */ +int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size) +{ + struct amdgpu_gtt_mgr *mgr = &adev->mman.gtt_mgr; + struct ttm_resource_manager *man = &mgr->manager; + uint64_t start, size; + int ret; + + man->use_tt = true; + man->func = &amdgpu_gtt_mgr_func; + + ttm_resource_manager_init(man, gtt_size >> PAGE_SHIFT); + + start = AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GTT_NUM_TRANSFER_WINDOWS; + size = (adev->gmc.gart_size >> PAGE_SHIFT) - start; + drm_mm_init(&mgr->mm, start, size); + spin_lock_init(&mgr->lock); + atomic64_set(&mgr->available, gtt_size >> PAGE_SHIFT); + + ret = device_create_file(adev->dev, &dev_attr_mem_info_gtt_total); + if (ret) { + DRM_ERROR("Failed to create device file mem_info_gtt_total\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_mem_info_gtt_used); + if (ret) { + DRM_ERROR("Failed to create device file mem_info_gtt_used\n"); + return ret; + } + + ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_TT, &mgr->manager); + ttm_resource_manager_set_used(man, true); + return 0; +} + +/** + * amdgpu_gtt_mgr_fini - free and destroy GTT manager + * + * @adev: amdgpu_device pointer + * + * Destroy and free the GTT manager, returns -EBUSY if ranges are still + * allocated inside it. + */ +void amdgpu_gtt_mgr_fini(struct amdgpu_device *adev) +{ + struct amdgpu_gtt_mgr *mgr = &adev->mman.gtt_mgr; + struct ttm_resource_manager *man = &mgr->manager; + int ret; + + ttm_resource_manager_set_used(man, false); + + ret = ttm_resource_manager_evict_all(&adev->mman.bdev, man); + if (ret) + return; + + spin_lock(&mgr->lock); + drm_mm_takedown(&mgr->mm); + spin_unlock(&mgr->lock); + + device_remove_file(adev->dev, &dev_attr_mem_info_gtt_total); + device_remove_file(adev->dev, &dev_attr_mem_info_gtt_used); + + ttm_resource_manager_cleanup(man); + ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_TT, NULL); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c new file mode 100644 index 000000000000..1d50d534d77c --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c @@ -0,0 +1,69 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "amdgpu_ras.h" + +int amdgpu_hdp_ras_late_init(struct amdgpu_device *adev) +{ + int r; + struct ras_ih_if ih_info = { + .cb = NULL, + }; + struct ras_fs_if fs_info = { + .sysfs_name = "hdp_err_count", + }; + + if (!adev->hdp.ras_if) { + adev->hdp.ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); + if (!adev->hdp.ras_if) + return -ENOMEM; + adev->hdp.ras_if->block = AMDGPU_RAS_BLOCK__HDP; + adev->hdp.ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; + adev->hdp.ras_if->sub_block_index = 0; + strcpy(adev->hdp.ras_if->name, "hdp"); + } + ih_info.head = fs_info.head = *adev->hdp.ras_if; + r = amdgpu_ras_late_init(adev, adev->hdp.ras_if, + &fs_info, &ih_info); + if (r || !amdgpu_ras_is_supported(adev, adev->hdp.ras_if->block)) { + kfree(adev->hdp.ras_if); + adev->hdp.ras_if = NULL; + } + + return r; +} + +void amdgpu_hdp_ras_fini(struct amdgpu_device *adev) +{ + if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__HDP) && + adev->hdp.ras_if) { + struct ras_common_if *ras_if = adev->hdp.ras_if; + struct ras_ih_if ih_info = { + .cb = NULL, + }; + + amdgpu_ras_late_fini(adev, ras_if, &ih_info); + kfree(ras_if); + } +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h index 43caf9f8cc11..7ec99d591584 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h @@ -23,18 +23,29 @@ #ifndef __AMDGPU_HDP_H__ #define __AMDGPU_HDP_H__ +struct amdgpu_hdp_ras_funcs { + int (*ras_late_init)(struct amdgpu_device *adev); + void (*ras_fini)(struct amdgpu_device *adev); + void (*query_ras_error_count)(struct amdgpu_device *adev, + void *ras_error_status); + void (*reset_ras_error_count)(struct amdgpu_device *adev); +}; + struct amdgpu_hdp_funcs { void (*flush_hdp)(struct amdgpu_device *adev, struct amdgpu_ring *ring); void (*invalidate_hdp)(struct amdgpu_device *adev, struct amdgpu_ring *ring); - void (*reset_ras_error_count)(struct amdgpu_device *adev); void (*update_clock_gating)(struct amdgpu_device *adev, bool enable); void (*get_clock_gating_state)(struct amdgpu_device *adev, u32 *flags); void (*init_registers)(struct amdgpu_device *adev); }; struct amdgpu_hdp { + struct ras_common_if *ras_if; const struct amdgpu_hdp_funcs *funcs; + const struct amdgpu_hdp_ras_funcs *ras_funcs; }; +int amdgpu_hdp_ras_late_init(struct amdgpu_device *adev); +void amdgpu_hdp_ras_fini(struct amdgpu_device *adev); #endif /* __AMDGPU_HDP_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index a2fe2dac32c1..2e6789a7dc46 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -328,7 +328,7 @@ int amdgpu_ib_pool_init(struct amdgpu_device *adev) for (i = 0; i < AMDGPU_IB_POOL_MAX; i++) { if (i == AMDGPU_IB_POOL_DIRECT) - size = PAGE_SIZE * 2; + size = PAGE_SIZE * 6; else size = AMDGPU_IB_POOL_SIZE; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c index faaa6aa2faaf..a36e191cf086 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c @@ -175,7 +175,9 @@ static bool amdgpu_ih_has_checkpoint_processed(struct amdgpu_device *adev, cur_rptr += ih->ptr_mask + 1; *prev_rptr = cur_rptr; - return cur_rptr >= checkpoint_wptr; + /* check ring is empty to workaround missing wptr overflow flag */ + return cur_rptr >= checkpoint_wptr || + (cur_rptr & ih->ptr_mask) == amdgpu_ih_get_wptr(adev, ih); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 39ee88d29cca..8d12e474745a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -986,7 +986,7 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) if (!ras) return -EINVAL; - ras_mask = (uint64_t)ras->supported << 32 | ras->features; + ras_mask = (uint64_t)adev->ras_enabled << 32 | ras->features; return copy_to_user(out, &ras_mask, min_t(u64, size, sizeof(ras_mask))) ? @@ -1114,7 +1114,8 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) dev_warn(adev->dev, "No more PASIDs available!"); pasid = 0; } - r = amdgpu_vm_init(adev, &fpriv->vm, AMDGPU_VM_CONTEXT_GFX, pasid); + + r = amdgpu_vm_init(adev, &fpriv->vm, pasid); if (r) goto error_pasid; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h index 11aa29933c1f..b27fcbccce2b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h @@ -28,6 +28,7 @@ struct amdgpu_mmhub_ras_funcs { void *ras_error_status); void (*query_ras_error_status)(struct amdgpu_device *adev); void (*reset_ras_error_count)(struct amdgpu_device *adev); + void (*reset_ras_error_status)(struct amdgpu_device *adev); }; struct amdgpu_mmhub_funcs { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c index 828b5167ff12..2741c28ff1b5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c @@ -155,3 +155,89 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo) mmu_interval_notifier_remove(&bo->notifier); bo->notifier.mm = NULL; } + +int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier, + struct mm_struct *mm, struct page **pages, + uint64_t start, uint64_t npages, + struct hmm_range **phmm_range, bool readonly, + bool mmap_locked) +{ + struct hmm_range *hmm_range; + unsigned long timeout; + unsigned long i; + unsigned long *pfns; + int r = 0; + + hmm_range = kzalloc(sizeof(*hmm_range), GFP_KERNEL); + if (unlikely(!hmm_range)) + return -ENOMEM; + + pfns = kvmalloc_array(npages, sizeof(*pfns), GFP_KERNEL); + if (unlikely(!pfns)) { + r = -ENOMEM; + goto out_free_range; + } + + hmm_range->notifier = notifier; + hmm_range->default_flags = HMM_PFN_REQ_FAULT; + if (!readonly) + hmm_range->default_flags |= HMM_PFN_REQ_WRITE; + hmm_range->hmm_pfns = pfns; + hmm_range->start = start; + hmm_range->end = start + npages * PAGE_SIZE; + + /* Assuming 512MB takes maxmium 1 second to fault page address */ + timeout = max(npages >> 17, 1ULL) * HMM_RANGE_DEFAULT_TIMEOUT; + timeout = jiffies + msecs_to_jiffies(timeout); + +retry: + hmm_range->notifier_seq = mmu_interval_read_begin(notifier); + + if (likely(!mmap_locked)) + mmap_read_lock(mm); + + r = hmm_range_fault(hmm_range); + + if (likely(!mmap_locked)) + mmap_read_unlock(mm); + if (unlikely(r)) { + /* + * FIXME: This timeout should encompass the retry from + * mmu_interval_read_retry() as well. + */ + if (r == -EBUSY && !time_after(jiffies, timeout)) + goto retry; + goto out_free_pfns; + } + + /* + * Due to default_flags, all pages are HMM_PFN_VALID or + * hmm_range_fault() fails. FIXME: The pages cannot be touched outside + * the notifier_lock, and mmu_interval_read_retry() must be done first. + */ + for (i = 0; pages && i < npages; i++) + pages[i] = hmm_pfn_to_page(pfns[i]); + + *phmm_range = hmm_range; + + return 0; + +out_free_pfns: + kvfree(pfns); +out_free_range: + kfree(hmm_range); + + return r; +} + +int amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range) +{ + int r; + + r = mmu_interval_read_retry(hmm_range->notifier, + hmm_range->notifier_seq); + kvfree(hmm_range->hmm_pfns); + kfree(hmm_range); + + return r; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h index a292238f75eb..7f7d37a457c3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h @@ -30,6 +30,13 @@ #include <linux/workqueue.h> #include <linux/interval_tree.h> +int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier, + struct mm_struct *mm, struct page **pages, + uint64_t start, uint64_t npages, + struct hmm_range **phmm_range, bool readonly, + bool mmap_locked); +int amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range); + #if defined(CONFIG_HMM_MIRROR) int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr); void amdgpu_mn_unregister(struct amdgpu_bo *bo); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index cb0b581bbce7..89fb372ed49c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -344,7 +344,7 @@ struct amdgpu_mode_info { /* pointer to fbdev info structure */ struct amdgpu_fbdev *rfbdev; /* firmware flags */ - u16 firmware_flags; + u32 firmware_flags; /* pointer to backlight encoder */ struct amdgpu_encoder *bl_encoder; u8 bl_level; /* saved backlight level */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 1345f7eba011..8714d50c5b20 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -52,36 +52,12 @@ * */ -/** - * amdgpu_bo_subtract_pin_size - Remove BO from pin_size accounting - * - * @bo: &amdgpu_bo buffer object - * - * This function is called when a BO stops being pinned, and updates the - * &amdgpu_device pin_size values accordingly. - */ -static void amdgpu_bo_subtract_pin_size(struct amdgpu_bo *bo) -{ - struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - - if (bo->tbo.mem.mem_type == TTM_PL_VRAM) { - atomic64_sub(amdgpu_bo_size(bo), &adev->vram_pin_size); - atomic64_sub(amdgpu_vram_mgr_bo_visible_size(bo), - &adev->visible_pin_size); - } else if (bo->tbo.mem.mem_type == TTM_PL_TT) { - atomic64_sub(amdgpu_bo_size(bo), &adev->gart_pin_size); - } -} - static void amdgpu_bo_destroy(struct ttm_buffer_object *tbo) { struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev); struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo); struct amdgpu_bo_user *ubo; - if (bo->tbo.pin_count > 0) - amdgpu_bo_subtract_pin_size(bo); - amdgpu_bo_kunmap(bo); if (bo->tbo.base.import_attach) @@ -515,7 +491,18 @@ bool amdgpu_bo_support_uswc(u64 bo_flags) #endif } -static int amdgpu_bo_do_create(struct amdgpu_device *adev, +/** + * amdgpu_bo_create - create an &amdgpu_bo buffer object + * @adev: amdgpu device object + * @bp: parameters to be used for the buffer object + * @bo_ptr: pointer to the buffer object pointer + * + * Creates an &amdgpu_bo buffer object. + * + * Returns: + * 0 for success or a negative error code on failure. + */ +int amdgpu_bo_create(struct amdgpu_device *adev, struct amdgpu_bo_param *bp, struct amdgpu_bo **bo_ptr) { @@ -625,9 +612,9 @@ fail_unreserve: return r; } -static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, - unsigned long size, - struct amdgpu_bo *bo) +int amdgpu_bo_create_shadow(struct amdgpu_device *adev, + unsigned long size, + struct amdgpu_bo *bo) { struct amdgpu_bo_param bp; int r; @@ -638,13 +625,12 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, memset(&bp, 0, sizeof(bp)); bp.size = size; bp.domain = AMDGPU_GEM_DOMAIN_GTT; - bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC | - AMDGPU_GEM_CREATE_SHADOW; + bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC; bp.type = ttm_bo_type_kernel; bp.resv = bo->tbo.base.resv; bp.bo_ptr_size = sizeof(struct amdgpu_bo); - r = amdgpu_bo_do_create(adev, &bp, &bo->shadow); + r = amdgpu_bo_create(adev, &bp, &bo->shadow); if (!r) { bo->shadow->parent = amdgpu_bo_ref(bo); mutex_lock(&adev->shadow_list_lock); @@ -656,50 +642,6 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, } /** - * amdgpu_bo_create - create an &amdgpu_bo buffer object - * @adev: amdgpu device object - * @bp: parameters to be used for the buffer object - * @bo_ptr: pointer to the buffer object pointer - * - * Creates an &amdgpu_bo buffer object; and if requested, also creates a - * shadow object. - * Shadow object is used to backup the original buffer object, and is always - * in GTT. - * - * Returns: - * 0 for success or a negative error code on failure. - */ -int amdgpu_bo_create(struct amdgpu_device *adev, - struct amdgpu_bo_param *bp, - struct amdgpu_bo **bo_ptr) -{ - u64 flags = bp->flags; - int r; - - bp->flags = bp->flags & ~AMDGPU_GEM_CREATE_SHADOW; - - r = amdgpu_bo_do_create(adev, bp, bo_ptr); - if (r) - return r; - - if ((flags & AMDGPU_GEM_CREATE_SHADOW) && !(adev->flags & AMD_IS_APU)) { - if (!bp->resv) - WARN_ON(dma_resv_lock((*bo_ptr)->tbo.base.resv, - NULL)); - - r = amdgpu_bo_create_shadow(adev, bp->size, *bo_ptr); - - if (!bp->resv) - dma_resv_unlock((*bo_ptr)->tbo.base.resv); - - if (r) - amdgpu_bo_unref(bo_ptr); - } - - return r; -} - -/** * amdgpu_bo_create_user - create an &amdgpu_bo_user buffer object * @adev: amdgpu device object * @bp: parameters to be used for the buffer object @@ -718,9 +660,8 @@ int amdgpu_bo_create_user(struct amdgpu_device *adev, struct amdgpu_bo *bo_ptr; int r; - bp->flags = bp->flags & ~AMDGPU_GEM_CREATE_SHADOW; bp->bo_ptr_size = sizeof(struct amdgpu_bo_user); - r = amdgpu_bo_do_create(adev, bp, &bo_ptr); + r = amdgpu_bo_create(adev, bp, &bo_ptr); if (r) return r; @@ -1037,14 +978,22 @@ int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain) */ void amdgpu_bo_unpin(struct amdgpu_bo *bo) { + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + ttm_bo_unpin(&bo->tbo); if (bo->tbo.pin_count) return; - amdgpu_bo_subtract_pin_size(bo); - if (bo->tbo.base.import_attach) dma_buf_unpin(bo->tbo.base.import_attach); + + if (bo->tbo.mem.mem_type == TTM_PL_VRAM) { + atomic64_sub(amdgpu_bo_size(bo), &adev->vram_pin_size); + atomic64_sub(amdgpu_vram_mgr_bo_visible_size(bo), + &adev->visible_pin_size); + } else if (bo->tbo.mem.mem_type == TTM_PL_TT) { + atomic64_sub(amdgpu_bo_size(bo), &adev->gart_pin_size); + } } /** @@ -1304,6 +1253,26 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, trace_amdgpu_bo_move(abo, new_mem->mem_type, old_mem->mem_type); } +void amdgpu_bo_get_memory(struct amdgpu_bo *bo, uint64_t *vram_mem, + uint64_t *gtt_mem, uint64_t *cpu_mem) +{ + unsigned int domain; + + domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type); + switch (domain) { + case AMDGPU_GEM_DOMAIN_VRAM: + *vram_mem += amdgpu_bo_size(bo); + break; + case AMDGPU_GEM_DOMAIN_GTT: + *gtt_mem += amdgpu_bo_size(bo); + break; + case AMDGPU_GEM_DOMAIN_CPU: + default: + *cpu_mem += amdgpu_bo_size(bo); + break; + } +} + /** * amdgpu_bo_release_notify - notification about a BO being released * @bo: pointer to a buffer object @@ -1362,7 +1331,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct ttm_operation_ctx ctx = { false, false }; struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo); - unsigned long offset, size; + unsigned long offset; int r; /* Remember that this BO was accessed by the CPU */ @@ -1371,9 +1340,8 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) if (bo->mem.mem_type != TTM_PL_VRAM) return 0; - size = bo->mem.num_pages << PAGE_SHIFT; offset = bo->mem.start << PAGE_SHIFT; - if ((offset + size) <= adev->gmc.visible_vram_size) + if ((offset + bo->base.size) <= adev->gmc.visible_vram_size) return 0; /* Can't move a pinned BO to visible VRAM */ @@ -1398,7 +1366,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) offset = bo->mem.start << PAGE_SHIFT; /* this should never happen */ if (bo->mem.mem_type == TTM_PL_VRAM && - (offset + size) > adev->gmc.visible_vram_size) + (offset + bo->base.size) > adev->gmc.visible_vram_size) return VM_FAULT_SIGBUS; ttm_bo_move_to_lru_tail_unlocked(bo); @@ -1592,7 +1560,6 @@ u64 amdgpu_bo_print_info(int id, struct amdgpu_bo *bo, struct seq_file *m) amdgpu_bo_print_flag(m, bo, NO_CPU_ACCESS); amdgpu_bo_print_flag(m, bo, CPU_GTT_USWC); amdgpu_bo_print_flag(m, bo, VRAM_CLEARED); - amdgpu_bo_print_flag(m, bo, SHADOW); amdgpu_bo_print_flag(m, bo, VRAM_CONTIGUOUS); amdgpu_bo_print_flag(m, bo, VM_ALWAYS_VALID); amdgpu_bo_print_flag(m, bo, EXPLICIT_SYNC); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 2d1fefbe1e99..a44779d3e0a7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -30,6 +30,8 @@ #include <drm/amdgpu_drm.h> #include "amdgpu.h" +#include "amdgpu_res_cursor.h" + #ifdef CONFIG_MMU_NOTIFIER #include <linux/mmu_notifier.h> #endif @@ -37,6 +39,10 @@ #define AMDGPU_BO_INVALID_OFFSET LONG_MAX #define AMDGPU_BO_MAX_PLACEMENTS 3 +/* BO flag to indicate a KFD userptr BO */ +#define AMDGPU_AMDKFD_CREATE_USERPTR_BO (1ULL << 63) +#define AMDGPU_AMDKFD_CREATE_SVM_BO (1ULL << 62) + #define to_amdgpu_bo_user(abo) container_of((abo), struct amdgpu_bo_user, bo) struct amdgpu_bo_param { @@ -191,7 +197,7 @@ static inline unsigned amdgpu_bo_ngpu_pages(struct amdgpu_bo *bo) static inline unsigned amdgpu_bo_gpu_page_alignment(struct amdgpu_bo *bo) { - return (bo->tbo.mem.page_alignment << PAGE_SHIFT) / AMDGPU_GPU_PAGE_SIZE; + return (bo->tbo.page_alignment << PAGE_SHIFT) / AMDGPU_GPU_PAGE_SIZE; } /** @@ -211,18 +217,19 @@ static inline u64 amdgpu_bo_mmap_offset(struct amdgpu_bo *bo) static inline bool amdgpu_bo_in_cpu_visible_vram(struct amdgpu_bo *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - unsigned fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT; - struct drm_mm_node *node = bo->tbo.mem.mm_node; - unsigned long pages_left; + struct amdgpu_res_cursor cursor; if (bo->tbo.mem.mem_type != TTM_PL_VRAM) return false; - for (pages_left = bo->tbo.mem.num_pages; pages_left; - pages_left -= node->size, node++) - if (node->start < fpfn) + amdgpu_res_first(&bo->tbo.mem, 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; } @@ -267,6 +274,9 @@ int amdgpu_bo_create_user(struct amdgpu_device *adev, struct amdgpu_bo_user **ubo_ptr); void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr, void **cpu_addr); +int amdgpu_bo_create_shadow(struct amdgpu_device *adev, + unsigned long size, + struct amdgpu_bo *bo); int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr); void *amdgpu_bo_kptr(struct amdgpu_bo *bo); void amdgpu_bo_kunmap(struct amdgpu_bo *bo); @@ -300,6 +310,8 @@ int amdgpu_bo_sync_wait(struct amdgpu_bo *bo, void *owner, bool intr); u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo); u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo); int amdgpu_bo_validate(struct amdgpu_bo *bo); +void amdgpu_bo_get_memory(struct amdgpu_bo *bo, uint64_t *vram_mem, + uint64_t *gtt_mem, uint64_t *cpu_mem); int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow, struct dma_fence **fence); uint32_t amdgpu_bo_get_preferred_pin_domain(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index a09483beb968..55378c6b9722 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -38,6 +38,7 @@ #include "amdgpu_ras.h" #include "amdgpu_securedisplay.h" +#include "amdgpu_atomfirmware.h" static int psp_sysfs_init(struct amdgpu_device *adev); static void psp_sysfs_fini(struct amdgpu_device *adev); @@ -104,6 +105,7 @@ static int psp_early_init(void *handle) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: psp_v11_0_set_psp_funcs(psp); psp->autoload_supported = true; break; @@ -417,31 +419,12 @@ static int psp_tmr_init(struct psp_context *psp) return ret; } -static int psp_clear_vf_fw(struct psp_context *psp) -{ - int ret; - struct psp_gfx_cmd_resp *cmd; - - if (!amdgpu_sriov_vf(psp->adev) || psp->adev->asic_type != CHIP_NAVI12) - return 0; - - cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); - if (!cmd) - return -ENOMEM; - - cmd->cmd_id = GFX_CMD_ID_CLEAR_VF_FW; - - ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); - kfree(cmd); - - return ret; -} - static bool psp_skip_tmr(struct psp_context *psp) { switch (psp->adev->asic_type) { case CHIP_NAVI12: case CHIP_SIENNA_CICHLID: + case CHIP_ALDEBARAN: return true; default: return false; @@ -557,7 +540,7 @@ static int psp_boot_config_set(struct amdgpu_device *adev) struct psp_context *psp = &adev->psp; struct psp_gfx_cmd_resp *cmd = psp->cmd; - if (adev->asic_type != CHIP_SIENNA_CICHLID || amdgpu_sriov_vf(adev)) + if (amdgpu_sriov_vf(adev)) return 0; memset(cmd, 0, sizeof(struct psp_gfx_cmd_resp)); @@ -1037,6 +1020,13 @@ static int psp_ras_load(struct psp_context *psp) memset(psp->fw_pri_buf, 0, PSP_1_MEG); memcpy(psp->fw_pri_buf, psp->ta_ras_start_addr, psp->ta_ras_ucode_size); + ras_cmd = (struct ta_ras_shared_memory *)psp->ras.ras_shared_buf; + + if (psp->adev->gmc.xgmi.connected_to_cpu) + ras_cmd->ras_in_message.init_flags.poison_mode_en = 1; + else + ras_cmd->ras_in_message.init_flags.dgpu_mode = 1; + psp_prep_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr, psp->ta_ras_ucode_size, @@ -1046,8 +1036,6 @@ static int psp_ras_load(struct psp_context *psp) ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); - ras_cmd = (struct ta_ras_shared_memory *)psp->ras.ras_shared_buf; - if (!ret) { psp->ras.session_id = cmd->resp.session_id; @@ -1128,6 +1116,31 @@ int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id) return ret; } +static int psp_ras_status_to_errno(struct amdgpu_device *adev, + enum ta_ras_status ras_status) +{ + int ret = -EINVAL; + + switch (ras_status) { + case TA_RAS_STATUS__SUCCESS: + ret = 0; + break; + case TA_RAS_STATUS__RESET_NEEDED: + ret = -EAGAIN; + break; + case TA_RAS_STATUS__ERROR_RAS_NOT_AVAILABLE: + dev_warn(adev->dev, "RAS WARN: ras function unavailable\n"); + break; + case TA_RAS_STATUS__ERROR_ASD_READ_WRITE: + dev_warn(adev->dev, "RAS WARN: asd read or write failed\n"); + break; + default: + dev_err(adev->dev, "RAS ERROR: ras function failed ret 0x%X\n", ret); + } + + return ret; +} + int psp_ras_enable_features(struct psp_context *psp, union ta_ras_cmd_input *info, bool enable) { @@ -1151,7 +1164,7 @@ int psp_ras_enable_features(struct psp_context *psp, if (ret) return -EINVAL; - return ras_cmd->ras_status; + return psp_ras_status_to_errno(psp->adev, ras_cmd->ras_status); } static int psp_ras_terminate(struct psp_context *psp) @@ -1234,7 +1247,7 @@ int psp_ras_trigger_error(struct psp_context *psp, if (amdgpu_ras_intr_triggered()) return 0; - return ras_cmd->ras_status; + return psp_ras_status_to_errno(psp->adev, ras_cmd->ras_status); } // ras end @@ -1920,15 +1933,10 @@ static int psp_hw_start(struct psp_context *psp) return ret; } - ret = psp_clear_vf_fw(psp); - if (ret) { - DRM_ERROR("PSP clear vf fw!\n"); - return ret; - } - - ret = psp_boot_config_set(adev); - if (ret) { - DRM_WARN("PSP set boot config@\n"); + if (amdgpu_atomfirmware_dynamic_boot_config_supported(adev)) { + ret = psp_boot_config_set(adev); + if (ret) + dev_warn(adev->dev, "PSP set boot config failed\n"); } ret = psp_tmr_init(psp); @@ -2166,7 +2174,7 @@ static int psp_load_smu_fw(struct psp_context *psp) return 0; if ((amdgpu_in_reset(adev) && - ras && ras->supported && + ras && adev->ras_enabled && (adev->asic_type == CHIP_ARCTURUS || adev->asic_type == CHIP_VEGA20)) || (adev->in_runpm && @@ -2434,7 +2442,6 @@ static int psp_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct psp_context *psp = &adev->psp; - int ret; if (psp->adev->psp.ta_fw) { psp_ras_terminate(psp); @@ -2445,11 +2452,6 @@ static int psp_hw_fini(void *handle) } psp_asd_unload(psp); - ret = psp_clear_vf_fw(psp); - if (ret) { - DRM_ERROR("PSP clear vf fw!\n"); - return ret; - } psp_tmr_terminate(psp); psp_ring_destroy(psp, PSP_RING_TYPE__KM); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index b0d2fc9454ca..e3a4c3a7635a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -33,6 +33,7 @@ #include "amdgpu_atomfirmware.h" #include "amdgpu_xgmi.h" #include "ivsrcid/nbio/irqsrcs_nbif_7_4.h" +#include "atom.h" static const char *RAS_FS_NAME = "ras"; @@ -320,11 +321,14 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f, * "disable" requires only the block. * "enable" requires the block and error type. * "inject" requires the block, error type, address, and value. + * * The block is one of: umc, sdma, gfx, etc. * see ras_block_string[] for details + * * The error type is one of: ue, ce, where, * ue is multi-uncorrectable * ce is single-correctable + * * 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. * @@ -531,7 +535,7 @@ static struct ras_manager *amdgpu_ras_create_obj(struct amdgpu_device *adev, struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj; - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return NULL; if (head->block >= AMDGPU_RAS_BLOCK_COUNT) @@ -558,7 +562,7 @@ struct ras_manager *amdgpu_ras_find_obj(struct amdgpu_device *adev, struct ras_manager *obj; int i; - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return NULL; if (head) { @@ -585,36 +589,11 @@ struct ras_manager *amdgpu_ras_find_obj(struct amdgpu_device *adev, } /* obj end */ -static void amdgpu_ras_parse_status_code(struct amdgpu_device *adev, - const char* invoke_type, - const char* block_name, - enum ta_ras_status ret) -{ - switch (ret) { - case TA_RAS_STATUS__SUCCESS: - return; - case TA_RAS_STATUS__ERROR_RAS_NOT_AVAILABLE: - dev_warn(adev->dev, - "RAS WARN: %s %s currently unavailable\n", - invoke_type, - block_name); - break; - default: - dev_err(adev->dev, - "RAS ERROR: %s %s error failed ret 0x%X\n", - invoke_type, - block_name, - ret); - } -} - /* feature ctl begin */ static int amdgpu_ras_is_feature_allowed(struct amdgpu_device *adev, - struct ras_common_if *head) + struct ras_common_if *head) { - struct amdgpu_ras *con = amdgpu_ras_get_context(adev); - - return con->hw_supported & BIT(head->block); + return adev->ras_hw_enabled & BIT(head->block); } static int amdgpu_ras_is_feature_enabled(struct amdgpu_device *adev, @@ -658,11 +637,7 @@ static int __amdgpu_ras_feature_enable(struct amdgpu_device *adev, con->features |= BIT(head->block); } else { if (obj && amdgpu_ras_is_feature_enabled(adev, head)) { - /* skip clean gfx ras context feature for VEGA20 Gaming. - * will clean later - */ - if (!(!adev->ras_features && con->features & BIT(AMDGPU_RAS_BLOCK__GFX))) - con->features &= ~BIT(head->block); + con->features &= ~BIT(head->block); put_obj(obj); } } @@ -708,15 +683,10 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev, if (!amdgpu_ras_intr_triggered()) { ret = psp_ras_enable_features(&adev->psp, info, enable); if (ret) { - amdgpu_ras_parse_status_code(adev, - enable ? "enable":"disable", - ras_block_str(head->block), - (enum ta_ras_status)ret); - if (ret == TA_RAS_STATUS__RESET_NEEDED) - ret = -EAGAIN; - else - ret = -EINVAL; - + dev_err(adev->dev, "ras %s %s failed %d\n", + enable ? "enable":"disable", + ras_block_str(head->block), + ret); goto out; } } @@ -770,6 +740,10 @@ int amdgpu_ras_feature_enable_on_boot(struct amdgpu_device *adev, con->features |= BIT(head->block); ret = amdgpu_ras_feature_enable(adev, head, 0); + + /* clean gfx block ras features flag */ + if (adev->ras_enabled && head->block == AMDGPU_RAS_BLOCK__GFX) + con->features &= ~BIT(head->block); } } else ret = amdgpu_ras_feature_enable(adev, head, enable); @@ -890,6 +864,11 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, adev->gmc.xgmi.ras_funcs->query_ras_error_count) adev->gmc.xgmi.ras_funcs->query_ras_error_count(adev, &err_data); break; + case AMDGPU_RAS_BLOCK__HDP: + if (adev->hdp.ras_funcs && + adev->hdp.ras_funcs->query_ras_error_count) + adev->hdp.ras_funcs->query_ras_error_count(adev, &err_data); + break; default: break; } @@ -901,17 +880,42 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, info->ce_count = obj->err_data.ce_count; if (err_data.ce_count) { - dev_info(adev->dev, "%ld correctable hardware errors " + if (adev->smuio.funcs && + adev->smuio.funcs->get_socket_id && + 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", + adev->smuio.funcs->get_socket_id(adev), + adev->smuio.funcs->get_die_id(adev), + obj->err_data.ce_count, + ras_block_str(info->head.block)); + } else { + dev_info(adev->dev, "%ld correctable hardware errors " "detected in %s block, no user " "action is needed.\n", obj->err_data.ce_count, ras_block_str(info->head.block)); + } } if (err_data.ue_count) { - dev_info(adev->dev, "%ld uncorrectable hardware errors " + if (adev->smuio.funcs && + adev->smuio.funcs->get_socket_id && + adev->smuio.funcs->get_die_id) { + dev_info(adev->dev, "socket: %d, die: %d " + "%ld uncorrectable hardware errors " + "detected in %s block\n", + adev->smuio.funcs->get_socket_id(adev), + adev->smuio.funcs->get_die_id(adev), + obj->err_data.ue_count, + ras_block_str(info->head.block)); + } else { + dev_info(adev->dev, "%ld uncorrectable hardware errors " "detected in %s block\n", obj->err_data.ue_count, ras_block_str(info->head.block)); + } } return 0; @@ -937,11 +941,20 @@ int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, if (adev->mmhub.ras_funcs && adev->mmhub.ras_funcs->reset_ras_error_count) adev->mmhub.ras_funcs->reset_ras_error_count(adev); + + if (adev->mmhub.ras_funcs && + adev->mmhub.ras_funcs->reset_ras_error_status) + adev->mmhub.ras_funcs->reset_ras_error_status(adev); break; case AMDGPU_RAS_BLOCK__SDMA: if (adev->sdma.funcs->reset_ras_error_count) adev->sdma.funcs->reset_ras_error_count(adev); break; + case AMDGPU_RAS_BLOCK__HDP: + if (adev->hdp.ras_funcs && + adev->hdp.ras_funcs->reset_ras_error_count) + adev->hdp.ras_funcs->reset_ras_error_count(adev); + break; default: break; } @@ -1022,10 +1035,9 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, ret = -EINVAL; } - amdgpu_ras_parse_status_code(adev, - "inject", - ras_block_str(info->head.block), - (enum ta_ras_status)ret); + if (ret) + dev_err(adev->dev, "ras inject %s failed %d\n", + ras_block_str(info->head.block), ret); return ret; } @@ -1038,7 +1050,7 @@ unsigned long amdgpu_ras_query_error_count(struct amdgpu_device *adev, struct ras_manager *obj; struct ras_err_data data = {0, 0}; - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return 0; list_for_each_entry(obj, &con->head, node) { @@ -1265,8 +1277,8 @@ static int amdgpu_ras_sysfs_remove_all(struct amdgpu_device *adev) static struct dentry *amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); - struct dentry *dir; - struct drm_minor *minor = adev_to_drm(adev)->primary; + struct drm_minor *minor = adev_to_drm(adev)->primary; + struct dentry *dir; dir = debugfs_create_dir(RAS_FS_NAME, minor->debugfs_root); debugfs_create_file("ras_ctrl", S_IWUGO | S_IRUGO, dir, adev, @@ -1275,6 +1287,8 @@ static struct dentry *amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device * &amdgpu_ras_debugfs_eeprom_ops); debugfs_create_u32("bad_page_cnt_threshold", 0444, dir, &con->bad_page_cnt_threshold); + debugfs_create_x32("ras_hw_enabled", 0444, dir, &adev->ras_hw_enabled); + debugfs_create_x32("ras_enabled", 0444, dir, &adev->ras_enabled); /* * After one uncorrectable error happens, usually GPU recovery will @@ -1561,7 +1575,7 @@ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev) struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj; - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return; list_for_each_entry(obj, &con->head, node) { @@ -1611,7 +1625,7 @@ static void amdgpu_ras_query_err_status(struct amdgpu_device *adev) struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj; - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return; list_for_each_entry(obj, &con->head, node) { @@ -1925,7 +1939,7 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev) bool exc_err_limit = false; int ret; - if (adev->ras_features && con) + if (adev->ras_enabled && con) data = &con->eh_data; else return 0; @@ -2029,6 +2043,23 @@ static bool amdgpu_ras_asic_supported(struct amdgpu_device *adev) } /* + * this is workaround for vega20 workstation sku, + * force enable gfx ras, ignore vbios gfx ras flag + * due to GC EDC can not write + */ +static void amdgpu_ras_get_quirks(struct amdgpu_device *adev) +{ + struct atom_context *ctx = adev->mode_info.atom_context; + + if (!ctx) + return; + + if (strnstr(ctx->vbios_version, "D16406", + sizeof(ctx->vbios_version))) + adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__GFX); +} + +/* * check hardware's ras ability which will be saved in hw_supported. * if hardware does not support ras, we can skip some ras initializtion and * forbid some ras operations from IP. @@ -2037,11 +2068,9 @@ static bool amdgpu_ras_asic_supported(struct amdgpu_device *adev) * we have to initialize ras as normal. but need check if operation is * allowed or not in each function. */ -static void amdgpu_ras_check_supported(struct amdgpu_device *adev, - uint32_t *hw_supported, uint32_t *supported) +static void amdgpu_ras_check_supported(struct amdgpu_device *adev) { - *hw_supported = 0; - *supported = 0; + adev->ras_hw_enabled = adev->ras_enabled = 0; if (amdgpu_sriov_vf(adev) || !adev->is_atom_fw || !amdgpu_ras_asic_supported(adev)) @@ -2050,33 +2079,34 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev, if (!adev->gmc.xgmi.connected_to_cpu) { if (amdgpu_atomfirmware_mem_ecc_supported(adev)) { dev_info(adev->dev, "MEM ECC is active.\n"); - *hw_supported |= (1 << AMDGPU_RAS_BLOCK__UMC | - 1 << AMDGPU_RAS_BLOCK__DF); + adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__UMC | + 1 << AMDGPU_RAS_BLOCK__DF); } else { dev_info(adev->dev, "MEM ECC is not presented.\n"); } if (amdgpu_atomfirmware_sram_ecc_supported(adev)) { dev_info(adev->dev, "SRAM ECC is active.\n"); - *hw_supported |= ~(1 << AMDGPU_RAS_BLOCK__UMC | - 1 << AMDGPU_RAS_BLOCK__DF); + adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC | + 1 << AMDGPU_RAS_BLOCK__DF); } else { dev_info(adev->dev, "SRAM ECC is not presented.\n"); } } else { /* driver only manages a few IP blocks RAS feature * when GPU is connected cpu through XGMI */ - *hw_supported |= (1 << AMDGPU_RAS_BLOCK__GFX | - 1 << AMDGPU_RAS_BLOCK__SDMA | - 1 << AMDGPU_RAS_BLOCK__MMHUB); + adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__GFX | + 1 << AMDGPU_RAS_BLOCK__SDMA | + 1 << AMDGPU_RAS_BLOCK__MMHUB); } + amdgpu_ras_get_quirks(adev); + /* hw_supported needs to be aligned with RAS block mask. */ - *hw_supported &= AMDGPU_RAS_BLOCK_MASK; + adev->ras_hw_enabled &= AMDGPU_RAS_BLOCK_MASK; - *supported = amdgpu_ras_enable == 0 ? - 0 : *hw_supported & amdgpu_ras_mask; - adev->ras_features = *supported; + adev->ras_enabled = amdgpu_ras_enable == 0 ? 0 : + adev->ras_hw_enabled & amdgpu_ras_mask; } int amdgpu_ras_init(struct amdgpu_device *adev) @@ -2097,13 +2127,13 @@ int amdgpu_ras_init(struct amdgpu_device *adev) amdgpu_ras_set_context(adev, con); - amdgpu_ras_check_supported(adev, &con->hw_supported, - &con->supported); - if (!con->hw_supported || (adev->asic_type == CHIP_VEGA10)) { + amdgpu_ras_check_supported(adev); + + if (!adev->ras_enabled || adev->asic_type == CHIP_VEGA10) { /* set gfx block ras context feature for VEGA20 Gaming * send ras disable cmd to ras ta during ras late init. */ - if (!adev->ras_features && adev->asic_type == CHIP_VEGA20) { + if (!adev->ras_enabled && adev->asic_type == CHIP_VEGA20) { con->features |= BIT(AMDGPU_RAS_BLOCK__GFX); return 0; @@ -2153,8 +2183,9 @@ int amdgpu_ras_init(struct amdgpu_device *adev) } dev_info(adev->dev, "RAS INFO: ras initialized successfully, " - "hardware ability[%x] ras_mask[%x]\n", - con->hw_supported, con->supported); + "hardware ability[%x] ras_mask[%x]\n", + adev->ras_hw_enabled, adev->ras_enabled); + return 0; release_con: amdgpu_ras_set_context(adev, NULL); @@ -2163,7 +2194,7 @@ release_con: return r; } -static int amdgpu_persistent_edc_harvesting_supported(struct amdgpu_device *adev) +int amdgpu_persistent_edc_harvesting_supported(struct amdgpu_device *adev) { if (adev->gmc.xgmi.connected_to_cpu) return 1; @@ -2268,7 +2299,7 @@ void amdgpu_ras_resume(struct amdgpu_device *adev) struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj, *tmp; - if (!adev->ras_features || !con) { + if (!adev->ras_enabled || !con) { /* clean ras context for VEGA20 Gaming after send ras disable cmd */ amdgpu_release_ras_context(adev); @@ -2314,7 +2345,7 @@ void amdgpu_ras_suspend(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return; amdgpu_ras_disable_all_features(adev, 0); @@ -2328,7 +2359,7 @@ int amdgpu_ras_pre_fini(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return 0; /* Need disable ras on all IPs here before ip [hw/sw]fini */ @@ -2341,7 +2372,7 @@ int amdgpu_ras_fini(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return 0; amdgpu_ras_fs_fini(adev); @@ -2360,10 +2391,8 @@ int amdgpu_ras_fini(struct amdgpu_device *adev) void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev) { - uint32_t hw_supported, supported; - - amdgpu_ras_check_supported(adev, &hw_supported, &supported); - if (!hw_supported) + amdgpu_ras_check_supported(adev); + if (!adev->ras_hw_enabled) return; if (atomic_cmpxchg(&amdgpu_ras_in_intr, 0, 1) == 0) { @@ -2392,7 +2421,7 @@ void amdgpu_release_ras_context(struct amdgpu_device *adev) if (!con) return; - if (!adev->ras_features && con->features & BIT(AMDGPU_RAS_BLOCK__GFX)) { + if (!adev->ras_enabled && con->features & BIT(AMDGPU_RAS_BLOCK__GFX)) { con->features &= ~BIT(AMDGPU_RAS_BLOCK__GFX); amdgpu_ras_set_context(adev, NULL); kfree(con); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 60df268a0c66..bfa40c8ecc94 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -313,9 +313,6 @@ struct ras_common_if { struct amdgpu_ras { /* ras infrastructure */ /* for ras itself. */ - uint32_t hw_supported; - /* for IP to check its ras ability. */ - uint32_t supported; uint32_t features; struct list_head head; /* sysfs */ @@ -478,7 +475,7 @@ static inline int amdgpu_ras_is_supported(struct amdgpu_device *adev, if (block >= AMDGPU_RAS_BLOCK_COUNT) return 0; - return ras && (ras->supported & (1 << block)); + return ras && (adev->ras_enabled & (1 << block)); } int amdgpu_ras_recovery_init(struct amdgpu_device *adev); @@ -628,4 +625,7 @@ void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready); bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev); void amdgpu_release_ras_context(struct amdgpu_device *adev); + +int amdgpu_persistent_edc_harvesting_supported(struct amdgpu_device *adev); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 688624ebe421..7b634a1517f9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -158,6 +158,7 @@ void amdgpu_ring_undo(struct amdgpu_ring *ring) * @irq_src: interrupt source to use for this ring * @irq_type: interrupt type to use for this ring * @hw_prio: ring priority (NORMAL/HIGH) + * @sched_score: optional score atomic shared with other schedulers * * Initialize the driver information for the selected ring (all asics). * Returns 0 on success, error on failure. diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c index 5369c8dd0764..123453999093 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c @@ -86,6 +86,8 @@ void psp_prep_securedisplay_cmd_buf(struct psp_context *psp, struct securedispla (*cmd)->cmd_id = command_id; } +#if defined(CONFIG_DEBUG_FS) + static ssize_t amdgpu_securedisplay_debugfs_write(struct file *f, const char __user *buf, size_t size, loff_t *pos) { @@ -162,6 +164,8 @@ static const struct file_operations amdgpu_securedisplay_debugfs_ops = { .llseek = default_llseek }; +#endif + void amdgpu_securedisplay_debugfs_init(struct amdgpu_device *adev) { #if defined(CONFIG_DEBUG_FS) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h index b860ec913ac5..484bb3dcec47 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h @@ -29,6 +29,7 @@ struct amdgpu_smuio_funcs { void (*update_rom_clock_gating)(struct amdgpu_device *adev, bool enable); void (*get_clock_gating_state)(struct amdgpu_device *adev, u32 *flags); u32 (*get_die_id)(struct amdgpu_device *adev); + u32 (*get_socket_id)(struct amdgpu_device *adev); bool (*is_host_gpu_xgmi_supported)(struct amdgpu_device *adev); }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 3bef0432cac2..e039f5b8bc87 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -32,7 +32,6 @@ #include <linux/dma-mapping.h> #include <linux/iommu.h> -#include <linux/hmm.h> #include <linux/pagemap.h> #include <linux/sched/task.h> #include <linux/sched/mm.h> @@ -112,6 +111,20 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, } abo = ttm_to_amdgpu_bo(bo); + if (abo->flags & AMDGPU_AMDKFD_CREATE_SVM_BO) { + struct dma_fence *fence; + struct dma_resv *resv = &bo->base._resv; + + rcu_read_lock(); + fence = rcu_dereference(resv->fence_excl); + if (fence && !fence->ops->signaled) + dma_fence_enable_sw_signaling(fence); + + placement->num_placement = 0; + placement->num_busy_placement = 0; + rcu_read_unlock(); + return; + } switch (bo->mem.mem_type) { case AMDGPU_PL_GDS: case AMDGPU_PL_GWS: @@ -165,13 +178,6 @@ static int amdgpu_verify_access(struct ttm_buffer_object *bo, struct file *filp) { struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo); - /* - * Don't verify access for KFD BOs. They don't have a GEM - * object associated with them. - */ - if (abo->kfd_bo) - return 0; - if (amdgpu_ttm_tt_get_usermm(bo->ttm)) return -EPERM; return drm_vma_node_verify_access(&abo->tbo.base.vma_node, @@ -225,7 +231,7 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo, *addr += mm_cur->start & ~PAGE_MASK; num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8); - num_bytes = num_pages * 8; + num_bytes = num_pages * 8 * AMDGPU_GPU_PAGES_IN_CPU_PAGE; r = amdgpu_job_alloc_with_ib(adev, num_dw * 4 + num_bytes, AMDGPU_IB_POOL_DELAYED, &job); @@ -288,7 +294,7 @@ error_free: } /** - * amdgpu_copy_ttm_mem_to_mem - Helper function for copy + * amdgpu_ttm_copy_mem_to_mem - Helper function for copy * @adev: amdgpu device * @src: buffer/address where to read from * @dst: buffer/address where to write to @@ -570,10 +576,10 @@ out: * * Called by ttm_mem_io_reserve() ultimately via ttm_bo_vm_fault() */ -static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev, struct ttm_resource *mem) +static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev, + struct ttm_resource *mem) { struct amdgpu_device *adev = amdgpu_ttm_adev(bdev); - struct drm_mm_node *mm_node = mem->mm_node; size_t bus_size = (size_t)mem->num_pages << PAGE_SHIFT; switch (mem->mem_type) { @@ -587,12 +593,9 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev, struct ttm_resourc /* check if it's visible */ if ((mem->bus.offset + bus_size) > adev->gmc.visible_vram_size) return -EINVAL; - /* Only physically contiguous buffers apply. In a contiguous - * buffer, size of the first mm_node would match the number of - * pages in ttm_resource. - */ + if (adev->mman.aper_base_kaddr && - (mm_node->size == mem->num_pages)) + mem->placement & TTM_PL_FLAG_CONTIGUOUS) mem->bus.addr = (u8 *)adev->mman.aper_base_kaddr + mem->bus.offset; @@ -670,10 +673,8 @@ int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages) struct amdgpu_ttm_tt *gtt = (void *)ttm; unsigned long start = gtt->userptr; struct vm_area_struct *vma; - struct hmm_range *range; - unsigned long timeout; struct mm_struct *mm; - unsigned long i; + bool readonly; int r = 0; mm = bo->notifier.mm; @@ -689,76 +690,26 @@ int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages) if (!mmget_not_zero(mm)) /* Happens during process shutdown */ return -ESRCH; - range = kzalloc(sizeof(*range), GFP_KERNEL); - if (unlikely(!range)) { - r = -ENOMEM; - goto out; - } - range->notifier = &bo->notifier; - range->start = bo->notifier.interval_tree.start; - range->end = bo->notifier.interval_tree.last + 1; - range->default_flags = HMM_PFN_REQ_FAULT; - if (!amdgpu_ttm_tt_is_readonly(ttm)) - range->default_flags |= HMM_PFN_REQ_WRITE; - - range->hmm_pfns = kvmalloc_array(ttm->num_pages, - sizeof(*range->hmm_pfns), GFP_KERNEL); - if (unlikely(!range->hmm_pfns)) { - r = -ENOMEM; - goto out_free_ranges; - } - mmap_read_lock(mm); vma = find_vma(mm, start); + mmap_read_unlock(mm); if (unlikely(!vma || start < vma->vm_start)) { r = -EFAULT; - goto out_unlock; + goto out_putmm; } if (unlikely((gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) && vma->vm_file)) { r = -EPERM; - goto out_unlock; - } - mmap_read_unlock(mm); - timeout = jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT); - -retry: - range->notifier_seq = mmu_interval_read_begin(&bo->notifier); - - mmap_read_lock(mm); - r = hmm_range_fault(range); - mmap_read_unlock(mm); - if (unlikely(r)) { - /* - * FIXME: This timeout should encompass the retry from - * mmu_interval_read_retry() as well. - */ - if (r == -EBUSY && !time_after(jiffies, timeout)) - goto retry; - goto out_free_pfns; + goto out_putmm; } - /* - * Due to default_flags, all pages are HMM_PFN_VALID or - * hmm_range_fault() fails. FIXME: The pages cannot be touched outside - * the notifier_lock, and mmu_interval_read_retry() must be done first. - */ - for (i = 0; i < ttm->num_pages; i++) - pages[i] = hmm_pfn_to_page(range->hmm_pfns[i]); - - gtt->range = range; + readonly = amdgpu_ttm_tt_is_readonly(ttm); + r = amdgpu_hmm_range_get_pages(&bo->notifier, mm, pages, start, + ttm->num_pages, >t->range, readonly, + false); +out_putmm: mmput(mm); - return 0; - -out_unlock: - mmap_read_unlock(mm); -out_free_pfns: - kvfree(range->hmm_pfns); -out_free_ranges: - kfree(range); -out: - mmput(mm); return r; } @@ -787,10 +738,7 @@ bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm) * FIXME: Must always hold notifier_lock for this, and must * not ignore the return code. */ - r = mmu_interval_read_retry(gtt->range->notifier, - gtt->range->notifier_seq); - kvfree(gtt->range->hmm_pfns); - kfree(gtt->range); + r = amdgpu_hmm_range_get_pages_done(gtt->range); gtt->range = NULL; } @@ -959,7 +907,23 @@ static int amdgpu_ttm_backend_bind(struct ttm_device *bdev, DRM_ERROR("failed to pin userptr\n"); return r; } + } else if (ttm->page_flags & TTM_PAGE_FLAG_SG) { + if (!ttm->sg) { + struct dma_buf_attachment *attach; + struct sg_table *sgt; + + attach = gtt->gobj->import_attach; + sgt = dma_buf_map_attachment(attach, DMA_BIDIRECTIONAL); + if (IS_ERR(sgt)) + return PTR_ERR(sgt); + + ttm->sg = sgt; + } + + drm_prime_sg_to_dma_addr_array(ttm->sg, gtt->ttm.dma_address, + ttm->num_pages); } + if (!ttm->num_pages) { WARN(1, "nothing to bind %u pages for mreg %p back %p!\n", ttm->num_pages, bo_mem, ttm); @@ -1018,8 +982,6 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo) } else { /* allocate GART space */ - tmp = bo->mem; - tmp.mm_node = NULL; placement.num_placement = 1; placement.placement = &placements; placement.num_busy_placement = 1; @@ -1086,8 +1048,15 @@ static void amdgpu_ttm_backend_unbind(struct ttm_device *bdev, int r; /* if the pages have userptr pinning then clear that first */ - if (gtt->userptr) + if (gtt->userptr) { amdgpu_ttm_tt_unpin_userptr(bdev, ttm); + } else if (ttm->sg && gtt->gobj->import_attach) { + struct dma_buf_attachment *attach; + + attach = gtt->gobj->import_attach; + dma_buf_unmap_attachment(attach, ttm->sg, DMA_BIDIRECTIONAL); + ttm->sg = NULL; + } if (!gtt->bound) return; @@ -1174,23 +1143,8 @@ static int amdgpu_ttm_tt_populate(struct ttm_device *bdev, return 0; } - if (ttm->page_flags & TTM_PAGE_FLAG_SG) { - if (!ttm->sg) { - struct dma_buf_attachment *attach; - struct sg_table *sgt; - - attach = gtt->gobj->import_attach; - sgt = dma_buf_map_attachment(attach, DMA_BIDIRECTIONAL); - if (IS_ERR(sgt)) - return PTR_ERR(sgt); - - ttm->sg = sgt; - } - - drm_prime_sg_to_dma_addr_array(ttm->sg, gtt->ttm.dma_address, - ttm->num_pages); + if (ttm->page_flags & TTM_PAGE_FLAG_SG) return 0; - } return ttm_pool_alloc(&adev->mman.bdev.pool, ttm, ctx); } @@ -1210,16 +1164,8 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_device *bdev, if (gtt && gtt->userptr) { amdgpu_ttm_tt_set_user_pages(ttm, NULL); kfree(ttm->sg); - ttm->page_flags &= ~TTM_PAGE_FLAG_SG; - return; - } - - if (ttm->sg && gtt->gobj->import_attach) { - struct dma_buf_attachment *attach; - - attach = gtt->gobj->import_attach; - dma_buf_unmap_attachment(attach, ttm->sg, DMA_BIDIRECTIONAL); ttm->sg = NULL; + ttm->page_flags &= ~TTM_PAGE_FLAG_SG; return; } @@ -1632,11 +1578,8 @@ static int amdgpu_ttm_reserve_tmr(struct amdgpu_device *adev) bool mem_train_support = false; if (!amdgpu_sriov_vf(adev)) { - ret = amdgpu_mem_train_support(adev); - if (ret == 1) + if (amdgpu_atomfirmware_mem_training_supported(adev)) mem_train_support = true; - else if (ret == -1) - return -EINVAL; else DRM_DEBUG("memory training does not support!\n"); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index 9e38475e0f8d..eb84a69c4b74 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -37,11 +37,6 @@ #define AMDGPU_POISON 0xd0bed0be -struct amdgpu_vram_reservation { - struct list_head node; - struct drm_mm_node mm_node; -}; - struct amdgpu_vram_mgr { struct ttm_resource_manager manager; struct drm_mm mm; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index 9733224117e3..8b502ffd58dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -403,6 +403,7 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type) case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: case CHIP_ALDEBARAN: + case CHIP_BEIGE_GOBY: if (!load_type) return AMDGPU_FW_LOAD_DIRECT; else diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c index 201645963ba5..d101cb9697a2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c @@ -48,6 +48,7 @@ #define FIRMWARE_VANGOGH "amdgpu/vangogh_vcn.bin" #define FIRMWARE_DIMGREY_CAVEFISH "amdgpu/dimgrey_cavefish_vcn.bin" #define FIRMWARE_ALDEBARAN "amdgpu/aldebaran_vcn.bin" +#define FIRMWARE_BEIGE_GOBY "amdgpu/beige_goby_vcn.bin" MODULE_FIRMWARE(FIRMWARE_RAVEN); MODULE_FIRMWARE(FIRMWARE_PICASSO); @@ -63,6 +64,7 @@ MODULE_FIRMWARE(FIRMWARE_SIENNA_CICHLID); MODULE_FIRMWARE(FIRMWARE_NAVY_FLOUNDER); MODULE_FIRMWARE(FIRMWARE_VANGOGH); MODULE_FIRMWARE(FIRMWARE_DIMGREY_CAVEFISH); +MODULE_FIRMWARE(FIRMWARE_BEIGE_GOBY); static void amdgpu_vcn_idle_work_handler(struct work_struct *work); @@ -151,6 +153,12 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev) (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)) adev->vcn.indirect_sram = true; break; + case CHIP_BEIGE_GOBY: + fw_name = FIRMWARE_BEIGE_GOBY; + if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) && + (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)) + adev->vcn.indirect_sram = true; + break; default: return -EINVAL; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c index 0c9c5255aa42..b71dd1deeb2d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c @@ -50,9 +50,12 @@ void amdgpu_virt_init_setting(struct amdgpu_device *adev) struct drm_device *ddev = adev_to_drm(adev); /* enable virtual display */ - if (adev->mode_info.num_crtc == 0) - adev->mode_info.num_crtc = 1; - adev->enable_virtual_display = true; + if (adev->asic_type != CHIP_ALDEBARAN && + adev->asic_type != CHIP_ARCTURUS) { + if (adev->mode_info.num_crtc == 0) + adev->mode_info.num_crtc = 1; + adev->enable_virtual_display = true; + } ddev->driver_features &= ~DRIVER_ATOMIC; adev->cg_flags = 0; adev->pg_flags = 0; @@ -429,6 +432,9 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev) uint32_t checksum; uint32_t checkval; + uint32_t i; + uint32_t tmp; + if (adev->virt.fw_reserve.p_pf2vf == NULL) return -EINVAL; @@ -469,6 +475,29 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev) adev->virt.reg_access = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->reg_access_flags.all; + adev->virt.decode_max_dimension_pixels = 0; + adev->virt.decode_max_frame_pixels = 0; + adev->virt.encode_max_dimension_pixels = 0; + adev->virt.encode_max_frame_pixels = 0; + adev->virt.is_mm_bw_enabled = false; + for (i = 0; i < AMD_SRIOV_MSG_RESERVE_VCN_INST; i++) { + tmp = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->mm_bw_management[i].decode_max_dimension_pixels; + adev->virt.decode_max_dimension_pixels = max(tmp, adev->virt.decode_max_dimension_pixels); + + tmp = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->mm_bw_management[i].decode_max_frame_pixels; + adev->virt.decode_max_frame_pixels = max(tmp, adev->virt.decode_max_frame_pixels); + + tmp = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->mm_bw_management[i].encode_max_dimension_pixels; + adev->virt.encode_max_dimension_pixels = max(tmp, adev->virt.encode_max_dimension_pixels); + + tmp = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->mm_bw_management[i].encode_max_frame_pixels; + adev->virt.encode_max_frame_pixels = max(tmp, adev->virt.encode_max_frame_pixels); + } + if((adev->virt.decode_max_dimension_pixels > 0) || (adev->virt.encode_max_dimension_pixels > 0)) + adev->virt.is_mm_bw_enabled = true; + + adev->unique_id = + ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->uuid; break; default: DRM_ERROR("invalid pf2vf version\n"); @@ -679,6 +708,7 @@ void amdgpu_detect_virtualization(struct amdgpu_device *adev) case CHIP_VEGA10: case CHIP_VEGA20: case CHIP_ARCTURUS: + case CHIP_ALDEBARAN: soc15_set_virt_ops(adev); break; case CHIP_NAVI10: @@ -740,3 +770,35 @@ enum amdgpu_sriov_vf_mode amdgpu_virt_get_sriov_vf_mode(struct amdgpu_device *ad return mode; } + +void amdgpu_virt_update_sriov_video_codec(struct amdgpu_device *adev, + struct amdgpu_video_codec_info *encode, uint32_t encode_array_size, + struct amdgpu_video_codec_info *decode, uint32_t decode_array_size) +{ + uint32_t i; + + if (!adev->virt.is_mm_bw_enabled) + return; + + if (encode) { + for (i = 0; i < encode_array_size; i++) { + encode[i].max_width = adev->virt.encode_max_dimension_pixels; + encode[i].max_pixels_per_frame = adev->virt.encode_max_frame_pixels; + if (encode[i].max_width > 0) + encode[i].max_height = encode[i].max_pixels_per_frame / encode[i].max_width; + else + encode[i].max_height = 0; + } + } + + if (decode) { + for (i = 0; i < decode_array_size; i++) { + decode[i].max_width = adev->virt.decode_max_dimension_pixels; + decode[i].max_pixels_per_frame = adev->virt.decode_max_frame_pixels; + if (decode[i].max_width > 0) + decode[i].max_height = decode[i].max_pixels_per_frame / decode[i].max_width; + else + decode[i].max_height = 0; + } + } +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index 383d4bdc3fb5..8d4c20bb71c5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -233,8 +233,17 @@ struct amdgpu_virt { /* vf2pf message */ struct delayed_work vf2pf_work; uint32_t vf2pf_update_interval_ms; + + /* multimedia bandwidth config */ + bool is_mm_bw_enabled; + uint32_t decode_max_dimension_pixels; + uint32_t decode_max_frame_pixels; + uint32_t encode_max_dimension_pixels; + uint32_t encode_max_frame_pixels; }; +struct amdgpu_video_codec_info; + #define amdgpu_sriov_enabled(adev) \ ((adev)->virt.caps & AMDGPU_SRIOV_CAPS_ENABLE_IOV) @@ -307,4 +316,8 @@ int amdgpu_virt_enable_access_debugfs(struct amdgpu_device *adev); void amdgpu_virt_disable_access_debugfs(struct amdgpu_device *adev); enum amdgpu_sriov_vf_mode amdgpu_virt_get_sriov_vf_mode(struct amdgpu_device *adev); + +void amdgpu_virt_update_sriov_video_codec(struct amdgpu_device *adev, + struct amdgpu_video_codec_info *encode, uint32_t encode_array_size, + struct amdgpu_video_codec_info *decode, uint32_t decode_array_size); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 9acee4a5b2ba..55991f393481 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -25,6 +25,7 @@ * Alex Deucher * Jerome Glisse */ + #include <linux/dma-fence-array.h> #include <linux/interval_tree_generic.h> #include <linux/idr.h> @@ -37,6 +38,8 @@ #include "amdgpu_gmc.h" #include "amdgpu_xgmi.h" #include "amdgpu_dma_buf.h" +#include "amdgpu_res_cursor.h" +#include "kfd_svm.h" /** * DOC: GPUVM @@ -849,35 +852,60 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, } /** - * amdgpu_vm_bo_param - fill in parameters for PD/PT allocation + * amdgpu_vm_pt_create - create bo for PD/PT * * @adev: amdgpu_device pointer * @vm: requesting vm * @level: the page table level * @immediate: use a immediate update - * @bp: resulting BO allocation parameters + * @bo: pointer to the buffer object pointer */ -static void amdgpu_vm_bo_param(struct amdgpu_device *adev, struct amdgpu_vm *vm, +static int amdgpu_vm_pt_create(struct amdgpu_device *adev, + struct amdgpu_vm *vm, int level, bool immediate, - struct amdgpu_bo_param *bp) + struct amdgpu_bo **bo) { - memset(bp, 0, sizeof(*bp)); + struct amdgpu_bo_param bp; + int r; + + memset(&bp, 0, sizeof(bp)); - bp->size = amdgpu_vm_bo_size(adev, level); - bp->byte_align = AMDGPU_GPU_PAGE_SIZE; - bp->domain = AMDGPU_GEM_DOMAIN_VRAM; - bp->domain = amdgpu_bo_get_preferred_pin_domain(adev, bp->domain); - bp->flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | + bp.size = amdgpu_vm_bo_size(adev, level); + bp.byte_align = AMDGPU_GPU_PAGE_SIZE; + bp.domain = AMDGPU_GEM_DOMAIN_VRAM; + bp.domain = amdgpu_bo_get_preferred_pin_domain(adev, bp.domain); + bp.flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | AMDGPU_GEM_CREATE_CPU_GTT_USWC; - bp->bo_ptr_size = sizeof(struct amdgpu_bo); + bp.bo_ptr_size = sizeof(struct amdgpu_bo); if (vm->use_cpu_for_update) - bp->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; - else if (!vm->root.base.bo || vm->root.base.bo->shadow) - bp->flags |= AMDGPU_GEM_CREATE_SHADOW; - bp->type = ttm_bo_type_kernel; - bp->no_wait_gpu = immediate; + bp.flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; + + bp.type = ttm_bo_type_kernel; + bp.no_wait_gpu = immediate; if (vm->root.base.bo) - bp->resv = vm->root.base.bo->tbo.base.resv; + bp.resv = vm->root.base.bo->tbo.base.resv; + + r = amdgpu_bo_create(adev, &bp, bo); + if (r) + return r; + + if (vm->is_compute_context && (adev->flags & AMD_IS_APU)) + return 0; + + if (!bp.resv) + WARN_ON(dma_resv_lock((*bo)->tbo.base.resv, + NULL)); + r = amdgpu_bo_create_shadow(adev, bp.size, *bo); + + if (!bp.resv) + dma_resv_unlock((*bo)->tbo.base.resv); + + if (r) { + amdgpu_bo_unref(bo); + return r; + } + + return 0; } /** @@ -900,7 +928,6 @@ static int amdgpu_vm_alloc_pts(struct amdgpu_device *adev, bool immediate) { struct amdgpu_vm_pt *entry = cursor->entry; - struct amdgpu_bo_param bp; struct amdgpu_bo *pt; int r; @@ -918,9 +945,7 @@ static int amdgpu_vm_alloc_pts(struct amdgpu_device *adev, if (entry->base.bo) return 0; - amdgpu_vm_bo_param(adev, vm, cursor->level, immediate, &bp); - - r = amdgpu_bo_create(adev, &bp, &pt); + r = amdgpu_vm_pt_create(adev, vm, cursor->level, immediate, &pt); if (r) return r; @@ -1559,6 +1584,7 @@ static int amdgpu_vm_update_ptes(struct amdgpu_vm_update_params *params, while (cursor.pfn < frag_start) { amdgpu_vm_free_pts(adev, params->vm, &cursor); amdgpu_vm_pt_next(adev, &cursor); + params->table_freed = true; } } else if (frag >= shift) { @@ -1583,28 +1609,30 @@ static int amdgpu_vm_update_ptes(struct amdgpu_vm_update_params *params, * @last: last mapped entry * @flags: flags for the entries * @offset: offset into nodes and pages_addr - * @nodes: array of drm_mm_nodes with the MC addresses + * @res: ttm_resource to map * @pages_addr: DMA addresses to use for mapping * @fence: optional resulting fence + * @table_freed: return true if page table is freed * * Fill in the page table entries between @start and @last. * * Returns: * 0 for success, -EINVAL for failure. */ -static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, - struct amdgpu_device *bo_adev, - struct amdgpu_vm *vm, bool immediate, - bool unlocked, struct dma_resv *resv, - uint64_t start, uint64_t last, - uint64_t flags, uint64_t offset, - struct drm_mm_node *nodes, - dma_addr_t *pages_addr, - struct dma_fence **fence) +int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, + struct amdgpu_device *bo_adev, + struct amdgpu_vm *vm, bool immediate, + bool unlocked, struct dma_resv *resv, + uint64_t start, uint64_t last, + uint64_t flags, uint64_t offset, + struct ttm_resource *res, + dma_addr_t *pages_addr, + struct dma_fence **fence, + bool *table_freed) { struct amdgpu_vm_update_params params; + struct amdgpu_res_cursor cursor; enum amdgpu_sync_mode sync_mode; - uint64_t pfn; int r; memset(¶ms, 0, sizeof(params)); @@ -1622,14 +1650,6 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, else sync_mode = AMDGPU_SYNC_EXPLICIT; - pfn = offset >> PAGE_SHIFT; - if (nodes) { - while (pfn >= nodes->size) { - pfn -= nodes->size; - ++nodes; - } - } - amdgpu_vm_eviction_lock(vm); if (vm->evicting) { r = -EBUSY; @@ -1648,23 +1668,17 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (r) goto error_unlock; - do { + amdgpu_res_first(res, offset, (last - start + 1) * AMDGPU_GPU_PAGE_SIZE, + &cursor); + while (cursor.remaining) { uint64_t tmp, num_entries, addr; - - num_entries = last - start + 1; - if (nodes) { - addr = nodes->start << PAGE_SHIFT; - num_entries = min((nodes->size - pfn) * - AMDGPU_GPU_PAGES_IN_CPU_PAGE, num_entries); - } else { - addr = 0; - } - + num_entries = cursor.size >> AMDGPU_GPU_PAGE_SHIFT; if (pages_addr) { bool contiguous = true; if (num_entries > AMDGPU_GPU_PAGES_IN_CPU_PAGE) { + uint64_t pfn = cursor.start >> PAGE_SHIFT; uint64_t count; contiguous = pages_addr[pfn + 1] == @@ -1684,16 +1698,18 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, } if (!contiguous) { - addr = pfn << PAGE_SHIFT; + addr = cursor.start; params.pages_addr = pages_addr; } else { - addr = pages_addr[pfn]; + addr = pages_addr[cursor.start >> PAGE_SHIFT]; params.pages_addr = NULL; } } else if (flags & (AMDGPU_PTE_VALID | AMDGPU_PTE_PRT)) { - addr += bo_adev->vm_manager.vram_base_offset; - addr += pfn << PAGE_SHIFT; + addr = bo_adev->vm_manager.vram_base_offset + + cursor.start; + } else { + addr = 0; } tmp = start + num_entries; @@ -1701,22 +1717,64 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (r) goto error_unlock; - pfn += num_entries / AMDGPU_GPU_PAGES_IN_CPU_PAGE; - if (nodes && nodes->size == pfn) { - pfn = 0; - ++nodes; - } + amdgpu_res_next(&cursor, num_entries * AMDGPU_GPU_PAGE_SIZE); start = tmp; - - } while (unlikely(start != last + 1)); + }; r = vm->update_funcs->commit(¶ms, fence); + if (table_freed) + *table_freed = params.table_freed; + error_unlock: amdgpu_vm_eviction_unlock(vm); return r; } +void amdgpu_vm_get_memory(struct amdgpu_vm *vm, uint64_t *vram_mem, + uint64_t *gtt_mem, uint64_t *cpu_mem) +{ + struct amdgpu_bo_va *bo_va, *tmp; + + list_for_each_entry_safe(bo_va, tmp, &vm->idle, base.vm_status) { + if (!bo_va->base.bo) + continue; + amdgpu_bo_get_memory(bo_va->base.bo, vram_mem, + gtt_mem, cpu_mem); + } + list_for_each_entry_safe(bo_va, tmp, &vm->evicted, base.vm_status) { + if (!bo_va->base.bo) + continue; + amdgpu_bo_get_memory(bo_va->base.bo, vram_mem, + gtt_mem, cpu_mem); + } + list_for_each_entry_safe(bo_va, tmp, &vm->relocated, base.vm_status) { + if (!bo_va->base.bo) + continue; + amdgpu_bo_get_memory(bo_va->base.bo, vram_mem, + gtt_mem, cpu_mem); + } + list_for_each_entry_safe(bo_va, tmp, &vm->moved, base.vm_status) { + if (!bo_va->base.bo) + continue; + amdgpu_bo_get_memory(bo_va->base.bo, vram_mem, + gtt_mem, cpu_mem); + } + spin_lock(&vm->invalidated_lock); + list_for_each_entry_safe(bo_va, tmp, &vm->invalidated, base.vm_status) { + if (!bo_va->base.bo) + continue; + amdgpu_bo_get_memory(bo_va->base.bo, vram_mem, + gtt_mem, cpu_mem); + } + list_for_each_entry_safe(bo_va, tmp, &vm->done, base.vm_status) { + if (!bo_va->base.bo) + continue; + amdgpu_bo_get_memory(bo_va->base.bo, vram_mem, + gtt_mem, cpu_mem); + } + spin_unlock(&vm->invalidated_lock); +} /** * amdgpu_vm_bo_update - update all BO mappings in the vm page table * @@ -1737,7 +1795,6 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, struct amdgpu_bo_va_mapping *mapping; dma_addr_t *pages_addr = NULL; struct ttm_resource *mem; - struct drm_mm_node *nodes; struct dma_fence **last_update; struct dma_resv *resv; uint64_t flags; @@ -1746,7 +1803,6 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, if (clear || !bo) { mem = NULL; - nodes = NULL; resv = vm->root.base.bo->tbo.base.resv; } else { struct drm_gem_object *obj = &bo->tbo.base; @@ -1761,7 +1817,6 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, bo = gem_to_amdgpu_bo(gobj); } mem = &bo->tbo.mem; - nodes = mem->mm_node; if (mem->mem_type == TTM_PL_TT) pages_addr = bo->tbo.ttm->dma_address; } @@ -1810,8 +1865,8 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, r = amdgpu_vm_bo_update_mapping(adev, bo_adev, vm, false, false, resv, mapping->start, mapping->last, update_flags, - mapping->offset, nodes, - pages_addr, last_update); + mapping->offset, mem, + pages_addr, last_update, NULL); if (r) return r; } @@ -2022,7 +2077,7 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, r = amdgpu_vm_bo_update_mapping(adev, adev, vm, false, false, resv, mapping->start, mapping->last, init_pte_value, - 0, NULL, NULL, &f); + 0, NULL, NULL, &f, NULL); amdgpu_vm_free_mapping(adev, vm, mapping, f); if (r) { dma_fence_put(f); @@ -2773,7 +2828,6 @@ long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout) * * @adev: amdgpu_device pointer * @vm: requested vm - * @vm_context: Indicates if it GFX or Compute context * @pasid: Process address space identifier * * Init @vm fields. @@ -2781,10 +2835,8 @@ long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout) * Returns: * 0 for success, error for failure. */ -int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, - int vm_context, u32 pasid) +int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, u32 pasid) { - struct amdgpu_bo_param bp; struct amdgpu_bo *root; int r, i; @@ -2816,16 +2868,9 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, vm->pte_support_ats = false; vm->is_compute_context = false; - if (vm_context == AMDGPU_VM_CONTEXT_COMPUTE) { - vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode & - AMDGPU_VM_USE_CPU_FOR_COMPUTE); + vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode & + AMDGPU_VM_USE_CPU_FOR_GFX); - if (adev->asic_type == CHIP_RAVEN) - vm->pte_support_ats = true; - } else { - vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode & - AMDGPU_VM_USE_CPU_FOR_GFX); - } DRM_DEBUG_DRIVER("VM update mode is %s\n", vm->use_cpu_for_update ? "CPU" : "SDMA"); WARN_ONCE((vm->use_cpu_for_update && @@ -2842,10 +2887,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, mutex_init(&vm->eviction_lock); vm->evicting = false; - amdgpu_vm_bo_param(adev, vm, adev->vm_manager.root_level, false, &bp); - if (vm_context == AMDGPU_VM_CONTEXT_COMPUTE) - bp.flags &= ~AMDGPU_GEM_CREATE_SHADOW; - r = amdgpu_bo_create(adev, &bp, &root); + r = amdgpu_vm_pt_create(adev, vm, adev->vm_manager.root_level, + false, &root); if (r) goto error_free_delayed; @@ -3304,6 +3347,7 @@ void amdgpu_vm_set_task_info(struct amdgpu_vm *vm) bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid, uint64_t addr) { + bool is_compute_context = false; struct amdgpu_bo *root; uint64_t value, flags; struct amdgpu_vm *vm; @@ -3311,15 +3355,25 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid, spin_lock(&adev->vm_manager.pasid_lock); vm = idr_find(&adev->vm_manager.pasid_idr, pasid); - if (vm) + if (vm) { root = amdgpu_bo_ref(vm->root.base.bo); - else + is_compute_context = vm->is_compute_context; + } else { root = NULL; + } spin_unlock(&adev->vm_manager.pasid_lock); if (!root) return false; + addr /= AMDGPU_GPU_PAGE_SIZE; + + if (is_compute_context && + !svm_range_restore_pages(adev, pasid, addr)) { + amdgpu_bo_unref(&root); + return true; + } + r = amdgpu_bo_reserve(root, true); if (r) goto error_unref; @@ -3333,18 +3387,16 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid, if (!vm) goto error_unlock; - addr /= AMDGPU_GPU_PAGE_SIZE; flags = AMDGPU_PTE_VALID | AMDGPU_PTE_SNOOPED | AMDGPU_PTE_SYSTEM; - if (vm->is_compute_context) { + if (is_compute_context) { /* Intentionally setting invalid PTE flag * combination to force a no-retry-fault */ flags = AMDGPU_PTE_EXECUTABLE | AMDGPU_PDE_PTE | AMDGPU_PTE_TF; value = 0; - } else if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_NEVER) { /* Redirect the access to the dummy page */ value = adev->dummy_page_addr; @@ -3363,7 +3415,7 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid, } r = amdgpu_vm_bo_update_mapping(adev, adev, vm, true, false, NULL, addr, - addr, flags, value, NULL, NULL, + addr, flags, value, NULL, NULL, NULL, NULL); if (r) goto error_unlock; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 4e140288159c..39d60e3c4e32 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -121,9 +121,6 @@ struct amdgpu_bo_list_entry; /* max vmids dedicated for process */ #define AMDGPU_VM_MAX_RESERVED_VMID 1 -#define AMDGPU_VM_CONTEXT_GFX 0 -#define AMDGPU_VM_CONTEXT_COMPUTE 1 - /* See vm_update_mode */ #define AMDGPU_VM_USE_CPU_FOR_GFX (1 << 0) #define AMDGPU_VM_USE_CPU_FOR_COMPUTE (1 << 1) @@ -234,6 +231,11 @@ struct amdgpu_vm_update_params { * @num_dw_left: number of dw left for the IB */ unsigned int num_dw_left; + + /** + * @table_freed: return true if page table is freed when updating + */ + bool table_freed; }; struct amdgpu_vm_update_funcs { @@ -367,6 +369,8 @@ struct amdgpu_vm_manager { spinlock_t pasid_lock; }; +struct amdgpu_bo_va_mapping; + #define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count))) #define amdgpu_vm_write_pte(adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr))) #define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags))) @@ -378,8 +382,7 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev); void amdgpu_vm_manager_fini(struct amdgpu_device *adev); long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout); -int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, - int vm_context, u32 pasid); +int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, u32 pasid); int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm, u32 pasid); void amdgpu_vm_release_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm); @@ -398,6 +401,15 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, struct dma_fence **fence); int amdgpu_vm_handle_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm); +int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, + struct amdgpu_device *bo_adev, + struct amdgpu_vm *vm, bool immediate, + bool unlocked, struct dma_resv *resv, + uint64_t start, uint64_t last, + uint64_t flags, uint64_t offset, + struct ttm_resource *res, + dma_addr_t *pages_addr, + struct dma_fence **fence, bool *free_table); int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, bool clear); @@ -447,6 +459,8 @@ void amdgpu_vm_set_task_info(struct amdgpu_vm *vm); void amdgpu_vm_move_to_lru_tail(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_vm_del_from_lru_notify(struct ttm_buffer_object *bo); +void amdgpu_vm_get_memory(struct amdgpu_vm *vm, uint64_t *vram_mem, + uint64_t *gtt_mem, uint64_t *cpu_mem); #if defined(CONFIG_DEBUG_FS) void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c index bce105e2973e..0f897a343073 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c @@ -29,12 +29,19 @@ #include "amdgpu_atomfirmware.h" #include "atom.h" -static inline struct amdgpu_vram_mgr *to_vram_mgr(struct ttm_resource_manager *man) +struct amdgpu_vram_reservation { + struct list_head node; + struct drm_mm_node mm_node; +}; + +static inline struct amdgpu_vram_mgr * +to_vram_mgr(struct ttm_resource_manager *man) { return container_of(man, struct amdgpu_vram_mgr, manager); } -static inline struct amdgpu_device *to_amdgpu_device(struct amdgpu_vram_mgr *mgr) +static inline struct amdgpu_device * +to_amdgpu_device(struct amdgpu_vram_mgr *mgr) { return container_of(mgr, struct amdgpu_device, mman.vram_mgr); } @@ -82,12 +89,14 @@ static ssize_t amdgpu_mem_info_vis_vram_total_show(struct device *dev, * amount of currently used VRAM in bytes */ static ssize_t amdgpu_mem_info_vram_used_show(struct device *dev, - struct device_attribute *attr, char *buf) + struct device_attribute *attr, + char *buf) { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); + struct ttm_resource_manager *man; + man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); return sysfs_emit(buf, "%llu\n", amdgpu_vram_mgr_usage(man)); } @@ -100,18 +109,28 @@ static ssize_t amdgpu_mem_info_vram_used_show(struct device *dev, * amount of currently used visible VRAM in bytes */ static ssize_t amdgpu_mem_info_vis_vram_used_show(struct device *dev, - struct device_attribute *attr, char *buf) + struct device_attribute *attr, + char *buf) { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); + struct ttm_resource_manager *man; + man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); return sysfs_emit(buf, "%llu\n", amdgpu_vram_mgr_vis_usage(man)); } +/** + * DOC: mem_info_vram_vendor + * + * The amdgpu driver provides a sysfs API for reporting the vendor of the + * installed VRAM + * The file mem_info_vram_vendor is used for this and returns the name of the + * vendor. + */ static ssize_t amdgpu_mem_info_vram_vendor(struct device *dev, - struct device_attribute *attr, - char *buf) + struct device_attribute *attr, + char *buf) { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); @@ -162,78 +181,6 @@ static const struct attribute *amdgpu_vram_mgr_attributes[] = { NULL }; -static const struct ttm_resource_manager_func amdgpu_vram_mgr_func; - -/** - * amdgpu_vram_mgr_init - init VRAM manager and DRM MM - * - * @adev: amdgpu_device pointer - * - * Allocate and initialize the VRAM manager. - */ -int amdgpu_vram_mgr_init(struct amdgpu_device *adev) -{ - struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr; - struct ttm_resource_manager *man = &mgr->manager; - int ret; - - ttm_resource_manager_init(man, adev->gmc.real_vram_size >> PAGE_SHIFT); - - man->func = &amdgpu_vram_mgr_func; - - drm_mm_init(&mgr->mm, 0, man->size); - spin_lock_init(&mgr->lock); - INIT_LIST_HEAD(&mgr->reservations_pending); - INIT_LIST_HEAD(&mgr->reserved_pages); - - /* Add the two VRAM-related sysfs files */ - ret = sysfs_create_files(&adev->dev->kobj, amdgpu_vram_mgr_attributes); - if (ret) - DRM_ERROR("Failed to register sysfs\n"); - - ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, &mgr->manager); - ttm_resource_manager_set_used(man, true); - return 0; -} - -/** - * amdgpu_vram_mgr_fini - free and destroy VRAM manager - * - * @adev: amdgpu_device pointer - * - * Destroy and free the VRAM manager, returns -EBUSY if ranges are still - * allocated inside it. - */ -void amdgpu_vram_mgr_fini(struct amdgpu_device *adev) -{ - struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr; - struct ttm_resource_manager *man = &mgr->manager; - int ret; - struct amdgpu_vram_reservation *rsv, *temp; - - ttm_resource_manager_set_used(man, false); - - ret = ttm_resource_manager_evict_all(&adev->mman.bdev, man); - if (ret) - return; - - spin_lock(&mgr->lock); - list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, node) - kfree(rsv); - - list_for_each_entry_safe(rsv, temp, &mgr->reserved_pages, node) { - drm_mm_remove_node(&rsv->mm_node); - kfree(rsv); - } - drm_mm_takedown(&mgr->mm); - spin_unlock(&mgr->lock); - - sysfs_remove_files(&adev->dev->kobj, amdgpu_vram_mgr_attributes); - - ttm_resource_manager_cleanup(man); - ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, NULL); -} - /** * amdgpu_vram_mgr_vis_size - Calculate visible node size * @@ -283,6 +230,7 @@ u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo) return usage; } +/* Commit the reservation of VRAM pages */ static void amdgpu_vram_mgr_do_reserve(struct ttm_resource_manager *man) { struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); @@ -415,13 +363,13 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, const struct ttm_place *place, struct ttm_resource *mem) { + unsigned long lpfn, num_nodes, pages_per_node, pages_left, pages; struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); struct amdgpu_device *adev = to_amdgpu_device(mgr); + uint64_t vis_usage = 0, mem_bytes, max_bytes; struct drm_mm *mm = &mgr->mm; - struct drm_mm_node *nodes; enum drm_mm_insert_mode mode; - unsigned long lpfn, num_nodes, pages_per_node, pages_left; - uint64_t vis_usage = 0, mem_bytes, max_bytes; + struct drm_mm_node *nodes; unsigned i; int r; @@ -448,9 +396,10 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, pages_per_node = HPAGE_PMD_NR; #else /* default to 2MB */ - pages_per_node = (2UL << (20UL - PAGE_SHIFT)); + pages_per_node = 2UL << (20UL - PAGE_SHIFT); #endif - pages_per_node = max((uint32_t)pages_per_node, mem->page_alignment); + pages_per_node = max_t(uint32_t, pages_per_node, + tbo->page_alignment); num_nodes = DIV_ROUND_UP(mem->num_pages, pages_per_node); } @@ -468,49 +417,45 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, mem->start = 0; pages_left = mem->num_pages; - spin_lock(&mgr->lock); - for (i = 0; pages_left >= pages_per_node; ++i) { - unsigned long pages = rounddown_pow_of_two(pages_left); - - /* Limit maximum size to 2GB due to SG table limitations */ - pages = min(pages, (2UL << (30 - PAGE_SHIFT))); - - r = drm_mm_insert_node_in_range(mm, &nodes[i], pages, - pages_per_node, 0, - place->fpfn, lpfn, - mode); - if (unlikely(r)) - break; + /* Limit maximum size to 2GB due to SG table limitations */ + pages = min(pages_left, 2UL << (30 - PAGE_SHIFT)); - vis_usage += amdgpu_vram_mgr_vis_size(adev, &nodes[i]); - amdgpu_vram_mgr_virt_start(mem, &nodes[i]); - pages_left -= pages; - } - - for (; pages_left; ++i) { - unsigned long pages = min(pages_left, pages_per_node); - uint32_t alignment = mem->page_alignment; + i = 0; + spin_lock(&mgr->lock); + while (pages_left) { + uint32_t alignment = tbo->page_alignment; - if (pages == pages_per_node) + if (pages >= pages_per_node) alignment = pages_per_node; - r = drm_mm_insert_node_in_range(mm, &nodes[i], - pages, alignment, 0, - place->fpfn, lpfn, - mode); - if (unlikely(r)) + r = drm_mm_insert_node_in_range(mm, &nodes[i], pages, alignment, + 0, place->fpfn, lpfn, mode); + if (unlikely(r)) { + if (pages > pages_per_node) { + if (is_power_of_2(pages)) + pages = pages / 2; + else + pages = rounddown_pow_of_two(pages); + continue; + } goto error; + } vis_usage += amdgpu_vram_mgr_vis_size(adev, &nodes[i]); amdgpu_vram_mgr_virt_start(mem, &nodes[i]); pages_left -= pages; + ++i; + + if (pages > pages_left) + pages = pages_left; } spin_unlock(&mgr->lock); - atomic64_add(vis_usage, &mgr->vis_usage); + if (i == 1) + mem->placement |= TTM_PL_FLAG_CONTIGUOUS; + atomic64_add(vis_usage, &mgr->vis_usage); mem->mm_node = nodes; - return 0; error: @@ -727,3 +672,73 @@ static const struct ttm_resource_manager_func amdgpu_vram_mgr_func = { .free = amdgpu_vram_mgr_del, .debug = amdgpu_vram_mgr_debug }; + +/** + * amdgpu_vram_mgr_init - init VRAM manager and DRM MM + * + * @adev: amdgpu_device pointer + * + * Allocate and initialize the VRAM manager. + */ +int amdgpu_vram_mgr_init(struct amdgpu_device *adev) +{ + struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr; + struct ttm_resource_manager *man = &mgr->manager; + int ret; + + ttm_resource_manager_init(man, adev->gmc.real_vram_size >> PAGE_SHIFT); + + man->func = &amdgpu_vram_mgr_func; + + drm_mm_init(&mgr->mm, 0, man->size); + spin_lock_init(&mgr->lock); + INIT_LIST_HEAD(&mgr->reservations_pending); + INIT_LIST_HEAD(&mgr->reserved_pages); + + /* Add the two VRAM-related sysfs files */ + ret = sysfs_create_files(&adev->dev->kobj, amdgpu_vram_mgr_attributes); + if (ret) + DRM_ERROR("Failed to register sysfs\n"); + + ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, &mgr->manager); + ttm_resource_manager_set_used(man, true); + return 0; +} + +/** + * amdgpu_vram_mgr_fini - free and destroy VRAM manager + * + * @adev: amdgpu_device pointer + * + * Destroy and free the VRAM manager, returns -EBUSY if ranges are still + * allocated inside it. + */ +void amdgpu_vram_mgr_fini(struct amdgpu_device *adev) +{ + struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr; + struct ttm_resource_manager *man = &mgr->manager; + int ret; + struct amdgpu_vram_reservation *rsv, *temp; + + ttm_resource_manager_set_used(man, false); + + ret = ttm_resource_manager_evict_all(&adev->mman.bdev, man); + if (ret) + return; + + spin_lock(&mgr->lock); + list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, node) + kfree(rsv); + + list_for_each_entry_safe(rsv, temp, &mgr->reserved_pages, node) { + drm_mm_remove_node(&rsv->mm_node); + kfree(rsv); + } + drm_mm_takedown(&mgr->mm); + spin_unlock(&mgr->lock); + + sysfs_remove_files(&adev->dev->kobj, amdgpu_vram_mgr_attributes); + + ttm_resource_manager_cleanup(man); + ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, NULL); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h index 1a8f6d4baab2..a434c71fde8e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h @@ -56,6 +56,8 @@ #define AMD_SRIOV_MSG_RESERVE_UCODE 24 +#define AMD_SRIOV_MSG_RESERVE_VCN_INST 4 + enum amd_sriov_ucode_engine_id { AMD_SRIOV_UCODE_ID_VCE = 0, AMD_SRIOV_UCODE_ID_UVD, @@ -98,10 +100,10 @@ union amd_sriov_msg_feature_flags { union amd_sriov_reg_access_flags { struct { - uint32_t vf_reg_access_ih : 1; + uint32_t vf_reg_access_ih : 1; uint32_t vf_reg_access_mmhub : 1; - uint32_t vf_reg_access_gc : 1; - uint32_t reserved : 29; + uint32_t vf_reg_access_gc : 1; + uint32_t reserved : 29; } flags; uint32_t all; }; @@ -114,6 +116,37 @@ union amd_sriov_msg_os_info { uint32_t all; }; +struct amd_sriov_msg_uuid_info { + union { + struct { + uint32_t did : 16; + uint32_t fcn : 8; + uint32_t asic_7 : 8; + }; + uint32_t time_low; + }; + + struct { + uint32_t time_mid : 16; + uint32_t time_high : 12; + uint32_t version : 4; + }; + + struct { + struct { + uint8_t clk_seq_hi : 6; + uint8_t variant : 2; + }; + union { + uint8_t clk_seq_low; + uint8_t asic_6; + }; + uint16_t asic_4; + }; + + uint32_t asic_0; +}; + struct amd_sriov_msg_pf2vf_info_header { /* the total structure size in byte */ uint32_t size; @@ -160,10 +193,19 @@ struct amd_sriov_msg_pf2vf_info { /* identification in ROCm SMI */ uint64_t uuid; uint32_t fcn_idx; - /* flags which indicate the register access method VF should use */ + /* flags to indicate which register access method VF should use */ union amd_sriov_reg_access_flags reg_access_flags; + /* MM BW management */ + struct { + uint32_t decode_max_dimension_pixels; + uint32_t decode_max_frame_pixels; + uint32_t encode_max_dimension_pixels; + uint32_t encode_max_frame_pixels; + } mm_bw_management[AMD_SRIOV_MSG_RESERVE_VCN_INST]; + /* UUID info */ + struct amd_sriov_msg_uuid_info uuid_info; /* reserved */ - uint32_t reserved[256-27]; + uint32_t reserved[256 - 47]; }; struct amd_sriov_msg_vf2pf_info_header { diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c index 2ac4988ea0ff..c12c2900732b 100644 --- a/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c +++ b/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c @@ -74,6 +74,7 @@ int athub_v2_1_set_clockgating(struct amdgpu_device *adev, case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: athub_v2_1_update_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE); athub_v2_1_update_medium_grain_light_sleep(adev, state == AMD_CG_STATE_GATE); break; diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c index a3ba9ca11e98..f327becb022f 100644 --- a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c +++ b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c @@ -188,6 +188,8 @@ void amdgpu_atombios_dp_aux_init(struct amdgpu_connector *amdgpu_connector) { amdgpu_connector->ddc_bus->rec.hpd = amdgpu_connector->hpd.hpd; amdgpu_connector->ddc_bus->aux.transfer = amdgpu_atombios_dp_aux_transfer; + amdgpu_connector->ddc_bus->aux.drm_dev = amdgpu_connector->base.dev; + drm_dp_aux_init(&amdgpu_connector->ddc_bus->aux); amdgpu_connector->ddc_bus->has_aux = true; } @@ -610,7 +612,7 @@ amdgpu_atombios_dp_link_train_cr(struct amdgpu_atombios_dp_link_train_info *dp_i dp_info->tries = 0; voltage = 0xff; while (1) { - drm_dp_link_train_clock_recovery_delay(dp_info->dpcd); + drm_dp_link_train_clock_recovery_delay(dp_info->aux, dp_info->dpcd); if (drm_dp_dpcd_read_link_status(dp_info->aux, dp_info->link_status) <= 0) { @@ -675,7 +677,7 @@ amdgpu_atombios_dp_link_train_ce(struct amdgpu_atombios_dp_link_train_info *dp_i dp_info->tries = 0; channel_eq = false; while (1) { - drm_dp_link_train_channel_eq_delay(dp_info->dpcd); + drm_dp_link_train_channel_eq_delay(dp_info->aux, dp_info->dpcd); if (drm_dp_dpcd_read_link_status(dp_info->aux, dp_info->link_status) <= 0) { diff --git a/drivers/gpu/drm/amd/amdgpu/beige_goby_reg_init.c b/drivers/gpu/drm/amd/amdgpu/beige_goby_reg_init.c new file mode 100644 index 000000000000..608a113ce354 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/beige_goby_reg_init.c @@ -0,0 +1,54 @@ +/* + * Copyright 2020 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 "nv.h" + +#include "soc15_common.h" +#include "soc15_hw_ip.h" +#include "beige_goby_ip_offset.h" + +int beige_goby_reg_base_init(struct amdgpu_device *adev) +{ + /* HW has more IP blocks, only initialize the block needed by driver */ + uint32_t i; + for (i = 0 ; i < MAX_INSTANCE ; ++i) { + adev->reg_offset[GC_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i])); + adev->reg_offset[HDP_HWIP][i] = (uint32_t *)(&(HDP_BASE.instance[i])); + adev->reg_offset[MMHUB_HWIP][i] = (uint32_t *)(&(MMHUB_BASE.instance[i])); + adev->reg_offset[ATHUB_HWIP][i] = (uint32_t *)(&(ATHUB_BASE.instance[i])); + adev->reg_offset[NBIO_HWIP][i] = (uint32_t *)(&(NBIO_BASE.instance[i])); + adev->reg_offset[MP0_HWIP][i] = (uint32_t *)(&(MP0_BASE.instance[i])); + adev->reg_offset[MP1_HWIP][i] = (uint32_t *)(&(MP1_BASE.instance[i])); + adev->reg_offset[VCN_HWIP][i] = (uint32_t *)(&(VCN0_BASE.instance[i])); + adev->reg_offset[DF_HWIP][i] = (uint32_t *)(&(DF_BASE.instance[i])); + adev->reg_offset[DCE_HWIP][i] = (uint32_t *)(&(DCN_BASE.instance[i])); + adev->reg_offset[OSSSYS_HWIP][i] = (uint32_t *)(&(OSSSYS_BASE.instance[i])); + adev->reg_offset[SDMA0_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i])); + adev->reg_offset[SDMA1_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i])); + adev->reg_offset[SDMA2_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i])); + adev->reg_offset[SDMA3_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i])); + adev->reg_offset[SMUIO_HWIP][i] = (uint32_t *)(&(SMUIO_BASE.instance[i])); + adev->reg_offset[THM_HWIP][i] = (uint32_t *)(&(THM_BASE.instance[i])); + } + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c index 5c11144da051..33324427b555 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c @@ -421,6 +421,11 @@ static int dce_virtual_sw_init(void *handle) static int dce_virtual_sw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i = 0; + + for (i = 0; i < adev->mode_info.num_crtc; i++) + if (adev->mode_info.crtcs[i]) + hrtimer_cancel(&adev->mode_info.crtcs[i]->vblank_timer); kfree(adev->mode_info.bios_hardcoded_edid); @@ -480,13 +485,6 @@ static int dce_virtual_hw_init(void *handle) static int dce_virtual_hw_fini(void *handle) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - int i = 0; - - for (i = 0; i<adev->mode_info.num_crtc; i++) - if (adev->mode_info.crtcs[i]) - hrtimer_cancel(&adev->mode_info.crtcs[i]->vblank_timer); - return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c index 0d8459d63bac..14514a145c17 100644 --- a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c +++ b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c @@ -219,11 +219,11 @@ static void df_v3_6_query_hashes(struct amdgpu_device *adev) adev->df.hash_status.hash_2m = false; adev->df.hash_status.hash_1g = false; - if (adev->asic_type != CHIP_ARCTURUS) - return; - - /* encoding for hash-enabled on Arcturus */ - if (adev->df.funcs->get_fb_channel_number(adev) == 0xe) { + /* encoding for hash-enabled on Arcturus and Aldebaran */ + if ((adev->asic_type == CHIP_ARCTURUS && + adev->df.funcs->get_fb_channel_number(adev) == 0xe) || + (adev->asic_type == CHIP_ALDEBARAN && + adev->df.funcs->get_fb_channel_number(adev) == 0x1e)) { tmp = RREG32_SOC15(DF, 0, mmDF_CS_UMC_AON0_DfGlobalCtrl); adev->df.hash_status.hash_64k = REG_GET_FIELD(tmp, DF_CS_UMC_AON0_DfGlobalCtrl, @@ -277,8 +277,14 @@ static u32 df_v3_6_get_fb_channel_number(struct amdgpu_device *adev) { u32 tmp; - tmp = RREG32_SOC15(DF, 0, mmDF_CS_UMC_AON0_DramBaseAddress0); - tmp &= DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan_MASK; + if (adev->asic_type == CHIP_ALDEBARAN) { + tmp = RREG32_SOC15(DF, 0, mmDF_GCM_AON0_DramMegaBaseAddress0); + tmp &= + ALDEBARAN_DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan_MASK; + } else { + tmp = RREG32_SOC15(DF, 0, mmDF_CS_UMC_AON0_DramBaseAddress0); + tmp &= DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan_MASK; + } tmp >>= DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan__SHIFT; return tmp; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c index 2408ed4c7d84..fc12e3c3e9ca 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c @@ -232,6 +232,13 @@ MODULE_FIRMWARE("amdgpu/dimgrey_cavefish_mec.bin"); MODULE_FIRMWARE("amdgpu/dimgrey_cavefish_mec2.bin"); MODULE_FIRMWARE("amdgpu/dimgrey_cavefish_rlc.bin"); +MODULE_FIRMWARE("amdgpu/beige_goby_ce.bin"); +MODULE_FIRMWARE("amdgpu/beige_goby_pfp.bin"); +MODULE_FIRMWARE("amdgpu/beige_goby_me.bin"); +MODULE_FIRMWARE("amdgpu/beige_goby_mec.bin"); +MODULE_FIRMWARE("amdgpu/beige_goby_mec2.bin"); +MODULE_FIRMWARE("amdgpu/beige_goby_rlc.bin"); + static const struct soc15_reg_golden golden_settings_gc_10_1[] = { SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_4, 0xffffffff, 0x00400014), @@ -1395,9 +1402,10 @@ static const struct soc15_reg_golden golden_settings_gc_10_1_2[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG, 0xffffffff, 0x20000000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG2, 0xffffffff, 0x00000420), SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG3, 0xffffffff, 0x00000200), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG4, 0xffffffff, 0x04800000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG4, 0xffffffff, 0x04900000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DFSM_TILES_IN_FLIGHT, 0x0000ffff, 0x0000003f), SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_LAST_OF_BURST_CONFIG, 0xffffffff, 0x03860204), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_ADDR_CONFIG, 0x0c1800ff, 0x00000044), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCR_GENERAL_CNTL, 0x1ff0ffff, 0x00000500), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGE_PRIV_CONTROL, 0x00007fff, 0x000001fe), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL1_PIPE_STEER, 0xffffffff, 0xe4e4e4e4), @@ -1415,12 +1423,13 @@ static const struct soc15_reg_golden golden_settings_gc_10_1_2[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE_2, 0x00000820, 0x00000820), SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_LINE_STIPPLE_STATE, 0x0000ff0f, 0x00000000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmRMI_SPARE, 0xffffffff, 0xffff3101), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_CONFIG_CNTL_1, 0x001f0000, 0x00070104), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_ALU_CLK_CTRL, 0xffffffff, 0xffffffff), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_ARB_CONFIG, 0x00000133, 0x00000130), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_LDS_CLK_CTRL, 0xffffffff, 0xffffffff), SOC15_REG_GOLDEN_VALUE(GC, 0, mmTA_CNTL_AUX, 0xfff7ffff, 0x01030000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CNTL, 0xffdf80ff, 0x479c0010), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffffffff, 0x00800000) + SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffffffff, 0x00c00000) }; static bool gfx_v10_is_rlcg_rw(struct amdgpu_device *adev, u32 offset, uint32_t *flag, bool write) @@ -3404,6 +3413,41 @@ static const struct soc15_reg_golden golden_settings_gc_10_3_4[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmLDS_CONFIG, 0x00000020, 0x00000020) }; +static const struct soc15_reg_golden golden_settings_gc_10_3_5[] = { + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0x78000000, 0x78000100), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA0_CLK_CTRL, 0xb0000ff0, 0x30000100), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA1_CLK_CTRL, 0xff000000, 0x7e000100), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPF_GCR_CNTL, 0x0007ffff, 0x0000c000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG3, 0xffffffff, 0x00000280), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG4, 0xffffffff, 0x00800000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCR_GENERAL_CNTL_Sienna_Cichlid, 0x1ff1ffff, 0x00000500), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2A_ADDR_MATCH_MASK, 0xffffffff, 0xffffffcf), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_ADDR_MATCH_MASK, 0xffffffff, 0xffffffcf), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CM_CTRL1, 0xff8fff0f, 0x580f1008), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CTRL3, 0xf7ffffff, 0x00f80988), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmLDS_CONFIG, 0x000001ff, 0x00000020), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_CL_ENHANCE, 0xf17fffff, 0x01200007), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_CONFIG, 0xe07df47f, 0x00180070), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER0_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER1_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER10_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER11_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER12_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER13_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER14_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER15_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER2_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER3_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER4_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER5_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER6_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER7_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER8_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER9_SELECT, 0xf0f001ff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmTA_CNTL_AUX,0xfff7ffff, 0x01030000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffbfffff, 0x00a00000) +}; + #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) | \ @@ -3624,6 +3668,11 @@ static void gfx_v10_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_gc_10_3_4, (const u32)ARRAY_SIZE(golden_settings_gc_10_3_4)); break; + case CHIP_BEIGE_GOBY: + soc15_program_register_sequence(adev, + golden_settings_gc_10_3_5, + (const u32)ARRAY_SIZE(golden_settings_gc_10_3_5)); + break; default: break; } @@ -3809,6 +3858,7 @@ static void gfx_v10_0_check_fw_write_wait(struct amdgpu_device *adev) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: adev->gfx.cp_fw_write_wait = true; break; default: @@ -3924,6 +3974,9 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev) case CHIP_DIMGREY_CAVEFISH: chip_name = "dimgrey_cavefish"; break; + case CHIP_BEIGE_GOBY: + chip_name = "beige_goby"; + break; default: BUG(); } @@ -4492,6 +4545,7 @@ static void gfx_v10_0_gpu_early_init(struct amdgpu_device *adev) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: adev->gfx.config.max_hw_contexts = 8; adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -4616,6 +4670,7 @@ static int gfx_v10_0_sw_init(void *handle) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: adev->gfx.me.num_me = 1; adev->gfx.me.num_pipe_per_me = 1; adev->gfx.me.num_queue_per_pipe = 1; @@ -6123,6 +6178,7 @@ static void gfx_v10_0_cp_gfx_set_doorbell(struct amdgpu_device *adev, case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: tmp = REG_SET_FIELD(0, CP_RB_DOORBELL_RANGE_LOWER, DOORBELL_RANGE_LOWER_Sienna_Cichlid, ring->doorbell_index); WREG32_SOC15(GC, 0, mmCP_RB_DOORBELL_RANGE_LOWER, tmp); @@ -6258,6 +6314,7 @@ static void gfx_v10_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: WREG32_SOC15(GC, 0, mmCP_MEC_CNTL_Sienna_Cichlid, 0); break; default: @@ -6270,6 +6327,7 @@ static void gfx_v10_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: WREG32_SOC15(GC, 0, mmCP_MEC_CNTL_Sienna_Cichlid, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK)); @@ -6366,6 +6424,7 @@ static void gfx_v10_0_kiq_setting(struct amdgpu_ring *ring) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid); tmp &= 0xffffff00; tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue); @@ -7080,6 +7139,7 @@ static bool gfx_v10_0_check_grbm_cam_remapping(struct amdgpu_device *adev) case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: data = RREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_Sienna_Cichlid); WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_Sienna_Cichlid, 0); WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_UMD, pattern); @@ -7114,6 +7174,9 @@ static void gfx_v10_0_setup_grbm_cam_remapping(struct amdgpu_device *adev) { uint32_t data; + if (amdgpu_sriov_vf(adev)) + return; + /* initialize cam_index to 0 * index will auto-inc after each data writting */ WREG32_SOC15(GC, 0, mmGRBM_CAM_INDEX, 0); @@ -7123,6 +7186,7 @@ static void gfx_v10_0_setup_grbm_cam_remapping(struct amdgpu_device *adev) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: /* mmVGT_TF_RING_SIZE_UMD -> mmVGT_TF_RING_SIZE */ data = (SOC15_REG_OFFSET(GC, 0, mmVGT_TF_RING_SIZE_UMD) << GRBM_CAM_DATA__CAM_ADDR__SHIFT) | @@ -7432,6 +7496,7 @@ static int gfx_v10_0_soft_reset(void *handle) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: if (REG_GET_FIELD(tmp, GRBM_STATUS2, RLC_BUSY_Sienna_Cichlid)) grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, @@ -7541,6 +7606,7 @@ static int gfx_v10_0_early_init(void *handle) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: adev->gfx.num_gfx_rings = GFX10_NUM_GFX_RINGS_Sienna_Cichlid; break; default: @@ -7597,6 +7663,7 @@ static void gfx_v10_0_set_safe_mode(struct amdgpu_device *adev) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE_Sienna_Cichlid, data); /* wait for RLC_SAFE_MODE */ @@ -7631,6 +7698,7 @@ static void gfx_v10_0_unset_safe_mode(struct amdgpu_device *adev) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE_Sienna_Cichlid, data); break; default: @@ -7998,6 +8066,7 @@ static int gfx_v10_0_set_powergating_state(void *handle, case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: amdgpu_gfx_off_ctrl(adev, enable); break; case CHIP_VANGOGH: @@ -8026,6 +8095,7 @@ static int gfx_v10_0_set_clockgating_state(void *handle, case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: gfx_v10_0_update_gfx_clock_gating(adev, state == AMD_CG_STATE_GATE); break; @@ -9136,6 +9206,7 @@ static void gfx_v10_0_set_rlc_funcs(struct amdgpu_device *adev) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: adev->gfx.rlc.funcs = &gfx_v10_0_rlc_funcs; break; case CHIP_NAVI12: diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index a078a38c2cee..feaa5e4a5538 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -3937,7 +3937,8 @@ static void gfx_v9_0_init_tcp_config(struct amdgpu_device *adev) { u32 tmp; - if (adev->asic_type != CHIP_ARCTURUS) + if (adev->asic_type != CHIP_ARCTURUS && + adev->asic_type != CHIP_ALDEBARAN) return; tmp = RREG32_SOC15(GC, 0, mmTCP_ADDR_CONFIG); @@ -4559,8 +4560,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) if (!ring->sched.ready) return 0; - if (adev->asic_type == CHIP_ARCTURUS || - adev->asic_type == CHIP_ALDEBARAN) { + if (adev->asic_type == CHIP_ARCTURUS) { vgpr_init_shader_ptr = vgpr_init_compute_shader_arcturus; vgpr_init_shader_size = sizeof(vgpr_init_compute_shader_arcturus); vgpr_init_regs_ptr = vgpr_init_regs_arcturus; @@ -4745,7 +4745,11 @@ static int gfx_v9_0_ecc_late_init(void *handle) } /* requires IBs so do in late init after IB pool is initialized */ - r = gfx_v9_0_do_edc_gpr_workarounds(adev); + if (adev->asic_type == CHIP_ALDEBARAN) + r = gfx_v9_4_2_do_edc_gpr_workarounds(adev); + else + r = gfx_v9_0_do_edc_gpr_workarounds(adev); + if (r) return r; @@ -4943,7 +4947,7 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, amdgpu_gfx_rlc_enter_safe_mode(adev); /* Enable 3D CGCG/CGLS */ - if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG)) { + if (enable) { /* write cmd to clear cgcg/cgls ov */ def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); /* unset CGCG override */ @@ -4955,8 +4959,12 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, /* enable 3Dcgcg FSM(0x0000363f) */ def = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D); - data = (0x36 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | - RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK; + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG) + data = (0x36 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | + RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK; + else + data = 0x0 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT; + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS) data |= (0x000F << RLC_CGCG_CGLS_CTRL_3D__CGLS_REP_COMPANSAT_DELAY__SHIFT) | RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c index a30c7c10cd9a..dbad9ef002d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c @@ -22,6 +22,7 @@ */ #include "amdgpu.h" #include "soc15.h" +#include "soc15d.h" #include "gc/gc_9_4_2_offset.h" #include "gc/gc_9_4_2_sh_mask.h" @@ -31,6 +32,11 @@ #include "amdgpu_ras.h" #include "amdgpu_gfx.h" +#define SE_ID_MAX 8 +#define CU_ID_MAX 16 +#define SIMD_ID_MAX 4 +#define WAVE_ID_MAX 10 + enum gfx_v9_4_2_utc_type { VML2_MEM, VML2_WALKER_MEM, @@ -79,6 +85,634 @@ static const struct soc15_reg_golden golden_settings_gc_9_4_2_alde[] = { SOC15_REG_GOLDEN_VALUE(GC, 0, regTCI_CNTL_3, 0xff, 0x20), }; +/** + * This shader is used to clear VGPRS and LDS, and also write the input + * pattern into the write back buffer, which will be used by driver to + * check whether all SIMDs have been covered. +*/ +static const u32 vgpr_init_compute_shader_aldebaran[] = { + 0xb8840904, 0xb8851a04, 0xb8861344, 0xb8831804, 0x9208ff06, 0x00000280, + 0x9209a805, 0x920a8a04, 0x81080908, 0x81080a08, 0x81080308, 0x8e078208, + 0x81078407, 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8a0000, 0xd3d94000, + 0x18000080, 0xd3d94001, 0x18000080, 0xd3d94002, 0x18000080, 0xd3d94003, + 0x18000080, 0xd3d94004, 0x18000080, 0xd3d94005, 0x18000080, 0xd3d94006, + 0x18000080, 0xd3d94007, 0x18000080, 0xd3d94008, 0x18000080, 0xd3d94009, + 0x18000080, 0xd3d9400a, 0x18000080, 0xd3d9400b, 0x18000080, 0xd3d9400c, + 0x18000080, 0xd3d9400d, 0x18000080, 0xd3d9400e, 0x18000080, 0xd3d9400f, + 0x18000080, 0xd3d94010, 0x18000080, 0xd3d94011, 0x18000080, 0xd3d94012, + 0x18000080, 0xd3d94013, 0x18000080, 0xd3d94014, 0x18000080, 0xd3d94015, + 0x18000080, 0xd3d94016, 0x18000080, 0xd3d94017, 0x18000080, 0xd3d94018, + 0x18000080, 0xd3d94019, 0x18000080, 0xd3d9401a, 0x18000080, 0xd3d9401b, + 0x18000080, 0xd3d9401c, 0x18000080, 0xd3d9401d, 0x18000080, 0xd3d9401e, + 0x18000080, 0xd3d9401f, 0x18000080, 0xd3d94020, 0x18000080, 0xd3d94021, + 0x18000080, 0xd3d94022, 0x18000080, 0xd3d94023, 0x18000080, 0xd3d94024, + 0x18000080, 0xd3d94025, 0x18000080, 0xd3d94026, 0x18000080, 0xd3d94027, + 0x18000080, 0xd3d94028, 0x18000080, 0xd3d94029, 0x18000080, 0xd3d9402a, + 0x18000080, 0xd3d9402b, 0x18000080, 0xd3d9402c, 0x18000080, 0xd3d9402d, + 0x18000080, 0xd3d9402e, 0x18000080, 0xd3d9402f, 0x18000080, 0xd3d94030, + 0x18000080, 0xd3d94031, 0x18000080, 0xd3d94032, 0x18000080, 0xd3d94033, + 0x18000080, 0xd3d94034, 0x18000080, 0xd3d94035, 0x18000080, 0xd3d94036, + 0x18000080, 0xd3d94037, 0x18000080, 0xd3d94038, 0x18000080, 0xd3d94039, + 0x18000080, 0xd3d9403a, 0x18000080, 0xd3d9403b, 0x18000080, 0xd3d9403c, + 0x18000080, 0xd3d9403d, 0x18000080, 0xd3d9403e, 0x18000080, 0xd3d9403f, + 0x18000080, 0xd3d94040, 0x18000080, 0xd3d94041, 0x18000080, 0xd3d94042, + 0x18000080, 0xd3d94043, 0x18000080, 0xd3d94044, 0x18000080, 0xd3d94045, + 0x18000080, 0xd3d94046, 0x18000080, 0xd3d94047, 0x18000080, 0xd3d94048, + 0x18000080, 0xd3d94049, 0x18000080, 0xd3d9404a, 0x18000080, 0xd3d9404b, + 0x18000080, 0xd3d9404c, 0x18000080, 0xd3d9404d, 0x18000080, 0xd3d9404e, + 0x18000080, 0xd3d9404f, 0x18000080, 0xd3d94050, 0x18000080, 0xd3d94051, + 0x18000080, 0xd3d94052, 0x18000080, 0xd3d94053, 0x18000080, 0xd3d94054, + 0x18000080, 0xd3d94055, 0x18000080, 0xd3d94056, 0x18000080, 0xd3d94057, + 0x18000080, 0xd3d94058, 0x18000080, 0xd3d94059, 0x18000080, 0xd3d9405a, + 0x18000080, 0xd3d9405b, 0x18000080, 0xd3d9405c, 0x18000080, 0xd3d9405d, + 0x18000080, 0xd3d9405e, 0x18000080, 0xd3d9405f, 0x18000080, 0xd3d94060, + 0x18000080, 0xd3d94061, 0x18000080, 0xd3d94062, 0x18000080, 0xd3d94063, + 0x18000080, 0xd3d94064, 0x18000080, 0xd3d94065, 0x18000080, 0xd3d94066, + 0x18000080, 0xd3d94067, 0x18000080, 0xd3d94068, 0x18000080, 0xd3d94069, + 0x18000080, 0xd3d9406a, 0x18000080, 0xd3d9406b, 0x18000080, 0xd3d9406c, + 0x18000080, 0xd3d9406d, 0x18000080, 0xd3d9406e, 0x18000080, 0xd3d9406f, + 0x18000080, 0xd3d94070, 0x18000080, 0xd3d94071, 0x18000080, 0xd3d94072, + 0x18000080, 0xd3d94073, 0x18000080, 0xd3d94074, 0x18000080, 0xd3d94075, + 0x18000080, 0xd3d94076, 0x18000080, 0xd3d94077, 0x18000080, 0xd3d94078, + 0x18000080, 0xd3d94079, 0x18000080, 0xd3d9407a, 0x18000080, 0xd3d9407b, + 0x18000080, 0xd3d9407c, 0x18000080, 0xd3d9407d, 0x18000080, 0xd3d9407e, + 0x18000080, 0xd3d9407f, 0x18000080, 0xd3d94080, 0x18000080, 0xd3d94081, + 0x18000080, 0xd3d94082, 0x18000080, 0xd3d94083, 0x18000080, 0xd3d94084, + 0x18000080, 0xd3d94085, 0x18000080, 0xd3d94086, 0x18000080, 0xd3d94087, + 0x18000080, 0xd3d94088, 0x18000080, 0xd3d94089, 0x18000080, 0xd3d9408a, + 0x18000080, 0xd3d9408b, 0x18000080, 0xd3d9408c, 0x18000080, 0xd3d9408d, + 0x18000080, 0xd3d9408e, 0x18000080, 0xd3d9408f, 0x18000080, 0xd3d94090, + 0x18000080, 0xd3d94091, 0x18000080, 0xd3d94092, 0x18000080, 0xd3d94093, + 0x18000080, 0xd3d94094, 0x18000080, 0xd3d94095, 0x18000080, 0xd3d94096, + 0x18000080, 0xd3d94097, 0x18000080, 0xd3d94098, 0x18000080, 0xd3d94099, + 0x18000080, 0xd3d9409a, 0x18000080, 0xd3d9409b, 0x18000080, 0xd3d9409c, + 0x18000080, 0xd3d9409d, 0x18000080, 0xd3d9409e, 0x18000080, 0xd3d9409f, + 0x18000080, 0xd3d940a0, 0x18000080, 0xd3d940a1, 0x18000080, 0xd3d940a2, + 0x18000080, 0xd3d940a3, 0x18000080, 0xd3d940a4, 0x18000080, 0xd3d940a5, + 0x18000080, 0xd3d940a6, 0x18000080, 0xd3d940a7, 0x18000080, 0xd3d940a8, + 0x18000080, 0xd3d940a9, 0x18000080, 0xd3d940aa, 0x18000080, 0xd3d940ab, + 0x18000080, 0xd3d940ac, 0x18000080, 0xd3d940ad, 0x18000080, 0xd3d940ae, + 0x18000080, 0xd3d940af, 0x18000080, 0xd3d940b0, 0x18000080, 0xd3d940b1, + 0x18000080, 0xd3d940b2, 0x18000080, 0xd3d940b3, 0x18000080, 0xd3d940b4, + 0x18000080, 0xd3d940b5, 0x18000080, 0xd3d940b6, 0x18000080, 0xd3d940b7, + 0x18000080, 0xd3d940b8, 0x18000080, 0xd3d940b9, 0x18000080, 0xd3d940ba, + 0x18000080, 0xd3d940bb, 0x18000080, 0xd3d940bc, 0x18000080, 0xd3d940bd, + 0x18000080, 0xd3d940be, 0x18000080, 0xd3d940bf, 0x18000080, 0xd3d940c0, + 0x18000080, 0xd3d940c1, 0x18000080, 0xd3d940c2, 0x18000080, 0xd3d940c3, + 0x18000080, 0xd3d940c4, 0x18000080, 0xd3d940c5, 0x18000080, 0xd3d940c6, + 0x18000080, 0xd3d940c7, 0x18000080, 0xd3d940c8, 0x18000080, 0xd3d940c9, + 0x18000080, 0xd3d940ca, 0x18000080, 0xd3d940cb, 0x18000080, 0xd3d940cc, + 0x18000080, 0xd3d940cd, 0x18000080, 0xd3d940ce, 0x18000080, 0xd3d940cf, + 0x18000080, 0xd3d940d0, 0x18000080, 0xd3d940d1, 0x18000080, 0xd3d940d2, + 0x18000080, 0xd3d940d3, 0x18000080, 0xd3d940d4, 0x18000080, 0xd3d940d5, + 0x18000080, 0xd3d940d6, 0x18000080, 0xd3d940d7, 0x18000080, 0xd3d940d8, + 0x18000080, 0xd3d940d9, 0x18000080, 0xd3d940da, 0x18000080, 0xd3d940db, + 0x18000080, 0xd3d940dc, 0x18000080, 0xd3d940dd, 0x18000080, 0xd3d940de, + 0x18000080, 0xd3d940df, 0x18000080, 0xd3d940e0, 0x18000080, 0xd3d940e1, + 0x18000080, 0xd3d940e2, 0x18000080, 0xd3d940e3, 0x18000080, 0xd3d940e4, + 0x18000080, 0xd3d940e5, 0x18000080, 0xd3d940e6, 0x18000080, 0xd3d940e7, + 0x18000080, 0xd3d940e8, 0x18000080, 0xd3d940e9, 0x18000080, 0xd3d940ea, + 0x18000080, 0xd3d940eb, 0x18000080, 0xd3d940ec, 0x18000080, 0xd3d940ed, + 0x18000080, 0xd3d940ee, 0x18000080, 0xd3d940ef, 0x18000080, 0xd3d940f0, + 0x18000080, 0xd3d940f1, 0x18000080, 0xd3d940f2, 0x18000080, 0xd3d940f3, + 0x18000080, 0xd3d940f4, 0x18000080, 0xd3d940f5, 0x18000080, 0xd3d940f6, + 0x18000080, 0xd3d940f7, 0x18000080, 0xd3d940f8, 0x18000080, 0xd3d940f9, + 0x18000080, 0xd3d940fa, 0x18000080, 0xd3d940fb, 0x18000080, 0xd3d940fc, + 0x18000080, 0xd3d940fd, 0x18000080, 0xd3d940fe, 0x18000080, 0xd3d940ff, + 0x18000080, 0xb07c0000, 0xbe8a00ff, 0x000000f8, 0xbf11080a, 0x7e000280, + 0x7e020280, 0x7e040280, 0x7e060280, 0x7e080280, 0x7e0a0280, 0x7e0c0280, + 0x7e0e0280, 0x808a880a, 0xbe80320a, 0xbf84fff5, 0xbf9c0000, 0xd28c0001, + 0x0001007f, 0xd28d0001, 0x0002027e, 0x10020288, 0xbe8b0004, 0xb78b4000, + 0xd1196a01, 0x00001701, 0xbe8a0087, 0xbefc00c1, 0xd89c4000, 0x00020201, + 0xd89cc080, 0x00040401, 0x320202ff, 0x00000800, 0x808a810a, 0xbf84fff8, + 0xbf810000, +}; + +const struct soc15_reg_entry vgpr_init_regs_aldebaran[] = { + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Y), 4 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Z), 1 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC1), 0xbf }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC2), 0x400006 }, /* 64KB LDS */ + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC3), 0x3F }, /* 63 - accum-offset = 256 */ + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE4), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE5), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE6), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE7), 0xffffffff }, +}; + +/** + * The below shaders are used to clear SGPRS, and also write the input + * pattern into the write back buffer. The first two dispatch should be + * scheduled simultaneously which make sure that all SGPRS could be + * allocated, so the dispatch 1 need check write back buffer before scheduled, + * make sure that waves of dispatch 0 are all dispacthed to all simds + * balanced. both dispatch 0 and dispatch 1 should be halted until all waves + * are dispatched, and then driver write a pattern to the shared memory to make + * all waves continue. +*/ +static const u32 sgpr112_init_compute_shader_aldebaran[] = { + 0xb8840904, 0xb8851a04, 0xb8861344, 0xb8831804, 0x9208ff06, 0x00000280, + 0x9209a805, 0x920a8a04, 0x81080908, 0x81080a08, 0x81080308, 0x8e078208, + 0x81078407, 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8e003f, 0xc0030200, + 0x00000000, 0xbf8c0000, 0xbf06ff08, 0xdeadbeaf, 0xbf84fff9, 0x81028102, + 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8a0000, 0xbefc0080, 0xbeea0080, + 0xbeeb0080, 0xbf00f280, 0xbee60080, 0xbee70080, 0xbee80080, 0xbee90080, + 0xbefe0080, 0xbeff0080, 0xbe880080, 0xbe890080, 0xbe8a0080, 0xbe8b0080, + 0xbe8c0080, 0xbe8d0080, 0xbe8e0080, 0xbe8f0080, 0xbe900080, 0xbe910080, + 0xbe920080, 0xbe930080, 0xbe940080, 0xbe950080, 0xbe960080, 0xbe970080, + 0xbe980080, 0xbe990080, 0xbe9a0080, 0xbe9b0080, 0xbe9c0080, 0xbe9d0080, + 0xbe9e0080, 0xbe9f0080, 0xbea00080, 0xbea10080, 0xbea20080, 0xbea30080, + 0xbea40080, 0xbea50080, 0xbea60080, 0xbea70080, 0xbea80080, 0xbea90080, + 0xbeaa0080, 0xbeab0080, 0xbeac0080, 0xbead0080, 0xbeae0080, 0xbeaf0080, + 0xbeb00080, 0xbeb10080, 0xbeb20080, 0xbeb30080, 0xbeb40080, 0xbeb50080, + 0xbeb60080, 0xbeb70080, 0xbeb80080, 0xbeb90080, 0xbeba0080, 0xbebb0080, + 0xbebc0080, 0xbebd0080, 0xbebe0080, 0xbebf0080, 0xbec00080, 0xbec10080, + 0xbec20080, 0xbec30080, 0xbec40080, 0xbec50080, 0xbec60080, 0xbec70080, + 0xbec80080, 0xbec90080, 0xbeca0080, 0xbecb0080, 0xbecc0080, 0xbecd0080, + 0xbece0080, 0xbecf0080, 0xbed00080, 0xbed10080, 0xbed20080, 0xbed30080, + 0xbed40080, 0xbed50080, 0xbed60080, 0xbed70080, 0xbed80080, 0xbed90080, + 0xbeda0080, 0xbedb0080, 0xbedc0080, 0xbedd0080, 0xbede0080, 0xbedf0080, + 0xbee00080, 0xbee10080, 0xbee20080, 0xbee30080, 0xbee40080, 0xbee50080, + 0xbf810000 +}; + +const struct soc15_reg_entry sgpr112_init_regs_aldebaran[] = { + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Y), 8 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Z), 1 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC1), 0x340 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC2), 0x6 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC3), 0x0 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE4), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE5), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE6), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE7), 0xffffffff }, +}; + +static const u32 sgpr96_init_compute_shader_aldebaran[] = { + 0xb8840904, 0xb8851a04, 0xb8861344, 0xb8831804, 0x9208ff06, 0x00000280, + 0x9209a805, 0x920a8a04, 0x81080908, 0x81080a08, 0x81080308, 0x8e078208, + 0x81078407, 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8e003f, 0xc0030200, + 0x00000000, 0xbf8c0000, 0xbf06ff08, 0xdeadbeaf, 0xbf84fff9, 0x81028102, + 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8a0000, 0xbefc0080, 0xbeea0080, + 0xbeeb0080, 0xbf00f280, 0xbee60080, 0xbee70080, 0xbee80080, 0xbee90080, + 0xbefe0080, 0xbeff0080, 0xbe880080, 0xbe890080, 0xbe8a0080, 0xbe8b0080, + 0xbe8c0080, 0xbe8d0080, 0xbe8e0080, 0xbe8f0080, 0xbe900080, 0xbe910080, + 0xbe920080, 0xbe930080, 0xbe940080, 0xbe950080, 0xbe960080, 0xbe970080, + 0xbe980080, 0xbe990080, 0xbe9a0080, 0xbe9b0080, 0xbe9c0080, 0xbe9d0080, + 0xbe9e0080, 0xbe9f0080, 0xbea00080, 0xbea10080, 0xbea20080, 0xbea30080, + 0xbea40080, 0xbea50080, 0xbea60080, 0xbea70080, 0xbea80080, 0xbea90080, + 0xbeaa0080, 0xbeab0080, 0xbeac0080, 0xbead0080, 0xbeae0080, 0xbeaf0080, + 0xbeb00080, 0xbeb10080, 0xbeb20080, 0xbeb30080, 0xbeb40080, 0xbeb50080, + 0xbeb60080, 0xbeb70080, 0xbeb80080, 0xbeb90080, 0xbeba0080, 0xbebb0080, + 0xbebc0080, 0xbebd0080, 0xbebe0080, 0xbebf0080, 0xbec00080, 0xbec10080, + 0xbec20080, 0xbec30080, 0xbec40080, 0xbec50080, 0xbec60080, 0xbec70080, + 0xbec80080, 0xbec90080, 0xbeca0080, 0xbecb0080, 0xbecc0080, 0xbecd0080, + 0xbece0080, 0xbecf0080, 0xbed00080, 0xbed10080, 0xbed20080, 0xbed30080, + 0xbed40080, 0xbed50080, 0xbed60080, 0xbed70080, 0xbed80080, 0xbed90080, + 0xbf810000, +}; + +const struct soc15_reg_entry sgpr96_init_regs_aldebaran[] = { + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Y), 0xc }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Z), 1 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC1), 0x2c0 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC2), 0x6 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC3), 0x0 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE4), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE5), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE6), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE7), 0xffffffff }, +}; + +/** + * This shader is used to clear the uninitiated sgprs after the above + * two dispatches, because of hardware feature, dispath 0 couldn't clear + * top hole sgprs. Therefore need 4 waves per SIMD to cover these sgprs +*/ +static const u32 sgpr64_init_compute_shader_aldebaran[] = { + 0xb8840904, 0xb8851a04, 0xb8861344, 0xb8831804, 0x9208ff06, 0x00000280, + 0x9209a805, 0x920a8a04, 0x81080908, 0x81080a08, 0x81080308, 0x8e078208, + 0x81078407, 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8e003f, 0xc0030200, + 0x00000000, 0xbf8c0000, 0xbf06ff08, 0xdeadbeaf, 0xbf84fff9, 0x81028102, + 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8a0000, 0xbefc0080, 0xbeea0080, + 0xbeeb0080, 0xbf00f280, 0xbee60080, 0xbee70080, 0xbee80080, 0xbee90080, + 0xbefe0080, 0xbeff0080, 0xbe880080, 0xbe890080, 0xbe8a0080, 0xbe8b0080, + 0xbe8c0080, 0xbe8d0080, 0xbe8e0080, 0xbe8f0080, 0xbe900080, 0xbe910080, + 0xbe920080, 0xbe930080, 0xbe940080, 0xbe950080, 0xbe960080, 0xbe970080, + 0xbe980080, 0xbe990080, 0xbe9a0080, 0xbe9b0080, 0xbe9c0080, 0xbe9d0080, + 0xbe9e0080, 0xbe9f0080, 0xbea00080, 0xbea10080, 0xbea20080, 0xbea30080, + 0xbea40080, 0xbea50080, 0xbea60080, 0xbea70080, 0xbea80080, 0xbea90080, + 0xbeaa0080, 0xbeab0080, 0xbeac0080, 0xbead0080, 0xbeae0080, 0xbeaf0080, + 0xbeb00080, 0xbeb10080, 0xbeb20080, 0xbeb30080, 0xbeb40080, 0xbeb50080, + 0xbeb60080, 0xbeb70080, 0xbeb80080, 0xbeb90080, 0xbf810000, +}; + +const struct soc15_reg_entry sgpr64_init_regs_aldebaran[] = { + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Y), 0x10 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Z), 1 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC1), 0x1c0 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC2), 0x6 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC3), 0x0 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE4), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE5), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE6), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE7), 0xffffffff }, +}; + +static int gfx_v9_4_2_run_shader(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_ib *ib, + const u32 *shader_ptr, u32 shader_size, + const struct soc15_reg_entry *init_regs, u32 regs_size, + u32 compute_dim_x, u64 wb_gpu_addr, u32 pattern, + struct dma_fence **fence_ptr) +{ + int r, i; + uint32_t total_size, shader_offset; + u64 gpu_addr; + + total_size = (regs_size * 3 + 4 + 5 + 5) * 4; + total_size = ALIGN(total_size, 256); + shader_offset = total_size; + total_size += ALIGN(shader_size, 256); + + /* allocate an indirect buffer to put the commands in */ + memset(ib, 0, sizeof(*ib)); + r = amdgpu_ib_get(adev, NULL, total_size, + AMDGPU_IB_POOL_DIRECT, ib); + if (r) { + dev_err(adev->dev, "failed to get ib (%d).\n", r); + return r; + } + + /* load the compute shaders */ + for (i = 0; i < shader_size/sizeof(u32); i++) + ib->ptr[i + (shader_offset / 4)] = shader_ptr[i]; + + /* init the ib length to 0 */ + ib->length_dw = 0; + + /* write the register state for the compute dispatch */ + for (i = 0; i < regs_size; i++) { + ib->ptr[ib->length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1); + ib->ptr[ib->length_dw++] = SOC15_REG_ENTRY_OFFSET(init_regs[i]) + - PACKET3_SET_SH_REG_START; + ib->ptr[ib->length_dw++] = init_regs[i].reg_value; + } + + /* write the shader start address: mmCOMPUTE_PGM_LO, mmCOMPUTE_PGM_HI */ + gpu_addr = (ib->gpu_addr + (u64)shader_offset) >> 8; + ib->ptr[ib->length_dw++] = PACKET3(PACKET3_SET_SH_REG, 2); + ib->ptr[ib->length_dw++] = SOC15_REG_OFFSET(GC, 0, regCOMPUTE_PGM_LO) + - PACKET3_SET_SH_REG_START; + ib->ptr[ib->length_dw++] = lower_32_bits(gpu_addr); + ib->ptr[ib->length_dw++] = upper_32_bits(gpu_addr); + + /* write the wb buffer address */ + ib->ptr[ib->length_dw++] = PACKET3(PACKET3_SET_SH_REG, 3); + ib->ptr[ib->length_dw++] = SOC15_REG_OFFSET(GC, 0, regCOMPUTE_USER_DATA_0) + - PACKET3_SET_SH_REG_START; + ib->ptr[ib->length_dw++] = lower_32_bits(wb_gpu_addr); + ib->ptr[ib->length_dw++] = upper_32_bits(wb_gpu_addr); + ib->ptr[ib->length_dw++] = pattern; + + /* write dispatch packet */ + ib->ptr[ib->length_dw++] = PACKET3(PACKET3_DISPATCH_DIRECT, 3); + ib->ptr[ib->length_dw++] = compute_dim_x; /* x */ + ib->ptr[ib->length_dw++] = 1; /* y */ + ib->ptr[ib->length_dw++] = 1; /* z */ + ib->ptr[ib->length_dw++] = + REG_SET_FIELD(0, COMPUTE_DISPATCH_INITIATOR, COMPUTE_SHADER_EN, 1); + + /* shedule the ib on the ring */ + r = amdgpu_ib_schedule(ring, 1, ib, NULL, fence_ptr); + if (r) { + dev_err(adev->dev, "ib submit failed (%d).\n", r); + amdgpu_ib_free(adev, ib, NULL); + } + return r; +} + +static void gfx_v9_4_2_log_wave_assignment(struct amdgpu_device *adev, uint32_t *wb_ptr) +{ + uint32_t se, cu, simd, wave; + uint32_t offset = 0; + char *str; + int size; + + str = kmalloc(256, GFP_KERNEL); + if (!str) + return; + + dev_dbg(adev->dev, "wave assignment:\n"); + + for (se = 0; se < adev->gfx.config.max_shader_engines; se++) { + for (cu = 0; cu < CU_ID_MAX; cu++) { + memset(str, 0, 256); + size = sprintf(str, "SE[%02d]CU[%02d]: ", se, cu); + for (simd = 0; simd < SIMD_ID_MAX; simd++) { + size += sprintf(str + size, "["); + for (wave = 0; wave < WAVE_ID_MAX; wave++) { + size += sprintf(str + size, "%x", wb_ptr[offset]); + offset++; + } + size += sprintf(str + size, "] "); + } + dev_dbg(adev->dev, "%s\n", str); + } + } + + kfree(str); +} + +static int gfx_v9_4_2_wait_for_waves_assigned(struct amdgpu_device *adev, + uint32_t *wb_ptr, uint32_t mask, + uint32_t pattern, uint32_t num_wave, bool wait) +{ + uint32_t se, cu, simd, wave; + uint32_t loop = 0; + uint32_t wave_cnt; + uint32_t offset; + + do { + wave_cnt = 0; + offset = 0; + + for (se = 0; se < adev->gfx.config.max_shader_engines; se++) + for (cu = 0; cu < CU_ID_MAX; cu++) + for (simd = 0; simd < SIMD_ID_MAX; simd++) + for (wave = 0; wave < WAVE_ID_MAX; wave++) { + if (((1 << wave) & mask) && + (wb_ptr[offset] == pattern)) + wave_cnt++; + + offset++; + } + + if (wave_cnt == num_wave) + return 0; + + mdelay(1); + } while (++loop < 2000 && wait); + + dev_err(adev->dev, "actual wave num: %d, expected wave num: %d\n", + wave_cnt, num_wave); + + gfx_v9_4_2_log_wave_assignment(adev, wb_ptr); + + return -EBADSLT; +} + +static int gfx_v9_4_2_do_sgprs_init(struct amdgpu_device *adev) +{ + int r; + int wb_size = adev->gfx.config.max_shader_engines * + CU_ID_MAX * SIMD_ID_MAX * WAVE_ID_MAX; + struct amdgpu_ib wb_ib; + struct amdgpu_ib disp_ibs[3]; + struct dma_fence *fences[3]; + u32 pattern[3] = { 0x1, 0x5, 0xa }; + + /* bail if the compute ring is not ready */ + if (!adev->gfx.compute_ring[0].sched.ready || + !adev->gfx.compute_ring[1].sched.ready) + return 0; + + /* allocate the write-back buffer from IB */ + memset(&wb_ib, 0, sizeof(wb_ib)); + r = amdgpu_ib_get(adev, NULL, (1 + wb_size) * sizeof(uint32_t), + AMDGPU_IB_POOL_DIRECT, &wb_ib); + if (r) { + dev_err(adev->dev, "failed to get ib (%d) for wb\n", r); + return r; + } + memset(wb_ib.ptr, 0, (1 + wb_size) * sizeof(uint32_t)); + + r = gfx_v9_4_2_run_shader(adev, + &adev->gfx.compute_ring[0], + &disp_ibs[0], + sgpr112_init_compute_shader_aldebaran, + sizeof(sgpr112_init_compute_shader_aldebaran), + sgpr112_init_regs_aldebaran, + ARRAY_SIZE(sgpr112_init_regs_aldebaran), + adev->gfx.cu_info.number, + wb_ib.gpu_addr, pattern[0], &fences[0]); + if (r) { + dev_err(adev->dev, "failed to clear first 224 sgprs\n"); + goto pro_end; + } + + r = gfx_v9_4_2_wait_for_waves_assigned(adev, + &wb_ib.ptr[1], 0b11, + pattern[0], + adev->gfx.cu_info.number * SIMD_ID_MAX * 2, + true); + if (r) { + dev_err(adev->dev, "wave coverage failed when clear first 224 sgprs\n"); + wb_ib.ptr[0] = 0xdeadbeaf; /* stop waves */ + goto disp0_failed; + } + + r = gfx_v9_4_2_run_shader(adev, + &adev->gfx.compute_ring[1], + &disp_ibs[1], + sgpr96_init_compute_shader_aldebaran, + sizeof(sgpr96_init_compute_shader_aldebaran), + sgpr96_init_regs_aldebaran, + ARRAY_SIZE(sgpr96_init_regs_aldebaran), + adev->gfx.cu_info.number * 2, + wb_ib.gpu_addr, pattern[1], &fences[1]); + if (r) { + dev_err(adev->dev, "failed to clear next 576 sgprs\n"); + goto disp0_failed; + } + + r = gfx_v9_4_2_wait_for_waves_assigned(adev, + &wb_ib.ptr[1], 0b11111100, + pattern[1], adev->gfx.cu_info.number * SIMD_ID_MAX * 6, + true); + if (r) { + dev_err(adev->dev, "wave coverage failed when clear first 576 sgprs\n"); + wb_ib.ptr[0] = 0xdeadbeaf; /* stop waves */ + goto disp1_failed; + } + + wb_ib.ptr[0] = 0xdeadbeaf; /* stop waves */ + + /* wait for the GPU to finish processing the IB */ + r = dma_fence_wait(fences[0], false); + if (r) { + dev_err(adev->dev, "timeout to clear first 224 sgprs\n"); + goto disp1_failed; + } + + r = dma_fence_wait(fences[1], false); + if (r) { + dev_err(adev->dev, "timeout to clear first 576 sgprs\n"); + goto disp1_failed; + } + + memset(wb_ib.ptr, 0, (1 + wb_size) * sizeof(uint32_t)); + r = gfx_v9_4_2_run_shader(adev, + &adev->gfx.compute_ring[0], + &disp_ibs[2], + sgpr64_init_compute_shader_aldebaran, + sizeof(sgpr64_init_compute_shader_aldebaran), + sgpr64_init_regs_aldebaran, + ARRAY_SIZE(sgpr64_init_regs_aldebaran), + adev->gfx.cu_info.number, + wb_ib.gpu_addr, pattern[2], &fences[2]); + if (r) { + dev_err(adev->dev, "failed to clear first 256 sgprs\n"); + goto disp1_failed; + } + + r = gfx_v9_4_2_wait_for_waves_assigned(adev, + &wb_ib.ptr[1], 0b1111, + pattern[2], + adev->gfx.cu_info.number * SIMD_ID_MAX * 4, + true); + if (r) { + dev_err(adev->dev, "wave coverage failed when clear first 256 sgprs\n"); + wb_ib.ptr[0] = 0xdeadbeaf; /* stop waves */ + goto disp2_failed; + } + + wb_ib.ptr[0] = 0xdeadbeaf; /* stop waves */ + + r = dma_fence_wait(fences[2], false); + if (r) { + dev_err(adev->dev, "timeout to clear first 256 sgprs\n"); + goto disp2_failed; + } + +disp2_failed: + amdgpu_ib_free(adev, &disp_ibs[2], NULL); + dma_fence_put(fences[2]); +disp1_failed: + amdgpu_ib_free(adev, &disp_ibs[1], NULL); + dma_fence_put(fences[1]); +disp0_failed: + amdgpu_ib_free(adev, &disp_ibs[0], NULL); + dma_fence_put(fences[0]); +pro_end: + amdgpu_ib_free(adev, &wb_ib, NULL); + + if (r) + dev_info(adev->dev, "Init SGPRS Failed\n"); + else + dev_info(adev->dev, "Init SGPRS Successfully\n"); + + return r; +} + +static int gfx_v9_4_2_do_vgprs_init(struct amdgpu_device *adev) +{ + int r; + /* CU_ID: 0~15, SIMD_ID: 0~3, WAVE_ID: 0 ~ 9 */ + int wb_size = adev->gfx.config.max_shader_engines * + CU_ID_MAX * SIMD_ID_MAX * WAVE_ID_MAX; + struct amdgpu_ib wb_ib; + struct amdgpu_ib disp_ib; + struct dma_fence *fence; + u32 pattern = 0xa; + + /* bail if the compute ring is not ready */ + if (!adev->gfx.compute_ring[0].sched.ready) + return 0; + + /* allocate the write-back buffer from IB */ + memset(&wb_ib, 0, sizeof(wb_ib)); + r = amdgpu_ib_get(adev, NULL, (1 + wb_size) * sizeof(uint32_t), + AMDGPU_IB_POOL_DIRECT, &wb_ib); + if (r) { + dev_err(adev->dev, "failed to get ib (%d) for wb.\n", r); + return r; + } + memset(wb_ib.ptr, 0, (1 + wb_size) * sizeof(uint32_t)); + + r = gfx_v9_4_2_run_shader(adev, + &adev->gfx.compute_ring[0], + &disp_ib, + vgpr_init_compute_shader_aldebaran, + sizeof(vgpr_init_compute_shader_aldebaran), + vgpr_init_regs_aldebaran, + ARRAY_SIZE(vgpr_init_regs_aldebaran), + adev->gfx.cu_info.number, + wb_ib.gpu_addr, pattern, &fence); + if (r) { + dev_err(adev->dev, "failed to clear vgprs\n"); + goto pro_end; + } + + /* wait for the GPU to finish processing the IB */ + r = dma_fence_wait(fence, false); + if (r) { + dev_err(adev->dev, "timeout to clear vgprs\n"); + goto disp_failed; + } + + r = gfx_v9_4_2_wait_for_waves_assigned(adev, + &wb_ib.ptr[1], 0b1, + pattern, + adev->gfx.cu_info.number * SIMD_ID_MAX, + false); + if (r) { + dev_err(adev->dev, "failed to cover all simds when clearing vgprs\n"); + goto disp_failed; + } + +disp_failed: + amdgpu_ib_free(adev, &disp_ib, NULL); + dma_fence_put(fence); +pro_end: + amdgpu_ib_free(adev, &wb_ib, NULL); + + if (r) + dev_info(adev->dev, "Init VGPRS Failed\n"); + else + dev_info(adev->dev, "Init VGPRS Successfully\n"); + + return r; +} + +int gfx_v9_4_2_do_edc_gpr_workarounds(struct amdgpu_device *adev) +{ + /* only support when RAS is enabled */ + if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) + return 0; + + gfx_v9_4_2_do_sgprs_init(adev); + + gfx_v9_4_2_do_vgprs_init(adev); + + return 0; +} + static void gfx_v9_4_2_query_sq_timeout_status(struct amdgpu_device *adev); static void gfx_v9_4_2_reset_sq_timeout_status(struct amdgpu_device *adev); @@ -808,8 +1442,9 @@ static struct gfx_v9_4_2_utc_block gfx_v9_4_2_utc_blocks[] = { REG_SET_FIELD(0, ATC_L2_CACHE_4K_DSM_CNTL, WRITE_COUNTERS, 1) }, }; -static const struct soc15_reg_entry gfx_v9_4_2_ea_err_status_regs = - { SOC15_REG_ENTRY(GC, 0, regGCEA_ERR_STATUS), 0, 1, 16 }; +static const struct soc15_reg_entry gfx_v9_4_2_ea_err_status_regs = { + SOC15_REG_ENTRY(GC, 0, regGCEA_ERR_STATUS), 0, 1, 16 +}; static int gfx_v9_4_2_get_reg_error_count(struct amdgpu_device *adev, const struct soc15_reg_entry *reg, @@ -1039,13 +1674,16 @@ static void gfx_v9_4_2_reset_utc_err_status(struct amdgpu_device *adev) static void gfx_v9_4_2_reset_ea_err_status(struct amdgpu_device *adev) { uint32_t i, j; + uint32_t value; + + value = REG_SET_FIELD(0, GCEA_ERR_STATUS, CLEAR_ERROR_STATUS, 0x1); mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < gfx_v9_4_2_ea_err_status_regs.se_num; i++) { for (j = 0; j < gfx_v9_4_2_ea_err_status_regs.instance; j++) { gfx_v9_4_2_select_se_sh(adev, i, 0, j); - WREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_4_2_ea_err_status_regs), 0x10); + WREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_4_2_ea_err_status_regs), value); } } gfx_v9_4_2_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h index 81c5833b6b9f..6db1f88509af 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h @@ -29,6 +29,7 @@ void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev, void gfx_v9_4_2_init_golden_registers(struct amdgpu_device *adev, uint32_t die_id); void gfx_v9_4_2_set_power_brake_sequence(struct amdgpu_device *adev); +int gfx_v9_4_2_do_edc_gpr_workarounds(struct amdgpu_device *adev); extern const struct amdgpu_gfx_ras_funcs gfx_v9_4_2_ras_funcs; diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c index 1e4678cb98f0..a03fdd41212b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c @@ -283,10 +283,14 @@ static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_BLOCK_SIZE, block_size); - /* Send no-retry XNACK on fault to suppress VM fault storm. */ + /* Send no-retry XNACK on fault to suppress VM fault storm. + * On Aldebaran, XNACK can be enabled in the SQ per-process. + * Retry faults need to be enabled for that to work. + */ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, - !adev->gmc.noretry); + !adev->gmc.noretry || + adev->asic_type == CHIP_ALDEBARAN); WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_CNTL, i * hub->ctx_distance, tmp); WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c index 498b28a35f5b..f02dc904e4cf 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c @@ -288,7 +288,7 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid, if (i < adev->usec_timeout) return; - DRM_ERROR("Timeout waiting for VM flush ACK!\n"); + DRM_ERROR("Timeout waiting for VM flush hub: %d!\n", vmhub); } /** @@ -681,6 +681,7 @@ static void gmc_v10_0_set_gfxhub_funcs(struct amdgpu_device *adev) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: adev->gfxhub.funcs = &gfxhub_v2_1_funcs; break; default: @@ -796,6 +797,7 @@ static int gmc_v10_0_mc_init(struct amdgpu_device *adev) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: default: adev->gmc.gart_size = 512ULL << 20; break; @@ -863,6 +865,7 @@ static int gmc_v10_0_sw_init(void *handle) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: adev->num_vmhubs = 2; /* * To fulfill 4-level page support, @@ -978,6 +981,7 @@ static void gmc_v10_0_init_golden_registers(struct amdgpu_device *adev) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: break; default: break; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 455bb91060d0..fb3c7588dde3 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -53,6 +53,7 @@ #include "mmhub_v1_7.h" #include "umc_v6_1.h" #include "umc_v6_0.h" +#include "hdp_v4_0.h" #include "ivsrcid/vmc/irqsrcs_vmc_1_0.h" @@ -1210,6 +1211,11 @@ static void gmc_v9_0_set_gfxhub_funcs(struct amdgpu_device *adev) adev->gfxhub.funcs = &gfxhub_v1_0_funcs; } +static void gmc_v9_0_set_hdp_ras_funcs(struct amdgpu_device *adev) +{ + adev->hdp.ras_funcs = &hdp_v4_0_ras_funcs; +} + static int gmc_v9_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -1230,6 +1236,7 @@ static int gmc_v9_0_early_init(void *handle) gmc_v9_0_set_mmhub_funcs(adev); gmc_v9_0_set_mmhub_ras_funcs(adev); gmc_v9_0_set_gfxhub_funcs(adev); + gmc_v9_0_set_hdp_ras_funcs(adev); adev->gmc.shared_aperture_start = 0x2000000000000000ULL; adev->gmc.shared_aperture_end = @@ -1255,15 +1262,21 @@ static int gmc_v9_0_late_init(void *handle) * writes, while disables HBM ECC for vega10. */ if (!amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_VEGA10)) { - if (!(adev->ras_features & (1 << AMDGPU_RAS_BLOCK__UMC))) { + if (!(adev->ras_enabled & (1 << AMDGPU_RAS_BLOCK__UMC))) { if (adev->df.funcs->enable_ecc_force_par_wr_rmw) adev->df.funcs->enable_ecc_force_par_wr_rmw(adev, false); } } - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->reset_ras_error_count) - adev->mmhub.ras_funcs->reset_ras_error_count(adev); + if (!amdgpu_persistent_edc_harvesting_supported(adev)) { + if (adev->mmhub.ras_funcs && + adev->mmhub.ras_funcs->reset_ras_error_count) + adev->mmhub.ras_funcs->reset_ras_error_count(adev); + + if (adev->hdp.ras_funcs && + adev->hdp.ras_funcs->reset_ras_error_count) + adev->hdp.ras_funcs->reset_ras_error_count(adev); + } r = amdgpu_gmc_ras_late_init(adev); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c index edbd35d293eb..74b90cc2bf48 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c @@ -59,12 +59,31 @@ static void hdp_v4_0_invalidate_hdp(struct amdgpu_device *adev, HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1); } +static void hdp_v4_0_query_ras_error_count(struct amdgpu_device *adev, + void *ras_error_status) +{ + struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; + + err_data->ue_count = 0; + err_data->ce_count = 0; + + if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__HDP)) + return; + + /* HDP SRAM errors are uncorrectable ones (i.e. fatal errors) */ + err_data->ue_count += RREG32_SOC15(HDP, 0, mmHDP_EDC_CNT); +}; + static void hdp_v4_0_reset_ras_error_count(struct amdgpu_device *adev) { if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__HDP)) return; - /*read back hdp ras counter to reset it to 0 */ - RREG32_SOC15(HDP, 0, mmHDP_EDC_CNT); + + if (adev->asic_type >= CHIP_ALDEBARAN) + WREG32_SOC15(HDP, 0, mmHDP_EDC_CNT, 0); + else + /*read back hdp ras counter to reset it to 0 */ + RREG32_SOC15(HDP, 0, mmHDP_EDC_CNT); } static void hdp_v4_0_update_clock_gating(struct amdgpu_device *adev, @@ -130,10 +149,16 @@ static void hdp_v4_0_init_registers(struct amdgpu_device *adev) WREG32_SOC15(HDP, 0, mmHDP_NONSURFACE_BASE_HI, (adev->gmc.vram_start >> 40)); } +const struct amdgpu_hdp_ras_funcs hdp_v4_0_ras_funcs = { + .ras_late_init = amdgpu_hdp_ras_late_init, + .ras_fini = amdgpu_hdp_ras_fini, + .query_ras_error_count = hdp_v4_0_query_ras_error_count, + .reset_ras_error_count = hdp_v4_0_reset_ras_error_count, +}; + const struct amdgpu_hdp_funcs hdp_v4_0_funcs = { .flush_hdp = hdp_v4_0_flush_hdp, .invalidate_hdp = hdp_v4_0_invalidate_hdp, - .reset_ras_error_count = hdp_v4_0_reset_ras_error_count, .update_clock_gating = hdp_v4_0_update_clock_gating, .get_clock_gating_state = hdp_v4_0_get_clockgating_state, .init_registers = hdp_v4_0_init_registers, diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h index d1e6399e8c46..dc3a1b81dd62 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h @@ -27,5 +27,6 @@ #include "soc15_common.h" extern const struct amdgpu_hdp_funcs hdp_v4_0_funcs; +extern const struct amdgpu_hdp_ras_funcs hdp_v4_0_ras_funcs; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c index 83531997aeba..938ef4ce5b76 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c @@ -198,8 +198,6 @@ static int jpeg_v2_5_hw_fini(void *handle) if (adev->jpeg.cur_state != AMD_PG_STATE_GATE && RREG32_SOC15(JPEG, i, mmUVD_JRBC_STATUS)) jpeg_v2_5_set_powergating_state(adev, AMD_PG_STATE_GATE); - - ring->sched.ready = false; } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c index de5dfcfb3859..94be35357f7d 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c @@ -166,8 +166,6 @@ static int jpeg_v3_0_hw_fini(void *handle) RREG32_SOC15(JPEG, 0, mmUVD_JRBC_STATUS)) jpeg_v3_0_set_powergating_state(adev, AMD_PG_STATE_GATE); - ring->sched.ready = false; - return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c index 0103a5ab28e6..6264934b67ac 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c @@ -296,10 +296,12 @@ static void mmhub_v1_7_setup_vmid_config(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_BLOCK_SIZE, block_size); - /* Send no-retry XNACK on fault to suppress VM fault storm. */ + /* On Aldebaran, XNACK can be enabled in the SQ per-process. + * Retry faults need to be enabled for that to work. + */ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, - !adev->gmc.noretry); + 1); WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_CNTL, i * hub->ctx_distance, tmp); WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, @@ -1313,12 +1315,31 @@ static void mmhub_v1_7_query_ras_error_status(struct amdgpu_device *adev) } } +static void mmhub_v1_7_reset_ras_error_status(struct amdgpu_device *adev) +{ + int i; + uint32_t reg_value; + + if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__MMHUB)) + return; + + for (i = 0; i < ARRAY_SIZE(mmhub_v1_7_ea_err_status_regs); i++) { + reg_value = RREG32(SOC15_REG_ENTRY_OFFSET( + mmhub_v1_7_ea_err_status_regs[i])); + reg_value = REG_SET_FIELD(reg_value, MMEA0_ERR_STATUS, + CLEAR_ERROR_STATUS, 0x01); + WREG32(SOC15_REG_ENTRY_OFFSET(mmhub_v1_7_ea_err_status_regs[i]), + reg_value); + } +} + const struct amdgpu_mmhub_ras_funcs mmhub_v1_7_ras_funcs = { .ras_late_init = amdgpu_mmhub_ras_late_init, .ras_fini = amdgpu_mmhub_ras_fini, .query_ras_error_count = mmhub_v1_7_query_ras_error_count, .reset_ras_error_count = mmhub_v1_7_reset_ras_error_count, .query_ras_error_status = mmhub_v1_7_query_ras_error_status, + .reset_ras_error_status = mmhub_v1_7_reset_ras_error_status, }; const struct amdgpu_mmhub_funcs mmhub_v1_7_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c index ac76081b91d5..655c3d5f3d35 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c @@ -93,6 +93,30 @@ static const char *mmhub_client_ids_sienna_cichlid[][2] = { [15][1] = "OSS", }; +static const char *mmhub_client_ids_beige_goby[][2] = { + [3][0] = "DCEDMC", + [4][0] = "DCEVGA", + [5][0] = "MP0", + [6][0] = "MP1", + [8][0] = "VMC", + [9][0] = "VCNU0", + [11][0] = "VCN0", + [14][0] = "HDP", + [15][0] = "OSS", + [0][1] = "DBGU0", + [1][1] = "DBGU1", + [2][1] = "DCEDWB", + [3][1] = "DCEDMC", + [4][1] = "DCEVGA", + [5][1] = "MP0", + [6][1] = "MP1", + [7][1] = "XDP", + [9][1] = "VCNU0", + [11][1] = "VCN0", + [14][1] = "HDP", + [15][1] = "OSS", +}; + static uint32_t mmhub_v2_0_get_invalidate_req(unsigned int vmid, uint32_t flush_type) { @@ -139,6 +163,9 @@ mmhub_v2_0_print_l2_protection_fault_status(struct amdgpu_device *adev, case CHIP_DIMGREY_CAVEFISH: mmhub_cid = mmhub_client_ids_sienna_cichlid[cid][rw]; break; + case CHIP_BEIGE_GOBY: + mmhub_cid = mmhub_client_ids_beige_goby[cid][rw]; + break; default: mmhub_cid = NULL; break; @@ -544,6 +571,7 @@ static void mmhub_v2_0_update_medium_grain_clock_gating(struct amdgpu_device *ad case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: def = data = RREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG_Sienna_Cichlid); def1 = data1 = RREG32_SOC15(MMHUB, 0, mmDAGB0_CNTL_MISC2_Sienna_Cichlid); break; @@ -578,6 +606,7 @@ static void mmhub_v2_0_update_medium_grain_clock_gating(struct amdgpu_device *ad case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: if (def != data) WREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG_Sienna_Cichlid, data); if (def1 != data1) @@ -601,6 +630,7 @@ static void mmhub_v2_0_update_medium_grain_light_sleep(struct amdgpu_device *ade case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: def = data = RREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG_Sienna_Cichlid); break; default: @@ -618,6 +648,7 @@ static void mmhub_v2_0_update_medium_grain_light_sleep(struct amdgpu_device *ade case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: WREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG_Sienna_Cichlid, data); break; default: @@ -640,6 +671,7 @@ static int mmhub_v2_0_set_clockgating(struct amdgpu_device *adev, case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: mmhub_v2_0_update_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE); mmhub_v2_0_update_medium_grain_light_sleep(adev, @@ -663,6 +695,7 @@ static void mmhub_v2_0_get_clockgating(struct amdgpu_device *adev, u32 *flags) case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: data = RREG32_SOC15(MMHUB, 0, mmMM_ATC_L2_MISC_CG_Sienna_Cichlid); data1 = RREG32_SOC15(MMHUB, 0, mmDAGB0_CNTL_MISC2_Sienna_Cichlid); break; diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c index a9899335d0b1..709ac576ac7e 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c @@ -569,9 +569,9 @@ static int mmhub_v2_3_set_clockgating(struct amdgpu_device *adev, return 0; mmhub_v2_3_update_medium_grain_clock_gating(adev, - state == AMD_CG_STATE_GATE ? true : false); + state == AMD_CG_STATE_GATE); mmhub_v2_3_update_medium_grain_light_sleep(adev, - state == AMD_CG_STATE_GATE ? true : false); + state == AMD_CG_STATE_GATE); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/navi10_ih.c b/drivers/gpu/drm/amd/amdgpu/navi10_ih.c index f4e4040bbd25..17ed61a34fbf 100644 --- a/drivers/gpu/drm/amd/amdgpu/navi10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/navi10_ih.c @@ -151,7 +151,15 @@ static int navi10_ih_toggle_ring_interrupts(struct amdgpu_device *adev, /* enable_intr field is only valid in ring0 */ if (ih == &adev->irq.ih) tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0)); - WREG32(ih_regs->ih_rb_cntl, tmp); + + if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) { + if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp)) { + DRM_ERROR("PSP program IH_RB_CNTL failed!\n"); + return -ETIMEDOUT; + } + } else { + WREG32(ih_regs->ih_rb_cntl, tmp); + } if (enable) { ih->enabled = true; @@ -261,7 +269,15 @@ static int navi10_ih_enable_ring(struct amdgpu_device *adev, tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_ENABLE, 0); tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_FULL_DRAIN_ENABLE, 1); } - WREG32(ih_regs->ih_rb_cntl, tmp); + + if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) { + if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp)) { + DRM_ERROR("PSP program IH_RB_CNTL failed!\n"); + return -ETIMEDOUT; + } + } else { + WREG32(ih_regs->ih_rb_cntl, tmp); + } if (ih == &adev->irq.ih) { /* set the ih ring 0 writeback address whether it's enabled or not */ @@ -311,6 +327,7 @@ static int navi10_ih_irq_init(struct amdgpu_device *adev) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: ih_chicken = RREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN_Sienna_Cichlid); ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN, MC_SPACE_GPA_ENABLE, 1); diff --git a/drivers/gpu/drm/amd/amdgpu/nv.c b/drivers/gpu/drm/amd/amdgpu/nv.c index d54af7f8801b..ec5f835f60f2 100644 --- a/drivers/gpu/drm/amd/amdgpu/nv.c +++ b/drivers/gpu/drm/amd/amdgpu/nv.c @@ -218,11 +218,114 @@ static const struct amdgpu_video_codecs sc_video_codecs_decode = .codec_array = sc_video_codecs_decode_array, }; +/* SRIOV Sienna Cichlid, not const since data is controlled by host */ +static struct amdgpu_video_codec_info sriov_sc_video_codecs_encode_array[] = +{ + { + .codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, + .max_width = 4096, + .max_height = 2304, + .max_pixels_per_frame = 4096 * 2304, + .max_level = 0, + }, + { + .codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, + .max_width = 4096, + .max_height = 2304, + .max_pixels_per_frame = 4096 * 2304, + .max_level = 0, + }, +}; + +static struct amdgpu_video_codec_info sriov_sc_video_codecs_decode_array[] = +{ + { + .codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, + .max_width = 4096, + .max_height = 4096, + .max_pixels_per_frame = 4096 * 4096, + .max_level = 3, + }, + { + .codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, + .max_width = 4096, + .max_height = 4096, + .max_pixels_per_frame = 4096 * 4096, + .max_level = 5, + }, + { + .codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, + .max_width = 4096, + .max_height = 4096, + .max_pixels_per_frame = 4096 * 4096, + .max_level = 52, + }, + { + .codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, + .max_width = 4096, + .max_height = 4096, + .max_pixels_per_frame = 4096 * 4096, + .max_level = 4, + }, + { + .codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, + .max_width = 8192, + .max_height = 4352, + .max_pixels_per_frame = 8192 * 4352, + .max_level = 186, + }, + { + .codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, + .max_width = 4096, + .max_height = 4096, + .max_pixels_per_frame = 4096 * 4096, + .max_level = 0, + }, + { + .codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, + .max_width = 8192, + .max_height = 4352, + .max_pixels_per_frame = 8192 * 4352, + .max_level = 0, + }, + { + .codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, + .max_width = 8192, + .max_height = 4352, + .max_pixels_per_frame = 8192 * 4352, + .max_level = 0, + }, +}; + +static struct amdgpu_video_codecs sriov_sc_video_codecs_encode = +{ + .codec_count = ARRAY_SIZE(sriov_sc_video_codecs_encode_array), + .codec_array = sriov_sc_video_codecs_encode_array, +}; + +static struct amdgpu_video_codecs sriov_sc_video_codecs_decode = +{ + .codec_count = ARRAY_SIZE(sriov_sc_video_codecs_decode_array), + .codec_array = sriov_sc_video_codecs_decode_array, +}; + static int nv_query_video_codecs(struct amdgpu_device *adev, bool encode, const struct amdgpu_video_codecs **codecs) { switch (adev->asic_type) { case CHIP_SIENNA_CICHLID: + if (amdgpu_sriov_vf(adev)) { + if (encode) + *codecs = &sriov_sc_video_codecs_encode; + else + *codecs = &sriov_sc_video_codecs_decode; + } else { + if (encode) + *codecs = &nv_video_codecs_encode; + else + *codecs = &sc_video_codecs_decode; + } + return 0; case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: case CHIP_VANGOGH: @@ -534,6 +637,7 @@ nv_asic_reset_method(struct amdgpu_device *adev) case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: return AMD_RESET_METHOD_MODE1; default: if (amdgpu_dpm_is_baco_supported(adev)) @@ -598,7 +702,7 @@ static void nv_pcie_gen3_enable(struct amdgpu_device *adev) static void nv_program_aspm(struct amdgpu_device *adev) { - if (amdgpu_aspm != 1) + if (!amdgpu_aspm) return; if (!(adev->flags & AMD_IS_APU) && @@ -623,6 +727,16 @@ static const struct amdgpu_ip_block_version nv_common_ip_block = .funcs = &nv_common_ip_funcs, }; +static bool nv_is_headless_sku(struct pci_dev *pdev) +{ + if ((pdev->device == 0x731E && + (pdev->revision == 0xC6 || pdev->revision == 0xC7)) || + (pdev->device == 0x7340 && pdev->revision == 0xC9) || + (pdev->device == 0x7360 && pdev->revision == 0xC7)) + return true; + return false; +} + static int nv_reg_base_init(struct amdgpu_device *adev) { int r; @@ -635,6 +749,12 @@ static int nv_reg_base_init(struct amdgpu_device *adev) goto legacy_init; } + amdgpu_discovery_harvest_ip(adev); + if (nv_is_headless_sku(adev->pdev)) { + adev->harvest_ip_mask |= AMD_HARVEST_IP_VCN_MASK; + adev->harvest_ip_mask |= AMD_HARVEST_IP_JPEG_MASK; + } + return 0; } @@ -659,6 +779,9 @@ legacy_init: case CHIP_DIMGREY_CAVEFISH: dimgrey_cavefish_reg_base_init(adev); break; + case CHIP_BEIGE_GOBY: + beige_goby_reg_base_init(adev); + break; default: return -EINVAL; } @@ -671,16 +794,6 @@ void nv_set_virt_ops(struct amdgpu_device *adev) adev->virt.ops = &xgpu_nv_virt_ops; } -static bool nv_is_headless_sku(struct pci_dev *pdev) -{ - if ((pdev->device == 0x731E && - (pdev->revision == 0xC6 || pdev->revision == 0xC7)) || - (pdev->device == 0x7340 && pdev->revision == 0xC9) || - (pdev->device == 0x7360 && pdev->revision == 0xC7)) - return true; - return false; -} - int nv_set_ip_blocks(struct amdgpu_device *adev) { int r; @@ -728,8 +841,7 @@ int nv_set_ip_blocks(struct amdgpu_device *adev) if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT && !amdgpu_sriov_vf(adev)) amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block); - if (!nv_is_headless_sku(adev->pdev)) - amdgpu_device_ip_block_add(adev, &vcn_v2_0_ip_block); + amdgpu_device_ip_block_add(adev, &vcn_v2_0_ip_block); amdgpu_device_ip_block_add(adev, &jpeg_v2_0_ip_block); if (adev->enable_mes) amdgpu_device_ip_block_add(adev, &mes_v10_1_ip_block); @@ -752,17 +864,22 @@ int nv_set_ip_blocks(struct amdgpu_device *adev) if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT && !amdgpu_sriov_vf(adev)) amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block); - if (!nv_is_headless_sku(adev->pdev)) - amdgpu_device_ip_block_add(adev, &vcn_v2_0_ip_block); + amdgpu_device_ip_block_add(adev, &vcn_v2_0_ip_block); if (!amdgpu_sriov_vf(adev)) amdgpu_device_ip_block_add(adev, &jpeg_v2_0_ip_block); break; case CHIP_SIENNA_CICHLID: amdgpu_device_ip_block_add(adev, &nv_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v10_0_ip_block); - amdgpu_device_ip_block_add(adev, &navi10_ih_ip_block); - if (likely(adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)) - amdgpu_device_ip_block_add(adev, &psp_v11_0_ip_block); + if (!amdgpu_sriov_vf(adev)) { + amdgpu_device_ip_block_add(adev, &navi10_ih_ip_block); + if (likely(adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)) + amdgpu_device_ip_block_add(adev, &psp_v11_0_ip_block); + } else { + if (likely(adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)) + amdgpu_device_ip_block_add(adev, &psp_v11_0_ip_block); + amdgpu_device_ip_block_add(adev, &navi10_ih_ip_block); + } if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP && is_support_sw_smu(adev)) amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block); @@ -777,7 +894,6 @@ int nv_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &vcn_v3_0_ip_block); if (!amdgpu_sriov_vf(adev)) amdgpu_device_ip_block_add(adev, &jpeg_v3_0_ip_block); - if (adev->enable_mes) amdgpu_device_ip_block_add(adev, &mes_v10_1_ip_block); break; @@ -842,6 +958,28 @@ int nv_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &vcn_v3_0_ip_block); amdgpu_device_ip_block_add(adev, &jpeg_v3_0_ip_block); break; + case CHIP_BEIGE_GOBY: + amdgpu_device_ip_block_add(adev, &nv_common_ip_block); + amdgpu_device_ip_block_add(adev, &gmc_v10_0_ip_block); + amdgpu_device_ip_block_add(adev, &navi10_ih_ip_block); + if (likely(adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)) + amdgpu_device_ip_block_add(adev, &psp_v11_0_ip_block); + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP && + is_support_sw_smu(adev)) + amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v10_0_ip_block); + amdgpu_device_ip_block_add(adev, &sdma_v5_2_ip_block); + if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) + amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_device_ip_block_add(adev, &dm_ip_block); +#endif + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT && + is_support_sw_smu(adev)) + amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block); + amdgpu_device_ip_block_add(adev, &vcn_v3_0_ip_block); + break; default: return -EINVAL; } @@ -1065,6 +1203,7 @@ static int nv_common_early_init(void *handle) case CHIP_SIENNA_CICHLID: adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_MC_MGCG | AMD_CG_SUPPORT_VCN_MGCG | @@ -1088,6 +1227,7 @@ static int nv_common_early_init(void *handle) case CHIP_NAVY_FLOUNDER: adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_VCN_MGCG | AMD_CG_SUPPORT_JPEG_MGCG | @@ -1118,6 +1258,8 @@ static int nv_common_early_init(void *handle) AMD_CG_SUPPORT_MC_LS | AMD_CG_SUPPORT_GFX_FGCG | AMD_CG_SUPPORT_VCN_MGCG | + AMD_CG_SUPPORT_SDMA_MGCG | + AMD_CG_SUPPORT_SDMA_LS | AMD_CG_SUPPORT_JPEG_MGCG; adev->pg_flags = AMD_PG_SUPPORT_GFX_PG | AMD_PG_SUPPORT_VCN | @@ -1129,6 +1271,7 @@ static int nv_common_early_init(void *handle) case CHIP_DIMGREY_CAVEFISH: adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_VCN_MGCG | AMD_CG_SUPPORT_JPEG_MGCG | @@ -1144,11 +1287,33 @@ static int nv_common_early_init(void *handle) AMD_PG_SUPPORT_MMHUB; adev->external_rev_id = adev->rev_id + 0x3c; break; + case CHIP_BEIGE_GOBY: + adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_GFX_3D_CGCG | + AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_MC_LS | + AMD_CG_SUPPORT_HDP_MGCG | + AMD_CG_SUPPORT_HDP_LS | + AMD_CG_SUPPORT_IH_CG | + AMD_CG_SUPPORT_VCN_MGCG; + adev->pg_flags = AMD_PG_SUPPORT_VCN | + AMD_PG_SUPPORT_VCN_DPG | + AMD_PG_SUPPORT_ATHUB | + AMD_PG_SUPPORT_MMHUB; + adev->external_rev_id = adev->rev_id + 0x46; + break; default: /* FIXME: not supported yet */ return -EINVAL; } + if (adev->harvest_ip_mask & AMD_HARVEST_IP_VCN_MASK) + adev->pg_flags &= ~(AMD_PG_SUPPORT_VCN | + AMD_PG_SUPPORT_VCN_DPG | + AMD_PG_SUPPORT_JPEG); + if (amdgpu_sriov_vf(adev)) { amdgpu_virt_init_setting(adev); xgpu_nv_mailbox_set_irq_funcs(adev); @@ -1161,8 +1326,12 @@ static int nv_common_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (amdgpu_sriov_vf(adev)) + if (amdgpu_sriov_vf(adev)) { xgpu_nv_mailbox_get_irq(adev); + amdgpu_virt_update_sriov_video_codec(adev, + sriov_sc_video_codecs_encode_array, ARRAY_SIZE(sriov_sc_video_codecs_encode_array), + sriov_sc_video_codecs_decode_array, ARRAY_SIZE(sriov_sc_video_codecs_decode_array)); + } return 0; } @@ -1258,6 +1427,7 @@ static int nv_common_set_clockgating_state(void *handle, case CHIP_SIENNA_CICHLID: case CHIP_NAVY_FLOUNDER: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: adev->nbio.funcs->update_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE); adev->nbio.funcs->update_medium_grain_light_sleep(adev, diff --git a/drivers/gpu/drm/amd/amdgpu/nv.h b/drivers/gpu/drm/amd/amdgpu/nv.h index 515d67bf249f..330438388fde 100644 --- a/drivers/gpu/drm/amd/amdgpu/nv.h +++ b/drivers/gpu/drm/amd/amdgpu/nv.h @@ -36,4 +36,5 @@ int navi12_reg_base_init(struct amdgpu_device *adev); int sienna_cichlid_reg_base_init(struct amdgpu_device *adev); void vangogh_reg_base_init(struct amdgpu_device *adev); int dimgrey_cavefish_reg_base_init(struct amdgpu_device *adev); +int beige_goby_reg_base_init(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h index 96064c343163..f6d3180febc4 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h @@ -97,7 +97,6 @@ enum psp_gfx_cmd_id GFX_CMD_ID_SETUP_VMR = 0x00000009, /* setup VMR region */ GFX_CMD_ID_DESTROY_VMR = 0x0000000A, /* destroy VMR region */ GFX_CMD_ID_PROG_REG = 0x0000000B, /* program regs */ - GFX_CMD_ID_CLEAR_VF_FW = 0x0000000D, /* Clear VF FW, to be used on VF shutdown. */ GFX_CMD_ID_GET_FW_ATTESTATION = 0x0000000F, /* Query GPUVA of the Fw Attestation DB */ /* IDs upto 0x1F are reserved for older programs (Raven, Vega 10/12/20) */ GFX_CMD_ID_LOAD_TOC = 0x00000020, /* Load TOC and obtain TMR size */ diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c index 589410c32d09..1f2e7e35c91e 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c @@ -63,6 +63,8 @@ MODULE_FIRMWARE("amdgpu/vangogh_asd.bin"); MODULE_FIRMWARE("amdgpu/vangogh_toc.bin"); MODULE_FIRMWARE("amdgpu/dimgrey_cavefish_sos.bin"); MODULE_FIRMWARE("amdgpu/dimgrey_cavefish_ta.bin"); +MODULE_FIRMWARE("amdgpu/beige_goby_sos.bin"); +MODULE_FIRMWARE("amdgpu/beige_goby_ta.bin"); /* address block */ #define smnMP1_FIRMWARE_FLAGS 0x3010024 @@ -115,6 +117,9 @@ static int psp_v11_0_init_microcode(struct psp_context *psp) case CHIP_DIMGREY_CAVEFISH: chip_name = "dimgrey_cavefish"; break; + case CHIP_BEIGE_GOBY: + chip_name = "beige_goby"; + break; default: BUG(); } @@ -200,6 +205,14 @@ static int psp_v11_0_init_microcode(struct psp_context *psp) if (err) return err; break; + case CHIP_BEIGE_GOBY: + err = psp_init_sos_microcode(psp, chip_name); + if (err) + return err; + err = psp_init_ta_microcode(psp, chip_name); + if (err) + return err; + break; case CHIP_VANGOGH: err = psp_init_asd_microcode(psp, chip_name); if (err) diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 5715be6770ec..d197185f7789 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -1109,6 +1109,8 @@ static void sdma_v4_0_ctx_switch_enable(struct amdgpu_device *adev, bool enable) if (adev->asic_type == CHIP_ARCTURUS && adev->sdma.instance[i].fw_version >= 14) WREG32_SDMA(i, mmSDMA0_PUB_DUMMY_REG2, enable); + /* Extend page fault timeout to avoid interrupt storm */ + WREG32_SDMA(i, mmSDMA0_UTCL1_TIMEOUT, 0x00800080); } } @@ -2227,7 +2229,7 @@ static int sdma_v4_0_print_iv_entry(struct amdgpu_device *adev, memset(&task_info, 0, sizeof(struct amdgpu_task_info)); amdgpu_vm_get_task_info(adev, entry->pasid, &task_info); - dev_info(adev->dev, + dev_dbg_ratelimited(adev->dev, "[sdma%d] address:0x%016llx src_id:%u ring:%u vmid:%u " "pasid:%u, for process %s pid %d thread %s pid %d\n", instance, addr, entry->src_id, entry->ring_id, entry->vmid, @@ -2240,7 +2242,7 @@ static int sdma_v4_0_process_vm_hole_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - dev_err(adev->dev, "MC or SEM address in VM hole\n"); + dev_dbg_ratelimited(adev->dev, "MC or SEM address in VM hole\n"); sdma_v4_0_print_iv_entry(adev, entry); return 0; } @@ -2249,7 +2251,7 @@ static int sdma_v4_0_process_doorbell_invalid_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - dev_err(adev->dev, "SDMA received a doorbell from BIF with byte_enable !=0xff\n"); + dev_dbg_ratelimited(adev->dev, "SDMA received a doorbell from BIF with byte_enable !=0xff\n"); sdma_v4_0_print_iv_entry(adev, entry); return 0; } @@ -2258,7 +2260,7 @@ static int sdma_v4_0_process_pool_timeout_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - dev_err(adev->dev, + dev_dbg_ratelimited(adev->dev, "Polling register/memory timeout executing POLL_REG/MEM with finite timer\n"); sdma_v4_0_print_iv_entry(adev, entry); return 0; @@ -2268,7 +2270,7 @@ static int sdma_v4_0_process_srbm_write_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - dev_err(adev->dev, + dev_dbg_ratelimited(adev->dev, "SDMA gets an Register Write SRBM_WRITE command in non-privilege command buffer\n"); sdma_v4_0_print_iv_entry(adev, entry); return 0; @@ -2597,27 +2599,18 @@ static const struct amdgpu_irq_src_funcs sdma_v4_0_srbm_write_irq_funcs = { static void sdma_v4_0_set_irq_funcs(struct amdgpu_device *adev) { + adev->sdma.trap_irq.num_types = adev->sdma.num_instances; + adev->sdma.ecc_irq.num_types = adev->sdma.num_instances; + /*For Arcturus and Aldebaran, add another 4 irq handler*/ switch (adev->sdma.num_instances) { - case 1: - adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE1; - adev->sdma.ecc_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE1; - break; case 5: - adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE5; - adev->sdma.ecc_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE5; - break; case 8: - adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_LAST; - adev->sdma.ecc_irq.num_types = AMDGPU_SDMA_IRQ_LAST; - adev->sdma.vm_hole_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE5; - adev->sdma.doorbell_invalid_irq.num_types = AMDGPU_SDMA_IRQ_LAST; - adev->sdma.pool_timeout_irq.num_types = AMDGPU_SDMA_IRQ_LAST; - adev->sdma.srbm_write_irq.num_types = AMDGPU_SDMA_IRQ_LAST; + adev->sdma.vm_hole_irq.num_types = adev->sdma.num_instances; + adev->sdma.doorbell_invalid_irq.num_types = adev->sdma.num_instances; + adev->sdma.pool_timeout_irq.num_types = adev->sdma.num_instances; + adev->sdma.srbm_write_irq.num_types = adev->sdma.num_instances; break; - case 2: default: - adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE2; - adev->sdma.ecc_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE2; break; } adev->sdma.trap_irq.funcs = &sdma_v4_0_trap_irq_funcs; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c index 920fc6d4a127..75d7310f8439 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c @@ -123,6 +123,10 @@ static const struct soc15_reg_golden golden_settings_sdma_nv14[] = { static const struct soc15_reg_golden golden_settings_sdma_nv12[] = { SOC15_REG_GOLDEN_VALUE(GC, 0, mmSDMA0_RLC3_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSDMA0_GB_ADDR_CONFIG, 0x001877ff, 0x00000044), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSDMA0_GB_ADDR_CONFIG_READ, 0x001877ff, 0x00000044), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSDMA1_GB_ADDR_CONFIG, 0x001877ff, 0x00000044), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSDMA1_GB_ADDR_CONFIG_READ, 0x001877ff, 0x00000044), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSDMA1_RLC3_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), }; @@ -405,18 +409,6 @@ static void sdma_v5_0_ring_emit_ib(struct amdgpu_ring *ring, unsigned vmid = AMDGPU_JOB_GET_VMID(job); uint64_t csa_mc_addr = amdgpu_sdma_get_csa_mc_addr(ring, vmid); - /* Invalidate L2, because if we don't do it, we might get stale cache - * lines from previous IBs. - */ - amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_GCR_REQ)); - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, (SDMA_GCR_GL2_INV | - SDMA_GCR_GL2_WB | - SDMA_GCR_GLM_INV | - SDMA_GCR_GLM_WB) << 16); - amdgpu_ring_write(ring, 0xffffff80); - amdgpu_ring_write(ring, 0xffff); - /* An IB packet must end on a 8 DW boundary--the next dword * must be on a 8-dword boundary. Our IB packet below is 6 * dwords long, thus add x number of NOPs, such that, in @@ -438,6 +430,33 @@ static void sdma_v5_0_ring_emit_ib(struct amdgpu_ring *ring, } /** + * sdma_v5_0_ring_emit_mem_sync - flush the IB by graphics cache rinse + * + * @ring: amdgpu ring pointer + * @job: job to retrieve vmid from + * @ib: IB object to schedule + * + * flush the IB by graphics cache rinse. + */ +static void sdma_v5_0_ring_emit_mem_sync(struct amdgpu_ring *ring) +{ + uint32_t gcr_cntl = + SDMA_GCR_GL2_INV | SDMA_GCR_GL2_WB | SDMA_GCR_GLM_INV | + SDMA_GCR_GL1_INV | SDMA_GCR_GLV_INV | SDMA_GCR_GLK_INV | + SDMA_GCR_GLI_INV(1); + + /* flush entire cache L0/L1/L2, this can be optimized by performance requirement */ + amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_GCR_REQ)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD1_BASE_VA_31_7(0)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD2_GCR_CONTROL_15_0(gcr_cntl) | + SDMA_PKT_GCR_REQ_PAYLOAD2_BASE_VA_47_32(0)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD3_LIMIT_VA_31_7(0) | + SDMA_PKT_GCR_REQ_PAYLOAD3_GCR_CONTROL_18_16(gcr_cntl >> 16)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD4_LIMIT_VA_47_32(0) | + SDMA_PKT_GCR_REQ_PAYLOAD4_VMID(0)); +} + +/** * sdma_v5_0_ring_emit_hdp_flush - emit an hdp flush on the DMA ring * * @ring: amdgpu ring pointer @@ -1643,6 +1662,7 @@ static const struct amdgpu_ring_funcs sdma_v5_0_ring_funcs = { 10 + 10 + 10, /* sdma_v5_0_ring_emit_fence x3 for user fence, vm fence */ .emit_ib_size = 5 + 7 + 6, /* sdma_v5_0_ring_emit_ib */ .emit_ib = sdma_v5_0_ring_emit_ib, + .emit_mem_sync = sdma_v5_0_ring_emit_mem_sync, .emit_fence = sdma_v5_0_ring_emit_fence, .emit_pipeline_sync = sdma_v5_0_ring_emit_pipeline_sync, .emit_vm_flush = sdma_v5_0_ring_emit_vm_flush, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c index b1ad9e52b234..ecb82c39b106 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c @@ -47,6 +47,7 @@ MODULE_FIRMWARE("amdgpu/sienna_cichlid_sdma.bin"); MODULE_FIRMWARE("amdgpu/navy_flounder_sdma.bin"); MODULE_FIRMWARE("amdgpu/dimgrey_cavefish_sdma.bin"); +MODULE_FIRMWARE("amdgpu/beige_goby_sdma.bin"); MODULE_FIRMWARE("amdgpu/vangogh_sdma.bin"); @@ -92,6 +93,7 @@ static void sdma_v5_2_init_golden_registers(struct amdgpu_device *adev) case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: break; default: break; @@ -163,6 +165,9 @@ static int sdma_v5_2_init_microcode(struct amdgpu_device *adev) case CHIP_DIMGREY_CAVEFISH: chip_name = "dimgrey_cavefish"; break; + case CHIP_BEIGE_GOBY: + chip_name = "beige_goby"; + break; default: BUG(); } @@ -497,11 +502,6 @@ static void sdma_v5_2_gfx_stop(struct amdgpu_device *adev) ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 0); WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL), ib_cntl); } - - sdma0->sched.ready = false; - sdma1->sched.ready = false; - sdma2->sched.ready = false; - sdma3->sched.ready = false; } /** @@ -1228,6 +1228,7 @@ static int sdma_v5_2_early_init(void *handle) adev->sdma.num_instances = 2; break; case CHIP_VANGOGH: + case CHIP_BEIGE_GOBY: adev->sdma.num_instances = 1; break; default: @@ -1556,6 +1557,10 @@ static void sdma_v5_2_update_medium_grain_clock_gating(struct amdgpu_device *ade int i; for (i = 0; i < adev->sdma.num_instances; i++) { + + if (adev->sdma.instance[i].fw_version < 70 && adev->asic_type == CHIP_VANGOGH) + adev->cg_flags &= ~AMD_CG_SUPPORT_SDMA_MGCG; + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG)) { /* Enable sdma clock gating */ def = data = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CLK_CTRL)); @@ -1589,6 +1594,10 @@ static void sdma_v5_2_update_medium_grain_light_sleep(struct amdgpu_device *adev int i; for (i = 0; i < adev->sdma.num_instances; i++) { + + if (adev->sdma.instance[i].fw_version < 70 && adev->asic_type == CHIP_VANGOGH) + adev->cg_flags &= ~AMD_CG_SUPPORT_SDMA_LS; + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS)) { /* Enable sdma mem light sleep */ def = data = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_POWER_CNTL)); @@ -1620,6 +1629,7 @@ static int sdma_v5_2_set_clockgating_state(void *handle, case CHIP_NAVY_FLOUNDER: case CHIP_VANGOGH: case CHIP_DIMGREY_CAVEFISH: + case CHIP_BEIGE_GOBY: sdma_v5_2_update_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE); sdma_v5_2_update_medium_grain_light_sleep(adev, diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c index 079b094c48ad..3c47c94846d6 100644 --- a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c +++ b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c @@ -89,6 +89,23 @@ static u32 smuio_v13_0_get_die_id(struct amdgpu_device *adev) } /** + * smuio_v13_0_get_socket_id - query socket id from FCH + * + * @adev: amdgpu device pointer + * + * Returns socket id + */ +static u32 smuio_v13_0_get_socket_id(struct amdgpu_device *adev) +{ + u32 data, socket_id; + + data = RREG32_SOC15(SMUIO, 0, regSMUIO_MCM_CONFIG); + socket_id = REG_GET_FIELD(data, SMUIO_MCM_CONFIG, SOCKET_ID); + + return socket_id; +} + +/** * smuio_v13_0_supports_host_gpu_xgmi - detect xgmi interface between cpu and gpu/s. * * @adev: amdgpu device pointer @@ -115,6 +132,7 @@ const struct amdgpu_smuio_funcs smuio_v13_0_funcs = { .get_rom_index_offset = smuio_v13_0_get_rom_index_offset, .get_rom_data_offset = smuio_v13_0_get_rom_data_offset, .get_die_id = smuio_v13_0_get_die_id, + .get_socket_id = smuio_v13_0_get_socket_id, .is_host_gpu_xgmi_supported = smuio_v13_0_is_host_gpu_xgmi_supported, .update_rom_clock_gating = smuio_v13_0_update_rom_clock_gating, .get_clock_gating_state = smuio_v13_0_get_clock_gating_state, diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index d80e12b80c7e..75008cc5f189 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -302,6 +302,7 @@ static int soc15_query_video_codecs(struct amdgpu_device *adev, bool encode, *codecs = &rv_video_codecs_decode; return 0; case CHIP_ARCTURUS: + case CHIP_ALDEBARAN: case CHIP_RENOIR: if (encode) *codecs = &vega_video_codecs_encode; @@ -655,7 +656,7 @@ static int soc15_asic_baco_reset(struct amdgpu_device *adev) int ret = 0; /* avoid NBIF got stuck when do RAS recovery in BACO reset */ - if (ras && ras->supported) + if (ras && adev->ras_enabled) adev->nbio.funcs->enable_doorbell_interrupt(adev, false); ret = amdgpu_dpm_baco_reset(adev); @@ -663,7 +664,7 @@ static int soc15_asic_baco_reset(struct amdgpu_device *adev) return ret; /* re-enable doorbell interrupt after BACO exit */ - if (ras && ras->supported) + if (ras && adev->ras_enabled) adev->nbio.funcs->enable_doorbell_interrupt(adev, true); return 0; @@ -710,7 +711,8 @@ soc15_asic_reset_method(struct amdgpu_device *adev) * 1. PMFW version > 0x284300: all cases use baco * 2. PMFW version <= 0x284300: only sGPU w/o RAS use baco */ - if ((ras && ras->supported) && adev->pm.fw_version <= 0x283400) + if (ras && adev->ras_enabled && + adev->pm.fw_version <= 0x283400) baco_reset = false; break; case CHIP_ALDEBARAN: @@ -816,7 +818,7 @@ static void soc15_pcie_gen3_enable(struct amdgpu_device *adev) static void soc15_program_aspm(struct amdgpu_device *adev) { - if (amdgpu_aspm != 1) + if (!amdgpu_aspm) return; if (!(adev->flags & AMD_IS_APU) && @@ -1392,7 +1394,6 @@ static int soc15_common_early_init(void *handle) adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS | AMD_CG_SUPPORT_GFX_CP_LS | - AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_GFX_3D_CGLS | AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_CGLS | @@ -1401,7 +1402,8 @@ static int soc15_common_early_init(void *handle) AMD_CG_SUPPORT_MC_MGCG | AMD_CG_SUPPORT_MC_LS | AMD_CG_SUPPORT_SDMA_MGCG | - AMD_CG_SUPPORT_SDMA_LS; + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_VCN_MGCG; adev->pg_flags = AMD_PG_SUPPORT_SDMA | AMD_PG_SUPPORT_MMHUB | @@ -1411,7 +1413,6 @@ static int soc15_common_early_init(void *handle) AMD_CG_SUPPORT_GFX_MGLS | AMD_CG_SUPPORT_GFX_RLC_LS | AMD_CG_SUPPORT_GFX_CP_LS | - AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_GFX_3D_CGLS | AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_CGLS | @@ -1521,9 +1522,6 @@ static int soc15_common_late_init(void *handle) if (amdgpu_sriov_vf(adev)) xgpu_ai_mailbox_get_irq(adev); - if (adev->hdp.funcs->reset_ras_error_count) - adev->hdp.funcs->reset_ras_error_count(adev); - if (adev->nbio.ras_funcs && adev->nbio.ras_funcs->ras_late_init) r = adev->nbio.ras_funcs->ras_late_init(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h b/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h index 745ed0fba1ed..0f214a398dd8 100644 --- a/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h +++ b/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h @@ -105,6 +105,12 @@ struct ta_ras_trigger_error_input { uint64_t value; // method if error injection. i.e persistent, coherent etc. }; +struct ta_ras_init_flags +{ + uint8_t poison_mode_en; + uint8_t dgpu_mode; +}; + struct ta_ras_output_flags { uint8_t ras_init_success_flag; @@ -115,6 +121,7 @@ struct ta_ras_output_flags /* Common input structure for RAS callbacks */ /**********************************************************/ union ta_ras_cmd_input { + struct ta_ras_init_flags init_flags; struct ta_ras_enable_features_input enable_features; struct ta_ras_disable_features_input disable_features; struct ta_ras_trigger_error_input trigger_error; diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index 0cd98fcb1f9f..939bcfa2a4ec 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -363,6 +363,7 @@ static int uvd_v7_0_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) error: dma_fence_put(fence); + amdgpu_bo_unpin(bo); amdgpu_bo_unreserve(bo); amdgpu_bo_unref(&bo); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c index 51a773a37a35..0c1beefa3e49 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c @@ -1119,10 +1119,10 @@ static int vcn_v1_0_stop_spg_mode(struct amdgpu_device *adev) UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK; SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_LMI_STATUS, tmp, tmp); - /* put VCPU into reset */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), - UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK, - ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); + /* stall UMC channel */ + WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), + UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, + ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK | UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK; @@ -1141,6 +1141,11 @@ static int vcn_v1_0_stop_spg_mode(struct amdgpu_device *adev) UVD_SOFT_RESET__LMI_SOFT_RESET_MASK, ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK); + /* put VCPU into reset */ + WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), + UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK, + ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); + WREG32_SOC15(UVD, 0, mmUVD_STATUS, 0); vcn_v1_0_enable_clock_gating(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c index cf165ab5dd26..946335d0f19c 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c @@ -91,6 +91,11 @@ static int vcn_v3_0_early_init(void *handle) adev->vcn.harvest_config = 0; adev->vcn.num_enc_rings = 1; + if (adev->asic_type == CHIP_BEIGE_GOBY) { + adev->vcn.num_vcn_inst = 1; + adev->vcn.num_enc_rings = 0; + } + } else { if (adev->asic_type == CHIP_SIENNA_CICHLID) { u32 harvest; @@ -110,7 +115,10 @@ static int vcn_v3_0_early_init(void *handle) } else adev->vcn.num_vcn_inst = 1; - adev->vcn.num_enc_rings = 2; + if (adev->asic_type == CHIP_BEIGE_GOBY) + adev->vcn.num_enc_rings = 0; + else + adev->vcn.num_enc_rings = 2; } vcn_v3_0_set_dec_ring_funcs(adev); @@ -373,7 +381,7 @@ static int vcn_v3_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_ring *ring; - int i, j; + int i; for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { if (adev->vcn.harvest_config & (1 << i)) @@ -388,12 +396,6 @@ static int vcn_v3_0_hw_fini(void *handle) vcn_v3_0_set_powergating_state(adev, AMD_PG_STATE_GATE); } } - ring->sched.ready = false; - - for (j = 0; j < adev->vcn.num_enc_rings; ++j) { - ring = &adev->vcn.inst[i].ring_enc[j]; - ring->sched.ready = false; - } } return 0; @@ -1261,23 +1263,25 @@ static int vcn_v3_0_start(struct amdgpu_device *adev) fw_shared->rb.wptr = lower_32_bits(ring->wptr); fw_shared->multi_queue.decode_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET); - fw_shared->multi_queue.encode_generalpurpose_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET); - ring = &adev->vcn.inst[i].ring_enc[0]; - WREG32_SOC15(VCN, i, mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); - WREG32_SOC15(VCN, i, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); - WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO, ring->gpu_addr); - WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); - WREG32_SOC15(VCN, i, mmUVD_RB_SIZE, ring->ring_size / 4); - fw_shared->multi_queue.encode_generalpurpose_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET); - - fw_shared->multi_queue.encode_lowlatency_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET); - ring = &adev->vcn.inst[i].ring_enc[1]; - WREG32_SOC15(VCN, i, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); - WREG32_SOC15(VCN, i, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); - WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO2, ring->gpu_addr); - WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); - WREG32_SOC15(VCN, i, mmUVD_RB_SIZE2, ring->ring_size / 4); - fw_shared->multi_queue.encode_lowlatency_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET); + if (adev->asic_type != CHIP_BEIGE_GOBY) { + fw_shared->multi_queue.encode_generalpurpose_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET); + ring = &adev->vcn.inst[i].ring_enc[0]; + WREG32_SOC15(VCN, i, mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); + WREG32_SOC15(VCN, i, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); + WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO, ring->gpu_addr); + WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(VCN, i, mmUVD_RB_SIZE, ring->ring_size / 4); + fw_shared->multi_queue.encode_generalpurpose_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET); + + fw_shared->multi_queue.encode_lowlatency_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET); + ring = &adev->vcn.inst[i].ring_enc[1]; + WREG32_SOC15(VCN, i, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); + WREG32_SOC15(VCN, i, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); + WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO2, ring->gpu_addr); + WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(VCN, i, mmUVD_RB_SIZE2, ring->ring_size / 4); + fw_shared->multi_queue.encode_lowlatency_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET); + } } return 0; @@ -1657,31 +1661,33 @@ static int vcn_v3_0_pause_dpg_mode(struct amdgpu_device *adev, UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK, ~UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK); - /* Restore */ - fw_shared = adev->vcn.inst[inst_idx].fw_shared_cpu_addr; - fw_shared->multi_queue.encode_generalpurpose_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET); - ring = &adev->vcn.inst[inst_idx].ring_enc[0]; - ring->wptr = 0; - WREG32_SOC15(VCN, inst_idx, mmUVD_RB_BASE_LO, ring->gpu_addr); - WREG32_SOC15(VCN, inst_idx, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); - WREG32_SOC15(VCN, inst_idx, mmUVD_RB_SIZE, ring->ring_size / 4); - WREG32_SOC15(VCN, inst_idx, mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); - WREG32_SOC15(VCN, inst_idx, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); - fw_shared->multi_queue.encode_generalpurpose_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET); - - fw_shared->multi_queue.encode_lowlatency_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET); - ring = &adev->vcn.inst[inst_idx].ring_enc[1]; - ring->wptr = 0; - WREG32_SOC15(VCN, inst_idx, mmUVD_RB_BASE_LO2, ring->gpu_addr); - WREG32_SOC15(VCN, inst_idx, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); - WREG32_SOC15(VCN, inst_idx, mmUVD_RB_SIZE2, ring->ring_size / 4); - WREG32_SOC15(VCN, inst_idx, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); - WREG32_SOC15(VCN, inst_idx, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); - fw_shared->multi_queue.encode_lowlatency_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET); - - /* restore wptr/rptr with pointers saved in FW shared memory*/ - WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_RPTR, fw_shared->rb.rptr); - WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_WPTR, fw_shared->rb.wptr); + if (adev->asic_type != CHIP_BEIGE_GOBY) { + /* Restore */ + fw_shared = adev->vcn.inst[inst_idx].fw_shared_cpu_addr; + fw_shared->multi_queue.encode_generalpurpose_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET); + ring = &adev->vcn.inst[inst_idx].ring_enc[0]; + ring->wptr = 0; + WREG32_SOC15(VCN, inst_idx, mmUVD_RB_BASE_LO, ring->gpu_addr); + WREG32_SOC15(VCN, inst_idx, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(VCN, inst_idx, mmUVD_RB_SIZE, ring->ring_size / 4); + WREG32_SOC15(VCN, inst_idx, mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); + WREG32_SOC15(VCN, inst_idx, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); + fw_shared->multi_queue.encode_generalpurpose_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET); + + fw_shared->multi_queue.encode_lowlatency_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET); + ring = &adev->vcn.inst[inst_idx].ring_enc[1]; + ring->wptr = 0; + WREG32_SOC15(VCN, inst_idx, mmUVD_RB_BASE_LO2, ring->gpu_addr); + WREG32_SOC15(VCN, inst_idx, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(VCN, inst_idx, mmUVD_RB_SIZE2, ring->ring_size / 4); + WREG32_SOC15(VCN, inst_idx, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); + WREG32_SOC15(VCN, inst_idx, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); + fw_shared->multi_queue.encode_lowlatency_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET); + + /* restore wptr/rptr with pointers saved in FW shared memory*/ + WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_RPTR, fw_shared->rb.rptr); + WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_WPTR, fw_shared->rb.wptr); + } /* Unstall DPG */ WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS), @@ -2138,7 +2144,8 @@ static void vcn_v3_0_set_enc_ring_funcs(struct amdgpu_device *adev) adev->vcn.inst[i].ring_enc[j].funcs = &vcn_v3_0_enc_ring_vm_funcs; adev->vcn.inst[i].ring_enc[j].me = i; } - DRM_INFO("VCN(%d) encode is enabled in VM mode\n", i); + if (adev->vcn.num_enc_rings > 0) + DRM_INFO("VCN(%d) encode is enabled in VM mode\n", i); } } diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index ca8efa5c6978..2f17c8a57015 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -104,6 +104,7 @@ static int vega10_ih_toggle_ring_interrupts(struct amdgpu_device *adev, tmp = RREG32(ih_regs->ih_rb_cntl); tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0)); + tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_GPU_TS_ENABLE, 1); /* enable_intr field is only valid in ring0 */ if (ih == &adev->irq.ih) tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0)); diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index 735ebbd1148f..3d21c0799037 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -1136,7 +1136,7 @@ static void vi_program_aspm(struct amdgpu_device *adev) bool bL1SS = false; bool bClkReqSupport = true; - if (amdgpu_aspm != 1) + if (!amdgpu_aspm) return; if (adev->flags & AMD_IS_APU || |