diff options
101 files changed, 1717 insertions, 868 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index c8ad6bf6618a..88db3c263e5f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -82,6 +82,7 @@ #include "amdgpu_bo_list.h" #include "amdgpu_gem.h" #include "amdgpu_doorbell.h" +#include "amdgpu_amdkfd.h" #define MAX_GPU_INSTANCE 16 @@ -163,6 +164,7 @@ extern int amdgpu_si_support; extern int amdgpu_cik_support; #endif +#define AMDGPU_VM_MAX_NUM_CTX 4096 #define AMDGPU_SG_THRESHOLD (256*1024*1024) #define AMDGPU_DEFAULT_GTT_SIZE_MB 3072ULL /* 3GB by default */ #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS 3000 @@ -862,6 +864,9 @@ struct amdgpu_device { /* GDS */ struct amdgpu_gds gds; + /* KFD */ + struct amdgpu_kfd_dev kfd; + /* display related functionality */ struct amdgpu_display_manager dm; @@ -875,9 +880,6 @@ struct amdgpu_device { atomic64_t visible_pin_size; atomic64_t gart_pin_size; - /* amdkfd interface */ - struct kfd_dev *kfd; - /* soc15 register offset based on ip, instance and segment */ uint32_t *reg_offset[MAX_HWIP][HWIP_MAX_INSTANCE]; @@ -910,7 +912,9 @@ struct amdgpu_device { bool in_gpu_reset; struct mutex lock_reset; struct amdgpu_doorbell_index doorbell_index; + int asic_reset_res; + struct work_struct xgmi_reset_work; }; static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_bo_device *bdev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c index 47db65926d71..4376b17ca594 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c @@ -886,6 +886,5 @@ void amdgpu_acpi_get_backlight_caps(struct amdgpu_device *adev, void amdgpu_acpi_fini(struct amdgpu_device *adev) { unregister_acpi_notifier(&adev->acpi_nb); - if (adev->atif) - kfree(adev->atif); + kfree(adev->atif); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index d693b8047653..2dfaf158ef07 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -26,15 +26,26 @@ #include "amdgpu.h" #include "amdgpu_gfx.h" #include <linux/module.h> +#include <linux/dma-buf.h> const struct kgd2kfd_calls *kgd2kfd; static const unsigned int compute_vmid_bitmap = 0xFF00; +/* Total memory size in system memory and all GPU VRAM. Used to + * estimate worst case amount of memory to reserve for page tables + */ +uint64_t amdgpu_amdkfd_total_mem_size; + int amdgpu_amdkfd_init(void) { + struct sysinfo si; int ret; + si_meminfo(&si); + amdgpu_amdkfd_total_mem_size = si.totalram - si.totalhigh; + amdgpu_amdkfd_total_mem_size *= si.mem_unit; + #ifdef CONFIG_HSA_AMD ret = kgd2kfd_init(KFD_INTERFACE_VERSION, &kgd2kfd); if (ret) @@ -87,8 +98,11 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) return; } - adev->kfd = kgd2kfd->probe((struct kgd_dev *)adev, - adev->pdev, kfd2kgd); + adev->kfd.dev = kgd2kfd->probe((struct kgd_dev *)adev, + adev->pdev, kfd2kgd); + + if (adev->kfd.dev) + amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; } /** @@ -128,7 +142,8 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) { int i, n; int last_valid_bit; - if (adev->kfd) { + + if (adev->kfd.dev) { struct kgd2kfd_shared_resources gpu_resources = { .compute_vmid_bitmap = compute_vmid_bitmap, .num_pipe_per_mec = adev->gfx.mec.num_pipe_per_mec, @@ -167,7 +182,7 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) &gpu_resources.doorbell_start_offset); if (adev->asic_type < CHIP_VEGA10) { - kgd2kfd->device_init(adev->kfd, &gpu_resources); + kgd2kfd->device_init(adev->kfd.dev, &gpu_resources); return; } @@ -196,37 +211,37 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) gpu_resources.reserved_doorbell_mask = 0x1e0; gpu_resources.reserved_doorbell_val = 0x0e0; - kgd2kfd->device_init(adev->kfd, &gpu_resources); + kgd2kfd->device_init(adev->kfd.dev, &gpu_resources); } } void amdgpu_amdkfd_device_fini(struct amdgpu_device *adev) { - if (adev->kfd) { - kgd2kfd->device_exit(adev->kfd); - adev->kfd = NULL; + if (adev->kfd.dev) { + kgd2kfd->device_exit(adev->kfd.dev); + adev->kfd.dev = NULL; } } void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev, const void *ih_ring_entry) { - if (adev->kfd) - kgd2kfd->interrupt(adev->kfd, ih_ring_entry); + if (adev->kfd.dev) + kgd2kfd->interrupt(adev->kfd.dev, ih_ring_entry); } void amdgpu_amdkfd_suspend(struct amdgpu_device *adev) { - if (adev->kfd) - kgd2kfd->suspend(adev->kfd); + if (adev->kfd.dev) + kgd2kfd->suspend(adev->kfd.dev); } int amdgpu_amdkfd_resume(struct amdgpu_device *adev) { int r = 0; - if (adev->kfd) - r = kgd2kfd->resume(adev->kfd); + if (adev->kfd.dev) + r = kgd2kfd->resume(adev->kfd.dev); return r; } @@ -235,8 +250,8 @@ int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev) { int r = 0; - if (adev->kfd) - r = kgd2kfd->pre_reset(adev->kfd); + if (adev->kfd.dev) + r = kgd2kfd->pre_reset(adev->kfd.dev); return r; } @@ -245,8 +260,8 @@ int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev) { int r = 0; - if (adev->kfd) - r = kgd2kfd->post_reset(adev->kfd); + if (adev->kfd.dev) + r = kgd2kfd->post_reset(adev->kfd.dev); return r; } @@ -419,6 +434,62 @@ void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info) cu_info->lds_size = acu_info.lds_size; } +int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd, + struct kgd_dev **dma_buf_kgd, + uint64_t *bo_size, void *metadata_buffer, + size_t buffer_size, uint32_t *metadata_size, + uint32_t *flags) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + struct dma_buf *dma_buf; + struct drm_gem_object *obj; + struct amdgpu_bo *bo; + uint64_t metadata_flags; + int r = -EINVAL; + + dma_buf = dma_buf_get(dma_buf_fd); + if (IS_ERR(dma_buf)) + return PTR_ERR(dma_buf); + + if (dma_buf->ops != &amdgpu_dmabuf_ops) + /* Can't handle non-graphics buffers */ + goto out_put; + + obj = dma_buf->priv; + if (obj->dev->driver != adev->ddev->driver) + /* Can't handle buffers from different drivers */ + goto out_put; + + adev = obj->dev->dev_private; + bo = gem_to_amdgpu_bo(obj); + if (!(bo->preferred_domains & (AMDGPU_GEM_DOMAIN_VRAM | + AMDGPU_GEM_DOMAIN_GTT))) + /* Only VRAM and GTT BOs are supported */ + goto out_put; + + r = 0; + if (dma_buf_kgd) + *dma_buf_kgd = (struct kgd_dev *)adev; + if (bo_size) + *bo_size = amdgpu_bo_size(bo); + if (metadata_size) + *metadata_size = bo->metadata_size; + if (metadata_buffer) + r = amdgpu_bo_get_metadata(bo, metadata_buffer, buffer_size, + metadata_size, &metadata_flags); + if (flags) { + *flags = (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) ? + ALLOC_MEM_FLAGS_VRAM : ALLOC_MEM_FLAGS_GTT; + + if (bo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) + *flags |= ALLOC_MEM_FLAGS_PUBLIC; + } + +out_put: + dma_buf_put(dma_buf); + return r; +} + uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; @@ -501,7 +572,7 @@ void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle) bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid) { - if (adev->kfd) { + if (adev->kfd.dev) { if ((1 << vmid) & compute_vmid_bitmap) return true; } @@ -515,7 +586,7 @@ bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm) return false; } -void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo) +void amdgpu_amdkfd_unreserve_memory_limit(struct amdgpu_bo *bo) { } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index bcf587b4ba98..70429f7aa9a8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -27,7 +27,6 @@ #include <linux/types.h> #include <linux/mm.h> -#include <linux/mmu_context.h> #include <linux/workqueue.h> #include <kgd_kfd_interface.h> #include <drm/ttm/ttm_execbuf_util.h> @@ -35,6 +34,7 @@ #include "amdgpu_vm.h" extern const struct kgd2kfd_calls *kgd2kfd; +extern uint64_t amdgpu_amdkfd_total_mem_size; struct amdgpu_device; @@ -77,6 +77,11 @@ struct amdgpu_amdkfd_fence { char timeline_name[TASK_COMM_LEN]; }; +struct amdgpu_kfd_dev { + struct kfd_dev *dev; + uint64_t vram_used; +}; + struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, struct mm_struct *mm); bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm); @@ -144,6 +149,11 @@ uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd); uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd); void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info); +int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd, + struct kgd_dev **dmabuf_kgd, + uint64_t *bo_size, void *metadata_buffer, + size_t buffer_size, uint32_t *metadata_size, + uint32_t *flags); uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd); uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd); @@ -195,7 +205,13 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info, 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, + struct kgd_mem **mem, uint64_t *size, + uint64_t *mmap_offset); + void amdgpu_amdkfd_gpuvm_init_mem_limits(void); -void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo); +void amdgpu_amdkfd_unreserve_memory_limit(struct amdgpu_bo *bo); #endif /* AMDGPU_AMDKFD_H_INCLUDED */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index 72a357dae070..ff7fac7df34b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -23,6 +23,7 @@ #include <linux/fdtable.h> #include <linux/uaccess.h> #include <linux/firmware.h> +#include <linux/mmu_context.h> #include <drm/drmP.h> #include "amdgpu.h" #include "amdgpu_amdkfd.h" diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index 0e2a56b6a9b6..56ea929f524b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -24,6 +24,7 @@ #include <linux/fdtable.h> #include <linux/uaccess.h> #include <linux/firmware.h> +#include <linux/mmu_context.h> #include <drm/drmP.h> #include "amdgpu.h" #include "amdgpu_amdkfd.h" 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 03b604c96d94..5c51d4910650 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -26,6 +26,7 @@ #include <linux/fdtable.h> #include <linux/uaccess.h> #include <linux/firmware.h> +#include <linux/mmu_context.h> #include <drm/drmP.h> #include "amdgpu.h" #include "amdgpu_amdkfd.h" diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index f3129b912714..be1ab43473c6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -25,6 +25,7 @@ #include <linux/list.h> #include <linux/pagemap.h> #include <linux/sched/mm.h> +#include <linux/dma-buf.h> #include <drm/drmP.h> #include "amdgpu_object.h" #include "amdgpu_vm.h" @@ -110,17 +111,17 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void) (kfd_mem_limit.max_ttm_mem_limit >> 20)); } -static int amdgpu_amdkfd_reserve_system_mem_limit(struct amdgpu_device *adev, +static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, uint64_t size, u32 domain, bool sg) { - size_t acc_size, system_mem_needed, ttm_mem_needed; + size_t acc_size, system_mem_needed, ttm_mem_needed, vram_needed; + uint64_t reserved_for_pt = amdgpu_amdkfd_total_mem_size >> 9; int ret = 0; acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size, sizeof(struct amdgpu_bo)); - spin_lock(&kfd_mem_limit.mem_limit_lock); - + vram_needed = 0; if (domain == AMDGPU_GEM_DOMAIN_GTT) { /* TTM GTT memory */ system_mem_needed = acc_size + size; @@ -133,23 +134,30 @@ static int amdgpu_amdkfd_reserve_system_mem_limit(struct amdgpu_device *adev, /* VRAM and SG */ system_mem_needed = acc_size; ttm_mem_needed = acc_size; + if (domain == AMDGPU_GEM_DOMAIN_VRAM) + vram_needed = size; } + spin_lock(&kfd_mem_limit.mem_limit_lock); + if ((kfd_mem_limit.system_mem_used + system_mem_needed > - kfd_mem_limit.max_system_mem_limit) || - (kfd_mem_limit.ttm_mem_used + ttm_mem_needed > - kfd_mem_limit.max_ttm_mem_limit)) + kfd_mem_limit.max_system_mem_limit) || + (kfd_mem_limit.ttm_mem_used + ttm_mem_needed > + kfd_mem_limit.max_ttm_mem_limit) || + (adev->kfd.vram_used + vram_needed > + adev->gmc.real_vram_size - reserved_for_pt)) { ret = -ENOMEM; - else { + } else { kfd_mem_limit.system_mem_used += system_mem_needed; kfd_mem_limit.ttm_mem_used += ttm_mem_needed; + adev->kfd.vram_used += vram_needed; } spin_unlock(&kfd_mem_limit.mem_limit_lock); return ret; } -static void unreserve_system_mem_limit(struct amdgpu_device *adev, +static void unreserve_mem_limit(struct amdgpu_device *adev, uint64_t size, u32 domain, bool sg) { size_t acc_size; @@ -167,6 +175,11 @@ static void unreserve_system_mem_limit(struct amdgpu_device *adev, } else { kfd_mem_limit.system_mem_used -= acc_size; kfd_mem_limit.ttm_mem_used -= acc_size; + if (domain == AMDGPU_GEM_DOMAIN_VRAM) { + adev->kfd.vram_used -= size; + WARN_ONCE(adev->kfd.vram_used < 0, + "kfd VRAM memory accounting unbalanced"); + } } WARN_ONCE(kfd_mem_limit.system_mem_used < 0, "kfd system memory accounting unbalanced"); @@ -176,29 +189,18 @@ static void unreserve_system_mem_limit(struct amdgpu_device *adev, spin_unlock(&kfd_mem_limit.mem_limit_lock); } -void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo) +void amdgpu_amdkfd_unreserve_memory_limit(struct amdgpu_bo *bo) { - spin_lock(&kfd_mem_limit.mem_limit_lock); + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + u32 domain = bo->preferred_domains; + bool sg = (bo->preferred_domains == AMDGPU_GEM_DOMAIN_CPU); if (bo->flags & AMDGPU_AMDKFD_USERPTR_BO) { - kfd_mem_limit.system_mem_used -= - (bo->tbo.acc_size + amdgpu_bo_size(bo)); - kfd_mem_limit.ttm_mem_used -= bo->tbo.acc_size; - } else if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GTT) { - kfd_mem_limit.system_mem_used -= - (bo->tbo.acc_size + amdgpu_bo_size(bo)); - kfd_mem_limit.ttm_mem_used -= - (bo->tbo.acc_size + amdgpu_bo_size(bo)); - } else { - kfd_mem_limit.system_mem_used -= bo->tbo.acc_size; - kfd_mem_limit.ttm_mem_used -= bo->tbo.acc_size; + domain = AMDGPU_GEM_DOMAIN_CPU; + sg = false; } - WARN_ONCE(kfd_mem_limit.system_mem_used < 0, - "kfd system memory accounting unbalanced"); - WARN_ONCE(kfd_mem_limit.ttm_mem_used < 0, - "kfd TTM memory accounting unbalanced"); - spin_unlock(&kfd_mem_limit.mem_limit_lock); + unreserve_mem_limit(adev, amdgpu_bo_size(bo), domain, sg); } @@ -535,7 +537,7 @@ static void add_kgd_mem_to_kfd_bo_list(struct kgd_mem *mem, struct amdgpu_bo *bo = mem->bo; INIT_LIST_HEAD(&entry->head); - entry->shared = true; + entry->num_shared = 1; entry->bo = &bo->tbo; mutex_lock(&process_info->lock); if (userptr) @@ -676,7 +678,7 @@ static int reserve_bo_and_vm(struct kgd_mem *mem, ctx->kfd_bo.priority = 0; ctx->kfd_bo.tv.bo = &bo->tbo; - ctx->kfd_bo.tv.shared = true; + ctx->kfd_bo.tv.num_shared = 1; ctx->kfd_bo.user_pages = NULL; list_add(&ctx->kfd_bo.tv.head, &ctx->list); @@ -740,7 +742,7 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem, ctx->kfd_bo.priority = 0; ctx->kfd_bo.tv.bo = &bo->tbo; - ctx->kfd_bo.tv.shared = true; + ctx->kfd_bo.tv.num_shared = 1; ctx->kfd_bo.user_pages = NULL; list_add(&ctx->kfd_bo.tv.head, &ctx->list); @@ -885,6 +887,24 @@ update_gpuvm_pte_failed: return ret; } +static struct sg_table *create_doorbell_sg(uint64_t addr, uint32_t size) +{ + struct sg_table *sg = kmalloc(sizeof(*sg), GFP_KERNEL); + + if (!sg) + return NULL; + if (sg_alloc_table(sg, 1, GFP_KERNEL)) { + kfree(sg); + return NULL; + } + sg->sgl->dma_address = addr; + sg->sgl->length = size; +#ifdef CONFIG_NEED_SG_DMA_LENGTH + sg->sgl->dma_length = size; +#endif + return sg; +} + static int process_validate_vms(struct amdkfd_process_info *process_info) { struct amdgpu_vm *peer_vm; @@ -1168,6 +1188,8 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( { struct amdgpu_device *adev = get_amdgpu_device(kgd); struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + enum ttm_bo_type bo_type = ttm_bo_type_device; + struct sg_table *sg = NULL; uint64_t user_addr = 0; struct amdgpu_bo *bo; struct amdgpu_bo_param bp; @@ -1196,13 +1218,25 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( if (!offset || !*offset) return -EINVAL; user_addr = *offset; + } else if (flags & ALLOC_MEM_FLAGS_DOORBELL) { + domain = AMDGPU_GEM_DOMAIN_GTT; + alloc_domain = AMDGPU_GEM_DOMAIN_CPU; + bo_type = ttm_bo_type_sg; + alloc_flags = 0; + if (size > UINT_MAX) + return -EINVAL; + sg = create_doorbell_sg(*offset, size); + if (!sg) + return -ENOMEM; } else { return -EINVAL; } *mem = kzalloc(sizeof(struct kgd_mem), GFP_KERNEL); - if (!*mem) - return -ENOMEM; + if (!*mem) { + ret = -ENOMEM; + goto err; + } INIT_LIST_HEAD(&(*mem)->bo_va_list); mutex_init(&(*mem)->lock); (*mem)->aql_queue = !!(flags & ALLOC_MEM_FLAGS_AQL_QUEUE_MEM); @@ -1235,8 +1269,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( amdgpu_sync_create(&(*mem)->sync); - ret = amdgpu_amdkfd_reserve_system_mem_limit(adev, size, - alloc_domain, false); + ret = amdgpu_amdkfd_reserve_mem_limit(adev, size, alloc_domain, !!sg); if (ret) { pr_debug("Insufficient system memory\n"); goto err_reserve_limit; @@ -1250,7 +1283,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( bp.byte_align = byte_align; bp.domain = alloc_domain; bp.flags = alloc_flags; - bp.type = ttm_bo_type_device; + bp.type = bo_type; bp.resv = NULL; ret = amdgpu_bo_create(adev, &bp, &bo); if (ret) { @@ -1258,6 +1291,10 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( domain_string(alloc_domain), ret); goto err_bo_create; } + if (bo_type == ttm_bo_type_sg) { + bo->tbo.sg = sg; + bo->tbo.ttm->sg = sg; + } bo->kfd_bo = *mem; (*mem)->bo = bo; if (user_addr) @@ -1289,10 +1326,15 @@ allocate_init_user_pages_failed: /* Don't unreserve system mem limit twice */ goto err_reserve_limit; err_bo_create: - unreserve_system_mem_limit(adev, size, alloc_domain, false); + unreserve_mem_limit(adev, size, alloc_domain, !!sg); err_reserve_limit: mutex_destroy(&(*mem)->lock); kfree(*mem); +err: + if (sg) { + sg_free_table(sg); + kfree(sg); + } return ret; } @@ -1362,6 +1404,14 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( /* Free the sync object */ amdgpu_sync_free(&mem->sync); + /* If the SG is not NULL, it's one we created for a doorbell + * BO. We need to free it. + */ + if (mem->bo->tbo.sg) { + sg_free_table(mem->bo->tbo.sg); + kfree(mem->bo->tbo.sg); + } + /* Free the BO*/ amdgpu_bo_unref(&mem->bo); mutex_destroy(&mem->lock); @@ -1664,6 +1714,60 @@ int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, return 0; } +int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd, + struct dma_buf *dma_buf, + uint64_t va, void *vm, + struct kgd_mem **mem, uint64_t *size, + uint64_t *mmap_offset) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + struct drm_gem_object *obj; + struct amdgpu_bo *bo; + struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + + if (dma_buf->ops != &amdgpu_dmabuf_ops) + /* Can't handle non-graphics buffers */ + return -EINVAL; + + obj = dma_buf->priv; + if (obj->dev->dev_private != adev) + /* Can't handle buffers from other devices */ + return -EINVAL; + + bo = gem_to_amdgpu_bo(obj); + if (!(bo->preferred_domains & (AMDGPU_GEM_DOMAIN_VRAM | + AMDGPU_GEM_DOMAIN_GTT))) + /* Only VRAM and GTT BOs are supported */ + return -EINVAL; + + *mem = kzalloc(sizeof(struct kgd_mem), GFP_KERNEL); + if (!*mem) + return -ENOMEM; + + if (size) + *size = amdgpu_bo_size(bo); + + if (mmap_offset) + *mmap_offset = amdgpu_bo_mmap_offset(bo); + + INIT_LIST_HEAD(&(*mem)->bo_va_list); + mutex_init(&(*mem)->lock); + (*mem)->mapping_flags = + AMDGPU_VM_PAGE_READABLE | AMDGPU_VM_PAGE_WRITEABLE | + AMDGPU_VM_PAGE_EXECUTABLE | AMDGPU_VM_MTYPE_NC; + + (*mem)->bo = amdgpu_bo_ref(bo); + (*mem)->va = va; + (*mem)->domain = (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) ? + AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT; + (*mem)->mapped_to_gpu_memory = 0; + (*mem)->process_info = avm->process_info; + add_kgd_mem_to_kfd_bo_list(*mem, avm->process_info, false); + amdgpu_sync_create(&(*mem)->sync); + + return 0; +} + /* Evict a userptr BO by stopping the queues if necessary * * Runs in MMU notifier, may be in RECLAIM_FS context. This means it @@ -1830,7 +1934,7 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) validate_list.head) { list_add_tail(&mem->resv_list.head, &resv_list); mem->resv_list.bo = mem->validate_list.bo; - mem->resv_list.shared = mem->validate_list.shared; + mem->resv_list.num_shared = mem->validate_list.num_shared; } /* Reserve all BOs and page tables for validation */ @@ -2049,7 +2153,7 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) list_add_tail(&mem->resv_list.head, &ctx.list); mem->resv_list.bo = mem->validate_list.bo; - mem->resv_list.shared = mem->validate_list.shared; + mem->resv_list.num_shared = mem->validate_list.num_shared; } ret = ttm_eu_reserve_buffers(&ctx.ticket, &ctx.list, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c index 14d2982a47cc..5c79da8e1150 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c @@ -118,7 +118,6 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp, entry->priority = min(info[i].bo_priority, AMDGPU_BO_LIST_MAX_PRIORITY); entry->tv.bo = &bo->tbo; - entry->tv.shared = !bo->prime_shared_count; if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GDS) list->gds_obj = bo; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index dc54e9efd910..5dc3ee372e2f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -50,7 +50,8 @@ static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p, bo = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj)); p->uf_entry.priority = 0; p->uf_entry.tv.bo = &bo->tbo; - p->uf_entry.tv.shared = true; + /* One for TTM and one for the CS job */ + p->uf_entry.tv.num_shared = 2; p->uf_entry.user_pages = NULL; drm_gem_object_put_unlocked(gobj); @@ -598,6 +599,10 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, return r; } + /* One for TTM and one for the CS job */ + amdgpu_bo_list_for_each_entry(e, p->bo_list) + e->tv.num_shared = 2; + amdgpu_bo_list_get_list(p->bo_list, &p->validated); if (p->bo_list->first_userptr != p->bo_list->num_entries) p->mn = amdgpu_mn_get(p->adev, AMDGPU_MN_TYPE_GFX); @@ -717,8 +722,14 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, gws = p->bo_list->gws_obj; oa = p->bo_list->oa_obj; - amdgpu_bo_list_for_each_entry(e, p->bo_list) - e->bo_va = amdgpu_vm_bo_find(vm, ttm_to_amdgpu_bo(e->tv.bo)); + amdgpu_bo_list_for_each_entry(e, p->bo_list) { + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); + + /* Make sure we use the exclusive slot for shared BOs */ + if (bo->prime_shared_count) + e->tv.num_shared = 0; + e->bo_va = amdgpu_vm_bo_find(vm, bo); + } if (gds) { p->job->gds_base = amdgpu_bo_gpu_offset(gds) >> PAGE_SHIFT; @@ -955,10 +966,6 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) if (r) return r; - r = reservation_object_reserve_shared(vm->root.base.bo->tbo.resv, 1); - if (r) - return r; - p->job->vm_pd_addr = amdgpu_gmc_pd_addr(vm->root.base.bo); if (amdgpu_vm_debug) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c index 5b550706ee76..7e22be7ca68a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c @@ -74,7 +74,7 @@ int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm, INIT_LIST_HEAD(&list); INIT_LIST_HEAD(&csa_tv.head); csa_tv.bo = &bo->tbo; - csa_tv.shared = true; + csa_tv.num_shared = 1; list_add(&csa_tv.head, &list); amdgpu_vm_get_pd_bo(vm, &list, &pd); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index 95f4c4139fc6..d85184b5b35c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -248,7 +248,7 @@ static int amdgpu_ctx_alloc(struct amdgpu_device *adev, return -ENOMEM; mutex_lock(&mgr->lock); - r = idr_alloc(&mgr->ctx_handles, ctx, 1, 0, GFP_KERNEL); + r = idr_alloc(&mgr->ctx_handles, ctx, 1, AMDGPU_VM_MAX_NUM_CTX, GFP_KERNEL); if (r < 0) { mutex_unlock(&mgr->lock); kfree(ctx); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index c75badfa5c4c..b60afeade50a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -515,7 +515,6 @@ void amdgpu_device_pci_config_reset(struct amdgpu_device *adev) */ static int amdgpu_device_doorbell_init(struct amdgpu_device *adev) { - amdgpu_asic_init_doorbell_index(adev); /* No doorbell on SI hardware generation */ if (adev->asic_type < CHIP_BONAIRE) { @@ -529,6 +528,8 @@ static int amdgpu_device_doorbell_init(struct amdgpu_device *adev) if (pci_resource_flags(adev->pdev, 2) & IORESOURCE_UNSET) return -EINVAL; + amdgpu_asic_init_doorbell_index(adev); + /* doorbell bar mapping */ adev->doorbell.base = pci_resource_start(adev->pdev, 2); adev->doorbell.size = pci_resource_len(adev->pdev, 2); @@ -1864,6 +1865,9 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) { int i, r; + if (adev->gmc.xgmi.num_physical_nodes > 1) + amdgpu_xgmi_remove_device(adev); + amdgpu_amdkfd_device_fini(adev); amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE); @@ -2353,6 +2357,19 @@ 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 = + container_of(__work, struct amdgpu_device, xgmi_reset_work); + + adev->asic_reset_res = amdgpu_asic_reset(adev); + if (adev->asic_reset_res) + DRM_WARN("ASIC reset failed with err r, %d for drm dev, %s", + adev->asic_reset_res, adev->ddev->unique); +} + + /** * amdgpu_device_init - initialize the driver * @@ -2451,6 +2468,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, INIT_DELAYED_WORK(&adev->gfx.gfx_off_delay_work, amdgpu_device_delay_enable_gfx_off); + INIT_WORK(&adev->xgmi_reset_work, amdgpu_device_xgmi_reset_func); + adev->gfx.gfx_off_req_count = 1; adev->pm.ac_power = power_supply_is_system_supplied() > 0 ? true : false; @@ -3239,6 +3258,8 @@ bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev) if (amdgpu_gpu_recovery == -1) { switch (adev->asic_type) { + case CHIP_BONAIRE: + case CHIP_HAWAII: case CHIP_TOPAZ: case CHIP_TONGA: case CHIP_FIJI: @@ -3328,10 +3349,31 @@ static int amdgpu_do_asic_reset(struct amdgpu_hive_info *hive, */ if (need_full_reset) { list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) { - r = amdgpu_asic_reset(tmp_adev); - if (r) - DRM_WARN("ASIC reset failed with err r, %d for drm dev, %s", + /* For XGMI run all resets in parallel to speed up the process */ + if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) { + if (!queue_work(system_highpri_wq, &tmp_adev->xgmi_reset_work)) + r = -EALREADY; + } else + r = amdgpu_asic_reset(tmp_adev); + + if (r) { + DRM_ERROR("ASIC reset failed with err r, %d for drm dev, %s", r, tmp_adev->ddev->unique); + break; + } + } + + /* For XGMI wait for all PSP resets to complete before proceed */ + if (!r) { + list_for_each_entry(tmp_adev, device_list_handle, + gmc.xgmi.head) { + if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) { + flush_work(&tmp_adev->xgmi_reset_work); + r = tmp_adev->asic_reset_res; + if (r) + break; + } + } } } @@ -3518,8 +3560,6 @@ retry: /* Rest of adevs pre asic reset from XGMI hive. */ if (tmp_adev == adev) continue; - dev_info(tmp_adev->dev, "GPU reset begin for drm dev %s!\n", adev->ddev->unique); - amdgpu_device_lock_adev(tmp_adev); r = amdgpu_device_pre_asic_reset(tmp_adev, NULL, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 7b3d1ebda9df..f4f00217546e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -169,7 +169,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj, INIT_LIST_HEAD(&duplicates); tv.bo = &bo->tbo; - tv.shared = true; + tv.num_shared = 1; list_add(&tv.head, &list); amdgpu_vm_get_pd_bo(vm, &list, &vm_pd); @@ -604,7 +604,10 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, return -ENOENT; abo = gem_to_amdgpu_bo(gobj); tv.bo = &abo->tbo; - tv.shared = !!(abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID); + if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) + tv.num_shared = 1; + else + tv.num_shared = 0; list_add(&tv.head, &list); } else { gobj = NULL; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h index d63daba9b17c..f1ddfc50bcc7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h @@ -54,6 +54,8 @@ void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj); void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr); int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma); +extern const struct dma_buf_ops amdgpu_dmabuf_ops; + /* * GEM objects. */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index 8c57924c075f..81e6070d255b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -99,6 +99,7 @@ struct amdgpu_xgmi { unsigned num_physical_nodes; /* gpu list in the same hive */ struct list_head head; + bool supported; }; struct amdgpu_gmc { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h index 9ce8c93ec19b..f877bb78d10a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h @@ -51,14 +51,12 @@ struct amdgpu_ih_ring { struct amdgpu_ih_funcs { /* ring read/write ptr handling, called from interrupt context */ u32 (*get_wptr)(struct amdgpu_device *adev); - bool (*prescreen_iv)(struct amdgpu_device *adev); void (*decode_iv)(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry); void (*set_rptr)(struct amdgpu_device *adev); }; #define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev)) -#define amdgpu_ih_prescreen_iv(adev) (adev)->irq.ih_funcs->prescreen_iv((adev)) #define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv)) #define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c index 6b6524f04ce0..b7968f426862 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c @@ -145,13 +145,6 @@ static void amdgpu_irq_callback(struct amdgpu_device *adev, u32 ring_index = ih->rptr >> 2; struct amdgpu_iv_entry entry; - /* Prescreening of high-frequency interrupts */ - if (!amdgpu_ih_prescreen_iv(adev)) - return; - - /* Before dispatching irq to IP blocks, send it to amdkfd */ - amdgpu_amdkfd_interrupt(adev, (const void *) &ih->ring[ring_index]); - entry.iv_entry = (const uint32_t *)&ih->ring[ring_index]; amdgpu_ih_decode_iv(adev, &entry); @@ -371,39 +364,38 @@ void amdgpu_irq_dispatch(struct amdgpu_device *adev, unsigned client_id = entry->client_id; unsigned src_id = entry->src_id; struct amdgpu_irq_src *src; + bool handled = false; int r; trace_amdgpu_iv(entry); if (client_id >= AMDGPU_IRQ_CLIENTID_MAX) { DRM_DEBUG("Invalid client_id in IV: %d\n", client_id); - return; - } - if (src_id >= AMDGPU_MAX_IRQ_SRC_ID) { + } else if (src_id >= AMDGPU_MAX_IRQ_SRC_ID) { DRM_DEBUG("Invalid src_id in IV: %d\n", src_id); - return; - } - if (adev->irq.virq[src_id]) { + } else if (adev->irq.virq[src_id]) { generic_handle_irq(irq_find_mapping(adev->irq.domain, src_id)); - } else { - if (!adev->irq.client[client_id].sources) { - DRM_DEBUG("Unregistered interrupt client_id: %d src_id: %d\n", - client_id, src_id); - return; - } - src = adev->irq.client[client_id].sources[src_id]; - if (!src) { - DRM_DEBUG("Unhandled interrupt src_id: %d\n", src_id); - return; - } + } else if (!adev->irq.client[client_id].sources) { + DRM_DEBUG("Unregistered interrupt client_id: %d src_id: %d\n", + client_id, src_id); + } else if ((src = adev->irq.client[client_id].sources[src_id])) { r = src->funcs->process(adev, src, entry); - if (r) + if (r < 0) DRM_ERROR("error processing interrupt (%d)\n", r); + else if (r) + handled = true; + + } else { + DRM_DEBUG("Unhandled interrupt src_id: %d\n", src_id); } + + /* Send it to amdkfd as well if it isn't already handled */ + if (!handled) + amdgpu_amdkfd_interrupt(adev, entry->iv_entry); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index cf768acb51dc..fd271f9746a2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -81,7 +81,7 @@ static void amdgpu_bo_destroy(struct ttm_buffer_object *tbo) amdgpu_bo_subtract_pin_size(bo); if (bo->kfd_bo) - amdgpu_amdkfd_unreserve_system_memory_limit(bo); + amdgpu_amdkfd_unreserve_memory_limit(bo); amdgpu_bo_kunmap(bo); @@ -608,53 +608,6 @@ int amdgpu_bo_create(struct amdgpu_device *adev, } /** - * amdgpu_bo_backup_to_shadow - Backs up an &amdgpu_bo buffer object - * @adev: amdgpu device object - * @ring: amdgpu_ring for the engine handling the buffer operations - * @bo: &amdgpu_bo buffer to be backed up - * @resv: reservation object with embedded fence - * @fence: dma_fence associated with the operation - * @direct: whether to submit the job directly - * - * Copies an &amdgpu_bo buffer object to its shadow object. - * Not used for now. - * - * Returns: - * 0 for success or a negative error code on failure. - */ -int amdgpu_bo_backup_to_shadow(struct amdgpu_device *adev, - struct amdgpu_ring *ring, - struct amdgpu_bo *bo, - struct reservation_object *resv, - struct dma_fence **fence, - bool direct) - -{ - struct amdgpu_bo *shadow = bo->shadow; - uint64_t bo_addr, shadow_addr; - int r; - - if (!shadow) - return -EINVAL; - - bo_addr = amdgpu_bo_gpu_offset(bo); - shadow_addr = amdgpu_bo_gpu_offset(bo->shadow); - - r = reservation_object_reserve_shared(bo->tbo.resv, 1); - if (r) - goto err; - - r = amdgpu_copy_buffer(ring, bo_addr, shadow_addr, - amdgpu_bo_size(bo), resv, fence, - direct, false); - if (!r) - amdgpu_bo_fence(bo, *fence, true); - -err: - return r; -} - -/** * amdgpu_bo_validate - validate an &amdgpu_bo buffer object * @bo: pointer to the buffer object * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 7d3312d0da11..9291c2f837e9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -267,11 +267,6 @@ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo); void amdgpu_bo_fence(struct amdgpu_bo *bo, struct dma_fence *fence, bool shared); u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo); -int amdgpu_bo_backup_to_shadow(struct amdgpu_device *adev, - struct amdgpu_ring *ring, - struct amdgpu_bo *bo, - struct reservation_object *resv, - struct dma_fence **fence, bool direct); int amdgpu_bo_validate(struct amdgpu_bo *bo); int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow, struct dma_fence **fence); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index 3e44d889f7af..71913a18d142 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -39,8 +39,6 @@ #include <drm/amdgpu_drm.h> #include <linux/dma-buf.h> -static const struct dma_buf_ops amdgpu_dmabuf_ops; - /** * amdgpu_gem_prime_get_sg_table - &drm_driver.gem_prime_get_sg_table * implementation @@ -332,7 +330,7 @@ static int amdgpu_gem_begin_cpu_access(struct dma_buf *dma_buf, return ret; } -static const struct dma_buf_ops amdgpu_dmabuf_ops = { +const struct dma_buf_ops amdgpu_dmabuf_ops = { .attach = amdgpu_gem_map_attach, .detach = amdgpu_gem_map_detach, .map_dma_buf = drm_gem_map_dma_buf, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index e05dc66b1090..6759d898b3ab 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -155,10 +155,22 @@ psp_cmd_submit_buf(struct psp_context *psp, return ret; } -static void psp_prep_tmr_cmd_buf(struct psp_gfx_cmd_resp *cmd, +bool psp_support_vmr_ring(struct psp_context *psp) +{ + if (amdgpu_sriov_vf(psp->adev) && psp->sos_fw_version > 0x80045) + return true; + else + return false; +} + +static void psp_prep_tmr_cmd_buf(struct psp_context *psp, + struct psp_gfx_cmd_resp *cmd, uint64_t tmr_mc, uint32_t size) { - cmd->cmd_id = GFX_CMD_ID_SETUP_TMR; + if (psp_support_vmr_ring(psp)) + cmd->cmd_id = GFX_CMD_ID_SETUP_VMR; + else + cmd->cmd_id = GFX_CMD_ID_SETUP_TMR; cmd->cmd.cmd_setup_tmr.buf_phy_addr_lo = lower_32_bits(tmr_mc); cmd->cmd.cmd_setup_tmr.buf_phy_addr_hi = upper_32_bits(tmr_mc); cmd->cmd.cmd_setup_tmr.buf_size = size; @@ -192,7 +204,7 @@ static int psp_tmr_load(struct psp_context *psp) if (!cmd) return -ENOMEM; - psp_prep_tmr_cmd_buf(cmd, psp->tmr_mc_addr, PSP_TMR_SIZE); + psp_prep_tmr_cmd_buf(psp, cmd, psp->tmr_mc_addr, PSP_TMR_SIZE); DRM_INFO("reserve 0x%x from 0x%llx for PSP TMR SIZE\n", PSP_TMR_SIZE, psp->tmr_mc_addr); @@ -536,8 +548,10 @@ static int psp_load_fw(struct amdgpu_device *adev) int ret; struct psp_context *psp = &adev->psp; - if (amdgpu_sriov_vf(adev) && adev->in_gpu_reset != 0) + if (amdgpu_sriov_vf(adev) && adev->in_gpu_reset) { + psp_ring_destroy(psp, PSP_RING_TYPE__KM); goto skip_memalloc; + } psp->cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); if (!psp->cmd) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h index 9ec5d1a666a6..10decf70c9aa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h @@ -217,6 +217,7 @@ extern const struct amdgpu_ip_block_version psp_v10_0_ip_block; int psp_gpu_reset(struct amdgpu_device *adev); int psp_xgmi_invoke(struct psp_context *psp, uint32_t ta_cmd_id); +bool psp_support_vmr_ring(struct psp_context *psp); extern const struct amdgpu_ip_block_version psp_v11_0_ip_block; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 5b75bdc8dc28..335a0edf114b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -397,7 +397,7 @@ bool amdgpu_ring_soft_recovery(struct amdgpu_ring *ring, unsigned int vmid, { ktime_t deadline = ktime_add_us(ktime_get(), 10000); - if (!ring->funcs->soft_recovery) + if (!ring->funcs->soft_recovery || !fence) return false; atomic_inc(&ring->adev->gpu_reset_counter); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 58a2363040dd..fc91f3e54a87 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -617,7 +617,8 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, { entry->priority = 0; entry->tv.bo = &vm->root.base.bo->tbo; - entry->tv.shared = true; + /* One for the VM updates, one for TTM and one for the CS job */ + entry->tv.num_shared = 3; entry->user_pages = NULL; list_add(&entry->tv.head, validated); } @@ -773,10 +774,6 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, ring = container_of(vm->entity.rq->sched, struct amdgpu_ring, sched); - r = reservation_object_reserve_shared(bo->tbo.resv, 1); - if (r) - return r; - r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (r) goto error; @@ -1842,10 +1839,6 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (r) goto error_free; - r = reservation_object_reserve_shared(vm->root.base.bo->tbo.resv, 1); - if (r) - goto error_free; - r = amdgpu_vm_update_ptes(¶ms, start, last + 1, addr, flags); if (r) goto error_free; @@ -3026,6 +3019,10 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, if (r) goto error_free_root; + r = reservation_object_reserve_shared(root->tbo.resv, 1); + if (r) + goto error_unreserve; + r = amdgpu_vm_clear_bo(adev, vm, root, adev->vm_manager.root_level, vm->pte_support_ats); @@ -3055,7 +3052,6 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, } INIT_KFIFO(vm->faults); - vm->fault_credit = 16; return 0; @@ -3268,42 +3264,6 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) } /** - * amdgpu_vm_pasid_fault_credit - Check fault credit for given PASID - * - * @adev: amdgpu_device pointer - * @pasid: PASID do identify the VM - * - * This function is expected to be called in interrupt context. - * - * Returns: - * True if there was fault credit, false otherwise - */ -bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev, - unsigned int pasid) -{ - struct amdgpu_vm *vm; - - spin_lock(&adev->vm_manager.pasid_lock); - vm = idr_find(&adev->vm_manager.pasid_idr, pasid); - if (!vm) { - /* VM not found, can't track fault credit */ - spin_unlock(&adev->vm_manager.pasid_lock); - return true; - } - - /* No lock needed. only accessed by IRQ handler */ - if (!vm->fault_credit) { - /* Too many faults in this VM */ - spin_unlock(&adev->vm_manager.pasid_lock); - return false; - } - - vm->fault_credit--; - spin_unlock(&adev->vm_manager.pasid_lock); - return true; -} - -/** * amdgpu_vm_manager_init - init the VM manager * * @adev: amdgpu_device pointer diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 2a8898d19c8b..e8dcfd59fc93 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -229,9 +229,6 @@ struct amdgpu_vm { /* Up to 128 pending retry page faults */ DECLARE_KFIFO(faults, u64, 128); - /* Limit non-retry fault storms */ - unsigned int fault_credit; - /* Points to the KFD process VM info */ struct amdkfd_process_info *process_info; @@ -299,8 +296,6 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned int 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); -bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev, - unsigned int pasid); void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, struct list_head *validated, struct amdgpu_bo_list_entry *entry); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index fb37e69f1bba..0b263a9857c6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -78,7 +78,7 @@ int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_dev adev->gmc.xgmi.node_id, adev->gmc.xgmi.hive_id, ret); else - dev_info(adev->dev, "XGMI: Add node %d to hive 0x%llx.\n", + dev_info(adev->dev, "XGMI: Set topology for node %d, hive 0x%llx.\n", adev->gmc.xgmi.physical_node_id, adev->gmc.xgmi.hive_id); @@ -94,9 +94,9 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev) int count = 0, ret = -EINVAL; - if ((adev->asic_type < CHIP_VEGA20) || - (adev->flags & AMD_IS_APU) ) + if (!adev->gmc.xgmi.supported) return 0; + adev->gmc.xgmi.node_id = psp_xgmi_get_node_id(&adev->psp); adev->gmc.xgmi.hive_id = psp_xgmi_get_hive_id(&adev->psp); @@ -135,3 +135,23 @@ exit: mutex_unlock(&xgmi_mutex); return ret; } + +void amdgpu_xgmi_remove_device(struct amdgpu_device *adev) +{ + struct amdgpu_hive_info *hive; + + if (!adev->gmc.xgmi.supported) + return; + + mutex_lock(&xgmi_mutex); + + hive = amdgpu_get_xgmi_hive(adev); + if (!hive) + goto exit; + + if (!(hive->number_devices--)) + mutex_destroy(&hive->hive_lock); + +exit: + mutex_unlock(&xgmi_mutex); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h index 6335bfdcc51d..6151eb9c8ad3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h @@ -35,5 +35,6 @@ struct amdgpu_hive_info { struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev); int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_device *adev); int amdgpu_xgmi_add_device(struct amdgpu_device *adev); +void amdgpu_xgmi_remove_device(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/cik_ih.c b/drivers/gpu/drm/amd/amdgpu/cik_ih.c index b5775c6a857b..8a8b4967a101 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_ih.c @@ -228,34 +228,6 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev) * [127:96] - reserved */ -/** - * cik_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool cik_ih_prescreen_iv(struct amdgpu_device *adev) -{ - u32 ring_index = adev->irq.ih.rptr >> 2; - u16 pasid; - - switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { - case 146: - case 147: - pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; - if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) - return true; - break; - default: - /* Not a VM fault */ - return true; - } - - adev->irq.ih.rptr += 16; - return false; -} - /** * cik_ih_decode_iv - decode an interrupt vector * @@ -461,7 +433,6 @@ static const struct amd_ip_funcs cik_ih_ip_funcs = { static const struct amdgpu_ih_funcs cik_ih_funcs = { .get_wptr = cik_ih_get_wptr, - .prescreen_iv = cik_ih_prescreen_iv, .decode_iv = cik_ih_decode_iv, .set_rptr = cik_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/cz_ih.c b/drivers/gpu/drm/amd/amdgpu/cz_ih.c index df5ac4d85a00..9d3ea298e116 100644 --- a/drivers/gpu/drm/amd/amdgpu/cz_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cz_ih.c @@ -208,34 +208,6 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev) } /** - * cz_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool cz_ih_prescreen_iv(struct amdgpu_device *adev) -{ - u32 ring_index = adev->irq.ih.rptr >> 2; - u16 pasid; - - switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { - case 146: - case 147: - pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; - if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) - return true; - break; - default: - /* Not a VM fault */ - return true; - } - - adev->irq.ih.rptr += 16; - return false; -} - -/** * cz_ih_decode_iv - decode an interrupt vector * * @adev: amdgpu_device pointer @@ -442,7 +414,6 @@ static const struct amd_ip_funcs cz_ih_ip_funcs = { static const struct amdgpu_ih_funcs cz_ih_funcs = { .get_wptr = cz_ih_get_wptr, - .prescreen_iv = cz_ih_prescreen_iv, .decode_iv = cz_ih_decode_iv, .set_rptr = cz_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 1454fc306783..381f593b0cda 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -4068,6 +4068,11 @@ static void gfx_v8_0_rlc_start(struct amdgpu_device *adev) static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev) { + if (amdgpu_sriov_vf(adev)) { + gfx_v8_0_init_csb(adev); + return 0; + } + adev->gfx.rlc.funcs->stop(adev); adev->gfx.rlc.funcs->reset(adev); gfx_v8_0_init_pg(adev); @@ -4947,14 +4952,13 @@ static bool gfx_v8_0_check_soft_reset(void *handle) static int gfx_v8_0_pre_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + u32 grbm_soft_reset = 0; if ((!adev->gfx.grbm_soft_reset) && (!adev->gfx.srbm_soft_reset)) return 0; grbm_soft_reset = adev->gfx.grbm_soft_reset; - srbm_soft_reset = adev->gfx.srbm_soft_reset; /* stop the rlc */ adev->gfx.rlc.funcs->stop(adev); @@ -5051,14 +5055,13 @@ static int gfx_v8_0_soft_reset(void *handle) static int gfx_v8_0_post_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + u32 grbm_soft_reset = 0; if ((!adev->gfx.grbm_soft_reset) && (!adev->gfx.srbm_soft_reset)) return 0; grbm_soft_reset = adev->gfx.grbm_soft_reset; - srbm_soft_reset = adev->gfx.srbm_soft_reset; if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPF) || diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index af8ccb014be3..f62d570a81a8 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -86,6 +86,7 @@ MODULE_FIRMWARE("amdgpu/picasso_me.bin"); MODULE_FIRMWARE("amdgpu/picasso_mec.bin"); MODULE_FIRMWARE("amdgpu/picasso_mec2.bin"); MODULE_FIRMWARE("amdgpu/picasso_rlc.bin"); +MODULE_FIRMWARE("amdgpu/picasso_rlc_am4.bin"); MODULE_FIRMWARE("amdgpu/raven2_ce.bin"); MODULE_FIRMWARE("amdgpu/raven2_pfp.bin"); @@ -645,7 +646,20 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) adev->gfx.ce_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); adev->gfx.ce_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name); + /* + * For Picasso && AM4 SOCKET board, we use picasso_rlc_am4.bin + * instead of picasso_rlc.bin. + * Judgment method: + * PCO AM4: revision >= 0xC8 && revision <= 0xCF + * or revision >= 0xD8 && revision <= 0xDF + * otherwise is PCO FP5 + */ + if (!strcmp(chip_name, "picasso") && + (((adev->pdev->revision >= 0xC8) && (adev->pdev->revision <= 0xCF)) || + ((adev->pdev->revision >= 0xD8) && (adev->pdev->revision <= 0xDF)))) + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc_am4.bin", chip_name); + else + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name); err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev); if (err) goto out; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 3a4e5d8d5162..ce150de723c9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -244,6 +244,62 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev, return 0; } +/** + * vega10_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool gmc_v9_0_prescreen_iv(struct amdgpu_device *adev, + struct amdgpu_iv_entry *entry, + uint64_t addr) +{ + struct amdgpu_vm *vm; + u64 key; + int r; + + /* No PASID, can't identify faulting process */ + if (!entry->pasid) + return true; + + /* Not a retry fault */ + if (!(entry->src_data[1] & 0x80)) + return true; + + /* Track retry faults in per-VM fault FIFO. */ + spin_lock(&adev->vm_manager.pasid_lock); + vm = idr_find(&adev->vm_manager.pasid_idr, entry->pasid); + if (!vm) { + /* VM not found, process it normally */ + spin_unlock(&adev->vm_manager.pasid_lock); + return true; + } + + key = AMDGPU_VM_FAULT(entry->pasid, addr); + r = amdgpu_vm_add_fault(vm->fault_hash, key); + + /* Hash table is full or the fault is already being processed, + * ignore further page faults + */ + if (r != 0) { + spin_unlock(&adev->vm_manager.pasid_lock); + return false; + } + /* No locking required with single writer and single reader */ + r = kfifo_put(&vm->faults, key); + if (!r) { + /* FIFO is full. Ignore it until there is space */ + amdgpu_vm_clear_fault(vm->fault_hash, key); + spin_unlock(&adev->vm_manager.pasid_lock); + return false; + } + + spin_unlock(&adev->vm_manager.pasid_lock); + /* It's the first fault for this address, process it normally */ + return true; +} + static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -255,6 +311,9 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev, addr = (u64)entry->src_data[0] << 12; addr |= ((u64)entry->src_data[1] & 0xf) << 44; + if (!gmc_v9_0_prescreen_iv(adev, entry, addr)) + return 1; /* This also prevents sending it to KFD */ + if (!amdgpu_sriov_vf(adev)) { status = RREG32(hub->vm_l2_pro_fault_status); WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1); @@ -902,6 +961,9 @@ static int gmc_v9_0_sw_init(void *handle) /* This interrupt is VMC page fault.*/ r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VMC, VMC_1_0__SRCID__VM_FAULT, &adev->gmc.vm_fault); + if (r) + return r; + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_UTCL2, UTCL2_1_0__SRCID__FAULT, &adev->gmc.vm_fault); @@ -934,7 +996,7 @@ static int gmc_v9_0_sw_init(void *handle) } adev->need_swiotlb = drm_get_max_iomem() > ((u64)1 << dma_bits); - if (adev->asic_type == CHIP_VEGA20) { + if (adev->gmc.xgmi.supported) { r = gfxhub_v1_1_get_xgmi_info(adev); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c index cf0fc61aebe6..a3984d10b604 100644 --- a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c @@ -208,34 +208,6 @@ static u32 iceland_ih_get_wptr(struct amdgpu_device *adev) } /** - * iceland_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool iceland_ih_prescreen_iv(struct amdgpu_device *adev) -{ - u32 ring_index = adev->irq.ih.rptr >> 2; - u16 pasid; - - switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { - case 146: - case 147: - pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; - if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) - return true; - break; - default: - /* Not a VM fault */ - return true; - } - - adev->irq.ih.rptr += 16; - return false; -} - -/** * iceland_ih_decode_iv - decode an interrupt vector * * @adev: amdgpu_device pointer @@ -440,7 +412,6 @@ static const struct amd_ip_funcs iceland_ih_ip_funcs = { static const struct amdgpu_ih_funcs iceland_ih_funcs = { .get_wptr = iceland_ih_get_wptr, - .prescreen_iv = iceland_ih_prescreen_iv, .decode_iv = iceland_ih_decode_iv, .set_rptr = iceland_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h index 882bd83a28c4..0de00fbe9233 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h @@ -43,6 +43,8 @@ enum psp_gfx_crtl_cmd_id GFX_CTRL_CMD_ID_ENABLE_INT = 0x00050000, /* enable PSP-to-Gfx interrupt */ GFX_CTRL_CMD_ID_DISABLE_INT = 0x00060000, /* disable PSP-to-Gfx interrupt */ GFX_CTRL_CMD_ID_MODE1_RST = 0x00070000, /* trigger the Mode 1 reset */ + GFX_CTRL_CMD_ID_CONSUME_CMD = 0x000A0000, /* send interrupt to psp for updating write pointer of vf */ + GFX_CTRL_CMD_ID_DESTROY_GPCOM_RING = 0x000C0000, /* destroy GPCOM ring */ GFX_CTRL_CMD_ID_MAX = 0x000F0000, /* max command ID */ }; @@ -89,7 +91,8 @@ enum psp_gfx_cmd_id GFX_CMD_ID_LOAD_IP_FW = 0x00000006, /* load HW IP FW */ GFX_CMD_ID_DESTROY_TMR = 0x00000007, /* destroy TMR region */ GFX_CMD_ID_SAVE_RESTORE = 0x00000008, /* save/restore HW IP FW */ - + GFX_CMD_ID_SETUP_VMR = 0x00000009, /* setup VMR region */ + GFX_CMD_ID_DESTROY_VMR = 0x0000000A, /* destroy VMR region */ }; diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c index e5dd052d9e06..6c9a1b748ca7 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c @@ -171,8 +171,11 @@ static int psp_v11_0_bootloader_load_sysdrv(struct psp_context *psp) * are already been loaded. */ sol_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81); - if (sol_reg) + if (sol_reg) { + psp->sos_fw_version = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_58); + printk("sos fw version = 0x%x.\n", psp->sos_fw_version); return 0; + } /* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1 */ ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35), @@ -296,26 +299,47 @@ static int psp_v11_0_ring_create(struct psp_context *psp, struct psp_ring *ring = &psp->km_ring; struct amdgpu_device *adev = psp->adev; - /* Write low address of the ring to C2PMSG_69 */ - psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr); - WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, psp_ring_reg); - /* Write high address of the ring to C2PMSG_70 */ - psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr); - WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, psp_ring_reg); - /* Write size of ring to C2PMSG_71 */ - psp_ring_reg = ring->ring_size; - WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_71, psp_ring_reg); - /* Write the ring initialization command to C2PMSG_64 */ - psp_ring_reg = ring_type; - psp_ring_reg = psp_ring_reg << 16; - WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, psp_ring_reg); - - /* there might be handshake issue with hardware which needs delay */ - mdelay(20); - - /* Wait for response flag (bit 31) in C2PMSG_64 */ - ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), - 0x80000000, 0x8000FFFF, false); + if (psp_support_vmr_ring(psp)) { + /* Write low address of the ring to C2PMSG_102 */ + psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_102, psp_ring_reg); + /* Write high address of the ring to C2PMSG_103 */ + psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_103, psp_ring_reg); + + /* Write the ring initialization command to C2PMSG_101 */ + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_101, + GFX_CTRL_CMD_ID_INIT_GPCOM_RING); + + /* there might be handshake issue with hardware which needs delay */ + mdelay(20); + + /* Wait for response flag (bit 31) in C2PMSG_101 */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101), + 0x80000000, 0x8000FFFF, false); + + } else { + /* Write low address of the ring to C2PMSG_69 */ + psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, psp_ring_reg); + /* Write high address of the ring to C2PMSG_70 */ + psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, psp_ring_reg); + /* Write size of ring to C2PMSG_71 */ + psp_ring_reg = ring->ring_size; + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_71, psp_ring_reg); + /* Write the ring initialization command to C2PMSG_64 */ + psp_ring_reg = ring_type; + psp_ring_reg = psp_ring_reg << 16; + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, psp_ring_reg); + + /* there might be handshake issue with hardware which needs delay */ + mdelay(20); + + /* Wait for response flag (bit 31) in C2PMSG_64 */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), + 0x80000000, 0x8000FFFF, false); + } return ret; } @@ -326,15 +350,24 @@ static int psp_v11_0_ring_stop(struct psp_context *psp, int ret = 0; struct amdgpu_device *adev = psp->adev; - /* Write the ring destroy command to C2PMSG_64 */ - WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, GFX_CTRL_CMD_ID_DESTROY_RINGS); + /* Write the ring destroy command*/ + if (psp_support_vmr_ring(psp)) + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_101, + GFX_CTRL_CMD_ID_DESTROY_GPCOM_RING); + else + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, + GFX_CTRL_CMD_ID_DESTROY_RINGS); /* there might be handshake issue with hardware which needs delay */ mdelay(20); - /* Wait for response flag (bit 31) in C2PMSG_64 */ - ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), - 0x80000000, 0x80000000, false); + /* Wait for response flag (bit 31) */ + if (psp_support_vmr_ring(psp)) + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101), + 0x80000000, 0x80000000, false); + else + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), + 0x80000000, 0x80000000, false); return ret; } @@ -373,7 +406,10 @@ static int psp_v11_0_cmd_submit(struct psp_context *psp, uint32_t rb_frame_size_dw = sizeof(struct psp_gfx_rb_frame) / 4; /* KM (GPCOM) prepare write pointer */ - psp_write_ptr_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67); + if (psp_support_vmr_ring(psp)) + psp_write_ptr_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_102); + else + psp_write_ptr_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67); /* Update KM RB frame pointer to new frame */ /* write_frame ptr increments by size of rb_frame in bytes */ @@ -402,7 +438,11 @@ static int psp_v11_0_cmd_submit(struct psp_context *psp, /* Update the write Pointer in DWORDs */ psp_write_ptr_reg = (psp_write_ptr_reg + rb_frame_size_dw) % ring_size_dw; - WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67, psp_write_ptr_reg); + if (psp_support_vmr_ring(psp)) { + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_102, psp_write_ptr_reg); + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_101, GFX_CTRL_CMD_ID_CONSUME_CMD); + } else + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67, psp_write_ptr_reg); return 0; } @@ -547,7 +587,7 @@ static int psp_v11_0_mode1_reset(struct psp_context *psp) /*send the mode 1 reset command*/ WREG32(offset, GFX_CTRL_CMD_ID_MODE1_RST); - mdelay(1000); + msleep(500); offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33); diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c index 7efb823dd3b1..7357fd56e614 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c @@ -592,7 +592,7 @@ static int psp_v3_1_mode1_reset(struct psp_context *psp) /*send the mode 1 reset command*/ WREG32(offset, GFX_CTRL_CMD_ID_MODE1_RST); - mdelay(1000); + msleep(500); offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33); diff --git a/drivers/gpu/drm/amd/amdgpu/si_ih.c b/drivers/gpu/drm/amd/amdgpu/si_ih.c index b3d7d9f83202..2938fb9f17cc 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/si_ih.c @@ -118,19 +118,6 @@ static u32 si_ih_get_wptr(struct amdgpu_device *adev) return (wptr & adev->irq.ih.ptr_mask); } -/** - * si_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool si_ih_prescreen_iv(struct amdgpu_device *adev) -{ - /* Process all interrupts */ - return true; -} - static void si_ih_decode_iv(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { @@ -301,7 +288,6 @@ static const struct amd_ip_funcs si_ih_ip_funcs = { static const struct amdgpu_ih_funcs si_ih_funcs = { .get_wptr = si_ih_get_wptr, - .prescreen_iv = si_ih_prescreen_iv, .decode_iv = si_ih_decode_iv, .set_rptr = si_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 83624e150ca7..8849b74078d6 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -507,6 +507,9 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) return -EINVAL; } + if (adev->asic_type == CHIP_VEGA20) + adev->gmc.xgmi.supported = true; + if (adev->flags & AMD_IS_APU) adev->nbio_funcs = &nbio_v7_0_funcs; else if (adev->asic_type == CHIP_VEGA20) diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c index dcdbb4d72472..15da06ddeb75 100644 --- a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c @@ -219,34 +219,6 @@ static u32 tonga_ih_get_wptr(struct amdgpu_device *adev) } /** - * tonga_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool tonga_ih_prescreen_iv(struct amdgpu_device *adev) -{ - u32 ring_index = adev->irq.ih.rptr >> 2; - u16 pasid; - - switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { - case 146: - case 147: - pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; - if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) - return true; - break; - default: - /* Not a VM fault */ - return true; - } - - adev->irq.ih.rptr += 16; - return false; -} - -/** * tonga_ih_decode_iv - decode an interrupt vector * * @adev: amdgpu_device pointer @@ -506,7 +478,6 @@ static const struct amd_ip_funcs tonga_ih_ip_funcs = { static const struct amdgpu_ih_funcs tonga_ih_funcs = { .get_wptr = tonga_ih_get_wptr, - .prescreen_iv = tonga_ih_prescreen_iv, .decode_iv = tonga_ih_decode_iv, .set_rptr = tonga_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index d84b687240d1..2c250b01a903 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -220,90 +220,6 @@ static u32 vega10_ih_get_wptr(struct amdgpu_device *adev) } /** - * vega10_ih_prescreen_iv - prescreen an interrupt vector - * - * @adev: amdgpu_device pointer - * - * Returns true if the interrupt vector should be further processed. - */ -static bool vega10_ih_prescreen_iv(struct amdgpu_device *adev) -{ - u32 ring_index = adev->irq.ih.rptr >> 2; - u32 dw0, dw3, dw4, dw5; - u16 pasid; - u64 addr, key; - struct amdgpu_vm *vm; - int r; - - dw0 = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); - dw3 = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); - dw4 = le32_to_cpu(adev->irq.ih.ring[ring_index + 4]); - dw5 = le32_to_cpu(adev->irq.ih.ring[ring_index + 5]); - - /* Filter retry page faults, let only the first one pass. If - * there are too many outstanding faults, ignore them until - * some faults get cleared. - */ - switch (dw0 & 0xff) { - case SOC15_IH_CLIENTID_VMC: - case SOC15_IH_CLIENTID_UTCL2: - break; - default: - /* Not a VM fault */ - return true; - } - - pasid = dw3 & 0xffff; - /* No PASID, can't identify faulting process */ - if (!pasid) - return true; - - /* Not a retry fault, check fault credit */ - if (!(dw5 & 0x80)) { - if (!amdgpu_vm_pasid_fault_credit(adev, pasid)) - goto ignore_iv; - return true; - } - - /* Track retry faults in per-VM fault FIFO. */ - spin_lock(&adev->vm_manager.pasid_lock); - vm = idr_find(&adev->vm_manager.pasid_idr, pasid); - addr = ((u64)(dw5 & 0xf) << 44) | ((u64)dw4 << 12); - key = AMDGPU_VM_FAULT(pasid, addr); - if (!vm) { - /* VM not found, process it normally */ - spin_unlock(&adev->vm_manager.pasid_lock); - return true; - } else { - r = amdgpu_vm_add_fault(vm->fault_hash, key); - - /* Hash table is full or the fault is already being processed, - * ignore further page faults - */ - if (r != 0) { - spin_unlock(&adev->vm_manager.pasid_lock); - goto ignore_iv; - } - } - /* No locking required with single writer and single reader */ - r = kfifo_put(&vm->faults, key); - if (!r) { - /* FIFO is full. Ignore it until there is space */ - amdgpu_vm_clear_fault(vm->fault_hash, key); - spin_unlock(&adev->vm_manager.pasid_lock); - goto ignore_iv; - } - - spin_unlock(&adev->vm_manager.pasid_lock); - /* It's the first fault for this address, process it normally */ - return true; - -ignore_iv: - adev->irq.ih.rptr += 32; - return false; -} - -/** * vega10_ih_decode_iv - decode an interrupt vector * * @adev: amdgpu_device pointer @@ -487,7 +403,6 @@ const struct amd_ip_funcs vega10_ih_ip_funcs = { static const struct amdgpu_ih_funcs vega10_ih_funcs = { .get_wptr = vega10_ih_get_wptr, - .prescreen_iv = vega10_ih_prescreen_iv, .decode_iv = vega10_ih_decode_iv, .set_rptr = vega10_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 5f4062b41add..3623538baf6f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -33,6 +33,7 @@ #include <linux/time.h> #include <linux/mm.h> #include <linux/mman.h> +#include <linux/dma-buf.h> #include <asm/processor.h> #include "kfd_priv.h" #include "kfd_device_queue_manager.h" @@ -1273,6 +1274,12 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep, return -EINVAL; } + if (flags & KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL) { + if (args->size != kfd_doorbell_process_slice(dev)) + return -EINVAL; + offset = kfd_get_process_doorbells(dev, p); + } + mutex_lock(&p->mutex); pdd = kfd_bind_process_to_device(dev, p); @@ -1550,6 +1557,115 @@ copy_from_user_failed: return err; } +static int kfd_ioctl_get_dmabuf_info(struct file *filep, + struct kfd_process *p, void *data) +{ + struct kfd_ioctl_get_dmabuf_info_args *args = data; + struct kfd_dev *dev = NULL; + struct kgd_dev *dma_buf_kgd; + void *metadata_buffer = NULL; + uint32_t flags; + unsigned int i; + int r; + + /* Find a KFD GPU device that supports the get_dmabuf_info query */ + for (i = 0; kfd_topology_enum_kfd_devices(i, &dev) == 0; i++) + if (dev) + break; + if (!dev) + return -EINVAL; + + if (args->metadata_ptr) { + metadata_buffer = kzalloc(args->metadata_size, GFP_KERNEL); + if (!metadata_buffer) + return -ENOMEM; + } + + /* Get dmabuf info from KGD */ + r = amdgpu_amdkfd_get_dmabuf_info(dev->kgd, args->dmabuf_fd, + &dma_buf_kgd, &args->size, + metadata_buffer, args->metadata_size, + &args->metadata_size, &flags); + if (r) + goto exit; + + /* Reverse-lookup gpu_id from kgd pointer */ + dev = kfd_device_by_kgd(dma_buf_kgd); + if (!dev) { + r = -EINVAL; + goto exit; + } + args->gpu_id = dev->id; + args->flags = flags; + + /* Copy metadata buffer to user mode */ + if (metadata_buffer) { + r = copy_to_user((void __user *)args->metadata_ptr, + metadata_buffer, args->metadata_size); + if (r != 0) + r = -EFAULT; + } + +exit: + kfree(metadata_buffer); + + return r; +} + +static int kfd_ioctl_import_dmabuf(struct file *filep, + struct kfd_process *p, void *data) +{ + struct kfd_ioctl_import_dmabuf_args *args = data; + struct kfd_process_device *pdd; + struct dma_buf *dmabuf; + struct kfd_dev *dev; + int idr_handle; + uint64_t size; + void *mem; + int r; + + dev = kfd_device_by_id(args->gpu_id); + if (!dev) + return -EINVAL; + + dmabuf = dma_buf_get(args->dmabuf_fd); + if (!dmabuf) + return -EINVAL; + + mutex_lock(&p->mutex); + + pdd = kfd_bind_process_to_device(dev, p); + if (IS_ERR(pdd)) { + r = PTR_ERR(pdd); + goto err_unlock; + } + + r = amdgpu_amdkfd_gpuvm_import_dmabuf(dev->kgd, dmabuf, + args->va_addr, pdd->vm, + (struct kgd_mem **)&mem, &size, + NULL); + if (r) + goto err_unlock; + + idr_handle = kfd_process_device_create_obj_handle(pdd, mem); + if (idr_handle < 0) { + r = -EFAULT; + goto err_free; + } + + mutex_unlock(&p->mutex); + + args->handle = MAKE_HANDLE(args->gpu_id, idr_handle); + + return 0; + +err_free: + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem); +err_unlock: + mutex_unlock(&p->mutex); + return r; +} + #define AMDKFD_IOCTL_DEF(ioctl, _func, _flags) \ [_IOC_NR(ioctl)] = {.cmd = ioctl, .func = _func, .flags = _flags, \ .cmd_drv = 0, .name = #ioctl} @@ -1635,7 +1751,13 @@ static const struct amdkfd_ioctl_desc amdkfd_ioctls[] = { kfd_ioctl_set_cu_mask, 0), AMDKFD_IOCTL_DEF(AMDKFD_IOC_GET_QUEUE_WAVE_STATE, - kfd_ioctl_get_queue_wave_state, 0) + kfd_ioctl_get_queue_wave_state, 0), + + AMDKFD_IOCTL_DEF(AMDKFD_IOC_GET_DMABUF_INFO, + kfd_ioctl_get_dmabuf_info, 0), + + AMDKFD_IOCTL_DEF(AMDKFD_IOC_IMPORT_DMABUF, + kfd_ioctl_import_dmabuf, 0), }; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index dec8e64f36bd..0689d4ccbbc0 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -793,6 +793,7 @@ struct kfd_topology_device *kfd_topology_device_by_proximity_domain( struct kfd_topology_device *kfd_topology_device_by_id(uint32_t gpu_id); struct kfd_dev *kfd_device_by_id(uint32_t gpu_id); struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev); +struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd); int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_dev **kdev); int kfd_numa_node_to_apic_id(int numa_node_id); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c index aa793fcbbdcc..5f5b2acedbac 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c @@ -101,7 +101,25 @@ struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev) down_read(&topology_lock); list_for_each_entry(top_dev, &topology_device_list, list) - if (top_dev->gpu->pdev == pdev) { + if (top_dev->gpu && top_dev->gpu->pdev == pdev) { + device = top_dev->gpu; + break; + } + + up_read(&topology_lock); + + return device; +} + +struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd) +{ + struct kfd_topology_device *top_dev; + struct kfd_dev *device = NULL; + + down_read(&topology_lock); + + list_for_each_entry(top_dev, &topology_device_list, list) + if (top_dev->gpu && top_dev->gpu->kgd == kgd) { device = top_dev->gpu; break; } diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index 32e791d9b9a8..c13856a46d8e 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -23,6 +23,9 @@ * */ +/* The caprices of the preprocessor require that this be declared right here */ +#define CREATE_TRACE_POINTS + #include "dm_services_types.h" #include "dc.h" #include "dc/inc/core_types.h" @@ -54,6 +57,7 @@ #include <drm/drmP.h> #include <drm/drm_atomic.h> +#include <drm/drm_atomic_uapi.h> #include <drm/drm_atomic_helper.h> #include <drm/drm_dp_mst_helper.h> #include <drm/drm_fb_helper.h> @@ -72,6 +76,7 @@ #include "modules/inc/mod_freesync.h" #include "modules/power/power_helpers.h" +#include "modules/inc/mod_info_packet.h" #define FIRMWARE_RAVEN_DMCU "amdgpu/raven_dmcu.bin" MODULE_FIRMWARE(FIRMWARE_RAVEN_DMCU); @@ -129,6 +134,8 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state); static int amdgpu_dm_atomic_check(struct drm_device *dev, struct drm_atomic_state *state); +static void handle_cursor_update(struct drm_plane *plane, + struct drm_plane_state *old_plane_state); @@ -398,6 +405,8 @@ static int amdgpu_dm_init(struct amdgpu_device *adev) /* Zero all the fields */ memset(&init_data, 0, sizeof(init_data)); + mutex_init(&adev->dm.dc_lock); + if(amdgpu_dm_irq_init(adev)) { DRM_ERROR("amdgpu: failed to initialize DM IRQ support.\n"); goto error; @@ -512,6 +521,9 @@ static void amdgpu_dm_fini(struct amdgpu_device *adev) /* DC Destroy TODO: Replace destroy DAL */ if (adev->dm.dc) dc_destroy(&adev->dm.dc); + + mutex_destroy(&adev->dm.dc_lock); + return; } @@ -2930,6 +2942,7 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector, if (dm_state && dm_state->freesync_capable) stream->ignore_msa_timing_param = true; + finish: if (sink && sink->sink_signal == SIGNAL_TYPE_VIRTUAL && aconnector->base.force != DRM_FORCE_ON) dc_sink_release(sink); @@ -3001,6 +3014,7 @@ dm_crtc_duplicate_state(struct drm_crtc *crtc) state->abm_level = cur->abm_level; state->vrr_supported = cur->vrr_supported; state->freesync_config = cur->freesync_config; + state->crc_enabled = cur->crc_enabled; /* TODO Duplicate dc_stream after objects are stream object is flattened */ @@ -3094,10 +3108,8 @@ int amdgpu_dm_connector_atomic_set_property(struct drm_connector *connector, rmx_type = RMX_FULL; break; case DRM_MODE_SCALE_NONE: - rmx_type = RMX_OFF; - break; default: - rmx_type = RMX_ASPECT; + rmx_type = RMX_OFF; break; } @@ -3210,10 +3222,11 @@ void amdgpu_dm_connector_funcs_reset(struct drm_connector *connector) state = kzalloc(sizeof(*state), GFP_KERNEL); if (state) { - state->scaling = RMX_ASPECT; + state->scaling = RMX_OFF; state->underscan_enable = false; state->underscan_hborder = 0; state->underscan_vborder = 0; + state->max_bpc = 8; __drm_atomic_helper_connector_reset(connector, &state->base); } @@ -3235,6 +3248,11 @@ amdgpu_dm_connector_atomic_duplicate_state(struct drm_connector *connector) new_state->freesync_capable = state->freesync_capable; new_state->abm_level = state->abm_level; + new_state->scaling = state->scaling; + new_state->underscan_enable = state->underscan_enable; + new_state->underscan_hborder = state->underscan_hborder; + new_state->underscan_vborder = state->underscan_vborder; + new_state->max_bpc = state->max_bpc; return &new_state->base; } @@ -3607,10 +3625,43 @@ static int dm_plane_atomic_check(struct drm_plane *plane, return -EINVAL; } +static int dm_plane_atomic_async_check(struct drm_plane *plane, + struct drm_plane_state *new_plane_state) +{ + /* Only support async updates on cursor planes. */ + if (plane->type != DRM_PLANE_TYPE_CURSOR) + return -EINVAL; + + return 0; +} + +static void dm_plane_atomic_async_update(struct drm_plane *plane, + struct drm_plane_state *new_state) +{ + struct drm_plane_state *old_state = + drm_atomic_get_old_plane_state(new_state->state, plane); + + if (plane->state->fb != new_state->fb) + drm_atomic_set_fb_for_plane(plane->state, new_state->fb); + + plane->state->src_x = new_state->src_x; + plane->state->src_y = new_state->src_y; + plane->state->src_w = new_state->src_w; + plane->state->src_h = new_state->src_h; + plane->state->crtc_x = new_state->crtc_x; + plane->state->crtc_y = new_state->crtc_y; + plane->state->crtc_w = new_state->crtc_w; + plane->state->crtc_h = new_state->crtc_h; + + handle_cursor_update(plane, old_state); +} + static const struct drm_plane_helper_funcs dm_plane_helper_funcs = { .prepare_fb = dm_plane_helper_prepare_fb, .cleanup_fb = dm_plane_helper_cleanup_fb, .atomic_check = dm_plane_atomic_check, + .atomic_async_check = dm_plane_atomic_async_check, + .atomic_async_update = dm_plane_atomic_async_update }; /* @@ -4299,6 +4350,7 @@ static int get_cursor_position(struct drm_plane *plane, struct drm_crtc *crtc, static void handle_cursor_update(struct drm_plane *plane, struct drm_plane_state *old_plane_state) { + struct amdgpu_device *adev = plane->dev->dev_private; struct amdgpu_framebuffer *afb = to_amdgpu_framebuffer(plane->state->fb); struct drm_crtc *crtc = afb ? plane->state->crtc : old_plane_state->crtc; struct dm_crtc_state *crtc_state = crtc ? to_dm_crtc_state(crtc->state) : NULL; @@ -4323,9 +4375,12 @@ static void handle_cursor_update(struct drm_plane *plane, if (!position.enable) { /* turn off cursor */ - if (crtc_state && crtc_state->stream) + if (crtc_state && crtc_state->stream) { + mutex_lock(&adev->dm.dc_lock); dc_stream_set_cursor_position(crtc_state->stream, &position); + mutex_unlock(&adev->dm.dc_lock); + } return; } @@ -4343,6 +4398,7 @@ static void handle_cursor_update(struct drm_plane *plane, attributes.pitch = attributes.width; if (crtc_state->stream) { + mutex_lock(&adev->dm.dc_lock); if (!dc_stream_set_cursor_attributes(crtc_state->stream, &attributes)) DRM_ERROR("DC failed to set cursor attributes\n"); @@ -4350,6 +4406,7 @@ static void handle_cursor_update(struct drm_plane *plane, if (!dc_stream_set_cursor_position(crtc_state->stream, &position)) DRM_ERROR("DC failed to set cursor position\n"); + mutex_unlock(&adev->dm.dc_lock); } } @@ -4423,8 +4480,8 @@ static void update_freesync_state_on_stream( dm->freesync_module, new_stream, &vrr, - packet_type_vrr, - transfer_func_unknown, + PACKET_TYPE_VRR, + TRANSFER_FUNC_UNKNOWN, &vrr_infopacket); new_crtc_state->freesync_timing_changed = @@ -4565,6 +4622,7 @@ static void amdgpu_dm_do_flip(struct drm_crtc *crtc, &acrtc_state->stream->vrr_infopacket; } + mutex_lock(&adev->dm.dc_lock); dc_commit_updates_for_stream(adev->dm.dc, surface_updates, 1, @@ -4572,6 +4630,7 @@ static void amdgpu_dm_do_flip(struct drm_crtc *crtc, &stream_update, &surface_updates->surface, state); + mutex_unlock(&adev->dm.dc_lock); DRM_DEBUG_DRIVER("%s Flipping to hi: 0x%x, low: 0x%x \n", __func__, @@ -4586,6 +4645,7 @@ static void amdgpu_dm_do_flip(struct drm_crtc *crtc, * with a dc_plane_state and follow the atomic model a bit more closely here. */ static bool commit_planes_to_stream( + struct amdgpu_display_manager *dm, struct dc *dc, struct dc_plane_state **plane_states, uint8_t new_plane_count, @@ -4662,11 +4722,13 @@ static bool commit_planes_to_stream( updates[i].scaling_info = &scaling_info[i]; } + mutex_lock(&dm->dc_lock); dc_commit_updates_for_stream( dc, updates, new_plane_count, dc_stream, stream_update, plane_states, state); + mutex_unlock(&dm->dc_lock); kfree(flip_addr); kfree(plane_info); @@ -4772,7 +4834,8 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, dc_stream_attach->abm_level = acrtc_state->abm_level; - if (false == commit_planes_to_stream(dm->dc, + if (false == commit_planes_to_stream(dm, + dm->dc, plane_states_constructed, planes_count, acrtc_state, @@ -4942,7 +5005,9 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) if (dc_state) { dm_enable_per_frame_crtc_master_sync(dc_state); + mutex_lock(&dm->dc_lock); WARN_ON(!dc_commit_state(dm->dc, dc_state)); + mutex_unlock(&dm->dc_lock); } for_each_new_crtc_in_state(state, crtc, new_crtc_state, i) { @@ -5004,6 +5069,7 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) /*TODO How it works with MPO ?*/ if (!commit_planes_to_stream( + dm, dm->dc, status->plane_states, status->plane_count, @@ -5896,6 +5962,13 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, ret = -EINVAL; goto fail; } + } else if (state->legacy_cursor_update) { + /* + * This is a fast cursor update coming from the plane update + * helper, check if it can be done asynchronously for better + * performance. + */ + state->async_update = !drm_atomic_helper_async_check(dev, state); } /* Must be success */ diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h index 4326dc256491..25bb91ee80ba 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h @@ -135,6 +135,14 @@ struct amdgpu_display_manager { struct drm_modeset_lock atomic_obj_lock; /** + * @dc_lock: + * + * Guards access to DC functions that can issue register write + * sequences. + */ + struct mutex dc_lock; + + /** * @irq_handler_list_low_tab: * * Low priority IRQ handler table. diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_trace.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_trace.h new file mode 100644 index 000000000000..d898981684d5 --- /dev/null +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_trace.h @@ -0,0 +1,104 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: AMD + * + */ + +#undef TRACE_SYSTEM +#define TRACE_SYSTEM amdgpu_dm + +#if !defined(_AMDGPU_DM_TRACE_H) || defined(TRACE_HEADER_MULTI_READ) +#define _AMDGPU_DM_TRACE_H_ + +#include <linux/tracepoint.h> + +TRACE_EVENT(amdgpu_dc_rreg, + TP_PROTO(unsigned long *read_count, uint32_t reg, uint32_t value), + TP_ARGS(read_count, reg, value), + TP_STRUCT__entry( + __field(uint32_t, reg) + __field(uint32_t, value) + ), + TP_fast_assign( + __entry->reg = reg; + __entry->value = value; + *read_count = *read_count + 1; + ), + TP_printk("reg=0x%08lx, value=0x%08lx", + (unsigned long)__entry->reg, + (unsigned long)__entry->value) +); + +TRACE_EVENT(amdgpu_dc_wreg, + TP_PROTO(unsigned long *write_count, uint32_t reg, uint32_t value), + TP_ARGS(write_count, reg, value), + TP_STRUCT__entry( + __field(uint32_t, reg) + __field(uint32_t, value) + ), + TP_fast_assign( + __entry->reg = reg; + __entry->value = value; + *write_count = *write_count + 1; + ), + TP_printk("reg=0x%08lx, value=0x%08lx", + (unsigned long)__entry->reg, + (unsigned long)__entry->value) +); + + +TRACE_EVENT(amdgpu_dc_performance, + TP_PROTO(unsigned long read_count, unsigned long write_count, + unsigned long *last_read, unsigned long *last_write, + const char *func, unsigned int line), + TP_ARGS(read_count, write_count, last_read, last_write, func, line), + TP_STRUCT__entry( + __field(uint32_t, reads) + __field(uint32_t, writes) + __field(uint32_t, read_delta) + __field(uint32_t, write_delta) + __string(func, func) + __field(uint32_t, line) + ), + TP_fast_assign( + __entry->reads = read_count; + __entry->writes = write_count; + __entry->read_delta = read_count - *last_read; + __entry->write_delta = write_count - *last_write; + __assign_str(func, func); + __entry->line = line; + *last_read = read_count; + *last_write = write_count; + ), + TP_printk("%s:%d reads=%08ld (%08ld total), writes=%08ld (%08ld total)", + __get_str(func), __entry->line, + (unsigned long)__entry->read_delta, + (unsigned long)__entry->reads, + (unsigned long)__entry->write_delta, + (unsigned long)__entry->writes) +); +#endif /* _AMDGPU_DM_TRACE_H_ */ + +#undef TRACE_INCLUDE_PATH +#define TRACE_INCLUDE_PATH . +#define TRACE_INCLUDE_FILE amdgpu_dm_trace +#include <trace/define_trace.h> diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index dba6b57830c7..d9c57984394b 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -62,6 +62,55 @@ const static char DC_BUILD_ID[] = "production-build"; +/** + * DOC: Overview + * + * DC is the OS-agnostic component of the amdgpu DC driver. + * + * DC maintains and validates a set of structs representing the state of the + * driver and writes that state to AMD hardware + * + * Main DC HW structs: + * + * struct dc - The central struct. One per driver. Created on driver load, + * destroyed on driver unload. + * + * struct dc_context - One per driver. + * Used as a backpointer by most other structs in dc. + * + * struct dc_link - One per connector (the physical DP, HDMI, miniDP, or eDP + * plugpoints). Created on driver load, destroyed on driver unload. + * + * struct dc_sink - One per display. Created on boot or hotplug. + * Destroyed on shutdown or hotunplug. A dc_link can have a local sink + * (the display directly attached). It may also have one or more remote + * sinks (in the Multi-Stream Transport case) + * + * struct resource_pool - One per driver. Represents the hw blocks not in the + * main pipeline. Not directly accessible by dm. + * + * Main dc state structs: + * + * These structs can be created and destroyed as needed. There is a full set of + * these structs in dc->current_state representing the currently programmed state. + * + * struct dc_state - The global DC state to track global state information, + * such as bandwidth values. + * + * struct dc_stream_state - Represents the hw configuration for the pipeline from + * a framebuffer to a display. Maps one-to-one with dc_sink. + * + * struct dc_plane_state - Represents a framebuffer. Each stream has at least one, + * and may have more in the Multi-Plane Overlay case. + * + * struct resource_context - Represents the programmable state of everything in + * the resource_pool. Not directly accessible by dm. + * + * struct pipe_ctx - A member of struct resource_context. Represents the + * internal hardware pipeline components. Each dc_plane_state has either + * one or two (in the pipe-split case). + */ + /******************************************************************************* * Private functions ******************************************************************************/ @@ -175,6 +224,17 @@ failed_alloc: return false; } +static struct dc_perf_trace *dc_perf_trace_create(void) +{ + return kzalloc(sizeof(struct dc_perf_trace), GFP_KERNEL); +} + +static void dc_perf_trace_destroy(struct dc_perf_trace **perf_trace) +{ + kfree(*perf_trace); + *perf_trace = NULL; +} + /** ***************************************************************************** * Function: dc_stream_adjust_vmin_vmax @@ -240,7 +300,7 @@ bool dc_stream_get_crtc_position(struct dc *dc, } /** - * dc_stream_configure_crc: Configure CRC capture for the given stream. + * dc_stream_configure_crc() - Configure CRC capture for the given stream. * @dc: DC Object * @stream: The stream to configure CRC on. * @enable: Enable CRC if true, disable otherwise. @@ -292,7 +352,7 @@ bool dc_stream_configure_crc(struct dc *dc, struct dc_stream_state *stream, } /** - * dc_stream_get_crc: Get CRC values for the given stream. + * dc_stream_get_crc() - Get CRC values for the given stream. * @dc: DC object * @stream: The DC stream state of the stream to get CRCs from. * @r_cr, g_y, b_cb: CRC values for the three channels are stored here. @@ -536,6 +596,8 @@ static void destruct(struct dc *dc) if (dc->ctx->created_bios) dal_bios_parser_destroy(&dc->ctx->dc_bios); + dc_perf_trace_destroy(&dc->ctx->perf_trace); + kfree(dc->ctx); dc->ctx = NULL; @@ -659,6 +721,12 @@ static bool construct(struct dc *dc, goto fail; } + dc_ctx->perf_trace = dc_perf_trace_create(); + if (!dc_ctx->perf_trace) { + ASSERT_CRITICAL(false); + goto fail; + } + /* Create GPIO service */ dc_ctx->gpio_service = dal_gpio_service_create( dc_version, @@ -1329,6 +1397,11 @@ static enum surface_update_type check_update_surfaces_for_stream( return overall_type; } +/** + * dc_check_update_surfaces_for_stream() - Determine update type (fast, med, or full) + * + * See :c:type:`enum surface_update_type <surface_update_type>` for explanation of update types + */ enum surface_update_type dc_check_update_surfaces_for_stream( struct dc *dc, struct dc_surface_update *updates, @@ -1409,6 +1482,14 @@ static void commit_planes_do_stream_update(struct dc *dc, if (stream_update->output_csc_transform) dc_stream_program_csc_matrix(dc, stream); + if (stream_update->dither_option) { + resource_build_bit_depth_reduction_params(pipe_ctx->stream, + &pipe_ctx->stream->bit_depth_params); + pipe_ctx->stream_res.opp->funcs->opp_program_fmt(pipe_ctx->stream_res.opp, + &stream->bit_depth_params, + &stream->clamping); + } + /* Full fe update*/ if (update_type == UPDATE_TYPE_FAST) continue; @@ -1631,6 +1712,9 @@ enum dc_irq_source dc_interrupt_to_irq_source( return dal_irq_service_to_irq_source(dc->res_pool->irqs, src_id, ext_id); } +/** + * dc_interrupt_set() - Enable/disable an AMD hw interrupt source + */ bool dc_interrupt_set(struct dc *dc, enum dc_irq_source src, bool enable) { @@ -1724,6 +1808,11 @@ static bool link_add_remote_sink_helper(struct dc_link *dc_link, struct dc_sink return true; } +/** + * dc_link_add_remote_sink() - Create a sink and attach it to an existing link + * + * EDID length is in bytes + */ struct dc_sink *dc_link_add_remote_sink( struct dc_link *link, const uint8_t *edid, @@ -1782,6 +1871,12 @@ fail_add_sink: return NULL; } +/** + * dc_link_remove_remote_sink() - Remove a remote sink from a dc_link + * + * Note that this just removes the struct dc_sink - it doesn't + * program hardware or alter other members of dc_link + */ void dc_link_remove_remote_sink(struct dc_link *link, struct dc_sink *sink) { int i; diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link.c b/drivers/gpu/drm/amd/display/dc/core/dc_link.c index 948596a02392..4dc5846de5c4 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link.c @@ -198,6 +198,13 @@ static bool program_hpd_filter( return result; } +/** + * dc_link_detect_sink() - Determine if there is a sink connected + * + * @type: Returned connection type + * Does not detect downstream devices, such as MST sinks + * or display connected through active dongles + */ bool dc_link_detect_sink(struct dc_link *link, enum dc_connection_type *type) { uint32_t is_hpd_high = 0; @@ -324,9 +331,9 @@ static enum signal_type get_basic_signal_type( return SIGNAL_TYPE_NONE; } -/* - * @brief - * Check whether there is a dongle on DP connector +/** + * dc_link_is_dp_sink_present() - Check if there is a native DP + * or passive DP-HDMI dongle connected */ bool dc_link_is_dp_sink_present(struct dc_link *link) { @@ -593,6 +600,14 @@ static bool is_same_edid(struct dc_edid *old_edid, struct dc_edid *new_edid) return (memcmp(old_edid->raw_edid, new_edid->raw_edid, new_edid->length) == 0); } +/** + * dc_link_detect() - Detect if a sink is attached to a given link + * + * link->local_sink is created or destroyed as needed. + * + * This does not create remote sinks but will trigger DM + * to start MST detection if a branch is detected. + */ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) { struct dc_sink_init_data sink_init_data = { 0 }; diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c index 4d1f8ac069c1..849a3a3032f7 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c @@ -2196,7 +2196,7 @@ static void get_active_converter_info( } if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_11) { - uint8_t det_caps[4]; + uint8_t det_caps[16]; /* CTS 4.2.2.7 expects source to read Detailed Capabilities Info : 00080h-0008F.*/ union dwnstream_port_caps_byte0 *port_caps = (union dwnstream_port_caps_byte0 *)det_caps; core_link_read_dpcd(link, DP_DOWNSTREAM_PORT_0, diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c index 0bb844a7b990..c347afd1030f 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c @@ -1447,6 +1447,14 @@ static bool are_stream_backends_same( return true; } +/** + * dc_is_stream_unchanged() - Compare two stream states for equivalence. + * + * Checks if there a difference between the two states + * that would require a mode change. + * + * Does not compare cursor position or attributes. + */ bool dc_is_stream_unchanged( struct dc_stream_state *old_stream, struct dc_stream_state *stream) { @@ -1457,6 +1465,9 @@ bool dc_is_stream_unchanged( return true; } +/** + * dc_is_stream_scaling_unchanged() - Compare scaling rectangles of two streams. + */ bool dc_is_stream_scaling_unchanged( struct dc_stream_state *old_stream, struct dc_stream_state *stream) { @@ -1616,6 +1627,9 @@ bool resource_is_stream_unchanged( return false; } +/** + * dc_add_stream_to_ctx() - Add a new dc_stream_state to a dc_state. + */ enum dc_status dc_add_stream_to_ctx( struct dc *dc, struct dc_state *new_ctx, @@ -1640,6 +1654,9 @@ enum dc_status dc_add_stream_to_ctx( return res; } +/** + * dc_remove_stream_from_ctx() - Remove a stream from a dc_state. + */ enum dc_status dc_remove_stream_from_ctx( struct dc *dc, struct dc_state *new_ctx, @@ -1860,6 +1877,12 @@ enum dc_status resource_map_pool_resources( return DC_ERROR_UNEXPECTED; } +/** + * dc_resource_state_copy_construct_current() - Creates a new dc_state from existing state + * Is a shallow copy. Increments refcounts on existing streams and planes. + * @dc: copy out of dc->current_state + * @dst_ctx: copy into this + */ void dc_resource_state_copy_construct_current( const struct dc *dc, struct dc_state *dst_ctx) @@ -1875,6 +1898,14 @@ void dc_resource_state_construct( dst_ctx->dccg = dc->res_pool->clk_mgr; } +/** + * dc_validate_global_state() - Determine if HW can support a given state + * Checks HW resource availability and bandwidth requirement. + * @dc: dc struct for this driver + * @new_ctx: state to be validated + * + * Return: DC_OK if the result can be programmed. Otherwise, an error code. + */ enum dc_status dc_validate_global_state( struct dc *dc, struct dc_state *new_ctx) @@ -2202,113 +2233,15 @@ static void set_vendor_info_packet( struct dc_info_packet *info_packet, struct dc_stream_state *stream) { - uint32_t length = 0; - bool hdmi_vic_mode = false; - uint8_t checksum = 0; - uint32_t i = 0; - enum dc_timing_3d_format format; - // Can be different depending on packet content /*todo*/ - // unsigned int length = pPathMode->dolbyVision ? 24 : 5; - - info_packet->valid = false; - - format = stream->timing.timing_3d_format; - if (stream->view_format == VIEW_3D_FORMAT_NONE) - format = TIMING_3D_FORMAT_NONE; - - /* Can be different depending on packet content */ - length = 5; - - if (stream->timing.hdmi_vic != 0 - && stream->timing.h_total >= 3840 - && stream->timing.v_total >= 2160) - hdmi_vic_mode = true; - - /* According to HDMI 1.4a CTS, VSIF should be sent - * for both 3D stereo and HDMI VIC modes. - * For all other modes, there is no VSIF sent. */ + /* SPD info packet for FreeSync */ - if (format == TIMING_3D_FORMAT_NONE && !hdmi_vic_mode) + /* Check if Freesync is supported. Return if false. If true, + * set the corresponding bit in the info packet + */ + if (!stream->vsp_infopacket.valid) return; - /* 24bit IEEE Registration identifier (0x000c03). LSB first. */ - info_packet->sb[1] = 0x03; - info_packet->sb[2] = 0x0C; - info_packet->sb[3] = 0x00; - - /*PB4: 5 lower bytes = 0 (reserved). 3 higher bits = HDMI_Video_Format. - * The value for HDMI_Video_Format are: - * 0x0 (0b000) - No additional HDMI video format is presented in this - * packet - * 0x1 (0b001) - Extended resolution format present. 1 byte of HDMI_VIC - * parameter follows - * 0x2 (0b010) - 3D format indication present. 3D_Structure and - * potentially 3D_Ext_Data follows - * 0x3..0x7 (0b011..0b111) - reserved for future use */ - if (format != TIMING_3D_FORMAT_NONE) - info_packet->sb[4] = (2 << 5); - else if (hdmi_vic_mode) - info_packet->sb[4] = (1 << 5); - - /* PB5: If PB4 claims 3D timing (HDMI_Video_Format = 0x2): - * 4 lower bites = 0 (reserved). 4 higher bits = 3D_Structure. - * The value for 3D_Structure are: - * 0x0 - Frame Packing - * 0x1 - Field Alternative - * 0x2 - Line Alternative - * 0x3 - Side-by-Side (full) - * 0x4 - L + depth - * 0x5 - L + depth + graphics + graphics-depth - * 0x6 - Top-and-Bottom - * 0x7 - Reserved for future use - * 0x8 - Side-by-Side (Half) - * 0x9..0xE - Reserved for future use - * 0xF - Not used */ - switch (format) { - case TIMING_3D_FORMAT_HW_FRAME_PACKING: - case TIMING_3D_FORMAT_SW_FRAME_PACKING: - info_packet->sb[5] = (0x0 << 4); - break; - - case TIMING_3D_FORMAT_SIDE_BY_SIDE: - case TIMING_3D_FORMAT_SBS_SW_PACKED: - info_packet->sb[5] = (0x8 << 4); - length = 6; - break; - - case TIMING_3D_FORMAT_TOP_AND_BOTTOM: - case TIMING_3D_FORMAT_TB_SW_PACKED: - info_packet->sb[5] = (0x6 << 4); - break; - - default: - break; - } - - /*PB5: If PB4 is set to 0x1 (extended resolution format) - * fill PB5 with the correct HDMI VIC code */ - if (hdmi_vic_mode) - info_packet->sb[5] = stream->timing.hdmi_vic; - - /* Header */ - info_packet->hb0 = HDMI_INFOFRAME_TYPE_VENDOR; /* VSIF packet type. */ - info_packet->hb1 = 0x01; /* Version */ - - /* 4 lower bits = Length, 4 higher bits = 0 (reserved) */ - info_packet->hb2 = (uint8_t) (length); - - /* Calculate checksum */ - checksum = 0; - checksum += info_packet->hb0; - checksum += info_packet->hb1; - checksum += info_packet->hb2; - - for (i = 1; i <= length; i++) - checksum += info_packet->sb[i]; - - info_packet->sb[0] = (uint8_t) (0x100 - checksum); - - info_packet->valid = true; + *info_packet = stream->vsp_infopacket; } static void set_spd_info_packet( @@ -2364,10 +2297,6 @@ void dc_resource_state_destruct(struct dc_state *context) } } -/* - * Copy src_ctx into dst_ctx and retain all surfaces and streams referenced - * by the src_ctx - */ void dc_resource_state_copy_construct( const struct dc_state *src_ctx, struct dc_state *dst_ctx) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_stream.c b/drivers/gpu/drm/amd/display/dc/core/dc_stream.c index 780838a05f44..66e5c4623a49 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_stream.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_stream.c @@ -170,7 +170,7 @@ struct dc_stream_status *dc_stream_get_status( } /** - * Update the cursor attributes and set cursor surface address + * dc_stream_set_cursor_attributes() - Update cursor attributes and set cursor surface address */ bool dc_stream_set_cursor_attributes( struct dc_stream_state *stream, diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index dea8bc39c688..4b5bbb13ce7f 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -39,7 +39,7 @@ #include "inc/hw/dmcu.h" #include "dml/display_mode_lib.h" -#define DC_VER "3.2.06" +#define DC_VER "3.2.08" #define MAX_SURFACES 3 #define MAX_STREAMS 6 diff --git a/drivers/gpu/drm/amd/display/dc/dc_hw_types.h b/drivers/gpu/drm/amd/display/dc/dc_hw_types.h index 7825e4b5e97c..9ddfe4c6938b 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_hw_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_hw_types.h @@ -358,15 +358,16 @@ union dc_tiling_info { } gfx8; struct { + enum swizzle_mode_values swizzle; unsigned int num_pipes; - unsigned int num_banks; + unsigned int max_compressed_frags; unsigned int pipe_interleave; + + unsigned int num_banks; unsigned int num_shader_engines; unsigned int num_rb_per_se; - unsigned int max_compressed_frags; bool shaderEnable; - enum swizzle_mode_values swizzle; bool meta_linear; bool rb_aligned; bool pipe_aligned; diff --git a/drivers/gpu/drm/amd/display/dc/dc_link.h b/drivers/gpu/drm/amd/display/dc/dc_link.h index 8738f27a8708..29f19d57ff7a 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_link.h +++ b/drivers/gpu/drm/amd/display/dc/dc_link.h @@ -128,8 +128,10 @@ struct dc_link { const struct dc_link_status *dc_link_get_status(const struct dc_link *dc_link); -/* - * Return an enumerated dc_link. dc_link order is constant and determined at +/** + * dc_get_link_at_index() - Return an enumerated dc_link. + * + * dc_link order is constant and determined at * boot time. They cannot be created or destroyed. * Use dc_get_caps() to get number of links. */ diff --git a/drivers/gpu/drm/amd/display/dc/dc_stream.h b/drivers/gpu/drm/amd/display/dc/dc_stream.h index 771d9f17e26e..be34d638e15d 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_stream.h +++ b/drivers/gpu/drm/amd/display/dc/dc_stream.h @@ -56,6 +56,7 @@ struct dc_stream_state { struct dc_crtc_timing_adjust adjust; struct dc_info_packet vrr_infopacket; struct dc_info_packet vsc_infopacket; + struct dc_info_packet vsp_infopacket; struct rect src; /* composition area */ struct rect dst; /* stream addressable area */ @@ -129,11 +130,13 @@ struct dc_stream_update { struct dc_crtc_timing_adjust *adjust; struct dc_info_packet *vrr_infopacket; struct dc_info_packet *vsc_infopacket; + struct dc_info_packet *vsp_infopacket; bool *dpms_off; struct colorspace_transform *gamut_remap; enum dc_color_space *output_color_space; + enum dc_dither_option *dither_option; struct dc_csc_transform *output_csc_transform; diff --git a/drivers/gpu/drm/amd/display/dc/dc_types.h b/drivers/gpu/drm/amd/display/dc/dc_types.h index 6e12d640d020..0b20ae23f169 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_types.h @@ -73,10 +73,18 @@ struct hw_asic_id { void *atombios_base_address; }; +struct dc_perf_trace { + unsigned long read_count; + unsigned long write_count; + unsigned long last_entry_read; + unsigned long last_entry_write; +}; + struct dc_context { struct dc *dc; void *driver_context; /* e.g. amdgpu_device */ + struct dc_perf_trace *perf_trace; void *cgs_device; enum dce_environment dce_environment; @@ -191,7 +199,6 @@ union display_content_support { }; struct dc_panel_patch { - unsigned int disconnect_delay; unsigned int dppowerup_delay; unsigned int extra_t12_ms; }; diff --git a/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.c index bc50a8e25f4f..87771676acac 100644 --- a/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dce100/dce100_hw_sequencer.c @@ -117,6 +117,18 @@ void dce100_prepare_bandwidth( false); } +void dce100_optimize_bandwidth( + struct dc *dc, + struct dc_state *context) +{ + dce110_set_safe_displaymarks(&context->res_ctx, dc->res_pool); + + dc->res_pool->clk_mgr->funcs->update_clocks( + dc->res_pool->clk_mgr, + context, + true); +} + /**************************************************************************/ void dce100_hw_sequencer_construct(struct dc *dc) @@ -125,6 +137,6 @@ void dce100_hw_sequencer_construct(struct dc *dc) dc->hwss.enable_display_power_gating = dce100_enable_display_power_gating; dc->hwss.prepare_bandwidth = dce100_prepare_bandwidth; - dc->hwss.optimize_bandwidth = dce100_prepare_bandwidth; + dc->hwss.optimize_bandwidth = dce100_optimize_bandwidth; } diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_compressor.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_compressor.c index 1f7f25013217..52d50e24a995 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_compressor.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_compressor.c @@ -64,65 +64,37 @@ static const struct dce110_compressor_reg_offsets reg_offsets[] = { static const uint32_t dce11_one_lpt_channel_max_resolution = 2560 * 1600; -enum fbc_idle_force { - /* Bit 0 - Display registers updated */ - FBC_IDLE_FORCE_DISPLAY_REGISTER_UPDATE = 0x00000001, - - /* Bit 2 - FBC_GRPH_COMP_EN register updated */ - FBC_IDLE_FORCE_GRPH_COMP_EN = 0x00000002, - /* Bit 3 - FBC_SRC_SEL register updated */ - FBC_IDLE_FORCE_SRC_SEL_CHANGE = 0x00000004, - /* Bit 4 - FBC_MIN_COMPRESSION register updated */ - FBC_IDLE_FORCE_MIN_COMPRESSION_CHANGE = 0x00000008, - /* Bit 5 - FBC_ALPHA_COMP_EN register updated */ - FBC_IDLE_FORCE_ALPHA_COMP_EN = 0x00000010, - /* Bit 6 - FBC_ZERO_ALPHA_CHUNK_SKIP_EN register updated */ - FBC_IDLE_FORCE_ZERO_ALPHA_CHUNK_SKIP_EN = 0x00000020, - /* Bit 7 - FBC_FORCE_COPY_TO_COMP_BUF register updated */ - FBC_IDLE_FORCE_FORCE_COPY_TO_COMP_BUF = 0x00000040, - - /* Bit 24 - Memory write to region 0 defined by MC registers. */ - FBC_IDLE_FORCE_MEMORY_WRITE_TO_REGION0 = 0x01000000, - /* Bit 25 - Memory write to region 1 defined by MC registers */ - FBC_IDLE_FORCE_MEMORY_WRITE_TO_REGION1 = 0x02000000, - /* Bit 26 - Memory write to region 2 defined by MC registers */ - FBC_IDLE_FORCE_MEMORY_WRITE_TO_REGION2 = 0x04000000, - /* Bit 27 - Memory write to region 3 defined by MC registers. */ - FBC_IDLE_FORCE_MEMORY_WRITE_TO_REGION3 = 0x08000000, - - /* Bit 28 - Memory write from any client other than MCIF */ - FBC_IDLE_FORCE_MEMORY_WRITE_OTHER_THAN_MCIF = 0x10000000, - /* Bit 29 - CG statics screen signal is inactive */ - FBC_IDLE_FORCE_CG_STATIC_SCREEN_IS_INACTIVE = 0x20000000, -}; - - static uint32_t align_to_chunks_number_per_line(uint32_t pixels) { return 256 * ((pixels + 255) / 256); } -static void reset_lb_on_vblank(struct dc_context *ctx) +static void reset_lb_on_vblank(struct compressor *compressor, uint32_t crtc_inst) { - uint32_t value, frame_count; + uint32_t value; + uint32_t frame_count; + uint32_t status_pos; uint32_t retry = 0; - uint32_t status_pos = - dm_read_reg(ctx, mmCRTC_STATUS_POSITION); + struct dce110_compressor *cp110 = TO_DCE110_COMPRESSOR(compressor); + + cp110->offsets = reg_offsets[crtc_inst]; + + status_pos = dm_read_reg(compressor->ctx, DCP_REG(mmCRTC_STATUS_POSITION)); /* Only if CRTC is enabled and counter is moving we wait for one frame. */ - if (status_pos != dm_read_reg(ctx, mmCRTC_STATUS_POSITION)) { + if (status_pos != dm_read_reg(compressor->ctx, DCP_REG(mmCRTC_STATUS_POSITION))) { /* Resetting LB on VBlank */ - value = dm_read_reg(ctx, mmLB_SYNC_RESET_SEL); + value = dm_read_reg(compressor->ctx, DCP_REG(mmLB_SYNC_RESET_SEL)); set_reg_field_value(value, 3, LB_SYNC_RESET_SEL, LB_SYNC_RESET_SEL); set_reg_field_value(value, 1, LB_SYNC_RESET_SEL, LB_SYNC_RESET_SEL2); - dm_write_reg(ctx, mmLB_SYNC_RESET_SEL, value); + dm_write_reg(compressor->ctx, DCP_REG(mmLB_SYNC_RESET_SEL), value); - frame_count = dm_read_reg(ctx, mmCRTC_STATUS_FRAME_COUNT); + frame_count = dm_read_reg(compressor->ctx, DCP_REG(mmCRTC_STATUS_FRAME_COUNT)); for (retry = 10000; retry > 0; retry--) { - if (frame_count != dm_read_reg(ctx, mmCRTC_STATUS_FRAME_COUNT)) + if (frame_count != dm_read_reg(compressor->ctx, DCP_REG(mmCRTC_STATUS_FRAME_COUNT))) break; udelay(10); } @@ -130,13 +102,11 @@ static void reset_lb_on_vblank(struct dc_context *ctx) dm_error("Frame count did not increase for 100ms.\n"); /* Resetting LB on VBlank */ - value = dm_read_reg(ctx, mmLB_SYNC_RESET_SEL); + value = dm_read_reg(compressor->ctx, DCP_REG(mmLB_SYNC_RESET_SEL)); set_reg_field_value(value, 2, LB_SYNC_RESET_SEL, LB_SYNC_RESET_SEL); set_reg_field_value(value, 0, LB_SYNC_RESET_SEL, LB_SYNC_RESET_SEL2); - dm_write_reg(ctx, mmLB_SYNC_RESET_SEL, value); - + dm_write_reg(compressor->ctx, DCP_REG(mmLB_SYNC_RESET_SEL), value); } - } static void wait_for_fbc_state_changed( @@ -226,10 +196,10 @@ void dce110_compressor_enable_fbc( uint32_t addr; uint32_t value, misc_value; - addr = mmFBC_CNTL; value = dm_read_reg(compressor->ctx, addr); set_reg_field_value(value, 1, FBC_CNTL, FBC_GRPH_COMP_EN); + /* params->inst is valid HW CRTC instance start from 0 */ set_reg_field_value( value, params->inst, @@ -238,8 +208,10 @@ void dce110_compressor_enable_fbc( /* Keep track of enum controller_id FBC is attached to */ compressor->is_enabled = true; - compressor->attached_inst = params->inst; - cp110->offsets = reg_offsets[params->inst]; + /* attached_inst is SW CRTC instance start from 1 + * 0 = CONTROLLER_ID_UNDEFINED means not attached crtc + */ + compressor->attached_inst = params->inst + CONTROLLER_ID_D0; /* Toggle it as there is bug in HW */ set_reg_field_value(value, 0, FBC_CNTL, FBC_GRPH_COMP_EN); @@ -268,9 +240,10 @@ void dce110_compressor_enable_fbc( void dce110_compressor_disable_fbc(struct compressor *compressor) { struct dce110_compressor *cp110 = TO_DCE110_COMPRESSOR(compressor); + uint32_t crtc_inst = 0; if (compressor->options.bits.FBC_SUPPORT) { - if (dce110_compressor_is_fbc_enabled_in_hw(compressor, NULL)) { + if (dce110_compressor_is_fbc_enabled_in_hw(compressor, &crtc_inst)) { uint32_t reg_data; /* Turn off compression */ reg_data = dm_read_reg(compressor->ctx, mmFBC_CNTL); @@ -284,8 +257,10 @@ void dce110_compressor_disable_fbc(struct compressor *compressor) wait_for_fbc_state_changed(cp110, false); } - /* Sync line buffer - dce100/110 only*/ - reset_lb_on_vblank(compressor->ctx); + /* Sync line buffer which fbc was attached to dce100/110 only */ + if (crtc_inst > CONTROLLER_ID_UNDEFINED && crtc_inst < CONTROLLER_ID_D3) + reset_lb_on_vblank(compressor, + crtc_inst - CONTROLLER_ID_D0); } } @@ -328,6 +303,8 @@ void dce110_compressor_program_compressed_surface_address_and_pitch( uint32_t compressed_surf_address_low_part = compressor->compr_surface_address.addr.low_part; + cp110->offsets = reg_offsets[params->inst]; + /* Clear content first. */ dm_write_reg( compressor->ctx, @@ -410,13 +387,7 @@ void dce110_compressor_set_fbc_invalidation_triggers( value = dm_read_reg(compressor->ctx, addr); set_reg_field_value( value, - fbc_trigger | - FBC_IDLE_FORCE_GRPH_COMP_EN | - FBC_IDLE_FORCE_SRC_SEL_CHANGE | - FBC_IDLE_FORCE_MIN_COMPRESSION_CHANGE | - FBC_IDLE_FORCE_ALPHA_COMP_EN | - FBC_IDLE_FORCE_ZERO_ALPHA_CHUNK_SKIP_EN | - FBC_IDLE_FORCE_FORCE_COPY_TO_COMP_BUF, + fbc_trigger, FBC_IDLE_FORCE_CLEAR_MASK, FBC_IDLE_FORCE_CLEAR_MASK); dm_write_reg(compressor->ctx, addr, value); @@ -549,7 +520,7 @@ void dce110_compressor_construct(struct dce110_compressor *compressor, compressor->base.channel_interleave_size = 0; compressor->base.dram_channels_num = 0; compressor->base.lpt_channels_num = 0; - compressor->base.attached_inst = 0; + compressor->base.attached_inst = CONTROLLER_ID_UNDEFINED; compressor->base.is_enabled = false; compressor->base.funcs = &dce110_compressor_funcs; diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c index 2f062bacd78a..6349ba7bec7c 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c @@ -1766,12 +1766,13 @@ static void set_static_screen_control(struct pipe_ctx **pipe_ctx, * Check if FBC can be enabled */ static bool should_enable_fbc(struct dc *dc, - struct dc_state *context, - uint32_t *pipe_idx) + struct dc_state *context, + uint32_t *pipe_idx) { uint32_t i; struct pipe_ctx *pipe_ctx = NULL; struct resource_context *res_ctx = &context->res_ctx; + unsigned int underlay_idx = dc->res_pool->underlay_pipe_index; ASSERT(dc->fbc_compressor); @@ -1786,14 +1787,28 @@ static bool should_enable_fbc(struct dc *dc, for (i = 0; i < dc->res_pool->pipe_count; i++) { if (res_ctx->pipe_ctx[i].stream) { + pipe_ctx = &res_ctx->pipe_ctx[i]; - *pipe_idx = i; - break; + + if (!pipe_ctx) + continue; + + /* fbc not applicable on underlay pipe */ + if (pipe_ctx->pipe_idx != underlay_idx) { + *pipe_idx = i; + break; + } } } - /* Pipe context should be found */ - ASSERT(pipe_ctx); + if (i == dc->res_pool->pipe_count) + return false; + + if (!pipe_ctx->stream->sink) + return false; + + if (!pipe_ctx->stream->sink->link) + return false; /* Only supports eDP */ if (pipe_ctx->stream->sink->link->connector_signal != SIGNAL_TYPE_EDP) @@ -1817,8 +1832,9 @@ static bool should_enable_fbc(struct dc *dc, /* * Enable FBC */ -static void enable_fbc(struct dc *dc, - struct dc_state *context) +static void enable_fbc( + struct dc *dc, + struct dc_state *context) { uint32_t pipe_idx = 0; @@ -1828,10 +1844,9 @@ static void enable_fbc(struct dc *dc, struct compressor *compr = dc->fbc_compressor; struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[pipe_idx]; - params.source_view_width = pipe_ctx->stream->timing.h_addressable; params.source_view_height = pipe_ctx->stream->timing.v_addressable; - + params.inst = pipe_ctx->stream_res.tg->inst; compr->compr_surface_address.quad_part = dc->ctx->fbc_gpu_addr; compr->funcs->surface_address_and_pitch(compr, ¶ms); @@ -2046,10 +2061,10 @@ enum dc_status dce110_apply_ctx_to_hw( return status; } - dcb->funcs->set_scratch_critical_state(dcb, false); - if (dc->fbc_compressor) - enable_fbc(dc, context); + enable_fbc(dc, dc->current_state); + + dcb->funcs->set_scratch_critical_state(dcb, false); return DC_OK; } @@ -2408,7 +2423,6 @@ static void dce110_program_front_end_for_pipe( struct dc_plane_state *plane_state = pipe_ctx->plane_state; struct xfm_grph_csc_adjustment adjust; struct out_csc_color_matrix tbl_entry; - unsigned int underlay_idx = dc->res_pool->underlay_pipe_index; unsigned int i; DC_LOGGER_INIT(); memset(&tbl_entry, 0, sizeof(tbl_entry)); @@ -2449,15 +2463,6 @@ static void dce110_program_front_end_for_pipe( program_scaler(dc, pipe_ctx); - /* fbc not applicable on Underlay pipe */ - if (dc->fbc_compressor && old_pipe->stream && - pipe_ctx->pipe_idx != underlay_idx) { - if (plane_state->tiling_info.gfx8.array_mode == DC_ARRAY_LINEAR_GENERAL) - dc->fbc_compressor->funcs->disable_fbc(dc->fbc_compressor); - else - enable_fbc(dc, dc->current_state); - } - mi->funcs->mem_input_program_surface_config( mi, plane_state->format, @@ -2534,6 +2539,9 @@ static void dce110_apply_ctx_for_surface( if (num_planes == 0) return; + if (dc->fbc_compressor) + dc->fbc_compressor->funcs->disable_fbc(dc->fbc_compressor); + for (i = 0; i < dc->res_pool->pipe_count; i++) { struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; struct pipe_ctx *old_pipe_ctx = &dc->current_state->res_ctx.pipe_ctx[i]; @@ -2576,6 +2584,9 @@ static void dce110_apply_ctx_for_surface( (pipe_ctx->plane_state || old_pipe_ctx->plane_state)) dc->hwss.pipe_control_lock(dc, pipe_ctx, false); } + + if (dc->fbc_compressor) + enable_fbc(dc, dc->current_state); } static void dce110_power_down_fe(struct dc *dc, struct pipe_ctx *pipe_ctx) diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.c index f9d7d2c26cc2..54abedbf1b43 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.c @@ -328,12 +328,10 @@ static void dcn1_update_clocks(struct clk_mgr *clk_mgr, *smu_req_cur = smu_req; } - static const struct clk_mgr_funcs dcn1_funcs = { .get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz, .update_clocks = dcn1_update_clocks }; - struct clk_mgr *dcn1_clk_mgr_create(struct dc_context *ctx) { struct dc_debug_options *debug = &ctx->dc->debug; @@ -373,3 +371,5 @@ struct clk_mgr *dcn1_clk_mgr_create(struct dc_context *ctx) return &clk_mgr_dce->base; } + + diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.h index 9dbaf6578006..a995eda443a3 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_clk_mgr.h @@ -28,6 +28,12 @@ #include "../dce/dce_clk_mgr.h" +struct clk_bypass { + uint32_t dcfclk_bypass; + uint32_t dispclk_pypass; + uint32_t dprefclk_bypass; +}; + void dcn1_pplib_apply_display_requirements( struct dc *dc, struct dc_state *context); diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c index 3eea44092a04..7469333a2c8a 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c @@ -324,7 +324,7 @@ bool cm_helper_translate_curve_to_hw_format( if (output_tf == NULL || lut_params == NULL || output_tf->type == TF_TYPE_BYPASS) return false; - PERF_TRACE(); + PERF_TRACE_CTX(output_tf->ctx); corner_points = lut_params->corner_points; rgb_resulted = lut_params->rgb_resulted; @@ -513,7 +513,7 @@ bool cm_helper_translate_curve_to_degamma_hw_format( if (output_tf == NULL || lut_params == NULL || output_tf->type == TF_TYPE_BYPASS) return false; - PERF_TRACE(); + PERF_TRACE_CTX(output_tf->ctx); corner_points = lut_params->corner_points; rgb_resulted = lut_params->rgb_resulted; diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer_debug.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer_debug.c index 211bb240a720..cd469014baa3 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer_debug.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer_debug.c @@ -44,6 +44,7 @@ #include "dcn10_hubp.h" #include "dcn10_hubbub.h" #include "dcn10_cm_common.h" +#include "dcn10_clk_mgr.h" static unsigned int snprintf_count(char *pBuf, unsigned int bufSize, char *fmt, ...) { @@ -463,19 +464,22 @@ static unsigned int dcn10_get_otg_states(struct dc *dc, char *pBuf, unsigned int static unsigned int dcn10_get_clock_states(struct dc *dc, char *pBuf, unsigned int bufSize) { unsigned int chars_printed = 0; + unsigned int remaining_buffer = bufSize; - chars_printed = snprintf_count(pBuf, bufSize, "dcfclk_khz,dcfclk_deep_sleep_khz,dispclk_khz," - "dppclk_khz,max_supported_dppclk_khz,fclk_khz,socclk_khz\n" - "%d,%d,%d,%d,%d,%d,%d\n", + chars_printed = snprintf_count(pBuf, bufSize, "dcfclk,dcfclk_deep_sleep,dispclk," + "dppclk,fclk,socclk\n" + "%d,%d,%d,%d,%d,%d\n", dc->current_state->bw.dcn.clk.dcfclk_khz, dc->current_state->bw.dcn.clk.dcfclk_deep_sleep_khz, dc->current_state->bw.dcn.clk.dispclk_khz, dc->current_state->bw.dcn.clk.dppclk_khz, - dc->current_state->bw.dcn.clk.max_supported_dppclk_khz, dc->current_state->bw.dcn.clk.fclk_khz, dc->current_state->bw.dcn.clk.socclk_khz); - return chars_printed; + remaining_buffer -= chars_printed; + pBuf += chars_printed; + + return bufSize - remaining_buffer; } static void dcn10_clear_otpc_underflow(struct dc *dc) @@ -538,16 +542,16 @@ void dcn10_get_hw_state(struct dc *dc, char *pBuf, unsigned int bufSize, unsigne * Bit 0 - 15: Hardware block mask * Bit 15: 1 = Invariant Only, 0 = All */ - const unsigned int DC_HW_STATE_MASK_HUBBUB = 0x1; - const unsigned int DC_HW_STATE_MASK_HUBP = 0x2; - const unsigned int DC_HW_STATE_MASK_RQ = 0x4; - const unsigned int DC_HW_STATE_MASK_DLG = 0x8; - const unsigned int DC_HW_STATE_MASK_TTU = 0x10; - const unsigned int DC_HW_STATE_MASK_CM = 0x20; - const unsigned int DC_HW_STATE_MASK_MPCC = 0x40; - const unsigned int DC_HW_STATE_MASK_OTG = 0x80; - const unsigned int DC_HW_STATE_MASK_CLOCKS = 0x100; - const unsigned int DC_HW_STATE_INVAR_ONLY = 0x8000; + const unsigned int DC_HW_STATE_MASK_HUBBUB = 0x1; + const unsigned int DC_HW_STATE_MASK_HUBP = 0x2; + const unsigned int DC_HW_STATE_MASK_RQ = 0x4; + const unsigned int DC_HW_STATE_MASK_DLG = 0x8; + const unsigned int DC_HW_STATE_MASK_TTU = 0x10; + const unsigned int DC_HW_STATE_MASK_CM = 0x20; + const unsigned int DC_HW_STATE_MASK_MPCC = 0x40; + const unsigned int DC_HW_STATE_MASK_OTG = 0x80; + const unsigned int DC_HW_STATE_MASK_CLOCKS = 0x100; + const unsigned int DC_HW_STATE_INVAR_ONLY = 0x8000; unsigned int chars_printed = 0; unsigned int remaining_buf_size = bufSize; @@ -603,6 +607,9 @@ void dcn10_get_hw_state(struct dc *dc, char *pBuf, unsigned int bufSize, unsigne remaining_buf_size -= chars_printed; } - if ((mask & DC_HW_STATE_MASK_CLOCKS) && remaining_buf_size > 0) + if ((mask & DC_HW_STATE_MASK_CLOCKS) && remaining_buf_size > 0) { chars_printed = dcn10_get_clock_states(dc, pBuf, remaining_buf_size); + pBuf += chars_printed; + remaining_buf_size -= chars_printed; + } } diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c index 47dbe4bb294a..5d4772dec0ba 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c @@ -202,7 +202,6 @@ enum dcn10_clk_src_array_id { #define MMHUB_SR(reg_name)\ .reg_name = MMHUB_BASE(mm ## reg_name ## _BASE_IDX) + \ mm ## reg_name - /* macros to expend register list macro defined in HW object header file * end *********************/ @@ -436,7 +435,6 @@ static const struct dcn_optc_mask tg_mask = { TG_COMMON_MASK_SH_LIST_DCN1_0(_MASK) }; - static const struct bios_registers bios_regs = { NBIO_SR(BIOS_SCRATCH_0), NBIO_SR(BIOS_SCRATCH_3), @@ -497,7 +495,6 @@ static const struct dce110_clk_src_mask cs_mask = { CS_COMMON_MASK_SH_LIST_DCN1_0(_MASK) }; - static const struct resource_caps res_cap = { .num_timing_generator = 4, .num_opp = 4, @@ -1277,7 +1274,6 @@ static bool construct( goto fail; } } - pool->base.clk_mgr = dcn1_clk_mgr_create(ctx); if (pool->base.clk_mgr == NULL) { dm_error("DC: failed to create display clock!\n"); diff --git a/drivers/gpu/drm/amd/display/dc/dm_event_log.h b/drivers/gpu/drm/amd/display/dc/dm_event_log.h index 34a701ca879e..65663f4d93e1 100644 --- a/drivers/gpu/drm/amd/display/dc/dm_event_log.h +++ b/drivers/gpu/drm/amd/display/dc/dm_event_log.h @@ -33,6 +33,7 @@ #define EVENT_LOG_AUX_REQ(ddc, type, action, address, len, data) #define EVENT_LOG_AUX_REP(ddc, type, replyStatus, len, data) +#define EVENT_LOG_CUST_MSG(tag, a, ...) #endif diff --git a/drivers/gpu/drm/amd/display/dc/dm_pp_smu.h b/drivers/gpu/drm/amd/display/dc/dm_pp_smu.h index beb08fd12b1d..0029a39efb1c 100644 --- a/drivers/gpu/drm/amd/display/dc/dm_pp_smu.h +++ b/drivers/gpu/drm/amd/display/dc/dm_pp_smu.h @@ -102,7 +102,7 @@ struct pp_smu_funcs_rv { */ void (*set_display_count)(struct pp_smu *pp, int count); - /* which SMU message? are reader and writer WM separate SMU msg? */ + /* reader and writer WM's are sent together as part of one table*/ /* * PPSMC_MSG_SetDriverDramAddrHigh * PPSMC_MSG_SetDriverDramAddrLow diff --git a/drivers/gpu/drm/amd/display/dc/dm_services.h b/drivers/gpu/drm/amd/display/dc/dm_services.h index 28128c02de00..1961cc6d9143 100644 --- a/drivers/gpu/drm/amd/display/dc/dm_services.h +++ b/drivers/gpu/drm/amd/display/dc/dm_services.h @@ -31,6 +31,8 @@ #define __DM_SERVICES_H__ +#include "amdgpu_dm_trace.h" + /* TODO: remove when DC is complete. */ #include "dm_services_types.h" #include "logger_interface.h" @@ -70,6 +72,7 @@ static inline uint32_t dm_read_reg_func( } #endif value = cgs_read_register(ctx->cgs_device, address); + trace_amdgpu_dc_rreg(&ctx->perf_trace->read_count, address, value); return value; } @@ -90,6 +93,7 @@ static inline void dm_write_reg_func( } #endif cgs_write_register(ctx->cgs_device, address, value); + trace_amdgpu_dc_wreg(&ctx->perf_trace->write_count, address, value); } static inline uint32_t dm_read_index_reg( @@ -351,8 +355,12 @@ unsigned long long dm_get_elapse_time_in_ns(struct dc_context *ctx, /* * performance tracing */ -void dm_perf_trace_timestamp(const char *func_name, unsigned int line); -#define PERF_TRACE() dm_perf_trace_timestamp(__func__, __LINE__) +#define PERF_TRACE() trace_amdgpu_dc_performance(CTX->perf_trace->read_count,\ + CTX->perf_trace->write_count, &CTX->perf_trace->last_entry_read,\ + &CTX->perf_trace->last_entry_write, __func__, __LINE__) +#define PERF_TRACE_CTX(__CTX) trace_amdgpu_dc_performance(__CTX->perf_trace->read_count,\ + __CTX->perf_trace->write_count, &__CTX->perf_trace->last_entry_read,\ + &__CTX->perf_trace->last_entry_write, __func__, __LINE__) /* diff --git a/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.c b/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.c index f20161c5706d..dada04296025 100644 --- a/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.c +++ b/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.c @@ -56,7 +56,6 @@ struct gpio_service *dal_gpio_service_create( struct dc_context *ctx) { struct gpio_service *service; - uint32_t index_of_id; service = kzalloc(sizeof(struct gpio_service), GFP_KERNEL); @@ -78,44 +77,33 @@ struct gpio_service *dal_gpio_service_create( goto failure_1; } - /* allocate and initialize business storage */ + /* allocate and initialize busyness storage */ { - const uint32_t bits_per_uint = sizeof(uint32_t) << 3; - index_of_id = 0; service->ctx = ctx; do { uint32_t number_of_bits = service->factory.number_of_pins[index_of_id]; + uint32_t i = 0; - uint32_t number_of_uints = - (number_of_bits + bits_per_uint - 1) / - bits_per_uint; - - uint32_t *slot; - - if (number_of_bits) { - uint32_t index_of_uint = 0; + if (number_of_bits) { + service->busyness[index_of_id] = + kcalloc(number_of_bits, sizeof(char), + GFP_KERNEL); - slot = kcalloc(number_of_uints, - sizeof(uint32_t), - GFP_KERNEL); - - if (!slot) { + if (!service->busyness[index_of_id]) { BREAK_TO_DEBUGGER(); goto failure_2; } do { - slot[index_of_uint] = 0; - - ++index_of_uint; - } while (index_of_uint < number_of_uints); - } else - slot = NULL; - - service->busyness[index_of_id] = slot; + service->busyness[index_of_id][i] = 0; + ++i; + } while (i < number_of_bits); + } else { + service->busyness[index_of_id] = NULL; + } ++index_of_id; } while (index_of_id < GPIO_ID_COUNT); @@ -125,13 +113,8 @@ struct gpio_service *dal_gpio_service_create( failure_2: while (index_of_id) { - uint32_t *slot; - --index_of_id; - - slot = service->busyness[index_of_id]; - - kfree(slot); + kfree(service->busyness[index_of_id]); } failure_1: @@ -169,9 +152,7 @@ void dal_gpio_service_destroy( uint32_t index_of_id = 0; do { - uint32_t *slot = (*ptr)->busyness[index_of_id]; - - kfree(slot); + kfree((*ptr)->busyness[index_of_id]); ++index_of_id; } while (index_of_id < GPIO_ID_COUNT); @@ -192,11 +173,7 @@ static bool is_pin_busy( enum gpio_id id, uint32_t en) { - const uint32_t bits_per_uint = sizeof(uint32_t) << 3; - - const uint32_t *slot = service->busyness[id] + (en / bits_per_uint); - - return 0 != (*slot & (1 << (en % bits_per_uint))); + return service->busyness[id][en]; } static void set_pin_busy( @@ -204,10 +181,7 @@ static void set_pin_busy( enum gpio_id id, uint32_t en) { - const uint32_t bits_per_uint = sizeof(uint32_t) << 3; - - service->busyness[id][en / bits_per_uint] |= - (1 << (en % bits_per_uint)); + service->busyness[id][en] = true; } static void set_pin_free( @@ -215,10 +189,7 @@ static void set_pin_free( enum gpio_id id, uint32_t en) { - const uint32_t bits_per_uint = sizeof(uint32_t) << 3; - - service->busyness[id][en / bits_per_uint] &= - ~(1 << (en % bits_per_uint)); + service->busyness[id][en] = false; } enum gpio_result dal_gpio_service_open( diff --git a/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.h b/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.h index c7f3081f59cc..1d501a43d13b 100644 --- a/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.h +++ b/drivers/gpu/drm/amd/display/dc/gpio/gpio_service.h @@ -36,10 +36,9 @@ struct gpio_service { /* * @brief * Business storage. - * For each member of 'enum gpio_id', - * store array of bits (packed into uint32_t slots), - * index individual bit by 'en' value */ - uint32_t *busyness[GPIO_ID_COUNT]; + * one byte For each member of 'enum gpio_id' + */ + char *busyness[GPIO_ID_COUNT]; }; enum gpio_result dal_gpio_service_open( diff --git a/drivers/gpu/drm/amd/display/dc/inc/compressor.h b/drivers/gpu/drm/amd/display/dc/inc/compressor.h index bcb18f5e1e60..7a147a9762a0 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/compressor.h +++ b/drivers/gpu/drm/amd/display/dc/inc/compressor.h @@ -77,6 +77,7 @@ struct compressor_funcs { }; struct compressor { struct dc_context *ctx; + /* CONTROLLER_ID_D0 + instance, CONTROLLER_ID_UNDEFINED = 0 */ uint32_t attached_inst; bool is_enabled; const struct compressor_funcs *funcs; diff --git a/drivers/gpu/drm/amd/display/dc/inc/core_types.h b/drivers/gpu/drm/amd/display/dc/inc/core_types.h index e3ee96afa60e..b168a5e9dd9d 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/core_types.h +++ b/drivers/gpu/drm/amd/display/dc/inc/core_types.h @@ -272,6 +272,17 @@ union bw_context { struct dce_bw_output dce; }; +/** + * struct dc_state - The full description of a state requested by a user + * + * @streams: Stream properties + * @stream_status: The planes on a given stream + * @res_ctx: Persistent state of resources + * @bw: The output from bandwidth and watermark calculations + * @pp_display_cfg: PowerPlay clocks and settings + * @dcn_bw_vars: non-stack memory to support bandwidth calculations + * + */ struct dc_state { struct dc_stream_state *streams[MAX_PIPES]; struct dc_stream_status stream_status[MAX_PIPES]; @@ -279,7 +290,6 @@ struct dc_state { struct resource_context res_ctx; - /* The output from BW and WM calculations. */ union bw_context bw; /* Note: these are big structures, do *not* put on stack! */ diff --git a/drivers/gpu/drm/amd/display/modules/color/color_gamma.c b/drivers/gpu/drm/amd/display/modules/color/color_gamma.c index bbecbaefb741..479b77c2e89e 100644 --- a/drivers/gpu/drm/amd/display/modules/color/color_gamma.c +++ b/drivers/gpu/drm/amd/display/modules/color/color_gamma.c @@ -1761,7 +1761,7 @@ bool mod_color_calculate_degamma_params(struct dc_transfer_func *input_tf, struct pwl_float_data *rgb_user = NULL; struct pwl_float_data_ex *curve = NULL; - struct gamma_pixel *axix_x = NULL; + struct gamma_pixel *axis_x = NULL; struct pixel_gamma_point *coeff = NULL; enum dc_transfer_func_predefined tf = TRANSFER_FUNCTION_SRGB; bool ret = false; @@ -1787,10 +1787,10 @@ bool mod_color_calculate_degamma_params(struct dc_transfer_func *input_tf, GFP_KERNEL); if (!curve) goto curve_alloc_fail; - axix_x = kvcalloc(ramp->num_entries + _EXTRA_POINTS, sizeof(*axix_x), + axis_x = kvcalloc(ramp->num_entries + _EXTRA_POINTS, sizeof(*axis_x), GFP_KERNEL); - if (!axix_x) - goto axix_x_alloc_fail; + if (!axis_x) + goto axis_x_alloc_fail; coeff = kvcalloc(MAX_HW_POINTS + _EXTRA_POINTS, sizeof(*coeff), GFP_KERNEL); if (!coeff) @@ -1803,7 +1803,7 @@ bool mod_color_calculate_degamma_params(struct dc_transfer_func *input_tf, tf = input_tf->tf; build_evenly_distributed_points( - axix_x, + axis_x, ramp->num_entries, dividers); @@ -1828,7 +1828,7 @@ bool mod_color_calculate_degamma_params(struct dc_transfer_func *input_tf, tf_pts->x_point_at_y1_blue = 1; map_regamma_hw_to_x_user(ramp, coeff, rgb_user, - coordinates_x, axix_x, curve, + coordinates_x, axis_x, curve, MAX_HW_POINTS, tf_pts, mapUserRamp && ramp->type != GAMMA_CUSTOM); if (ramp->type == GAMMA_CUSTOM) @@ -1838,8 +1838,8 @@ bool mod_color_calculate_degamma_params(struct dc_transfer_func *input_tf, kvfree(coeff); coeff_alloc_fail: - kvfree(axix_x); -axix_x_alloc_fail: + kvfree(axis_x); +axis_x_alloc_fail: kvfree(curve); curve_alloc_fail: kvfree(rgb_user); diff --git a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c index 620a171620ee..1544ed3f1747 100644 --- a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c +++ b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c @@ -608,12 +608,12 @@ static void build_vrr_infopacket_data(const struct mod_vrr_params *vrr, static void build_vrr_infopacket_fs2_data(enum color_transfer_func app_tf, struct dc_info_packet *infopacket) { - if (app_tf != transfer_func_unknown) { + if (app_tf != TRANSFER_FUNC_UNKNOWN) { infopacket->valid = true; infopacket->sb[6] |= 0x08; // PB6 = [Bit 3 = Native Color Active] - if (app_tf == transfer_func_gamma_22) { + if (app_tf == TRANSFER_FUNC_GAMMA_22) { infopacket->sb[9] |= 0x04; // PB6 = [Bit 2 = Gamma 2.2 EOTF Active] } } @@ -688,11 +688,11 @@ void mod_freesync_build_vrr_infopacket(struct mod_freesync *mod_freesync, return; switch (packet_type) { - case packet_type_fs2: + case PACKET_TYPE_FS2: build_vrr_infopacket_v2(stream->signal, vrr, app_tf, infopacket); break; - case packet_type_vrr: - case packet_type_fs1: + case PACKET_TYPE_VRR: + case PACKET_TYPE_FS1: default: build_vrr_infopacket_v1(stream->signal, vrr, infopacket); } diff --git a/drivers/gpu/drm/amd/display/modules/inc/mod_info_packet.h b/drivers/gpu/drm/amd/display/modules/inc/mod_info_packet.h index 786b34380f85..5b1c9a4c7643 100644 --- a/drivers/gpu/drm/amd/display/modules/inc/mod_info_packet.h +++ b/drivers/gpu/drm/amd/display/modules/inc/mod_info_packet.h @@ -26,15 +26,13 @@ #ifndef MOD_INFO_PACKET_H_ #define MOD_INFO_PACKET_H_ -struct info_packet_inputs { - const struct dc_stream_state *pStream; -}; +#include "mod_shared.h" -struct info_packets { - struct dc_info_packet *pVscInfoPacket; -}; +//Forward Declarations +struct dc_stream_state; +struct dc_info_packet; -void mod_build_infopackets(struct info_packet_inputs *inputs, - struct info_packets *info_packets); +void mod_build_vsc_infopacket(const struct dc_stream_state *stream, + struct dc_info_packet *info_packet); #endif diff --git a/drivers/gpu/drm/amd/display/modules/inc/mod_shared.h b/drivers/gpu/drm/amd/display/modules/inc/mod_shared.h index 238c431ae483..1bd02c0ac30c 100644 --- a/drivers/gpu/drm/amd/display/modules/inc/mod_shared.h +++ b/drivers/gpu/drm/amd/display/modules/inc/mod_shared.h @@ -23,27 +23,26 @@ * */ - #ifndef MOD_SHARED_H_ #define MOD_SHARED_H_ enum color_transfer_func { - transfer_func_unknown, - transfer_func_srgb, - transfer_func_bt709, - transfer_func_pq2084, - transfer_func_pq2084_interim, - transfer_func_linear_0_1, - transfer_func_linear_0_125, - transfer_func_dolbyvision, - transfer_func_gamma_22, - transfer_func_gamma_26 + TRANSFER_FUNC_UNKNOWN, + TRANSFER_FUNC_SRGB, + TRANSFER_FUNC_BT709, + TRANSFER_FUNC_PQ2084, + TRANSFER_FUNC_PQ2084_INTERIM, + TRANSFER_FUNC_LINEAR_0_1, + TRANSFER_FUNC_LINEAR_0_125, + TRANSFER_FUNC_GAMMA_22, + TRANSFER_FUNC_GAMMA_26 }; enum vrr_packet_type { - packet_type_vrr, - packet_type_fs1, - packet_type_fs2 + PACKET_TYPE_VRR, + PACKET_TYPE_FS1, + PACKET_TYPE_FS2 }; + #endif /* MOD_SHARED_H_ */ diff --git a/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c b/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c index ff8bfb9b43b0..db06fab2ad5c 100644 --- a/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c +++ b/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c @@ -25,6 +25,10 @@ #include "mod_info_packet.h" #include "core_types.h" +#include "dc_types.h" +#include "mod_shared.h" + +#define HDMI_INFOFRAME_TYPE_VENDOR 0x81 enum ColorimetryRGBDP { ColorimetryRGB_DP_sRGB = 0, @@ -41,7 +45,7 @@ enum ColorimetryYCCDP { ColorimetryYCC_DP_ITU2020YCbCr = 7, }; -static void mod_build_vsc_infopacket(const struct dc_stream_state *stream, +void mod_build_vsc_infopacket(const struct dc_stream_state *stream, struct dc_info_packet *info_packet) { unsigned int vscPacketRevision = 0; @@ -159,7 +163,7 @@ static void mod_build_vsc_infopacket(const struct dc_stream_state *stream, * DPCD register is exposed in the new Extended Receiver Capability field for DPCD Rev. 1.4 * (and higher). When MISC1. bit 6. is Set to 1, a Source device uses a VSC SDP to indicate * the Pixel Encoding/Colorimetry Format and that a Sink device must ignore MISC1, bit 7, and - * MISC0, bits 7:1 (MISC1, bit 7. and MISC0, bits 7:1 become “don’t care”).) + * MISC0, bits 7:1 (MISC1, bit 7. and MISC0, bits 7:1 become "don't care").) */ if (vscPacketRevision == 0x5) { /* Secondary-data Packet ID = 0 */ @@ -320,10 +324,3 @@ static void mod_build_vsc_infopacket(const struct dc_stream_state *stream, } -void mod_build_infopackets(struct info_packet_inputs *inputs, - struct info_packets *info_packets) -{ - if (info_packets->pVscInfoPacket != NULL) - mod_build_vsc_infopacket(inputs->pStream, info_packets->pVscInfoPacket); -} - diff --git a/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_offset.h new file mode 100644 index 000000000000..8f515875a34d --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_offset.h @@ -0,0 +1,32 @@ +/* + * Copyright (C) 2018 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) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ +#ifndef _mmhub_9_4_0_OFFSET_HEADER +#define _mmhub_9_4_0_OFFSET_HEADER + + +// addressBlock: mmhub_utcl2_vmsharedpfdec +// base address: 0x6a040 +#define mmMC_VM_XGMI_LFB_CNTL 0x0823 +#define mmMC_VM_XGMI_LFB_CNTL_BASE_IDX 0 +#define mmMC_VM_XGMI_LFB_SIZE 0x0824 +#define mmMC_VM_XGMI_LFB_SIZE_BASE_IDX 0 + +#endif diff --git a/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_sh_mask.h new file mode 100644 index 000000000000..0a6b072d191e --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_9_4_0_sh_mask.h @@ -0,0 +1,35 @@ +/* + * Copyright (C) 2018 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) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ +#ifndef _mmhub_9_4_0_SH_MASK_HEADER +#define _mmhub_9_4_0_SH_MASK_HEADER + + +// addressBlock: mmhub_utcl2_vmsharedpfdec +//MC_VM_XGMI_LFB_CNTL +#define MC_VM_XGMI_LFB_CNTL__PF_LFB_REGION__SHIFT 0x0 +#define MC_VM_XGMI_LFB_CNTL__PF_MAX_REGION__SHIFT 0x4 +#define MC_VM_XGMI_LFB_CNTL__PF_LFB_REGION_MASK 0x00000007L +#define MC_VM_XGMI_LFB_CNTL__PF_MAX_REGION_MASK 0x00000070L +//MC_VM_XGMI_LFB_SIZE +#define MC_VM_XGMI_LFB_SIZE__PF_LFB_SIZE__SHIFT 0x0 +#define MC_VM_XGMI_LFB_SIZE__PF_LFB_SIZE_MASK 0x0000FFFFL + +#endif diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h index 58ac0b90c310..8154d67388cc 100644 --- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h @@ -188,8 +188,8 @@ struct tile_config { */ #define ALLOC_MEM_FLAGS_VRAM (1 << 0) #define ALLOC_MEM_FLAGS_GTT (1 << 1) -#define ALLOC_MEM_FLAGS_USERPTR (1 << 2) /* TODO */ -#define ALLOC_MEM_FLAGS_DOORBELL (1 << 3) /* TODO */ +#define ALLOC_MEM_FLAGS_USERPTR (1 << 2) +#define ALLOC_MEM_FLAGS_DOORBELL (1 << 3) /* * Allocation flags attributes/access options. diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index 980e696989b1..1479ea1dc3e7 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -276,6 +276,10 @@ struct amd_pm_funcs { struct amd_pp_simple_clock_info *clocks); int (*notify_smu_enable_pwe)(void *handle); int (*enable_mgpu_fan_boost)(void *handle); + int (*set_active_display_count)(void *handle, uint32_t count); + int (*set_hard_min_dcefclk_by_freq)(void *handle, uint32_t clock); + int (*set_hard_min_fclk_by_freq)(void *handle, uint32_t clock); + int (*set_min_deep_sleep_dcefclk)(void *handle, uint32_t clock); }; #endif diff --git a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c index b68c2e0fef01..9bc27f468d5b 100644 --- a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c +++ b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c @@ -725,7 +725,7 @@ static int pp_dpm_force_clock_level(void *handle, } if (hwmgr->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL) { - pr_info("force clock level is for dpm manual mode only.\n"); + pr_debug("force clock level is for dpm manual mode only.\n"); return -EINVAL; } @@ -899,7 +899,7 @@ static int pp_set_power_profile_mode(void *handle, long *input, uint32_t size) } if (hwmgr->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL) { - pr_info("power profile setting is for manual dpm mode only.\n"); + pr_debug("power profile setting is for manual dpm mode only.\n"); return ret; } @@ -1072,7 +1072,7 @@ static int pp_get_current_clocks(void *handle, &hw_clocks, PHM_PerformanceLevelDesignation_Activity); if (ret) { - pr_info("Error in phm_get_clock_info \n"); + pr_debug("Error in phm_get_clock_info \n"); mutex_unlock(&hwmgr->smu_lock); return -EINVAL; } @@ -1332,6 +1332,78 @@ static int pp_enable_mgpu_fan_boost(void *handle) return 0; } +static int pp_set_min_deep_sleep_dcefclk(void *handle, uint32_t clock) +{ + struct pp_hwmgr *hwmgr = handle; + + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; + + if (hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk == NULL) { + pr_debug("%s was not implemented.\n", __func__); + return -EINVAL;; + } + + mutex_lock(&hwmgr->smu_lock); + hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk(hwmgr, clock); + mutex_unlock(&hwmgr->smu_lock); + + return 0; +} + +static int pp_set_hard_min_dcefclk_by_freq(void *handle, uint32_t clock) +{ + struct pp_hwmgr *hwmgr = handle; + + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; + + if (hwmgr->hwmgr_func->set_hard_min_dcefclk_by_freq == NULL) { + pr_debug("%s was not implemented.\n", __func__); + return -EINVAL;; + } + + mutex_lock(&hwmgr->smu_lock); + hwmgr->hwmgr_func->set_hard_min_dcefclk_by_freq(hwmgr, clock); + mutex_unlock(&hwmgr->smu_lock); + + return 0; +} + +static int pp_set_hard_min_fclk_by_freq(void *handle, uint32_t clock) +{ + struct pp_hwmgr *hwmgr = handle; + + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; + + if (hwmgr->hwmgr_func->set_hard_min_fclk_by_freq == NULL) { + pr_debug("%s was not implemented.\n", __func__); + return -EINVAL;; + } + + mutex_lock(&hwmgr->smu_lock); + hwmgr->hwmgr_func->set_hard_min_fclk_by_freq(hwmgr, clock); + mutex_unlock(&hwmgr->smu_lock); + + return 0; +} + +static int pp_set_active_display_count(void *handle, uint32_t count) +{ + struct pp_hwmgr *hwmgr = handle; + int ret = 0; + + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; + + mutex_lock(&hwmgr->smu_lock); + ret = phm_set_active_display_count(hwmgr, count); + mutex_unlock(&hwmgr->smu_lock); + + return ret; +} + static const struct amd_pm_funcs pp_dpm_funcs = { .load_firmware = pp_dpm_load_fw, .wait_for_fw_loading_complete = pp_dpm_fw_loading_complete, @@ -1378,4 +1450,8 @@ static const struct amd_pm_funcs pp_dpm_funcs = { .get_display_mode_validation_clocks = pp_get_display_mode_validation_clocks, .notify_smu_enable_pwe = pp_notify_smu_enable_pwe, .enable_mgpu_fan_boost = pp_enable_mgpu_fan_boost, + .set_active_display_count = pp_set_active_display_count, + .set_min_deep_sleep_dcefclk = pp_set_min_deep_sleep_dcefclk, + .set_hard_min_dcefclk_by_freq = pp_set_hard_min_dcefclk_by_freq, + .set_hard_min_fclk_by_freq = pp_set_hard_min_fclk_by_freq, }; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c b/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c index 85119c2bdcc8..333b9b845971 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c @@ -286,8 +286,8 @@ int phm_store_dal_configuration_data(struct pp_hwmgr *hwmgr, if (display_config == NULL) return -EINVAL; - if (NULL != hwmgr->hwmgr_func->set_deep_sleep_dcefclk) - hwmgr->hwmgr_func->set_deep_sleep_dcefclk(hwmgr, display_config->min_dcef_deep_sleep_set_clk); + if (NULL != hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk) + hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk(hwmgr, display_config->min_dcef_deep_sleep_set_clk); for (index = 0; index < display_config->num_path_including_non_display; index++) { if (display_config->displays[index].controller_id != 0) @@ -478,3 +478,44 @@ int phm_disable_smc_firmware_ctf(struct pp_hwmgr *hwmgr) return hwmgr->hwmgr_func->disable_smc_firmware_ctf(hwmgr); } + +int phm_set_active_display_count(struct pp_hwmgr *hwmgr, uint32_t count) +{ + PHM_FUNC_CHECK(hwmgr); + + if (!hwmgr->hwmgr_func->set_active_display_count) + return -EINVAL; + + return hwmgr->hwmgr_func->set_active_display_count(hwmgr, count); +} + +int phm_set_min_deep_sleep_dcefclk(struct pp_hwmgr *hwmgr, uint32_t clock) +{ + PHM_FUNC_CHECK(hwmgr); + + if (!hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk) + return -EINVAL; + + return hwmgr->hwmgr_func->set_min_deep_sleep_dcefclk(hwmgr, clock); +} + +int phm_set_hard_min_dcefclk_by_freq(struct pp_hwmgr *hwmgr, uint32_t clock) +{ + PHM_FUNC_CHECK(hwmgr); + + if (!hwmgr->hwmgr_func->set_hard_min_dcefclk_by_freq) + return -EINVAL; + + return hwmgr->hwmgr_func->set_hard_min_dcefclk_by_freq(hwmgr, clock); +} + +int phm_set_hard_min_fclk_by_freq(struct pp_hwmgr *hwmgr, uint32_t clock) +{ + PHM_FUNC_CHECK(hwmgr); + + if (!hwmgr->hwmgr_func->set_hard_min_fclk_by_freq) + return -EINVAL; + + return hwmgr->hwmgr_func->set_hard_min_fclk_by_freq(hwmgr, clock); +} + diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c index dd18cb710391..f95c5f50eb0f 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c @@ -216,12 +216,12 @@ static inline uint32_t convert_10k_to_mhz(uint32_t clock) return (clock + 99) / 100; } -static int smu10_set_deep_sleep_dcefclk(struct pp_hwmgr *hwmgr, uint32_t clock) +static int smu10_set_min_deep_sleep_dcefclk(struct pp_hwmgr *hwmgr, uint32_t clock) { struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); if (smu10_data->need_min_deep_sleep_dcefclk && - smu10_data->deep_sleep_dcefclk != convert_10k_to_mhz(clock)) { + smu10_data->deep_sleep_dcefclk != convert_10k_to_mhz(clock)) { smu10_data->deep_sleep_dcefclk = convert_10k_to_mhz(clock); smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetMinDeepSleepDcefclk, @@ -230,6 +230,34 @@ static int smu10_set_deep_sleep_dcefclk(struct pp_hwmgr *hwmgr, uint32_t clock) return 0; } +static int smu10_set_hard_min_dcefclk_by_freq(struct pp_hwmgr *hwmgr, uint32_t clock) +{ + struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); + + if (smu10_data->dcf_actual_hard_min_freq && + smu10_data->dcf_actual_hard_min_freq != convert_10k_to_mhz(clock)) { + smu10_data->dcf_actual_hard_min_freq = convert_10k_to_mhz(clock); + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetHardMinDcefclkByFreq, + smu10_data->dcf_actual_hard_min_freq); + } + return 0; +} + +static int smu10_set_hard_min_fclk_by_freq(struct pp_hwmgr *hwmgr, uint32_t clock) +{ + struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); + + if (smu10_data->f_actual_hard_min_freq && + smu10_data->f_actual_hard_min_freq != convert_10k_to_mhz(clock)) { + smu10_data->f_actual_hard_min_freq = convert_10k_to_mhz(clock); + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetHardMinFclkByFreq, + smu10_data->f_actual_hard_min_freq); + } + return 0; +} + static int smu10_set_active_display_count(struct pp_hwmgr *hwmgr, uint32_t count) { struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); @@ -1206,7 +1234,7 @@ static const struct pp_hwmgr_func smu10_hwmgr_funcs = { .get_max_high_clocks = smu10_get_max_high_clocks, .read_sensor = smu10_read_sensor, .set_active_display_count = smu10_set_active_display_count, - .set_deep_sleep_dcefclk = smu10_set_deep_sleep_dcefclk, + .set_min_deep_sleep_dcefclk = smu10_set_min_deep_sleep_dcefclk, .dynamic_state_management_enable = smu10_enable_dpm_tasks, .power_off_asic = smu10_power_off_asic, .asic_setup = smu10_setup_asic_task, @@ -1217,6 +1245,8 @@ static const struct pp_hwmgr_func smu10_hwmgr_funcs = { .display_clock_voltage_request = smu10_display_clock_voltage_request, .powergate_gfx = smu10_gfx_off_control, .powergate_sdma = smu10_powergate_sdma, + .set_hard_min_dcefclk_by_freq = smu10_set_hard_min_dcefclk_by_freq, + .set_hard_min_fclk_by_freq = smu10_set_hard_min_fclk_by_freq, }; int smu10_init_function_pointers(struct pp_hwmgr *hwmgr) diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c index 5dcd21d29dbf..3958729d6265 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c @@ -2859,7 +2859,10 @@ static int smu7_vblank_too_short(struct pp_hwmgr *hwmgr, case CHIP_POLARIS10: case CHIP_POLARIS11: case CHIP_POLARIS12: - switch_limit_us = data->is_memory_gddr5 ? 190 : 150; + if (hwmgr->is_kicker) + switch_limit_us = data->is_memory_gddr5 ? 450 : 150; + else + switch_limit_us = data->is_memory_gddr5 ? 190 : 150; break; case CHIP_VEGAM: switch_limit_us = 30; @@ -4219,9 +4222,17 @@ static int smu7_check_mc_firmware(struct pp_hwmgr *hwmgr) if (tmp & (1 << 23)) { data->mem_latency_high = MEM_LATENCY_HIGH; data->mem_latency_low = MEM_LATENCY_LOW; + if ((hwmgr->chip_id == CHIP_POLARIS10) || + (hwmgr->chip_id == CHIP_POLARIS11) || + (hwmgr->chip_id == CHIP_POLARIS12)) + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_EnableFFC); } else { data->mem_latency_high = 330; data->mem_latency_low = 330; + if ((hwmgr->chip_id == CHIP_POLARIS10) || + (hwmgr->chip_id == CHIP_POLARIS11) || + (hwmgr->chip_id == CHIP_POLARIS12)) + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_DisableFFC); } return 0; diff --git a/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h b/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h index 54fd0125d9cf..f4dab979a3a1 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h +++ b/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h @@ -463,5 +463,8 @@ extern int phm_display_clock_voltage_request(struct pp_hwmgr *hwmgr, extern int phm_get_max_high_clocks(struct pp_hwmgr *hwmgr, struct amd_pp_simple_clock_info *clocks); extern int phm_disable_smc_firmware_ctf(struct pp_hwmgr *hwmgr); + +extern int phm_set_active_display_count(struct pp_hwmgr *hwmgr, uint32_t count); + #endif /* _HARDWARE_MANAGER_H_ */ diff --git a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h index fb0f96f7cdbc..0d298a0409f5 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h @@ -309,7 +309,7 @@ struct pp_hwmgr_func { int (*avfs_control)(struct pp_hwmgr *hwmgr, bool enable); int (*disable_smc_firmware_ctf)(struct pp_hwmgr *hwmgr); int (*set_active_display_count)(struct pp_hwmgr *hwmgr, uint32_t count); - int (*set_deep_sleep_dcefclk)(struct pp_hwmgr *hwmgr, uint32_t clock); + int (*set_min_deep_sleep_dcefclk)(struct pp_hwmgr *hwmgr, uint32_t clock); int (*start_thermal_controller)(struct pp_hwmgr *hwmgr, struct PP_TemperatureRange *range); int (*notify_cac_buffer_info)(struct pp_hwmgr *hwmgr, uint32_t virtual_addr_low, @@ -332,6 +332,8 @@ struct pp_hwmgr_func { int (*smus_notify_pwe)(struct pp_hwmgr *hwmgr); int (*powergate_sdma)(struct pp_hwmgr *hwmgr, bool bgate); int (*enable_mgpu_fan_boost)(struct pp_hwmgr *hwmgr); + int (*set_hard_min_dcefclk_by_freq)(struct pp_hwmgr *hwmgr, uint32_t clock); + int (*set_hard_min_fclk_by_freq)(struct pp_hwmgr *hwmgr, uint32_t clock); }; struct pp_table_func { diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu7_ppsmc.h b/drivers/gpu/drm/amd/powerplay/inc/smu7_ppsmc.h index 62f36ba2435b..d11d6a797ce4 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/smu7_ppsmc.h +++ b/drivers/gpu/drm/amd/powerplay/inc/smu7_ppsmc.h @@ -395,6 +395,9 @@ typedef uint16_t PPSMC_Result; #define PPSMC_MSG_SetVBITimeout ((uint16_t) 0x306) +#define PPSMC_MSG_EnableFFC ((uint16_t) 0x307) +#define PPSMC_MSG_DisableFFC ((uint16_t) 0x308) + #define PPSMC_MSG_EnableDpmDidt ((uint16_t) 0x309) #define PPSMC_MSG_DisableDpmDidt ((uint16_t) 0x30A) diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c index 2b2c26616902..b3e06e498834 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c @@ -1528,8 +1528,21 @@ static int polaris10_populate_clock_stretcher_data_table(struct pp_hwmgr *hwmgr) efuse = efuse >> 24; if (hwmgr->chip_id == CHIP_POLARIS10) { - min = 1000; - max = 2300; + if (hwmgr->is_kicker) { + min = 1200; + max = 2500; + } else { + min = 1000; + max = 2300; + } + } else if (hwmgr->chip_id == CHIP_POLARIS11) { + if (hwmgr->is_kicker) { + min = 900; + max = 2100; + } else { + min = 1100; + max = 2100; + } } else { min = 1100; max = 2100; @@ -1626,6 +1639,7 @@ static int polaris10_populate_avfs_parameters(struct pp_hwmgr *hwmgr) { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); struct polaris10_smumgr *smu_data = (struct polaris10_smumgr *)(hwmgr->smu_backend); + struct amdgpu_device *adev = hwmgr->adev; SMU74_Discrete_DpmTable *table = &(smu_data->smc_state_table); int result = 0; @@ -1646,6 +1660,59 @@ static int polaris10_populate_avfs_parameters(struct pp_hwmgr *hwmgr) result = atomctrl_get_avfs_information(hwmgr, &avfs_params); if (0 == result) { + if (((adev->pdev->device == 0x67ef) && + ((adev->pdev->revision == 0xe0) || + (adev->pdev->revision == 0xe5))) || + ((adev->pdev->device == 0x67ff) && + ((adev->pdev->revision == 0xcf) || + (adev->pdev->revision == 0xef) || + (adev->pdev->revision == 0xff)))) { + avfs_params.ucEnableApplyAVFS_CKS_OFF_Voltage = 1; + if ((adev->pdev->device == 0x67ef && adev->pdev->revision == 0xe5) || + (adev->pdev->device == 0x67ff && adev->pdev->revision == 0xef)) { + if ((avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a0 == 0xEA522DD3) && + (avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a1 == 0x5645A) && + (avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a2 == 0x33F9E) && + (avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_m1 == 0xFFFFC5CC) && + (avfs_params.usAVFSGB_FUSE_TABLE_CKSOFF_m2 == 0x1B1A) && + (avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_b == 0xFFFFFCED)) { + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a0 = 0xF718F1D4; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a1 = 0x323FD; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a2 = 0x1E455; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_m1 = 0; + avfs_params.usAVFSGB_FUSE_TABLE_CKSOFF_m2 = 0; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_b = 0x23; + } + } + } else if (hwmgr->chip_id == CHIP_POLARIS12 && !hwmgr->is_kicker) { + avfs_params.ucEnableApplyAVFS_CKS_OFF_Voltage = 1; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a0 = 0xF6B024DD; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a1 = 0x3005E; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a2 = 0x18A5F; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_m1 = 0x315; + avfs_params.usAVFSGB_FUSE_TABLE_CKSOFF_m2 = 0xFED1; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_b = 0x3B; + } else if (((adev->pdev->device == 0x67df) && + ((adev->pdev->revision == 0xe0) || + (adev->pdev->revision == 0xe3) || + (adev->pdev->revision == 0xe4) || + (adev->pdev->revision == 0xe5) || + (adev->pdev->revision == 0xe7) || + (adev->pdev->revision == 0xef))) || + ((adev->pdev->device == 0x6fdf) && + ((adev->pdev->revision == 0xef) || + (adev->pdev->revision == 0xff)))) { + avfs_params.ucEnableApplyAVFS_CKS_OFF_Voltage = 1; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a0 = 0xF843B66B; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a1 = 0x59CB5; + avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a2 = 0xFFFF287F; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_m1 = 0; + avfs_params.usAVFSGB_FUSE_TABLE_CKSOFF_m2 = 0xFF23; + avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_b = 0x58; + } + } + + if (0 == result) { table->BTCGB_VDROOP_TABLE[0].a0 = PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSON_a0); table->BTCGB_VDROOP_TABLE[0].a1 = PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSON_a1); table->BTCGB_VDROOP_TABLE[0].a2 = PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSON_a2); diff --git a/drivers/gpu/drm/etnaviv/etnaviv_dump.c b/drivers/gpu/drm/etnaviv/etnaviv_dump.c index 9146e30e24a6..3fbb4855396c 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_dump.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_dump.c @@ -118,6 +118,7 @@ void etnaviv_core_dump(struct etnaviv_gpu *gpu) unsigned int n_obj, n_bomap_pages; size_t file_size, mmu_size; __le64 *bomap, *bomap_start; + unsigned long flags; /* Only catch the first event, or when manually re-armed */ if (!etnaviv_dump_core) @@ -134,13 +135,13 @@ void etnaviv_core_dump(struct etnaviv_gpu *gpu) mmu_size + gpu->buffer.size; /* Add in the active command buffers */ - spin_lock(&gpu->sched.job_list_lock); + spin_lock_irqsave(&gpu->sched.job_list_lock, flags); list_for_each_entry(s_job, &gpu->sched.ring_mirror_list, node) { submit = to_etnaviv_submit(s_job); file_size += submit->cmdbuf.size; n_obj++; } - spin_unlock(&gpu->sched.job_list_lock); + spin_unlock_irqrestore(&gpu->sched.job_list_lock, flags); /* Add in the active buffer objects */ list_for_each_entry(vram, &gpu->mmu->mappings, mmu_node) { @@ -182,14 +183,14 @@ void etnaviv_core_dump(struct etnaviv_gpu *gpu) gpu->buffer.size, etnaviv_cmdbuf_get_va(&gpu->buffer)); - spin_lock(&gpu->sched.job_list_lock); + spin_lock_irqsave(&gpu->sched.job_list_lock, flags); list_for_each_entry(s_job, &gpu->sched.ring_mirror_list, node) { submit = to_etnaviv_submit(s_job); etnaviv_core_dump_mem(&iter, ETDUMP_BUF_CMD, submit->cmdbuf.vaddr, submit->cmdbuf.size, etnaviv_cmdbuf_get_va(&submit->cmdbuf)); } - spin_unlock(&gpu->sched.job_list_lock); + spin_unlock_irqrestore(&gpu->sched.job_list_lock, flags); /* Reserve space for the bomap */ if (n_bomap_pages) { diff --git a/drivers/gpu/drm/qxl/qxl_release.c b/drivers/gpu/drm/qxl/qxl_release.c index 0a693fede05e..30f85f0130cb 100644 --- a/drivers/gpu/drm/qxl/qxl_release.c +++ b/drivers/gpu/drm/qxl/qxl_release.c @@ -217,7 +217,7 @@ int qxl_release_list_add(struct qxl_release *release, struct qxl_bo *bo) qxl_bo_ref(bo); entry->tv.bo = &bo->tbo; - entry->tv.shared = false; + entry->tv.num_shared = 0; list_add_tail(&entry->tv.head, &release->bos); return 0; } diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c index 1ae31dbc61c6..f43305329939 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -178,7 +178,7 @@ static int radeon_cs_parser_relocs(struct radeon_cs_parser *p) } p->relocs[i].tv.bo = &p->relocs[i].robj->tbo; - p->relocs[i].tv.shared = !r->write_domain; + p->relocs[i].tv.num_shared = !r->write_domain; radeon_cs_buckets_add(&buckets, &p->relocs[i].tv.head, priority); @@ -253,7 +253,7 @@ static int radeon_cs_sync_rings(struct radeon_cs_parser *p) resv = reloc->robj->tbo.resv; r = radeon_sync_resv(p->rdev, &p->ib.sync, resv, - reloc->tv.shared); + reloc->tv.num_shared); if (r) return r; } diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c index 27d8e7dd2d06..44617dec8183 100644 --- a/drivers/gpu/drm/radeon/radeon_gem.c +++ b/drivers/gpu/drm/radeon/radeon_gem.c @@ -552,7 +552,7 @@ static void radeon_gem_va_update_vm(struct radeon_device *rdev, INIT_LIST_HEAD(&list); tv.bo = &bo_va->bo->tbo; - tv.shared = true; + tv.num_shared = 1; list_add(&tv.head, &list); vm_bos = radeon_vm_get_bos(rdev, bo_va->vm, &list); diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index a3d2ca07a058..0d374211661c 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -142,7 +142,7 @@ struct radeon_bo_list *radeon_vm_get_bos(struct radeon_device *rdev, list[0].preferred_domains = RADEON_GEM_DOMAIN_VRAM; list[0].allowed_domains = RADEON_GEM_DOMAIN_VRAM; list[0].tv.bo = &vm->page_directory->tbo; - list[0].tv.shared = true; + list[0].tv.num_shared = 1; list[0].tiling_flags = 0; list_add(&list[0].tv.head, head); @@ -154,7 +154,7 @@ struct radeon_bo_list *radeon_vm_get_bos(struct radeon_device *rdev, list[idx].preferred_domains = RADEON_GEM_DOMAIN_VRAM; list[idx].allowed_domains = RADEON_GEM_DOMAIN_VRAM; list[idx].tv.bo = &list[idx].robj->tbo; - list[idx].tv.shared = true; + list[idx].tv.num_shared = 1; list[idx].tiling_flags = 0; list_add(&list[idx++].tv.head, head); } diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c index 9d4cd196037a..dbb69063b3d5 100644 --- a/drivers/gpu/drm/scheduler/sched_main.c +++ b/drivers/gpu/drm/scheduler/sched_main.c @@ -211,6 +211,62 @@ void drm_sched_fault(struct drm_gpu_scheduler *sched) } EXPORT_SYMBOL(drm_sched_fault); +/** + * drm_sched_suspend_timeout - Suspend scheduler job timeout + * + * @sched: scheduler instance for which to suspend the timeout + * + * Suspend the delayed work timeout for the scheduler. This is done by + * modifying the delayed work timeout to an arbitrary large value, + * MAX_SCHEDULE_TIMEOUT in this case. Note that this function can be + * called from an IRQ context. + * + * Returns the timeout remaining + * + */ +unsigned long drm_sched_suspend_timeout(struct drm_gpu_scheduler *sched) +{ + unsigned long sched_timeout, now = jiffies; + + sched_timeout = sched->work_tdr.timer.expires; + + /* + * Modify the timeout to an arbitrarily large value. This also prevents + * the timeout to be restarted when new submissions arrive + */ + if (mod_delayed_work(system_wq, &sched->work_tdr, MAX_SCHEDULE_TIMEOUT) + && time_after(sched_timeout, now)) + return sched_timeout - now; + else + return sched->timeout; +} +EXPORT_SYMBOL(drm_sched_suspend_timeout); + +/** + * drm_sched_resume_timeout - Resume scheduler job timeout + * + * @sched: scheduler instance for which to resume the timeout + * @remaining: remaining timeout + * + * Resume the delayed work timeout for the scheduler. Note that + * this function can be called from an IRQ context. + */ +void drm_sched_resume_timeout(struct drm_gpu_scheduler *sched, + unsigned long remaining) +{ + unsigned long flags; + + spin_lock_irqsave(&sched->job_list_lock, flags); + + if (list_empty(&sched->ring_mirror_list)) + cancel_delayed_work(&sched->work_tdr); + else + mod_delayed_work(system_wq, &sched->work_tdr, remaining); + + spin_unlock_irqrestore(&sched->job_list_lock, flags); +} +EXPORT_SYMBOL(drm_sched_resume_timeout); + /* job_finish is called after hw fence signaled */ static void drm_sched_job_finish(struct work_struct *work) @@ -218,6 +274,7 @@ static void drm_sched_job_finish(struct work_struct *work) struct drm_sched_job *s_job = container_of(work, struct drm_sched_job, finish_work); struct drm_gpu_scheduler *sched = s_job->sched; + unsigned long flags; /* * Canceling the timeout without removing our job from the ring mirror @@ -228,12 +285,12 @@ static void drm_sched_job_finish(struct work_struct *work) */ cancel_delayed_work_sync(&sched->work_tdr); - spin_lock(&sched->job_list_lock); + spin_lock_irqsave(&sched->job_list_lock, flags); /* remove job from ring_mirror_list */ list_del_init(&s_job->node); /* queue TDR for next job */ drm_sched_start_timeout(sched); - spin_unlock(&sched->job_list_lock); + spin_unlock_irqrestore(&sched->job_list_lock, flags); sched->ops->free_job(s_job); } @@ -249,20 +306,22 @@ static void drm_sched_job_finish_cb(struct dma_fence *f, static void drm_sched_job_begin(struct drm_sched_job *s_job) { struct drm_gpu_scheduler *sched = s_job->sched; + unsigned long flags; dma_fence_add_callback(&s_job->s_fence->finished, &s_job->finish_cb, drm_sched_job_finish_cb); - spin_lock(&sched->job_list_lock); + spin_lock_irqsave(&sched->job_list_lock, flags); list_add_tail(&s_job->node, &sched->ring_mirror_list); drm_sched_start_timeout(sched); - spin_unlock(&sched->job_list_lock); + spin_unlock_irqrestore(&sched->job_list_lock, flags); } static void drm_sched_job_timedout(struct work_struct *work) { struct drm_gpu_scheduler *sched; struct drm_sched_job *job; + unsigned long flags; sched = container_of(work, struct drm_gpu_scheduler, work_tdr.work); job = list_first_entry_or_null(&sched->ring_mirror_list, @@ -271,9 +330,9 @@ static void drm_sched_job_timedout(struct work_struct *work) if (job) job->sched->ops->timedout_job(job); - spin_lock(&sched->job_list_lock); + spin_lock_irqsave(&sched->job_list_lock, flags); drm_sched_start_timeout(sched); - spin_unlock(&sched->job_list_lock); + spin_unlock_irqrestore(&sched->job_list_lock, flags); } /** @@ -287,9 +346,10 @@ void drm_sched_hw_job_reset(struct drm_gpu_scheduler *sched, struct drm_sched_jo { struct drm_sched_job *s_job; struct drm_sched_entity *entity, *tmp; + unsigned long flags; int i; - spin_lock(&sched->job_list_lock); + spin_lock_irqsave(&sched->job_list_lock, flags); list_for_each_entry_reverse(s_job, &sched->ring_mirror_list, node) { if (s_job->s_fence->parent && dma_fence_remove_callback(s_job->s_fence->parent, @@ -299,7 +359,7 @@ void drm_sched_hw_job_reset(struct drm_gpu_scheduler *sched, struct drm_sched_jo atomic_dec(&sched->hw_rq_count); } } - spin_unlock(&sched->job_list_lock); + spin_unlock_irqrestore(&sched->job_list_lock, flags); if (bad && bad->s_priority != DRM_SCHED_PRIORITY_KERNEL) { atomic_inc(&bad->karma); @@ -337,9 +397,10 @@ void drm_sched_job_recovery(struct drm_gpu_scheduler *sched) { struct drm_sched_job *s_job, *tmp; bool found_guilty = false; + unsigned long flags; int r; - spin_lock(&sched->job_list_lock); + spin_lock_irqsave(&sched->job_list_lock, flags); list_for_each_entry_safe(s_job, tmp, &sched->ring_mirror_list, node) { struct drm_sched_fence *s_fence = s_job->s_fence; struct dma_fence *fence; @@ -353,7 +414,7 @@ void drm_sched_job_recovery(struct drm_gpu_scheduler *sched) if (found_guilty && s_job->s_fence->scheduled.context == guilty_context) dma_fence_set_error(&s_fence->finished, -ECANCELED); - spin_unlock(&sched->job_list_lock); + spin_unlock_irqrestore(&sched->job_list_lock, flags); fence = sched->ops->run_job(s_job); atomic_inc(&sched->hw_rq_count); @@ -372,10 +433,10 @@ void drm_sched_job_recovery(struct drm_gpu_scheduler *sched) drm_sched_expel_job_unlocked(s_job); drm_sched_process_job(NULL, &s_fence->cb); } - spin_lock(&sched->job_list_lock); + spin_lock_irqsave(&sched->job_list_lock, flags); } drm_sched_start_timeout(sched); - spin_unlock(&sched->job_list_lock); + spin_unlock_irqrestore(&sched->job_list_lock, flags); } EXPORT_SYMBOL(drm_sched_job_recovery); @@ -612,7 +673,7 @@ int drm_sched_init(struct drm_gpu_scheduler *sched, long timeout, const char *name) { - int i; + int i, ret; sched->ops = ops; sched->hw_submission_limit = hw_submission; sched->name = name; @@ -633,8 +694,10 @@ int drm_sched_init(struct drm_gpu_scheduler *sched, /* Each scheduler will run on a seperate kernel thread */ sched->thread = kthread_run(drm_sched_main, sched, sched->name); if (IS_ERR(sched->thread)) { + ret = PTR_ERR(sched->thread); + sched->thread = NULL; DRM_ERROR("Failed to create scheduler for %s.\n", name); - return PTR_ERR(sched->thread); + return ret; } sched->ready = true; diff --git a/drivers/gpu/drm/ttm/ttm_execbuf_util.c b/drivers/gpu/drm/ttm/ttm_execbuf_util.c index efa005a1c1b7..93860346c426 100644 --- a/drivers/gpu/drm/ttm/ttm_execbuf_util.c +++ b/drivers/gpu/drm/ttm/ttm_execbuf_util.c @@ -126,10 +126,11 @@ int ttm_eu_reserve_buffers(struct ww_acquire_ctx *ticket, } if (!ret) { - if (!entry->shared) + if (!entry->num_shared) continue; - ret = reservation_object_reserve_shared(bo->resv, 1); + ret = reservation_object_reserve_shared(bo->resv, + entry->num_shared); if (!ret) continue; } @@ -150,8 +151,9 @@ int ttm_eu_reserve_buffers(struct ww_acquire_ctx *ticket, } } - if (!ret && entry->shared) - ret = reservation_object_reserve_shared(bo->resv, 1); + if (!ret && entry->num_shared) + ret = reservation_object_reserve_shared(bo->resv, + entry->num_shared); if (unlikely(ret != 0)) { if (ret == -EINTR) @@ -199,7 +201,7 @@ void ttm_eu_fence_buffer_objects(struct ww_acquire_ctx *ticket, list_for_each_entry(entry, list, head) { bo = entry->bo; - if (entry->shared) + if (entry->num_shared) reservation_object_add_shared_fence(bo->resv, fence); else reservation_object_add_excl_fence(bo->resv, fence); diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c index 8a029bade32a..3025bfc001a1 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c @@ -85,7 +85,7 @@ static void vmw_resource_release(struct kref *kref) struct ttm_validate_buffer val_buf; val_buf.bo = bo; - val_buf.shared = false; + val_buf.num_shared = 0; res->func->unbind(res, false, &val_buf); } res->backup_dirty = false; @@ -462,7 +462,7 @@ vmw_resource_check_buffer(struct ww_acquire_ctx *ticket, INIT_LIST_HEAD(&val_list); val_buf->bo = ttm_bo_reference(&res->backup->base); - val_buf->shared = false; + val_buf->num_shared = 0; list_add_tail(&val_buf->head, &val_list); ret = ttm_eu_reserve_buffers(ticket, &val_list, interruptible, NULL); if (unlikely(ret != 0)) @@ -565,7 +565,7 @@ static int vmw_resource_do_evict(struct ww_acquire_ctx *ticket, BUG_ON(!func->may_evict); val_buf.bo = NULL; - val_buf.shared = false; + val_buf.num_shared = 0; ret = vmw_resource_check_buffer(ticket, res, interruptible, &val_buf); if (unlikely(ret != 0)) return ret; @@ -614,7 +614,7 @@ int vmw_resource_validate(struct vmw_resource *res, bool intr) return 0; val_buf.bo = NULL; - val_buf.shared = false; + val_buf.num_shared = 0; if (res->backup) val_buf.bo = &res->backup->base; do { @@ -685,7 +685,7 @@ void vmw_resource_unbind_list(struct vmw_buffer_object *vbo) struct vmw_resource *res, *next; struct ttm_validate_buffer val_buf = { .bo = &vbo->base, - .shared = false + .num_shared = 0 }; lockdep_assert_held(&vbo->base.resv->lock.base); diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c b/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c index 184025fa938e..fef22753f4de 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c @@ -266,7 +266,7 @@ int vmw_validation_add_bo(struct vmw_validation_context *ctx, val_buf->bo = ttm_bo_get_unless_zero(&vbo->base); if (!val_buf->bo) return -ESRCH; - val_buf->shared = false; + val_buf->num_shared = 0; list_add_tail(&val_buf->head, &ctx->bo_list); bo_node->as_mob = as_mob; bo_node->cpu_blit = cpu_blit; diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h index 926379d53484..47e19796c450 100644 --- a/include/drm/gpu_scheduler.h +++ b/include/drm/gpu_scheduler.h @@ -331,4 +331,8 @@ struct drm_sched_fence *drm_sched_fence_create( void drm_sched_fence_scheduled(struct drm_sched_fence *fence); void drm_sched_fence_finished(struct drm_sched_fence *fence); +unsigned long drm_sched_suspend_timeout(struct drm_gpu_scheduler *sched); +void drm_sched_resume_timeout(struct drm_gpu_scheduler *sched, + unsigned long remaining); + #endif diff --git a/include/drm/ttm/ttm_execbuf_util.h b/include/drm/ttm/ttm_execbuf_util.h index b0fdd1980034..621615fa7728 100644 --- a/include/drm/ttm/ttm_execbuf_util.h +++ b/include/drm/ttm/ttm_execbuf_util.h @@ -40,13 +40,13 @@ * * @head: list head for thread-private list. * @bo: refcounted buffer object pointer. - * @shared: should the fence be added shared? + * @num_shared: How many shared fences we want to add. */ struct ttm_validate_buffer { struct list_head head; struct ttm_buffer_object *bo; - bool shared; + unsigned int num_shared; }; /** diff --git a/include/uapi/linux/kfd_ioctl.h b/include/uapi/linux/kfd_ioctl.h index b01eb502d49c..e622fd1fbd46 100644 --- a/include/uapi/linux/kfd_ioctl.h +++ b/include/uapi/linux/kfd_ioctl.h @@ -398,6 +398,24 @@ struct kfd_ioctl_unmap_memory_from_gpu_args { __u32 n_success; /* to/from KFD */ }; +struct kfd_ioctl_get_dmabuf_info_args { + __u64 size; /* from KFD */ + __u64 metadata_ptr; /* to KFD */ + __u32 metadata_size; /* to KFD (space allocated by user) + * from KFD (actual metadata size) + */ + __u32 gpu_id; /* from KFD */ + __u32 flags; /* from KFD (KFD_IOC_ALLOC_MEM_FLAGS) */ + __u32 dmabuf_fd; /* to KFD */ +}; + +struct kfd_ioctl_import_dmabuf_args { + __u64 va_addr; /* to KFD */ + __u64 handle; /* from KFD */ + __u32 gpu_id; /* to KFD */ + __u32 dmabuf_fd; /* to KFD */ +}; + #define AMDKFD_IOCTL_BASE 'K' #define AMDKFD_IO(nr) _IO(AMDKFD_IOCTL_BASE, nr) #define AMDKFD_IOR(nr, type) _IOR(AMDKFD_IOCTL_BASE, nr, type) @@ -486,7 +504,13 @@ struct kfd_ioctl_unmap_memory_from_gpu_args { #define AMDKFD_IOC_GET_QUEUE_WAVE_STATE \ AMDKFD_IOWR(0x1B, struct kfd_ioctl_get_queue_wave_state_args) +#define AMDKFD_IOC_GET_DMABUF_INFO \ + AMDKFD_IOWR(0x1C, struct kfd_ioctl_get_dmabuf_info_args) + +#define AMDKFD_IOC_IMPORT_DMABUF \ + AMDKFD_IOWR(0x1D, struct kfd_ioctl_import_dmabuf_args) + #define AMDKFD_COMMAND_START 0x01 -#define AMDKFD_COMMAND_END 0x1C +#define AMDKFD_COMMAND_END 0x1E #endif |