diff options
| author | Dave Airlie <airlied@redhat.com> | 2021-05-21 15:29:34 +1000 |
|---|---|---|
| committer | Dave Airlie <airlied@redhat.com> | 2021-05-21 15:29:40 +1000 |
| commit | c99c4d0ca57c978dcc2a2f41ab8449684ea154cc (patch) | |
| tree | 3fd20557381e99063293ae5d399a54d0108bcdde /drivers/gpu/drm | |
| parent | 2ba047855096fff551402a87272b520fe97323f5 (diff) | |
| parent | 2bb5b5f688cbbd5030629905d3ed8032ab46e79f (diff) | |
| download | linux-c99c4d0ca57c978dcc2a2f41ab8449684ea154cc.tar.gz | |
Merge tag 'amd-drm-next-5.14-2021-05-19' of https://gitlab.freedesktop.org/agd5f/linux into drm-next
amd-drm-next-5.14-2021-05-19:
amdgpu:
- Aldebaran updates
- More LTTPR display work
- Vangogh updates
- SDMA 5.x GCR fixes
- RAS fixes
- PCIe ASPM support
- Modifier fixes
- Enable TMZ on Renoir
- Buffer object code cleanup
- Display overlay fixes
- Initial support for multiple eDP panels
- Initial SR-IOV support for Aldebaran
- DP link training refactor
- Misc code cleanups and bug fixes
- SMU regression fixes for variable sized arrays
- MAINTAINERS fixes for amdgpu
amdkfd:
- Initial SR-IOV support for Aldebaran
- Topology fixes
- Initial HMM SVM support
- Misc code cleanups and bug fixes
radeon:
- Misc code cleanups and bug fixes
- SMU regression fixes for variable sized arrays
- Flickering fix for Oland with multiple 4K displays
UAPI:
- amdgpu: Drop AMDGPU_GEM_CREATE_SHADOW flag.
This was always a kernel internal flag and userspace use of it has always been blocked.
It's no longer needed so remove it.
- amdkgd: HMM SVM support
Overview: https://patchwork.freedesktop.org/series/85562/
Porposed userspace: https://github.com/RadeonOpenCompute/ROCT-Thunk-Interface/tree/fxkamd/hmm-wip
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210520031258.231896-1-alexander.deucher@amd.com
Diffstat (limited to 'drivers/gpu/drm')
156 files changed, 8706 insertions, 1550 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index d216b7ecb5d1..cc36570b0d2f 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -56,7 +56,7 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \ amdgpu_gmc.o amdgpu_mmhub.o amdgpu_xgmi.o amdgpu_csa.o amdgpu_ras.o amdgpu_vm_cpu.o \ amdgpu_vm_sdma.o amdgpu_discovery.o amdgpu_ras_eeprom.o amdgpu_nbio.o \ amdgpu_umc.o smu_v11_0_i2c.o amdgpu_fru_eeprom.o amdgpu_rap.o \ - amdgpu_fw_attestation.o amdgpu_securedisplay.o + amdgpu_fw_attestation.o amdgpu_securedisplay.o amdgpu_hdp.o amdgpu-$(CONFIG_PROC_FS) += amdgpu_fdinfo.o diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 6cabecc74007..ad5f508924b8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1075,7 +1075,8 @@ struct amdgpu_device { atomic_t throttling_logging_enabled; struct ratelimit_state throttling_logging_rs; - uint32_t ras_features; + uint32_t ras_hw_enabled; + uint32_t ras_enabled; bool in_pci_err_recovery; struct pci_saved_state *pci_state; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c index 2e9b16fb3fcd..bf2939b6eb43 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c @@ -76,7 +76,7 @@ struct amdgpu_atif { /** * amdgpu_atif_call - call an ATIF method * - * @handle: acpi handle + * @atif: acpi handle * @function: the ATIF function to execute * @params: ATIF function params * @@ -166,7 +166,6 @@ static void amdgpu_atif_parse_functions(struct amdgpu_atif_functions *f, u32 mas /** * amdgpu_atif_verify_interface - verify ATIF * - * @handle: acpi handle * @atif: amdgpu atif struct * * Execute the ATIF_FUNCTION_VERIFY_INTERFACE ATIF function @@ -240,8 +239,7 @@ out: /** * amdgpu_atif_get_notification_params - determine notify configuration * - * @handle: acpi handle - * @n: atif notification configuration struct + * @atif: acpi handle * * Execute the ATIF_FUNCTION_GET_SYSTEM_PARAMETERS ATIF function * to determine if a notifier is used and if so which one @@ -304,7 +302,7 @@ out: /** * amdgpu_atif_query_backlight_caps - get min and max backlight input signal * - * @handle: acpi handle + * @atif: acpi handle * * Execute the QUERY_BRIGHTNESS_TRANSFER_CHARACTERISTICS ATIF function * to determine the acceptable range of backlight values @@ -363,7 +361,7 @@ out: /** * amdgpu_atif_get_sbios_requests - get requested sbios event * - * @handle: acpi handle + * @atif: acpi handle * @req: atif sbios request struct * * Execute the ATIF_FUNCTION_GET_SYSTEM_BIOS_REQUESTS ATIF function @@ -899,6 +897,8 @@ void amdgpu_acpi_fini(struct amdgpu_device *adev) /** * amdgpu_acpi_is_s0ix_supported * + * @adev: amdgpu_device_pointer + * * returns true if supported, false if not. */ bool amdgpu_acpi_is_s0ix_supported(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index 5ffb07b02810..313ee49b9f17 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -75,6 +75,7 @@ struct amdgpu_amdkfd_fence { struct mm_struct *mm; spinlock_t lock; char timeline_name[TASK_COMM_LEN]; + struct svm_range_bo *svm_bo; }; struct amdgpu_kfd_dev { @@ -148,7 +149,8 @@ int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev, int queue_bit); struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, - struct mm_struct *mm); + struct mm_struct *mm, + struct svm_range_bo *svm_bo); #if IS_ENABLED(CONFIG_HSA_AMD) bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm); struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f); @@ -234,22 +236,27 @@ uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *s }) /* GPUVM API */ +#define drm_priv_to_vm(drm_priv) \ + (&((struct amdgpu_fpriv *) \ + ((struct drm_file *)(drm_priv))->driver_priv)->vm) + int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, struct file *filp, u32 pasid, - void **vm, void **process_info, + void **process_info, struct dma_fence **ef); -void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm); -uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm); +void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *drm_priv); +uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv); int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( struct kgd_dev *kgd, uint64_t va, uint64_t size, - void *vm, struct kgd_mem **mem, + void *drm_priv, struct kgd_mem **mem, uint64_t *offset, uint32_t flags); int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, uint64_t *size); + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv, + uint64_t *size); int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *vm); + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv); int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *vm); + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv); int amdgpu_amdkfd_gpuvm_sync_memory( struct kgd_dev *kgd, struct kgd_mem *mem, bool intr); int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd, @@ -260,7 +267,7 @@ int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, struct kfd_vm_fault_info *info); int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd, struct dma_buf *dmabuf, - uint64_t va, void *vm, + uint64_t va, void *drm_priv, struct kgd_mem **mem, uint64_t *size, uint64_t *mmap_offset); int amdgpu_amdkfd_get_tile_config(struct kgd_dev *kgd, @@ -270,6 +277,7 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void); void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_amdkfd_unreserve_memory_limit(struct amdgpu_bo *bo); +void amdgpu_amdkfd_reserve_system_mem(uint64_t size); #else static inline void amdgpu_amdkfd_gpuvm_init_mem_limits(void) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c index 9ef9f3ddad48..6409d6b1b2df 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c @@ -25,6 +25,7 @@ #include <linux/firmware.h> #include "amdgpu.h" #include "amdgpu_amdkfd.h" +#include "amdgpu_amdkfd_arcturus.h" #include "sdma0/sdma0_4_2_2_offset.h" #include "sdma0/sdma0_4_2_2_sh_mask.h" #include "sdma1/sdma1_4_2_2_offset.h" diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c index 5af464933976..1d0dbff87d3f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c @@ -28,6 +28,7 @@ #include <linux/slab.h> #include <linux/sched/mm.h> #include "amdgpu_amdkfd.h" +#include "kfd_svm.h" static const struct dma_fence_ops amdkfd_fence_ops; static atomic_t fence_seq = ATOMIC_INIT(0); @@ -60,7 +61,8 @@ static atomic_t fence_seq = ATOMIC_INIT(0); */ struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, - struct mm_struct *mm) + struct mm_struct *mm, + struct svm_range_bo *svm_bo) { struct amdgpu_amdkfd_fence *fence; @@ -73,7 +75,7 @@ struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, fence->mm = mm; get_task_comm(fence->timeline_name, current); spin_lock_init(&fence->lock); - + fence->svm_bo = svm_bo; dma_fence_init(&fence->base, &amdkfd_fence_ops, &fence->lock, context, atomic_inc_return(&fence_seq)); @@ -111,6 +113,8 @@ static const char *amdkfd_fence_get_timeline_name(struct dma_fence *f) * a KFD BO and schedules a job to move the BO. * If fence is already signaled return true. * If fence is not signaled schedule a evict KFD process work item. + * + * @f: dma_fence */ static bool amdkfd_fence_enable_signaling(struct dma_fence *f) { @@ -122,16 +126,20 @@ static bool amdkfd_fence_enable_signaling(struct dma_fence *f) if (dma_fence_is_signaled(f)) return true; - if (!kgd2kfd_schedule_evict_and_restore_process(fence->mm, f)) - return true; - + if (!fence->svm_bo) { + if (!kgd2kfd_schedule_evict_and_restore_process(fence->mm, f)) + return true; + } else { + if (!svm_range_schedule_evict_svm_bo(fence)) + return true; + } return false; } /** * amdkfd_fence_release - callback that fence can be freed * - * @fence: fence + * @f: dma_fence * * This function is called when the reference count becomes zero. * Drops the mm_struct reference and RCU schedules freeing up the fence. diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c index b43e68fc1378..ed3014fbb563 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -719,7 +719,7 @@ static void unlock_spi_csq_mutexes(struct amdgpu_device *adev) } /** - * @get_wave_count: Read device registers to get number of waves in flight for + * get_wave_count: Read device registers to get number of waves in flight for * a particular queue. The method also returns the VMID associated with the * queue. * @@ -755,19 +755,19 @@ static void get_wave_count(struct amdgpu_device *adev, int queue_idx, } /** - * @kgd_gfx_v9_get_cu_occupancy: Reads relevant registers associated with each + * kgd_gfx_v9_get_cu_occupancy: Reads relevant registers associated with each * shader engine and aggregates the number of waves that are in flight for the * process whose pasid is provided as a parameter. The process could have ZERO * or more queues running and submitting waves to compute units. * * @kgd: Handle of device from which to get number of waves in flight * @pasid: Identifies the process for which this query call is invoked - * @wave_cnt: Output parameter updated with number of waves in flight that + * @pasid_wave_cnt: Output parameter updated with number of waves in flight that * belong to process with given pasid * @max_waves_per_cu: Output parameter updated with maximum number of waves * possible per Compute Unit * - * @note: It's possible that the device has too many queues (oversubscription) + * Note: It's possible that the device has too many queues (oversubscription) * in which case a VMID could be remapped to a different PASID. This could lead * to an iaccurate wave count. Following is a high-level sequence: * Time T1: vmid = getVmid(); vmid is associated with Pasid P1 diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index 7d4118c8128a..dfa025d694f8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -33,9 +33,6 @@ #include <uapi/linux/kfd_ioctl.h> #include "amdgpu_xgmi.h" -/* BO flag to indicate a KFD userptr BO */ -#define AMDGPU_AMDKFD_USERPTR_BO (1ULL << 63) - /* Userptr restore delay, just long enough to allow consecutive VM * changes to accumulate */ @@ -108,6 +105,11 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void) (kfd_mem_limit.max_ttm_mem_limit >> 20)); } +void amdgpu_amdkfd_reserve_system_mem(uint64_t size) +{ + kfd_mem_limit.system_mem_used += size; +} + /* Estimate page table size needed to represent a given memory size * * With 4KB pages, we need one 8 byte PTE for each 4KB of memory @@ -217,7 +219,7 @@ void amdgpu_amdkfd_unreserve_memory_limit(struct amdgpu_bo *bo) u32 domain = bo->preferred_domains; bool sg = (bo->preferred_domains == AMDGPU_GEM_DOMAIN_CPU); - if (bo->flags & AMDGPU_AMDKFD_USERPTR_BO) { + if (bo->flags & AMDGPU_AMDKFD_CREATE_USERPTR_BO) { domain = AMDGPU_GEM_DOMAIN_CPU; sg = false; } @@ -967,7 +969,8 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, info->eviction_fence = amdgpu_amdkfd_fence_create(dma_fence_context_alloc(1), - current->mm); + current->mm, + NULL); if (!info->eviction_fence) { pr_err("Failed to create eviction fence\n"); ret = -ENOMEM; @@ -1036,15 +1039,19 @@ create_evict_fence_fail: int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, struct file *filp, u32 pasid, - void **vm, void **process_info, + void **process_info, struct dma_fence **ef) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct drm_file *drm_priv = filp->private_data; - struct amdgpu_fpriv *drv_priv = drm_priv->driver_priv; - struct amdgpu_vm *avm = &drv_priv->vm; + struct amdgpu_fpriv *drv_priv; + struct amdgpu_vm *avm; int ret; + ret = amdgpu_file_to_fpriv(filp, &drv_priv); + if (ret) + return ret; + avm = &drv_priv->vm; + /* Already a compute VM? */ if (avm->process_info) return -EINVAL; @@ -1059,7 +1066,7 @@ int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, if (ret) return ret; - *vm = (void *)avm; + amdgpu_vm_set_task_info(avm); return 0; } @@ -1100,15 +1107,17 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, } } -void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm) +void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *drm_priv) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + struct amdgpu_vm *avm; - if (WARN_ON(!kgd || !vm)) + if (WARN_ON(!kgd || !drm_priv)) return; - pr_debug("Releasing process vm %p\n", vm); + avm = drm_priv_to_vm(drm_priv); + + pr_debug("Releasing process vm %p\n", avm); /* The original pasid of amdgpu vm has already been * released during making a amdgpu vm to a compute vm @@ -1119,9 +1128,9 @@ void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm) amdgpu_vm_release_compute(adev, avm); } -uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm) +uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv) { - struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); struct amdgpu_bo *pd = avm->root.base.bo; struct amdgpu_device *adev = amdgpu_ttm_adev(pd->tbo.bdev); @@ -1132,11 +1141,11 @@ uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm) int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( struct kgd_dev *kgd, uint64_t va, uint64_t size, - void *vm, struct kgd_mem **mem, + void *drm_priv, struct kgd_mem **mem, uint64_t *offset, uint32_t flags) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); enum ttm_bo_type bo_type = ttm_bo_type_device; struct sg_table *sg = NULL; uint64_t user_addr = 0; @@ -1216,6 +1225,11 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( domain_string(alloc_domain), ret); goto err_bo_create; } + ret = drm_vma_node_allow(&gobj->vma_node, drm_priv); + if (ret) { + pr_debug("Failed to allow vma node access. ret %d\n", ret); + goto err_node_allow; + } bo = gem_to_amdgpu_bo(gobj); if (bo_type == ttm_bo_type_sg) { bo->tbo.sg = sg; @@ -1224,7 +1238,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( bo->kfd_bo = *mem; (*mem)->bo = bo; if (user_addr) - bo->flags |= AMDGPU_AMDKFD_USERPTR_BO; + bo->flags |= AMDGPU_AMDKFD_CREATE_USERPTR_BO; (*mem)->va = va; (*mem)->domain = domain; @@ -1245,6 +1259,8 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( allocate_init_user_pages_failed: remove_kgd_mem_from_kfd_bo_list(*mem, avm->process_info); + drm_vma_node_revoke(&gobj->vma_node, drm_priv); +err_node_allow: amdgpu_bo_unref(&bo); /* Don't unreserve system mem limit twice */ goto err_reserve_limit; @@ -1262,7 +1278,8 @@ err: } int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, uint64_t *size) + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv, + uint64_t *size) { struct amdkfd_process_info *process_info = mem->process_info; unsigned long bo_size = mem->bo->tbo.base.size; @@ -1339,6 +1356,7 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( } /* Free the BO*/ + drm_vma_node_revoke(&mem->bo->tbo.base.vma_node, drm_priv); drm_gem_object_put(&mem->bo->tbo.base); mutex_destroy(&mem->lock); kfree(mem); @@ -1347,10 +1365,10 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( } int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *vm) + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); int ret; struct amdgpu_bo *bo; uint32_t domain; @@ -1391,9 +1409,9 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( pr_debug("Map VA 0x%llx - 0x%llx to vm %p domain %s\n", mem->va, mem->va + bo_size * (1 + mem->aql_queue), - vm, domain_string(domain)); + avm, domain_string(domain)); - ret = reserve_bo_and_vm(mem, vm, &ctx); + ret = reserve_bo_and_vm(mem, avm, &ctx); if (unlikely(ret)) goto out; @@ -1437,7 +1455,7 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( } list_for_each_entry(entry, &mem->bo_va_list, bo_list) { - if (entry->bo_va->base.vm == vm && !entry->is_mapped) { + if (entry->bo_va->base.vm == avm && !entry->is_mapped) { pr_debug("\t map VA 0x%llx - 0x%llx in entry %p\n", entry->va, entry->va + bo_size, entry); @@ -1449,7 +1467,7 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( goto map_bo_to_gpuvm_failed; } - ret = vm_update_pds(vm, ctx.sync); + ret = vm_update_pds(avm, ctx.sync); if (ret) { pr_err("Failed to update page directories\n"); goto map_bo_to_gpuvm_failed; @@ -1485,11 +1503,11 @@ out: } int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( - struct kgd_dev *kgd, struct kgd_mem *mem, void *vm) + struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct amdkfd_process_info *process_info = - ((struct amdgpu_vm *)vm)->process_info; + struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); + struct amdkfd_process_info *process_info = avm->process_info; unsigned long bo_size = mem->bo->tbo.base.size; struct kfd_bo_va_list *entry; struct bo_vm_reservation_context ctx; @@ -1497,7 +1515,7 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( mutex_lock(&mem->lock); - ret = reserve_bo_and_cond_vms(mem, vm, BO_VM_MAPPED, &ctx); + ret = reserve_bo_and_cond_vms(mem, avm, BO_VM_MAPPED, &ctx); if (unlikely(ret)) goto out; /* If no VMs were reserved, it means the BO wasn't actually mapped */ @@ -1506,17 +1524,17 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( goto unreserve_out; } - ret = vm_validate_pt_pd_bos((struct amdgpu_vm *)vm); + ret = vm_validate_pt_pd_bos(avm); if (unlikely(ret)) goto unreserve_out; pr_debug("Unmap VA 0x%llx - 0x%llx from vm %p\n", mem->va, mem->va + bo_size * (1 + mem->aql_queue), - vm); + avm); list_for_each_entry(entry, &mem->bo_va_list, bo_list) { - if (entry->bo_va->base.vm == vm && entry->is_mapped) { + if (entry->bo_va->base.vm == avm && entry->is_mapped) { pr_debug("\t unmap VA 0x%llx - 0x%llx from entry %p\n", entry->va, entry->va + bo_size, @@ -1642,14 +1660,15 @@ int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd, struct dma_buf *dma_buf, - uint64_t va, void *vm, + uint64_t va, void *drm_priv, struct kgd_mem **mem, uint64_t *size, uint64_t *mmap_offset) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv); struct drm_gem_object *obj; struct amdgpu_bo *bo; - struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + int ret; if (dma_buf->ops != &amdgpu_dmabuf_ops) /* Can't handle non-graphics buffers */ @@ -1670,6 +1689,12 @@ int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd, if (!*mem) return -ENOMEM; + ret = drm_vma_node_allow(&obj->vma_node, drm_priv); + if (ret) { + kfree(mem); + return ret; + } + if (size) *size = amdgpu_bo_size(bo); @@ -2135,7 +2160,8 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef) */ new_fence = amdgpu_amdkfd_fence_create( process_info->eviction_fence->base.context, - process_info->eviction_fence->mm); + process_info->eviction_fence->mm, + NULL); if (!new_fence) { pr_err("Failed to create eviction fence\n"); ret = -ENOMEM; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index b5c766998045..90136f9dedd6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -672,7 +672,7 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p) } /** - * cs_parser_fini() - clean parser states + * amdgpu_cs_parser_fini() - clean parser states * @parser: parser structure holding parsing context. * @error: error number * @backoff: indicator to backoff the reservation @@ -1488,7 +1488,7 @@ int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data, } /** - * amdgpu_cs_wait_all_fence - wait on all fences to signal + * amdgpu_cs_wait_all_fences - wait on all fences to signal * * @adev: amdgpu device * @filp: file private @@ -1639,7 +1639,7 @@ err_free_fences: } /** - * amdgpu_cs_find_bo_va - find bo_va for VM address + * amdgpu_cs_find_mapping - find bo_va for VM address * * @parser: command submission parser context * @addr: VM address diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 8b2a37bf2adf..2360a9c518eb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2856,7 +2856,7 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev) AMD_IP_BLOCK_TYPE_IH, }; - for (i = 0; i < ARRAY_SIZE(ip_order); i++) { + for (i = 0; i < adev->num_ip_blocks; i++) { int j; struct amdgpu_ip_block *block; @@ -3179,8 +3179,8 @@ static int amdgpu_device_get_job_timeout_settings(struct amdgpu_device *adev) int ret = 0; /* - * By default timeout for non compute jobs is 10000. - * And there is no timeout enforced on compute jobs. + * By default timeout for non compute jobs is 10000 + * and 60000 for compute jobs. * In SR-IOV or passthrough mode, timeout for compute * jobs are 60000 by default. */ @@ -3189,10 +3189,8 @@ static int amdgpu_device_get_job_timeout_settings(struct amdgpu_device *adev) if (amdgpu_sriov_vf(adev)) adev->compute_timeout = amdgpu_sriov_is_pp_one_vf(adev) ? msecs_to_jiffies(60000) : msecs_to_jiffies(10000); - else if (amdgpu_passthrough(adev)) - adev->compute_timeout = msecs_to_jiffies(60000); else - adev->compute_timeout = MAX_SCHEDULE_TIMEOUT; + adev->compute_timeout = msecs_to_jiffies(60000); if (strnlen(input, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) { while ((timeout_setting = strsep(&input, ",")) && @@ -3741,7 +3739,6 @@ void amdgpu_device_fini(struct amdgpu_device *adev) int amdgpu_device_suspend(struct drm_device *dev, bool fbcon) { struct amdgpu_device *adev = drm_to_adev(dev); - int r; if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; @@ -3756,7 +3753,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool fbcon) amdgpu_ras_suspend(adev); - r = amdgpu_device_ip_suspend_phase1(adev); + amdgpu_device_ip_suspend_phase1(adev); if (!adev->in_s0ix) amdgpu_amdkfd_suspend(adev, adev->in_runpm); @@ -3766,7 +3763,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool fbcon) amdgpu_fence_driver_suspend(adev); - r = amdgpu_device_ip_suspend_phase2(adev); + amdgpu_device_ip_suspend_phase2(adev); /* evict remaining vram memory * This second call to evict vram is to evict the gart page table * using the CPU. @@ -5124,7 +5121,8 @@ int amdgpu_device_baco_enter(struct drm_device *dev) if (!amdgpu_device_supports_baco(adev_to_drm(adev))) return -ENOTSUPP; - if (ras && ras->supported && adev->nbio.funcs->enable_doorbell_interrupt) + if (ras && adev->ras_enabled && + adev->nbio.funcs->enable_doorbell_interrupt) adev->nbio.funcs->enable_doorbell_interrupt(adev, false); return amdgpu_dpm_baco_enter(adev); @@ -5143,7 +5141,8 @@ int amdgpu_device_baco_exit(struct drm_device *dev) if (ret) return ret; - if (ras && ras->supported && adev->nbio.funcs->enable_doorbell_interrupt) + if (ras && adev->ras_enabled && + adev->nbio.funcs->enable_doorbell_interrupt) adev->nbio.funcs->enable_doorbell_interrupt(adev, true); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 6fd20ea2935b..1ed9748b9bc7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -288,9 +288,9 @@ module_param_named(msi, amdgpu_msi, int, 0444); * for SDMA and Video. * * By default(with no lockup_timeout settings), the timeout for all non-compute(GFX, SDMA and Video) - * jobs is 10000. And there is no timeout enforced on compute jobs. + * jobs is 10000. The timeout for compute is 60000. */ -MODULE_PARM_DESC(lockup_timeout, "GPU lockup timeout in ms (default: for bare metal 10000 for non-compute jobs and infinity timeout for compute jobs; " +MODULE_PARM_DESC(lockup_timeout, "GPU lockup timeout in ms (default: for bare metal 10000 for non-compute jobs and 60000 for compute jobs; " "for passthrough or sriov, 10000 for all jobs." " 0: keep default value. negative: infinity timeout), " "format: for bare metal [Non-Compute] or [GFX,Compute,SDMA,Video]; " @@ -641,7 +641,8 @@ module_param_named(mes, amdgpu_mes, int, 0444); /** * DOC: noretry (int) - * Disable retry faults in the GPU memory controller. + * Disable XNACK retry in the SQ by default on GFXv9 hardware. On ASICs that + * do not support per-process XNACK this also disables retry page faults. * (0 = retry enabled, 1 = retry disabled, -1 auto (default)) */ MODULE_PARM_DESC(noretry, @@ -1186,6 +1187,7 @@ static const struct pci_device_id pciidlist[] = { {0x1002, 0x7408, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, {0x1002, 0x740C, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, {0x1002, 0x740F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, + {0x1002, 0x7410, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN|AMD_EXP_HW_SUPPORT}, {0, 0, 0} }; @@ -1598,17 +1600,15 @@ static int amdgpu_pmops_runtime_idle(struct device *dev) if (amdgpu_device_has_dc_support(adev)) { struct drm_crtc *crtc; - drm_modeset_lock_all(drm_dev); - drm_for_each_crtc(crtc, drm_dev) { - if (crtc->state->active) { + drm_modeset_lock(&crtc->mutex, NULL); + if (crtc->state->active) ret = -EBUSY; + drm_modeset_unlock(&crtc->mutex); + if (ret < 0) break; - } } - drm_modeset_unlock_all(drm_dev); - } else { struct drm_connector *list_connector; struct drm_connector_list_iter iter; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index 47ea46859618..30772608eac6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -434,6 +434,7 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring, * * @ring: ring to init the fence driver on * @num_hw_submission: number of entries on the hardware queue + * @sched_score: optional score atomic shared with other schedulers * * Init the fence driver for the requested ring (all asics). * Helper function for amdgpu_fence_driver_init(). diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index c5a9a4fb10d2..5562b5c90c03 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -60,7 +60,7 @@ */ /** - * amdgpu_dummy_page_init - init dummy page used by the driver + * amdgpu_gart_dummy_page_init - init dummy page used by the driver * * @adev: amdgpu_device pointer * @@ -86,7 +86,7 @@ static int amdgpu_gart_dummy_page_init(struct amdgpu_device *adev) } /** - * amdgpu_dummy_page_fini - free dummy page used by the driver + * amdgpu_gart_dummy_page_fini - free dummy page used by the driver * * @adev: amdgpu_device pointer * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index c39ed9eb0987..a129ecc73869 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -333,6 +333,17 @@ void amdgpu_gmc_agp_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc) } /** + * amdgpu_gmc_fault_key - get hask key from vm fault address and pasid + * + * @addr: 48 bit physical address, page aligned (36 significant bits) + * @pasid: 16 bit process address space identifier + */ +static inline uint64_t amdgpu_gmc_fault_key(uint64_t addr, uint16_t pasid) +{ + return addr << 4 | pasid; +} + +/** * amdgpu_gmc_filter_faults - filter VM faults * * @adev: amdgpu device structure @@ -348,8 +359,7 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr, uint16_t pasid, uint64_t timestamp) { struct amdgpu_gmc *gmc = &adev->gmc; - - uint64_t stamp, key = addr << 4 | pasid; + uint64_t stamp, key = amdgpu_gmc_fault_key(addr, pasid); struct amdgpu_gmc_fault *fault; uint32_t hash; @@ -365,7 +375,7 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr, while (fault->timestamp >= stamp) { uint64_t tmp; - if (fault->key == key) + if (atomic64_read(&fault->key) == key) return true; tmp = fault->timestamp; @@ -378,7 +388,7 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr, /* Add the fault to the ring */ fault = &gmc->fault_ring[gmc->last_fault]; - fault->key = key; + atomic64_set(&fault->key, key); fault->timestamp = timestamp; /* And update the hash */ @@ -387,6 +397,36 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr, return false; } +/** + * amdgpu_gmc_filter_faults_remove - remove address from VM faults filter + * + * @adev: amdgpu device structure + * @addr: address of the VM fault + * @pasid: PASID of the process causing the fault + * + * Remove the address from fault filter, then future vm fault on this address + * will pass to retry fault handler to recover. + */ +void amdgpu_gmc_filter_faults_remove(struct amdgpu_device *adev, uint64_t addr, + uint16_t pasid) +{ + struct amdgpu_gmc *gmc = &adev->gmc; + uint64_t key = amdgpu_gmc_fault_key(addr, pasid); + struct amdgpu_gmc_fault *fault; + uint32_t hash; + uint64_t tmp; + + hash = hash_64(key, AMDGPU_GMC_FAULT_HASH_ORDER); + fault = &gmc->fault_ring[gmc->fault_hash[hash].idx]; + do { + if (atomic64_cmpxchg(&fault->key, key, 0) == key) + break; + + tmp = fault->timestamp; + fault = &gmc->fault_ring[fault->next]; + } while (fault->timestamp < tmp); +} + int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) { int r; @@ -415,6 +455,13 @@ int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) return r; } + if (adev->hdp.ras_funcs && + adev->hdp.ras_funcs->ras_late_init) { + r = adev->hdp.ras_funcs->ras_late_init(adev); + if (r) + return r; + } + return 0; } @@ -426,11 +473,15 @@ void amdgpu_gmc_ras_fini(struct amdgpu_device *adev) if (adev->mmhub.ras_funcs && adev->mmhub.ras_funcs->ras_fini) - amdgpu_mmhub_ras_fini(adev); + adev->mmhub.ras_funcs->ras_fini(adev); if (adev->gmc.xgmi.ras_funcs && adev->gmc.xgmi.ras_funcs->ras_fini) adev->gmc.xgmi.ras_funcs->ras_fini(adev); + + if (adev->hdp.ras_funcs && + adev->hdp.ras_funcs->ras_fini) + adev->hdp.ras_funcs->ras_fini(adev); } /* diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index 9d11c02a3938..6aa1d52d3aee 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -66,9 +66,9 @@ struct firmware; * GMC page fault information */ struct amdgpu_gmc_fault { - uint64_t timestamp; + uint64_t timestamp:48; uint64_t next:AMDGPU_GMC_FAULT_RING_ORDER; - uint64_t key:52; + atomic64_t key; }; /* @@ -318,6 +318,8 @@ void amdgpu_gmc_agp_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc); bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr, uint16_t pasid, uint64_t timestamp); +void amdgpu_gmc_filter_faults_remove(struct amdgpu_device *adev, uint64_t addr, + uint16_t pasid); int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev); void amdgpu_gmc_ras_fini(struct amdgpu_device *adev); int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c index 72962de4c04c..c026972ca9a1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c @@ -24,7 +24,8 @@ #include "amdgpu.h" -static inline struct amdgpu_gtt_mgr *to_gtt_mgr(struct ttm_resource_manager *man) +static inline struct amdgpu_gtt_mgr * +to_gtt_mgr(struct ttm_resource_manager *man) { return container_of(man, struct amdgpu_gtt_mgr, manager); } @@ -43,12 +44,14 @@ struct amdgpu_gtt_node { * the GTT block, in bytes */ static ssize_t amdgpu_mem_info_gtt_total_show(struct device *dev, - struct device_attribute *attr, char *buf) + struct device_attribute *attr, + char *buf) { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); + struct ttm_resource_manager *man; + man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); return sysfs_emit(buf, "%llu\n", man->size * PAGE_SIZE); } @@ -61,12 +64,14 @@ static ssize_t amdgpu_mem_info_gtt_total_show(struct device *dev, * size of the GTT block, in bytes */ static ssize_t amdgpu_mem_info_gtt_used_show(struct device *dev, - struct device_attribute *attr, char *buf) + struct device_attribute *attr, + char *buf) { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); + struct ttm_resource_manager *man; + man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); return sysfs_emit(buf, "%llu\n", amdgpu_gtt_mgr_usage(man)); } @@ -75,80 +80,6 @@ static DEVICE_ATTR(mem_info_gtt_total, S_IRUGO, static DEVICE_ATTR(mem_info_gtt_used, S_IRUGO, amdgpu_mem_info_gtt_used_show, NULL); -static const struct ttm_resource_manager_func amdgpu_gtt_mgr_func; -/** - * amdgpu_gtt_mgr_init - init GTT manager and DRM MM - * - * @adev: amdgpu_device pointer - * @gtt_size: maximum size of GTT - * - * Allocate and initialize the GTT manager. - */ -int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size) -{ - struct amdgpu_gtt_mgr *mgr = &adev->mman.gtt_mgr; - struct ttm_resource_manager *man = &mgr->manager; - uint64_t start, size; - int ret; - - man->use_tt = true; - man->func = &amdgpu_gtt_mgr_func; - - ttm_resource_manager_init(man, gtt_size >> PAGE_SHIFT); - - start = AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GTT_NUM_TRANSFER_WINDOWS; - size = (adev->gmc.gart_size >> PAGE_SHIFT) - start; - drm_mm_init(&mgr->mm, start, size); - spin_lock_init(&mgr->lock); - atomic64_set(&mgr->available, gtt_size >> PAGE_SHIFT); - - ret = device_create_file(adev->dev, &dev_attr_mem_info_gtt_total); - if (ret) { - DRM_ERROR("Failed to create device file mem_info_gtt_total\n"); - return ret; - } - ret = device_create_file(adev->dev, &dev_attr_mem_info_gtt_used); - if (ret) { - DRM_ERROR("Failed to create device file mem_info_gtt_used\n"); - return ret; - } - - ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_TT, &mgr->manager); - ttm_resource_manager_set_used(man, true); - return 0; -} - -/** - * amdgpu_gtt_mgr_fini - free and destroy GTT manager - * - * @adev: amdgpu_device pointer - * - * Destroy and free the GTT manager, returns -EBUSY if ranges are still - * allocated inside it. - */ -void amdgpu_gtt_mgr_fini(struct amdgpu_device *adev) -{ - struct amdgpu_gtt_mgr *mgr = &adev->mman.gtt_mgr; - struct ttm_resource_manager *man = &mgr->manager; - int ret; - - ttm_resource_manager_set_used(man, false); - - ret = ttm_resource_manager_evict_all(&adev->mman.bdev, man); - if (ret) - return; - - spin_lock(&mgr->lock); - drm_mm_takedown(&mgr->mm); - spin_unlock(&mgr->lock); - - device_remove_file(adev->dev, &dev_attr_mem_info_gtt_total); - device_remove_file(adev->dev, &dev_attr_mem_info_gtt_used); - - ttm_resource_manager_cleanup(man); - ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_TT, NULL); -} - /** * amdgpu_gtt_mgr_has_gart_addr - Check if mem has address space * @@ -265,6 +196,13 @@ uint64_t amdgpu_gtt_mgr_usage(struct ttm_resource_manager *man) return (result > 0 ? result : 0) * PAGE_SIZE; } +/** + * amdgpu_gtt_mgr_recover - re-init gart + * + * @man: TTM memory type manager + * + * Re-init the gart for each known BO in the GTT. + */ int amdgpu_gtt_mgr_recover(struct ttm_resource_manager *man) { struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man); @@ -311,3 +249,76 @@ static const struct ttm_resource_manager_func amdgpu_gtt_mgr_func = { .free = amdgpu_gtt_mgr_del, .debug = amdgpu_gtt_mgr_debug }; + +/** + * amdgpu_gtt_mgr_init - init GTT manager and DRM MM + * + * @adev: amdgpu_device pointer + * @gtt_size: maximum size of GTT + * + * Allocate and initialize the GTT manager. + */ +int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size) +{ + struct amdgpu_gtt_mgr *mgr = &adev->mman.gtt_mgr; + struct ttm_resource_manager *man = &mgr->manager; + uint64_t start, size; + int ret; + + man->use_tt = true; + man->func = &amdgpu_gtt_mgr_func; + + ttm_resource_manager_init(man, gtt_size >> PAGE_SHIFT); + + start = AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GTT_NUM_TRANSFER_WINDOWS; + size = (adev->gmc.gart_size >> PAGE_SHIFT) - start; + drm_mm_init(&mgr->mm, start, size); + spin_lock_init(&mgr->lock); + atomic64_set(&mgr->available, gtt_size >> PAGE_SHIFT); + + ret = device_create_file(adev->dev, &dev_attr_mem_info_gtt_total); + if (ret) { + DRM_ERROR("Failed to create device file mem_info_gtt_total\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_mem_info_gtt_used); + if (ret) { + DRM_ERROR("Failed to create device file mem_info_gtt_used\n"); + return ret; + } + + ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_TT, &mgr->manager); + ttm_resource_manager_set_used(man, true); + return 0; +} + +/** + * amdgpu_gtt_mgr_fini - free and destroy GTT manager + * + * @adev: amdgpu_device pointer + * + * Destroy and free the GTT manager, returns -EBUSY if ranges are still + * allocated inside it. + */ +void amdgpu_gtt_mgr_fini(struct amdgpu_device *adev) +{ + struct amdgpu_gtt_mgr *mgr = &adev->mman.gtt_mgr; + struct ttm_resource_manager *man = &mgr->manager; + int ret; + + ttm_resource_manager_set_used(man, false); + + ret = ttm_resource_manager_evict_all(&adev->mman.bdev, man); + if (ret) + return; + + spin_lock(&mgr->lock); + drm_mm_takedown(&mgr->mm); + spin_unlock(&mgr->lock); + + device_remove_file(adev->dev, &dev_attr_mem_info_gtt_total); + device_remove_file(adev->dev, &dev_attr_mem_info_gtt_used); + + ttm_resource_manager_cleanup(man); + ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_TT, NULL); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c new file mode 100644 index 000000000000..1d50d534d77c --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c @@ -0,0 +1,69 @@ +/* + * Copyright 2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "amdgpu_ras.h" + +int amdgpu_hdp_ras_late_init(struct amdgpu_device *adev) +{ + int r; + struct ras_ih_if ih_info = { + .cb = NULL, + }; + struct ras_fs_if fs_info = { + .sysfs_name = "hdp_err_count", + }; + + if (!adev->hdp.ras_if) { + adev->hdp.ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); + if (!adev->hdp.ras_if) + return -ENOMEM; + adev->hdp.ras_if->block = AMDGPU_RAS_BLOCK__HDP; + adev->hdp.ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; + adev->hdp.ras_if->sub_block_index = 0; + strcpy(adev->hdp.ras_if->name, "hdp"); + } + ih_info.head = fs_info.head = *adev->hdp.ras_if; + r = amdgpu_ras_late_init(adev, adev->hdp.ras_if, + &fs_info, &ih_info); + if (r || !amdgpu_ras_is_supported(adev, adev->hdp.ras_if->block)) { + kfree(adev->hdp.ras_if); + adev->hdp.ras_if = NULL; + } + + return r; +} + +void amdgpu_hdp_ras_fini(struct amdgpu_device *adev) +{ + if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__HDP) && + adev->hdp.ras_if) { + struct ras_common_if *ras_if = adev->hdp.ras_if; + struct ras_ih_if ih_info = { + .cb = NULL, + }; + + amdgpu_ras_late_fini(adev, ras_if, &ih_info); + kfree(ras_if); + } +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h index 43caf9f8cc11..7ec99d591584 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h @@ -23,18 +23,29 @@ #ifndef __AMDGPU_HDP_H__ #define __AMDGPU_HDP_H__ +struct amdgpu_hdp_ras_funcs { + int (*ras_late_init)(struct amdgpu_device *adev); + void (*ras_fini)(struct amdgpu_device *adev); + void (*query_ras_error_count)(struct amdgpu_device *adev, + void *ras_error_status); + void (*reset_ras_error_count)(struct amdgpu_device *adev); +}; + struct amdgpu_hdp_funcs { void (*flush_hdp)(struct amdgpu_device *adev, struct amdgpu_ring *ring); void (*invalidate_hdp)(struct amdgpu_device *adev, struct amdgpu_ring *ring); - void (*reset_ras_error_count)(struct amdgpu_device *adev); void (*update_clock_gating)(struct amdgpu_device *adev, bool enable); void (*get_clock_gating_state)(struct amdgpu_device *adev, u32 *flags); void (*init_registers)(struct amdgpu_device *adev); }; struct amdgpu_hdp { + struct ras_common_if *ras_if; const struct amdgpu_hdp_funcs *funcs; + const struct amdgpu_hdp_ras_funcs *ras_funcs; }; +int amdgpu_hdp_ras_late_init(struct amdgpu_device *adev); +void amdgpu_hdp_ras_fini(struct amdgpu_device *adev); #endif /* __AMDGPU_HDP_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index a2fe2dac32c1..2e6789a7dc46 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -328,7 +328,7 @@ int amdgpu_ib_pool_init(struct amdgpu_device *adev) for (i = 0; i < AMDGPU_IB_POOL_MAX; i++) { if (i == AMDGPU_IB_POOL_DIRECT) - size = PAGE_SIZE * 2; + size = PAGE_SIZE * 6; else size = AMDGPU_IB_POOL_SIZE; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c index faaa6aa2faaf..a36e191cf086 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c @@ -175,7 +175,9 @@ static bool amdgpu_ih_has_checkpoint_processed(struct amdgpu_device *adev, cur_rptr += ih->ptr_mask + 1; *prev_rptr = cur_rptr; - return cur_rptr >= checkpoint_wptr; + /* check ring is empty to workaround missing wptr overflow flag */ + return cur_rptr >= checkpoint_wptr || + (cur_rptr & ih->ptr_mask) == amdgpu_ih_get_wptr(adev, ih); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 39ee88d29cca..8d12e474745a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -986,7 +986,7 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) if (!ras) return -EINVAL; - ras_mask = (uint64_t)ras->supported << 32 | ras->features; + ras_mask = (uint64_t)adev->ras_enabled << 32 | ras->features; return copy_to_user(out, &ras_mask, min_t(u64, size, sizeof(ras_mask))) ? @@ -1114,7 +1114,8 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) dev_warn(adev->dev, "No more PASIDs available!"); pasid = 0; } - r = amdgpu_vm_init(adev, &fpriv->vm, AMDGPU_VM_CONTEXT_GFX, pasid); + + r = amdgpu_vm_init(adev, &fpriv->vm, pasid); if (r) goto error_pasid; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h index 11aa29933c1f..b27fcbccce2b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h @@ -28,6 +28,7 @@ struct amdgpu_mmhub_ras_funcs { void *ras_error_status); void (*query_ras_error_status)(struct amdgpu_device *adev); void (*reset_ras_error_count)(struct amdgpu_device *adev); + void (*reset_ras_error_status)(struct amdgpu_device *adev); }; struct amdgpu_mmhub_funcs { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c index 828b5167ff12..2741c28ff1b5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c @@ -155,3 +155,89 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo) mmu_interval_notifier_remove(&bo->notifier); bo->notifier.mm = NULL; } + +int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier, + struct mm_struct *mm, struct page **pages, + uint64_t start, uint64_t npages, + struct hmm_range **phmm_range, bool readonly, + bool mmap_locked) +{ + struct hmm_range *hmm_range; + unsigned long timeout; + unsigned long i; + unsigned long *pfns; + int r = 0; + + hmm_range = kzalloc(sizeof(*hmm_range), GFP_KERNEL); + if (unlikely(!hmm_range)) + return -ENOMEM; + + pfns = kvmalloc_array(npages, sizeof(*pfns), GFP_KERNEL); + if (unlikely(!pfns)) { + r = -ENOMEM; + goto out_free_range; + } + + hmm_range->notifier = notifier; + hmm_range->default_flags = HMM_PFN_REQ_FAULT; + if (!readonly) + hmm_range->default_flags |= HMM_PFN_REQ_WRITE; + hmm_range->hmm_pfns = pfns; + hmm_range->start = start; + hmm_range->end = start + npages * PAGE_SIZE; + + /* Assuming 512MB takes maxmium 1 second to fault page address */ + timeout = max(npages >> 17, 1ULL) * HMM_RANGE_DEFAULT_TIMEOUT; + timeout = jiffies + msecs_to_jiffies(timeout); + +retry: + hmm_range->notifier_seq = mmu_interval_read_begin(notifier); + + if (likely(!mmap_locked)) + mmap_read_lock(mm); + + r = hmm_range_fault(hmm_range); + + if (likely(!mmap_locked)) + mmap_read_unlock(mm); + if (unlikely(r)) { + /* + * FIXME: This timeout should encompass the retry from + * mmu_interval_read_retry() as well. + */ + if (r == -EBUSY && !time_after(jiffies, timeout)) + goto retry; + goto out_free_pfns; + } + + /* + * Due to default_flags, all pages are HMM_PFN_VALID or + * hmm_range_fault() fails. FIXME: The pages cannot be touched outside + * the notifier_lock, and mmu_interval_read_retry() must be done first. + */ + for (i = 0; pages && i < npages; i++) + pages[i] = hmm_pfn_to_page(pfns[i]); + + *phmm_range = hmm_range; + + return 0; + +out_free_pfns: + kvfree(pfns); +out_free_range: + kfree(hmm_range); + + return r; +} + +int amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range) +{ + int r; + + r = mmu_interval_read_retry(hmm_range->notifier, + hmm_range->notifier_seq); + kvfree(hmm_range->hmm_pfns); + kfree(hmm_range); + + return r; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h index a292238f75eb..7f7d37a457c3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h @@ -30,6 +30,13 @@ #include <linux/workqueue.h> #include <linux/interval_tree.h> +int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier, + struct mm_struct *mm, struct page **pages, + uint64_t start, uint64_t npages, + struct hmm_range **phmm_range, bool readonly, + bool mmap_locked); +int amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range); + #if defined(CONFIG_HMM_MIRROR) int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr); void amdgpu_mn_unregister(struct amdgpu_bo *bo); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 0adffcace326..8714d50c5b20 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -491,7 +491,18 @@ bool amdgpu_bo_support_uswc(u64 bo_flags) #endif } -static int amdgpu_bo_do_create(struct amdgpu_device *adev, +/** + * amdgpu_bo_create - create an &amdgpu_bo buffer object + * @adev: amdgpu device object + * @bp: parameters to be used for the buffer object + * @bo_ptr: pointer to the buffer object pointer + * + * Creates an &amdgpu_bo buffer object. + * + * Returns: + * 0 for success or a negative error code on failure. + */ +int amdgpu_bo_create(struct amdgpu_device *adev, struct amdgpu_bo_param *bp, struct amdgpu_bo **bo_ptr) { @@ -601,9 +612,9 @@ fail_unreserve: return r; } -static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, - unsigned long size, - struct amdgpu_bo *bo) +int amdgpu_bo_create_shadow(struct amdgpu_device *adev, + unsigned long size, + struct amdgpu_bo *bo) { struct amdgpu_bo_param bp; int r; @@ -614,13 +625,12 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, memset(&bp, 0, sizeof(bp)); bp.size = size; bp.domain = AMDGPU_GEM_DOMAIN_GTT; - bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC | - AMDGPU_GEM_CREATE_SHADOW; + bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC; bp.type = ttm_bo_type_kernel; bp.resv = bo->tbo.base.resv; bp.bo_ptr_size = sizeof(struct amdgpu_bo); - r = amdgpu_bo_do_create(adev, &bp, &bo->shadow); + r = amdgpu_bo_create(adev, &bp, &bo->shadow); if (!r) { bo->shadow->parent = amdgpu_bo_ref(bo); mutex_lock(&adev->shadow_list_lock); @@ -632,50 +642,6 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, } /** - * amdgpu_bo_create - create an &amdgpu_bo buffer object - * @adev: amdgpu device object - * @bp: parameters to be used for the buffer object - * @bo_ptr: pointer to the buffer object pointer - * - * Creates an &amdgpu_bo buffer object; and if requested, also creates a - * shadow object. - * Shadow object is used to backup the original buffer object, and is always - * in GTT. - * - * Returns: - * 0 for success or a negative error code on failure. - */ -int amdgpu_bo_create(struct amdgpu_device *adev, - struct amdgpu_bo_param *bp, - struct amdgpu_bo **bo_ptr) -{ - u64 flags = bp->flags; - int r; - - bp->flags = bp->flags & ~AMDGPU_GEM_CREATE_SHADOW; - - r = amdgpu_bo_do_create(adev, bp, bo_ptr); - if (r) - return r; - - if ((flags & AMDGPU_GEM_CREATE_SHADOW) && !(adev->flags & AMD_IS_APU)) { - if (!bp->resv) - WARN_ON(dma_resv_lock((*bo_ptr)->tbo.base.resv, - NULL)); - - r = amdgpu_bo_create_shadow(adev, bp->size, *bo_ptr); - - if (!bp->resv) - dma_resv_unlock((*bo_ptr)->tbo.base.resv); - - if (r) - amdgpu_bo_unref(bo_ptr); - } - - return r; -} - -/** * amdgpu_bo_create_user - create an &amdgpu_bo_user buffer object * @adev: amdgpu device object * @bp: parameters to be used for the buffer object @@ -694,9 +660,8 @@ int amdgpu_bo_create_user(struct amdgpu_device *adev, struct amdgpu_bo *bo_ptr; int r; - bp->flags = bp->flags & ~AMDGPU_GEM_CREATE_SHADOW; bp->bo_ptr_size = sizeof(struct amdgpu_bo_user); - r = amdgpu_bo_do_create(adev, bp, &bo_ptr); + r = amdgpu_bo_create(adev, bp, &bo_ptr); if (r) return r; @@ -1595,7 +1560,6 @@ u64 amdgpu_bo_print_info(int id, struct amdgpu_bo *bo, struct seq_file *m) amdgpu_bo_print_flag(m, bo, NO_CPU_ACCESS); amdgpu_bo_print_flag(m, bo, CPU_GTT_USWC); amdgpu_bo_print_flag(m, bo, VRAM_CLEARED); - amdgpu_bo_print_flag(m, bo, SHADOW); amdgpu_bo_print_flag(m, bo, VRAM_CONTIGUOUS); amdgpu_bo_print_flag(m, bo, VM_ALWAYS_VALID); amdgpu_bo_print_flag(m, bo, EXPLICIT_SYNC); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index b37d36ac6b5a..46d22ab85492 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -37,6 +37,10 @@ #define AMDGPU_BO_INVALID_OFFSET LONG_MAX #define AMDGPU_BO_MAX_PLACEMENTS 3 +/* BO flag to indicate a KFD userptr BO */ +#define AMDGPU_AMDKFD_CREATE_USERPTR_BO (1ULL << 63) +#define AMDGPU_AMDKFD_CREATE_SVM_BO (1ULL << 62) + #define to_amdgpu_bo_user(abo) container_of((abo), struct amdgpu_bo_user, bo) struct amdgpu_bo_param { @@ -267,6 +271,9 @@ int amdgpu_bo_create_user(struct amdgpu_device *adev, struct amdgpu_bo_user **ubo_ptr); void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr, void **cpu_addr); +int amdgpu_bo_create_shadow(struct amdgpu_device *adev, + unsigned long size, + struct amdgpu_bo *bo); int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr); void *amdgpu_bo_kptr(struct amdgpu_bo *bo); void amdgpu_bo_kunmap(struct amdgpu_bo *bo); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index a09483beb968..f7bbb04d01ee 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -417,31 +417,12 @@ static int psp_tmr_init(struct psp_context *psp) return ret; } -static int psp_clear_vf_fw(struct psp_context *psp) -{ - int ret; - struct psp_gfx_cmd_resp *cmd; - - if (!amdgpu_sriov_vf(psp->adev) || psp->adev->asic_type != CHIP_NAVI12) - return 0; - - cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); - if (!cmd) - return -ENOMEM; - - cmd->cmd_id = GFX_CMD_ID_CLEAR_VF_FW; - - ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); - kfree(cmd); - - return ret; -} - static bool psp_skip_tmr(struct psp_context *psp) { switch (psp->adev->asic_type) { case CHIP_NAVI12: case CHIP_SIENNA_CICHLID: + case CHIP_ALDEBARAN: return true; default: return false; @@ -1037,6 +1018,13 @@ static int psp_ras_load(struct psp_context *psp) memset(psp->fw_pri_buf, 0, PSP_1_MEG); memcpy(psp->fw_pri_buf, psp->ta_ras_start_addr, psp->ta_ras_ucode_size); + ras_cmd = (struct ta_ras_shared_memory *)psp->ras.ras_shared_buf; + + if (psp->adev->gmc.xgmi.connected_to_cpu) + ras_cmd->ras_in_message.init_flags.poison_mode_en = 1; + else + ras_cmd->ras_in_message.init_flags.dgpu_mode = 1; + psp_prep_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr, psp->ta_ras_ucode_size, @@ -1046,8 +1034,6 @@ static int psp_ras_load(struct psp_context *psp) ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); - ras_cmd = (struct ta_ras_shared_memory *)psp->ras.ras_shared_buf; - if (!ret) { psp->ras.session_id = cmd->resp.session_id; @@ -1128,6 +1114,31 @@ int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id) return ret; } +static int psp_ras_status_to_errno(struct amdgpu_device *adev, + enum ta_ras_status ras_status) +{ + int ret = -EINVAL; + + switch (ras_status) { + case TA_RAS_STATUS__SUCCESS: + ret = 0; + break; + case TA_RAS_STATUS__RESET_NEEDED: + ret = -EAGAIN; + break; + case TA_RAS_STATUS__ERROR_RAS_NOT_AVAILABLE: + dev_warn(adev->dev, "RAS WARN: ras function unavailable\n"); + break; + case TA_RAS_STATUS__ERROR_ASD_READ_WRITE: + dev_warn(adev->dev, "RAS WARN: asd read or write failed\n"); + break; + default: + dev_err(adev->dev, "RAS ERROR: ras function failed ret 0x%X\n", ret); + } + + return ret; +} + int psp_ras_enable_features(struct psp_context *psp, union ta_ras_cmd_input *info, bool enable) { @@ -1151,7 +1162,7 @@ int psp_ras_enable_features(struct psp_context *psp, if (ret) return -EINVAL; - return ras_cmd->ras_status; + return psp_ras_status_to_errno(psp->adev, ras_cmd->ras_status); } static int psp_ras_terminate(struct psp_context *psp) @@ -1234,7 +1245,7 @@ int psp_ras_trigger_error(struct psp_context *psp, if (amdgpu_ras_intr_triggered()) return 0; - return ras_cmd->ras_status; + return psp_ras_status_to_errno(psp->adev, ras_cmd->ras_status); } // ras end @@ -1920,12 +1931,6 @@ static int psp_hw_start(struct psp_context *psp) return ret; } - ret = psp_clear_vf_fw(psp); - if (ret) { - DRM_ERROR("PSP clear vf fw!\n"); - return ret; - } - ret = psp_boot_config_set(adev); if (ret) { DRM_WARN("PSP set boot config@\n"); @@ -2166,7 +2171,7 @@ static int psp_load_smu_fw(struct psp_context *psp) return 0; if ((amdgpu_in_reset(adev) && - ras && ras->supported && + ras && adev->ras_enabled && (adev->asic_type == CHIP_ARCTURUS || adev->asic_type == CHIP_VEGA20)) || (adev->in_runpm && @@ -2434,7 +2439,6 @@ static int psp_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct psp_context *psp = &adev->psp; - int ret; if (psp->adev->psp.ta_fw) { psp_ras_terminate(psp); @@ -2445,11 +2449,6 @@ static int psp_hw_fini(void *handle) } psp_asd_unload(psp); - ret = psp_clear_vf_fw(psp); - if (ret) { - DRM_ERROR("PSP clear vf fw!\n"); - return ret; - } psp_tmr_terminate(psp); psp_ring_destroy(psp, PSP_RING_TYPE__KM); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index b0d2fc9454ca..b1c57a5b6e89 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -33,6 +33,7 @@ #include "amdgpu_atomfirmware.h" #include "amdgpu_xgmi.h" #include "ivsrcid/nbio/irqsrcs_nbif_7_4.h" +#include "atom.h" static const char *RAS_FS_NAME = "ras"; @@ -320,11 +321,14 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f, * "disable" requires only the block. * "enable" requires the block and error type. * "inject" requires the block, error type, address, and value. + * * The block is one of: umc, sdma, gfx, etc. * see ras_block_string[] for details + * * The error type is one of: ue, ce, where, * ue is multi-uncorrectable * ce is single-correctable + * * The sub-block is a the sub-block index, pass 0 if there is no sub-block. * The address and value are hexadecimal numbers, leading 0x is optional. * @@ -531,7 +535,7 @@ static struct ras_manager *amdgpu_ras_create_obj(struct amdgpu_device *adev, struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj; - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return NULL; if (head->block >= AMDGPU_RAS_BLOCK_COUNT) @@ -558,7 +562,7 @@ struct ras_manager *amdgpu_ras_find_obj(struct amdgpu_device *adev, struct ras_manager *obj; int i; - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return NULL; if (head) { @@ -585,36 +589,11 @@ struct ras_manager *amdgpu_ras_find_obj(struct amdgpu_device *adev, } /* obj end */ -static void amdgpu_ras_parse_status_code(struct amdgpu_device *adev, - const char* invoke_type, - const char* block_name, - enum ta_ras_status ret) -{ - switch (ret) { - case TA_RAS_STATUS__SUCCESS: - return; - case TA_RAS_STATUS__ERROR_RAS_NOT_AVAILABLE: - dev_warn(adev->dev, - "RAS WARN: %s %s currently unavailable\n", - invoke_type, - block_name); - break; - default: - dev_err(adev->dev, - "RAS ERROR: %s %s error failed ret 0x%X\n", - invoke_type, - block_name, - ret); - } -} - /* feature ctl begin */ static int amdgpu_ras_is_feature_allowed(struct amdgpu_device *adev, - struct ras_common_if *head) + struct ras_common_if *head) { - struct amdgpu_ras *con = amdgpu_ras_get_context(adev); - - return con->hw_supported & BIT(head->block); + return adev->ras_hw_enabled & BIT(head->block); } static int amdgpu_ras_is_feature_enabled(struct amdgpu_device *adev, @@ -658,11 +637,7 @@ static int __amdgpu_ras_feature_enable(struct amdgpu_device *adev, con->features |= BIT(head->block); } else { if (obj && amdgpu_ras_is_feature_enabled(adev, head)) { - /* skip clean gfx ras context feature for VEGA20 Gaming. - * will clean later - */ - if (!(!adev->ras_features && con->features & BIT(AMDGPU_RAS_BLOCK__GFX))) - con->features &= ~BIT(head->block); + con->features &= ~BIT(head->block); put_obj(obj); } } @@ -708,15 +683,10 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev, if (!amdgpu_ras_intr_triggered()) { ret = psp_ras_enable_features(&adev->psp, info, enable); if (ret) { - amdgpu_ras_parse_status_code(adev, - enable ? "enable":"disable", - ras_block_str(head->block), - (enum ta_ras_status)ret); - if (ret == TA_RAS_STATUS__RESET_NEEDED) - ret = -EAGAIN; - else - ret = -EINVAL; - + dev_err(adev->dev, "ras %s %s failed %d\n", + enable ? "enable":"disable", + ras_block_str(head->block), + ret); goto out; } } @@ -770,6 +740,10 @@ int amdgpu_ras_feature_enable_on_boot(struct amdgpu_device *adev, con->features |= BIT(head->block); ret = amdgpu_ras_feature_enable(adev, head, 0); + + /* clean gfx block ras features flag */ + if (adev->ras_enabled && head->block == AMDGPU_RAS_BLOCK__GFX) + con->features &= ~BIT(head->block); } } else ret = amdgpu_ras_feature_enable(adev, head, enable); @@ -890,6 +864,11 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, adev->gmc.xgmi.ras_funcs->query_ras_error_count) adev->gmc.xgmi.ras_funcs->query_ras_error_count(adev, &err_data); break; + case AMDGPU_RAS_BLOCK__HDP: + if (adev->hdp.ras_funcs && + adev->hdp.ras_funcs->query_ras_error_count) + adev->hdp.ras_funcs->query_ras_error_count(adev, &err_data); + break; default: break; } @@ -901,17 +880,42 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, info->ce_count = obj->err_data.ce_count; if (err_data.ce_count) { - dev_info(adev->dev, "%ld correctable hardware errors " + if (adev->smuio.funcs && + adev->smuio.funcs->get_socket_id && + adev->smuio.funcs->get_die_id) { + dev_info(adev->dev, "socket: %d, die: %d " + "%ld correctable hardware errors " + "detected in %s block, no user " + "action is needed.\n", + adev->smuio.funcs->get_socket_id(adev), + adev->smuio.funcs->get_die_id(adev), + obj->err_data.ce_count, + ras_block_str(info->head.block)); + } else { + dev_info(adev->dev, "%ld correctable hardware errors " "detected in %s block, no user " "action is needed.\n", obj->err_data.ce_count, ras_block_str(info->head.block)); + } } if (err_data.ue_count) { - dev_info(adev->dev, "%ld uncorrectable hardware errors " + if (adev->smuio.funcs && + adev->smuio.funcs->get_socket_id && + adev->smuio.funcs->get_die_id) { + dev_info(adev->dev, "socket: %d, die: %d " + "%ld uncorrectable hardware errors " + "detected in %s block\n", + adev->smuio.funcs->get_socket_id(adev), + adev->smuio.funcs->get_die_id(adev), + obj->err_data.ue_count, + ras_block_str(info->head.block)); + } else { + dev_info(adev->dev, "%ld uncorrectable hardware errors " "detected in %s block\n", obj->err_data.ue_count, ras_block_str(info->head.block)); + } } return 0; @@ -937,11 +941,20 @@ int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, if (adev->mmhub.ras_funcs && adev->mmhub.ras_funcs->reset_ras_error_count) adev->mmhub.ras_funcs->reset_ras_error_count(adev); + + if (adev->mmhub.ras_funcs && + adev->mmhub.ras_funcs->reset_ras_error_status) + adev->mmhub.ras_funcs->reset_ras_error_status(adev); break; case AMDGPU_RAS_BLOCK__SDMA: if (adev->sdma.funcs->reset_ras_error_count) adev->sdma.funcs->reset_ras_error_count(adev); break; + case AMDGPU_RAS_BLOCK__HDP: + if (adev->hdp.ras_funcs && + adev->hdp.ras_funcs->reset_ras_error_count) + adev->hdp.ras_funcs->reset_ras_error_count(adev); + break; default: break; } @@ -1022,10 +1035,9 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, ret = -EINVAL; } - amdgpu_ras_parse_status_code(adev, - "inject", - ras_block_str(info->head.block), - (enum ta_ras_status)ret); + if (ret) + dev_err(adev->dev, "ras inject %s failed %d\n", + ras_block_str(info->head.block), ret); return ret; } @@ -1038,7 +1050,7 @@ unsigned long amdgpu_ras_query_error_count(struct amdgpu_device *adev, struct ras_manager *obj; struct ras_err_data data = {0, 0}; - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return 0; list_for_each_entry(obj, &con->head, node) { @@ -1265,8 +1277,8 @@ static int amdgpu_ras_sysfs_remove_all(struct amdgpu_device *adev) static struct dentry *amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); - struct dentry *dir; - struct drm_minor *minor = adev_to_drm(adev)->primary; + struct drm_minor *minor = adev_to_drm(adev)->primary; + struct dentry *dir; dir = debugfs_create_dir(RAS_FS_NAME, minor->debugfs_root); debugfs_create_file("ras_ctrl", S_IWUGO | S_IRUGO, dir, adev, @@ -1275,6 +1287,8 @@ static struct dentry *amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device * &amdgpu_ras_debugfs_eeprom_ops); debugfs_create_u32("bad_page_cnt_threshold", 0444, dir, &con->bad_page_cnt_threshold); + debugfs_create_x32("ras_hw_enabled", 0444, dir, &adev->ras_hw_enabled); + debugfs_create_x32("ras_enabled", 0444, dir, &adev->ras_enabled); /* * After one uncorrectable error happens, usually GPU recovery will @@ -1561,7 +1575,7 @@ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev) struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj; - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return; list_for_each_entry(obj, &con->head, node) { @@ -1611,7 +1625,7 @@ static void amdgpu_ras_query_err_status(struct amdgpu_device *adev) struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj; - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return; list_for_each_entry(obj, &con->head, node) { @@ -1925,7 +1939,7 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev) bool exc_err_limit = false; int ret; - if (adev->ras_features && con) + if (adev->ras_enabled && con) data = &con->eh_data; else return 0; @@ -2029,6 +2043,23 @@ static bool amdgpu_ras_asic_supported(struct amdgpu_device *adev) } /* + * this is workaround for vega20 workstation sku, + * force enable gfx ras, ignore vbios gfx ras flag + * due to GC EDC can not write + */ +static void amdgpu_ras_get_quirks(struct amdgpu_device *adev) +{ + struct atom_context *ctx = adev->mode_info.atom_context; + + if (!ctx) + return; + + if (strnstr(ctx->vbios_version, "D16406", + sizeof(ctx->vbios_version))) + adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__GFX); +} + +/* * check hardware's ras ability which will be saved in hw_supported. * if hardware does not support ras, we can skip some ras initializtion and * forbid some ras operations from IP. @@ -2037,11 +2068,9 @@ static bool amdgpu_ras_asic_supported(struct amdgpu_device *adev) * we have to initialize ras as normal. but need check if operation is * allowed or not in each function. */ -static void amdgpu_ras_check_supported(struct amdgpu_device *adev, - uint32_t *hw_supported, uint32_t *supported) +static void amdgpu_ras_check_supported(struct amdgpu_device *adev) { - *hw_supported = 0; - *supported = 0; + adev->ras_hw_enabled = adev->ras_enabled = 0; if (amdgpu_sriov_vf(adev) || !adev->is_atom_fw || !amdgpu_ras_asic_supported(adev)) @@ -2050,33 +2079,34 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev, if (!adev->gmc.xgmi.connected_to_cpu) { if (amdgpu_atomfirmware_mem_ecc_supported(adev)) { dev_info(adev->dev, "MEM ECC is active.\n"); - *hw_supported |= (1 << AMDGPU_RAS_BLOCK__UMC | - 1 << AMDGPU_RAS_BLOCK__DF); + adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__UMC | + 1 << AMDGPU_RAS_BLOCK__DF); } else { dev_info(adev->dev, "MEM ECC is not presented.\n"); } if (amdgpu_atomfirmware_sram_ecc_supported(adev)) { dev_info(adev->dev, "SRAM ECC is active.\n"); - *hw_supported |= ~(1 << AMDGPU_RAS_BLOCK__UMC | - 1 << AMDGPU_RAS_BLOCK__DF); + adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC | + 1 << AMDGPU_RAS_BLOCK__DF); } else { dev_info(adev->dev, "SRAM ECC is not presented.\n"); } } else { /* driver only manages a few IP blocks RAS feature * when GPU is connected cpu through XGMI */ - *hw_supported |= (1 << AMDGPU_RAS_BLOCK__GFX | - 1 << AMDGPU_RAS_BLOCK__SDMA | - 1 << AMDGPU_RAS_BLOCK__MMHUB); + adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__GFX | + 1 << AMDGPU_RAS_BLOCK__SDMA | + 1 << AMDGPU_RAS_BLOCK__MMHUB); } + amdgpu_ras_get_quirks(adev); + /* hw_supported needs to be aligned with RAS block mask. */ - *hw_supported &= AMDGPU_RAS_BLOCK_MASK; + adev->ras_hw_enabled &= AMDGPU_RAS_BLOCK_MASK; - *supported = amdgpu_ras_enable == 0 ? - 0 : *hw_supported & amdgpu_ras_mask; - adev->ras_features = *supported; + adev->ras_enabled = amdgpu_ras_enable == 0 ? 0 : + adev->ras_hw_enabled & amdgpu_ras_mask; } int amdgpu_ras_init(struct amdgpu_device *adev) @@ -2097,13 +2127,13 @@ int amdgpu_ras_init(struct amdgpu_device *adev) amdgpu_ras_set_context(adev, con); - amdgpu_ras_check_supported(adev, &con->hw_supported, - &con->supported); - if (!con->hw_supported || (adev->asic_type == CHIP_VEGA10)) { + amdgpu_ras_check_supported(adev); + + if (!adev->ras_enabled || adev->asic_type == CHIP_VEGA10) { /* set gfx block ras context feature for VEGA20 Gaming * send ras disable cmd to ras ta during ras late init. */ - if (!adev->ras_features && adev->asic_type == CHIP_VEGA20) { + if (!adev->ras_enabled && adev->asic_type == CHIP_VEGA20) { con->features |= BIT(AMDGPU_RAS_BLOCK__GFX); return 0; @@ -2153,8 +2183,9 @@ int amdgpu_ras_init(struct amdgpu_device *adev) } dev_info(adev->dev, "RAS INFO: ras initialized successfully, " - "hardware ability[%x] ras_mask[%x]\n", - con->hw_supported, con->supported); + "hardware ability[%x] ras_mask[%x]\n", + adev->ras_hw_enabled, adev->ras_enabled); + return 0; release_con: amdgpu_ras_set_context(adev, NULL); @@ -2268,7 +2299,7 @@ void amdgpu_ras_resume(struct amdgpu_device *adev) struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj, *tmp; - if (!adev->ras_features || !con) { + if (!adev->ras_enabled || !con) { /* clean ras context for VEGA20 Gaming after send ras disable cmd */ amdgpu_release_ras_context(adev); @@ -2314,7 +2345,7 @@ void amdgpu_ras_suspend(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return; amdgpu_ras_disable_all_features(adev, 0); @@ -2328,7 +2359,7 @@ int amdgpu_ras_pre_fini(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return 0; /* Need disable ras on all IPs here before ip [hw/sw]fini */ @@ -2341,7 +2372,7 @@ int amdgpu_ras_fini(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); - if (!adev->ras_features || !con) + if (!adev->ras_enabled || !con) return 0; amdgpu_ras_fs_fini(adev); @@ -2360,10 +2391,8 @@ int amdgpu_ras_fini(struct amdgpu_device *adev) void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev) { - uint32_t hw_supported, supported; - - amdgpu_ras_check_supported(adev, &hw_supported, &supported); - if (!hw_supported) + amdgpu_ras_check_supported(adev); + if (!adev->ras_hw_enabled) return; if (atomic_cmpxchg(&amdgpu_ras_in_intr, 0, 1) == 0) { @@ -2392,7 +2421,7 @@ void amdgpu_release_ras_context(struct amdgpu_device *adev) if (!con) return; - if (!adev->ras_features && con->features & BIT(AMDGPU_RAS_BLOCK__GFX)) { + if (!adev->ras_enabled && con->features & BIT(AMDGPU_RAS_BLOCK__GFX)) { con->features &= ~BIT(AMDGPU_RAS_BLOCK__GFX); amdgpu_ras_set_context(adev, NULL); kfree(con); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 60df268a0c66..201fbdee1d09 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -313,9 +313,6 @@ struct ras_common_if { struct amdgpu_ras { /* ras infrastructure */ /* for ras itself. */ - uint32_t hw_supported; - /* for IP to check its ras ability. */ - uint32_t supported; uint32_t features; struct list_head head; /* sysfs */ @@ -478,7 +475,7 @@ static inline int amdgpu_ras_is_supported(struct amdgpu_device *adev, if (block >= AMDGPU_RAS_BLOCK_COUNT) return 0; - return ras && (ras->supported & (1 << block)); + return ras && (adev->ras_enabled & (1 << block)); } int amdgpu_ras_recovery_init(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 688624ebe421..7b634a1517f9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -158,6 +158,7 @@ void amdgpu_ring_undo(struct amdgpu_ring *ring) * @irq_src: interrupt source to use for this ring * @irq_type: interrupt type to use for this ring * @hw_prio: ring priority (NORMAL/HIGH) + * @sched_score: optional score atomic shared with other schedulers * * Initialize the driver information for the selected ring (all asics). * Returns 0 on success, error on failure. diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h index b860ec913ac5..484bb3dcec47 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h @@ -29,6 +29,7 @@ struct amdgpu_smuio_funcs { void (*update_rom_clock_gating)(struct amdgpu_device *adev, bool enable); void (*get_clock_gating_state)(struct amdgpu_device *adev, u32 *flags); u32 (*get_die_id)(struct amdgpu_device *adev); + u32 (*get_socket_id)(struct amdgpu_device *adev); bool (*is_host_gpu_xgmi_supported)(struct amdgpu_device *adev); }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 8c7ec09eb1a4..10391fcff343 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -32,7 +32,6 @@ #include <linux/dma-mapping.h> #include <linux/iommu.h> -#include <linux/hmm.h> #include <linux/pagemap.h> #include <linux/sched/task.h> #include <linux/sched/mm.h> @@ -112,6 +111,20 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, } abo = ttm_to_amdgpu_bo(bo); + if (abo->flags & AMDGPU_AMDKFD_CREATE_SVM_BO) { + struct dma_fence *fence; + struct dma_resv *resv = &bo->base._resv; + + rcu_read_lock(); + fence = rcu_dereference(resv->fence_excl); + if (fence && !fence->ops->signaled) + dma_fence_enable_sw_signaling(fence); + + placement->num_placement = 0; + placement->num_busy_placement = 0; + rcu_read_unlock(); + return; + } switch (bo->mem.mem_type) { case AMDGPU_PL_GDS: case AMDGPU_PL_GWS: @@ -165,13 +178,6 @@ static int amdgpu_verify_access(struct ttm_buffer_object *bo, struct file *filp) { struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo); - /* - * Don't verify access for KFD BOs. They don't have a GEM - * object associated with them. - */ - if (abo->kfd_bo) - return 0; - if (amdgpu_ttm_tt_get_usermm(bo->ttm)) return -EPERM; return drm_vma_node_verify_access(&abo->tbo.base.vma_node, @@ -288,7 +294,7 @@ error_free: } /** - * amdgpu_copy_ttm_mem_to_mem - Helper function for copy + * amdgpu_ttm_copy_mem_to_mem - Helper function for copy * @adev: amdgpu device * @src: buffer/address where to read from * @dst: buffer/address where to write to @@ -670,10 +676,8 @@ int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages) struct amdgpu_ttm_tt *gtt = (void *)ttm; unsigned long start = gtt->userptr; struct vm_area_struct *vma; - struct hmm_range *range; - unsigned long timeout; struct mm_struct *mm; - unsigned long i; + bool readonly; int r = 0; mm = bo->notifier.mm; @@ -689,76 +693,26 @@ int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages) if (!mmget_not_zero(mm)) /* Happens during process shutdown */ return -ESRCH; - range = kzalloc(sizeof(*range), GFP_KERNEL); - if (unlikely(!range)) { - r = -ENOMEM; - goto out; - } - range->notifier = &bo->notifier; - range->start = bo->notifier.interval_tree.start; - range->end = bo->notifier.interval_tree.last + 1; - range->default_flags = HMM_PFN_REQ_FAULT; - if (!amdgpu_ttm_tt_is_readonly(ttm)) - range->default_flags |= HMM_PFN_REQ_WRITE; - - range->hmm_pfns = kvmalloc_array(ttm->num_pages, - sizeof(*range->hmm_pfns), GFP_KERNEL); - if (unlikely(!range->hmm_pfns)) { - r = -ENOMEM; - goto out_free_ranges; - } - mmap_read_lock(mm); vma = find_vma(mm, start); + mmap_read_unlock(mm); if (unlikely(!vma || start < vma->vm_start)) { r = -EFAULT; - goto out_unlock; + goto out_putmm; } if (unlikely((gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) && vma->vm_file)) { r = -EPERM; - goto out_unlock; - } - mmap_read_unlock(mm); - timeout = jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT); - -retry: - range->notifier_seq = mmu_interval_read_begin(&bo->notifier); - - mmap_read_lock(mm); - r = hmm_range_fault(range); - mmap_read_unlock(mm); - if (unlikely(r)) { - /* - * FIXME: This timeout should encompass the retry from - * mmu_interval_read_retry() as well. - */ - if (r == -EBUSY && !time_after(jiffies, timeout)) - goto retry; - goto out_free_pfns; + goto out_putmm; } - /* - * Due to default_flags, all pages are HMM_PFN_VALID or - * hmm_range_fault() fails. FIXME: The pages cannot be touched outside - * the notifier_lock, and mmu_interval_read_retry() must be done first. - */ - for (i = 0; i < ttm->num_pages; i++) - pages[i] = hmm_pfn_to_page(range->hmm_pfns[i]); - - gtt->range = range; + readonly = amdgpu_ttm_tt_is_readonly(ttm); + r = amdgpu_hmm_range_get_pages(&bo->notifier, mm, pages, start, + ttm->num_pages, >t->range, readonly, + false); +out_putmm: mmput(mm); - return 0; - -out_unlock: - mmap_read_unlock(mm); -out_free_pfns: - kvfree(range->hmm_pfns); -out_free_ranges: - kfree(range); -out: - mmput(mm); return r; } @@ -787,10 +741,7 @@ bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm) * FIXME: Must always hold notifier_lock for this, and must * not ignore the return code. */ - r = mmu_interval_read_retry(gtt->range->notifier, - gtt->range->notifier_seq); - kvfree(gtt->range->hmm_pfns); - kfree(gtt->range); + r = amdgpu_hmm_range_get_pages_done(gtt->range); gtt->range = NULL; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c index 0c9c5255aa42..a57842689d42 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c @@ -50,9 +50,12 @@ void amdgpu_virt_init_setting(struct amdgpu_device *adev) struct drm_device *ddev = adev_to_drm(adev); /* enable virtual display */ - if (adev->mode_info.num_crtc == 0) - adev->mode_info.num_crtc = 1; - adev->enable_virtual_display = true; + if (adev->asic_type != CHIP_ALDEBARAN && + adev->asic_type != CHIP_ARCTURUS) { + if (adev->mode_info.num_crtc == 0) + adev->mode_info.num_crtc = 1; + adev->enable_virtual_display = true; + } ddev->driver_features &= ~DRIVER_ATOMIC; adev->cg_flags = 0; adev->pg_flags = 0; @@ -679,6 +682,7 @@ void amdgpu_detect_virtualization(struct amdgpu_device *adev) case CHIP_VEGA10: case CHIP_VEGA20: case CHIP_ARCTURUS: + case CHIP_ALDEBARAN: soc15_set_virt_ops(adev); break; case CHIP_NAVI10: diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 4a3e3f72e127..edc63d3e087e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -38,6 +38,7 @@ #include "amdgpu_gmc.h" #include "amdgpu_xgmi.h" #include "amdgpu_dma_buf.h" +#include "kfd_svm.h" /** * DOC: GPUVM @@ -850,35 +851,60 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, } /** - * amdgpu_vm_bo_param - fill in parameters for PD/PT allocation + * amdgpu_vm_pt_create - create bo for PD/PT * * @adev: amdgpu_device pointer * @vm: requesting vm * @level: the page table level * @immediate: use a immediate update - * @bp: resulting BO allocation parameters + * @bo: pointer to the buffer object pointer */ -static void amdgpu_vm_bo_param(struct amdgpu_device *adev, struct amdgpu_vm *vm, +static int amdgpu_vm_pt_create(struct amdgpu_device *adev, + struct amdgpu_vm *vm, int level, bool immediate, - struct amdgpu_bo_param *bp) + struct amdgpu_bo **bo) { - memset(bp, 0, sizeof(*bp)); + struct amdgpu_bo_param bp; + int r; - bp->size = amdgpu_vm_bo_size(adev, level); - bp->byte_align = AMDGPU_GPU_PAGE_SIZE; - bp->domain = AMDGPU_GEM_DOMAIN_VRAM; - bp->domain = amdgpu_bo_get_preferred_pin_domain(adev, bp->domain); - bp->flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | + memset(&bp, 0, sizeof(bp)); + + bp.size = amdgpu_vm_bo_size(adev, level); + bp.byte_align = AMDGPU_GPU_PAGE_SIZE; + bp.domain = AMDGPU_GEM_DOMAIN_VRAM; + bp.domain = amdgpu_bo_get_preferred_pin_domain(adev, bp.domain); + bp.flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | AMDGPU_GEM_CREATE_CPU_GTT_USWC; - bp->bo_ptr_size = sizeof(struct amdgpu_bo); + bp.bo_ptr_size = sizeof(struct amdgpu_bo); if (vm->use_cpu_for_update) - bp->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; - else if (!vm->root.base.bo || vm->root.base.bo->shadow) - bp->flags |= AMDGPU_GEM_CREATE_SHADOW; - bp->type = ttm_bo_type_kernel; - bp->no_wait_gpu = immediate; + bp.flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; + + bp.type = ttm_bo_type_kernel; + bp.no_wait_gpu = immediate; if (vm->root.base.bo) - bp->resv = vm->root.base.bo->tbo.base.resv; + bp.resv = vm->root.base.bo->tbo.base.resv; + + r = amdgpu_bo_create(adev, &bp, bo); + if (r) + return r; + + if (vm->is_compute_context && (adev->flags & AMD_IS_APU)) + return 0; + + if (!bp.resv) + WARN_ON(dma_resv_lock((*bo)->tbo.base.resv, + NULL)); + r = amdgpu_bo_create_shadow(adev, bp.size, *bo); + + if (!bp.resv) + dma_resv_unlock((*bo)->tbo.base.resv); + + if (r) { + amdgpu_bo_unref(bo); + return r; + } + + return 0; } /** @@ -901,7 +927,6 @@ static int amdgpu_vm_alloc_pts(struct amdgpu_device *adev, bool immediate) { struct amdgpu_vm_pt *entry = cursor->entry; - struct amdgpu_bo_param bp; struct amdgpu_bo *pt; int r; @@ -919,9 +944,7 @@ static int amdgpu_vm_alloc_pts(struct amdgpu_device *adev, if (entry->base.bo) return 0; - amdgpu_vm_bo_param(adev, vm, cursor->level, immediate, &bp); - - r = amdgpu_bo_create(adev, &bp, &pt); + r = amdgpu_vm_pt_create(adev, vm, cursor->level, immediate, &pt); if (r) return r; @@ -1593,15 +1616,15 @@ static int amdgpu_vm_update_ptes(struct amdgpu_vm_update_params *params, * Returns: * 0 for success, -EINVAL for failure. */ -static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, - struct amdgpu_device *bo_adev, - struct amdgpu_vm *vm, bool immediate, - bool unlocked, struct dma_resv *resv, - uint64_t start, uint64_t last, - uint64_t flags, uint64_t offset, - struct drm_mm_node *nodes, - dma_addr_t *pages_addr, - struct dma_fence **fence) +int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, + struct amdgpu_device *bo_adev, + struct amdgpu_vm *vm, bool immediate, + bool unlocked, struct dma_resv *resv, + uint64_t start, uint64_t last, + uint64_t flags, uint64_t offset, + struct drm_mm_node *nodes, + dma_addr_t *pages_addr, + struct dma_fence **fence) { struct amdgpu_vm_update_params params; enum amdgpu_sync_mode sync_mode; @@ -2818,7 +2841,6 @@ long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout) * * @adev: amdgpu_device pointer * @vm: requested vm - * @vm_context: Indicates if it GFX or Compute context * @pasid: Process address space identifier * * Init @vm fields. @@ -2826,10 +2848,8 @@ long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout) * Returns: * 0 for success, error for failure. */ -int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, - int vm_context, u32 pasid) +int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, u32 pasid) { - struct amdgpu_bo_param bp; struct amdgpu_bo *root; int r, i; @@ -2861,16 +2881,9 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, vm->pte_support_ats = false; vm->is_compute_context = false; - if (vm_context == AMDGPU_VM_CONTEXT_COMPUTE) { - vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode & - AMDGPU_VM_USE_CPU_FOR_COMPUTE); + vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode & + AMDGPU_VM_USE_CPU_FOR_GFX); - if (adev->asic_type == CHIP_RAVEN) - vm->pte_support_ats = true; - } else { - vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode & - AMDGPU_VM_USE_CPU_FOR_GFX); - } DRM_DEBUG_DRIVER("VM update mode is %s\n", vm->use_cpu_for_update ? "CPU" : "SDMA"); WARN_ONCE((vm->use_cpu_for_update && @@ -2887,10 +2900,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, mutex_init(&vm->eviction_lock); vm->evicting = false; - amdgpu_vm_bo_param(adev, vm, adev->vm_manager.root_level, false, &bp); - if (vm_context == AMDGPU_VM_CONTEXT_COMPUTE) - bp.flags &= ~AMDGPU_GEM_CREATE_SHADOW; - r = amdgpu_bo_create(adev, &bp, &root); + r = amdgpu_vm_pt_create(adev, vm, adev->vm_manager.root_level, + false, &root); if (r) goto error_free_delayed; @@ -3349,6 +3360,7 @@ void amdgpu_vm_set_task_info(struct amdgpu_vm *vm) bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid, uint64_t addr) { + bool is_compute_context = false; struct amdgpu_bo *root; uint64_t value, flags; struct amdgpu_vm *vm; @@ -3356,15 +3368,25 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid, spin_lock(&adev->vm_manager.pasid_lock); vm = idr_find(&adev->vm_manager.pasid_idr, pasid); - if (vm) + if (vm) { root = amdgpu_bo_ref(vm->root.base.bo); - else + is_compute_context = vm->is_compute_context; + } else { root = NULL; + } spin_unlock(&adev->vm_manager.pasid_lock); if (!root) return false; + addr /= AMDGPU_GPU_PAGE_SIZE; + + if (is_compute_context && + !svm_range_restore_pages(adev, pasid, addr)) { + amdgpu_bo_unref(&root); + return true; + } + r = amdgpu_bo_reserve(root, true); if (r) goto error_unref; @@ -3378,18 +3400,16 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid, if (!vm) goto error_unlock; - addr /= AMDGPU_GPU_PAGE_SIZE; flags = AMDGPU_PTE_VALID | AMDGPU_PTE_SNOOPED | AMDGPU_PTE_SYSTEM; - if (vm->is_compute_context) { + if (is_compute_context) { /* Intentionally setting invalid PTE flag * combination to force a no-retry-fault */ flags = AMDGPU_PTE_EXECUTABLE | AMDGPU_PDE_PTE | AMDGPU_PTE_TF; value = 0; - } else if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_NEVER) { /* Redirect the access to the dummy page */ value = adev->dummy_page_addr; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 6fd7dad0540a..ea60ec122b51 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -121,9 +121,6 @@ struct amdgpu_bo_list_entry; /* max vmids dedicated for process */ #define AMDGPU_VM_MAX_RESERVED_VMID 1 -#define AMDGPU_VM_CONTEXT_GFX 0 -#define AMDGPU_VM_CONTEXT_COMPUTE 1 - /* See vm_update_mode */ #define AMDGPU_VM_USE_CPU_FOR_GFX (1 << 0) #define AMDGPU_VM_USE_CPU_FOR_COMPUTE (1 << 1) @@ -367,6 +364,8 @@ struct amdgpu_vm_manager { spinlock_t pasid_lock; }; +struct amdgpu_bo_va_mapping; + #define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count))) #define amdgpu_vm_write_pte(adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr))) #define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags))) @@ -378,8 +377,7 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev); void amdgpu_vm_manager_fini(struct amdgpu_device *adev); long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout); -int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, - int vm_context, u32 pasid); +int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, u32 pasid); int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm, u32 pasid); void amdgpu_vm_release_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm); @@ -398,6 +396,15 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, struct dma_fence **fence); int amdgpu_vm_handle_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm); +int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, + struct amdgpu_device *bo_adev, + struct amdgpu_vm *vm, bool immediate, + bool unlocked, struct dma_resv *resv, + uint64_t start, uint64_t last, + uint64_t flags, uint64_t offset, + struct drm_mm_node *nodes, + dma_addr_t *pages_addr, + struct dma_fence **fence); int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, bool clear); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c index f7235438535f..f78d21910e07 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c @@ -29,12 +29,14 @@ #include "amdgpu_atomfirmware.h" #include "atom.h" -static inline struct amdgpu_vram_mgr *to_vram_mgr(struct ttm_resource_manager *man) +static inline struct amdgpu_vram_mgr * +to_vram_mgr(struct ttm_resource_manager *man) { return container_of(man, struct amdgpu_vram_mgr, manager); } -static inline struct amdgpu_device *to_amdgpu_device(struct amdgpu_vram_mgr *mgr) +static inline struct amdgpu_device * +to_amdgpu_device(struct amdgpu_vram_mgr *mgr) { return container_of(mgr, struct amdgpu_device, mman.vram_mgr); } @@ -82,12 +84,14 @@ static ssize_t amdgpu_mem_info_vis_vram_total_show(struct device *dev, * amount of currently used VRAM in bytes */ static ssize_t amdgpu_mem_info_vram_used_show(struct device *dev, - struct device_attribute *attr, char *buf) + struct device_attribute *attr, + char *buf) { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); + struct ttm_resource_manager *man; + man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); return sysfs_emit(buf, "%llu\n", amdgpu_vram_mgr_usage(man)); } @@ -100,18 +104,28 @@ static ssize_t amdgpu_mem_info_vram_used_show(struct device *dev, * amount of currently used visible VRAM in bytes */ static ssize_t amdgpu_mem_info_vis_vram_used_show(struct device *dev, - struct device_attribute *attr, char *buf) + struct device_attribute *attr, + char *buf) { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); - struct ttm_resource_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); + struct ttm_resource_manager *man; + man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); return sysfs_emit(buf, "%llu\n", amdgpu_vram_mgr_vis_usage(man)); } +/** + * DOC: mem_info_vram_vendor + * + * The amdgpu driver provides a sysfs API for reporting the vendor of the + * installed VRAM + * The file mem_info_vram_vendor is used for this and returns the name of the + * vendor. + */ static ssize_t amdgpu_mem_info_vram_vendor(struct device *dev, - struct device_attribute *attr, - char *buf) + struct device_attribute *attr, + char *buf) { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(ddev); @@ -162,78 +176,6 @@ static const struct attribute *amdgpu_vram_mgr_attributes[] = { NULL }; -static const struct ttm_resource_manager_func amdgpu_vram_mgr_func; - -/** - * amdgpu_vram_mgr_init - init VRAM manager and DRM MM - * - * @adev: amdgpu_device pointer - * - * Allocate and initialize the VRAM manager. - */ -int amdgpu_vram_mgr_init(struct amdgpu_device *adev) -{ - struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr; - struct ttm_resource_manager *man = &mgr->manager; - int ret; - - ttm_resource_manager_init(man, adev->gmc.real_vram_size >> PAGE_SHIFT); - - man->func = &amdgpu_vram_mgr_func; - - drm_mm_init(&mgr->mm, 0, man->size); - spin_lock_init(&mgr->lock); - INIT_LIST_HEAD(&mgr->reservations_pending); - INIT_LIST_HEAD(&mgr->reserved_pages); - - /* Add the two VRAM-related sysfs files */ - ret = sysfs_create_files(&adev->dev->kobj, amdgpu_vram_mgr_attributes); - if (ret) - DRM_ERROR("Failed to register sysfs\n"); - - ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, &mgr->manager); - ttm_resource_manager_set_used(man, true); - return 0; -} - -/** - * amdgpu_vram_mgr_fini - free and destroy VRAM manager - * - * @adev: amdgpu_device pointer - * - * Destroy and free the VRAM manager, returns -EBUSY if ranges are still - * allocated inside it. - */ -void amdgpu_vram_mgr_fini(struct amdgpu_device *adev) -{ - struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr; - struct ttm_resource_manager *man = &mgr->manager; - int ret; - struct amdgpu_vram_reservation *rsv, *temp; - - ttm_resource_manager_set_used(man, false); - - ret = ttm_resource_manager_evict_all(&adev->mman.bdev, man); - if (ret) - return; - - spin_lock(&mgr->lock); - list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, node) - kfree(rsv); - - list_for_each_entry_safe(rsv, temp, &mgr->reserved_pages, node) { - drm_mm_remove_node(&rsv->mm_node); - kfree(rsv); - } - drm_mm_takedown(&mgr->mm); - spin_unlock(&mgr->lock); - - sysfs_remove_files(&adev->dev->kobj, amdgpu_vram_mgr_attributes); - - ttm_resource_manager_cleanup(man); - ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, NULL); -} - /** * amdgpu_vram_mgr_vis_size - Calculate visible node size * @@ -283,6 +225,7 @@ u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo) return usage; } +/* Commit the reservation of VRAM pages */ static void amdgpu_vram_mgr_do_reserve(struct ttm_resource_manager *man) { struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); @@ -415,13 +358,13 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, const struct ttm_place *place, struct ttm_resource *mem) { + unsigned long lpfn, num_nodes, pages_per_node, pages_left, pages; struct amdgpu_vram_mgr *mgr = to_vram_mgr(man); struct amdgpu_device *adev = to_amdgpu_device(mgr); + uint64_t vis_usage = 0, mem_bytes, max_bytes; struct drm_mm *mm = &mgr->mm; - struct drm_mm_node *nodes; enum drm_mm_insert_mode mode; - unsigned long lpfn, num_nodes, pages_per_node, pages_left; - uint64_t vis_usage = 0, mem_bytes, max_bytes; + struct drm_mm_node *nodes; unsigned i; int r; @@ -448,10 +391,10 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, pages_per_node = HPAGE_PMD_NR; #else /* default to 2MB */ - pages_per_node = (2UL << (20UL - PAGE_SHIFT)); + pages_per_node = 2UL << (20UL - PAGE_SHIFT); #endif - pages_per_node = max((uint32_t)pages_per_node, - tbo->page_alignment); + pages_per_node = max_t(uint32_t, pages_per_node, + tbo->page_alignment); num_nodes = DIV_ROUND_UP(mem->num_pages, pages_per_node); } @@ -469,42 +412,37 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, mem->start = 0; pages_left = mem->num_pages; - spin_lock(&mgr->lock); - for (i = 0; pages_left >= pages_per_node; ++i) { - unsigned long pages = rounddown_pow_of_two(pages_left); + /* Limit maximum size to 2GB due to SG table limitations */ + pages = min(pages_left, 2UL << (30 - PAGE_SHIFT)); - /* Limit maximum size to 2GB due to SG table limitations */ - pages = min(pages, (2UL << (30 - PAGE_SHIFT))); - - r = drm_mm_insert_node_in_range(mm, &nodes[i], pages, - pages_per_node, 0, - place->fpfn, lpfn, - mode); - if (unlikely(r)) - break; - - vis_usage += amdgpu_vram_mgr_vis_size(adev, &nodes[i]); - amdgpu_vram_mgr_virt_start(mem, &nodes[i]); - pages_left -= pages; - } - - for (; pages_left; ++i) { - unsigned long pages = min(pages_left, pages_per_node); + i = 0; + spin_lock(&mgr->lock); + while (pages_left) { uint32_t alignment = tbo->page_alignment; - if (pages == pages_per_node) + if (pages >= pages_per_node) alignment = pages_per_node; - r = drm_mm_insert_node_in_range(mm, &nodes[i], - pages, alignment, 0, - place->fpfn, lpfn, - mode); - if (unlikely(r)) + r = drm_mm_insert_node_in_range(mm, &nodes[i], pages, alignment, + 0, place->fpfn, lpfn, mode); + if (unlikely(r)) { + if (pages > pages_per_node) { + if (is_power_of_2(pages)) + pages = pages / 2; + else + pages = rounddown_pow_of_two(pages); + continue; + } goto error; + } vis_usage += amdgpu_vram_mgr_vis_size(adev, &nodes[i]); amdgpu_vram_mgr_virt_start(mem, &nodes[i]); pages_left -= pages; + ++i; + + if (pages > pages_left) + pages = pages_left; } spin_unlock(&mgr->lock); @@ -728,3 +666,73 @@ static const struct ttm_resource_manager_func amdgpu_vram_mgr_func = { .free = amdgpu_vram_mgr_del, .debug = amdgpu_vram_mgr_debug }; + +/** + * amdgpu_vram_mgr_init - init VRAM manager and DRM MM + * + * @adev: amdgpu_device pointer + * + * Allocate and initialize the VRAM manager. + */ +int amdgpu_vram_mgr_init(struct amdgpu_device *adev) +{ + struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr; + struct ttm_resource_manager *man = &mgr->manager; + int ret; + + ttm_resource_manager_init(man, adev->gmc.real_vram_size >> PAGE_SHIFT); + + man->func = &amdgpu_vram_mgr_func; + + drm_mm_init(&mgr->mm, 0, man->size); + spin_lock_init(&mgr->lock); + INIT_LIST_HEAD(&mgr->reservations_pending); + INIT_LIST_HEAD(&mgr->reserved_pages); + + /* Add the two VRAM-related sysfs files */ + ret = sysfs_create_files(&adev->dev->kobj, amdgpu_vram_mgr_attributes); + if (ret) + DRM_ERROR("Failed to register sysfs\n"); + + ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, &mgr->manager); + ttm_resource_manager_set_used(man, true); + return 0; +} + +/** + * amdgpu_vram_mgr_fini - free and destroy VRAM manager + * + * @adev: amdgpu_device pointer + * + * Destroy and free the VRAM manager, returns -EBUSY if ranges are still + * allocated inside it. + */ +void amdgpu_vram_mgr_fini(struct amdgpu_device *adev) +{ + struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr; + struct ttm_resource_manager *man = &mgr->manager; + int ret; + struct amdgpu_vram_reservation *rsv, *temp; + + ttm_resource_manager_set_used(man, false); + + ret = ttm_resource_manager_evict_all(&adev->mman.bdev, man); + if (ret) + return; + + spin_lock(&mgr->lock); + list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, node) + kfree(rsv); + + list_for_each_entry_safe(rsv, temp, &mgr->reserved_pages, node) { + drm_mm_remove_node(&rsv->mm_node); + kfree(rsv); + } + drm_mm_takedown(&mgr->mm); + spin_unlock(&mgr->lock); + + sysfs_remove_files(&adev->dev->kobj, amdgpu_vram_mgr_attributes); + + ttm_resource_manager_cleanup(man); + ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, NULL); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h index 1a8f6d4baab2..befd0b4b7bea 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h @@ -98,9 +98,9 @@ union amd_sriov_msg_feature_flags { union amd_sriov_reg_access_flags { struct { - uint32_t vf_reg_access_ih : 1; - uint32_t vf_reg_access_mmhub : 1; - uint32_t vf_reg_access_gc : 1; + uint32_t vf_reg_psp_access_ih : 1; + uint32_t vf_reg_rlc_access_mmhub : 1; + uint32_t vf_reg_rlc_access_gc : 1; uint32_t reserved : 29; } flags; uint32_t all; diff --git a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c index 5c11144da051..33324427b555 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c @@ -421,6 +421,11 @@ static int dce_virtual_sw_init(void *handle) static int dce_virtual_sw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i = 0; + + for (i = 0; i < adev->mode_info.num_crtc; i++) + if (adev->mode_info.crtcs[i]) + hrtimer_cancel(&adev->mode_info.crtcs[i]->vblank_timer); kfree(adev->mode_info.bios_hardcoded_edid); @@ -480,13 +485,6 @@ static int dce_virtual_hw_init(void *handle) static int dce_virtual_hw_fini(void *handle) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - int i = 0; - - for (i = 0; i<adev->mode_info.num_crtc; i++) - if (adev->mode_info.crtcs[i]) - hrtimer_cancel(&adev->mode_info.crtcs[i]->vblank_timer); - return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c index 0d8459d63bac..36ba229576d8 100644 --- a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c +++ b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c @@ -219,11 +219,11 @@ static void df_v3_6_query_hashes(struct amdgpu_device *adev) adev->df.hash_status.hash_2m = false; adev->df.hash_status.hash_1g = false; - if (adev->asic_type != CHIP_ARCTURUS) - return; - - /* encoding for hash-enabled on Arcturus */ - if (adev->df.funcs->get_fb_channel_number(adev) == 0xe) { + /* encoding for hash-enabled on Arcturus and Aldebaran */ + if ((adev->asic_type == CHIP_ARCTURUS && + adev->df.funcs->get_fb_channel_number(adev) == 0xe) || + (adev->asic_type == CHIP_ALDEBARAN && + adev->df.funcs->get_fb_channel_number(adev) == 0x1e)) { tmp = RREG32_SOC15(DF, 0, mmDF_CS_UMC_AON0_DfGlobalCtrl); adev->df.hash_status.hash_64k = REG_GET_FIELD(tmp, DF_CS_UMC_AON0_DfGlobalCtrl, @@ -278,7 +278,12 @@ static u32 df_v3_6_get_fb_channel_number(struct amdgpu_device *adev) u32 tmp; tmp = RREG32_SOC15(DF, 0, mmDF_CS_UMC_AON0_DramBaseAddress0); - tmp &= DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan_MASK; + if (adev->asic_type == CHIP_ALDEBARAN) + tmp &= + ALDEBARAN_DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan_MASK; + else + tmp &= DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan_MASK; + tmp >>= DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan__SHIFT; return tmp; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index a078a38c2cee..22608c45f07c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -3937,7 +3937,8 @@ static void gfx_v9_0_init_tcp_config(struct amdgpu_device *adev) { u32 tmp; - if (adev->asic_type != CHIP_ARCTURUS) + if (adev->asic_type != CHIP_ARCTURUS && + adev->asic_type != CHIP_ALDEBARAN) return; tmp = RREG32_SOC15(GC, 0, mmTCP_ADDR_CONFIG); @@ -4559,8 +4560,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) if (!ring->sched.ready) return 0; - if (adev->asic_type == CHIP_ARCTURUS || - adev->asic_type == CHIP_ALDEBARAN) { + if (adev->asic_type == CHIP_ARCTURUS) { vgpr_init_shader_ptr = vgpr_init_compute_shader_arcturus; vgpr_init_shader_size = sizeof(vgpr_init_compute_shader_arcturus); vgpr_init_regs_ptr = vgpr_init_regs_arcturus; @@ -4745,7 +4745,11 @@ static int gfx_v9_0_ecc_late_init(void *handle) } /* requires IBs so do in late init after IB pool is initialized */ - r = gfx_v9_0_do_edc_gpr_workarounds(adev); + if (adev->asic_type == CHIP_ALDEBARAN) + r = gfx_v9_4_2_do_edc_gpr_workarounds(adev); + else + r = gfx_v9_0_do_edc_gpr_workarounds(adev); + if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c index a30c7c10cd9a..dbad9ef002d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c @@ -22,6 +22,7 @@ */ #include "amdgpu.h" #include "soc15.h" +#include "soc15d.h" #include "gc/gc_9_4_2_offset.h" #include "gc/gc_9_4_2_sh_mask.h" @@ -31,6 +32,11 @@ #include "amdgpu_ras.h" #include "amdgpu_gfx.h" +#define SE_ID_MAX 8 +#define CU_ID_MAX 16 +#define SIMD_ID_MAX 4 +#define WAVE_ID_MAX 10 + enum gfx_v9_4_2_utc_type { VML2_MEM, VML2_WALKER_MEM, @@ -79,6 +85,634 @@ static const struct soc15_reg_golden golden_settings_gc_9_4_2_alde[] = { SOC15_REG_GOLDEN_VALUE(GC, 0, regTCI_CNTL_3, 0xff, 0x20), }; +/** + * This shader is used to clear VGPRS and LDS, and also write the input + * pattern into the write back buffer, which will be used by driver to + * check whether all SIMDs have been covered. +*/ +static const u32 vgpr_init_compute_shader_aldebaran[] = { + 0xb8840904, 0xb8851a04, 0xb8861344, 0xb8831804, 0x9208ff06, 0x00000280, + 0x9209a805, 0x920a8a04, 0x81080908, 0x81080a08, 0x81080308, 0x8e078208, + 0x81078407, 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8a0000, 0xd3d94000, + 0x18000080, 0xd3d94001, 0x18000080, 0xd3d94002, 0x18000080, 0xd3d94003, + 0x18000080, 0xd3d94004, 0x18000080, 0xd3d94005, 0x18000080, 0xd3d94006, + 0x18000080, 0xd3d94007, 0x18000080, 0xd3d94008, 0x18000080, 0xd3d94009, + 0x18000080, 0xd3d9400a, 0x18000080, 0xd3d9400b, 0x18000080, 0xd3d9400c, + 0x18000080, 0xd3d9400d, 0x18000080, 0xd3d9400e, 0x18000080, 0xd3d9400f, + 0x18000080, 0xd3d94010, 0x18000080, 0xd3d94011, 0x18000080, 0xd3d94012, + 0x18000080, 0xd3d94013, 0x18000080, 0xd3d94014, 0x18000080, 0xd3d94015, + 0x18000080, 0xd3d94016, 0x18000080, 0xd3d94017, 0x18000080, 0xd3d94018, + 0x18000080, 0xd3d94019, 0x18000080, 0xd3d9401a, 0x18000080, 0xd3d9401b, + 0x18000080, 0xd3d9401c, 0x18000080, 0xd3d9401d, 0x18000080, 0xd3d9401e, + 0x18000080, 0xd3d9401f, 0x18000080, 0xd3d94020, 0x18000080, 0xd3d94021, + 0x18000080, 0xd3d94022, 0x18000080, 0xd3d94023, 0x18000080, 0xd3d94024, + 0x18000080, 0xd3d94025, 0x18000080, 0xd3d94026, 0x18000080, 0xd3d94027, + 0x18000080, 0xd3d94028, 0x18000080, 0xd3d94029, 0x18000080, 0xd3d9402a, + 0x18000080, 0xd3d9402b, 0x18000080, 0xd3d9402c, 0x18000080, 0xd3d9402d, + 0x18000080, 0xd3d9402e, 0x18000080, 0xd3d9402f, 0x18000080, 0xd3d94030, + 0x18000080, 0xd3d94031, 0x18000080, 0xd3d94032, 0x18000080, 0xd3d94033, + 0x18000080, 0xd3d94034, 0x18000080, 0xd3d94035, 0x18000080, 0xd3d94036, + 0x18000080, 0xd3d94037, 0x18000080, 0xd3d94038, 0x18000080, 0xd3d94039, + 0x18000080, 0xd3d9403a, 0x18000080, 0xd3d9403b, 0x18000080, 0xd3d9403c, + 0x18000080, 0xd3d9403d, 0x18000080, 0xd3d9403e, 0x18000080, 0xd3d9403f, + 0x18000080, 0xd3d94040, 0x18000080, 0xd3d94041, 0x18000080, 0xd3d94042, + 0x18000080, 0xd3d94043, 0x18000080, 0xd3d94044, 0x18000080, 0xd3d94045, + 0x18000080, 0xd3d94046, 0x18000080, 0xd3d94047, 0x18000080, 0xd3d94048, + 0x18000080, 0xd3d94049, 0x18000080, 0xd3d9404a, 0x18000080, 0xd3d9404b, + 0x18000080, 0xd3d9404c, 0x18000080, 0xd3d9404d, 0x18000080, 0xd3d9404e, + 0x18000080, 0xd3d9404f, 0x18000080, 0xd3d94050, 0x18000080, 0xd3d94051, + 0x18000080, 0xd3d94052, 0x18000080, 0xd3d94053, 0x18000080, 0xd3d94054, + 0x18000080, 0xd3d94055, 0x18000080, 0xd3d94056, 0x18000080, 0xd3d94057, + 0x18000080, 0xd3d94058, 0x18000080, 0xd3d94059, 0x18000080, 0xd3d9405a, + 0x18000080, 0xd3d9405b, 0x18000080, 0xd3d9405c, 0x18000080, 0xd3d9405d, + 0x18000080, 0xd3d9405e, 0x18000080, 0xd3d9405f, 0x18000080, 0xd3d94060, + 0x18000080, 0xd3d94061, 0x18000080, 0xd3d94062, 0x18000080, 0xd3d94063, + 0x18000080, 0xd3d94064, 0x18000080, 0xd3d94065, 0x18000080, 0xd3d94066, + 0x18000080, 0xd3d94067, 0x18000080, 0xd3d94068, 0x18000080, 0xd3d94069, + 0x18000080, 0xd3d9406a, 0x18000080, 0xd3d9406b, 0x18000080, 0xd3d9406c, + 0x18000080, 0xd3d9406d, 0x18000080, 0xd3d9406e, 0x18000080, 0xd3d9406f, + 0x18000080, 0xd3d94070, 0x18000080, 0xd3d94071, 0x18000080, 0xd3d94072, + 0x18000080, 0xd3d94073, 0x18000080, 0xd3d94074, 0x18000080, 0xd3d94075, + 0x18000080, 0xd3d94076, 0x18000080, 0xd3d94077, 0x18000080, 0xd3d94078, + 0x18000080, 0xd3d94079, 0x18000080, 0xd3d9407a, 0x18000080, 0xd3d9407b, + 0x18000080, 0xd3d9407c, 0x18000080, 0xd3d9407d, 0x18000080, 0xd3d9407e, + 0x18000080, 0xd3d9407f, 0x18000080, 0xd3d94080, 0x18000080, 0xd3d94081, + 0x18000080, 0xd3d94082, 0x18000080, 0xd3d94083, 0x18000080, 0xd3d94084, + 0x18000080, 0xd3d94085, 0x18000080, 0xd3d94086, 0x18000080, 0xd3d94087, + 0x18000080, 0xd3d94088, 0x18000080, 0xd3d94089, 0x18000080, 0xd3d9408a, + 0x18000080, 0xd3d9408b, 0x18000080, 0xd3d9408c, 0x18000080, 0xd3d9408d, + 0x18000080, 0xd3d9408e, 0x18000080, 0xd3d9408f, 0x18000080, 0xd3d94090, + 0x18000080, 0xd3d94091, 0x18000080, 0xd3d94092, 0x18000080, 0xd3d94093, + 0x18000080, 0xd3d94094, 0x18000080, 0xd3d94095, 0x18000080, 0xd3d94096, + 0x18000080, 0xd3d94097, 0x18000080, 0xd3d94098, 0x18000080, 0xd3d94099, + 0x18000080, 0xd3d9409a, 0x18000080, 0xd3d9409b, 0x18000080, 0xd3d9409c, + 0x18000080, 0xd3d9409d, 0x18000080, 0xd3d9409e, 0x18000080, 0xd3d9409f, + 0x18000080, 0xd3d940a0, 0x18000080, 0xd3d940a1, 0x18000080, 0xd3d940a2, + 0x18000080, 0xd3d940a3, 0x18000080, 0xd3d940a4, 0x18000080, 0xd3d940a5, + 0x18000080, 0xd3d940a6, 0x18000080, 0xd3d940a7, 0x18000080, 0xd3d940a8, + 0x18000080, 0xd3d940a9, 0x18000080, 0xd3d940aa, 0x18000080, 0xd3d940ab, + 0x18000080, 0xd3d940ac, 0x18000080, 0xd3d940ad, 0x18000080, 0xd3d940ae, + 0x18000080, 0xd3d940af, 0x18000080, 0xd3d940b0, 0x18000080, 0xd3d940b1, + 0x18000080, 0xd3d940b2, 0x18000080, 0xd3d940b3, 0x18000080, 0xd3d940b4, + 0x18000080, 0xd3d940b5, 0x18000080, 0xd3d940b6, 0x18000080, 0xd3d940b7, + 0x18000080, 0xd3d940b8, 0x18000080, 0xd3d940b9, 0x18000080, 0xd3d940ba, + 0x18000080, 0xd3d940bb, 0x18000080, 0xd3d940bc, 0x18000080, 0xd3d940bd, + 0x18000080, 0xd3d940be, 0x18000080, 0xd3d940bf, 0x18000080, 0xd3d940c0, + 0x18000080, 0xd3d940c1, 0x18000080, 0xd3d940c2, 0x18000080, 0xd3d940c3, + 0x18000080, 0xd3d940c4, 0x18000080, 0xd3d940c5, 0x18000080, 0xd3d940c6, + 0x18000080, 0xd3d940c7, 0x18000080, 0xd3d940c8, 0x18000080, 0xd3d940c9, + 0x18000080, 0xd3d940ca, 0x18000080, 0xd3d940cb, 0x18000080, 0xd3d940cc, + 0x18000080, 0xd3d940cd, 0x18000080, 0xd3d940ce, 0x18000080, 0xd3d940cf, + 0x18000080, 0xd3d940d0, 0x18000080, 0xd3d940d1, 0x18000080, 0xd3d940d2, + 0x18000080, 0xd3d940d3, 0x18000080, 0xd3d940d4, 0x18000080, 0xd3d940d5, + 0x18000080, 0xd3d940d6, 0x18000080, 0xd3d940d7, 0x18000080, 0xd3d940d8, + 0x18000080, 0xd3d940d9, 0x18000080, 0xd3d940da, 0x18000080, 0xd3d940db, + 0x18000080, 0xd3d940dc, 0x18000080, 0xd3d940dd, 0x18000080, 0xd3d940de, + 0x18000080, 0xd3d940df, 0x18000080, 0xd3d940e0, 0x18000080, 0xd3d940e1, + 0x18000080, 0xd3d940e2, 0x18000080, 0xd3d940e3, 0x18000080, 0xd3d940e4, + 0x18000080, 0xd3d940e5, 0x18000080, 0xd3d940e6, 0x18000080, 0xd3d940e7, + 0x18000080, 0xd3d940e8, 0x18000080, 0xd3d940e9, 0x18000080, 0xd3d940ea, + 0x18000080, 0xd3d940eb, 0x18000080, 0xd3d940ec, 0x18000080, 0xd3d940ed, + 0x18000080, 0xd3d940ee, 0x18000080, 0xd3d940ef, 0x18000080, 0xd3d940f0, + 0x18000080, 0xd3d940f1, 0x18000080, 0xd3d940f2, 0x18000080, 0xd3d940f3, + 0x18000080, 0xd3d940f4, 0x18000080, 0xd3d940f5, 0x18000080, 0xd3d940f6, + 0x18000080, 0xd3d940f7, 0x18000080, 0xd3d940f8, 0x18000080, 0xd3d940f9, + 0x18000080, 0xd3d940fa, 0x18000080, 0xd3d940fb, 0x18000080, 0xd3d940fc, + 0x18000080, 0xd3d940fd, 0x18000080, 0xd3d940fe, 0x18000080, 0xd3d940ff, + 0x18000080, 0xb07c0000, 0xbe8a00ff, 0x000000f8, 0xbf11080a, 0x7e000280, + 0x7e020280, 0x7e040280, 0x7e060280, 0x7e080280, 0x7e0a0280, 0x7e0c0280, + 0x7e0e0280, 0x808a880a, 0xbe80320a, 0xbf84fff5, 0xbf9c0000, 0xd28c0001, + 0x0001007f, 0xd28d0001, 0x0002027e, 0x10020288, 0xbe8b0004, 0xb78b4000, + 0xd1196a01, 0x00001701, 0xbe8a0087, 0xbefc00c1, 0xd89c4000, 0x00020201, + 0xd89cc080, 0x00040401, 0x320202ff, 0x00000800, 0x808a810a, 0xbf84fff8, + 0xbf810000, +}; + +const struct soc15_reg_entry vgpr_init_regs_aldebaran[] = { + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Y), 4 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Z), 1 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC1), 0xbf }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC2), 0x400006 }, /* 64KB LDS */ + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC3), 0x3F }, /* 63 - accum-offset = 256 */ + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE4), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE5), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE6), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE7), 0xffffffff }, +}; + +/** + * The below shaders are used to clear SGPRS, and also write the input + * pattern into the write back buffer. The first two dispatch should be + * scheduled simultaneously which make sure that all SGPRS could be + * allocated, so the dispatch 1 need check write back buffer before scheduled, + * make sure that waves of dispatch 0 are all dispacthed to all simds + * balanced. both dispatch 0 and dispatch 1 should be halted until all waves + * are dispatched, and then driver write a pattern to the shared memory to make + * all waves continue. +*/ +static const u32 sgpr112_init_compute_shader_aldebaran[] = { + 0xb8840904, 0xb8851a04, 0xb8861344, 0xb8831804, 0x9208ff06, 0x00000280, + 0x9209a805, 0x920a8a04, 0x81080908, 0x81080a08, 0x81080308, 0x8e078208, + 0x81078407, 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8e003f, 0xc0030200, + 0x00000000, 0xbf8c0000, 0xbf06ff08, 0xdeadbeaf, 0xbf84fff9, 0x81028102, + 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8a0000, 0xbefc0080, 0xbeea0080, + 0xbeeb0080, 0xbf00f280, 0xbee60080, 0xbee70080, 0xbee80080, 0xbee90080, + 0xbefe0080, 0xbeff0080, 0xbe880080, 0xbe890080, 0xbe8a0080, 0xbe8b0080, + 0xbe8c0080, 0xbe8d0080, 0xbe8e0080, 0xbe8f0080, 0xbe900080, 0xbe910080, + 0xbe920080, 0xbe930080, 0xbe940080, 0xbe950080, 0xbe960080, 0xbe970080, + 0xbe980080, 0xbe990080, 0xbe9a0080, 0xbe9b0080, 0xbe9c0080, 0xbe9d0080, + 0xbe9e0080, 0xbe9f0080, 0xbea00080, 0xbea10080, 0xbea20080, 0xbea30080, + 0xbea40080, 0xbea50080, 0xbea60080, 0xbea70080, 0xbea80080, 0xbea90080, + 0xbeaa0080, 0xbeab0080, 0xbeac0080, 0xbead0080, 0xbeae0080, 0xbeaf0080, + 0xbeb00080, 0xbeb10080, 0xbeb20080, 0xbeb30080, 0xbeb40080, 0xbeb50080, + 0xbeb60080, 0xbeb70080, 0xbeb80080, 0xbeb90080, 0xbeba0080, 0xbebb0080, + 0xbebc0080, 0xbebd0080, 0xbebe0080, 0xbebf0080, 0xbec00080, 0xbec10080, + 0xbec20080, 0xbec30080, 0xbec40080, 0xbec50080, 0xbec60080, 0xbec70080, + 0xbec80080, 0xbec90080, 0xbeca0080, 0xbecb0080, 0xbecc0080, 0xbecd0080, + 0xbece0080, 0xbecf0080, 0xbed00080, 0xbed10080, 0xbed20080, 0xbed30080, + 0xbed40080, 0xbed50080, 0xbed60080, 0xbed70080, 0xbed80080, 0xbed90080, + 0xbeda0080, 0xbedb0080, 0xbedc0080, 0xbedd0080, 0xbede0080, 0xbedf0080, + 0xbee00080, 0xbee10080, 0xbee20080, 0xbee30080, 0xbee40080, 0xbee50080, + 0xbf810000 +}; + +const struct soc15_reg_entry sgpr112_init_regs_aldebaran[] = { + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Y), 8 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Z), 1 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC1), 0x340 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC2), 0x6 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC3), 0x0 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE4), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE5), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE6), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE7), 0xffffffff }, +}; + +static const u32 sgpr96_init_compute_shader_aldebaran[] = { + 0xb8840904, 0xb8851a04, 0xb8861344, 0xb8831804, 0x9208ff06, 0x00000280, + 0x9209a805, 0x920a8a04, 0x81080908, 0x81080a08, 0x81080308, 0x8e078208, + 0x81078407, 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8e003f, 0xc0030200, + 0x00000000, 0xbf8c0000, 0xbf06ff08, 0xdeadbeaf, 0xbf84fff9, 0x81028102, + 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8a0000, 0xbefc0080, 0xbeea0080, + 0xbeeb0080, 0xbf00f280, 0xbee60080, 0xbee70080, 0xbee80080, 0xbee90080, + 0xbefe0080, 0xbeff0080, 0xbe880080, 0xbe890080, 0xbe8a0080, 0xbe8b0080, + 0xbe8c0080, 0xbe8d0080, 0xbe8e0080, 0xbe8f0080, 0xbe900080, 0xbe910080, + 0xbe920080, 0xbe930080, 0xbe940080, 0xbe950080, 0xbe960080, 0xbe970080, + 0xbe980080, 0xbe990080, 0xbe9a0080, 0xbe9b0080, 0xbe9c0080, 0xbe9d0080, + 0xbe9e0080, 0xbe9f0080, 0xbea00080, 0xbea10080, 0xbea20080, 0xbea30080, + 0xbea40080, 0xbea50080, 0xbea60080, 0xbea70080, 0xbea80080, 0xbea90080, + 0xbeaa0080, 0xbeab0080, 0xbeac0080, 0xbead0080, 0xbeae0080, 0xbeaf0080, + 0xbeb00080, 0xbeb10080, 0xbeb20080, 0xbeb30080, 0xbeb40080, 0xbeb50080, + 0xbeb60080, 0xbeb70080, 0xbeb80080, 0xbeb90080, 0xbeba0080, 0xbebb0080, + 0xbebc0080, 0xbebd0080, 0xbebe0080, 0xbebf0080, 0xbec00080, 0xbec10080, + 0xbec20080, 0xbec30080, 0xbec40080, 0xbec50080, 0xbec60080, 0xbec70080, + 0xbec80080, 0xbec90080, 0xbeca0080, 0xbecb0080, 0xbecc0080, 0xbecd0080, + 0xbece0080, 0xbecf0080, 0xbed00080, 0xbed10080, 0xbed20080, 0xbed30080, + 0xbed40080, 0xbed50080, 0xbed60080, 0xbed70080, 0xbed80080, 0xbed90080, + 0xbf810000, +}; + +const struct soc15_reg_entry sgpr96_init_regs_aldebaran[] = { + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Y), 0xc }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Z), 1 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC1), 0x2c0 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC2), 0x6 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC3), 0x0 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE4), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE5), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE6), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE7), 0xffffffff }, +}; + +/** + * This shader is used to clear the uninitiated sgprs after the above + * two dispatches, because of hardware feature, dispath 0 couldn't clear + * top hole sgprs. Therefore need 4 waves per SIMD to cover these sgprs +*/ +static const u32 sgpr64_init_compute_shader_aldebaran[] = { + 0xb8840904, 0xb8851a04, 0xb8861344, 0xb8831804, 0x9208ff06, 0x00000280, + 0x9209a805, 0x920a8a04, 0x81080908, 0x81080a08, 0x81080308, 0x8e078208, + 0x81078407, 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8e003f, 0xc0030200, + 0x00000000, 0xbf8c0000, 0xbf06ff08, 0xdeadbeaf, 0xbf84fff9, 0x81028102, + 0xc0410080, 0x00000007, 0xbf8c0000, 0xbf8a0000, 0xbefc0080, 0xbeea0080, + 0xbeeb0080, 0xbf00f280, 0xbee60080, 0xbee70080, 0xbee80080, 0xbee90080, + 0xbefe0080, 0xbeff0080, 0xbe880080, 0xbe890080, 0xbe8a0080, 0xbe8b0080, + 0xbe8c0080, 0xbe8d0080, 0xbe8e0080, 0xbe8f0080, 0xbe900080, 0xbe910080, + 0xbe920080, 0xbe930080, 0xbe940080, 0xbe950080, 0xbe960080, 0xbe970080, + 0xbe980080, 0xbe990080, 0xbe9a0080, 0xbe9b0080, 0xbe9c0080, 0xbe9d0080, + 0xbe9e0080, 0xbe9f0080, 0xbea00080, 0xbea10080, 0xbea20080, 0xbea30080, + 0xbea40080, 0xbea50080, 0xbea60080, 0xbea70080, 0xbea80080, 0xbea90080, + 0xbeaa0080, 0xbeab0080, 0xbeac0080, 0xbead0080, 0xbeae0080, 0xbeaf0080, + 0xbeb00080, 0xbeb10080, 0xbeb20080, 0xbeb30080, 0xbeb40080, 0xbeb50080, + 0xbeb60080, 0xbeb70080, 0xbeb80080, 0xbeb90080, 0xbf810000, +}; + +const struct soc15_reg_entry sgpr64_init_regs_aldebaran[] = { + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_X), 0x40 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Y), 0x10 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_NUM_THREAD_Z), 1 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC1), 0x1c0 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC2), 0x6 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_PGM_RSRC3), 0x0 }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE4), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE5), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE6), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, regCOMPUTE_STATIC_THREAD_MGMT_SE7), 0xffffffff }, +}; + +static int gfx_v9_4_2_run_shader(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_ib *ib, + const u32 *shader_ptr, u32 shader_size, + const struct soc15_reg_entry *init_regs, u32 regs_size, + u32 compute_dim_x, u64 wb_gpu_addr, u32 pattern, + struct dma_fence **fence_ptr) +{ + int r, i; + uint32_t total_size, shader_offset; + u64 gpu_addr; + + total_size = (regs_size * 3 + 4 + 5 + 5) * 4; + total_size = ALIGN(total_size, 256); + shader_offset = total_size; + total_size += ALIGN(shader_size, 256); + + /* allocate an indirect buffer to put the commands in */ + memset(ib, 0, sizeof(*ib)); + r = amdgpu_ib_get(adev, NULL, total_size, + AMDGPU_IB_POOL_DIRECT, ib); + if (r) { + dev_err(adev->dev, "failed to get ib (%d).\n", r); + return r; + } + + /* load the compute shaders */ + for (i = 0; i < shader_size/sizeof(u32); i++) + ib->ptr[i + (shader_offset / 4)] = shader_ptr[i]; + + /* init the ib length to 0 */ + ib->length_dw = 0; + + /* write the register state for the compute dispatch */ + for (i = 0; i < regs_size; i++) { + ib->ptr[ib->length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1); + ib->ptr[ib->length_dw++] = SOC15_REG_ENTRY_OFFSET(init_regs[i]) + - PACKET3_SET_SH_REG_START; + ib->ptr[ib->length_dw++] = init_regs[i].reg_value; + } + + /* write the shader start address: mmCOMPUTE_PGM_LO, mmCOMPUTE_PGM_HI */ + gpu_addr = (ib->gpu_addr + (u64)shader_offset) >> 8; + ib->ptr[ib->length_dw++] = PACKET3(PACKET3_SET_SH_REG, 2); + ib->ptr[ib->length_dw++] = SOC15_REG_OFFSET(GC, 0, regCOMPUTE_PGM_LO) + - PACKET3_SET_SH_REG_START; + ib->ptr[ib->length_dw++] = lower_32_bits(gpu_addr); + ib->ptr[ib->length_dw++] = upper_32_bits(gpu_addr); + + /* write the wb buffer address */ + ib->ptr[ib->length_dw++] = PACKET3(PACKET3_SET_SH_REG, 3); + ib->ptr[ib->length_dw++] = SOC15_REG_OFFSET(GC, 0, regCOMPUTE_USER_DATA_0) + - PACKET3_SET_SH_REG_START; + ib->ptr[ib->length_dw++] = lower_32_bits(wb_gpu_addr); + ib->ptr[ib->length_dw++] = upper_32_bits(wb_gpu_addr); + ib->ptr[ib->length_dw++] = pattern; + + /* write dispatch packet */ + ib->ptr[ib->length_dw++] = PACKET3(PACKET3_DISPATCH_DIRECT, 3); + ib->ptr[ib->length_dw++] = compute_dim_x; /* x */ + ib->ptr[ib->length_dw++] = 1; /* y */ + ib->ptr[ib->length_dw++] = 1; /* z */ + ib->ptr[ib->length_dw++] = + REG_SET_FIELD(0, COMPUTE_DISPATCH_INITIATOR, COMPUTE_SHADER_EN, 1); + + /* shedule the ib on the ring */ + r = amdgpu_ib_schedule(ring, 1, ib, NULL, fence_ptr); + if (r) { + dev_err(adev->dev, "ib submit failed (%d).\n", r); + amdgpu_ib_free(adev, ib, NULL); + } + return r; +} + +static void gfx_v9_4_2_log_wave_assignment(struct amdgpu_device *adev, uint32_t *wb_ptr) +{ + uint32_t se, cu, simd, wave; + uint32_t offset = 0; + char *str; + int size; + + str = kmalloc(256, GFP_KERNEL); + if (!str) + return; + + dev_dbg(adev->dev, "wave assignment:\n"); + + for (se = 0; se < adev->gfx.config.max_shader_engines; se++) { + for (cu = 0; cu < CU_ID_MAX; cu++) { + memset(str, 0, 256); + size = sprintf(str, "SE[%02d]CU[%02d]: ", se, cu); + for (simd = 0; simd < SIMD_ID_MAX; simd++) { + size += sprintf(str + size, "["); + for (wave = 0; wave < WAVE_ID_MAX; wave++) { + size += sprintf(str + size, "%x", wb_ptr[offset]); + offset++; + } + size += sprintf(str + size, "] "); + } + dev_dbg(adev->dev, "%s\n", str); + } + } + + kfree(str); +} + +static int gfx_v9_4_2_wait_for_waves_assigned(struct amdgpu_device *adev, + uint32_t *wb_ptr, uint32_t mask, + uint32_t pattern, uint32_t num_wave, bool wait) +{ + uint32_t se, cu, simd, wave; + uint32_t loop = 0; + uint32_t wave_cnt; + uint32_t offset; + + do { + wave_cnt = 0; + offset = 0; + + for (se = 0; se < adev->gfx.config.max_shader_engines; se++) + for (cu = 0; cu < CU_ID_MAX; cu++) + for (simd = 0; simd < SIMD_ID_MAX; simd++) + for (wave = 0; wave < WAVE_ID_MAX; wave++) { + if (((1 << wave) & mask) && + (wb_ptr[offset] == pattern)) + wave_cnt++; + + offset++; + } + + if (wave_cnt == num_wave) + return 0; + + mdelay(1); + } while (++loop < 2000 && wait); + + dev_err(adev->dev, "actual wave num: %d, expected wave num: %d\n", + wave_cnt, num_wave); + + gfx_v9_4_2_log_wave_assignment(adev, wb_ptr); + + return -EBADSLT; +} + +static int gfx_v9_4_2_do_sgprs_init(struct amdgpu_device *adev) +{ + int r; + int wb_size = adev->gfx.config.max_shader_engines * + CU_ID_MAX * SIMD_ID_MAX * WAVE_ID_MAX; + struct amdgpu_ib wb_ib; + struct amdgpu_ib disp_ibs[3]; + struct dma_fence *fences[3]; + u32 pattern[3] = { 0x1, 0x5, 0xa }; + + /* bail if the compute ring is not ready */ + if (!adev->gfx.compute_ring[0].sched.ready || + !adev->gfx.compute_ring[1].sched.ready) + return 0; + + /* allocate the write-back buffer from IB */ + memset(&wb_ib, 0, sizeof(wb_ib)); + r = amdgpu_ib_get(adev, NULL, (1 + wb_size) * sizeof(uint32_t), + AMDGPU_IB_POOL_DIRECT, &wb_ib); + if (r) { + dev_err(adev->dev, "failed to get ib (%d) for wb\n", r); + return r; + } + memset(wb_ib.ptr, 0, (1 + wb_size) * sizeof(uint32_t)); + + r = gfx_v9_4_2_run_shader(adev, + &adev->gfx.compute_ring[0], + &disp_ibs[0], + sgpr112_init_compute_shader_aldebaran, + sizeof(sgpr112_init_compute_shader_aldebaran), + sgpr112_init_regs_aldebaran, + ARRAY_SIZE(sgpr112_init_regs_aldebaran), + adev->gfx.cu_info.number, + wb_ib.gpu_addr, pattern[0], &fences[0]); + if (r) { + dev_err(adev->dev, "failed to clear first 224 sgprs\n"); + goto pro_end; + } + + r = gfx_v9_4_2_wait_for_waves_assigned(adev, + &wb_ib.ptr[1], 0b11, + pattern[0], + adev->gfx.cu_info.number * SIMD_ID_MAX * 2, + true); + if (r) { + dev_err(adev->dev, "wave coverage failed when clear first 224 sgprs\n"); + wb_ib.ptr[0] = 0xdeadbeaf; /* stop waves */ + goto disp0_failed; + } + + r = gfx_v9_4_2_run_shader(adev, + &adev->gfx.compute_ring[1], + &disp_ibs[1], + sgpr96_init_compute_shader_aldebaran, + sizeof(sgpr96_init_compute_shader_aldebaran), + sgpr96_init_regs_aldebaran, + ARRAY_SIZE(sgpr96_init_regs_aldebaran), + adev->gfx.cu_info.number * 2, + wb_ib.gpu_addr, pattern[1], &fences[1]); + if (r) { + dev_err(adev->dev, "failed to clear next 576 sgprs\n"); + goto disp0_failed; + } + + r = gfx_v9_4_2_wait_for_waves_assigned(adev, + &wb_ib.ptr[1], 0b11111100, + pattern[1], adev->gfx.cu_info.number * SIMD_ID_MAX * 6, + true); + if (r) { + dev_err(adev->dev, "wave coverage failed when clear first 576 sgprs\n"); + wb_ib.ptr[0] = 0xdeadbeaf; /* stop waves */ + goto disp1_failed; + } + + wb_ib.ptr[0] = 0xdeadbeaf; /* stop waves */ + + /* wait for the GPU to finish processing the IB */ + r = dma_fence_wait(fences[0], false); + if (r) { + dev_err(adev->dev, "timeout to clear first 224 sgprs\n"); + goto disp1_failed; + } + + r = dma_fence_wait(fences[1], false); + if (r) { + dev_err(adev->dev, "timeout to clear first 576 sgprs\n"); + goto disp1_failed; + } + + memset(wb_ib.ptr, 0, (1 + wb_size) * sizeof(uint32_t)); + r = gfx_v9_4_2_run_shader(adev, + &adev->gfx.compute_ring[0], + &disp_ibs[2], + sgpr64_init_compute_shader_aldebaran, + sizeof(sgpr64_init_compute_shader_aldebaran), + sgpr64_init_regs_aldebaran, + ARRAY_SIZE(sgpr64_init_regs_aldebaran), + adev->gfx.cu_info.number, + wb_ib.gpu_addr, pattern[2], &fences[2]); + if (r) { + dev_err(adev->dev, "failed to clear first 256 sgprs\n"); + goto disp1_failed; + } + + r = gfx_v9_4_2_wait_for_waves_assigned(adev, + &wb_ib.ptr[1], 0b1111, + pattern[2], + adev->gfx.cu_info.number * SIMD_ID_MAX * 4, + true); + if (r) { + dev_err(adev->dev, "wave coverage failed when clear first 256 sgprs\n"); + wb_ib.ptr[0] = 0xdeadbeaf; /* stop waves */ + goto disp2_failed; + } + + wb_ib.ptr[0] = 0xdeadbeaf; /* stop waves */ + + r = dma_fence_wait(fences[2], false); + if (r) { + dev_err(adev->dev, "timeout to clear first 256 sgprs\n"); + goto disp2_failed; + } + +disp2_failed: + amdgpu_ib_free(adev, &disp_ibs[2], NULL); + dma_fence_put(fences[2]); +disp1_failed: + amdgpu_ib_free(adev, &disp_ibs[1], NULL); + dma_fence_put(fences[1]); +disp0_failed: + amdgpu_ib_free(adev, &disp_ibs[0], NULL); + dma_fence_put(fences[0]); +pro_end: + amdgpu_ib_free(adev, &wb_ib, NULL); + + if (r) + dev_info(adev->dev, "Init SGPRS Failed\n"); + else + dev_info(adev->dev, "Init SGPRS Successfully\n"); + + return r; +} + +static int gfx_v9_4_2_do_vgprs_init(struct amdgpu_device *adev) +{ + int r; + /* CU_ID: 0~15, SIMD_ID: 0~3, WAVE_ID: 0 ~ 9 */ + int wb_size = adev->gfx.config.max_shader_engines * + CU_ID_MAX * SIMD_ID_MAX * WAVE_ID_MAX; + struct amdgpu_ib wb_ib; + struct amdgpu_ib disp_ib; + struct dma_fence *fence; + u32 pattern = 0xa; + + /* bail if the compute ring is not ready */ + if (!adev->gfx.compute_ring[0].sched.ready) + return 0; + + /* allocate the write-back buffer from IB */ + memset(&wb_ib, 0, sizeof(wb_ib)); + r = amdgpu_ib_get(adev, NULL, (1 + wb_size) * sizeof(uint32_t), + AMDGPU_IB_POOL_DIRECT, &wb_ib); + if (r) { + dev_err(adev->dev, "failed to get ib (%d) for wb.\n", r); + return r; + } + memset(wb_ib.ptr, 0, (1 + wb_size) * sizeof(uint32_t)); + + r = gfx_v9_4_2_run_shader(adev, + &adev->gfx.compute_ring[0], + &disp_ib, + vgpr_init_compute_shader_aldebaran, + sizeof(vgpr_init_compute_shader_aldebaran), + vgpr_init_regs_aldebaran, + ARRAY_SIZE(vgpr_init_regs_aldebaran), + adev->gfx.cu_info.number, + wb_ib.gpu_addr, pattern, &fence); + if (r) { + dev_err(adev->dev, "failed to clear vgprs\n"); + goto pro_end; + } + + /* wait for the GPU to finish processing the IB */ + r = dma_fence_wait(fence, false); + if (r) { + dev_err(adev->dev, "timeout to clear vgprs\n"); + goto disp_failed; + } + + r = gfx_v9_4_2_wait_for_waves_assigned(adev, + &wb_ib.ptr[1], 0b1, + pattern, + adev->gfx.cu_info.number * SIMD_ID_MAX, + false); + if (r) { + dev_err(adev->dev, "failed to cover all simds when clearing vgprs\n"); + goto disp_failed; + } + +disp_failed: + amdgpu_ib_free(adev, &disp_ib, NULL); + dma_fence_put(fence); +pro_end: + amdgpu_ib_free(adev, &wb_ib, NULL); + + if (r) + dev_info(adev->dev, "Init VGPRS Failed\n"); + else + dev_info(adev->dev, "Init VGPRS Successfully\n"); + + return r; +} + +int gfx_v9_4_2_do_edc_gpr_workarounds(struct amdgpu_device *adev) +{ + /* only support when RAS is enabled */ + if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) + return 0; + + gfx_v9_4_2_do_sgprs_init(adev); + + gfx_v9_4_2_do_vgprs_init(adev); + + return 0; +} + static void gfx_v9_4_2_query_sq_timeout_status(struct amdgpu_device *adev); static void gfx_v9_4_2_reset_sq_timeout_status(struct amdgpu_device *adev); @@ -808,8 +1442,9 @@ static struct gfx_v9_4_2_utc_block gfx_v9_4_2_utc_blocks[] = { REG_SET_FIELD(0, ATC_L2_CACHE_4K_DSM_CNTL, WRITE_COUNTERS, 1) }, }; -static const struct soc15_reg_entry gfx_v9_4_2_ea_err_status_regs = - { SOC15_REG_ENTRY(GC, 0, regGCEA_ERR_STATUS), 0, 1, 16 }; +static const struct soc15_reg_entry gfx_v9_4_2_ea_err_status_regs = { + SOC15_REG_ENTRY(GC, 0, regGCEA_ERR_STATUS), 0, 1, 16 +}; static int gfx_v9_4_2_get_reg_error_count(struct amdgpu_device *adev, const struct soc15_reg_entry *reg, @@ -1039,13 +1674,16 @@ static void gfx_v9_4_2_reset_utc_err_status(struct amdgpu_device *adev) static void gfx_v9_4_2_reset_ea_err_status(struct amdgpu_device *adev) { uint32_t i, j; + uint32_t value; + + value = REG_SET_FIELD(0, GCEA_ERR_STATUS, CLEAR_ERROR_STATUS, 0x1); mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < gfx_v9_4_2_ea_err_status_regs.se_num; i++) { for (j = 0; j < gfx_v9_4_2_ea_err_status_regs.instance; j++) { gfx_v9_4_2_select_se_sh(adev, i, 0, j); - WREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_4_2_ea_err_status_regs), 0x10); + WREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_4_2_ea_err_status_regs), value); } } gfx_v9_4_2_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h index 81c5833b6b9f..6db1f88509af 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h @@ -29,6 +29,7 @@ void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev, void gfx_v9_4_2_init_golden_registers(struct amdgpu_device *adev, uint32_t die_id); void gfx_v9_4_2_set_power_brake_sequence(struct amdgpu_device *adev); +int gfx_v9_4_2_do_edc_gpr_workarounds(struct amdgpu_device *adev); extern const struct amdgpu_gfx_ras_funcs gfx_v9_4_2_ras_funcs; diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c index 1e4678cb98f0..a03fdd41212b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c @@ -283,10 +283,14 @@ static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_BLOCK_SIZE, block_size); - /* Send no-retry XNACK on fault to suppress VM fault storm. */ + /* Send no-retry XNACK on fault to suppress VM fault storm. + * On Aldebaran, XNACK can be enabled in the SQ per-process. + * Retry faults need to be enabled for that to work. + */ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, - !adev->gmc.noretry); + !adev->gmc.noretry || + adev->asic_type == CHIP_ALDEBARAN); WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_CNTL, i * hub->ctx_distance, tmp); WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 455bb91060d0..093ab98c31bb 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -53,6 +53,7 @@ #include "mmhub_v1_7.h" #include "umc_v6_1.h" #include "umc_v6_0.h" +#include "hdp_v4_0.h" #include "ivsrcid/vmc/irqsrcs_vmc_1_0.h" @@ -1210,6 +1211,11 @@ static void gmc_v9_0_set_gfxhub_funcs(struct amdgpu_device *adev) adev->gfxhub.funcs = &gfxhub_v1_0_funcs; } +static void gmc_v9_0_set_hdp_ras_funcs(struct amdgpu_device *adev) +{ + adev->hdp.ras_funcs = &hdp_v4_0_ras_funcs; +} + static int gmc_v9_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -1230,6 +1236,7 @@ static int gmc_v9_0_early_init(void *handle) gmc_v9_0_set_mmhub_funcs(adev); gmc_v9_0_set_mmhub_ras_funcs(adev); gmc_v9_0_set_gfxhub_funcs(adev); + gmc_v9_0_set_hdp_ras_funcs(adev); adev->gmc.shared_aperture_start = 0x2000000000000000ULL; adev->gmc.shared_aperture_end = @@ -1255,7 +1262,7 @@ static int gmc_v9_0_late_init(void *handle) * writes, while disables HBM ECC for vega10. */ if (!amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_VEGA10)) { - if (!(adev->ras_features & (1 << AMDGPU_RAS_BLOCK__UMC))) { + if (!(adev->ras_enabled & (1 << AMDGPU_RAS_BLOCK__UMC))) { if (adev->df.funcs->enable_ecc_force_par_wr_rmw) adev->df.funcs->enable_ecc_force_par_wr_rmw(adev, false); } @@ -1265,6 +1272,10 @@ static int gmc_v9_0_late_init(void *handle) adev->mmhub.ras_funcs->reset_ras_error_count) adev->mmhub.ras_funcs->reset_ras_error_count(adev); + if (adev->hdp.ras_funcs && + adev->hdp.ras_funcs->reset_ras_error_count) + adev->hdp.ras_funcs->reset_ras_error_count(adev); + r = amdgpu_gmc_ras_late_init(adev); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c index edbd35d293eb..74b90cc2bf48 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c @@ -59,12 +59,31 @@ static void hdp_v4_0_invalidate_hdp(struct amdgpu_device *adev, HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1); } +static void hdp_v4_0_query_ras_error_count(struct amdgpu_device *adev, + void *ras_error_status) +{ + struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; + + err_data->ue_count = 0; + err_data->ce_count = 0; + + if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__HDP)) + return; + + /* HDP SRAM errors are uncorrectable ones (i.e. fatal errors) */ + err_data->ue_count += RREG32_SOC15(HDP, 0, mmHDP_EDC_CNT); +}; + static void hdp_v4_0_reset_ras_error_count(struct amdgpu_device *adev) { if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__HDP)) return; - /*read back hdp ras counter to reset it to 0 */ - RREG32_SOC15(HDP, 0, mmHDP_EDC_CNT); + + if (adev->asic_type >= CHIP_ALDEBARAN) + WREG32_SOC15(HDP, 0, mmHDP_EDC_CNT, 0); + else + /*read back hdp ras counter to reset it to 0 */ + RREG32_SOC15(HDP, 0, mmHDP_EDC_CNT); } static void hdp_v4_0_update_clock_gating(struct amdgpu_device *adev, @@ -130,10 +149,16 @@ static void hdp_v4_0_init_registers(struct amdgpu_device *adev) WREG32_SOC15(HDP, 0, mmHDP_NONSURFACE_BASE_HI, (adev->gmc.vram_start >> 40)); } +const struct amdgpu_hdp_ras_funcs hdp_v4_0_ras_funcs = { + .ras_late_init = amdgpu_hdp_ras_late_init, + .ras_fini = amdgpu_hdp_ras_fini, + .query_ras_error_count = hdp_v4_0_query_ras_error_count, + .reset_ras_error_count = hdp_v4_0_reset_ras_error_count, +}; + const struct amdgpu_hdp_funcs hdp_v4_0_funcs = { .flush_hdp = hdp_v4_0_flush_hdp, .invalidate_hdp = hdp_v4_0_invalidate_hdp, - .reset_ras_error_count = hdp_v4_0_reset_ras_error_count, .update_clock_gating = hdp_v4_0_update_clock_gating, .get_clock_gating_state = hdp_v4_0_get_clockgating_state, .init_registers = hdp_v4_0_init_registers, diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h index d1e6399e8c46..dc3a1b81dd62 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h @@ -27,5 +27,6 @@ #include "soc15_common.h" extern const struct amdgpu_hdp_funcs hdp_v4_0_funcs; +extern const struct amdgpu_hdp_ras_funcs hdp_v4_0_ras_funcs; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c index 0103a5ab28e6..6264934b67ac 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c @@ -296,10 +296,12 @@ static void mmhub_v1_7_setup_vmid_config(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_BLOCK_SIZE, block_size); - /* Send no-retry XNACK on fault to suppress VM fault storm. */ + /* On Aldebaran, XNACK can be enabled in the SQ per-process. + * Retry faults need to be enabled for that to work. + */ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, - !adev->gmc.noretry); + 1); WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_CNTL, i * hub->ctx_distance, tmp); WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, @@ -1313,12 +1315,31 @@ static void mmhub_v1_7_query_ras_error_status(struct amdgpu_device *adev) } } +static void mmhub_v1_7_reset_ras_error_status(struct amdgpu_device *adev) +{ + int i; + uint32_t reg_value; + + if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__MMHUB)) + return; + + for (i = 0; i < ARRAY_SIZE(mmhub_v1_7_ea_err_status_regs); i++) { + reg_value = RREG32(SOC15_REG_ENTRY_OFFSET( + mmhub_v1_7_ea_err_status_regs[i])); + reg_value = REG_SET_FIELD(reg_value, MMEA0_ERR_STATUS, + CLEAR_ERROR_STATUS, 0x01); + WREG32(SOC15_REG_ENTRY_OFFSET(mmhub_v1_7_ea_err_status_regs[i]), + reg_value); + } +} + const struct amdgpu_mmhub_ras_funcs mmhub_v1_7_ras_funcs = { .ras_late_init = amdgpu_mmhub_ras_late_init, .ras_fini = amdgpu_mmhub_ras_fini, .query_ras_error_count = mmhub_v1_7_query_ras_error_count, .reset_ras_error_count = mmhub_v1_7_reset_ras_error_count, .query_ras_error_status = mmhub_v1_7_query_ras_error_status, + .reset_ras_error_status = mmhub_v1_7_reset_ras_error_status, }; const struct amdgpu_mmhub_funcs mmhub_v1_7_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c index a9899335d0b1..709ac576ac7e 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c @@ -569,9 +569,9 @@ static int mmhub_v2_3_set_clockgating(struct amdgpu_device *adev, return 0; mmhub_v2_3_update_medium_grain_clock_gating(adev, - state == AMD_CG_STATE_GATE ? true : false); + state == AMD_CG_STATE_GATE); mmhub_v2_3_update_medium_grain_light_sleep(adev, - state == AMD_CG_STATE_GATE ? true : false); + state == AMD_CG_STATE_GATE); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/nv.c b/drivers/gpu/drm/amd/amdgpu/nv.c index d290ca0b06da..75d1f9b939b2 100644 --- a/drivers/gpu/drm/amd/amdgpu/nv.c +++ b/drivers/gpu/drm/amd/amdgpu/nv.c @@ -598,7 +598,7 @@ static void nv_pcie_gen3_enable(struct amdgpu_device *adev) static void nv_program_aspm(struct amdgpu_device *adev) { - if (amdgpu_aspm != 1) + if (!amdgpu_aspm) return; if (!(adev->flags & AMD_IS_APU) && @@ -1068,6 +1068,7 @@ static int nv_common_early_init(void *handle) case CHIP_SIENNA_CICHLID: adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_MC_MGCG | AMD_CG_SUPPORT_VCN_MGCG | @@ -1091,6 +1092,7 @@ static int nv_common_early_init(void *handle) case CHIP_NAVY_FLOUNDER: adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_VCN_MGCG | AMD_CG_SUPPORT_JPEG_MGCG | @@ -1121,6 +1123,8 @@ static int nv_common_early_init(void *handle) AMD_CG_SUPPORT_MC_LS | AMD_CG_SUPPORT_GFX_FGCG | AMD_CG_SUPPORT_VCN_MGCG | + AMD_CG_SUPPORT_SDMA_MGCG | + AMD_CG_SUPPORT_SDMA_LS | AMD_CG_SUPPORT_JPEG_MGCG; adev->pg_flags = AMD_PG_SUPPORT_GFX_PG | AMD_PG_SUPPORT_VCN | @@ -1132,6 +1136,7 @@ static int nv_common_early_init(void *handle) case CHIP_DIMGREY_CAVEFISH: adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_VCN_MGCG | AMD_CG_SUPPORT_JPEG_MGCG | diff --git a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h index 96064c343163..f6d3180febc4 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h @@ -97,7 +97,6 @@ enum psp_gfx_cmd_id GFX_CMD_ID_SETUP_VMR = 0x00000009, /* setup VMR region */ GFX_CMD_ID_DESTROY_VMR = 0x0000000A, /* destroy VMR region */ GFX_CMD_ID_PROG_REG = 0x0000000B, /* program regs */ - GFX_CMD_ID_CLEAR_VF_FW = 0x0000000D, /* Clear VF FW, to be used on VF shutdown. */ GFX_CMD_ID_GET_FW_ATTESTATION = 0x0000000F, /* Query GPUVA of the Fw Attestation DB */ /* IDs upto 0x1F are reserved for older programs (Raven, Vega 10/12/20) */ GFX_CMD_ID_LOAD_TOC = 0x00000020, /* Load TOC and obtain TMR size */ diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 5715be6770ec..d197185f7789 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -1109,6 +1109,8 @@ static void sdma_v4_0_ctx_switch_enable(struct amdgpu_device *adev, bool enable) if (adev->asic_type == CHIP_ARCTURUS && adev->sdma.instance[i].fw_version >= 14) WREG32_SDMA(i, mmSDMA0_PUB_DUMMY_REG2, enable); + /* Extend page fault timeout to avoid interrupt storm */ + WREG32_SDMA(i, mmSDMA0_UTCL1_TIMEOUT, 0x00800080); } } @@ -2227,7 +2229,7 @@ static int sdma_v4_0_print_iv_entry(struct amdgpu_device *adev, memset(&task_info, 0, sizeof(struct amdgpu_task_info)); amdgpu_vm_get_task_info(adev, entry->pasid, &task_info); - dev_info(adev->dev, + dev_dbg_ratelimited(adev->dev, "[sdma%d] address:0x%016llx src_id:%u ring:%u vmid:%u " "pasid:%u, for process %s pid %d thread %s pid %d\n", instance, addr, entry->src_id, entry->ring_id, entry->vmid, @@ -2240,7 +2242,7 @@ static int sdma_v4_0_process_vm_hole_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - dev_err(adev->dev, "MC or SEM address in VM hole\n"); + dev_dbg_ratelimited(adev->dev, "MC or SEM address in VM hole\n"); sdma_v4_0_print_iv_entry(adev, entry); return 0; } @@ -2249,7 +2251,7 @@ static int sdma_v4_0_process_doorbell_invalid_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - dev_err(adev->dev, "SDMA received a doorbell from BIF with byte_enable !=0xff\n"); + dev_dbg_ratelimited(adev->dev, "SDMA received a doorbell from BIF with byte_enable !=0xff\n"); sdma_v4_0_print_iv_entry(adev, entry); return 0; } @@ -2258,7 +2260,7 @@ static int sdma_v4_0_process_pool_timeout_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - dev_err(adev->dev, + dev_dbg_ratelimited(adev->dev, "Polling register/memory timeout executing POLL_REG/MEM with finite timer\n"); sdma_v4_0_print_iv_entry(adev, entry); return 0; @@ -2268,7 +2270,7 @@ static int sdma_v4_0_process_srbm_write_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - dev_err(adev->dev, + dev_dbg_ratelimited(adev->dev, "SDMA gets an Register Write SRBM_WRITE command in non-privilege command buffer\n"); sdma_v4_0_print_iv_entry(adev, entry); return 0; @@ -2597,27 +2599,18 @@ static const struct amdgpu_irq_src_funcs sdma_v4_0_srbm_write_irq_funcs = { static void sdma_v4_0_set_irq_funcs(struct amdgpu_device *adev) { + adev->sdma.trap_irq.num_types = adev->sdma.num_instances; + adev->sdma.ecc_irq.num_types = adev->sdma.num_instances; + /*For Arcturus and Aldebaran, add another 4 irq handler*/ switch (adev->sdma.num_instances) { - case 1: - adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE1; - adev->sdma.ecc_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE1; - break; case 5: - adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE5; - adev->sdma.ecc_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE5; - break; case 8: - adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_LAST; - adev->sdma.ecc_irq.num_types = AMDGPU_SDMA_IRQ_LAST; - adev->sdma.vm_hole_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE5; - adev->sdma.doorbell_invalid_irq.num_types = AMDGPU_SDMA_IRQ_LAST; - adev->sdma.pool_timeout_irq.num_types = AMDGPU_SDMA_IRQ_LAST; - adev->sdma.srbm_write_irq.num_types = AMDGPU_SDMA_IRQ_LAST; + adev->sdma.vm_hole_irq.num_types = adev->sdma.num_instances; + adev->sdma.doorbell_invalid_irq.num_types = adev->sdma.num_instances; + adev->sdma.pool_timeout_irq.num_types = adev->sdma.num_instances; + adev->sdma.srbm_write_irq.num_types = adev->sdma.num_instances; break; - case 2: default: - adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE2; - adev->sdma.ecc_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE2; break; } adev->sdma.trap_irq.funcs = &sdma_v4_0_trap_irq_funcs; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c index 920fc6d4a127..04c68a79eca4 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c @@ -405,18 +405,6 @@ static void sdma_v5_0_ring_emit_ib(struct amdgpu_ring *ring, unsigned vmid = AMDGPU_JOB_GET_VMID(job); uint64_t csa_mc_addr = amdgpu_sdma_get_csa_mc_addr(ring, vmid); - /* Invalidate L2, because if we don't do it, we might get stale cache - * lines from previous IBs. - */ - amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_GCR_REQ)); - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, (SDMA_GCR_GL2_INV | - SDMA_GCR_GL2_WB | - SDMA_GCR_GLM_INV | - SDMA_GCR_GLM_WB) << 16); - amdgpu_ring_write(ring, 0xffffff80); - amdgpu_ring_write(ring, 0xffff); - /* An IB packet must end on a 8 DW boundary--the next dword * must be on a 8-dword boundary. Our IB packet below is 6 * dwords long, thus add x number of NOPs, such that, in @@ -438,6 +426,33 @@ static void sdma_v5_0_ring_emit_ib(struct amdgpu_ring *ring, } /** + * sdma_v5_0_ring_emit_mem_sync - flush the IB by graphics cache rinse + * + * @ring: amdgpu ring pointer + * @job: job to retrieve vmid from + * @ib: IB object to schedule + * + * flush the IB by graphics cache rinse. + */ +static void sdma_v5_0_ring_emit_mem_sync(struct amdgpu_ring *ring) +{ + uint32_t gcr_cntl = + SDMA_GCR_GL2_INV | SDMA_GCR_GL2_WB | SDMA_GCR_GLM_INV | + SDMA_GCR_GL1_INV | SDMA_GCR_GLV_INV | SDMA_GCR_GLK_INV | + SDMA_GCR_GLI_INV(1); + + /* flush entire cache L0/L1/L2, this can be optimized by performance requirement */ + amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_GCR_REQ)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD1_BASE_VA_31_7(0)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD2_GCR_CONTROL_15_0(gcr_cntl) | + SDMA_PKT_GCR_REQ_PAYLOAD2_BASE_VA_47_32(0)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD3_LIMIT_VA_31_7(0) | + SDMA_PKT_GCR_REQ_PAYLOAD3_GCR_CONTROL_18_16(gcr_cntl >> 16)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD4_LIMIT_VA_47_32(0) | + SDMA_PKT_GCR_REQ_PAYLOAD4_VMID(0)); +} + +/** * sdma_v5_0_ring_emit_hdp_flush - emit an hdp flush on the DMA ring * * @ring: amdgpu ring pointer @@ -1643,6 +1658,7 @@ static const struct amdgpu_ring_funcs sdma_v5_0_ring_funcs = { 10 + 10 + 10, /* sdma_v5_0_ring_emit_fence x3 for user fence, vm fence */ .emit_ib_size = 5 + 7 + 6, /* sdma_v5_0_ring_emit_ib */ .emit_ib = sdma_v5_0_ring_emit_ib, + .emit_mem_sync = sdma_v5_0_ring_emit_mem_sync, .emit_fence = sdma_v5_0_ring_emit_fence, .emit_pipeline_sync = sdma_v5_0_ring_emit_pipeline_sync, .emit_vm_flush = sdma_v5_0_ring_emit_vm_flush, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c index b1ad9e52b234..7c4e0586e26d 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c @@ -1556,6 +1556,10 @@ static void sdma_v5_2_update_medium_grain_clock_gating(struct amdgpu_device *ade int i; for (i = 0; i < adev->sdma.num_instances; i++) { + + if (adev->sdma.instance[i].fw_version < 70 && adev->asic_type == CHIP_VANGOGH) + adev->cg_flags &= ~AMD_CG_SUPPORT_SDMA_MGCG; + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG)) { /* Enable sdma clock gating */ def = data = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CLK_CTRL)); @@ -1589,6 +1593,10 @@ static void sdma_v5_2_update_medium_grain_light_sleep(struct amdgpu_device *adev int i; for (i = 0; i < adev->sdma.num_instances; i++) { + + if (adev->sdma.instance[i].fw_version < 70 && adev->asic_type == CHIP_VANGOGH) + adev->cg_flags &= ~AMD_CG_SUPPORT_SDMA_LS; + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS)) { /* Enable sdma mem light sleep */ def = data = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_POWER_CNTL)); diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c index 079b094c48ad..3c47c94846d6 100644 --- a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c +++ b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c @@ -89,6 +89,23 @@ static u32 smuio_v13_0_get_die_id(struct amdgpu_device *adev) } /** + * smuio_v13_0_get_socket_id - query socket id from FCH + * + * @adev: amdgpu device pointer + * + * Returns socket id + */ +static u32 smuio_v13_0_get_socket_id(struct amdgpu_device *adev) +{ + u32 data, socket_id; + + data = RREG32_SOC15(SMUIO, 0, regSMUIO_MCM_CONFIG); + socket_id = REG_GET_FIELD(data, SMUIO_MCM_CONFIG, SOCKET_ID); + + return socket_id; +} + +/** * smuio_v13_0_supports_host_gpu_xgmi - detect xgmi interface between cpu and gpu/s. * * @adev: amdgpu device pointer @@ -115,6 +132,7 @@ const struct amdgpu_smuio_funcs smuio_v13_0_funcs = { .get_rom_index_offset = smuio_v13_0_get_rom_index_offset, .get_rom_data_offset = smuio_v13_0_get_rom_data_offset, .get_die_id = smuio_v13_0_get_die_id, + .get_socket_id = smuio_v13_0_get_socket_id, .is_host_gpu_xgmi_supported = smuio_v13_0_is_host_gpu_xgmi_supported, .update_rom_clock_gating = smuio_v13_0_update_rom_clock_gating, .get_clock_gating_state = smuio_v13_0_get_clock_gating_state, diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 8e1b9a40839f..4b660b2d1c22 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -655,7 +655,7 @@ static int soc15_asic_baco_reset(struct amdgpu_device *adev) int ret = 0; /* avoid NBIF got stuck when do RAS recovery in BACO reset */ - if (ras && ras->supported) + if (ras && adev->ras_enabled) adev->nbio.funcs->enable_doorbell_interrupt(adev, false); ret = amdgpu_dpm_baco_reset(adev); @@ -663,7 +663,7 @@ static int soc15_asic_baco_reset(struct amdgpu_device *adev) return ret; /* re-enable doorbell interrupt after BACO exit */ - if (ras && ras->supported) + if (ras && adev->ras_enabled) adev->nbio.funcs->enable_doorbell_interrupt(adev, true); return 0; @@ -710,7 +710,8 @@ soc15_asic_reset_method(struct amdgpu_device *adev) * 1. PMFW version > 0x284300: all cases use baco * 2. PMFW version <= 0x284300: only sGPU w/o RAS use baco */ - if ((ras && ras->supported) && adev->pm.fw_version <= 0x283400) + if (ras && adev->ras_enabled && + adev->pm.fw_version <= 0x283400) baco_reset = false; break; case CHIP_ALDEBARAN: @@ -816,7 +817,7 @@ static void soc15_pcie_gen3_enable(struct amdgpu_device *adev) static void soc15_program_aspm(struct amdgpu_device *adev) { - if (amdgpu_aspm != 1) + if (!amdgpu_aspm) return; if (!(adev->flags & AMD_IS_APU) && @@ -1522,9 +1523,6 @@ static int soc15_common_late_init(void *handle) if (amdgpu_sriov_vf(adev)) xgpu_ai_mailbox_get_irq(adev); - if (adev->hdp.funcs->reset_ras_error_count) - adev->hdp.funcs->reset_ras_error_count(adev); - if (adev->nbio.ras_funcs && adev->nbio.ras_funcs->ras_late_init) r = adev->nbio.ras_funcs->ras_late_init(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h b/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h index 745ed0fba1ed..0f214a398dd8 100644 --- a/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h +++ b/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h @@ -105,6 +105,12 @@ struct ta_ras_trigger_error_input { uint64_t value; // method if error injection. i.e persistent, coherent etc. }; +struct ta_ras_init_flags +{ + uint8_t poison_mode_en; + uint8_t dgpu_mode; +}; + struct ta_ras_output_flags { uint8_t ras_init_success_flag; @@ -115,6 +121,7 @@ struct ta_ras_output_flags /* Common input structure for RAS callbacks */ /**********************************************************/ union ta_ras_cmd_input { + struct ta_ras_init_flags init_flags; struct ta_ras_enable_features_input enable_features; struct ta_ras_disable_features_input disable_features; struct ta_ras_trigger_error_input trigger_error; diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index ca8efa5c6978..2f17c8a57015 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -104,6 +104,7 @@ static int vega10_ih_toggle_ring_interrupts(struct amdgpu_device *adev, tmp = RREG32(ih_regs->ih_rb_cntl); tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0)); + tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_GPU_TS_ENABLE, 1); /* enable_intr field is only valid in ring0 */ if (ih == &adev->irq.ih) tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0)); diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index 735ebbd1148f..3d21c0799037 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -1136,7 +1136,7 @@ static void vi_program_aspm(struct amdgpu_device *adev) bool bL1SS = false; bool bClkReqSupport = true; - if (amdgpu_aspm != 1) + if (!amdgpu_aspm) return; if (adev->flags & AMD_IS_APU || diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig b/drivers/gpu/drm/amd/amdkfd/Kconfig index f02c938f75da..8cc0a76ddf9f 100644 --- a/drivers/gpu/drm/amd/amdkfd/Kconfig +++ b/drivers/gpu/drm/amd/amdkfd/Kconfig @@ -12,3 +12,16 @@ config HSA_AMD select DRM_AMDGPU_USERPTR help Enable this if you want to use HSA features on AMD GPU devices. + +config HSA_AMD_SVM + bool "Enable HMM-based shared virtual memory manager" + depends on HSA_AMD && DEVICE_PRIVATE + default y + select HMM_MIRROR + select MMU_NOTIFIER + help + Enable this to use unified memory and managed memory in HIP. This + memory manager supports two modes of operation. One based on + preemptions and one based on page faults. To enable page fault + based memory management on most GFXv9 GPUs, set the module + parameter amdgpu.noretry=0. diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index e1e4115dcf78..c4f3aff11072 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -63,3 +63,8 @@ endif ifneq ($(CONFIG_DEBUG_FS),) AMDKFD_FILES += $(AMDKFD_PATH)/kfd_debugfs.o endif + +ifneq ($(CONFIG_HSA_AMD_SVM),) +AMDKFD_FILES += $(AMDKFD_PATH)/kfd_svm.o \ + $(AMDKFD_PATH)/kfd_migrate.o +endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 43de260b2230..059c3f1ca27d 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -38,6 +38,7 @@ #include "kfd_priv.h" #include "kfd_device_queue_manager.h" #include "kfd_dbgmgr.h" +#include "kfd_svm.h" #include "amdgpu_amdkfd.h" #include "kfd_smi_events.h" @@ -1297,7 +1298,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep, err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( dev->kgd, args->va_addr, args->size, - pdd->vm, (struct kgd_mem **) &mem, &offset, + pdd->drm_priv, (struct kgd_mem **) &mem, &offset, flags); if (err) @@ -1328,7 +1329,8 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep, return 0; err_free: - amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem, NULL); + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem, + pdd->drm_priv, NULL); err_unlock: mutex_unlock(&p->mutex); return err; @@ -1365,7 +1367,7 @@ static int kfd_ioctl_free_memory_of_gpu(struct file *filep, } ret = amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, - (struct kgd_mem *)mem, &size); + (struct kgd_mem *)mem, pdd->drm_priv, &size); /* If freeing the buffer failed, leave the handle in place for * clean-up during process tear-down. @@ -1448,7 +1450,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep, goto get_mem_obj_from_handle_failed; } err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu( - peer->kgd, (struct kgd_mem *)mem, peer_pdd->vm); + peer->kgd, (struct kgd_mem *)mem, peer_pdd->drm_priv); if (err) { pr_err("Failed to map to gpu %d/%d\n", i, args->n_devices); @@ -1555,7 +1557,7 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep, goto get_mem_obj_from_handle_failed; } err = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( - peer->kgd, (struct kgd_mem *)mem, peer_pdd->vm); + peer->kgd, (struct kgd_mem *)mem, peer_pdd->drm_priv); if (err) { pr_err("Failed to unmap from gpu %d/%d\n", i, args->n_devices); @@ -1701,7 +1703,7 @@ static int kfd_ioctl_import_dmabuf(struct file *filep, } r = amdgpu_amdkfd_gpuvm_import_dmabuf(dev->kgd, dmabuf, - args->va_addr, pdd->vm, + args->va_addr, pdd->drm_priv, (struct kgd_mem **)&mem, &size, NULL); if (r) @@ -1721,7 +1723,8 @@ static int kfd_ioctl_import_dmabuf(struct file *filep, return 0; err_free: - amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem, NULL); + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem, + pdd->drm_priv, NULL); err_unlock: mutex_unlock(&p->mutex); dma_buf_put(dmabuf); @@ -1742,6 +1745,64 @@ static int kfd_ioctl_smi_events(struct file *filep, return kfd_smi_event_open(dev, &args->anon_fd); } +static int kfd_ioctl_set_xnack_mode(struct file *filep, + struct kfd_process *p, void *data) +{ + struct kfd_ioctl_set_xnack_mode_args *args = data; + int r = 0; + + mutex_lock(&p->mutex); + if (args->xnack_enabled >= 0) { + if (!list_empty(&p->pqm.queues)) { + pr_debug("Process has user queues running\n"); + mutex_unlock(&p->mutex); + return -EBUSY; + } + if (args->xnack_enabled && !kfd_process_xnack_mode(p, true)) + r = -EPERM; + else + p->xnack_enabled = args->xnack_enabled; + } else { + args->xnack_enabled = p->xnack_enabled; + } + mutex_unlock(&p->mutex); + + return r; +} + +#if IS_ENABLED(CONFIG_HSA_AMD_SVM) +static int kfd_ioctl_svm(struct file *filep, struct kfd_process *p, void *data) +{ + struct kfd_ioctl_svm_args *args = data; + int r = 0; + + if (p->svm_disabled) + return -EPERM; + + pr_debug("start 0x%llx size 0x%llx op 0x%x nattr 0x%x\n", + args->start_addr, args->size, args->op, args->nattr); + + if ((args->start_addr & ~PAGE_MASK) || (args->size & ~PAGE_MASK)) + return -EINVAL; + if (!args->start_addr || !args->size) + return -EINVAL; + + mutex_lock(&p->mutex); + + r = svm_ioctl(p, args->op, args->start_addr, args->size, args->nattr, + args->attrs); + + mutex_unlock(&p->mutex); + + return r; +} +#else +static int kfd_ioctl_svm(struct file *filep, struct kfd_process *p, void *data) +{ + return -EPERM; +} +#endif + #define AMDKFD_IOCTL_DEF(ioctl, _func, _flags) \ [_IOC_NR(ioctl)] = {.cmd = ioctl, .func = _func, .flags = _flags, \ .cmd_drv = 0, .name = #ioctl} @@ -1840,6 +1901,11 @@ static const struct amdkfd_ioctl_desc amdkfd_ioctls[] = { AMDKFD_IOCTL_DEF(AMDKFD_IOC_SMI_EVENTS, kfd_ioctl_smi_events, 0), + + AMDKFD_IOCTL_DEF(AMDKFD_IOC_SVM, kfd_ioctl_svm, 0), + + AMDKFD_IOCTL_DEF(AMDKFD_IOC_SET_XNACK_MODE, + kfd_ioctl_set_xnack_mode, 0), }; #define AMDKFD_CORE_IOCTL_COUNT ARRAY_SIZE(amdkfd_ioctls) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c index c60e82697385..c1815b708ac8 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c @@ -55,7 +55,7 @@ struct kfd_gpu_cache_info { uint32_t cache_level; uint32_t flags; /* Indicates how many Compute Units share this cache - * Value = 1 indicates the cache is not shared + * within a SA. Value = 1 indicates the cache is not shared */ uint32_t num_cu_shared; }; @@ -69,7 +69,6 @@ static struct kfd_gpu_cache_info kaveri_cache_info[] = { CRAT_CACHE_FLAGS_DATA_CACHE | CRAT_CACHE_FLAGS_SIMD_CACHE), .num_cu_shared = 1, - }, { /* Scalar L1 Instruction Cache (in SQC module) per bank */ @@ -126,9 +125,6 @@ static struct kfd_gpu_cache_info carrizo_cache_info[] = { /* TODO: Add L2 Cache information */ }; -/* NOTE: In future if more information is added to struct kfd_gpu_cache_info - * the following ASICs may need a separate table. - */ #define hawaii_cache_info kaveri_cache_info #define tonga_cache_info carrizo_cache_info #define fiji_cache_info carrizo_cache_info @@ -136,13 +132,562 @@ static struct kfd_gpu_cache_info carrizo_cache_info[] = { #define polaris11_cache_info carrizo_cache_info #define polaris12_cache_info carrizo_cache_info #define vegam_cache_info carrizo_cache_info -/* TODO - check & update Vega10 cache details */ -#define vega10_cache_info carrizo_cache_info -#define raven_cache_info carrizo_cache_info -#define renoir_cache_info carrizo_cache_info -/* TODO - check & update Navi10 cache details */ -#define navi10_cache_info carrizo_cache_info -#define vangogh_cache_info carrizo_cache_info + +/* NOTE: L1 cache information has been updated and L2/L3 + * cache information has been added for Vega10 and + * newer ASICs. The unit for cache_size is KiB. + * In future, check & update cache details + * for every new ASIC is required. + */ + +static struct kfd_gpu_cache_info vega10_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 3, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 3, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 4096, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 16, + }, +}; + +static struct kfd_gpu_cache_info raven_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 3, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 3, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 1024, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 11, + }, +}; + +static struct kfd_gpu_cache_info renoir_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 3, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 3, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 1024, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 8, + }, +}; + +static struct kfd_gpu_cache_info vega12_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 3, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 3, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 2048, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 5, + }, +}; + +static struct kfd_gpu_cache_info vega20_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 3, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 3, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 8192, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 16, + }, +}; + +static struct kfd_gpu_cache_info aldebaran_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 8192, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 14, + }, +}; + +static struct kfd_gpu_cache_info navi10_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* GL1 Data Cache per SA */ + .cache_size = 128, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 10, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 4096, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 10, + }, +}; + +static struct kfd_gpu_cache_info vangogh_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* GL1 Data Cache per SA */ + .cache_size = 128, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 8, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 1024, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 8, + }, +}; + +static struct kfd_gpu_cache_info navi14_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* GL1 Data Cache per SA */ + .cache_size = 128, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 12, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 2048, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 12, + }, +}; + +static struct kfd_gpu_cache_info sienna_cichlid_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* GL1 Data Cache per SA */ + .cache_size = 128, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 10, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 4096, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 10, + }, + { + /* L3 Data Cache per GPU */ + .cache_size = 128*1024, + .cache_level = 3, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 10, + }, +}; + +static struct kfd_gpu_cache_info navy_flounder_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* GL1 Data Cache per SA */ + .cache_size = 128, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 10, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 3072, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 10, + }, + { + /* L3 Data Cache per GPU */ + .cache_size = 96*1024, + .cache_level = 3, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 10, + }, +}; + +static struct kfd_gpu_cache_info dimgrey_cavefish_cache_info[] = { + { + /* TCP L1 Cache per CU */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 1, + }, + { + /* Scalar L1 Instruction Cache per SQC */ + .cache_size = 32, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_INST_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* Scalar L1 Data Cache per SQC */ + .cache_size = 16, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 2, + }, + { + /* GL1 Data Cache per SA */ + .cache_size = 128, + .cache_level = 1, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 8, + }, + { + /* L2 Data Cache per GPU (Total Tex Cache) */ + .cache_size = 2048, + .cache_level = 2, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 8, + }, + { + /* L3 Data Cache per GPU */ + .cache_size = 32*1024, + .cache_level = 3, + .flags = (CRAT_CACHE_FLAGS_ENABLED | + CRAT_CACHE_FLAGS_DATA_CACHE | + CRAT_CACHE_FLAGS_SIMD_CACHE), + .num_cu_shared = 8, + }, +}; static void kfd_populated_cu_info_cpu(struct kfd_topology_device *dev, struct crat_subtype_computeunit *cu) @@ -544,7 +1089,7 @@ err: } /* Helper function. See kfd_fill_gpu_cache_info for parameter description */ -static int fill_in_pcache(struct crat_subtype_cache *pcache, +static int fill_in_l1_pcache(struct crat_subtype_cache *pcache, struct kfd_gpu_cache_info *pcache_info, struct kfd_cu_info *cu_info, int mem_available, @@ -597,6 +1142,70 @@ static int fill_in_pcache(struct crat_subtype_cache *pcache, return 1; } +/* Helper function. See kfd_fill_gpu_cache_info for parameter description */ +static int fill_in_l2_l3_pcache(struct crat_subtype_cache *pcache, + struct kfd_gpu_cache_info *pcache_info, + struct kfd_cu_info *cu_info, + int mem_available, + int cache_type, unsigned int cu_processor_id) +{ + unsigned int cu_sibling_map_mask; + int first_active_cu; + int i, j, k; + + /* First check if enough memory is available */ + if (sizeof(struct crat_subtype_cache) > mem_available) + return -ENOMEM; + + cu_sibling_map_mask = cu_info->cu_bitmap[0][0]; + cu_sibling_map_mask &= + ((1 << pcache_info[cache_type].num_cu_shared) - 1); + first_active_cu = ffs(cu_sibling_map_mask); + + /* CU could be inactive. In case of shared cache find the first active + * CU. and incase of non-shared cache check if the CU is inactive. If + * inactive active skip it + */ + if (first_active_cu) { + memset(pcache, 0, sizeof(struct crat_subtype_cache)); + pcache->type = CRAT_SUBTYPE_CACHE_AFFINITY; + pcache->length = sizeof(struct crat_subtype_cache); + pcache->flags = pcache_info[cache_type].flags; + pcache->processor_id_low = cu_processor_id + + (first_active_cu - 1); + pcache->cache_level = pcache_info[cache_type].cache_level; + pcache->cache_size = pcache_info[cache_type].cache_size; + + /* Sibling map is w.r.t processor_id_low, so shift out + * inactive CU + */ + cu_sibling_map_mask = + cu_sibling_map_mask >> (first_active_cu - 1); + k = 0; + for (i = 0; i < cu_info->num_shader_engines; i++) { + for (j = 0; j < cu_info->num_shader_arrays_per_engine; + j++) { + pcache->sibling_map[k] = + (uint8_t)(cu_sibling_map_mask & 0xFF); + pcache->sibling_map[k+1] = + (uint8_t)((cu_sibling_map_mask >> 8) & 0xFF); + pcache->sibling_map[k+2] = + (uint8_t)((cu_sibling_map_mask >> 16) & 0xFF); + pcache->sibling_map[k+3] = + (uint8_t)((cu_sibling_map_mask >> 24) & 0xFF); + k += 4; + cu_sibling_map_mask = + cu_info->cu_bitmap[i % 4][j + i / 4]; + cu_sibling_map_mask &= ( + (1 << pcache_info[cache_type].num_cu_shared) + - 1); + } + } + return 0; + } + return 1; +} + /* kfd_fill_gpu_cache_info - Fill GPU cache info using kfd_gpu_cache_info * tables * @@ -624,6 +1233,7 @@ static int kfd_fill_gpu_cache_info(struct kfd_dev *kdev, int mem_available = available_size; unsigned int cu_processor_id; int ret; + unsigned int num_cu_shared; switch (kdev->device_info->asic_family) { case CHIP_KAVERI: @@ -663,12 +1273,21 @@ static int kfd_fill_gpu_cache_info(struct kfd_dev *kdev, num_of_cache_types = ARRAY_SIZE(vegam_cache_info); break; case CHIP_VEGA10: + pcache_info = vega10_cache_info; + num_of_cache_types = ARRAY_SIZE(vega10_cache_info); + break; case CHIP_VEGA12: + pcache_info = vega12_cache_info; + num_of_cache_types = ARRAY_SIZE(vega12_cache_info); + break; case CHIP_VEGA20: case CHIP_ARCTURUS: + pcache_info = vega20_cache_info; + num_of_cache_types = ARRAY_SIZE(vega20_cache_info); + break; case CHIP_ALDEBARAN: - pcache_info = vega10_cache_info; - num_of_cache_types = ARRAY_SIZE(vega10_cache_info); + pcache_info = aldebaran_cache_info; + num_of_cache_types = ARRAY_SIZE(aldebaran_cache_info); break; case CHIP_RAVEN: pcache_info = raven_cache_info; @@ -680,12 +1299,24 @@ static int kfd_fill_gpu_cache_info(struct kfd_dev *kdev, break; case CHIP_NAVI10: case CHIP_NAVI12: + pcache_info = navi10_cache_info; + num_of_cache_types = ARRAY_SIZE(navi10_cache_info); + break; case CHIP_NAVI14: + pcache_info = navi14_cache_info; + num_of_cache_types = ARRAY_SIZE(navi14_cache_info); + break; case CHIP_SIENNA_CICHLID: + pcache_info = sienna_cichlid_cache_info; + num_of_cache_types = ARRAY_SIZE(sienna_cichlid_cache_info); + break; case CHIP_NAVY_FLOUNDER: + pcache_info = navy_flounder_cache_info; + num_of_cache_types = ARRAY_SIZE(navy_flounder_cache_info); + break; case CHIP_DIMGREY_CAVEFISH: - pcache_info = navi10_cache_info; - num_of_cache_types = ARRAY_SIZE(navi10_cache_info); + pcache_info = dimgrey_cavefish_cache_info; + num_of_cache_types = ARRAY_SIZE(dimgrey_cavefish_cache_info); break; case CHIP_VANGOGH: pcache_info = vangogh_cache_info; @@ -709,40 +1340,58 @@ static int kfd_fill_gpu_cache_info(struct kfd_dev *kdev, */ for (ct = 0; ct < num_of_cache_types; ct++) { - cu_processor_id = gpu_processor_id; - for (i = 0; i < cu_info->num_shader_engines; i++) { - for (j = 0; j < cu_info->num_shader_arrays_per_engine; - j++) { - for (k = 0; k < cu_info->num_cu_per_sh; - k += pcache_info[ct].num_cu_shared) { - - ret = fill_in_pcache(pcache, - pcache_info, - cu_info, - mem_available, - cu_info->cu_bitmap[i % 4][j + i / 4], - ct, - cu_processor_id, - k); - - if (ret < 0) - break; - - if (!ret) { - pcache++; - (*num_of_entries)++; - mem_available -= - sizeof(*pcache); - (*size_filled) += - sizeof(*pcache); - } - - /* Move to next CU block */ - cu_processor_id += - pcache_info[ct].num_cu_shared; - } - } + cu_processor_id = gpu_processor_id; + if (pcache_info[ct].cache_level == 1) { + for (i = 0; i < cu_info->num_shader_engines; i++) { + for (j = 0; j < cu_info->num_shader_arrays_per_engine; j++) { + for (k = 0; k < cu_info->num_cu_per_sh; + k += pcache_info[ct].num_cu_shared) { + ret = fill_in_l1_pcache(pcache, + pcache_info, + cu_info, + mem_available, + cu_info->cu_bitmap[i % 4][j + i / 4], + ct, + cu_processor_id, + k); + + if (ret < 0) + break; + + if (!ret) { + pcache++; + (*num_of_entries)++; + mem_available -= sizeof(*pcache); + (*size_filled) += sizeof(*pcache); + } + + /* Move to next CU block */ + num_cu_shared = ((k + pcache_info[ct].num_cu_shared) <= + cu_info->num_cu_per_sh) ? + pcache_info[ct].num_cu_shared : + (cu_info->num_cu_per_sh - k); + cu_processor_id += num_cu_shared; } + } + } + } else { + ret = fill_in_l2_l3_pcache(pcache, + pcache_info, + cu_info, + mem_available, + ct, + cu_processor_id); + + if (ret < 0) + break; + + if (!ret) { + pcache++; + (*num_of_entries)++; + mem_available -= sizeof(*pcache); + (*size_filled) += sizeof(*pcache); + } + } } pr_debug("Added [%d] GPU cache entries\n", *num_of_entries); @@ -1100,6 +1749,92 @@ static int kfd_fill_gpu_memory_affinity(int *avail_size, return 0; } +#ifdef CONFIG_ACPI_NUMA +static void kfd_find_numa_node_in_srat(struct kfd_dev *kdev) +{ + struct acpi_table_header *table_header = NULL; + struct acpi_subtable_header *sub_header = NULL; + unsigned long table_end, subtable_len; + u32 pci_id = pci_domain_nr(kdev->pdev->bus) << 16 | + pci_dev_id(kdev->pdev); + u32 bdf; + acpi_status status; + struct acpi_srat_cpu_affinity *cpu; + struct acpi_srat_generic_affinity *gpu; + int pxm = 0, max_pxm = 0; + int numa_node = NUMA_NO_NODE; + bool found = false; + + /* Fetch the SRAT table from ACPI */ + status = acpi_get_table(ACPI_SIG_SRAT, 0, &table_header); + if (status == AE_NOT_FOUND) { + pr_warn("SRAT table not found\n"); + return; + } else if (ACPI_FAILURE(status)) { + const char *err = acpi_format_exception(status); + pr_err("SRAT table error: %s\n", err); + return; + } + + table_end = (unsigned long)table_header + table_header->length; + + /* Parse all entries looking for a match. */ + sub_header = (struct acpi_subtable_header *) + ((unsigned long)table_header + + sizeof(struct acpi_table_srat)); + subtable_len = sub_header->length; + + while (((unsigned long)sub_header) + subtable_len < table_end) { + /* + * If length is 0, break from this loop to avoid + * infinite loop. + */ + if (subtable_len == 0) { + pr_err("SRAT invalid zero length\n"); + break; + } + + switch (sub_header->type) { + case ACPI_SRAT_TYPE_CPU_AFFINITY: + cpu = (struct acpi_srat_cpu_affinity *)sub_header; + pxm = *((u32 *)cpu->proximity_domain_hi) << 8 | + cpu->proximity_domain_lo; + if (pxm > max_pxm) + max_pxm = pxm; + break; + case ACPI_SRAT_TYPE_GENERIC_AFFINITY: + gpu = (struct acpi_srat_generic_affinity *)sub_header; + bdf = *((u16 *)(&gpu->device_handle[0])) << 16 | + *((u16 *)(&gpu->device_handle[2])); + if (bdf == pci_id) { + found = true; + numa_node = pxm_to_node(gpu->proximity_domain); + } + break; + default: + break; + } + + if (found) + break; + + sub_header = (struct acpi_subtable_header *) + ((unsigned long)sub_header + subtable_len); + subtable_len = sub_header->length; + } + + acpi_put_table(table_header); + + /* Workaround bad cpu-gpu binding case */ + if (found && (numa_node < 0 || + numa_node > pxm_to_node(max_pxm))) + numa_node = 0; + + if (numa_node != NUMA_NO_NODE) + set_dev_node(&kdev->pdev->dev, numa_node); +} +#endif + /* kfd_fill_gpu_direct_io_link - Fill in direct io link from GPU * to its NUMA node * @avail_size: Available size in the memory @@ -1140,11 +1875,17 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size, */ sub_type_hdr->flags |= CRAT_IOLINK_FLAGS_BI_DIRECTIONAL; sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_XGMI; + sub_type_hdr->num_hops_xgmi = 1; } else { sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_PCIEXPRESS; } sub_type_hdr->proximity_domain_from = proximity_domain; + +#ifdef CONFIG_ACPI_NUMA + if (kdev->pdev->dev.numa_node == NUMA_NO_NODE) + kfd_find_numa_node_in_srat(kdev); +#endif #ifdef CONFIG_NUMA if (kdev->pdev->dev.numa_node == NUMA_NO_NODE) sub_type_hdr->proximity_domain_to = 0; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index 357b9bf62a1c..dedb8e33b953 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -26,10 +26,12 @@ #include "kfd_priv.h" #include "kfd_device_queue_manager.h" #include "kfd_pm4_headers_vi.h" +#include "kfd_pm4_headers_aldebaran.h" #include "cwsr_trap_handler.h" #include "kfd_iommu.h" #include "amdgpu_amdkfd.h" #include "kfd_smi_events.h" +#include "kfd_migrate.h" #define MQD_SIZE_ALIGNED 768 @@ -576,7 +578,7 @@ static const struct kfd_device_info *kfd_supported_devices[][2] = { [CHIP_VEGA20] = {&vega20_device_info, NULL}, [CHIP_RENOIR] = {&renoir_device_info, NULL}, [CHIP_ARCTURUS] = {&arcturus_device_info, &arcturus_device_info}, - [CHIP_ALDEBARAN] = {&aldebaran_device_info, NULL}, + [CHIP_ALDEBARAN] = {&aldebaran_device_info, &aldebaran_device_info}, [CHIP_NAVI10] = {&navi10_device_info, NULL}, [CHIP_NAVI12] = {&navi12_device_info, &navi12_device_info}, [CHIP_NAVI14] = {&navi14_device_info, NULL}, @@ -697,7 +699,9 @@ static int kfd_gws_init(struct kfd_dev *kfd) && kfd->device_info->asic_family <= CHIP_RAVEN && kfd->mec2_fw_version >= 0x1b3) || (kfd->device_info->asic_family == CHIP_ARCTURUS - && kfd->mec2_fw_version >= 0x30)) + && kfd->mec2_fw_version >= 0x30) + || (kfd->device_info->asic_family == CHIP_ALDEBARAN + && kfd->mec2_fw_version >= 0x28)) ret = amdgpu_amdkfd_alloc_gws(kfd->kgd, amdgpu_amdkfd_get_num_gws(kfd->kgd), &kfd->gws); @@ -713,7 +717,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, struct drm_device *ddev, const struct kgd2kfd_shared_resources *gpu_resources) { - unsigned int size; + unsigned int size, map_process_packet_size; kfd->ddev = ddev; kfd->mec_fw_version = amdgpu_amdkfd_get_fw_version(kfd->kgd, @@ -748,7 +752,11 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, * calculate max size of runlist packet. * There can be only 2 packets at once */ - size += (KFD_MAX_NUM_OF_PROCESSES * sizeof(struct pm4_mes_map_process) + + map_process_packet_size = + kfd->device_info->asic_family == CHIP_ALDEBARAN ? + sizeof(struct pm4_mes_map_process_aldebaran) : + sizeof(struct pm4_mes_map_process); + size += (KFD_MAX_NUM_OF_PROCESSES * map_process_packet_size + max_num_of_queues_per_device * sizeof(struct pm4_mes_map_queues) + sizeof(struct pm4_mes_runlist)) * 2; @@ -814,6 +822,8 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, kfd_cwsr_init(kfd); + svm_migrate_init((struct amdgpu_device *)kfd->kgd); + if (kfd_resume(kfd)) goto kfd_resume_error; @@ -862,6 +872,7 @@ void kgd2kfd_device_exit(struct kfd_dev *kfd) { if (kfd->init_complete) { kgd2kfd_suspend(kfd, false); + svm_migrate_fini((struct amdgpu_device *)kfd->kgd); device_queue_manager_uninit(kfd->dqm); kfd_interrupt_exit(kfd); kfd_topology_remove_device(kfd); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index d3eaa1549bd7..98c2046c7331 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -738,7 +738,7 @@ static int restore_process_queues_nocpsch(struct device_queue_manager *dqm, pdd = qpd_to_pdd(qpd); /* Retrieve PD base */ - pd_base = amdgpu_amdkfd_gpuvm_get_process_page_dir(pdd->vm); + pd_base = amdgpu_amdkfd_gpuvm_get_process_page_dir(pdd->drm_priv); dqm_lock(dqm); if (WARN_ON_ONCE(!qpd->evicted)) /* already restored, do nothing */ @@ -821,7 +821,7 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm, pdd = qpd_to_pdd(qpd); /* Retrieve PD base */ - pd_base = amdgpu_amdkfd_gpuvm_get_process_page_dir(pdd->vm); + pd_base = amdgpu_amdkfd_gpuvm_get_process_page_dir(pdd->drm_priv); dqm_lock(dqm); if (WARN_ON_ONCE(!qpd->evicted)) /* already restored, do nothing */ @@ -873,7 +873,7 @@ static int register_process(struct device_queue_manager *dqm, pdd = qpd_to_pdd(qpd); /* Retrieve PD base */ - pd_base = amdgpu_amdkfd_gpuvm_get_process_page_dir(pdd->vm); + pd_base = amdgpu_amdkfd_gpuvm_get_process_page_dir(pdd->drm_priv); dqm_lock(dqm); list_add(&n->list, &dqm->queues); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c index eca6331efa94..b5c3d13643f1 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c @@ -61,10 +61,19 @@ static int update_qpd_v9(struct device_queue_manager *dqm, qpd->sh_mem_config = SH_MEM_ALIGNMENT_MODE_UNALIGNED << SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT; - if (dqm->dev->noretry && - !dqm->dev->use_iommu_v2) + + if (dqm->dev->device_info->asic_family == CHIP_ALDEBARAN) { + /* Aldebaran can safely support different XNACK modes + * per process + */ + if (!pdd->process->xnack_enabled) + qpd->sh_mem_config |= + 1 << SH_MEM_CONFIG__RETRY_DISABLE__SHIFT; + } else if (dqm->dev->noretry && + !dqm->dev->use_iommu_v2) { qpd->sh_mem_config |= 1 << SH_MEM_CONFIG__RETRY_DISABLE__SHIFT; + } qpd->sh_mem_ape1_limit = 0; qpd->sh_mem_ape1_base = 0; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c index a2c9063076cc..72815e86a3b8 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c @@ -405,6 +405,10 @@ int kfd_init_apertures(struct kfd_process *process) case CHIP_POLARIS12: case CHIP_VEGAM: kfd_init_apertures_vi(pdd, id); + /* VI GPUs cannot support SVM with only + * 40 bits of virtual address space. + */ + process->svm_disabled = true; break; case CHIP_VEGA10: case CHIP_VEGA12: diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c index 696944fa0177..97c36e3c8c80 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c @@ -25,7 +25,6 @@ #include "soc15_int.h" #include "kfd_device_queue_manager.h" #include "kfd_smi_events.h" -#include "amdgpu.h" enum SQ_INTERRUPT_WORD_ENCODING { SQ_INTERRUPT_WORD_ENCODING_AUTO = 0x0, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c new file mode 100644 index 000000000000..fd8f544f0de2 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c @@ -0,0 +1,937 @@ +// SPDX-License-Identifier: GPL-2.0 OR MIT +/* + * Copyright 2020-2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include <linux/types.h> +#include <linux/hmm.h> +#include <linux/dma-direction.h> +#include <linux/dma-mapping.h> +#include "amdgpu_sync.h" +#include "amdgpu_object.h" +#include "amdgpu_vm.h" +#include "amdgpu_mn.h" +#include "kfd_priv.h" +#include "kfd_svm.h" +#include "kfd_migrate.h" + +static uint64_t +svm_migrate_direct_mapping_addr(struct amdgpu_device *adev, uint64_t addr) +{ + return addr + amdgpu_ttm_domain_start(adev, TTM_PL_VRAM); +} + +static int +svm_migrate_gart_map(struct amdgpu_ring *ring, uint64_t npages, + dma_addr_t *addr, uint64_t *gart_addr, uint64_t flags) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_job *job; + unsigned int num_dw, num_bytes; + struct dma_fence *fence; + uint64_t src_addr, dst_addr; + uint64_t pte_flags; + void *cpu_addr; + int r; + + /* use gart window 0 */ + *gart_addr = adev->gmc.gart_start; + + num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8); + num_bytes = npages * 8; + + r = amdgpu_job_alloc_with_ib(adev, num_dw * 4 + num_bytes, + AMDGPU_IB_POOL_DELAYED, &job); + if (r) + return r; + + src_addr = num_dw * 4; + src_addr += job->ibs[0].gpu_addr; + + dst_addr = amdgpu_bo_gpu_offset(adev->gart.bo); + amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_addr, + dst_addr, num_bytes, false); + + amdgpu_ring_pad_ib(ring, &job->ibs[0]); + WARN_ON(job->ibs[0].length_dw > num_dw); + + pte_flags = AMDGPU_PTE_VALID | AMDGPU_PTE_READABLE; + pte_flags |= AMDGPU_PTE_SYSTEM | AMDGPU_PTE_SNOOPED; + if (!(flags & KFD_IOCTL_SVM_FLAG_GPU_RO)) + pte_flags |= AMDGPU_PTE_WRITEABLE; + pte_flags |= adev->gart.gart_pte_flags; + + cpu_addr = &job->ibs[0].ptr[num_dw]; + + r = amdgpu_gart_map(adev, 0, npages, addr, pte_flags, cpu_addr); + if (r) + goto error_free; + + r = amdgpu_job_submit(job, &adev->mman.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, &fence); + if (r) + goto error_free; + + dma_fence_put(fence); + + return r; + +error_free: + amdgpu_job_free(job); + return r; +} + +/** + * svm_migrate_copy_memory_gart - sdma copy data between ram and vram + * + * @adev: amdgpu device the sdma ring running + * @src: source page address array + * @dst: destination page address array + * @npages: number of pages to copy + * @direction: enum MIGRATION_COPY_DIR + * @mfence: output, sdma fence to signal after sdma is done + * + * ram address uses GART table continuous entries mapping to ram pages, + * vram address uses direct mapping of vram pages, which must have npages + * number of continuous pages. + * GART update and sdma uses same buf copy function ring, sdma is splited to + * multiple GTT_MAX_PAGES transfer, all sdma operations are serialized, wait for + * the last sdma finish fence which is returned to check copy memory is done. + * + * Context: Process context, takes and releases gtt_window_lock + * + * Return: + * 0 - OK, otherwise error code + */ + +static int +svm_migrate_copy_memory_gart(struct amdgpu_device *adev, dma_addr_t *sys, + uint64_t *vram, uint64_t npages, + enum MIGRATION_COPY_DIR direction, + struct dma_fence **mfence) +{ + const uint64_t GTT_MAX_PAGES = AMDGPU_GTT_MAX_TRANSFER_SIZE; + struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; + uint64_t gart_s, gart_d; + struct dma_fence *next; + uint64_t size; + int r; + + mutex_lock(&adev->mman.gtt_window_lock); + + while (npages) { + size = min(GTT_MAX_PAGES, npages); + + if (direction == FROM_VRAM_TO_RAM) { + gart_s = svm_migrate_direct_mapping_addr(adev, *vram); + r = svm_migrate_gart_map(ring, size, sys, &gart_d, 0); + + } else if (direction == FROM_RAM_TO_VRAM) { + r = svm_migrate_gart_map(ring, size, sys, &gart_s, + KFD_IOCTL_SVM_FLAG_GPU_RO); + gart_d = svm_migrate_direct_mapping_addr(adev, *vram); + } + if (r) { + pr_debug("failed %d to create gart mapping\n", r); + goto out_unlock; + } + + r = amdgpu_copy_buffer(ring, gart_s, gart_d, size * PAGE_SIZE, + NULL, &next, false, true, false); + if (r) { + pr_debug("failed %d to copy memory\n", r); + goto out_unlock; + } + + dma_fence_put(*mfence); + *mfence = next; + npages -= size; + if (npages) { + sys += size; + vram += size; + } + } + +out_unlock: + mutex_unlock(&adev->mman.gtt_window_lock); + + return r; +} + +/** + * svm_migrate_copy_done - wait for memory copy sdma is done + * + * @adev: amdgpu device the sdma memory copy is executing on + * @mfence: migrate fence + * + * Wait for dma fence is signaled, if the copy ssplit into multiple sdma + * operations, this is the last sdma operation fence. + * + * Context: called after svm_migrate_copy_memory + * + * Return: + * 0 - success + * otherwise - error code from dma fence signal + */ +static int +svm_migrate_copy_done(struct amdgpu_device *adev, struct dma_fence *mfence) +{ + int r = 0; + + if (mfence) { + r = dma_fence_wait(mfence, false); + dma_fence_put(mfence); + pr_debug("sdma copy memory fence done\n"); + } + + return r; +} + +static uint64_t +svm_migrate_node_physical_addr(struct amdgpu_device *adev, + struct drm_mm_node **mm_node, uint64_t *offset) +{ + struct drm_mm_node *node = *mm_node; + uint64_t pos = *offset; + + if (node->start == AMDGPU_BO_INVALID_OFFSET) { + pr_debug("drm node is not validated\n"); + return 0; + } + + pr_debug("vram node start 0x%llx npages 0x%llx\n", node->start, + node->size); + + if (pos >= node->size) { + do { + pos -= node->size; + node++; + } while (pos >= node->size); + + *mm_node = node; + *offset = pos; + } + + return (node->start + pos) << PAGE_SHIFT; +} + +unsigned long +svm_migrate_addr_to_pfn(struct amdgpu_device *adev, unsigned long addr) +{ + return (addr + adev->kfd.dev->pgmap.range.start) >> PAGE_SHIFT; +} + +static void +svm_migrate_get_vram_page(struct svm_range *prange, unsigned long pfn) +{ + struct page *page; + + page = pfn_to_page(pfn); + page->zone_device_data = prange; + get_page(page); + lock_page(page); +} + +static void +svm_migrate_put_vram_page(struct amdgpu_device *adev, unsigned long addr) +{ + struct page *page; + + page = pfn_to_page(svm_migrate_addr_to_pfn(adev, addr)); + unlock_page(page); + put_page(page); +} + +static unsigned long +svm_migrate_addr(struct amdgpu_device *adev, struct page *page) +{ + unsigned long addr; + + addr = page_to_pfn(page) << PAGE_SHIFT; + return (addr - adev->kfd.dev->pgmap.range.start); +} + +static struct page * +svm_migrate_get_sys_page(struct vm_area_struct *vma, unsigned long addr) +{ + struct page *page; + + page = alloc_page_vma(GFP_HIGHUSER, vma, addr); + if (page) + lock_page(page); + + return page; +} + +static void svm_migrate_put_sys_page(unsigned long addr) +{ + struct page *page; + + page = pfn_to_page(addr >> PAGE_SHIFT); + unlock_page(page); + put_page(page); +} + +static int +svm_migrate_copy_to_vram(struct amdgpu_device *adev, struct svm_range *prange, + struct migrate_vma *migrate, struct dma_fence **mfence, + dma_addr_t *scratch) +{ + uint64_t npages = migrate->cpages; + struct device *dev = adev->dev; + struct drm_mm_node *node; + dma_addr_t *src; + uint64_t *dst; + uint64_t vram_addr; + uint64_t offset; + uint64_t i, j; + int r; + + pr_debug("svms 0x%p [0x%lx 0x%lx]\n", prange->svms, prange->start, + prange->last); + + src = scratch; + dst = (uint64_t *)(scratch + npages); + + r = svm_range_vram_node_new(adev, prange, true); + if (r) { + pr_debug("failed %d get 0x%llx pages from vram\n", r, npages); + goto out; + } + + node = prange->ttm_res->mm_node; + offset = prange->offset; + vram_addr = svm_migrate_node_physical_addr(adev, &node, &offset); + if (!vram_addr) { + WARN_ONCE(1, "vram node address is 0\n"); + r = -ENOMEM; + goto out; + } + + for (i = j = 0; i < npages; i++) { + struct page *spage; + + dst[i] = vram_addr + (j << PAGE_SHIFT); + migrate->dst[i] = svm_migrate_addr_to_pfn(adev, dst[i]); + svm_migrate_get_vram_page(prange, migrate->dst[i]); + + migrate->dst[i] = migrate_pfn(migrate->dst[i]); + migrate->dst[i] |= MIGRATE_PFN_LOCKED; + + if (migrate->src[i] & MIGRATE_PFN_VALID) { + spage = migrate_pfn_to_page(migrate->src[i]); + src[i] = dma_map_page(dev, spage, 0, PAGE_SIZE, + DMA_TO_DEVICE); + r = dma_mapping_error(dev, src[i]); + if (r) { + pr_debug("failed %d dma_map_page\n", r); + goto out_free_vram_pages; + } + } else { + if (j) { + r = svm_migrate_copy_memory_gart( + adev, src + i - j, + dst + i - j, j, + FROM_RAM_TO_VRAM, + mfence); + if (r) + goto out_free_vram_pages; + offset += j; + vram_addr = (node->start + offset) << PAGE_SHIFT; + j = 0; + } else { + offset++; + vram_addr += PAGE_SIZE; + } + if (offset >= node->size) { + node++; + pr_debug("next node size 0x%llx\n", node->size); + vram_addr = node->start << PAGE_SHIFT; + offset = 0; + } + continue; + } + + pr_debug("dma mapping src to 0x%llx, page_to_pfn 0x%lx\n", + src[i] >> PAGE_SHIFT, page_to_pfn(spage)); + + if (j + offset >= node->size - 1 && i < npages - 1) { + r = svm_migrate_copy_memory_gart(adev, src + i - j, + dst + i - j, j + 1, + FROM_RAM_TO_VRAM, + mfence); + if (r) + goto out_free_vram_pages; + + node++; + pr_debug("next node size 0x%llx\n", node->size); + vram_addr = node->start << PAGE_SHIFT; + offset = 0; + j = 0; + } else { + j++; + } + } + + r = svm_migrate_copy_memory_gart(adev, src + i - j, dst + i - j, j, + FROM_RAM_TO_VRAM, mfence); + +out_free_vram_pages: + if (r) { + pr_debug("failed %d to copy memory to vram\n", r); + while (i--) { + svm_migrate_put_vram_page(adev, dst[i]); + migrate->dst[i] = 0; + } + } + +out: + return r; +} + +static int +svm_migrate_vma_to_vram(struct amdgpu_device *adev, struct svm_range *prange, + struct vm_area_struct *vma, uint64_t start, + uint64_t end) +{ + uint64_t npages = (end - start) >> PAGE_SHIFT; + struct dma_fence *mfence = NULL; + struct migrate_vma migrate; + dma_addr_t *scratch; + size_t size; + void *buf; + int r = -ENOMEM; + int retry = 0; + + memset(&migrate, 0, sizeof(migrate)); + migrate.vma = vma; + migrate.start = start; + migrate.end = end; + migrate.flags = MIGRATE_VMA_SELECT_SYSTEM; + migrate.pgmap_owner = adev; + + size = 2 * sizeof(*migrate.src) + sizeof(uint64_t) + sizeof(dma_addr_t); + size *= npages; + buf = kvmalloc(size, GFP_KERNEL | __GFP_ZERO); + if (!buf) + goto out; + + migrate.src = buf; + migrate.dst = migrate.src + npages; + scratch = (dma_addr_t *)(migrate.dst + npages); + +retry: + r = migrate_vma_setup(&migrate); + if (r) { + pr_debug("failed %d prepare migrate svms 0x%p [0x%lx 0x%lx]\n", + r, prange->svms, prange->start, prange->last); + goto out_free; + } + if (migrate.cpages != npages) { + pr_debug("collect 0x%lx/0x%llx pages, retry\n", migrate.cpages, + npages); + migrate_vma_finalize(&migrate); + if (retry++ >= 3) { + r = -ENOMEM; + pr_debug("failed %d migrate svms 0x%p [0x%lx 0x%lx]\n", + r, prange->svms, prange->start, prange->last); + goto out_free; + } + + goto retry; + } + + if (migrate.cpages) { + r = svm_migrate_copy_to_vram(adev, prange, &migrate, &mfence, + scratch); + migrate_vma_pages(&migrate); + svm_migrate_copy_done(adev, mfence); + migrate_vma_finalize(&migrate); + } + + svm_range_dma_unmap(adev->dev, scratch, 0, npages); + svm_range_free_dma_mappings(prange); + +out_free: + kvfree(buf); +out: + return r; +} + +/** + * svm_migrate_ram_to_vram - migrate svm range from system to device + * @prange: range structure + * @best_loc: the device to migrate to + * @mm: the process mm structure + * + * Context: Process context, caller hold mmap read lock, svms lock, prange lock + * + * Return: + * 0 - OK, otherwise error code + */ +static int +svm_migrate_ram_to_vram(struct svm_range *prange, uint32_t best_loc, + struct mm_struct *mm) +{ + unsigned long addr, start, end; + struct vm_area_struct *vma; + struct amdgpu_device *adev; + int r = 0; + + if (prange->actual_loc == best_loc) { + pr_debug("svms 0x%p [0x%lx 0x%lx] already on best_loc 0x%x\n", + prange->svms, prange->start, prange->last, best_loc); + return 0; + } + + adev = svm_range_get_adev_by_id(prange, best_loc); + if (!adev) { + pr_debug("failed to get device by id 0x%x\n", best_loc); + return -ENODEV; + } + + pr_debug("svms 0x%p [0x%lx 0x%lx] to gpu 0x%x\n", prange->svms, + prange->start, prange->last, best_loc); + + /* FIXME: workaround for page locking bug with invalid pages */ + svm_range_prefault(prange, mm); + + start = prange->start << PAGE_SHIFT; + end = (prange->last + 1) << PAGE_SHIFT; + + for (addr = start; addr < end;) { + unsigned long next; + + vma = find_vma(mm, addr); + if (!vma || addr < vma->vm_start) + break; + + next = min(vma->vm_end, end); + r = svm_migrate_vma_to_vram(adev, prange, vma, addr, next); + if (r) { + pr_debug("failed to migrate\n"); + break; + } + addr = next; + } + + if (!r) + prange->actual_loc = best_loc; + + return r; +} + +static void svm_migrate_page_free(struct page *page) +{ + /* Keep this function to avoid warning */ +} + +static int +svm_migrate_copy_to_ram(struct amdgpu_device *adev, struct svm_range *prange, + struct migrate_vma *migrate, struct dma_fence **mfence, + dma_addr_t *scratch) +{ + uint64_t npages = migrate->cpages; + struct device *dev = adev->dev; + uint64_t *src; + dma_addr_t *dst; + struct page *dpage; + uint64_t i = 0, j; + uint64_t addr; + int r = 0; + + pr_debug("svms 0x%p [0x%lx 0x%lx]\n", prange->svms, prange->start, + prange->last); + + addr = prange->start << PAGE_SHIFT; + + src = (uint64_t *)(scratch + npages); + dst = scratch; + + for (i = 0, j = 0; i < npages; i++, j++, addr += PAGE_SIZE) { + struct page *spage; + + spage = migrate_pfn_to_page(migrate->src[i]); + if (!spage) { + pr_debug("failed get spage svms 0x%p [0x%lx 0x%lx]\n", + prange->svms, prange->start, prange->last); + r = -ENOMEM; + goto out_oom; + } + src[i] = svm_migrate_addr(adev, spage); + if (i > 0 && src[i] != src[i - 1] + PAGE_SIZE) { + r = svm_migrate_copy_memory_gart(adev, dst + i - j, + src + i - j, j, + FROM_VRAM_TO_RAM, + mfence); + if (r) + goto out_oom; + j = 0; + } + + dpage = svm_migrate_get_sys_page(migrate->vma, addr); + if (!dpage) { + pr_debug("failed get page svms 0x%p [0x%lx 0x%lx]\n", + prange->svms, prange->start, prange->last); + r = -ENOMEM; + goto out_oom; + } + + dst[i] = dma_map_page(dev, dpage, 0, PAGE_SIZE, DMA_FROM_DEVICE); + r = dma_mapping_error(dev, dst[i]); + if (r) { + pr_debug("failed %d dma_map_page\n", r); + goto out_oom; + } + + pr_debug("dma mapping dst to 0x%llx, page_to_pfn 0x%lx\n", + dst[i] >> PAGE_SHIFT, page_to_pfn(dpage)); + + migrate->dst[i] = migrate_pfn(page_to_pfn(dpage)); + migrate->dst[i] |= MIGRATE_PFN_LOCKED; + } + + r = svm_migrate_copy_memory_gart(adev, dst + i - j, src + i - j, j, + FROM_VRAM_TO_RAM, mfence); + +out_oom: + if (r) { + pr_debug("failed %d copy to ram\n", r); + while (i--) { + svm_migrate_put_sys_page(dst[i]); + migrate->dst[i] = 0; + } + } + + return r; +} + +static int +svm_migrate_vma_to_ram(struct amdgpu_device *adev, struct svm_range *prange, + struct vm_area_struct *vma, uint64_t start, uint64_t end) +{ + uint64_t npages = (end - start) >> PAGE_SHIFT; + struct dma_fence *mfence = NULL; + struct migrate_vma migrate; + dma_addr_t *scratch; + size_t size; + void *buf; + int r = -ENOMEM; + + memset(&migrate, 0, sizeof(migrate)); + migrate.vma = vma; + migrate.start = start; + migrate.end = end; + migrate.flags = MIGRATE_VMA_SELECT_DEVICE_PRIVATE; + migrate.pgmap_owner = adev; + + size = 2 * sizeof(*migrate.src) + sizeof(uint64_t) + sizeof(dma_addr_t); + size *= npages; + buf = kvmalloc(size, GFP_KERNEL | __GFP_ZERO); + if (!buf) + goto out; + + migrate.src = buf; + migrate.dst = migrate.src + npages; + scratch = (dma_addr_t *)(migrate.dst + npages); + + r = migrate_vma_setup(&migrate); + if (r) { + pr_debug("failed %d prepare migrate svms 0x%p [0x%lx 0x%lx]\n", + r, prange->svms, prange->start, prange->last); + goto out_free; + } + + pr_debug("cpages %ld\n", migrate.cpages); + + if (migrate.cpages) { + r = svm_migrate_copy_to_ram(adev, prange, &migrate, &mfence, + scratch); + migrate_vma_pages(&migrate); + svm_migrate_copy_done(adev, mfence); + migrate_vma_finalize(&migrate); + } else { + pr_debug("failed collect migrate device pages [0x%lx 0x%lx]\n", + prange->start, prange->last); + } + + svm_range_dma_unmap(adev->dev, scratch, 0, npages); + +out_free: + kvfree(buf); +out: + return r; +} + +/** + * svm_migrate_vram_to_ram - migrate svm range from device to system + * @prange: range structure + * @mm: process mm, use current->mm if NULL + * + * Context: Process context, caller hold mmap read lock, svms lock, prange lock + * + * Return: + * 0 - OK, otherwise error code + */ +int svm_migrate_vram_to_ram(struct svm_range *prange, struct mm_struct *mm) +{ + struct amdgpu_device *adev; + struct vm_area_struct *vma; + unsigned long addr; + unsigned long start; + unsigned long end; + int r = 0; + + if (!prange->actual_loc) { + pr_debug("[0x%lx 0x%lx] already migrated to ram\n", + prange->start, prange->last); + return 0; + } + + adev = svm_range_get_adev_by_id(prange, prange->actual_loc); + if (!adev) { + pr_debug("failed to get device by id 0x%x\n", + prange->actual_loc); + return -ENODEV; + } + + pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx] from gpu 0x%x to ram\n", + prange->svms, prange, prange->start, prange->last, + prange->actual_loc); + + start = prange->start << PAGE_SHIFT; + end = (prange->last + 1) << PAGE_SHIFT; + + for (addr = start; addr < end;) { + unsigned long next; + + vma = find_vma(mm, addr); + if (!vma || addr < vma->vm_start) + break; + + next = min(vma->vm_end, end); + r = svm_migrate_vma_to_ram(adev, prange, vma, addr, next); + if (r) { + pr_debug("failed %d to migrate\n", r); + break; + } + addr = next; + } + + if (!r) { + svm_range_vram_node_free(prange); + prange->actual_loc = 0; + } + return r; +} + +/** + * svm_migrate_vram_to_vram - migrate svm range from device to device + * @prange: range structure + * @best_loc: the device to migrate to + * @mm: process mm, use current->mm if NULL + * + * Context: Process context, caller hold mmap read lock, svms lock, prange lock + * + * Return: + * 0 - OK, otherwise error code + */ +static int +svm_migrate_vram_to_vram(struct svm_range *prange, uint32_t best_loc, + struct mm_struct *mm) +{ + int r; + + /* + * TODO: for both devices with PCIe large bar or on same xgmi hive, skip + * system memory as migration bridge + */ + + pr_debug("from gpu 0x%x to gpu 0x%x\n", prange->actual_loc, best_loc); + + r = svm_migrate_vram_to_ram(prange, mm); + if (r) + return r; + + return svm_migrate_ram_to_vram(prange, best_loc, mm); +} + +int +svm_migrate_to_vram(struct svm_range *prange, uint32_t best_loc, + struct mm_struct *mm) +{ + if (!prange->actual_loc) + return svm_migrate_ram_to_vram(prange, best_loc, mm); + else + return svm_migrate_vram_to_vram(prange, best_loc, mm); + +} + +/** + * svm_migrate_to_ram - CPU page fault handler + * @vmf: CPU vm fault vma, address + * + * Context: vm fault handler, caller holds the mmap read lock + * + * Return: + * 0 - OK + * VM_FAULT_SIGBUS - notice application to have SIGBUS page fault + */ +static vm_fault_t svm_migrate_to_ram(struct vm_fault *vmf) +{ + unsigned long addr = vmf->address; + struct vm_area_struct *vma; + enum svm_work_list_ops op; + struct svm_range *parent; + struct svm_range *prange; + struct kfd_process *p; + struct mm_struct *mm; + int r = 0; + + vma = vmf->vma; + mm = vma->vm_mm; + + p = kfd_lookup_process_by_mm(vma->vm_mm); + if (!p) { + pr_debug("failed find process at fault address 0x%lx\n", addr); + return VM_FAULT_SIGBUS; + } + addr >>= PAGE_SHIFT; + pr_debug("CPU page fault svms 0x%p address 0x%lx\n", &p->svms, addr); + + mutex_lock(&p->svms.lock); + + prange = svm_range_from_addr(&p->svms, addr, &parent); + if (!prange) { + pr_debug("cannot find svm range at 0x%lx\n", addr); + r = -EFAULT; + goto out; + } + + mutex_lock(&parent->migrate_mutex); + if (prange != parent) + mutex_lock_nested(&prange->migrate_mutex, 1); + + if (!prange->actual_loc) + goto out_unlock_prange; + + svm_range_lock(parent); + if (prange != parent) + mutex_lock_nested(&prange->lock, 1); + r = svm_range_split_by_granularity(p, mm, addr, parent, prange); + if (prange != parent) + mutex_unlock(&prange->lock); + svm_range_unlock(parent); + if (r) { + pr_debug("failed %d to split range by granularity\n", r); + goto out_unlock_prange; + } + + r = svm_migrate_vram_to_ram(prange, mm); + if (r) + pr_debug("failed %d migrate 0x%p [0x%lx 0x%lx] to ram\n", r, + prange, prange->start, prange->last); + + /* xnack on, update mapping on GPUs with ACCESS_IN_PLACE */ + if (p->xnack_enabled && parent == prange) + op = SVM_OP_UPDATE_RANGE_NOTIFIER_AND_MAP; + else + op = SVM_OP_UPDATE_RANGE_NOTIFIER; + svm_range_add_list_work(&p->svms, parent, mm, op); + schedule_deferred_list_work(&p->svms); + +out_unlock_prange: + if (prange != parent) + mutex_unlock(&prange->migrate_mutex); + mutex_unlock(&parent->migrate_mutex); +out: + mutex_unlock(&p->svms.lock); + kfd_unref_process(p); + + pr_debug("CPU fault svms 0x%p address 0x%lx done\n", &p->svms, addr); + + return r ? VM_FAULT_SIGBUS : 0; +} + +static const struct dev_pagemap_ops svm_migrate_pgmap_ops = { + .page_free = svm_migrate_page_free, + .migrate_to_ram = svm_migrate_to_ram, +}; + +/* Each VRAM page uses sizeof(struct page) on system memory */ +#define SVM_HMM_PAGE_STRUCT_SIZE(size) ((size)/PAGE_SIZE * sizeof(struct page)) + +int svm_migrate_init(struct amdgpu_device *adev) +{ + struct kfd_dev *kfddev = adev->kfd.dev; + struct dev_pagemap *pgmap; + struct resource *res; + unsigned long size; + void *r; + + /* Page migration works on Vega10 or newer */ + if (kfddev->device_info->asic_family < CHIP_VEGA10) + return -EINVAL; + + pgmap = &kfddev->pgmap; + memset(pgmap, 0, sizeof(*pgmap)); + + /* TODO: register all vram to HMM for now. + * should remove reserved size + */ + size = ALIGN(adev->gmc.real_vram_size, 2ULL << 20); + res = devm_request_free_mem_region(adev->dev, &iomem_resource, size); + if (IS_ERR(res)) + return -ENOMEM; + + pgmap->type = MEMORY_DEVICE_PRIVATE; + pgmap->nr_range = 1; + pgmap->range.start = res->start; + pgmap->range.end = res->end; + pgmap->ops = &svm_migrate_pgmap_ops; + pgmap->owner = adev; + pgmap->flags = MIGRATE_VMA_SELECT_DEVICE_PRIVATE; + r = devm_memremap_pages(adev->dev, pgmap); + if (IS_ERR(r)) { + pr_err("failed to register HMM device memory\n"); + devm_release_mem_region(adev->dev, res->start, + res->end - res->start + 1); + return PTR_ERR(r); + } + + pr_debug("reserve %ldMB system memory for VRAM pages struct\n", + SVM_HMM_PAGE_STRUCT_SIZE(size) >> 20); + + amdgpu_amdkfd_reserve_system_mem(SVM_HMM_PAGE_STRUCT_SIZE(size)); + + pr_info("HMM registered %ldMB device memory\n", size >> 20); + + return 0; +} + +void svm_migrate_fini(struct amdgpu_device *adev) +{ + struct dev_pagemap *pgmap = &adev->kfd.dev->pgmap; + + devm_memunmap_pages(adev->dev, pgmap); + devm_release_mem_region(adev->dev, pgmap->range.start, + pgmap->range.end - pgmap->range.start + 1); +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.h b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.h new file mode 100644 index 000000000000..0de76b5d4973 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.h @@ -0,0 +1,65 @@ +/* SPDX-License-Identifier: GPL-2.0 OR MIT */ +/* + * Copyright 2020-2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef KFD_MIGRATE_H_ +#define KFD_MIGRATE_H_ + +#if IS_ENABLED(CONFIG_HSA_AMD_SVM) + +#include <linux/rwsem.h> +#include <linux/list.h> +#include <linux/mutex.h> +#include <linux/sched/mm.h> +#include <linux/hmm.h> +#include "kfd_priv.h" +#include "kfd_svm.h" + +enum MIGRATION_COPY_DIR { + FROM_RAM_TO_VRAM = 0, + FROM_VRAM_TO_RAM +}; + +int svm_migrate_to_vram(struct svm_range *prange, uint32_t best_loc, + struct mm_struct *mm); +int svm_migrate_vram_to_ram(struct svm_range *prange, struct mm_struct *mm); +unsigned long +svm_migrate_addr_to_pfn(struct amdgpu_device *adev, unsigned long addr); + +int svm_migrate_init(struct amdgpu_device *adev); +void svm_migrate_fini(struct amdgpu_device *adev); + +#else + +static inline int svm_migrate_init(struct amdgpu_device *adev) +{ + return 0; +} +static inline void svm_migrate_fini(struct amdgpu_device *adev) +{ + /* empty */ +} + +#endif /* IS_ENABLED(CONFIG_HSA_AMD_SVM) */ + +#endif /* KFD_MIGRATE_H_ */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c index e840dd581719..f688451cb299 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c @@ -124,14 +124,14 @@ static int pm_create_runlist_ib(struct packet_manager *pm, { unsigned int alloc_size_bytes; unsigned int *rl_buffer, rl_wptr, i; - int retval, proccesses_mapped; + int retval, processes_mapped; struct device_process_node *cur; struct qcm_process_device *qpd; struct queue *q; struct kernel_queue *kq; bool is_over_subscription; - rl_wptr = retval = proccesses_mapped = 0; + rl_wptr = retval = processes_mapped = 0; retval = pm_allocate_runlist_ib(pm, &rl_buffer, rl_gpu_addr, &alloc_size_bytes, &is_over_subscription); @@ -148,7 +148,7 @@ static int pm_create_runlist_ib(struct packet_manager *pm, list_for_each_entry(cur, queues, list) { qpd = cur->qpd; /* build map process packet */ - if (proccesses_mapped >= pm->dqm->processes_count) { + if (processes_mapped >= pm->dqm->processes_count) { pr_debug("Not enough space left in runlist IB\n"); pm_release_ib(pm); return -ENOMEM; @@ -158,7 +158,7 @@ static int pm_create_runlist_ib(struct packet_manager *pm, if (retval) return retval; - proccesses_mapped++; + processes_mapped++; inc_wptr(&rl_wptr, pm->pmf->map_process_size, alloc_size_bytes); @@ -242,7 +242,6 @@ int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm) case CHIP_RAVEN: case CHIP_RENOIR: case CHIP_ARCTURUS: - case CHIP_ALDEBARAN: case CHIP_NAVI10: case CHIP_NAVI12: case CHIP_NAVI14: @@ -252,6 +251,9 @@ int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm) case CHIP_DIMGREY_CAVEFISH: pm->pmf = &kfd_v9_pm_funcs; break; + case CHIP_ALDEBARAN: + pm->pmf = &kfd_aldebaran_pm_funcs; + break; default: WARN(1, "Unexpected ASIC family %u", dqm->dev->device_info->asic_family); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c index e3ba0cd3b6fa..7ea3f671b325 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c @@ -24,6 +24,7 @@ #include "kfd_kernel_queue.h" #include "kfd_device_queue_manager.h" #include "kfd_pm4_headers_ai.h" +#include "kfd_pm4_headers_aldebaran.h" #include "kfd_pm4_opcodes.h" #include "gc/gc_10_1_0_sh_mask.h" @@ -35,7 +36,6 @@ static int pm_map_process_v9(struct packet_manager *pm, packet = (struct pm4_mes_map_process *)buffer; memset(buffer, 0, sizeof(struct pm4_mes_map_process)); - packet->header.u32All = pm_build_pm4_header(IT_MAP_PROCESS, sizeof(struct pm4_mes_map_process)); packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0; @@ -73,6 +73,45 @@ static int pm_map_process_v9(struct packet_manager *pm, return 0; } +static int pm_map_process_aldebaran(struct packet_manager *pm, + uint32_t *buffer, struct qcm_process_device *qpd) +{ + struct pm4_mes_map_process_aldebaran *packet; + uint64_t vm_page_table_base_addr = qpd->page_table_base; + + packet = (struct pm4_mes_map_process_aldebaran *)buffer; + memset(buffer, 0, sizeof(struct pm4_mes_map_process_aldebaran)); + packet->header.u32All = pm_build_pm4_header(IT_MAP_PROCESS, + sizeof(struct pm4_mes_map_process_aldebaran)); + packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0; + packet->bitfields2.process_quantum = 10; + packet->bitfields2.pasid = qpd->pqm->process->pasid; + packet->bitfields14.gds_size = qpd->gds_size & 0x3F; + packet->bitfields14.gds_size_hi = (qpd->gds_size >> 6) & 0xF; + packet->bitfields14.num_gws = (qpd->mapped_gws_queue) ? qpd->num_gws : 0; + packet->bitfields14.num_oac = qpd->num_oac; + packet->bitfields14.sdma_enable = 1; + packet->bitfields14.num_queues = (qpd->is_debug) ? 0 : qpd->queue_count; + + packet->sh_mem_config = qpd->sh_mem_config; + packet->sh_mem_bases = qpd->sh_mem_bases; + if (qpd->tba_addr) { + packet->sq_shader_tba_lo = lower_32_bits(qpd->tba_addr >> 8); + packet->sq_shader_tma_lo = lower_32_bits(qpd->tma_addr >> 8); + packet->sq_shader_tma_hi = upper_32_bits(qpd->tma_addr >> 8); + } + + packet->gds_addr_lo = lower_32_bits(qpd->gds_context_area); + packet->gds_addr_hi = upper_32_bits(qpd->gds_context_area); + + packet->vm_context_page_table_base_addr_lo32 = + lower_32_bits(vm_page_table_base_addr); + packet->vm_context_page_table_base_addr_hi32 = + upper_32_bits(vm_page_table_base_addr); + + return 0; +} + static int pm_runlist_v9(struct packet_manager *pm, uint32_t *buffer, uint64_t ib, size_t ib_size_in_dwords, bool chain) { @@ -324,3 +363,20 @@ const struct packet_manager_funcs kfd_v9_pm_funcs = { .query_status_size = sizeof(struct pm4_mes_query_status), .release_mem_size = 0, }; + +const struct packet_manager_funcs kfd_aldebaran_pm_funcs = { + .map_process = pm_map_process_aldebaran, + .runlist = pm_runlist_v9, + .set_resources = pm_set_resources_v9, + .map_queues = pm_map_queues_v9, + .unmap_queues = pm_unmap_queues_v9, + .query_status = pm_query_status_v9, + .release_mem = NULL, + .map_process_size = sizeof(struct pm4_mes_map_process_aldebaran), + .runlist_size = sizeof(struct pm4_mes_runlist), + .set_resources_size = sizeof(struct pm4_mes_set_resources), + .map_queues_size = sizeof(struct pm4_mes_map_queues), + .unmap_queues_size = sizeof(struct pm4_mes_unmap_queues), + .query_status_size = sizeof(struct pm4_mes_query_status), + .release_mem_size = 0, +}; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h new file mode 100644 index 000000000000..f795ec815e2a --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h @@ -0,0 +1,93 @@ +/* + * Copyright 2020 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +/*--------------------MES_MAP_PROCESS (PER DEBUG VMID)--------------------*/ + +#ifndef PM4_MES_MAP_PROCESS_PER_DEBUG_VMID_DEFINED +#define PM4_MES_MAP_PROCESS_PER_DEBUG_VMID_DEFINED + +struct pm4_mes_map_process_aldebaran { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + uint32_t pasid:16; /* 0 - 15 */ + uint32_t single_memops:1; /* 16 */ + uint32_t reserved1:1; /* 17 */ + uint32_t debug_vmid:4; /* 18 - 21 */ + uint32_t new_debug:1; /* 22 */ + uint32_t tmz:1; /* 23 */ + uint32_t diq_enable:1; /* 24 */ + uint32_t process_quantum:7; /* 25 - 31 */ + } bitfields2; + uint32_t ordinal2; + }; + + uint32_t vm_context_page_table_base_addr_lo32; + + uint32_t vm_context_page_table_base_addr_hi32; + + uint32_t sh_mem_bases; + + uint32_t sh_mem_config; + + uint32_t sq_shader_tba_lo; + + uint32_t sq_shader_tba_hi; + + uint32_t sq_shader_tma_lo; + + uint32_t sq_shader_tma_hi; + + uint32_t reserved6; + + uint32_t gds_addr_lo; + + uint32_t gds_addr_hi; + + union { + struct { + uint32_t num_gws:7; + uint32_t sdma_enable:1; + uint32_t num_oac:4; + uint32_t gds_size_hi:4; + uint32_t gds_size:6; + uint32_t num_queues:10; + } bitfields14; + uint32_t ordinal14; + }; + + uint32_t spi_gdbg_per_vmid_cntl; + + uint32_t tcp_watch_cntl[4]; + + uint32_t completion_signal_lo; + + uint32_t completion_signal_hi; + +}; + +#endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 0b6595f7acda..64552f6b8ba4 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -322,6 +322,9 @@ struct kfd_dev { unsigned int max_doorbell_slices; int noretry; + + /* HMM page migration MEMORY_DEVICE_PRIVATE mapping */ + struct dev_pagemap pgmap; }; enum kfd_mempool { @@ -669,7 +672,7 @@ struct kfd_process_device { /* VM context for GPUVM allocations */ struct file *drm_file; - void *vm; + void *drm_priv; /* GPUVM allocations storage */ struct idr alloc_idr; @@ -731,6 +734,17 @@ struct kfd_process_device { #define qpd_to_pdd(x) container_of(x, struct kfd_process_device, qpd) +struct svm_range_list { + struct mutex lock; + struct rb_root_cached objects; + struct list_head list; + struct work_struct deferred_list_work; + struct list_head deferred_range_list; + spinlock_t deferred_list_lock; + atomic_t evicted_ranges; + struct delayed_work restore_work; +}; + /* Process data */ struct kfd_process { /* @@ -809,6 +823,12 @@ struct kfd_process { struct kobject *kobj; struct kobject *kobj_queues; struct attribute attr_pasid; + + /* shared virtual memory registered by this process */ + struct svm_range_list svms; + bool svm_disabled; + + bool xnack_enabled; }; #define KFD_PROCESS_TABLE_SIZE 5 /* bits: 32 entries */ @@ -842,6 +862,20 @@ struct kfd_process *kfd_create_process(struct file *filep); struct kfd_process *kfd_get_process(const struct task_struct *); struct kfd_process *kfd_lookup_process_by_pasid(u32 pasid); struct kfd_process *kfd_lookup_process_by_mm(const struct mm_struct *mm); + +int kfd_process_gpuidx_from_gpuid(struct kfd_process *p, uint32_t gpu_id); +int kfd_process_gpuid_from_kgd(struct kfd_process *p, + struct amdgpu_device *adev, uint32_t *gpuid, + uint32_t *gpuidx); +static inline int kfd_process_gpuid_from_gpuidx(struct kfd_process *p, + uint32_t gpuidx, uint32_t *gpuid) { + return gpuidx < p->n_pdds ? p->pdds[gpuidx]->dev->id : -EINVAL; +} +static inline struct kfd_process_device *kfd_process_device_from_gpuidx( + struct kfd_process *p, uint32_t gpuidx) { + return gpuidx < p->n_pdds ? p->pdds[gpuidx] : NULL; +} + void kfd_unref_process(struct kfd_process *p); int kfd_process_evict_queues(struct kfd_process *p); int kfd_process_restore_queues(struct kfd_process *p); @@ -857,6 +891,8 @@ struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev, struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev, struct kfd_process *p); +bool kfd_process_xnack_mode(struct kfd_process *p, bool supported); + int kfd_reserved_mem_mmap(struct kfd_dev *dev, struct kfd_process *process, struct vm_area_struct *vma); @@ -1052,6 +1088,7 @@ struct packet_manager_funcs { extern const struct packet_manager_funcs kfd_vi_pm_funcs; extern const struct packet_manager_funcs kfd_v9_pm_funcs; +extern const struct packet_manager_funcs kfd_aldebaran_pm_funcs; int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm); void pm_uninit(struct packet_manager *pm, bool hanging); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index d97e330a5022..9d4f527bda7c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -35,6 +35,7 @@ #include <linux/pm_runtime.h> #include "amdgpu_amdkfd.h" #include "amdgpu.h" +#include "kfd_svm.h" struct mm_struct; @@ -42,6 +43,7 @@ struct mm_struct; #include "kfd_device_queue_manager.h" #include "kfd_dbgmgr.h" #include "kfd_iommu.h" +#include "kfd_svm.h" /* * List of struct kfd_process (field kfd_process). @@ -250,7 +252,7 @@ cleanup: } /** - * @kfd_get_cu_occupancy() - Collect number of waves in-flight on this device + * @kfd_get_cu_occupancy - Collect number of waves in-flight on this device * by current process. Translates acquired wave count into number of compute units * that are occupied. * @@ -647,8 +649,9 @@ static void kfd_process_free_gpuvm(struct kgd_mem *mem, { struct kfd_dev *dev = pdd->dev; - amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(dev->kgd, mem, pdd->vm); - amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, mem, NULL); + amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(dev->kgd, mem, pdd->drm_priv); + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, mem, pdd->drm_priv, + NULL); } /* kfd_process_alloc_gpuvm - Allocate GPU VM for the KFD process @@ -667,11 +670,11 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd, int err; err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(kdev->kgd, gpu_va, size, - pdd->vm, &mem, NULL, flags); + pdd->drm_priv, &mem, NULL, flags); if (err) goto err_alloc_mem; - err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->kgd, mem, pdd->vm); + err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->kgd, mem, pdd->drm_priv); if (err) goto err_map_mem; @@ -712,7 +715,8 @@ sync_memory_failed: return err; err_map_mem: - amdgpu_amdkfd_gpuvm_free_memory_of_gpu(kdev->kgd, mem, NULL); + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(kdev->kgd, mem, pdd->drm_priv, + NULL); err_alloc_mem: *kptr = NULL; return err; @@ -901,13 +905,14 @@ static void kfd_process_device_free_bos(struct kfd_process_device *pdd) for (i = 0; i < p->n_pdds; i++) { struct kfd_process_device *peer_pdd = p->pdds[i]; - if (!peer_pdd->vm) + if (!peer_pdd->drm_priv) continue; amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu( - peer_pdd->dev->kgd, mem, peer_pdd->vm); + peer_pdd->dev->kgd, mem, peer_pdd->drm_priv); } - amdgpu_amdkfd_gpuvm_free_memory_of_gpu(pdd->dev->kgd, mem, NULL); + amdgpu_amdkfd_gpuvm_free_memory_of_gpu(pdd->dev->kgd, mem, + pdd->drm_priv, NULL); kfd_process_device_remove_obj_handle(pdd, id); } } @@ -932,7 +937,7 @@ static void kfd_process_destroy_pdds(struct kfd_process *p) if (pdd->drm_file) { amdgpu_amdkfd_gpuvm_release_process_vm( - pdd->dev->kgd, pdd->vm); + pdd->dev->kgd, pdd->drm_priv); fput(pdd->drm_file); } @@ -1000,6 +1005,7 @@ static void kfd_process_wq_release(struct work_struct *work) kfd_iommu_unbind_process(p); kfd_process_free_outstanding_kfd_bos(p); + svm_range_list_fini(p); kfd_process_destroy_pdds(p); dma_fence_put(p->ef); @@ -1058,6 +1064,7 @@ static void kfd_process_notifier_release(struct mmu_notifier *mn, cancel_delayed_work_sync(&p->eviction_work); cancel_delayed_work_sync(&p->restore_work); + cancel_delayed_work_sync(&p->svms.restore_work); mutex_lock(&p->mutex); @@ -1186,6 +1193,56 @@ void kfd_process_set_trap_handler(struct qcm_process_device *qpd, } } +bool kfd_process_xnack_mode(struct kfd_process *p, bool supported) +{ + int i; + + /* On most GFXv9 GPUs, the retry mode in the SQ must match the + * boot time retry setting. Mixing processes with different + * XNACK/retry settings can hang the GPU. + * + * Different GPUs can have different noretry settings depending + * on HW bugs or limitations. We need to find at least one + * XNACK mode for this process that's compatible with all GPUs. + * Fortunately GPUs with retry enabled (noretry=0) can run code + * built for XNACK-off. On GFXv9 it may perform slower. + * + * Therefore applications built for XNACK-off can always be + * supported and will be our fallback if any GPU does not + * support retry. + */ + for (i = 0; i < p->n_pdds; i++) { + struct kfd_dev *dev = p->pdds[i]->dev; + + /* Only consider GFXv9 and higher GPUs. Older GPUs don't + * support the SVM APIs and don't need to be considered + * for the XNACK mode selection. + */ + if (dev->device_info->asic_family < CHIP_VEGA10) + continue; + /* Aldebaran can always support XNACK because it can support + * per-process XNACK mode selection. But let the dev->noretry + * setting still influence the default XNACK mode. + */ + if (supported && + dev->device_info->asic_family == CHIP_ALDEBARAN) + continue; + + /* GFXv10 and later GPUs do not support shader preemption + * during page faults. This can lead to poor QoS for queue + * management and memory-manager-related preemptions or + * even deadlocks. + */ + if (dev->device_info->asic_family >= CHIP_NAVI10) + return false; + + if (dev->noretry) + return false; + } + + return true; +} + /* * On return the kfd_process is fully operational and will be freed when the * mm is released @@ -1205,6 +1262,7 @@ static struct kfd_process *create_process(const struct task_struct *thread) process->mm = thread->mm; process->lead_thread = thread->group_leader; process->n_pdds = 0; + process->svm_disabled = false; INIT_DELAYED_WORK(&process->eviction_work, evict_process_worker); INIT_DELAYED_WORK(&process->restore_work, restore_process_worker); process->last_restore_timestamp = get_jiffies_64(); @@ -1224,6 +1282,13 @@ static struct kfd_process *create_process(const struct task_struct *thread) if (err != 0) goto err_init_apertures; + /* Check XNACK support after PDDs are created in kfd_init_apertures */ + process->xnack_enabled = kfd_process_xnack_mode(process, false); + + err = svm_range_list_init(process); + if (err) + goto err_init_svm_range_list; + /* alloc_notifier needs to find the process in the hash table */ hash_add_rcu(kfd_processes_table, &process->kfd_processes, (uintptr_t)process->mm); @@ -1246,6 +1311,8 @@ static struct kfd_process *create_process(const struct task_struct *thread) err_register_notifier: hash_del_rcu(&process->kfd_processes); + svm_range_list_fini(process); +err_init_svm_range_list: kfd_process_free_outstanding_kfd_bos(process); kfd_process_destroy_pdds(process); err_init_apertures: @@ -1375,7 +1442,7 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd, if (!drm_file) return -EINVAL; - if (pdd->vm) + if (pdd->drm_priv) return -EBUSY; p = pdd->process; @@ -1383,13 +1450,12 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd, ret = amdgpu_amdkfd_gpuvm_acquire_process_vm( dev->kgd, drm_file, p->pasid, - &pdd->vm, &p->kgd_process_info, &p->ef); + &p->kgd_process_info, &p->ef); if (ret) { pr_err("Failed to create process VM object\n"); return ret; } - - amdgpu_vm_set_task_info(pdd->vm); + pdd->drm_priv = drm_file->private_data; ret = kfd_process_device_reserve_ib_mem(pdd); if (ret) @@ -1405,7 +1471,7 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd, err_init_cwsr: err_reserve_ib_mem: kfd_process_device_free_bos(pdd); - pdd->vm = NULL; + pdd->drm_priv = NULL; return ret; } @@ -1429,7 +1495,7 @@ struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev, return ERR_PTR(-ENOMEM); } - if (!pdd->vm) + if (!pdd->drm_priv) return ERR_PTR(-ENODEV); /* @@ -1600,6 +1666,32 @@ int kfd_process_restore_queues(struct kfd_process *p) return ret; } +int kfd_process_gpuidx_from_gpuid(struct kfd_process *p, uint32_t gpu_id) +{ + int i; + + for (i = 0; i < p->n_pdds; i++) + if (p->pdds[i] && gpu_id == p->pdds[i]->dev->id) + return i; + return -EINVAL; +} + +int +kfd_process_gpuid_from_kgd(struct kfd_process *p, struct amdgpu_device *adev, + uint32_t *gpuid, uint32_t *gpuidx) +{ + struct kgd_dev *kgd = (struct kgd_dev *)adev; + int i; + + for (i = 0; i < p->n_pdds; i++) + if (p->pdds[i] && p->pdds[i]->dev->kgd == kgd) { + *gpuid = p->pdds[i]->dev->id; + *gpuidx = i; + return 0; + } + return -EINVAL; +} + static void evict_process_worker(struct work_struct *work) { int ret; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c new file mode 100644 index 000000000000..b665e9ff77e3 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c @@ -0,0 +1,3085 @@ +// SPDX-License-Identifier: GPL-2.0 OR MIT +/* + * Copyright 2020-2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include <linux/types.h> +#include <linux/sched/task.h> +#include "amdgpu_sync.h" +#include "amdgpu_object.h" +#include "amdgpu_vm.h" +#include "amdgpu_mn.h" +#include "amdgpu.h" +#include "amdgpu_xgmi.h" +#include "kfd_priv.h" +#include "kfd_svm.h" +#include "kfd_migrate.h" + +#define AMDGPU_SVM_RANGE_RESTORE_DELAY_MS 1 + +/* Long enough to ensure no retry fault comes after svm range is restored and + * page table is updated. + */ +#define AMDGPU_SVM_RANGE_RETRY_FAULT_PENDING 2000 + +static void svm_range_evict_svm_bo_worker(struct work_struct *work); +static bool +svm_range_cpu_invalidate_pagetables(struct mmu_interval_notifier *mni, + const struct mmu_notifier_range *range, + unsigned long cur_seq); + +static const struct mmu_interval_notifier_ops svm_range_mn_ops = { + .invalidate = svm_range_cpu_invalidate_pagetables, +}; + +/** + * svm_range_unlink - unlink svm_range from lists and interval tree + * @prange: svm range structure to be removed + * + * Remove the svm_range from the svms and svm_bo lists and the svms + * interval tree. + * + * Context: The caller must hold svms->lock + */ +static void svm_range_unlink(struct svm_range *prange) +{ + pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx]\n", prange->svms, + prange, prange->start, prange->last); + + if (prange->svm_bo) { + spin_lock(&prange->svm_bo->list_lock); + list_del(&prange->svm_bo_list); + spin_unlock(&prange->svm_bo->list_lock); + } + + list_del(&prange->list); + if (prange->it_node.start != 0 && prange->it_node.last != 0) + interval_tree_remove(&prange->it_node, &prange->svms->objects); +} + +static void +svm_range_add_notifier_locked(struct mm_struct *mm, struct svm_range *prange) +{ + pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx]\n", prange->svms, + prange, prange->start, prange->last); + + mmu_interval_notifier_insert_locked(&prange->notifier, mm, + prange->start << PAGE_SHIFT, + prange->npages << PAGE_SHIFT, + &svm_range_mn_ops); +} + +/** + * svm_range_add_to_svms - add svm range to svms + * @prange: svm range structure to be added + * + * Add the svm range to svms interval tree and link list + * + * Context: The caller must hold svms->lock + */ +static void svm_range_add_to_svms(struct svm_range *prange) +{ + pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx]\n", prange->svms, + prange, prange->start, prange->last); + + list_add_tail(&prange->list, &prange->svms->list); + prange->it_node.start = prange->start; + prange->it_node.last = prange->last; + interval_tree_insert(&prange->it_node, &prange->svms->objects); +} + +static void svm_range_remove_notifier(struct svm_range *prange) +{ + pr_debug("remove notifier svms 0x%p prange 0x%p [0x%lx 0x%lx]\n", + prange->svms, prange, + prange->notifier.interval_tree.start >> PAGE_SHIFT, + prange->notifier.interval_tree.last >> PAGE_SHIFT); + + if (prange->notifier.interval_tree.start != 0 && + prange->notifier.interval_tree.last != 0) + mmu_interval_notifier_remove(&prange->notifier); +} + +static int +svm_range_dma_map_dev(struct device *dev, dma_addr_t **dma_addr, + unsigned long *hmm_pfns, uint64_t npages) +{ + enum dma_data_direction dir = DMA_BIDIRECTIONAL; + dma_addr_t *addr = *dma_addr; + struct page *page; + int i, r; + + if (!addr) { + addr = kvmalloc_array(npages, sizeof(*addr), + GFP_KERNEL | __GFP_ZERO); + if (!addr) + return -ENOMEM; + *dma_addr = addr; + } + + for (i = 0; i < npages; i++) { + if (WARN_ONCE(addr[i] && !dma_mapping_error(dev, addr[i]), + "leaking dma mapping\n")) + dma_unmap_page(dev, addr[i], PAGE_SIZE, dir); + + page = hmm_pfn_to_page(hmm_pfns[i]); + addr[i] = dma_map_page(dev, page, 0, PAGE_SIZE, dir); + r = dma_mapping_error(dev, addr[i]); + if (r) { + pr_debug("failed %d dma_map_page\n", r); + return r; + } + pr_debug("dma mapping 0x%llx for page addr 0x%lx\n", + addr[i] >> PAGE_SHIFT, page_to_pfn(page)); + } + return 0; +} + +static int +svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap, + unsigned long *hmm_pfns) +{ + struct kfd_process *p; + uint32_t gpuidx; + int r; + + p = container_of(prange->svms, struct kfd_process, svms); + + for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) { + struct kfd_process_device *pdd; + struct amdgpu_device *adev; + + pr_debug("mapping to gpu idx 0x%x\n", gpuidx); + pdd = kfd_process_device_from_gpuidx(p, gpuidx); + if (!pdd) { + pr_debug("failed to find device idx %d\n", gpuidx); + return -EINVAL; + } + adev = (struct amdgpu_device *)pdd->dev->kgd; + + r = svm_range_dma_map_dev(adev->dev, &prange->dma_addr[gpuidx], + hmm_pfns, prange->npages); + if (r) + break; + } + + return r; +} + +void svm_range_dma_unmap(struct device *dev, dma_addr_t *dma_addr, + unsigned long offset, unsigned long npages) +{ + enum dma_data_direction dir = DMA_BIDIRECTIONAL; + int i; + + if (!dma_addr) + return; + + for (i = offset; i < offset + npages; i++) { + if (!dma_addr[i] || dma_mapping_error(dev, dma_addr[i])) + continue; + pr_debug("dma unmapping 0x%llx\n", dma_addr[i] >> PAGE_SHIFT); + dma_unmap_page(dev, dma_addr[i], PAGE_SIZE, dir); + dma_addr[i] = 0; + } +} + +void svm_range_free_dma_mappings(struct svm_range *prange) +{ + struct kfd_process_device *pdd; + dma_addr_t *dma_addr; + struct device *dev; + struct kfd_process *p; + uint32_t gpuidx; + + p = container_of(prange->svms, struct kfd_process, svms); + + for (gpuidx = 0; gpuidx < MAX_GPU_INSTANCE; gpuidx++) { + dma_addr = prange->dma_addr[gpuidx]; + if (!dma_addr) + continue; + + pdd = kfd_process_device_from_gpuidx(p, gpuidx); + if (!pdd) { + pr_debug("failed to find device idx %d\n", gpuidx); + continue; + } + dev = &pdd->dev->pdev->dev; + svm_range_dma_unmap(dev, dma_addr, 0, prange->npages); + kvfree(dma_addr); + prange->dma_addr[gpuidx] = NULL; + } +} + +static void svm_range_free(struct svm_range *prange) +{ + pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx]\n", prange->svms, prange, + prange->start, prange->last); + + svm_range_vram_node_free(prange); + svm_range_free_dma_mappings(prange); + mutex_destroy(&prange->lock); + mutex_destroy(&prange->migrate_mutex); + kfree(prange); +} + +static void +svm_range_set_default_attributes(int32_t *location, int32_t *prefetch_loc, + uint8_t *granularity, uint32_t *flags) +{ + *location = KFD_IOCTL_SVM_LOCATION_UNDEFINED; + *prefetch_loc = KFD_IOCTL_SVM_LOCATION_UNDEFINED; + *granularity = 9; + *flags = + KFD_IOCTL_SVM_FLAG_HOST_ACCESS | KFD_IOCTL_SVM_FLAG_COHERENT; +} + +static struct +svm_range *svm_range_new(struct svm_range_list *svms, uint64_t start, + uint64_t last) +{ + uint64_t size = last - start + 1; + struct svm_range *prange; + struct kfd_process *p; + + prange = kzalloc(sizeof(*prange), GFP_KERNEL); + if (!prange) + return NULL; + prange->npages = size; + prange->svms = svms; + prange->start = start; + prange->last = last; + INIT_LIST_HEAD(&prange->list); + INIT_LIST_HEAD(&prange->update_list); + INIT_LIST_HEAD(&prange->remove_list); + INIT_LIST_HEAD(&prange->insert_list); + INIT_LIST_HEAD(&prange->svm_bo_list); + INIT_LIST_HEAD(&prange->deferred_list); + INIT_LIST_HEAD(&prange->child_list); + atomic_set(&prange->invalid, 0); + prange->validate_timestamp = 0; + mutex_init(&prange->migrate_mutex); + mutex_init(&prange->lock); + + p = container_of(svms, struct kfd_process, svms); + if (p->xnack_enabled) + bitmap_fill(prange->bitmap_access, MAX_GPU_INSTANCE); + + svm_range_set_default_attributes(&prange->preferred_loc, + &prange->prefetch_loc, + &prange->granularity, &prange->flags); + + pr_debug("svms 0x%p [0x%llx 0x%llx]\n", svms, start, last); + + return prange; +} + +static bool svm_bo_ref_unless_zero(struct svm_range_bo *svm_bo) +{ + if (!svm_bo || !kref_get_unless_zero(&svm_bo->kref)) + return false; + + return true; +} + +static struct svm_range_bo *svm_range_bo_ref(struct svm_range_bo *svm_bo) +{ + if (svm_bo) + kref_get(&svm_bo->kref); + + return svm_bo; +} + +static void svm_range_bo_release(struct kref *kref) +{ + struct svm_range_bo *svm_bo; + + svm_bo = container_of(kref, struct svm_range_bo, kref); + spin_lock(&svm_bo->list_lock); + while (!list_empty(&svm_bo->range_list)) { + struct svm_range *prange = + list_first_entry(&svm_bo->range_list, + struct svm_range, svm_bo_list); + /* list_del_init tells a concurrent svm_range_vram_node_new when + * it's safe to reuse the svm_bo pointer and svm_bo_list head. + */ + list_del_init(&prange->svm_bo_list); + spin_unlock(&svm_bo->list_lock); + + pr_debug("svms 0x%p [0x%lx 0x%lx]\n", prange->svms, + prange->start, prange->last); + mutex_lock(&prange->lock); + prange->svm_bo = NULL; + mutex_unlock(&prange->lock); + + spin_lock(&svm_bo->list_lock); + } + spin_unlock(&svm_bo->list_lock); + if (!dma_fence_is_signaled(&svm_bo->eviction_fence->base)) { + /* We're not in the eviction worker. + * Signal the fence and synchronize with any + * pending eviction work. + */ + dma_fence_signal(&svm_bo->eviction_fence->base); + cancel_work_sync(&svm_bo->eviction_work); + } + dma_fence_put(&svm_bo->eviction_fence->base); + amdgpu_bo_unref(&svm_bo->bo); + kfree(svm_bo); +} + +static void svm_range_bo_unref(struct svm_range_bo *svm_bo) +{ + if (!svm_bo) + return; + + kref_put(&svm_bo->kref, svm_range_bo_release); +} + +static bool +svm_range_validate_svm_bo(struct amdgpu_device *adev, struct svm_range *prange) +{ + struct amdgpu_device *bo_adev; + + mutex_lock(&prange->lock); + if (!prange->svm_bo) { + mutex_unlock(&prange->lock); + return false; + } + if (prange->ttm_res) { + /* We still have a reference, all is well */ + mutex_unlock(&prange->lock); + return true; + } + if (svm_bo_ref_unless_zero(prange->svm_bo)) { + /* + * Migrate from GPU to GPU, remove range from source bo_adev + * svm_bo range list, and return false to allocate svm_bo from + * destination adev. + */ + bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev); + if (bo_adev != adev) { + mutex_unlock(&prange->lock); + + spin_lock(&prange->svm_bo->list_lock); + list_del_init(&prange->svm_bo_list); + spin_unlock(&prange->svm_bo->list_lock); + + svm_range_bo_unref(prange->svm_bo); + return false; + } + if (READ_ONCE(prange->svm_bo->evicting)) { + struct dma_fence *f; + struct svm_range_bo *svm_bo; + /* The BO is getting evicted, + * we need to get a new one + */ + mutex_unlock(&prange->lock); + svm_bo = prange->svm_bo; + f = dma_fence_get(&svm_bo->eviction_fence->base); + svm_range_bo_unref(prange->svm_bo); + /* wait for the fence to avoid long spin-loop + * at list_empty_careful + */ + dma_fence_wait(f, false); + dma_fence_put(f); + } else { + /* The BO was still around and we got + * a new reference to it + */ + mutex_unlock(&prange->lock); + pr_debug("reuse old bo svms 0x%p [0x%lx 0x%lx]\n", + prange->svms, prange->start, prange->last); + + prange->ttm_res = &prange->svm_bo->bo->tbo.mem; + return true; + } + + } else { + mutex_unlock(&prange->lock); + } + + /* We need a new svm_bo. Spin-loop to wait for concurrent + * svm_range_bo_release to finish removing this range from + * its range list. After this, it is safe to reuse the + * svm_bo pointer and svm_bo_list head. + */ + while (!list_empty_careful(&prange->svm_bo_list)) + ; + + return false; +} + +static struct svm_range_bo *svm_range_bo_new(void) +{ + struct svm_range_bo *svm_bo; + + svm_bo = kzalloc(sizeof(*svm_bo), GFP_KERNEL); + if (!svm_bo) + return NULL; + + kref_init(&svm_bo->kref); + INIT_LIST_HEAD(&svm_bo->range_list); + spin_lock_init(&svm_bo->list_lock); + + return svm_bo; +} + +int +svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange, + bool clear) +{ + struct amdgpu_bo_param bp; + struct svm_range_bo *svm_bo; + struct amdgpu_bo_user *ubo; + struct amdgpu_bo *bo; + struct kfd_process *p; + struct mm_struct *mm; + int r; + + p = container_of(prange->svms, struct kfd_process, svms); + pr_debug("pasid: %x svms 0x%p [0x%lx 0x%lx]\n", p->pasid, prange->svms, + prange->start, prange->last); + + if (svm_range_validate_svm_bo(adev, prange)) + return 0; + + svm_bo = svm_range_bo_new(); + if (!svm_bo) { + pr_debug("failed to alloc svm bo\n"); + return -ENOMEM; + } + mm = get_task_mm(p->lead_thread); + if (!mm) { + pr_debug("failed to get mm\n"); + kfree(svm_bo); + return -ESRCH; + } + svm_bo->svms = prange->svms; + svm_bo->eviction_fence = + amdgpu_amdkfd_fence_create(dma_fence_context_alloc(1), + mm, + svm_bo); + mmput(mm); + INIT_WORK(&svm_bo->eviction_work, svm_range_evict_svm_bo_worker); + svm_bo->evicting = 0; + memset(&bp, 0, sizeof(bp)); + bp.size = prange->npages * PAGE_SIZE; + bp.byte_align = PAGE_SIZE; + bp.domain = AMDGPU_GEM_DOMAIN_VRAM; + bp.flags = AMDGPU_GEM_CREATE_NO_CPU_ACCESS; + bp.flags |= clear ? AMDGPU_GEM_CREATE_VRAM_CLEARED : 0; + bp.flags |= AMDGPU_AMDKFD_CREATE_SVM_BO; + bp.type = ttm_bo_type_device; + bp.resv = NULL; + + r = amdgpu_bo_create_user(adev, &bp, &ubo); + if (r) { + pr_debug("failed %d to create bo\n", r); + goto create_bo_failed; + } + bo = &ubo->bo; + r = amdgpu_bo_reserve(bo, true); + if (r) { + pr_debug("failed %d to reserve bo\n", r); + goto reserve_bo_failed; + } + + r = dma_resv_reserve_shared(bo->tbo.base.resv, 1); + if (r) { + pr_debug("failed %d to reserve bo\n", r); + amdgpu_bo_unreserve(bo); + goto reserve_bo_failed; + } + amdgpu_bo_fence(bo, &svm_bo->eviction_fence->base, true); + + amdgpu_bo_unreserve(bo); + + svm_bo->bo = bo; + prange->svm_bo = svm_bo; + prange->ttm_res = &bo->tbo.mem; + prange->offset = 0; + + spin_lock(&svm_bo->list_lock); + list_add(&prange->svm_bo_list, &svm_bo->range_list); + spin_unlock(&svm_bo->list_lock); + + return 0; + +reserve_bo_failed: + amdgpu_bo_unref(&bo); +create_bo_failed: + dma_fence_put(&svm_bo->eviction_fence->base); + kfree(svm_bo); + prange->ttm_res = NULL; + + return r; +} + +void svm_range_vram_node_free(struct svm_range *prange) +{ + svm_range_bo_unref(prange->svm_bo); + prange->ttm_res = NULL; +} + +struct amdgpu_device * +svm_range_get_adev_by_id(struct svm_range *prange, uint32_t gpu_id) +{ + struct kfd_process_device *pdd; + struct kfd_process *p; + int32_t gpu_idx; + + p = container_of(prange->svms, struct kfd_process, svms); + + gpu_idx = kfd_process_gpuidx_from_gpuid(p, gpu_id); + if (gpu_idx < 0) { + pr_debug("failed to get device by id 0x%x\n", gpu_id); + return NULL; + } + pdd = kfd_process_device_from_gpuidx(p, gpu_idx); + if (!pdd) { + pr_debug("failed to get device by idx 0x%x\n", gpu_idx); + return NULL; + } + + return (struct amdgpu_device *)pdd->dev->kgd; +} + +static int svm_range_bo_validate(void *param, struct amdgpu_bo *bo) +{ + struct ttm_operation_ctx ctx = { false, false }; + + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_VRAM); + + return ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); +} + +static int +svm_range_check_attr(struct kfd_process *p, + uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs) +{ + uint32_t i; + int gpuidx; + + for (i = 0; i < nattr; i++) { + switch (attrs[i].type) { + case KFD_IOCTL_SVM_ATTR_PREFERRED_LOC: + if (attrs[i].value != KFD_IOCTL_SVM_LOCATION_SYSMEM && + attrs[i].value != KFD_IOCTL_SVM_LOCATION_UNDEFINED && + kfd_process_gpuidx_from_gpuid(p, + attrs[i].value) < 0) { + pr_debug("no GPU 0x%x found\n", attrs[i].value); + return -EINVAL; + } + break; + case KFD_IOCTL_SVM_ATTR_PREFETCH_LOC: + if (attrs[i].value != KFD_IOCTL_SVM_LOCATION_SYSMEM && + kfd_process_gpuidx_from_gpuid(p, + attrs[i].value) < 0) { + pr_debug("no GPU 0x%x found\n", attrs[i].value); + return -EINVAL; + } + break; + case KFD_IOCTL_SVM_ATTR_ACCESS: + case KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE: + case KFD_IOCTL_SVM_ATTR_NO_ACCESS: + gpuidx = kfd_process_gpuidx_from_gpuid(p, + attrs[i].value); + if (gpuidx < 0) { + pr_debug("no GPU 0x%x found\n", attrs[i].value); + return -EINVAL; + } + break; + case KFD_IOCTL_SVM_ATTR_SET_FLAGS: + break; + case KFD_IOCTL_SVM_ATTR_CLR_FLAGS: + break; + case KFD_IOCTL_SVM_ATTR_GRANULARITY: + break; + default: + pr_debug("unknown attr type 0x%x\n", attrs[i].type); + return -EINVAL; + } + } + + return 0; +} + +static void +svm_range_apply_attrs(struct kfd_process *p, struct svm_range *prange, + uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs) +{ + uint32_t i; + int gpuidx; + + for (i = 0; i < nattr; i++) { + switch (attrs[i].type) { + case KFD_IOCTL_SVM_ATTR_PREFERRED_LOC: + prange->preferred_loc = attrs[i].value; + break; + case KFD_IOCTL_SVM_ATTR_PREFETCH_LOC: + prange->prefetch_loc = attrs[i].value; + break; + case KFD_IOCTL_SVM_ATTR_ACCESS: + case KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE: + case KFD_IOCTL_SVM_ATTR_NO_ACCESS: + gpuidx = kfd_process_gpuidx_from_gpuid(p, + attrs[i].value); + if (attrs[i].type == KFD_IOCTL_SVM_ATTR_NO_ACCESS) { + bitmap_clear(prange->bitmap_access, gpuidx, 1); + bitmap_clear(prange->bitmap_aip, gpuidx, 1); + } else if (attrs[i].type == KFD_IOCTL_SVM_ATTR_ACCESS) { + bitmap_set(prange->bitmap_access, gpuidx, 1); + bitmap_clear(prange->bitmap_aip, gpuidx, 1); + } else { + bitmap_clear(prange->bitmap_access, gpuidx, 1); + bitmap_set(prange->bitmap_aip, gpuidx, 1); + } + break; + case KFD_IOCTL_SVM_ATTR_SET_FLAGS: + prange->flags |= attrs[i].value; + break; + case KFD_IOCTL_SVM_ATTR_CLR_FLAGS: + prange->flags &= ~attrs[i].value; + break; + case KFD_IOCTL_SVM_ATTR_GRANULARITY: + prange->granularity = attrs[i].value; + break; + default: + WARN_ONCE(1, "svm_range_check_attrs wasn't called?"); + } + } +} + +/** + * svm_range_debug_dump - print all range information from svms + * @svms: svm range list header + * + * debug output svm range start, end, prefetch location from svms + * interval tree and link list + * + * Context: The caller must hold svms->lock + */ +static void svm_range_debug_dump(struct svm_range_list *svms) +{ + struct interval_tree_node *node; + struct svm_range *prange; + + pr_debug("dump svms 0x%p list\n", svms); + pr_debug("range\tstart\tpage\tend\t\tlocation\n"); + + list_for_each_entry(prange, &svms->list, list) { + pr_debug("0x%p 0x%lx\t0x%llx\t0x%llx\t0x%x\n", + prange, prange->start, prange->npages, + prange->start + prange->npages - 1, + prange->actual_loc); + } + + pr_debug("dump svms 0x%p interval tree\n", svms); + pr_debug("range\tstart\tpage\tend\t\tlocation\n"); + node = interval_tree_iter_first(&svms->objects, 0, ~0ULL); + while (node) { + prange = container_of(node, struct svm_range, it_node); + pr_debug("0x%p 0x%lx\t0x%llx\t0x%llx\t0x%x\n", + prange, prange->start, prange->npages, + prange->start + prange->npages - 1, + prange->actual_loc); + node = interval_tree_iter_next(node, 0, ~0ULL); + } +} + +static bool +svm_range_is_same_attrs(struct svm_range *old, struct svm_range *new) +{ + return (old->prefetch_loc == new->prefetch_loc && + old->flags == new->flags && + old->granularity == new->granularity); +} + +static int +svm_range_split_array(void *ppnew, void *ppold, size_t size, + uint64_t old_start, uint64_t old_n, + uint64_t new_start, uint64_t new_n) +{ + unsigned char *new, *old, *pold; + uint64_t d; + + if (!ppold) + return 0; + pold = *(unsigned char **)ppold; + if (!pold) + return 0; + + new = kvmalloc_array(new_n, size, GFP_KERNEL); + if (!new) + return -ENOMEM; + + d = (new_start - old_start) * size; + memcpy(new, pold + d, new_n * size); + + old = kvmalloc_array(old_n, size, GFP_KERNEL); + if (!old) { + kvfree(new); + return -ENOMEM; + } + + d = (new_start == old_start) ? new_n * size : 0; + memcpy(old, pold + d, old_n * size); + + kvfree(pold); + *(void **)ppold = old; + *(void **)ppnew = new; + + return 0; +} + +static int +svm_range_split_pages(struct svm_range *new, struct svm_range *old, + uint64_t start, uint64_t last) +{ + uint64_t npages = last - start + 1; + int i, r; + + for (i = 0; i < MAX_GPU_INSTANCE; i++) { + r = svm_range_split_array(&new->dma_addr[i], &old->dma_addr[i], + sizeof(*old->dma_addr[i]), old->start, + npages, new->start, new->npages); + if (r) + return r; + } + + return 0; +} + +static int +svm_range_split_nodes(struct svm_range *new, struct svm_range *old, + uint64_t start, uint64_t last) +{ + uint64_t npages = last - start + 1; + + pr_debug("svms 0x%p new prange 0x%p start 0x%lx [0x%llx 0x%llx]\n", + new->svms, new, new->start, start, last); + + if (new->start == old->start) { + new->offset = old->offset; + old->offset += new->npages; + } else { + new->offset = old->offset + npages; + } + + new->svm_bo = svm_range_bo_ref(old->svm_bo); + new->ttm_res = old->ttm_res; + + spin_lock(&new->svm_bo->list_lock); + list_add(&new->svm_bo_list, &new->svm_bo->range_list); + spin_unlock(&new->svm_bo->list_lock); + + return 0; +} + +/** + * svm_range_split_adjust - split range and adjust + * + * @new: new range + * @old: the old range + * @start: the old range adjust to start address in pages + * @last: the old range adjust to last address in pages + * + * Copy system memory dma_addr or vram ttm_res in old range to new + * range from new_start up to size new->npages, the remaining old range is from + * start to last + * + * Return: + * 0 - OK, -ENOMEM - out of memory + */ +static int +svm_range_split_adjust(struct svm_range *new, struct svm_range *old, + uint64_t start, uint64_t last) +{ + int r; + + pr_debug("svms 0x%p new 0x%lx old [0x%lx 0x%lx] => [0x%llx 0x%llx]\n", + new->svms, new->start, old->start, old->last, start, last); + + if (new->start < old->start || + new->last > old->last) { + WARN_ONCE(1, "invalid new range start or last\n"); + return -EINVAL; + } + + r = svm_range_split_pages(new, old, start, last); + if (r) + return r; + + if (old->actual_loc && old->ttm_res) { + r = svm_range_split_nodes(new, old, start, last); + if (r) + return r; + } + + old->npages = last - start + 1; + old->start = start; + old->last = last; + new->flags = old->flags; + new->preferred_loc = old->preferred_loc; + new->prefetch_loc = old->prefetch_loc; + new->actual_loc = old->actual_loc; + new->granularity = old->granularity; + bitmap_copy(new->bitmap_access, old->bitmap_access, MAX_GPU_INSTANCE); + bitmap_copy(new->bitmap_aip, old->bitmap_aip, MAX_GPU_INSTANCE); + + return 0; +} + +/** + * svm_range_split - split a range in 2 ranges + * + * @prange: the svm range to split + * @start: the remaining range start address in pages + * @last: the remaining range last address in pages + * @new: the result new range generated + * + * Two cases only: + * case 1: if start == prange->start + * prange ==> prange[start, last] + * new range [last + 1, prange->last] + * + * case 2: if last == prange->last + * prange ==> prange[start, last] + * new range [prange->start, start - 1] + * + * Return: + * 0 - OK, -ENOMEM - out of memory, -EINVAL - invalid start, last + */ +static int +svm_range_split(struct svm_range *prange, uint64_t start, uint64_t last, + struct svm_range **new) +{ + uint64_t old_start = prange->start; + uint64_t old_last = prange->last; + struct svm_range_list *svms; + int r = 0; + + pr_debug("svms 0x%p [0x%llx 0x%llx] to [0x%llx 0x%llx]\n", prange->svms, + old_start, old_last, start, last); + + if (old_start != start && old_last != last) + return -EINVAL; + if (start < old_start || last > old_last) + return -EINVAL; + + svms = prange->svms; + if (old_start == start) + *new = svm_range_new(svms, last + 1, old_last); + else + *new = svm_range_new(svms, old_start, start - 1); + if (!*new) + return -ENOMEM; + + r = svm_range_split_adjust(*new, prange, start, last); + if (r) { + pr_debug("failed %d split [0x%llx 0x%llx] to [0x%llx 0x%llx]\n", + r, old_start, old_last, start, last); + svm_range_free(*new); + *new = NULL; + } + + return r; +} + +static int +svm_range_split_tail(struct svm_range *prange, struct svm_range *new, + uint64_t new_last, struct list_head *insert_list) +{ + struct svm_range *tail; + int r = svm_range_split(prange, prange->start, new_last, &tail); + + if (!r) + list_add(&tail->insert_list, insert_list); + return r; +} + +static int +svm_range_split_head(struct svm_range *prange, struct svm_range *new, + uint64_t new_start, struct list_head *insert_list) +{ + struct svm_range *head; + int r = svm_range_split(prange, new_start, prange->last, &head); + + if (!r) + list_add(&head->insert_list, insert_list); + return r; +} + +static void +svm_range_add_child(struct svm_range *prange, struct mm_struct *mm, + struct svm_range *pchild, enum svm_work_list_ops op) +{ + pr_debug("add child 0x%p [0x%lx 0x%lx] to prange 0x%p child list %d\n", + pchild, pchild->start, pchild->last, prange, op); + + pchild->work_item.mm = mm; + pchild->work_item.op = op; + list_add_tail(&pchild->child_list, &prange->child_list); +} + +/** + * svm_range_split_by_granularity - collect ranges within granularity boundary + * + * @p: the process with svms list + * @mm: mm structure + * @addr: the vm fault address in pages, to split the prange + * @parent: parent range if prange is from child list + * @prange: prange to split + * + * Trims @prange to be a single aligned block of prange->granularity if + * possible. The head and tail are added to the child_list in @parent. + * + * Context: caller must hold mmap_read_lock and prange->lock + * + * Return: + * 0 - OK, otherwise error code + */ +int +svm_range_split_by_granularity(struct kfd_process *p, struct mm_struct *mm, + unsigned long addr, struct svm_range *parent, + struct svm_range *prange) +{ + struct svm_range *head, *tail; + unsigned long start, last, size; + int r; + + /* Align splited range start and size to granularity size, then a single + * PTE will be used for whole range, this reduces the number of PTE + * updated and the L1 TLB space used for translation. + */ + size = 1UL << prange->granularity; + start = ALIGN_DOWN(addr, size); + last = ALIGN(addr + 1, size) - 1; + + pr_debug("svms 0x%p split [0x%lx 0x%lx] to [0x%lx 0x%lx] size 0x%lx\n", + prange->svms, prange->start, prange->last, start, last, size); + + if (start > prange->start) { + r = svm_range_split(prange, start, prange->last, &head); + if (r) + return r; + svm_range_add_child(parent, mm, head, SVM_OP_ADD_RANGE); + } + + if (last < prange->last) { + r = svm_range_split(prange, prange->start, last, &tail); + if (r) + return r; + svm_range_add_child(parent, mm, tail, SVM_OP_ADD_RANGE); + } + + /* xnack on, update mapping on GPUs with ACCESS_IN_PLACE */ + if (p->xnack_enabled && prange->work_item.op == SVM_OP_ADD_RANGE) { + prange->work_item.op = SVM_OP_ADD_RANGE_AND_MAP; + pr_debug("change prange 0x%p [0x%lx 0x%lx] op %d\n", + prange, prange->start, prange->last, + SVM_OP_ADD_RANGE_AND_MAP); + } + return 0; +} + +static uint64_t +svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange) +{ + struct amdgpu_device *bo_adev; + uint32_t flags = prange->flags; + uint32_t mapping_flags = 0; + uint64_t pte_flags; + bool snoop = !prange->ttm_res; + bool coherent = flags & KFD_IOCTL_SVM_FLAG_COHERENT; + + if (prange->svm_bo && prange->ttm_res) + bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev); + + switch (adev->asic_type) { + case CHIP_ARCTURUS: + if (prange->svm_bo && prange->ttm_res) { + if (bo_adev == adev) { + mapping_flags |= coherent ? + AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW; + } else { + mapping_flags |= AMDGPU_VM_MTYPE_UC; + if (amdgpu_xgmi_same_hive(adev, bo_adev)) + snoop = true; + } + } else { + mapping_flags |= coherent ? + AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; + } + break; + case CHIP_ALDEBARAN: + if (prange->svm_bo && prange->ttm_res) { + if (bo_adev == adev) { + mapping_flags |= coherent ? + AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW; + if (adev->gmc.xgmi.connected_to_cpu) + snoop = true; + } else { + mapping_flags |= AMDGPU_VM_MTYPE_UC; + if (amdgpu_xgmi_same_hive(adev, bo_adev)) + snoop = true; + } + } else { + mapping_flags |= coherent ? + AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; + } + break; + default: + mapping_flags |= coherent ? + AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; + } + + mapping_flags |= AMDGPU_VM_PAGE_READABLE | AMDGPU_VM_PAGE_WRITEABLE; + + if (flags & KFD_IOCTL_SVM_FLAG_GPU_RO) + mapping_flags &= ~AMDGPU_VM_PAGE_WRITEABLE; + if (flags & KFD_IOCTL_SVM_FLAG_GPU_EXEC) + mapping_flags |= AMDGPU_VM_PAGE_EXECUTABLE; + + pte_flags = AMDGPU_PTE_VALID; + pte_flags |= prange->ttm_res ? 0 : AMDGPU_PTE_SYSTEM; + pte_flags |= snoop ? AMDGPU_PTE_SNOOPED : 0; + + pte_flags |= amdgpu_gem_va_map_flags(adev, mapping_flags); + + pr_debug("svms 0x%p [0x%lx 0x%lx] vram %d PTE 0x%llx mapping 0x%x\n", + prange->svms, prange->start, prange->last, + prange->ttm_res ? 1:0, pte_flags, mapping_flags); + + return pte_flags; +} + +static int +svm_range_unmap_from_gpu(struct amdgpu_device *adev, struct amdgpu_vm *vm, + uint64_t start, uint64_t last, + struct dma_fence **fence) +{ + uint64_t init_pte_value = 0; + + pr_debug("[0x%llx 0x%llx]\n", start, last); + + return amdgpu_vm_bo_update_mapping(adev, adev, vm, false, true, NULL, + start, last, init_pte_value, 0, + NULL, NULL, fence); +} + +static int +svm_range_unmap_from_gpus(struct svm_range *prange, unsigned long start, + unsigned long last) +{ + DECLARE_BITMAP(bitmap, MAX_GPU_INSTANCE); + struct kfd_process_device *pdd; + struct dma_fence *fence = NULL; + struct amdgpu_device *adev; + struct kfd_process *p; + uint32_t gpuidx; + int r = 0; + + bitmap_or(bitmap, prange->bitmap_access, prange->bitmap_aip, + MAX_GPU_INSTANCE); + p = container_of(prange->svms, struct kfd_process, svms); + + for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) { + pr_debug("unmap from gpu idx 0x%x\n", gpuidx); + pdd = kfd_process_device_from_gpuidx(p, gpuidx); + if (!pdd) { + pr_debug("failed to find device idx %d\n", gpuidx); + return -EINVAL; + } + adev = (struct amdgpu_device *)pdd->dev->kgd; + + r = svm_range_unmap_from_gpu(adev, drm_priv_to_vm(pdd->drm_priv), + start, last, &fence); + if (r) + break; + + if (fence) { + r = dma_fence_wait(fence, false); + dma_fence_put(fence); + fence = NULL; + if (r) + break; + } + amdgpu_amdkfd_flush_gpu_tlb_pasid((struct kgd_dev *)adev, + p->pasid); + } + + return r; +} + +static int +svm_range_map_to_gpu(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct svm_range *prange, dma_addr_t *dma_addr, + struct amdgpu_device *bo_adev, struct dma_fence **fence) +{ + struct amdgpu_bo_va bo_va; + uint64_t pte_flags; + int r = 0; + + pr_debug("svms 0x%p [0x%lx 0x%lx]\n", prange->svms, prange->start, + prange->last); + + if (prange->svm_bo && prange->ttm_res) { + bo_va.is_xgmi = amdgpu_xgmi_same_hive(adev, bo_adev); + prange->mapping.bo_va = &bo_va; + } + + prange->mapping.start = prange->start; + prange->mapping.last = prange->last; + prange->mapping.offset = prange->offset; + pte_flags = svm_range_get_pte_flags(adev, prange); + + r = amdgpu_vm_bo_update_mapping(adev, bo_adev, vm, false, false, NULL, + prange->mapping.start, + prange->mapping.last, pte_flags, + prange->mapping.offset, + prange->ttm_res ? + prange->ttm_res->mm_node : NULL, + dma_addr, &vm->last_update); + if (r) { + pr_debug("failed %d to map to gpu 0x%lx\n", r, prange->start); + goto out; + } + + r = amdgpu_vm_update_pdes(adev, vm, false); + if (r) { + pr_debug("failed %d to update directories 0x%lx\n", r, + prange->start); + goto out; + } + + if (fence) + *fence = dma_fence_get(vm->last_update); + +out: + prange->mapping.bo_va = NULL; + return r; +} + +static int svm_range_map_to_gpus(struct svm_range *prange, + unsigned long *bitmap, bool wait) +{ + struct kfd_process_device *pdd; + struct amdgpu_device *bo_adev; + struct amdgpu_device *adev; + struct kfd_process *p; + struct dma_fence *fence = NULL; + uint32_t gpuidx; + int r = 0; + + if (prange->svm_bo && prange->ttm_res) + bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev); + else + bo_adev = NULL; + + p = container_of(prange->svms, struct kfd_process, svms); + for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) { + pr_debug("mapping to gpu idx 0x%x\n", gpuidx); + pdd = kfd_process_device_from_gpuidx(p, gpuidx); + if (!pdd) { + pr_debug("failed to find device idx %d\n", gpuidx); + return -EINVAL; + } + adev = (struct amdgpu_device *)pdd->dev->kgd; + + pdd = kfd_bind_process_to_device(pdd->dev, p); + if (IS_ERR(pdd)) + return -EINVAL; + + if (bo_adev && adev != bo_adev && + !amdgpu_xgmi_same_hive(adev, bo_adev)) { + pr_debug("cannot map to device idx %d\n", gpuidx); + continue; + } + + r = svm_range_map_to_gpu(adev, drm_priv_to_vm(pdd->drm_priv), + prange, prange->dma_addr[gpuidx], + bo_adev, wait ? &fence : NULL); + if (r) + break; + + if (fence) { + r = dma_fence_wait(fence, false); + dma_fence_put(fence); + fence = NULL; + if (r) { + pr_debug("failed %d to dma fence wait\n", r); + break; + } + } + + amdgpu_amdkfd_flush_gpu_tlb_pasid((struct kgd_dev *)adev, + p->pasid); + } + + return r; +} + +struct svm_validate_context { + struct kfd_process *process; + struct svm_range *prange; + bool intr; + unsigned long bitmap[MAX_GPU_INSTANCE]; + struct ttm_validate_buffer tv[MAX_GPU_INSTANCE+1]; + struct list_head validate_list; + struct ww_acquire_ctx ticket; +}; + +static int svm_range_reserve_bos(struct svm_validate_context *ctx) +{ + struct kfd_process_device *pdd; + struct amdgpu_device *adev; + struct amdgpu_vm *vm; + uint32_t gpuidx; + int r; + + INIT_LIST_HEAD(&ctx->validate_list); + for_each_set_bit(gpuidx, ctx->bitmap, MAX_GPU_INSTANCE) { + pdd = kfd_process_device_from_gpuidx(ctx->process, gpuidx); + if (!pdd) { + pr_debug("failed to find device idx %d\n", gpuidx); + return -EINVAL; + } + adev = (struct amdgpu_device *)pdd->dev->kgd; + vm = drm_priv_to_vm(pdd->drm_priv); + + ctx->tv[gpuidx].bo = &vm->root.base.bo->tbo; + ctx->tv[gpuidx].num_shared = 4; + list_add(&ctx->tv[gpuidx].head, &ctx->validate_list); + } + if (ctx->prange->svm_bo && ctx->prange->ttm_res) { + ctx->tv[MAX_GPU_INSTANCE].bo = &ctx->prange->svm_bo->bo->tbo; + ctx->tv[MAX_GPU_INSTANCE].num_shared = 1; + list_add(&ctx->tv[MAX_GPU_INSTANCE].head, &ctx->validate_list); + } + + r = ttm_eu_reserve_buffers(&ctx->ticket, &ctx->validate_list, + ctx->intr, NULL); + if (r) { + pr_debug("failed %d to reserve bo\n", r); + return r; + } + + for_each_set_bit(gpuidx, ctx->bitmap, MAX_GPU_INSTANCE) { + pdd = kfd_process_device_from_gpuidx(ctx->process, gpuidx); + if (!pdd) { + pr_debug("failed to find device idx %d\n", gpuidx); + r = -EINVAL; + goto unreserve_out; + } + adev = (struct amdgpu_device *)pdd->dev->kgd; + + r = amdgpu_vm_validate_pt_bos(adev, drm_priv_to_vm(pdd->drm_priv), + svm_range_bo_validate, NULL); + if (r) { + pr_debug("failed %d validate pt bos\n", r); + goto unreserve_out; + } + } + + return 0; + +unreserve_out: + ttm_eu_backoff_reservation(&ctx->ticket, &ctx->validate_list); + return r; +} + +static void svm_range_unreserve_bos(struct svm_validate_context *ctx) +{ + ttm_eu_backoff_reservation(&ctx->ticket, &ctx->validate_list); +} + +/* + * Validation+GPU mapping with concurrent invalidation (MMU notifiers) + * + * To prevent concurrent destruction or change of range attributes, the + * svm_read_lock must be held. The caller must not hold the svm_write_lock + * because that would block concurrent evictions and lead to deadlocks. To + * serialize concurrent migrations or validations of the same range, the + * prange->migrate_mutex must be held. + * + * For VRAM ranges, the SVM BO must be allocated and valid (protected by its + * eviction fence. + * + * The following sequence ensures race-free validation and GPU mapping: + * + * 1. Reserve page table (and SVM BO if range is in VRAM) + * 2. hmm_range_fault to get page addresses (if system memory) + * 3. DMA-map pages (if system memory) + * 4-a. Take notifier lock + * 4-b. Check that pages still valid (mmu_interval_read_retry) + * 4-c. Check that the range was not split or otherwise invalidated + * 4-d. Update GPU page table + * 4.e. Release notifier lock + * 5. Release page table (and SVM BO) reservation + */ +static int svm_range_validate_and_map(struct mm_struct *mm, + struct svm_range *prange, + int32_t gpuidx, bool intr, bool wait) +{ + struct svm_validate_context ctx; + struct hmm_range *hmm_range; + int r = 0; + + ctx.process = container_of(prange->svms, struct kfd_process, svms); + ctx.prange = prange; + ctx.intr = intr; + + if (gpuidx < MAX_GPU_INSTANCE) { + bitmap_zero(ctx.bitmap, MAX_GPU_INSTANCE); + bitmap_set(ctx.bitmap, gpuidx, 1); + } else if (ctx.process->xnack_enabled) { + bitmap_copy(ctx.bitmap, prange->bitmap_aip, MAX_GPU_INSTANCE); + + /* If prefetch range to GPU, or GPU retry fault migrate range to + * GPU, which has ACCESS attribute to the range, create mapping + * on that GPU. + */ + if (prange->actual_loc) { + gpuidx = kfd_process_gpuidx_from_gpuid(ctx.process, + prange->actual_loc); + if (gpuidx < 0) { + WARN_ONCE(1, "failed get device by id 0x%x\n", + prange->actual_loc); + return -EINVAL; + } + if (test_bit(gpuidx, prange->bitmap_access)) + bitmap_set(ctx.bitmap, gpuidx, 1); + } + } else { + bitmap_or(ctx.bitmap, prange->bitmap_access, + prange->bitmap_aip, MAX_GPU_INSTANCE); + } + + if (bitmap_empty(ctx.bitmap, MAX_GPU_INSTANCE)) + return 0; + + if (prange->actual_loc && !prange->ttm_res) { + /* This should never happen. actual_loc gets set by + * svm_migrate_ram_to_vram after allocating a BO. + */ + WARN(1, "VRAM BO missing during validation\n"); + return -EINVAL; + } + + svm_range_reserve_bos(&ctx); + + if (!prange->actual_loc) { + r = amdgpu_hmm_range_get_pages(&prange->notifier, mm, NULL, + prange->start << PAGE_SHIFT, + prange->npages, &hmm_range, + false, true); + if (r) { + pr_debug("failed %d to get svm range pages\n", r); + goto unreserve_out; + } + + r = svm_range_dma_map(prange, ctx.bitmap, + hmm_range->hmm_pfns); + if (r) { + pr_debug("failed %d to dma map range\n", r); + goto unreserve_out; + } + + prange->validated_once = true; + } + + svm_range_lock(prange); + if (!prange->actual_loc) { + if (amdgpu_hmm_range_get_pages_done(hmm_range)) { + pr_debug("hmm update the range, need validate again\n"); + r = -EAGAIN; + goto unlock_out; + } + } + if (!list_empty(&prange->child_list)) { + pr_debug("range split by unmap in parallel, validate again\n"); + r = -EAGAIN; + goto unlock_out; + } + + r = svm_range_map_to_gpus(prange, ctx.bitmap, wait); + +unlock_out: + svm_range_unlock(prange); +unreserve_out: + svm_range_unreserve_bos(&ctx); + + if (!r) + prange->validate_timestamp = ktime_to_us(ktime_get()); + + return r; +} + +/** + * svm_range_list_lock_and_flush_work - flush pending deferred work + * + * @svms: the svm range list + * @mm: the mm structure + * + * Context: Returns with mmap write lock held, pending deferred work flushed + * + */ +static void +svm_range_list_lock_and_flush_work(struct svm_range_list *svms, + struct mm_struct *mm) +{ +retry_flush_work: + flush_work(&svms->deferred_list_work); + mmap_write_lock(mm); + + if (list_empty(&svms->deferred_range_list)) + return; + mmap_write_unlock(mm); + pr_debug("retry flush\n"); + goto retry_flush_work; +} + +static void svm_range_restore_work(struct work_struct *work) +{ + struct delayed_work *dwork = to_delayed_work(work); + struct amdkfd_process_info *process_info; + struct svm_range_list *svms; + struct svm_range *prange; + struct kfd_process *p; + struct mm_struct *mm; + int evicted_ranges; + int invalid; + int r; + + svms = container_of(dwork, struct svm_range_list, restore_work); + evicted_ranges = atomic_read(&svms->evicted_ranges); + if (!evicted_ranges) + return; + + pr_debug("restore svm ranges\n"); + + /* kfd_process_notifier_release destroys this worker thread. So during + * the lifetime of this thread, kfd_process and mm will be valid. + */ + p = container_of(svms, struct kfd_process, svms); + process_info = p->kgd_process_info; + mm = p->mm; + if (!mm) + return; + + mutex_lock(&process_info->lock); + svm_range_list_lock_and_flush_work(svms, mm); + mutex_lock(&svms->lock); + + evicted_ranges = atomic_read(&svms->evicted_ranges); + + list_for_each_entry(prange, &svms->list, list) { + invalid = atomic_read(&prange->invalid); + if (!invalid) + continue; + + pr_debug("restoring svms 0x%p prange 0x%p [0x%lx %lx] inv %d\n", + prange->svms, prange, prange->start, prange->last, + invalid); + + /* + * If range is migrating, wait for migration is done. + */ + mutex_lock(&prange->migrate_mutex); + + r = svm_range_validate_and_map(mm, prange, MAX_GPU_INSTANCE, + false, true); + if (r) + pr_debug("failed %d to map 0x%lx to gpus\n", r, + prange->start); + + mutex_unlock(&prange->migrate_mutex); + if (r) + goto out_reschedule; + + if (atomic_cmpxchg(&prange->invalid, invalid, 0) != invalid) + goto out_reschedule; + } + + if (atomic_cmpxchg(&svms->evicted_ranges, evicted_ranges, 0) != + evicted_ranges) + goto out_reschedule; + + evicted_ranges = 0; + + r = kgd2kfd_resume_mm(mm); + if (r) { + /* No recovery from this failure. Probably the CP is + * hanging. No point trying again. + */ + pr_debug("failed %d to resume KFD\n", r); + } + + pr_debug("restore svm ranges successfully\n"); + +out_reschedule: + mutex_unlock(&svms->lock); + mmap_write_unlock(mm); + mutex_unlock(&process_info->lock); + + /* If validation failed, reschedule another attempt */ + if (evicted_ranges) { + pr_debug("reschedule to restore svm range\n"); + schedule_delayed_work(&svms->restore_work, + msecs_to_jiffies(AMDGPU_SVM_RANGE_RESTORE_DELAY_MS)); + } +} + +/** + * svm_range_evict - evict svm range + * + * Stop all queues of the process to ensure GPU doesn't access the memory, then + * return to let CPU evict the buffer and proceed CPU pagetable update. + * + * Don't need use lock to sync cpu pagetable invalidation with GPU execution. + * If invalidation happens while restore work is running, restore work will + * restart to ensure to get the latest CPU pages mapping to GPU, then start + * the queues. + */ +static int +svm_range_evict(struct svm_range *prange, struct mm_struct *mm, + unsigned long start, unsigned long last) +{ + struct svm_range_list *svms = prange->svms; + struct kfd_process *p; + int r = 0; + + p = container_of(svms, struct kfd_process, svms); + + pr_debug("invalidate svms 0x%p prange [0x%lx 0x%lx] [0x%lx 0x%lx]\n", + svms, prange->start, prange->last, start, last); + + if (!p->xnack_enabled) { + int evicted_ranges; + + atomic_inc(&prange->invalid); + evicted_ranges = atomic_inc_return(&svms->evicted_ranges); + if (evicted_ranges != 1) + return r; + + pr_debug("evicting svms 0x%p range [0x%lx 0x%lx]\n", + prange->svms, prange->start, prange->last); + + /* First eviction, stop the queues */ + r = kgd2kfd_quiesce_mm(mm); + if (r) + pr_debug("failed to quiesce KFD\n"); + + pr_debug("schedule to restore svm %p ranges\n", svms); + schedule_delayed_work(&svms->restore_work, + msecs_to_jiffies(AMDGPU_SVM_RANGE_RESTORE_DELAY_MS)); + } else { + struct svm_range *pchild; + unsigned long s, l; + + pr_debug("invalidate unmap svms 0x%p [0x%lx 0x%lx] from GPUs\n", + prange->svms, start, last); + list_for_each_entry(pchild, &prange->child_list, child_list) { + mutex_lock_nested(&pchild->lock, 1); + s = max(start, pchild->start); + l = min(last, pchild->last); + if (l >= s) + svm_range_unmap_from_gpus(pchild, s, l); + mutex_unlock(&pchild->lock); + } + s = max(start, prange->start); + l = min(last, prange->last); + if (l >= s) + svm_range_unmap_from_gpus(prange, s, l); + } + + return r; +} + +static struct svm_range *svm_range_clone(struct svm_range *old) +{ + struct svm_range *new; + + new = svm_range_new(old->svms, old->start, old->last); + if (!new) + return NULL; + + if (old->svm_bo) { + new->ttm_res = old->ttm_res; + new->offset = old->offset; + new->svm_bo = svm_range_bo_ref(old->svm_bo); + spin_lock(&new->svm_bo->list_lock); + list_add(&new->svm_bo_list, &new->svm_bo->range_list); + spin_unlock(&new->svm_bo->list_lock); + } + new->flags = old->flags; + new->preferred_loc = old->preferred_loc; + new->prefetch_loc = old->prefetch_loc; + new->actual_loc = old->actual_loc; + new->granularity = old->granularity; + bitmap_copy(new->bitmap_access, old->bitmap_access, MAX_GPU_INSTANCE); + bitmap_copy(new->bitmap_aip, old->bitmap_aip, MAX_GPU_INSTANCE); + + return new; +} + +/** + * svm_range_handle_overlap - split overlap ranges + * @svms: svm range list header + * @new: range added with this attributes + * @start: range added start address, in pages + * @last: range last address, in pages + * @update_list: output, the ranges attributes are updated. For set_attr, this + * will do validation and map to GPUs. For unmap, this will be + * removed and unmap from GPUs + * @insert_list: output, the ranges will be inserted into svms, attributes are + * not changes. For set_attr, this will add into svms. + * @remove_list:output, the ranges will be removed from svms + * @left: the remaining range after overlap, For set_attr, this will be added + * as new range. + * + * Total have 5 overlap cases. + * + * This function handles overlap of an address interval with existing + * struct svm_ranges for applying new attributes. This may require + * splitting existing struct svm_ranges. All changes should be applied to + * the range_list and interval tree transactionally. If any split operation + * fails, the entire update fails. Therefore the existing overlapping + * svm_ranges are cloned and the original svm_ranges left unchanged. If the + * transaction succeeds, the modified clones are added and the originals + * freed. Otherwise the clones are removed and the old svm_ranges remain. + * + * Context: The caller must hold svms->lock + */ +static int +svm_range_handle_overlap(struct svm_range_list *svms, struct svm_range *new, + unsigned long start, unsigned long last, + struct list_head *update_list, + struct list_head *insert_list, + struct list_head *remove_list, + unsigned long *left) +{ + struct interval_tree_node *node; + struct svm_range *prange; + struct svm_range *tmp; + int r = 0; + + INIT_LIST_HEAD(update_list); + INIT_LIST_HEAD(insert_list); + INIT_LIST_HEAD(remove_list); + + node = interval_tree_iter_first(&svms->objects, start, last); + while (node) { + struct interval_tree_node *next; + struct svm_range *old; + unsigned long next_start; + + pr_debug("found overlap node [0x%lx 0x%lx]\n", node->start, + node->last); + + old = container_of(node, struct svm_range, it_node); + next = interval_tree_iter_next(node, start, last); + next_start = min(node->last, last) + 1; + + if (node->start < start || node->last > last) { + /* node intersects the updated range, clone+split it */ + prange = svm_range_clone(old); + if (!prange) { + r = -ENOMEM; + goto out; + } + + list_add(&old->remove_list, remove_list); + list_add(&prange->insert_list, insert_list); + + if (node->start < start) { + pr_debug("change old range start\n"); + r = svm_range_split_head(prange, new, start, + insert_list); + if (r) + goto out; + } + if (node->last > last) { + pr_debug("change old range last\n"); + r = svm_range_split_tail(prange, new, last, + insert_list); + if (r) + goto out; + } + } else { + /* The node is contained within start..last, + * just update it + */ + prange = old; + } + + if (!svm_range_is_same_attrs(prange, new)) + list_add(&prange->update_list, update_list); + + /* insert a new node if needed */ + if (node->start > start) { + prange = svm_range_new(prange->svms, start, + node->start - 1); + if (!prange) { + r = -ENOMEM; + goto out; + } + + list_add(&prange->insert_list, insert_list); + list_add(&prange->update_list, update_list); + } + + node = next; + start = next_start; + } + + if (left && start <= last) + *left = last - start + 1; + +out: + if (r) + list_for_each_entry_safe(prange, tmp, insert_list, insert_list) + svm_range_free(prange); + + return r; +} + +static void +svm_range_update_notifier_and_interval_tree(struct mm_struct *mm, + struct svm_range *prange) +{ + unsigned long start; + unsigned long last; + + start = prange->notifier.interval_tree.start >> PAGE_SHIFT; + last = prange->notifier.interval_tree.last >> PAGE_SHIFT; + + if (prange->start == start && prange->last == last) + return; + + pr_debug("up notifier 0x%p prange 0x%p [0x%lx 0x%lx] [0x%lx 0x%lx]\n", + prange->svms, prange, start, last, prange->start, + prange->last); + + if (start != 0 && last != 0) { + interval_tree_remove(&prange->it_node, &prange->svms->objects); + svm_range_remove_notifier(prange); + } + prange->it_node.start = prange->start; + prange->it_node.last = prange->last; + + interval_tree_insert(&prange->it_node, &prange->svms->objects); + svm_range_add_notifier_locked(mm, prange); +} + +static void +svm_range_handle_list_op(struct svm_range_list *svms, struct svm_range *prange) +{ + struct mm_struct *mm = prange->work_item.mm; + + switch (prange->work_item.op) { + case SVM_OP_NULL: + pr_debug("NULL OP 0x%p prange 0x%p [0x%lx 0x%lx]\n", + svms, prange, prange->start, prange->last); + break; + case SVM_OP_UNMAP_RANGE: + pr_debug("remove 0x%p prange 0x%p [0x%lx 0x%lx]\n", + svms, prange, prange->start, prange->last); + svm_range_unlink(prange); + svm_range_remove_notifier(prange); + svm_range_free(prange); + break; + case SVM_OP_UPDATE_RANGE_NOTIFIER: + pr_debug("update notifier 0x%p prange 0x%p [0x%lx 0x%lx]\n", + svms, prange, prange->start, prange->last); + svm_range_update_notifier_and_interval_tree(mm, prange); + break; + case SVM_OP_UPDATE_RANGE_NOTIFIER_AND_MAP: + pr_debug("update and map 0x%p prange 0x%p [0x%lx 0x%lx]\n", + svms, prange, prange->start, prange->last); + svm_range_update_notifier_and_interval_tree(mm, prange); + /* TODO: implement deferred validation and mapping */ + break; + case SVM_OP_ADD_RANGE: + pr_debug("add 0x%p prange 0x%p [0x%lx 0x%lx]\n", svms, prange, + prange->start, prange->last); + svm_range_add_to_svms(prange); + svm_range_add_notifier_locked(mm, prange); + break; + case SVM_OP_ADD_RANGE_AND_MAP: + pr_debug("add and map 0x%p prange 0x%p [0x%lx 0x%lx]\n", svms, + prange, prange->start, prange->last); + svm_range_add_to_svms(prange); + svm_range_add_notifier_locked(mm, prange); + /* TODO: implement deferred validation and mapping */ + break; + default: + WARN_ONCE(1, "Unknown prange 0x%p work op %d\n", prange, + prange->work_item.op); + } +} + +static void svm_range_drain_retry_fault(struct svm_range_list *svms) +{ + struct kfd_process_device *pdd; + struct amdgpu_device *adev; + struct kfd_process *p; + uint32_t i; + + p = container_of(svms, struct kfd_process, svms); + + for (i = 0; i < p->n_pdds; i++) { + pdd = p->pdds[i]; + if (!pdd) + continue; + + pr_debug("drain retry fault gpu %d svms %p\n", i, svms); + adev = (struct amdgpu_device *)pdd->dev->kgd; + + amdgpu_ih_wait_on_checkpoint_process(adev, &adev->irq.ih1); + pr_debug("drain retry fault gpu %d svms 0x%p done\n", i, svms); + } +} + +static void svm_range_deferred_list_work(struct work_struct *work) +{ + struct svm_range_list *svms; + struct svm_range *prange; + struct mm_struct *mm; + + svms = container_of(work, struct svm_range_list, deferred_list_work); + pr_debug("enter svms 0x%p\n", svms); + + spin_lock(&svms->deferred_list_lock); + while (!list_empty(&svms->deferred_range_list)) { + prange = list_first_entry(&svms->deferred_range_list, + struct svm_range, deferred_list); + spin_unlock(&svms->deferred_list_lock); + pr_debug("prange 0x%p [0x%lx 0x%lx] op %d\n", prange, + prange->start, prange->last, prange->work_item.op); + + /* Make sure no stale retry fault coming after range is freed */ + if (prange->work_item.op == SVM_OP_UNMAP_RANGE) + svm_range_drain_retry_fault(prange->svms); + + mm = prange->work_item.mm; + mmap_write_lock(mm); + mutex_lock(&svms->lock); + + /* Remove from deferred_list must be inside mmap write lock, + * otherwise, svm_range_list_lock_and_flush_work may hold mmap + * write lock, and continue because deferred_list is empty, then + * deferred_list handle is blocked by mmap write lock. + */ + spin_lock(&svms->deferred_list_lock); + list_del_init(&prange->deferred_list); + spin_unlock(&svms->deferred_list_lock); + + mutex_lock(&prange->migrate_mutex); + while (!list_empty(&prange->child_list)) { + struct svm_range *pchild; + + pchild = list_first_entry(&prange->child_list, + struct svm_range, child_list); + pr_debug("child prange 0x%p op %d\n", pchild, + pchild->work_item.op); + list_del_init(&pchild->child_list); + svm_range_handle_list_op(svms, pchild); + } + mutex_unlock(&prange->migrate_mutex); + + svm_range_handle_list_op(svms, prange); + mutex_unlock(&svms->lock); + mmap_write_unlock(mm); + + spin_lock(&svms->deferred_list_lock); + } + spin_unlock(&svms->deferred_list_lock); + + pr_debug("exit svms 0x%p\n", svms); +} + +void +svm_range_add_list_work(struct svm_range_list *svms, struct svm_range *prange, + struct mm_struct *mm, enum svm_work_list_ops op) +{ + spin_lock(&svms->deferred_list_lock); + /* if prange is on the deferred list */ + if (!list_empty(&prange->deferred_list)) { + pr_debug("update exist prange 0x%p work op %d\n", prange, op); + WARN_ONCE(prange->work_item.mm != mm, "unmatch mm\n"); + if (op != SVM_OP_NULL && + prange->work_item.op != SVM_OP_UNMAP_RANGE) + prange->work_item.op = op; + } else { + prange->work_item.op = op; + prange->work_item.mm = mm; + list_add_tail(&prange->deferred_list, + &prange->svms->deferred_range_list); + pr_debug("add prange 0x%p [0x%lx 0x%lx] to work list op %d\n", + prange, prange->start, prange->last, op); + } + spin_unlock(&svms->deferred_list_lock); +} + +void schedule_deferred_list_work(struct svm_range_list *svms) +{ + spin_lock(&svms->deferred_list_lock); + if (!list_empty(&svms->deferred_range_list)) + schedule_work(&svms->deferred_list_work); + spin_unlock(&svms->deferred_list_lock); +} + +static void +svm_range_unmap_split(struct mm_struct *mm, struct svm_range *parent, + struct svm_range *prange, unsigned long start, + unsigned long last) +{ + struct svm_range *head; + struct svm_range *tail; + + if (prange->work_item.op == SVM_OP_UNMAP_RANGE) { + pr_debug("prange 0x%p [0x%lx 0x%lx] is already freed\n", prange, + prange->start, prange->last); + return; + } + if (start > prange->last || last < prange->start) + return; + + head = tail = prange; + if (start > prange->start) + svm_range_split(prange, prange->start, start - 1, &tail); + if (last < tail->last) + svm_range_split(tail, last + 1, tail->last, &head); + + if (head != prange && tail != prange) { + svm_range_add_child(parent, mm, head, SVM_OP_UNMAP_RANGE); + svm_range_add_child(parent, mm, tail, SVM_OP_ADD_RANGE); + } else if (tail != prange) { + svm_range_add_child(parent, mm, tail, SVM_OP_UNMAP_RANGE); + } else if (head != prange) { + svm_range_add_child(parent, mm, head, SVM_OP_UNMAP_RANGE); + } else if (parent != prange) { + prange->work_item.op = SVM_OP_UNMAP_RANGE; + } +} + +static void +svm_range_unmap_from_cpu(struct mm_struct *mm, struct svm_range *prange, + unsigned long start, unsigned long last) +{ + struct svm_range_list *svms; + struct svm_range *pchild; + struct kfd_process *p; + unsigned long s, l; + bool unmap_parent; + + p = kfd_lookup_process_by_mm(mm); + if (!p) + return; + svms = &p->svms; + + pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx] [0x%lx 0x%lx]\n", svms, + prange, prange->start, prange->last, start, last); + + unmap_parent = start <= prange->start && last >= prange->last; + + list_for_each_entry(pchild, &prange->child_list, child_list) { + mutex_lock_nested(&pchild->lock, 1); + s = max(start, pchild->start); + l = min(last, pchild->last); + if (l >= s) + svm_range_unmap_from_gpus(pchild, s, l); + svm_range_unmap_split(mm, prange, pchild, start, last); + mutex_unlock(&pchild->lock); + } + s = max(start, prange->start); + l = min(last, prange->last); + if (l >= s) + svm_range_unmap_from_gpus(prange, s, l); + svm_range_unmap_split(mm, prange, prange, start, last); + + if (unmap_parent) + svm_range_add_list_work(svms, prange, mm, SVM_OP_UNMAP_RANGE); + else + svm_range_add_list_work(svms, prange, mm, + SVM_OP_UPDATE_RANGE_NOTIFIER); + schedule_deferred_list_work(svms); + + kfd_unref_process(p); +} + +/** + * svm_range_cpu_invalidate_pagetables - interval notifier callback + * + * If event is MMU_NOTIFY_UNMAP, this is from CPU unmap range, otherwise, it + * is from migration, or CPU page invalidation callback. + * + * For unmap event, unmap range from GPUs, remove prange from svms in a delayed + * work thread, and split prange if only part of prange is unmapped. + * + * For invalidation event, if GPU retry fault is not enabled, evict the queues, + * then schedule svm_range_restore_work to update GPU mapping and resume queues. + * If GPU retry fault is enabled, unmap the svm range from GPU, retry fault will + * update GPU mapping to recover. + * + * Context: mmap lock, notifier_invalidate_start lock are held + * for invalidate event, prange lock is held if this is from migration + */ +static bool +svm_range_cpu_invalidate_pagetables(struct mmu_interval_notifier *mni, + const struct mmu_notifier_range *range, + unsigned long cur_seq) +{ + struct svm_range *prange; + unsigned long start; + unsigned long last; + + if (range->event == MMU_NOTIFY_RELEASE) + return true; + + start = mni->interval_tree.start; + last = mni->interval_tree.last; + start = (start > range->start ? start : range->start) >> PAGE_SHIFT; + last = (last < (range->end - 1) ? last : range->end - 1) >> PAGE_SHIFT; + pr_debug("[0x%lx 0x%lx] range[0x%lx 0x%lx] notifier[0x%lx 0x%lx] %d\n", + start, last, range->start >> PAGE_SHIFT, + (range->end - 1) >> PAGE_SHIFT, + mni->interval_tree.start >> PAGE_SHIFT, + mni->interval_tree.last >> PAGE_SHIFT, range->event); + + prange = container_of(mni, struct svm_range, notifier); + + svm_range_lock(prange); + mmu_interval_set_seq(mni, cur_seq); + + switch (range->event) { + case MMU_NOTIFY_UNMAP: + svm_range_unmap_from_cpu(mni->mm, prange, start, last); + break; + default: + svm_range_evict(prange, mni->mm, start, last); + break; + } + + svm_range_unlock(prange); + + return true; +} + +/** + * svm_range_from_addr - find svm range from fault address + * @svms: svm range list header + * @addr: address to search range interval tree, in pages + * @parent: parent range if range is on child list + * + * Context: The caller must hold svms->lock + * + * Return: the svm_range found or NULL + */ +struct svm_range * +svm_range_from_addr(struct svm_range_list *svms, unsigned long addr, + struct svm_range **parent) +{ + struct interval_tree_node *node; + struct svm_range *prange; + struct svm_range *pchild; + + node = interval_tree_iter_first(&svms->objects, addr, addr); + if (!node) + return NULL; + + prange = container_of(node, struct svm_range, it_node); + pr_debug("address 0x%lx prange [0x%lx 0x%lx] node [0x%lx 0x%lx]\n", + addr, prange->start, prange->last, node->start, node->last); + + if (addr >= prange->start && addr <= prange->last) { + if (parent) + *parent = prange; + return prange; + } + list_for_each_entry(pchild, &prange->child_list, child_list) + if (addr >= pchild->start && addr <= pchild->last) { + pr_debug("found address 0x%lx pchild [0x%lx 0x%lx]\n", + addr, pchild->start, pchild->last); + if (parent) + *parent = prange; + return pchild; + } + + return NULL; +} + +/* svm_range_best_restore_location - decide the best fault restore location + * @prange: svm range structure + * @adev: the GPU on which vm fault happened + * + * This is only called when xnack is on, to decide the best location to restore + * the range mapping after GPU vm fault. Caller uses the best location to do + * migration if actual loc is not best location, then update GPU page table + * mapping to the best location. + * + * If vm fault gpu is range preferred loc, the best_loc is preferred loc. + * If vm fault gpu idx is on range ACCESSIBLE bitmap, best_loc is vm fault gpu + * If vm fault gpu idx is on range ACCESSIBLE_IN_PLACE bitmap, then + * if range actual loc is cpu, best_loc is cpu + * if vm fault gpu is on xgmi same hive of range actual loc gpu, best_loc is + * range actual loc. + * Otherwise, GPU no access, best_loc is -1. + * + * Return: + * -1 means vm fault GPU no access + * 0 for CPU or GPU id + */ +static int32_t +svm_range_best_restore_location(struct svm_range *prange, + struct amdgpu_device *adev, + int32_t *gpuidx) +{ + struct amdgpu_device *bo_adev; + struct kfd_process *p; + uint32_t gpuid; + int r; + + p = container_of(prange->svms, struct kfd_process, svms); + + r = kfd_process_gpuid_from_kgd(p, adev, &gpuid, gpuidx); + if (r < 0) { + pr_debug("failed to get gpuid from kgd\n"); + return -1; + } + + if (prange->preferred_loc == gpuid) + return prange->preferred_loc; + + if (test_bit(*gpuidx, prange->bitmap_access)) + return gpuid; + + if (test_bit(*gpuidx, prange->bitmap_aip)) { + if (!prange->actual_loc) + return 0; + + bo_adev = svm_range_get_adev_by_id(prange, prange->actual_loc); + if (amdgpu_xgmi_same_hive(adev, bo_adev)) + return prange->actual_loc; + else + return 0; + } + + return -1; +} +static int +svm_range_get_range_boundaries(struct kfd_process *p, int64_t addr, + unsigned long *start, unsigned long *last) +{ + struct vm_area_struct *vma; + struct interval_tree_node *node; + unsigned long start_limit, end_limit; + + vma = find_vma(p->mm, addr << PAGE_SHIFT); + if (!vma || (addr << PAGE_SHIFT) < vma->vm_start) { + pr_debug("VMA does not exist in address [0x%llx]\n", addr); + return -EFAULT; + } + start_limit = max(vma->vm_start >> PAGE_SHIFT, + (unsigned long)ALIGN_DOWN(addr, 2UL << 8)); + end_limit = min(vma->vm_end >> PAGE_SHIFT, + (unsigned long)ALIGN(addr + 1, 2UL << 8)); + /* First range that starts after the fault address */ + node = interval_tree_iter_first(&p->svms.objects, addr + 1, ULONG_MAX); + if (node) { + end_limit = min(end_limit, node->start); + /* Last range that ends before the fault address */ + node = container_of(rb_prev(&node->rb), + struct interval_tree_node, rb); + } else { + /* Last range must end before addr because + * there was no range after addr + */ + node = container_of(rb_last(&p->svms.objects.rb_root), + struct interval_tree_node, rb); + } + if (node) { + if (node->last >= addr) { + WARN(1, "Overlap with prev node and page fault addr\n"); + return -EFAULT; + } + start_limit = max(start_limit, node->last + 1); + } + + *start = start_limit; + *last = end_limit - 1; + + pr_debug("vma start: 0x%lx start: 0x%lx vma end: 0x%lx last: 0x%lx\n", + vma->vm_start >> PAGE_SHIFT, *start, + vma->vm_end >> PAGE_SHIFT, *last); + + return 0; + +} +static struct +svm_range *svm_range_create_unregistered_range(struct amdgpu_device *adev, + struct kfd_process *p, + struct mm_struct *mm, + int64_t addr) +{ + struct svm_range *prange = NULL; + unsigned long start, last; + uint32_t gpuid, gpuidx; + + if (svm_range_get_range_boundaries(p, addr, &start, &last)) + return NULL; + + prange = svm_range_new(&p->svms, start, last); + if (!prange) { + pr_debug("Failed to create prange in address [0x%llx]\n", addr); + return NULL; + } + if (kfd_process_gpuid_from_kgd(p, adev, &gpuid, &gpuidx)) { + pr_debug("failed to get gpuid from kgd\n"); + svm_range_free(prange); + return NULL; + } + + svm_range_add_to_svms(prange); + svm_range_add_notifier_locked(mm, prange); + + return prange; +} + +/* svm_range_skip_recover - decide if prange can be recovered + * @prange: svm range structure + * + * GPU vm retry fault handle skip recover the range for cases: + * 1. prange is on deferred list to be removed after unmap, it is stale fault, + * deferred list work will drain the stale fault before free the prange. + * 2. prange is on deferred list to add interval notifier after split, or + * 3. prange is child range, it is split from parent prange, recover later + * after interval notifier is added. + * + * Return: true to skip recover, false to recover + */ +static bool svm_range_skip_recover(struct svm_range *prange) +{ + struct svm_range_list *svms = prange->svms; + + spin_lock(&svms->deferred_list_lock); + if (list_empty(&prange->deferred_list) && + list_empty(&prange->child_list)) { + spin_unlock(&svms->deferred_list_lock); + return false; + } + spin_unlock(&svms->deferred_list_lock); + + if (prange->work_item.op == SVM_OP_UNMAP_RANGE) { + pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx] unmapped\n", + svms, prange, prange->start, prange->last); + return true; + } + if (prange->work_item.op == SVM_OP_ADD_RANGE_AND_MAP || + prange->work_item.op == SVM_OP_ADD_RANGE) { + pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx] not added yet\n", + svms, prange, prange->start, prange->last); + return true; + } + return false; +} + +int +svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid, + uint64_t addr) +{ + struct mm_struct *mm = NULL; + struct svm_range_list *svms; + struct svm_range *prange; + struct kfd_process *p; + uint64_t timestamp; + int32_t best_loc, gpuidx; + bool write_locked = false; + int r = 0; + + p = kfd_lookup_process_by_pasid(pasid); + if (!p) { + pr_debug("kfd process not founded pasid 0x%x\n", pasid); + return -ESRCH; + } + if (!p->xnack_enabled) { + pr_debug("XNACK not enabled for pasid 0x%x\n", pasid); + return -EFAULT; + } + svms = &p->svms; + + pr_debug("restoring svms 0x%p fault address 0x%llx\n", svms, addr); + + mm = get_task_mm(p->lead_thread); + if (!mm) { + pr_debug("svms 0x%p failed to get mm\n", svms); + r = -ESRCH; + goto out; + } + + mmap_read_lock(mm); +retry_write_locked: + mutex_lock(&svms->lock); + prange = svm_range_from_addr(svms, addr, NULL); + if (!prange) { + pr_debug("failed to find prange svms 0x%p address [0x%llx]\n", + svms, addr); + if (!write_locked) { + /* Need the write lock to create new range with MMU notifier. + * Also flush pending deferred work to make sure the interval + * tree is up to date before we add a new range + */ + mutex_unlock(&svms->lock); + mmap_read_unlock(mm); + mmap_write_lock(mm); + write_locked = true; + goto retry_write_locked; + } + prange = svm_range_create_unregistered_range(adev, p, mm, addr); + if (!prange) { + pr_debug("failed to create unregistered range svms 0x%p address [0x%llx]\n", + svms, addr); + mmap_write_downgrade(mm); + r = -EFAULT; + goto out_unlock_svms; + } + } + if (write_locked) + mmap_write_downgrade(mm); + + mutex_lock(&prange->migrate_mutex); + + if (svm_range_skip_recover(prange)) { + amdgpu_gmc_filter_faults_remove(adev, addr, pasid); + goto out_unlock_range; + } + + timestamp = ktime_to_us(ktime_get()) - prange->validate_timestamp; + /* skip duplicate vm fault on different pages of same range */ + if (timestamp < AMDGPU_SVM_RANGE_RETRY_FAULT_PENDING) { + pr_debug("svms 0x%p [0x%lx %lx] already restored\n", + svms, prange->start, prange->last); + goto out_unlock_range; + } + + best_loc = svm_range_best_restore_location(prange, adev, &gpuidx); + if (best_loc == -1) { + pr_debug("svms %p failed get best restore loc [0x%lx 0x%lx]\n", + svms, prange->start, prange->last); + r = -EACCES; + goto out_unlock_range; + } + + pr_debug("svms %p [0x%lx 0x%lx] best restore 0x%x, actual loc 0x%x\n", + svms, prange->start, prange->last, best_loc, + prange->actual_loc); + + if (prange->actual_loc != best_loc) { + if (best_loc) { + r = svm_migrate_to_vram(prange, best_loc, mm); + if (r) { + pr_debug("svm_migrate_to_vram failed (%d) at %llx, falling back to system memory\n", + r, addr); + /* Fallback to system memory if migration to + * VRAM failed + */ + if (prange->actual_loc) + r = svm_migrate_vram_to_ram(prange, mm); + else + r = 0; + } + } else { + r = svm_migrate_vram_to_ram(prange, mm); + } + if (r) { + pr_debug("failed %d to migrate svms %p [0x%lx 0x%lx]\n", + r, svms, prange->start, prange->last); + goto out_unlock_range; + } + } + + r = svm_range_validate_and_map(mm, prange, gpuidx, false, false); + if (r) + pr_debug("failed %d to map svms 0x%p [0x%lx 0x%lx] to gpus\n", + r, svms, prange->start, prange->last); + +out_unlock_range: + mutex_unlock(&prange->migrate_mutex); +out_unlock_svms: + mutex_unlock(&svms->lock); + mmap_read_unlock(mm); + mmput(mm); +out: + kfd_unref_process(p); + + if (r == -EAGAIN) { + pr_debug("recover vm fault later\n"); + amdgpu_gmc_filter_faults_remove(adev, addr, pasid); + r = 0; + } + return r; +} + +void svm_range_list_fini(struct kfd_process *p) +{ + struct svm_range *prange; + struct svm_range *next; + + pr_debug("pasid 0x%x svms 0x%p\n", p->pasid, &p->svms); + + /* Ensure list work is finished before process is destroyed */ + flush_work(&p->svms.deferred_list_work); + + list_for_each_entry_safe(prange, next, &p->svms.list, list) { + svm_range_unlink(prange); + svm_range_remove_notifier(prange); + svm_range_free(prange); + } + + mutex_destroy(&p->svms.lock); + + pr_debug("pasid 0x%x svms 0x%p done\n", p->pasid, &p->svms); +} + +int svm_range_list_init(struct kfd_process *p) +{ + struct svm_range_list *svms = &p->svms; + + svms->objects = RB_ROOT_CACHED; + mutex_init(&svms->lock); + INIT_LIST_HEAD(&svms->list); + atomic_set(&svms->evicted_ranges, 0); + INIT_DELAYED_WORK(&svms->restore_work, svm_range_restore_work); + INIT_WORK(&svms->deferred_list_work, svm_range_deferred_list_work); + INIT_LIST_HEAD(&svms->deferred_range_list); + spin_lock_init(&svms->deferred_list_lock); + + return 0; +} + +/** + * svm_range_is_valid - check if virtual address range is valid + * @mm: current process mm_struct + * @start: range start address, in pages + * @size: range size, in pages + * + * Valid virtual address range means it belongs to one or more VMAs + * + * Context: Process context + * + * Return: + * true - valid svm range + * false - invalid svm range + */ +static bool +svm_range_is_valid(struct mm_struct *mm, uint64_t start, uint64_t size) +{ + const unsigned long device_vma = VM_IO | VM_PFNMAP | VM_MIXEDMAP; + struct vm_area_struct *vma; + unsigned long end; + + start <<= PAGE_SHIFT; + end = start + (size << PAGE_SHIFT); + + do { + vma = find_vma(mm, start); + if (!vma || start < vma->vm_start || + (vma->vm_flags & device_vma)) + return false; + start = min(end, vma->vm_end); + } while (start < end); + + return true; +} + +/** + * svm_range_add - add svm range and handle overlap + * @p: the range add to this process svms + * @start: page size aligned + * @size: page size aligned + * @nattr: number of attributes + * @attrs: array of attributes + * @update_list: output, the ranges need validate and update GPU mapping + * @insert_list: output, the ranges need insert to svms + * @remove_list: output, the ranges are replaced and need remove from svms + * + * Check if the virtual address range has overlap with the registered ranges, + * split the overlapped range, copy and adjust pages address and vram nodes in + * old and new ranges. + * + * Context: Process context, caller must hold svms->lock + * + * Return: + * 0 - OK, otherwise error code + */ +static int +svm_range_add(struct kfd_process *p, uint64_t start, uint64_t size, + uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs, + struct list_head *update_list, struct list_head *insert_list, + struct list_head *remove_list) +{ + uint64_t last = start + size - 1UL; + struct svm_range_list *svms; + struct svm_range new = {0}; + struct svm_range *prange; + unsigned long left = 0; + int r = 0; + + pr_debug("svms 0x%p [0x%llx 0x%llx]\n", &p->svms, start, last); + + svm_range_apply_attrs(p, &new, nattr, attrs); + + svms = &p->svms; + + r = svm_range_handle_overlap(svms, &new, start, last, update_list, + insert_list, remove_list, &left); + if (r) + return r; + + if (left) { + prange = svm_range_new(svms, last - left + 1, last); + list_add(&prange->insert_list, insert_list); + list_add(&prange->update_list, update_list); + } + + return 0; +} + +/* svm_range_best_prefetch_location - decide the best prefetch location + * @prange: svm range structure + * + * For xnack off: + * If range map to single GPU, the best acutal location is prefetch loc, which + * can be CPU or GPU. + * + * If range map to multiple GPUs, only if mGPU connection on xgmi same hive, + * the best actual location could be prefetch_loc GPU. If mGPU connection on + * PCIe, the best actual location is always CPU, because GPU cannot access vram + * of other GPUs, assuming PCIe small bar (large bar support is not upstream). + * + * For xnack on: + * The best actual location is prefetch location. If mGPU connection on xgmi + * same hive, range map to multiple GPUs. Otherwise, the range only map to + * actual location GPU. Other GPU access vm fault will trigger migration. + * + * Context: Process context + * + * Return: + * 0 for CPU or GPU id + */ +static uint32_t +svm_range_best_prefetch_location(struct svm_range *prange) +{ + DECLARE_BITMAP(bitmap, MAX_GPU_INSTANCE); + uint32_t best_loc = prange->prefetch_loc; + struct kfd_process_device *pdd; + struct amdgpu_device *bo_adev; + struct amdgpu_device *adev; + struct kfd_process *p; + uint32_t gpuidx; + + p = container_of(prange->svms, struct kfd_process, svms); + + /* xnack on */ + if (p->xnack_enabled) + goto out; + + /* xnack off */ + if (!best_loc || best_loc == KFD_IOCTL_SVM_LOCATION_UNDEFINED) + goto out; + + bo_adev = svm_range_get_adev_by_id(prange, best_loc); + if (!bo_adev) { + WARN_ONCE(1, "failed to get device by id 0x%x\n", best_loc); + best_loc = 0; + goto out; + } + bitmap_or(bitmap, prange->bitmap_access, prange->bitmap_aip, + MAX_GPU_INSTANCE); + + for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) { + pdd = kfd_process_device_from_gpuidx(p, gpuidx); + if (!pdd) { + pr_debug("failed to get device by idx 0x%x\n", gpuidx); + continue; + } + adev = (struct amdgpu_device *)pdd->dev->kgd; + + if (adev == bo_adev) + continue; + + if (!amdgpu_xgmi_same_hive(adev, bo_adev)) { + best_loc = 0; + break; + } + } + +out: + pr_debug("xnack %d svms 0x%p [0x%lx 0x%lx] best loc 0x%x\n", + p->xnack_enabled, &p->svms, prange->start, prange->last, + best_loc); + + return best_loc; +} + +/* FIXME: This is a workaround for page locking bug when some pages are + * invalid during migration to VRAM + */ +void svm_range_prefault(struct svm_range *prange, struct mm_struct *mm) +{ + struct hmm_range *hmm_range; + int r; + + if (prange->validated_once) + return; + + r = amdgpu_hmm_range_get_pages(&prange->notifier, mm, NULL, + prange->start << PAGE_SHIFT, + prange->npages, &hmm_range, + false, true); + if (!r) { + amdgpu_hmm_range_get_pages_done(hmm_range); + prange->validated_once = true; + } +} + +/* svm_range_trigger_migration - start page migration if prefetch loc changed + * @mm: current process mm_struct + * @prange: svm range structure + * @migrated: output, true if migration is triggered + * + * If range perfetch_loc is GPU, actual loc is cpu 0, then migrate the range + * from ram to vram. + * If range prefetch_loc is cpu 0, actual loc is GPU, then migrate the range + * from vram to ram. + * + * If GPU vm fault retry is not enabled, migration interact with MMU notifier + * and restore work: + * 1. migrate_vma_setup invalidate pages, MMU notifier callback svm_range_evict + * stops all queues, schedule restore work + * 2. svm_range_restore_work wait for migration is done by + * a. svm_range_validate_vram takes prange->migrate_mutex + * b. svm_range_validate_ram HMM get pages wait for CPU fault handle returns + * 3. restore work update mappings of GPU, resume all queues. + * + * Context: Process context + * + * Return: + * 0 - OK, otherwise - error code of migration + */ +static int +svm_range_trigger_migration(struct mm_struct *mm, struct svm_range *prange, + bool *migrated) +{ + uint32_t best_loc; + int r = 0; + + *migrated = false; + best_loc = svm_range_best_prefetch_location(prange); + + if (best_loc == KFD_IOCTL_SVM_LOCATION_UNDEFINED || + best_loc == prange->actual_loc) + return 0; + + /* + * Prefetch to GPU without host access flag, set actual_loc to gpu, then + * validate on gpu and map to gpus will be handled afterwards. + */ + if (best_loc && !prange->actual_loc && + !(prange->flags & KFD_IOCTL_SVM_FLAG_HOST_ACCESS)) { + prange->actual_loc = best_loc; + return 0; + } + + if (!best_loc) { + r = svm_migrate_vram_to_ram(prange, mm); + *migrated = !r; + return r; + } + + r = svm_migrate_to_vram(prange, best_loc, mm); + *migrated = !r; + + return r; +} + +int svm_range_schedule_evict_svm_bo(struct amdgpu_amdkfd_fence *fence) +{ + if (!fence) + return -EINVAL; + + if (dma_fence_is_signaled(&fence->base)) + return 0; + + if (fence->svm_bo) { + WRITE_ONCE(fence->svm_bo->evicting, 1); + schedule_work(&fence->svm_bo->eviction_work); + } + + return 0; +} + +static void svm_range_evict_svm_bo_worker(struct work_struct *work) +{ + struct svm_range_bo *svm_bo; + struct kfd_process *p; + struct mm_struct *mm; + + svm_bo = container_of(work, struct svm_range_bo, eviction_work); + if (!svm_bo_ref_unless_zero(svm_bo)) + return; /* svm_bo was freed while eviction was pending */ + + /* svm_range_bo_release destroys this worker thread. So during + * the lifetime of this thread, kfd_process and mm will be valid. + */ + p = container_of(svm_bo->svms, struct kfd_process, svms); + mm = p->mm; + if (!mm) + return; + + mmap_read_lock(mm); + spin_lock(&svm_bo->list_lock); + while (!list_empty(&svm_bo->range_list)) { + struct svm_range *prange = + list_first_entry(&svm_bo->range_list, + struct svm_range, svm_bo_list); + list_del_init(&prange->svm_bo_list); + spin_unlock(&svm_bo->list_lock); + + pr_debug("svms 0x%p [0x%lx 0x%lx]\n", prange->svms, + prange->start, prange->last); + + mutex_lock(&prange->migrate_mutex); + svm_migrate_vram_to_ram(prange, svm_bo->eviction_fence->mm); + + mutex_lock(&prange->lock); + prange->svm_bo = NULL; + mutex_unlock(&prange->lock); + + mutex_unlock(&prange->migrate_mutex); + + spin_lock(&svm_bo->list_lock); + } + spin_unlock(&svm_bo->list_lock); + mmap_read_unlock(mm); + + dma_fence_signal(&svm_bo->eviction_fence->base); + /* This is the last reference to svm_bo, after svm_range_vram_node_free + * has been called in svm_migrate_vram_to_ram + */ + WARN_ONCE(kref_read(&svm_bo->kref) != 1, "This was not the last reference\n"); + svm_range_bo_unref(svm_bo); +} + +static int +svm_range_set_attr(struct kfd_process *p, uint64_t start, uint64_t size, + uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs) +{ + struct amdkfd_process_info *process_info = p->kgd_process_info; + struct mm_struct *mm = current->mm; + struct list_head update_list; + struct list_head insert_list; + struct list_head remove_list; + struct svm_range_list *svms; + struct svm_range *prange; + struct svm_range *next; + int r = 0; + + pr_debug("pasid 0x%x svms 0x%p [0x%llx 0x%llx] pages 0x%llx\n", + p->pasid, &p->svms, start, start + size - 1, size); + + r = svm_range_check_attr(p, nattr, attrs); + if (r) + return r; + + svms = &p->svms; + + mutex_lock(&process_info->lock); + + svm_range_list_lock_and_flush_work(svms, mm); + + if (!svm_range_is_valid(mm, start, size)) { + pr_debug("invalid range\n"); + r = -EFAULT; + mmap_write_unlock(mm); + goto out; + } + + mutex_lock(&svms->lock); + + /* Add new range and split existing ranges as needed */ + r = svm_range_add(p, start, size, nattr, attrs, &update_list, + &insert_list, &remove_list); + if (r) { + mutex_unlock(&svms->lock); + mmap_write_unlock(mm); + goto out; + } + /* Apply changes as a transaction */ + list_for_each_entry_safe(prange, next, &insert_list, insert_list) { + svm_range_add_to_svms(prange); + svm_range_add_notifier_locked(mm, prange); + } + list_for_each_entry(prange, &update_list, update_list) { + svm_range_apply_attrs(p, prange, nattr, attrs); + /* TODO: unmap ranges from GPU that lost access */ + } + list_for_each_entry_safe(prange, next, &remove_list, + remove_list) { + pr_debug("unlink old 0x%p prange 0x%p [0x%lx 0x%lx]\n", + prange->svms, prange, prange->start, + prange->last); + svm_range_unlink(prange); + svm_range_remove_notifier(prange); + svm_range_free(prange); + } + + mmap_write_downgrade(mm); + /* Trigger migrations and revalidate and map to GPUs as needed. If + * this fails we may be left with partially completed actions. There + * is no clean way of rolling back to the previous state in such a + * case because the rollback wouldn't be guaranteed to work either. + */ + list_for_each_entry(prange, &update_list, update_list) { + bool migrated; + + mutex_lock(&prange->migrate_mutex); + + r = svm_range_trigger_migration(mm, prange, &migrated); + if (r) + goto out_unlock_range; + + if (migrated && !p->xnack_enabled) { + pr_debug("restore_work will update mappings of GPUs\n"); + mutex_unlock(&prange->migrate_mutex); + continue; + } + + r = svm_range_validate_and_map(mm, prange, MAX_GPU_INSTANCE, + true, true); + if (r) + pr_debug("failed %d to map svm range\n", r); + +out_unlock_range: + mutex_unlock(&prange->migrate_mutex); + if (r) + break; + } + + svm_range_debug_dump(svms); + + mutex_unlock(&svms->lock); + mmap_read_unlock(mm); +out: + mutex_unlock(&process_info->lock); + + pr_debug("pasid 0x%x svms 0x%p [0x%llx 0x%llx] done, r=%d\n", p->pasid, + &p->svms, start, start + size - 1, r); + + return r; +} + +static int +svm_range_get_attr(struct kfd_process *p, uint64_t start, uint64_t size, + uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs) +{ + DECLARE_BITMAP(bitmap_access, MAX_GPU_INSTANCE); + DECLARE_BITMAP(bitmap_aip, MAX_GPU_INSTANCE); + bool get_preferred_loc = false; + bool get_prefetch_loc = false; + bool get_granularity = false; + bool get_accessible = false; + bool get_flags = false; + uint64_t last = start + size - 1UL; + struct mm_struct *mm = current->mm; + uint8_t granularity = 0xff; + struct interval_tree_node *node; + struct svm_range_list *svms; + struct svm_range *prange; + uint32_t prefetch_loc = KFD_IOCTL_SVM_LOCATION_UNDEFINED; + uint32_t location = KFD_IOCTL_SVM_LOCATION_UNDEFINED; + uint32_t flags = 0xffffffff; + int gpuidx; + uint32_t i; + + pr_debug("svms 0x%p [0x%llx 0x%llx] nattr 0x%x\n", &p->svms, start, + start + size - 1, nattr); + + mmap_read_lock(mm); + if (!svm_range_is_valid(mm, start, size)) { + pr_debug("invalid range\n"); + mmap_read_unlock(mm); + return -EINVAL; + } + mmap_read_unlock(mm); + + for (i = 0; i < nattr; i++) { + switch (attrs[i].type) { + case KFD_IOCTL_SVM_ATTR_PREFERRED_LOC: + get_preferred_loc = true; + break; + case KFD_IOCTL_SVM_ATTR_PREFETCH_LOC: + get_prefetch_loc = true; + break; + case KFD_IOCTL_SVM_ATTR_ACCESS: + get_accessible = true; + break; + case KFD_IOCTL_SVM_ATTR_SET_FLAGS: + get_flags = true; + break; + case KFD_IOCTL_SVM_ATTR_GRANULARITY: + get_granularity = true; + break; + case KFD_IOCTL_SVM_ATTR_CLR_FLAGS: + case KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE: + case KFD_IOCTL_SVM_ATTR_NO_ACCESS: + fallthrough; + default: + pr_debug("get invalid attr type 0x%x\n", attrs[i].type); + return -EINVAL; + } + } + + svms = &p->svms; + + mutex_lock(&svms->lock); + + node = interval_tree_iter_first(&svms->objects, start, last); + if (!node) { + pr_debug("range attrs not found return default values\n"); + svm_range_set_default_attributes(&location, &prefetch_loc, + &granularity, &flags); + if (p->xnack_enabled) + bitmap_fill(bitmap_access, MAX_GPU_INSTANCE); + else + bitmap_zero(bitmap_access, MAX_GPU_INSTANCE); + bitmap_zero(bitmap_aip, MAX_GPU_INSTANCE); + goto fill_values; + } + bitmap_fill(bitmap_access, MAX_GPU_INSTANCE); + bitmap_fill(bitmap_aip, MAX_GPU_INSTANCE); + + while (node) { + struct interval_tree_node *next; + + prange = container_of(node, struct svm_range, it_node); + next = interval_tree_iter_next(node, start, last); + + if (get_preferred_loc) { + if (prange->preferred_loc == + KFD_IOCTL_SVM_LOCATION_UNDEFINED || + (location != KFD_IOCTL_SVM_LOCATION_UNDEFINED && + location != prange->preferred_loc)) { + location = KFD_IOCTL_SVM_LOCATION_UNDEFINED; + get_preferred_loc = false; + } else { + location = prange->preferred_loc; + } + } + if (get_prefetch_loc) { + if (prange->prefetch_loc == + KFD_IOCTL_SVM_LOCATION_UNDEFINED || + (prefetch_loc != KFD_IOCTL_SVM_LOCATION_UNDEFINED && + prefetch_loc != prange->prefetch_loc)) { + prefetch_loc = KFD_IOCTL_SVM_LOCATION_UNDEFINED; + get_prefetch_loc = false; + } else { + prefetch_loc = prange->prefetch_loc; + } + } + if (get_accessible) { + bitmap_and(bitmap_access, bitmap_access, + prange->bitmap_access, MAX_GPU_INSTANCE); + bitmap_and(bitmap_aip, bitmap_aip, + prange->bitmap_aip, MAX_GPU_INSTANCE); + } + if (get_flags) + flags &= prange->flags; + + if (get_granularity && prange->granularity < granularity) + granularity = prange->granularity; + + node = next; + } +fill_values: + mutex_unlock(&svms->lock); + + for (i = 0; i < nattr; i++) { + switch (attrs[i].type) { + case KFD_IOCTL_SVM_ATTR_PREFERRED_LOC: + attrs[i].value = location; + break; + case KFD_IOCTL_SVM_ATTR_PREFETCH_LOC: + attrs[i].value = prefetch_loc; + break; + case KFD_IOCTL_SVM_ATTR_ACCESS: + gpuidx = kfd_process_gpuidx_from_gpuid(p, + attrs[i].value); + if (gpuidx < 0) { + pr_debug("invalid gpuid %x\n", attrs[i].value); + return -EINVAL; + } + if (test_bit(gpuidx, bitmap_access)) + attrs[i].type = KFD_IOCTL_SVM_ATTR_ACCESS; + else if (test_bit(gpuidx, bitmap_aip)) + attrs[i].type = + KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE; + else + attrs[i].type = KFD_IOCTL_SVM_ATTR_NO_ACCESS; + break; + case KFD_IOCTL_SVM_ATTR_SET_FLAGS: + attrs[i].value = flags; + break; + case KFD_IOCTL_SVM_ATTR_GRANULARITY: + attrs[i].value = (uint32_t)granularity; + break; + } + } + + return 0; +} + +int +svm_ioctl(struct kfd_process *p, enum kfd_ioctl_svm_op op, uint64_t start, + uint64_t size, uint32_t nattrs, struct kfd_ioctl_svm_attribute *attrs) +{ + int r; + + start >>= PAGE_SHIFT; + size >>= PAGE_SHIFT; + + switch (op) { + case KFD_IOCTL_SVM_OP_SET_ATTR: + r = svm_range_set_attr(p, start, size, nattrs, attrs); + break; + case KFD_IOCTL_SVM_OP_GET_ATTR: + r = svm_range_get_attr(p, start, size, nattrs, attrs); + break; + default: + r = EINVAL; + break; + } + + return r; +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h new file mode 100644 index 000000000000..573f984b81fe --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h @@ -0,0 +1,206 @@ +/* SPDX-License-Identifier: GPL-2.0 OR MIT */ +/* + * Copyright 2020-2021 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef KFD_SVM_H_ +#define KFD_SVM_H_ + +#if IS_ENABLED(CONFIG_HSA_AMD_SVM) + +#include <linux/rwsem.h> +#include <linux/list.h> +#include <linux/mutex.h> +#include <linux/sched/mm.h> +#include <linux/hmm.h> +#include "amdgpu.h" +#include "kfd_priv.h" + +struct svm_range_bo { + struct amdgpu_bo *bo; + struct kref kref; + struct list_head range_list; /* all svm ranges shared this bo */ + spinlock_t list_lock; + struct amdgpu_amdkfd_fence *eviction_fence; + struct work_struct eviction_work; + struct svm_range_list *svms; + uint32_t evicting; +}; + +enum svm_work_list_ops { + SVM_OP_NULL, + SVM_OP_UNMAP_RANGE, + SVM_OP_UPDATE_RANGE_NOTIFIER, + SVM_OP_UPDATE_RANGE_NOTIFIER_AND_MAP, + SVM_OP_ADD_RANGE, + SVM_OP_ADD_RANGE_AND_MAP +}; + +struct svm_work_list_item { + enum svm_work_list_ops op; + struct mm_struct *mm; +}; + +/** + * struct svm_range - shared virtual memory range + * + * @svms: list of svm ranges, structure defined in kfd_process + * @migrate_mutex: to serialize range migration, validation and mapping update + * @start: range start address in pages + * @last: range last address in pages + * @it_node: node [start, last] stored in interval tree, start, last are page + * aligned, page size is (last - start + 1) + * @list: link list node, used to scan all ranges of svms + * @update_list:link list node used to add to update_list + * @remove_list:link list node used to add to remove list + * @insert_list:link list node used to add to insert list + * @mapping: bo_va mapping structure to create and update GPU page table + * @npages: number of pages + * @dma_addr: dma mapping address on each GPU for system memory physical page + * @ttm_res: vram ttm resource map + * @offset: range start offset within mm_nodes + * @svm_bo: struct to manage splited amdgpu_bo + * @svm_bo_list:link list node, to scan all ranges which share same svm_bo + * @lock: protect prange start, last, child_list, svm_bo_list + * @saved_flags:save/restore current PF_MEMALLOC flags + * @flags: flags defined as KFD_IOCTL_SVM_FLAG_* + * @perferred_loc: perferred location, 0 for CPU, or GPU id + * @perfetch_loc: last prefetch location, 0 for CPU, or GPU id + * @actual_loc: the actual location, 0 for CPU, or GPU id + * @granularity:migration granularity, log2 num pages + * @invalid: not 0 means cpu page table is invalidated + * @validate_timestamp: system timestamp when range is validated + * @notifier: register mmu interval notifier + * @work_item: deferred work item information + * @deferred_list: list header used to add range to deferred list + * @child_list: list header for split ranges which are not added to svms yet + * @bitmap_access: index bitmap of GPUs which can access the range + * @bitmap_aip: index bitmap of GPUs which can access the range in place + * + * Data structure for virtual memory range shared by CPU and GPUs, it can be + * allocated from system memory ram or device vram, and migrate from ram to vram + * or from vram to ram. + */ +struct svm_range { + struct svm_range_list *svms; + struct mutex migrate_mutex; + unsigned long start; + unsigned long last; + struct interval_tree_node it_node; + struct list_head list; + struct list_head update_list; + struct list_head remove_list; + struct list_head insert_list; + struct amdgpu_bo_va_mapping mapping; + uint64_t npages; + dma_addr_t *dma_addr[MAX_GPU_INSTANCE]; + struct ttm_resource *ttm_res; + uint64_t offset; + struct svm_range_bo *svm_bo; + struct list_head svm_bo_list; + struct mutex lock; + unsigned int saved_flags; + uint32_t flags; + uint32_t preferred_loc; + uint32_t prefetch_loc; + uint32_t actual_loc; + uint8_t granularity; + atomic_t invalid; + uint64_t validate_timestamp; + struct mmu_interval_notifier notifier; + struct svm_work_list_item work_item; + struct list_head deferred_list; + struct list_head child_list; + DECLARE_BITMAP(bitmap_access, MAX_GPU_INSTANCE); + DECLARE_BITMAP(bitmap_aip, MAX_GPU_INSTANCE); + bool validated_once; +}; + +static inline void svm_range_lock(struct svm_range *prange) +{ + mutex_lock(&prange->lock); + prange->saved_flags = memalloc_noreclaim_save(); + +} +static inline void svm_range_unlock(struct svm_range *prange) +{ + memalloc_noreclaim_restore(prange->saved_flags); + mutex_unlock(&prange->lock); +} + +int svm_range_list_init(struct kfd_process *p); +void svm_range_list_fini(struct kfd_process *p); +int svm_ioctl(struct kfd_process *p, enum kfd_ioctl_svm_op op, uint64_t start, + uint64_t size, uint32_t nattrs, + struct kfd_ioctl_svm_attribute *attrs); +struct svm_range *svm_range_from_addr(struct svm_range_list *svms, + unsigned long addr, + struct svm_range **parent); +struct amdgpu_device *svm_range_get_adev_by_id(struct svm_range *prange, + uint32_t id); +int svm_range_vram_node_new(struct amdgpu_device *adev, + struct svm_range *prange, bool clear); +void svm_range_vram_node_free(struct svm_range *prange); +int svm_range_split_by_granularity(struct kfd_process *p, struct mm_struct *mm, + unsigned long addr, struct svm_range *parent, + struct svm_range *prange); +int svm_range_restore_pages(struct amdgpu_device *adev, + unsigned int pasid, uint64_t addr); +int svm_range_schedule_evict_svm_bo(struct amdgpu_amdkfd_fence *fence); +void svm_range_add_list_work(struct svm_range_list *svms, + struct svm_range *prange, struct mm_struct *mm, + enum svm_work_list_ops op); +void schedule_deferred_list_work(struct svm_range_list *svms); +void svm_range_dma_unmap(struct device *dev, dma_addr_t *dma_addr, + unsigned long offset, unsigned long npages); +void svm_range_free_dma_mappings(struct svm_range *prange); +void svm_range_prefault(struct svm_range *prange, struct mm_struct *mm); + +#else + +struct kfd_process; + +static inline int svm_range_list_init(struct kfd_process *p) +{ + return 0; +} +static inline void svm_range_list_fini(struct kfd_process *p) +{ + /* empty */ +} + +static inline int svm_range_restore_pages(struct amdgpu_device *adev, + unsigned int pasid, uint64_t addr) +{ + return -EFAULT; +} + +static inline int svm_range_schedule_evict_svm_bo( + struct amdgpu_amdkfd_fence *fence) +{ + WARN_ONCE(1, "SVM eviction fence triggered, but SVM is disabled"); + return -EINVAL; +} + +#endif /* IS_ENABLED(CONFIG_HSA_AMD_SVM) */ + +#endif /* KFD_SVM_H_ */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c index cdef608db4f4..7fae6a7e51f5 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c @@ -1192,40 +1192,60 @@ static void kfd_fill_mem_clk_max_info(struct kfd_topology_device *dev) mem->mem_clk_max = local_mem_info.mem_clk_max; } -static void kfd_fill_iolink_non_crat_info(struct kfd_topology_device *dev) +static void kfd_set_iolink_no_atomics(struct kfd_topology_device *dev, + struct kfd_topology_device *target_gpu_dev, + struct kfd_iolink_properties *link) { - struct kfd_iolink_properties *link, *cpu_link; - struct kfd_topology_device *cpu_dev; - uint32_t cap; - uint32_t cpu_flag = CRAT_IOLINK_FLAGS_ENABLED; - uint32_t flag = CRAT_IOLINK_FLAGS_ENABLED; - - if (!dev || !dev->gpu) + /* xgmi always supports atomics between links. */ + if (link->iolink_type == CRAT_IOLINK_TYPE_XGMI) return; - pcie_capability_read_dword(dev->gpu->pdev, - PCI_EXP_DEVCAP2, &cap); + /* check pcie support to set cpu(dev) flags for target_gpu_dev link. */ + if (target_gpu_dev) { + uint32_t cap; - if (!(cap & (PCI_EXP_DEVCAP2_ATOMIC_COMP32 | - PCI_EXP_DEVCAP2_ATOMIC_COMP64))) - cpu_flag |= CRAT_IOLINK_FLAGS_NO_ATOMICS_32_BIT | - CRAT_IOLINK_FLAGS_NO_ATOMICS_64_BIT; + pcie_capability_read_dword(target_gpu_dev->gpu->pdev, + PCI_EXP_DEVCAP2, &cap); + + if (!(cap & (PCI_EXP_DEVCAP2_ATOMIC_COMP32 | + PCI_EXP_DEVCAP2_ATOMIC_COMP64))) + link->flags |= CRAT_IOLINK_FLAGS_NO_ATOMICS_32_BIT | + CRAT_IOLINK_FLAGS_NO_ATOMICS_64_BIT; + /* set gpu (dev) flags. */ + } else { + if (!dev->gpu->pci_atomic_requested || + dev->gpu->device_info->asic_family == + CHIP_HAWAII) + link->flags |= CRAT_IOLINK_FLAGS_NO_ATOMICS_32_BIT | + CRAT_IOLINK_FLAGS_NO_ATOMICS_64_BIT; + } +} + +static void kfd_fill_iolink_non_crat_info(struct kfd_topology_device *dev) +{ + struct kfd_iolink_properties *link, *inbound_link; + struct kfd_topology_device *peer_dev; - if (!dev->gpu->pci_atomic_requested || - dev->gpu->device_info->asic_family == CHIP_HAWAII) - flag |= CRAT_IOLINK_FLAGS_NO_ATOMICS_32_BIT | - CRAT_IOLINK_FLAGS_NO_ATOMICS_64_BIT; + if (!dev || !dev->gpu) + return; /* GPU only creates direct links so apply flags setting to all */ list_for_each_entry(link, &dev->io_link_props, list) { - link->flags = flag; - cpu_dev = kfd_topology_device_by_proximity_domain( + link->flags = CRAT_IOLINK_FLAGS_ENABLED; + kfd_set_iolink_no_atomics(dev, NULL, link); + peer_dev = kfd_topology_device_by_proximity_domain( link->node_to); - if (cpu_dev) { - list_for_each_entry(cpu_link, - &cpu_dev->io_link_props, list) - if (cpu_link->node_to == link->node_from) - cpu_link->flags = cpu_flag; + + if (!peer_dev) + continue; + + list_for_each_entry(inbound_link, &peer_dev->io_link_props, + list) { + if (inbound_link->node_to != link->node_from) + continue; + + inbound_link->flags = CRAT_IOLINK_FLAGS_ENABLED; + kfd_set_iolink_no_atomics(peer_dev, dev, inbound_link); } } } @@ -1410,15 +1430,21 @@ int kfd_topology_add_device(struct kfd_dev *gpu) adev = (struct amdgpu_device *)(dev->gpu->kgd); /* kfd only concerns sram ecc on GFX and HBM ecc on UMC */ dev->node_props.capability |= - ((adev->ras_features & BIT(AMDGPU_RAS_BLOCK__GFX)) != 0) ? + ((adev->ras_enabled & BIT(AMDGPU_RAS_BLOCK__GFX)) != 0) ? HSA_CAP_SRAM_EDCSUPPORTED : 0; - dev->node_props.capability |= ((adev->ras_features & BIT(AMDGPU_RAS_BLOCK__UMC)) != 0) ? + dev->node_props.capability |= ((adev->ras_enabled & BIT(AMDGPU_RAS_BLOCK__UMC)) != 0) ? HSA_CAP_MEM_EDCSUPPORTED : 0; if (adev->asic_type != CHIP_VEGA10) - dev->node_props.capability |= (adev->ras_features != 0) ? + dev->node_props.capability |= (adev->ras_enabled != 0) ? HSA_CAP_RASEVENTNOTIFY : 0; + /* SVM API and HMM page migration work together, device memory type + * is initialized to not 0 when page migration register device memory. + */ + if (adev->kfd.dev->pgmap.type != 0) + dev->node_props.capability |= HSA_CAP_SVMAPI_SUPPORTED; + kfd_debug_print_topology(); if (!res) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h index b8b68087bd7a..6bd6380b0ee0 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h @@ -53,8 +53,9 @@ #define HSA_CAP_ASIC_REVISION_MASK 0x03c00000 #define HSA_CAP_ASIC_REVISION_SHIFT 22 #define HSA_CAP_SRAM_EDCSUPPORTED 0x04000000 +#define HSA_CAP_SVMAPI_SUPPORTED 0x08000000 -#define HSA_CAP_RESERVED 0xf80f8000 +#define HSA_CAP_RESERVED 0xf00f8000 struct kfd_node_properties { uint64_t hive_id; @@ -98,9 +99,10 @@ struct kfd_node_properties { #define HSA_MEM_HEAP_TYPE_GPU_LDS 4 #define HSA_MEM_HEAP_TYPE_GPU_SCRATCH 5 -#define HSA_MEM_FLAGS_HOT_PLUGGABLE 0x00000001 -#define HSA_MEM_FLAGS_NON_VOLATILE 0x00000002 -#define HSA_MEM_FLAGS_RESERVED 0xfffffffc +#define HSA_MEM_FLAGS_HOT_PLUGGABLE 0x00000001 +#define HSA_MEM_FLAGS_NON_VOLATILE 0x00000002 +#define HSA_MEM_FLAGS_COHERENTHOSTACCESS 0x00000004 +#define HSA_MEM_FLAGS_RESERVED 0xfffffff8 struct kfd_mem_properties { struct list_head list; 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 9ca517b65854..4823f8a32159 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -35,6 +35,7 @@ #include "dc/inc/hw/abm.h" #include "dc/dc_dmub_srv.h" #include "dc/dc_edid_parser.h" +#include "dc/dc_stat.h" #include "amdgpu_dm_trace.h" #include "vid.h" @@ -59,6 +60,7 @@ #include "ivsrcid/ivsrcid_vislands30.h" +#include "i2caux_interface.h" #include <linux/module.h> #include <linux/moduleparam.h> #include <linux/types.h> @@ -618,6 +620,58 @@ static void dm_dcn_vertical_interrupt0_high_irq(void *interrupt_params) amdgpu_dm_crtc_handle_crc_window_irq(&acrtc->base); } #endif + +/** + * dm_dmub_outbox1_low_irq() - Handles Outbox interrupt + * @interrupt_params: used for determining the Outbox instance + * + * Handles the Outbox Interrupt + * event handler. + */ +#define DMUB_TRACE_MAX_READ 64 +static void dm_dmub_outbox1_low_irq(void *interrupt_params) +{ + struct dmub_notification notify; + struct common_irq_params *irq_params = interrupt_params; + struct amdgpu_device *adev = irq_params->adev; + struct amdgpu_display_manager *dm = &adev->dm; + struct dmcub_trace_buf_entry entry = { 0 }; + uint32_t count = 0; + + if (dc_enable_dmub_notifications(adev->dm.dc)) { + if (irq_params->irq_src == DC_IRQ_SOURCE_DMCUB_OUTBOX) { + do { + dc_stat_get_dmub_notification(adev->dm.dc, ¬ify); + } while (notify.pending_notification); + + if (adev->dm.dmub_notify) + memcpy(adev->dm.dmub_notify, ¬ify, sizeof(struct dmub_notification)); + if (notify.type == DMUB_NOTIFICATION_AUX_REPLY) + complete(&adev->dm.dmub_aux_transfer_done); + // TODO : HPD Implementation + + } else { + DRM_ERROR("DM: Failed to receive correct outbox IRQ !"); + } + } + + + do { + if (dc_dmub_srv_get_dmub_outbox0_msg(dm->dc, &entry)) { + trace_amdgpu_dmub_trace_high_irq(entry.trace_code, entry.tick_count, + entry.param0, entry.param1); + + DRM_DEBUG_DRIVER("trace_code:%u, tick_count:%u, param0:%u, param1:%u\n", + entry.trace_code, entry.tick_count, entry.param0, entry.param1); + } else + break; + + count++; + + } while (count <= DMUB_TRACE_MAX_READ); + + ASSERT(count <= DMUB_TRACE_MAX_READ); +} #endif static int dm_set_clockgating_state(void *handle, @@ -938,32 +992,6 @@ static int dm_dmub_hw_init(struct amdgpu_device *adev) } #if defined(CONFIG_DRM_AMD_DC_DCN) -#define DMUB_TRACE_MAX_READ 64 -static void dm_dmub_trace_high_irq(void *interrupt_params) -{ - struct common_irq_params *irq_params = interrupt_params; - struct amdgpu_device *adev = irq_params->adev; - struct amdgpu_display_manager *dm = &adev->dm; - struct dmcub_trace_buf_entry entry = { 0 }; - uint32_t count = 0; - - do { - if (dc_dmub_srv_get_dmub_outbox0_msg(dm->dc, &entry)) { - trace_amdgpu_dmub_trace_high_irq(entry.trace_code, entry.tick_count, - entry.param0, entry.param1); - - DRM_DEBUG_DRIVER("trace_code:%u, tick_count:%u, param0:%u, param1:%u\n", - entry.trace_code, entry.tick_count, entry.param0, entry.param1); - } else - break; - - count++; - - } while (count <= DMUB_TRACE_MAX_READ); - - ASSERT(count <= DMUB_TRACE_MAX_READ); -} - static void mmhub_read_system_context(struct amdgpu_device *adev, struct dc_phy_addr_space_config *pa_config) { uint64_t pt_base; @@ -1220,6 +1248,16 @@ static int amdgpu_dm_init(struct amdgpu_device *adev) #if defined(CONFIG_DRM_AMD_SECURE_DISPLAY) adev->dm.crc_rd_wrk = amdgpu_dm_crtc_secure_display_create_work(); #endif + if (dc_enable_dmub_notifications(adev->dm.dc)) { + init_completion(&adev->dm.dmub_aux_transfer_done); + adev->dm.dmub_notify = kzalloc(sizeof(struct dmub_notification), GFP_KERNEL); + if (!adev->dm.dmub_notify) { + DRM_INFO("amdgpu: fail to allocate adev->dm.dmub_notify"); + goto error; + } + amdgpu_dm_outbox_init(adev); + } + if (amdgpu_dm_initialize_drm_device(adev)) { DRM_ERROR( "amdgpu: failed to initialize sw for display support.\n"); @@ -1293,6 +1331,11 @@ static void amdgpu_dm_fini(struct amdgpu_device *adev) adev->dm.dc->ctx->dmub_srv = NULL; } + if (dc_enable_dmub_notifications(adev->dm.dc)) { + kfree(adev->dm.dmub_notify); + adev->dm.dmub_notify = NULL; + } + if (adev->dm.dmub_bo) amdgpu_bo_free_kernel(&adev->dm.dmub_bo, &adev->dm.dmub_bo_gpu_addr, @@ -2708,8 +2751,7 @@ static void handle_hpd_rx_irq(void *param) * conflict, after implement i2c helper, this mutex should be * retired. */ - if (dc_link->type != dc_connection_mst_branch) - mutex_lock(&aconnector->hpd_lock); + mutex_lock(&aconnector->hpd_lock); read_hpd_rx_irq_data(dc_link, &hpd_irq_data); @@ -2726,13 +2768,15 @@ static void handle_hpd_rx_irq(void *param) } } - mutex_lock(&adev->dm.dc_lock); + if (!amdgpu_in_reset(adev)) { + mutex_lock(&adev->dm.dc_lock); #ifdef CONFIG_DRM_AMD_DC_HDCP result = dc_link_handle_hpd_rx_irq(dc_link, &hpd_irq_data, NULL); #else result = dc_link_handle_hpd_rx_irq(dc_link, NULL, NULL); #endif - mutex_unlock(&adev->dm.dc_lock); + mutex_unlock(&adev->dm.dc_lock); + } out: if (result && !is_mst_root_connector) { @@ -2776,10 +2820,10 @@ out: } #endif - if (dc_link->type != dc_connection_mst_branch) { + if (dc_link->type != dc_connection_mst_branch) drm_dp_cec_irq(&aconnector->dm_dp_aux.aux); - mutex_unlock(&aconnector->hpd_lock); - } + + mutex_unlock(&aconnector->hpd_lock); } static void register_hpd_handlers(struct amdgpu_device *adev) @@ -3151,38 +3195,51 @@ static int dcn10_register_irq_handlers(struct amdgpu_device *adev) } - if (dc->ctx->dmub_srv) { - i = DCN_1_0__SRCID__DMCUB_OUTBOX_HIGH_PRIORITY_READY_INT; - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DCE, i, &adev->dmub_trace_irq); + /* HPD */ + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DCE, DCN_1_0__SRCID__DC_HPD1_INT, + &adev->hpd_irq); + if (r) { + DRM_ERROR("Failed to add hpd irq id!\n"); + return r; + } - if (r) { - DRM_ERROR("Failed to add dmub trace irq id!\n"); - return r; - } + register_hpd_handlers(adev); - int_params.int_context = INTERRUPT_HIGH_IRQ_CONTEXT; + return 0; +} +/* Register Outbox IRQ sources and initialize IRQ callbacks */ +static int register_outbox_irq_handlers(struct amdgpu_device *adev) +{ + struct dc *dc = adev->dm.dc; + struct common_irq_params *c_irq_params; + struct dc_interrupt_params int_params = {0}; + int r, i; + + int_params.requested_polarity = INTERRUPT_POLARITY_DEFAULT; + int_params.current_polarity = INTERRUPT_POLARITY_DEFAULT; + + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DCE, DCN_1_0__SRCID__DMCUB_OUTBOX_LOW_PRIORITY_READY_INT, + &adev->dmub_outbox_irq); + if (r) { + DRM_ERROR("Failed to add outbox irq id!\n"); + return r; + } + + if (dc->ctx->dmub_srv) { + i = DCN_1_0__SRCID__DMCUB_OUTBOX_LOW_PRIORITY_READY_INT; + int_params.int_context = INTERRUPT_LOW_IRQ_CONTEXT; int_params.irq_source = - dc_interrupt_to_irq_source(dc, i, 0); + dc_interrupt_to_irq_source(dc, i, 0); - c_irq_params = &adev->dm.dmub_trace_params[0]; + c_irq_params = &adev->dm.dmub_outbox_params[0]; c_irq_params->adev = adev; c_irq_params->irq_src = int_params.irq_source; amdgpu_dm_irq_register_interrupt(adev, &int_params, - dm_dmub_trace_high_irq, c_irq_params); - } - - /* HPD */ - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DCE, DCN_1_0__SRCID__DC_HPD1_INT, - &adev->hpd_irq); - if (r) { - DRM_ERROR("Failed to add hpd irq id!\n"); - return r; + dm_dmub_outbox1_low_irq, c_irq_params); } - register_hpd_handlers(adev); - return 0; } #endif @@ -3414,22 +3471,37 @@ static int amdgpu_dm_backlight_update_status(struct backlight_device *bd) { struct amdgpu_display_manager *dm = bl_get_data(bd); struct amdgpu_dm_backlight_caps caps; - struct dc_link *link = NULL; + struct dc_link *link[AMDGPU_DM_MAX_NUM_EDP]; u32 brightness; bool rc; + int i; amdgpu_dm_update_backlight_caps(dm); caps = dm->backlight_caps; - link = (struct dc_link *)dm->backlight_link; + for (i = 0; i < dm->num_of_edps; i++) + link[i] = (struct dc_link *)dm->backlight_link[i]; brightness = convert_brightness_from_user(&caps, bd->props.brightness); // Change brightness based on AUX property - if (caps.aux_support) - rc = dc_link_set_backlight_level_nits(link, true, brightness, - AUX_BL_DEFAULT_TRANSITION_TIME_MS); - else - rc = dc_link_set_backlight_level(dm->backlight_link, brightness, 0); + if (caps.aux_support) { + for (i = 0; i < dm->num_of_edps; i++) { + rc = dc_link_set_backlight_level_nits(link[i], true, brightness, + AUX_BL_DEFAULT_TRANSITION_TIME_MS); + if (!rc) { + DRM_ERROR("DM: Failed to update backlight via AUX on eDP[%d]\n", i); + break; + } + } + } else { + for (i = 0; i < dm->num_of_edps; i++) { + rc = dc_link_set_backlight_level(dm->backlight_link[i], brightness, 0); + if (!rc) { + DRM_ERROR("DM: Failed to update backlight on eDP[%d]\n", i); + break; + } + } + } return rc ? 0 : 1; } @@ -3443,7 +3515,7 @@ static int amdgpu_dm_backlight_get_brightness(struct backlight_device *bd) caps = dm->backlight_caps; if (caps.aux_support) { - struct dc_link *link = (struct dc_link *)dm->backlight_link; + struct dc_link *link = (struct dc_link *)dm->backlight_link[0]; u32 avg, peak; bool rc; @@ -3452,7 +3524,7 @@ static int amdgpu_dm_backlight_get_brightness(struct backlight_device *bd) return bd->props.brightness; return convert_brightness_to_user(&caps, avg); } else { - int ret = dc_link_get_backlight_level(dm->backlight_link); + int ret = dc_link_get_backlight_level(dm->backlight_link[0]); if (ret == DC_ERROR_UNEXPECTED) return bd->props.brightness; @@ -3549,10 +3621,13 @@ static void register_backlight_device(struct amdgpu_display_manager *dm, * DM initialization because not having a backlight control * is better then a black screen. */ - amdgpu_dm_register_backlight_device(dm); + if (!dm->backlight_dev) + amdgpu_dm_register_backlight_device(dm); - if (dm->backlight_dev) - dm->backlight_link = link; + if (dm->backlight_dev) { + dm->backlight_link[dm->num_of_edps] = link; + dm->num_of_edps++; + } } #endif } @@ -3643,6 +3718,22 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev) goto fail; } +#if defined(CONFIG_DRM_AMD_DC_DCN) + /* Use Outbox interrupt */ + switch (adev->asic_type) { + case CHIP_SIENNA_CICHLID: + case CHIP_NAVY_FLOUNDER: + case CHIP_RENOIR: + if (register_outbox_irq_handlers(dm->adev)) { + DRM_ERROR("DM: Failed to initialize IRQ\n"); + goto fail; + } + break; + default: + DRM_DEBUG_KMS("Unsupported ASIC type for outbox: 0x%X\n", adev->asic_type); + } +#endif + /* loops over all connectors on the board */ for (i = 0; i < link_cnt; i++) { struct dc_link *link = NULL; @@ -6560,13 +6651,13 @@ static int dm_update_mst_vcpi_slots_for_dsc(struct drm_atomic_state *state, { struct dc_stream_state *stream = NULL; struct drm_connector *connector; - struct drm_connector_state *new_con_state, *old_con_state; + struct drm_connector_state *new_con_state; struct amdgpu_dm_connector *aconnector; struct dm_connector_state *dm_conn_state; int i, j, clock, bpp; int vcpi, pbn_div, pbn = 0; - for_each_oldnew_connector_in_state(state, connector, old_con_state, new_con_state, i) { + for_each_new_connector_in_state(state, connector, new_con_state, i) { aconnector = to_amdgpu_dm_connector(connector); @@ -8164,15 +8255,14 @@ static void amdgpu_dm_handle_vrr_transition(struct dm_crtc_state *old_state, static void amdgpu_dm_commit_cursors(struct drm_atomic_state *state) { struct drm_plane *plane; - struct drm_plane_state *old_plane_state, *new_plane_state; + struct drm_plane_state *old_plane_state; int i; /* * TODO: Make this per-stream so we don't issue redundant updates for * commits with multiple streams. */ - for_each_oldnew_plane_in_state(state, plane, old_plane_state, - new_plane_state, i) + for_each_old_plane_in_state(state, plane, old_plane_state, i) if (plane->type == DRM_PLANE_TYPE_CURSOR) handle_cursor_update(plane, old_plane_state); } @@ -10668,3 +10758,30 @@ uint32_t dm_read_reg_func(const struct dc_context *ctx, uint32_t address, return value; } + +int amdgpu_dm_process_dmub_aux_transfer_sync(struct dc_context *ctx, unsigned int linkIndex, + struct aux_payload *payload, enum aux_return_code_type *operation_result) +{ + struct amdgpu_device *adev = ctx->driver_context; + int ret = 0; + + dc_process_dmub_aux_transfer_async(ctx->dc, linkIndex, payload); + ret = wait_for_completion_interruptible_timeout(&adev->dm.dmub_aux_transfer_done, 10*HZ); + if (ret == 0) { + *operation_result = AUX_RET_ERROR_TIMEOUT; + return -1; + } + *operation_result = (enum aux_return_code_type)adev->dm.dmub_notify->result; + + if (adev->dm.dmub_notify->result == AUX_RET_SUCCESS) { + (*payload->reply) = adev->dm.dmub_notify->aux_reply.command; + + // For read case, Copy data to payload + if (!payload->write && adev->dm.dmub_notify->aux_reply.length && + (*payload->reply == AUX_TRANSACTION_REPLY_AUX_ACK)) + memcpy(payload->data, adev->dm.dmub_notify->aux_reply.data, + adev->dm.dmub_notify->aux_reply.length); + } + + return adev->dm.dmub_notify->aux_reply.length; +} 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 b2f2ccfc20bb..c6f79c7dfac4 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h @@ -46,6 +46,7 @@ #define AMDGPU_DM_MAX_CRTC 6 +#define AMDGPU_DM_MAX_NUM_EDP 2 /* #include "include/amdgpu_dal_power_if.h" #include "amdgpu_dm_irq.h" @@ -54,6 +55,8 @@ #include "irq_types.h" #include "signal_types.h" #include "amdgpu_dm_crc.h" +struct aux_payload; +enum aux_return_code_type; /* Forward declarations */ struct amdgpu_device; @@ -62,6 +65,7 @@ struct dc; struct amdgpu_bo; struct dmub_srv; struct dc_plane_state; +struct dmub_notification; struct common_irq_params { struct amdgpu_device *adev; @@ -135,6 +139,10 @@ struct amdgpu_dm_backlight_caps { /** * struct dal_allocation - Tracks mapped FB memory for SMU communication + * @list: list of dal allocations + * @bo: GPU buffer object + * @cpu_ptr: CPU virtual address of the GPU buffer object + * @gpu_addr: GPU virtual address of the GPU buffer object */ struct dal_allocation { struct list_head list; @@ -164,6 +172,7 @@ struct dal_allocation { * @compressor: Frame buffer compression buffer. See &struct dm_compressor_info * @force_timing_sync: set via debugfs. When set, indicates that all connected * displays will be forced to synchronize. + * @dmcub_trace_event_en: enable dmcub trace events */ struct amdgpu_display_manager { @@ -178,6 +187,8 @@ struct amdgpu_display_manager { */ struct dmub_srv *dmub_srv; + struct dmub_notification *dmub_notify; + /** * @dmub_fb_info: * @@ -349,11 +360,17 @@ struct amdgpu_display_manager { struct common_irq_params dmub_trace_params[1]; + struct common_irq_params + dmub_outbox_params[1]; + spinlock_t irq_handler_list_table_lock; struct backlight_device *backlight_dev; - const struct dc_link *backlight_link; + const struct dc_link *backlight_link[AMDGPU_DM_MAX_NUM_EDP]; + + uint8_t num_of_edps; + struct amdgpu_dm_backlight_caps backlight_caps; struct mod_freesync *freesync_module; @@ -418,6 +435,7 @@ struct amdgpu_display_manager { * DAL fb memory allocation list, for communication with SMU. */ struct list_head da_list; + struct completion dmub_aux_transfer_done; }; enum dsc_clock_force_state { @@ -600,4 +618,6 @@ void amdgpu_dm_update_connector_after_detect( extern const struct drm_encoder_helper_funcs amdgpu_dm_encoder_helper_funcs; +int amdgpu_dm_process_dmub_aux_transfer_sync(struct dc_context *ctx, unsigned int linkIndex, + struct aux_payload *payload, enum aux_return_code_type *operation_result); #endif /* __AMDGPU_DM_H__ */ diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c index 1b6b15708b96..9fbbd0159119 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c @@ -925,6 +925,22 @@ static int hdcp_sink_capability_show(struct seq_file *m, void *data) return 0; } #endif + +/* + * Returns whether the connected display is internal and not hotpluggable. + * Example usage: cat /sys/kernel/debug/dri/0/DP-1/internal_display + */ +static int internal_display_show(struct seq_file *m, void *data) +{ + struct drm_connector *connector = m->private; + struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector); + struct dc_link *link = aconnector->dc_link; + + seq_printf(m, "Internal: %u\n", link->is_internal_display); + + return 0; +} + /* function description * * generic SDP message access for testing @@ -2361,6 +2377,44 @@ unlock: return size; } +/* + * Backlight at this moment. Read only. + * As written to display, taking ABM and backlight lut into account. + * Ranges from 0x0 to 0x10000 (= 100% PWM) + * + * Example usage: cat /sys/kernel/debug/dri/0/eDP-1/current_backlight + */ +static int current_backlight_show(struct seq_file *m, void *unused) +{ + struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(m->private); + struct dc_link *link = aconnector->dc_link; + unsigned int backlight; + + backlight = dc_link_get_backlight_level(link); + seq_printf(m, "0x%x\n", backlight); + + return 0; +} + +/* + * Backlight value that is being approached. Read only. + * As written to display, taking ABM and backlight lut into account. + * Ranges from 0x0 to 0x10000 (= 100% PWM) + * + * Example usage: cat /sys/kernel/debug/dri/0/eDP-1/target_backlight + */ +static int target_backlight_show(struct seq_file *m, void *unused) +{ + struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(m->private); + struct dc_link *link = aconnector->dc_link; + unsigned int backlight; + + backlight = dc_link_get_target_backlight_pwm(link); + seq_printf(m, "0x%x\n", backlight); + + return 0; +} + DEFINE_SHOW_ATTRIBUTE(dp_dsc_fec_support); DEFINE_SHOW_ATTRIBUTE(dmub_fw_state); DEFINE_SHOW_ATTRIBUTE(dmub_tracebuffer); @@ -2369,6 +2423,7 @@ DEFINE_SHOW_ATTRIBUTE(dp_lttpr_status); #ifdef CONFIG_DRM_AMD_DC_HDCP DEFINE_SHOW_ATTRIBUTE(hdcp_sink_capability); #endif +DEFINE_SHOW_ATTRIBUTE(internal_display); static const struct file_operations dp_dsc_clock_en_debugfs_fops = { .owner = THIS_MODULE, @@ -2594,13 +2649,17 @@ DEFINE_DEBUGFS_ATTRIBUTE(dmcub_trace_event_state_fops, dmcub_trace_event_state_g DEFINE_DEBUGFS_ATTRIBUTE(psr_fops, psr_get, NULL, "%llu\n"); +DEFINE_SHOW_ATTRIBUTE(current_backlight); +DEFINE_SHOW_ATTRIBUTE(target_backlight); + static const struct { char *name; const struct file_operations *fops; } connector_debugfs_entries[] = { {"force_yuv420_output", &force_yuv420_output_fops}, {"output_bpc", &output_bpc_fops}, - {"trigger_hotplug", &trigger_hotplug_debugfs_fops} + {"trigger_hotplug", &trigger_hotplug_debugfs_fops}, + {"internal_display", &internal_display_fops} }; void connector_debugfs_init(struct amdgpu_dm_connector *connector) @@ -2616,8 +2675,13 @@ void connector_debugfs_init(struct amdgpu_dm_connector *connector) dp_debugfs_entries[i].fops); } } - if (connector->base.connector_type == DRM_MODE_CONNECTOR_eDP) + if (connector->base.connector_type == DRM_MODE_CONNECTOR_eDP) { debugfs_create_file_unsafe("psr_state", 0444, dir, connector, &psr_fops); + debugfs_create_file("amdgpu_current_backlight_pwm", 0444, dir, connector, + ¤t_backlight_fops); + debugfs_create_file("amdgpu_target_backlight_pwm", 0444, dir, connector, + &target_backlight_fops); + } for (i = 0; i < ARRAY_SIZE(connector_debugfs_entries); i++) { debugfs_create_file(connector_debugfs_entries[i].name, @@ -2920,38 +2984,6 @@ static ssize_t dtn_log_write( return size; } -/* - * Backlight at this moment. Read only. - * As written to display, taking ABM and backlight lut into account. - * Ranges from 0x0 to 0x10000 (= 100% PWM) - */ -static int current_backlight_show(struct seq_file *m, void *unused) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)m->private; - struct amdgpu_display_manager *dm = &adev->dm; - - unsigned int backlight = dc_link_get_backlight_level(dm->backlight_link); - - seq_printf(m, "0x%x\n", backlight); - return 0; -} - -/* - * Backlight value that is being approached. Read only. - * As written to display, taking ABM and backlight lut into account. - * Ranges from 0x0 to 0x10000 (= 100% PWM) - */ -static int target_backlight_show(struct seq_file *m, void *unused) -{ - struct amdgpu_device *adev = (struct amdgpu_device *)m->private; - struct amdgpu_display_manager *dm = &adev->dm; - - unsigned int backlight = dc_link_get_target_backlight_pwm(dm->backlight_link); - - seq_printf(m, "0x%x\n", backlight); - return 0; -} - static int mst_topo_show(struct seq_file *m, void *unused) { struct amdgpu_device *adev = (struct amdgpu_device *)m->private; @@ -3134,8 +3166,6 @@ static int visual_confirm_get(void *data, u64 *val) return 0; } -DEFINE_SHOW_ATTRIBUTE(current_backlight); -DEFINE_SHOW_ATTRIBUTE(target_backlight); DEFINE_SHOW_ATTRIBUTE(mst_topo); DEFINE_DEBUGFS_ATTRIBUTE(visual_confirm_fops, visual_confirm_get, visual_confirm_set, "%llu\n"); @@ -3215,10 +3245,6 @@ void dtn_debugfs_init(struct amdgpu_device *adev) struct drm_minor *minor = adev_to_drm(adev)->primary; struct dentry *root = minor->debugfs_root; - debugfs_create_file("amdgpu_current_backlight_pwm", 0444, - root, adev, ¤t_backlight_fops); - debugfs_create_file("amdgpu_target_backlight_pwm", 0444, - root, adev, &target_backlight_fops); debugfs_create_file("amdgpu_mst_topology", 0444, root, adev, &mst_topo_fops); debugfs_create_file("amdgpu_dm_dtn_log", 0644, root, adev, diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c index e8b325a828c1..4646b0d02939 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c @@ -544,8 +544,10 @@ bool dm_helpers_dp_write_dsc_enable( ret = drm_dp_dpcd_write(aconnector->dsc_aux, DP_DSC_ENABLE, &enable_dsc, 1); } - if (stream->signal == SIGNAL_TYPE_DISPLAY_PORT) - return dm_helpers_dp_write_dpcd(ctx, stream->link, DP_DSC_ENABLE, &enable_dsc, 1); + if (stream->signal == SIGNAL_TYPE_DISPLAY_PORT) { + ret = dm_helpers_dp_write_dpcd(ctx, stream->link, DP_DSC_ENABLE, &enable_dsc, 1); + DC_LOG_DC("Send DSC %s to sst display\n", enable_dsc ? "enable" : "disable"); + } return (ret > 0); } @@ -640,7 +642,14 @@ enum dc_edid_status dm_helpers_read_local_edid( return edid_status; } - +int dm_helper_dmub_aux_transfer_sync( + struct dc_context *ctx, + const struct dc_link *link, + struct aux_payload *payload, + enum aux_return_code_type *operation_result) +{ + return amdgpu_dm_process_dmub_aux_transfer_sync(ctx, link->link_index, payload, operation_result); +} void dm_set_dcn_clocks(struct dc_context *ctx, struct dc_clocks *clks) { /* TODO: something */ @@ -698,12 +707,12 @@ void dm_helpers_free_gpu_mem( } } -bool dm_helpers_dmub_outbox0_interrupt_control(struct dc_context *ctx, bool enable) +bool dm_helpers_dmub_outbox_interrupt_control(struct dc_context *ctx, bool enable) { enum dc_irq_source irq_source; bool ret; - irq_source = DC_IRQ_SOURCE_DMCUB_OUTBOX0; + irq_source = DC_IRQ_SOURCE_DMCUB_OUTBOX; ret = dc_interrupt_set(ctx->dc, irq_source, enable); diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c index b3ed7e777720..40f617bbb86f 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c @@ -769,6 +769,18 @@ static int amdgpu_dm_set_vline0_irq_state(struct amdgpu_device *adev, __func__); } +static int amdgpu_dm_set_dmub_outbox_irq_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned int crtc_id, + enum amdgpu_interrupt_state state) +{ + enum dc_irq_source irq_source = DC_IRQ_SOURCE_DMCUB_OUTBOX; + bool st = (state == AMDGPU_IRQ_STATE_ENABLE); + + dc_interrupt_set(adev->dm.dc, irq_source, st); + return 0; +} + static int amdgpu_dm_set_vupdate_irq_state(struct amdgpu_device *adev, struct amdgpu_irq_src *source, unsigned int crtc_id, @@ -805,6 +817,11 @@ static const struct amdgpu_irq_src_funcs dm_vline0_irq_funcs = { .process = amdgpu_dm_irq_handler, }; +static const struct amdgpu_irq_src_funcs dm_dmub_outbox_irq_funcs = { + .set = amdgpu_dm_set_dmub_outbox_irq_state, + .process = amdgpu_dm_irq_handler, +}; + static const struct amdgpu_irq_src_funcs dm_vupdate_irq_funcs = { .set = amdgpu_dm_set_vupdate_irq_state, .process = amdgpu_dm_irq_handler, @@ -827,13 +844,15 @@ static const struct amdgpu_irq_src_funcs dm_hpd_irq_funcs = { void amdgpu_dm_set_irq_funcs(struct amdgpu_device *adev) { - adev->crtc_irq.num_types = adev->mode_info.num_crtc; adev->crtc_irq.funcs = &dm_crtc_irq_funcs; adev->vline0_irq.num_types = adev->mode_info.num_crtc; adev->vline0_irq.funcs = &dm_vline0_irq_funcs; + adev->dmub_outbox_irq.num_types = 1; + adev->dmub_outbox_irq.funcs = &dm_dmub_outbox_irq_funcs; + adev->vupdate_irq.num_types = adev->mode_info.num_crtc; adev->vupdate_irq.funcs = &dm_vupdate_irq_funcs; @@ -846,6 +865,12 @@ void amdgpu_dm_set_irq_funcs(struct amdgpu_device *adev) adev->hpd_irq.num_types = adev->mode_info.num_hpd; adev->hpd_irq.funcs = &dm_hpd_irq_funcs; } +void amdgpu_dm_outbox_init(struct amdgpu_device *adev) +{ + dc_interrupt_set(adev->dm.dc, + DC_IRQ_SOURCE_DMCUB_OUTBOX, + true); +} /** * amdgpu_dm_hpd_init - hpd setup callback. diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.h index 82f8e761beca..2349238a626b 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.h +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.h @@ -82,6 +82,7 @@ void amdgpu_dm_irq_unregister_interrupt(struct amdgpu_device *adev, void amdgpu_dm_set_irq_funcs(struct amdgpu_device *adev); +void amdgpu_dm_outbox_init(struct amdgpu_device *adev); void amdgpu_dm_hpd_init(struct amdgpu_device *adev); void amdgpu_dm_hpd_fini(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c index 4a0c24ce5f7d..e6b2eec9fb59 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c @@ -278,6 +278,9 @@ dm_dp_mst_detect(struct drm_connector *connector, struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector); struct amdgpu_dm_connector *master = aconnector->mst_port; + if (drm_connector_is_unregistered(connector)) + return connector_status_disconnected; + return drm_dp_mst_detect_port(connector, ctx, &master->mst_mgr, aconnector->port); } diff --git a/drivers/gpu/drm/amd/display/dc/Makefile b/drivers/gpu/drm/amd/display/dc/Makefile index f33847299bca..5cd89f7e7772 100644 --- a/drivers/gpu/drm/amd/display/dc/Makefile +++ b/drivers/gpu/drm/amd/display/dc/Makefile @@ -54,7 +54,7 @@ AMD_DC = $(addsuffix /Makefile, $(addprefix $(FULL_AMD_DISPLAY_PATH)/dc/,$(DC_LI include $(AMD_DC) -DISPLAY_CORE = dc.o dc_link.o dc_resource.o dc_hw_sequencer.o dc_sink.o \ +DISPLAY_CORE = dc.o dc_stat.o dc_link.o dc_resource.o dc_hw_sequencer.o dc_sink.o \ dc_surface.o dc_link_hwss.o dc_link_dp.o dc_link_ddc.o dc_debug.o dc_stream.o \ dc_link_enc_cfg.o diff --git a/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c b/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c index d79f4fe06c47..49126a0f66af 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c +++ b/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c @@ -836,8 +836,10 @@ static enum bp_result bios_parser_get_spread_spectrum_info( return get_ss_info_v4_1(bp, signal, index, ss_info); case 2: case 3: + case 4: return get_ss_info_v4_2(bp, signal, index, ss_info); default: + ASSERT(0); break; } break; diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr.c index a06e86853bb9..f965914ea57c 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr.c @@ -106,10 +106,10 @@ static void rn_update_clocks_update_dpp_dto(struct clk_mgr_internal *clk_mgr, for (i = 0; i < clk_mgr->base.ctx->dc->res_pool->pipe_count; i++) { int dpp_inst, dppclk_khz, prev_dppclk_khz; - /* Loop index will match dpp->inst if resource exists, - * and we want to avoid dependency on dpp object + /* Loop index may not match dpp->inst if some pipes disabled, + * so select correct inst from res_pool */ - dpp_inst = i; + dpp_inst = clk_mgr->base.ctx->dc->res_pool->dpps[i]->inst; dppclk_khz = context->res_ctx.pipe_ctx[i].plane_res.bw.dppclk_khz; prev_dppclk_khz = clk_mgr->dccg->pipe_dppclk_khz[i]; @@ -128,7 +128,7 @@ void rn_update_clocks(struct clk_mgr *clk_mgr_base, struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base); struct dc_clocks *new_clocks = &context->bw_ctx.bw.dcn.clk; struct dc *dc = clk_mgr_base->ctx->dc; - int display_count, i; + int display_count; bool update_dppclk = false; bool update_dispclk = false; bool dpp_clock_lowered = false; @@ -210,14 +210,6 @@ void rn_update_clocks(struct clk_mgr *clk_mgr_base, clk_mgr_base->clks.dppclk_khz, safe_to_lower); - for (i = 0; i < context->stream_count; i++) { - if (context->streams[i]->signal == SIGNAL_TYPE_EDP && - context->streams[i]->apply_seamless_boot_optimization) { - dc_wait_for_vblank(dc, context->streams[i]); - break; - } - } - clk_mgr_base->clks.actual_dppclk_khz = rn_vbios_smu_set_dppclk(clk_mgr, clk_mgr_base->clks.dppclk_khz); @@ -769,43 +761,6 @@ static struct wm_table ddr4_wm_table_rn = { } }; -static struct wm_table ddr4_1R_wm_table_rn = { - .entries = { - { - .wm_inst = WM_A, - .wm_type = WM_TYPE_PSTATE_CHG, - .pstate_latency_us = 11.72, - .sr_exit_time_us = 13.90, - .sr_enter_plus_exit_time_us = 14.80, - .valid = true, - }, - { - .wm_inst = WM_B, - .wm_type = WM_TYPE_PSTATE_CHG, - .pstate_latency_us = 11.72, - .sr_exit_time_us = 13.90, - .sr_enter_plus_exit_time_us = 14.80, - .valid = true, - }, - { - .wm_inst = WM_C, - .wm_type = WM_TYPE_PSTATE_CHG, - .pstate_latency_us = 11.72, - .sr_exit_time_us = 13.90, - .sr_enter_plus_exit_time_us = 14.80, - .valid = true, - }, - { - .wm_inst = WM_D, - .wm_type = WM_TYPE_PSTATE_CHG, - .pstate_latency_us = 11.72, - .sr_exit_time_us = 13.90, - .sr_enter_plus_exit_time_us = 14.80, - .valid = true, - }, - } -}; - static struct wm_table lpddr4_wm_table_rn = { .entries = { { @@ -842,46 +797,67 @@ static struct wm_table lpddr4_wm_table_rn = { }, } }; -static unsigned int find_socclk_for_voltage(struct dpm_clocks *clock_table, unsigned int voltage) + +static unsigned int find_max_fclk_for_voltage(struct dpm_clocks *clock_table, + unsigned int voltage) { int i; + uint32_t max_clk = 0; - for (i = 0; i < PP_SMU_NUM_SOCCLK_DPM_LEVELS; i++) { - if (clock_table->SocClocks[i].Vol == voltage) - return clock_table->SocClocks[i].Freq; + for (i = 0; i < PP_SMU_NUM_FCLK_DPM_LEVELS; i++) { + if (clock_table->FClocks[i].Vol <= voltage) { + max_clk = clock_table->FClocks[i].Freq > max_clk ? + clock_table->FClocks[i].Freq : max_clk; + } + } + + return max_clk; +} + +static unsigned int find_max_memclk_for_voltage(struct dpm_clocks *clock_table, + unsigned int voltage) +{ + int i; + uint32_t max_clk = 0; + + for (i = 0; i < PP_SMU_NUM_MEMCLK_DPM_LEVELS; i++) { + if (clock_table->MemClocks[i].Vol <= voltage) { + max_clk = clock_table->MemClocks[i].Freq > max_clk ? + clock_table->MemClocks[i].Freq : max_clk; + } } - ASSERT(0); - return 0; + return max_clk; } -static unsigned int find_dcfclk_for_voltage(struct dpm_clocks *clock_table, unsigned int voltage) + +static unsigned int find_max_socclk_for_voltage(struct dpm_clocks *clock_table, + unsigned int voltage) { int i; + uint32_t max_clk = 0; - for (i = 0; i < PP_SMU_NUM_DCFCLK_DPM_LEVELS; i++) { - if (clock_table->DcfClocks[i].Vol == voltage) - return clock_table->DcfClocks[i].Freq; + for (i = 0; i < PP_SMU_NUM_SOCCLK_DPM_LEVELS; i++) { + if (clock_table->SocClocks[i].Vol <= voltage) { + max_clk = clock_table->SocClocks[i].Freq > max_clk ? + clock_table->SocClocks[i].Freq : max_clk; + } } - ASSERT(0); - return 0; + return max_clk; } static void rn_clk_mgr_helper_populate_bw_params(struct clk_bw_params *bw_params, struct dpm_clocks *clock_table, struct integrated_info *bios_info) { int i, j = 0; + unsigned int volt; j = -1; - ASSERT(PP_SMU_NUM_FCLK_DPM_LEVELS <= MAX_NUM_DPM_LVL); - - /* Find lowest DPM, FCLK is filled in reverse order*/ - - for (i = PP_SMU_NUM_FCLK_DPM_LEVELS - 1; i >= 0; i--) { - if (clock_table->FClocks[i].Freq != 0 && clock_table->FClocks[i].Vol != 0) { + /* Find max DPM */ + for (i = 0; i < PP_SMU_NUM_DCFCLK_DPM_LEVELS; ++i) { + if (clock_table->DcfClocks[i].Freq != 0 && + clock_table->DcfClocks[i].Vol != 0) j = i; - break; - } } if (j == -1) { @@ -892,13 +868,18 @@ static void rn_clk_mgr_helper_populate_bw_params(struct clk_bw_params *bw_params bw_params->clk_table.num_entries = j + 1; - for (i = 0; i < bw_params->clk_table.num_entries; i++, j--) { - bw_params->clk_table.entries[i].fclk_mhz = clock_table->FClocks[j].Freq; - bw_params->clk_table.entries[i].memclk_mhz = clock_table->MemClocks[j].Freq; - bw_params->clk_table.entries[i].voltage = clock_table->FClocks[j].Vol; - bw_params->clk_table.entries[i].dcfclk_mhz = find_dcfclk_for_voltage(clock_table, clock_table->FClocks[j].Vol); - bw_params->clk_table.entries[i].socclk_mhz = find_socclk_for_voltage(clock_table, - bw_params->clk_table.entries[i].voltage); + for (i = 0; i < bw_params->clk_table.num_entries; i++) { + volt = clock_table->DcfClocks[i].Vol; + + bw_params->clk_table.entries[i].voltage = volt; + bw_params->clk_table.entries[i].dcfclk_mhz = + clock_table->DcfClocks[i].Freq; + bw_params->clk_table.entries[i].fclk_mhz = + find_max_fclk_for_voltage(clock_table, volt); + bw_params->clk_table.entries[i].memclk_mhz = + find_max_memclk_for_voltage(clock_table, volt); + bw_params->clk_table.entries[i].socclk_mhz = + find_max_socclk_for_voltage(clock_table, volt); } bw_params->vram_type = bios_info->memory_type; @@ -990,12 +971,8 @@ void rn_clk_mgr_construct( } else { if (is_green_sardine) rn_bw_params.wm_table = ddr4_wm_table_gs; - else { - if (ctx->dc->config.is_single_rank_dimm) - rn_bw_params.wm_table = ddr4_1R_wm_table_rn; - else - rn_bw_params.wm_table = ddr4_wm_table_rn; - } + else + rn_bw_params.wm_table = ddr4_wm_table_rn; } /* Saved clocks configured at boot for debug purposes */ rn_dump_clk_registers(&clk_mgr->base.boot_snapshot, &clk_mgr->base, &log_info); @@ -1013,9 +990,6 @@ void rn_clk_mgr_construct( if (status == PP_SMU_RESULT_OK && ctx->dc_bios && ctx->dc_bios->integrated_info) { rn_clk_mgr_helper_populate_bw_params (clk_mgr->base.bw_params, &clock_table, ctx->dc_bios->integrated_info); - /* treat memory config as single channel if memory is asymmetrics. */ - if (ctx->dc->config.is_asymmetric_memory) - clk_mgr->base.bw_params->num_channels = 1; } } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 4713f09bcbf1..d6f11dbf1bd4 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -59,7 +59,6 @@ #include "dc_link_ddc.h" #include "dm_helpers.h" #include "mem_input.h" -#include "hubp.h" #include "dc_link_dp.h" #include "dc_dmub_srv.h" @@ -3219,19 +3218,6 @@ void dc_link_remove_remote_sink(struct dc_link *link, struct dc_sink *sink) } } -void dc_wait_for_vblank(struct dc *dc, struct dc_stream_state *stream) -{ - int i; - - for (i = 0; i < dc->res_pool->pipe_count; i++) - if (dc->current_state->res_ctx.pipe_ctx[i].stream == stream) { - struct timing_generator *tg = - dc->current_state->res_ctx.pipe_ctx[i].stream_res.tg; - tg->funcs->wait_for_state(tg, CRTC_STATE_VBLANK); - break; - } -} - void get_clock_requirements_for_state(struct dc_state *state, struct AsicStateEx *info) { info->displayClock = (unsigned int)state->bw_ctx.bw.dcn.clk.dispclk_khz; @@ -3287,7 +3273,7 @@ void dc_allow_idle_optimizations(struct dc *dc, bool allow) if (dc->debug.disable_idle_power_optimizations) return; - if (dc->clk_mgr->funcs->is_smu_present) + if (dc->clk_mgr != NULL && dc->clk_mgr->funcs->is_smu_present) if (!dc->clk_mgr->funcs->is_smu_present(dc->clk_mgr)) return; 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 f4374d83662a..81beff4e9434 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link.c @@ -48,6 +48,7 @@ #include "dce/dmub_psr.h" #include "dmub/dmub_srv.h" #include "inc/hw/panel_cntl.h" +#include "inc/link_enc_cfg.h" #define DC_LOGGER_INIT(logger) @@ -247,6 +248,16 @@ bool dc_link_detect_sink(struct dc_link *link, enum dc_connection_type *type) link->dc->hwss.edp_wait_for_hpd_ready(link, true); } + /* Link may not have physical HPD pin. */ + if (link->ep_type != DISPLAY_ENDPOINT_PHY) { + if (link->hpd_status) + *type = dc_connection_single; + else + *type = dc_connection_none; + + return true; + } + /* todo: may need to lock gpio access */ hpd_pin = get_hpd_gpio(link->ctx->dc_bios, link->link_id, link->ctx->gpio_service); @@ -432,8 +443,18 @@ bool dc_link_is_dp_sink_present(struct dc_link *link) static enum signal_type link_detect_sink(struct dc_link *link, enum dc_detect_reason reason) { - enum signal_type result = get_basic_signal_type(link->link_enc->id, - link->link_id); + enum signal_type result; + struct graphics_object_id enc_id; + + if (link->is_dig_mapping_flexible) + enc_id = (struct graphics_object_id){.id = ENCODER_ID_UNKNOWN}; + else + enc_id = link->link_enc->id; + result = get_basic_signal_type(enc_id, link->link_id); + + /* Use basic signal type for link without physical connector. */ + if (link->ep_type != DISPLAY_ENDPOINT_PHY) + return result; /* Internal digital encoder will detect only dongles * that require digital signal @@ -762,19 +783,20 @@ static bool detect_dp(struct dc_link *link, } if (link->type != dc_connection_mst_branch && - is_dp_active_dongle(link)) { - /* DP active dongles */ - link->type = dc_connection_active_dongle; + is_dp_branch_device(link)) { + /* DP SST branch */ + link->type = dc_connection_sst_branch; if (!link->dpcd_caps.sink_count.bits.SINK_COUNT) { /* - * active dongle unplug processing for short irq + * SST branch unplug processing for short irq */ link_disconnect_sink(link); return true; } - if (link->dpcd_caps.dongle_type != - DISPLAY_DONGLE_DP_HDMI_CONVERTER) + if (is_dp_active_dongle(link) && + (link->dpcd_caps.dongle_type != + DISPLAY_DONGLE_DP_HDMI_CONVERTER)) *converter_disable_audio = true; } } else { @@ -954,7 +976,8 @@ static bool dc_link_detect_helper(struct dc_link *link, case SIGNAL_TYPE_DISPLAY_PORT: { /* wa HPD high coming too early*/ - if (link->link_enc->features.flags.bits.DP_IS_USB_C == 1) { + if (link->ep_type == DISPLAY_ENDPOINT_PHY && + link->link_enc->features.flags.bits.DP_IS_USB_C == 1) { /* if alt mode times out, return false */ if (!wait_for_entering_dp_alt_mode(link)) return false; @@ -974,8 +997,8 @@ static bool dc_link_detect_helper(struct dc_link *link, sizeof(struct dpcd_caps))) same_dpcd = false; } - /* Active dongle downstream unplug*/ - if (link->type == dc_connection_active_dongle && + /* Active SST downstream branch device unplug*/ + if (link->type == dc_connection_sst_branch && link->dpcd_caps.sink_count.bits.SINK_COUNT == 0) { if (prev_sink) /* Downstream unplug */ @@ -1206,14 +1229,25 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) { const struct dc *dc = link->dc; bool ret; + bool can_apply_seamless_boot = false; + int i; + + for (i = 0; i < dc->current_state->stream_count; i++) { + if (dc->current_state->streams[i]->apply_seamless_boot_optimization) { + can_apply_seamless_boot = true; + break; + } + } /* get out of low power state */ - clk_mgr_exit_optimized_pwr_state(dc, dc->clk_mgr); + if (!can_apply_seamless_boot && reason != DETECT_REASON_BOOT) + clk_mgr_exit_optimized_pwr_state(dc, dc->clk_mgr); ret = dc_link_detect_helper(link, reason); /* Go back to power optimized state */ - clk_mgr_optimize_pwr_state(dc, dc->clk_mgr); + if (!can_apply_seamless_boot && reason != DETECT_REASON_BOOT) + clk_mgr_optimize_pwr_state(dc, dc->clk_mgr); return ret; } @@ -1716,6 +1750,8 @@ static enum dc_status enable_link_dp(struct dc_state *state, bool apply_seamless_boot_optimization = false; uint32_t bl_oled_enable_delay = 50; // in ms const uint32_t post_oui_delay = 30; // 30ms + /* Reduce link bandwidth between failed link training attempts. */ + bool do_fallback = false; // check for seamless boot for (i = 0; i < state->stream_count; i++) { @@ -1754,7 +1790,8 @@ static enum dc_status enable_link_dp(struct dc_state *state, skip_video_pattern, LINK_TRAINING_ATTEMPTS, pipe_ctx, - pipe_ctx->stream->signal)) { + pipe_ctx->stream->signal, + do_fallback)) { link->cur_link_settings = link_settings; status = DC_OK; } else { @@ -3475,9 +3512,11 @@ uint32_t dc_bandwidth_in_kbps_from_timing( uint32_t kbps; #if defined(CONFIG_DRM_AMD_DC_DCN) - if (timing->flags.DSC) { - return dc_dsc_stream_bandwidth_in_kbps(timing->pix_clk_100hz, timing->dsc_cfg.bits_per_pixel); - } + if (timing->flags.DSC) + return dc_dsc_stream_bandwidth_in_kbps(timing, + timing->dsc_cfg.bits_per_pixel, + timing->dsc_cfg.num_slices_h, + timing->dsc_cfg.is_dp); #endif switch (timing->display_color_depth) { @@ -3539,19 +3578,6 @@ void dc_link_set_drive_settings(struct dc *dc, dc_link_dp_set_drive_settings(dc->links[i], lt_settings); } -void dc_link_perform_link_training(struct dc *dc, - struct dc_link_settings *link_setting, - bool skip_video_pattern) -{ - int i; - - for (i = 0; i < dc->link_count; i++) - dc_link_dp_perform_link_training( - dc->links[i], - link_setting, - skip_video_pattern); -} - void dc_link_set_preferred_link_settings(struct dc *dc, struct dc_link_settings *link_setting, struct dc_link *link) @@ -3702,8 +3728,22 @@ void dc_link_overwrite_extended_receiver_cap( bool dc_link_is_fec_supported(const struct dc_link *link) { + struct link_encoder *link_enc = NULL; + + /* Links supporting dynamically assigned link encoder will be assigned next + * available encoder if one not already assigned. + */ + if (link->is_dig_mapping_flexible && + link->dc->res_pool->funcs->link_encs_assign) { + link_enc = link_enc_cfg_get_link_enc_used_by_link(link->dc->current_state, link); + if (link_enc == NULL) + link_enc = link_enc_cfg_get_next_avail_link_enc(link->dc, link->dc->current_state); + } else + link_enc = link->link_enc; + ASSERT(link_enc); + return (dc_is_dp_signal(link->connector_signal) && - link->link_enc->features.fec_supported && + link_enc->features.fec_supported && link->dpcd_caps.fec_cap.bits.FEC_CAPABLE && !IS_FPGA_MAXIMUS_DC(link->ctx->dce_environment)); } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c index 64414c51312d..ba6b56f20269 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c @@ -658,7 +658,10 @@ int dc_link_aux_transfer_raw(struct ddc_service *ddc, struct aux_payload *payload, enum aux_return_code_type *operation_result) { - return dce_aux_transfer_raw(ddc, payload, operation_result); + if (dc_enable_dmub_notifications(ddc->ctx->dc)) + return dce_aux_transfer_dmub_raw(ddc, payload, operation_result); + else + return dce_aux_transfer_raw(ddc, payload, operation_result); } /* dc_link_aux_transfer_with_retries() - Attempt to submit an @@ -682,6 +685,10 @@ bool dc_link_aux_try_to_configure_timeout(struct ddc_service *ddc, bool result = false; struct ddc *ddc_pin = ddc->ddc_pin; + /* Do not try to access nonexistent DDC pin. */ + if (ddc->link->ep_type != DISPLAY_ENDPOINT_PHY) + return true; + if (ddc->ctx->dc->res_pool->engines[ddc_pin->pin_data->en]->funcs->configure_timeout) { ddc->ctx->dc->res_pool->engines[ddc_pin->pin_data->en]->funcs->configure_timeout(ddc, timeout); result = true; 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 afa43181dec6..ba4883fca616 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 @@ -14,6 +14,7 @@ #include "dpcd_defs.h" #include "dc_dmub_srv.h" #include "dce/dmub_hw_lock_mgr.h" +#include "inc/link_enc_cfg.h" /*Travis*/ static const uint8_t DP_VGA_LVDS_CONVERTER_ID_2[] = "sivarT"; @@ -107,10 +108,50 @@ static void wait_for_training_aux_rd_interval( wait_in_micro_secs); } +static enum dpcd_training_patterns + dc_dp_training_pattern_to_dpcd_training_pattern( + struct dc_link *link, + enum dc_dp_training_pattern pattern) +{ + enum dpcd_training_patterns dpcd_tr_pattern = + DPCD_TRAINING_PATTERN_VIDEOIDLE; + + switch (pattern) { + case DP_TRAINING_PATTERN_SEQUENCE_1: + dpcd_tr_pattern = DPCD_TRAINING_PATTERN_1; + break; + case DP_TRAINING_PATTERN_SEQUENCE_2: + dpcd_tr_pattern = DPCD_TRAINING_PATTERN_2; + break; + case DP_TRAINING_PATTERN_SEQUENCE_3: + dpcd_tr_pattern = DPCD_TRAINING_PATTERN_3; + break; + case DP_TRAINING_PATTERN_SEQUENCE_4: + dpcd_tr_pattern = DPCD_TRAINING_PATTERN_4; + break; + case DP_TRAINING_PATTERN_VIDEOIDLE: + dpcd_tr_pattern = DPCD_TRAINING_PATTERN_VIDEOIDLE; + break; + default: + ASSERT(0); + DC_LOG_HW_LINK_TRAINING("%s: Invalid HW Training pattern: %d\n", + __func__, pattern); + break; + } + + return dpcd_tr_pattern; +} + static void dpcd_set_training_pattern( struct dc_link *link, - union dpcd_training_pattern dpcd_pattern) + enum dc_dp_training_pattern training_pattern) { + union dpcd_training_pattern dpcd_pattern = { {0} }; + + dpcd_pattern.v1_4.TRAINING_PATTERN_SET = + dc_dp_training_pattern_to_dpcd_training_pattern( + link, training_pattern); + core_link_write_dpcd( link, DP_TRAINING_PATTERN_SET, @@ -132,10 +173,22 @@ static enum dc_dp_training_pattern decide_cr_training_pattern( static enum dc_dp_training_pattern decide_eq_training_pattern(struct dc_link *link, const struct dc_link_settings *link_settings) { + struct link_encoder *link_enc; enum dc_dp_training_pattern highest_tp = DP_TRAINING_PATTERN_SEQUENCE_2; - struct encoder_feature_support *features = &link->link_enc->features; + struct encoder_feature_support *features; struct dpcd_caps *dpcd_caps = &link->dpcd_caps; + /* Access link encoder capability based on whether it is statically + * or dynamically assigned to a link. + */ + if (link->is_dig_mapping_flexible && + link->dc->res_pool->funcs->link_encs_assign) + link_enc = link_enc_cfg_get_link_enc_used_by_link(link->dc->current_state, link); + else + link_enc = link->link_enc; + ASSERT(link_enc); + features = &link_enc->features; + if (features->flags.bits.IS_TPS3_CAPABLE) highest_tp = DP_TRAINING_PATTERN_SEQUENCE_3; @@ -227,37 +280,6 @@ static void dpcd_set_link_settings( } } -static enum dpcd_training_patterns - dc_dp_training_pattern_to_dpcd_training_pattern( - struct dc_link *link, - enum dc_dp_training_pattern pattern) -{ - enum dpcd_training_patterns dpcd_tr_pattern = - DPCD_TRAINING_PATTERN_VIDEOIDLE; - - switch (pattern) { - case DP_TRAINING_PATTERN_SEQUENCE_1: - dpcd_tr_pattern = DPCD_TRAINING_PATTERN_1; - break; - case DP_TRAINING_PATTERN_SEQUENCE_2: - dpcd_tr_pattern = DPCD_TRAINING_PATTERN_2; - break; - case DP_TRAINING_PATTERN_SEQUENCE_3: - dpcd_tr_pattern = DPCD_TRAINING_PATTERN_3; - break; - case DP_TRAINING_PATTERN_SEQUENCE_4: - dpcd_tr_pattern = DPCD_TRAINING_PATTERN_4; - break; - default: - ASSERT(0); - DC_LOG_HW_LINK_TRAINING("%s: Invalid HW Training pattern: %d\n", - __func__, pattern); - break; - } - - return dpcd_tr_pattern; -} - static uint8_t dc_dp_initialize_scrambling_data_symbols( struct dc_link *link, enum dc_dp_training_pattern pattern) @@ -420,20 +442,30 @@ static bool is_cr_done(enum dc_lane_count ln_count, } static bool is_ch_eq_done(enum dc_lane_count ln_count, - union lane_status *dpcd_lane_status, - union lane_align_status_updated *lane_status_updated) + union lane_status *dpcd_lane_status) { + bool done = true; uint32_t lane; - if (!lane_status_updated->bits.INTERLANE_ALIGN_DONE) - return false; - else { - for (lane = 0; lane < (uint32_t)(ln_count); lane++) { - if (!dpcd_lane_status[lane].bits.SYMBOL_LOCKED_0 || - !dpcd_lane_status[lane].bits.CHANNEL_EQ_DONE_0) - return false; - } - } - return true; + for (lane = 0; lane < (uint32_t)(ln_count); lane++) + if (!dpcd_lane_status[lane].bits.CHANNEL_EQ_DONE_0) + done = false; + return done; +} + +static bool is_symbol_locked(enum dc_lane_count ln_count, + union lane_status *dpcd_lane_status) +{ + bool locked = true; + uint32_t lane; + for (lane = 0; lane < (uint32_t)(ln_count); lane++) + if (!dpcd_lane_status[lane].bits.SYMBOL_LOCKED_0) + locked = false; + return locked; +} + +static inline bool is_interlane_aligned(union lane_align_status_updated align_status) +{ + return align_status.bits.INTERLANE_ALIGN_DONE == 1; } static void update_drive_settings( @@ -835,10 +867,9 @@ static bool perform_post_lt_adj_req_sequence( if (!is_cr_done(lane_count, dpcd_lane_status)) return false; - if (!is_ch_eq_done( - lane_count, - dpcd_lane_status, - &dpcd_lane_status_updated)) + if (!is_ch_eq_done(lane_count, dpcd_lane_status) || + !is_symbol_locked(lane_count, dpcd_lane_status) || + !is_interlane_aligned(dpcd_lane_status_updated)) return false; for (lane = 0; lane < (uint32_t)(lane_count); lane++) { @@ -992,9 +1023,9 @@ static enum link_training_result perform_channel_equalization_sequence( return LINK_TRAINING_EQ_FAIL_CR; /* 6. check CHEQ done*/ - if (is_ch_eq_done(lane_count, - dpcd_lane_status, - &dpcd_lane_status_updated)) + if (is_ch_eq_done(lane_count, dpcd_lane_status) && + is_symbol_locked(lane_count, dpcd_lane_status) && + is_interlane_aligned(dpcd_lane_status_updated)) return LINK_TRAINING_SUCCESS; /* 7. update VS/PE/PC2 in lt_settings*/ @@ -1162,7 +1193,7 @@ static inline enum link_training_result perform_link_training_int( return status; } -static enum link_training_result check_link_loss_status( +enum link_training_result dp_check_link_loss_status( struct dc_link *link, const struct link_training_settings *link_training_setting) { @@ -1296,7 +1327,7 @@ static void initialize_training_settings( lt_settings->enhanced_framing = 1; } -static uint8_t convert_to_count(uint8_t lttpr_repeater_count) +uint8_t dp_convert_to_count(uint8_t lttpr_repeater_count) { switch (lttpr_repeater_count) { case 0x80: // 1 lttpr repeater @@ -1365,7 +1396,8 @@ static void configure_lttpr_mode_non_transparent(struct dc_link *link) link->dpcd_caps.lttpr_caps.mode = repeater_mode; } - repeater_cnt = convert_to_count(link->dpcd_caps.lttpr_caps.phy_repeater_cnt); + repeater_cnt = dp_convert_to_count(link->dpcd_caps.lttpr_caps.phy_repeater_cnt); + for (repeater_id = repeater_cnt; repeater_id > 0; repeater_id--) { aux_interval_address = DP_TRAINING_AUX_RD_INTERVAL_PHY_REPEATER1 + ((DP_REPEATER_CONFIGURATION_AND_STATUS_SIZE) * (repeater_id - 1)); @@ -1555,7 +1587,6 @@ enum link_training_result dc_link_dp_perform_link_training( { enum link_training_result status = LINK_TRAINING_SUCCESS; struct link_training_settings lt_settings; - union dpcd_training_pattern dpcd_pattern = { { 0 } }; bool fec_enable; uint8_t repeater_cnt; @@ -1591,7 +1622,7 @@ enum link_training_result dc_link_dp_perform_link_training( /* 2. perform link training (set link training done * to false is done as well) */ - repeater_cnt = convert_to_count(link->dpcd_caps.lttpr_caps.phy_repeater_cnt); + repeater_cnt = dp_convert_to_count(link->dpcd_caps.lttpr_caps.phy_repeater_cnt); for (repeater_id = repeater_cnt; (repeater_id > 0 && status == LINK_TRAINING_SUCCESS); repeater_id--) { @@ -1621,8 +1652,7 @@ enum link_training_result dc_link_dp_perform_link_training( } /* 3. set training not in progress*/ - dpcd_pattern.v1_4.TRAINING_PATTERN_SET = DPCD_TRAINING_PATTERN_VIDEOIDLE; - dpcd_set_training_pattern(link, dpcd_pattern); + dpcd_set_training_pattern(link, DP_TRAINING_PATTERN_VIDEOIDLE); if ((status == LINK_TRAINING_SUCCESS) || !skip_video_pattern) { status = perform_link_training_int(link, <_settings, @@ -1634,7 +1664,7 @@ enum link_training_result dc_link_dp_perform_link_training( */ if (link->connector_signal != SIGNAL_TYPE_EDP && status == LINK_TRAINING_SUCCESS) { msleep(5); - status = check_link_loss_status(link, <_settings); + status = dp_check_link_loss_status(link, <_settings); } /* 6. print status message*/ @@ -1687,18 +1717,31 @@ bool perform_link_training_with_retries( bool skip_video_pattern, int attempts, struct pipe_ctx *pipe_ctx, - enum signal_type signal) + enum signal_type signal, + bool do_fallback) { uint8_t j; uint8_t delay_between_attempts = LINK_TRAINING_RETRY_DELAY; struct dc_stream_state *stream = pipe_ctx->stream; struct dc_link *link = stream->link; enum dp_panel_mode panel_mode; + struct link_encoder *link_enc; + enum link_training_result status = LINK_TRAINING_CR_FAIL_LANE0; + struct dc_link_settings currnet_setting = *link_setting; + + /* Dynamically assigned link encoders associated with stream rather than + * link. + */ + if (link->dc->res_pool->funcs->link_encs_assign) + link_enc = stream->link_enc; + else + link_enc = link->link_enc; + ASSERT(link_enc); /* We need to do this before the link training to ensure the idle pattern in SST * mode will be sent right after the link training */ - link->link_enc->funcs->connect_dig_be_to_fe(link->link_enc, + link_enc->funcs->connect_dig_be_to_fe(link_enc, pipe_ctx->stream_res.stream_enc->id, true); for (j = 0; j < attempts; ++j) { @@ -1710,7 +1753,7 @@ bool perform_link_training_with_retries( link, signal, pipe_ctx->clock_source->id, - link_setting); + &currnet_setting); if (stream->sink_patches.dppowerup_delay > 0) { int delay_dp_power_up_in_ms = stream->sink_patches.dppowerup_delay; @@ -1725,14 +1768,12 @@ bool perform_link_training_with_retries( panel_mode != DP_PANEL_MODE_DEFAULT); if (link->aux_access_disabled) { - dc_link_dp_perform_link_training_skip_aux(link, link_setting); + dc_link_dp_perform_link_training_skip_aux(link, &currnet_setting); return true; } else { - enum link_training_result status = LINK_TRAINING_CR_FAIL_LANE0; - status = dc_link_dp_perform_link_training( link, - link_setting, + &currnet_setting, skip_video_pattern); if (status == LINK_TRAINING_SUCCESS) return true; @@ -1740,7 +1781,7 @@ bool perform_link_training_with_retries( /* latest link training still fail, skip delay and keep PHY on */ - if (j == (attempts - 1)) + if (j == (attempts - 1) && link->ep_type == DISPLAY_ENDPOINT_PHY) break; DC_LOG_WARNING("%s: Link training attempt %u of %d failed\n", @@ -1748,6 +1789,19 @@ bool perform_link_training_with_retries( dp_disable_link_phy(link, signal); + /* Abort link training if failure due to sink being unplugged. */ + if (status == LINK_TRAINING_ABORT) + break; + else if (do_fallback) { + decide_fallback_link_setting(*link_setting, &currnet_setting, status); + /* Fail link training if reduced link bandwidth no longer meets + * stream requirements. + */ + if (dc_bandwidth_in_kbps_from_timing(&stream->timing) < + dc_link_bandwidth_kbps(link, &currnet_setting)) + break; + } + msleep(delay_between_attempts); delay_between_attempts += LINK_TRAINING_RETRY_DELAY; @@ -2429,6 +2483,12 @@ bool dp_validate_mode_timing( const struct dc_link_settings *link_setting; + /* According to spec, VSC SDP should be used if pixel format is YCbCr420 */ + if (timing->pixel_encoding == PIXEL_ENCODING_YCBCR420 && + !link->dpcd_caps.dprx_feature.bits.VSC_SDP_COLORIMETRY_SUPPORTED && + dal_graphics_object_id_get_connector_id(link->link_id) != CONNECTOR_ID_VIRTUAL) + return false; + /*always DP fail safe mode*/ if ((timing->pix_clk_100hz / 10) == (uint32_t) 25175 && timing->h_addressable == (uint32_t) 640 && @@ -2611,13 +2671,11 @@ static bool allow_hpd_rx_irq(const struct dc_link *link) /* * Don't handle RX IRQ unless one of following is met: * 1) The link is established (cur_link_settings != unknown) - * 2) We kicked off MST detection - * 3) We know we're dealing with an active dongle + * 2) We know we're dealing with a branch device, SST or MST */ if ((link->cur_link_settings.lane_count != LANE_COUNT_UNKNOWN) || - (link->type == dc_connection_mst_branch) || - is_dp_active_dongle(link)) + is_dp_branch_device(link)) return true; return false; @@ -2917,6 +2975,22 @@ static void dp_test_send_link_test_pattern(struct dc_link *link) break; } + switch (dpcd_test_params.bits.CLR_FORMAT) { + case 0: + pipe_ctx->stream->timing.pixel_encoding = PIXEL_ENCODING_RGB; + break; + case 1: + pipe_ctx->stream->timing.pixel_encoding = PIXEL_ENCODING_YCBCR422; + break; + case 2: + pipe_ctx->stream->timing.pixel_encoding = PIXEL_ENCODING_YCBCR444; + break; + default: + pipe_ctx->stream->timing.pixel_encoding = PIXEL_ENCODING_RGB; + break; + } + + if (requestColorDepth != COLOR_DEPTH_UNDEFINED && pipe_ctx->stream->timing.display_color_depth != requestColorDepth) { DC_LOG_DEBUG("%s: original bpc %d, changing to %d\n", @@ -2924,9 +2998,10 @@ static void dp_test_send_link_test_pattern(struct dc_link *link) pipe_ctx->stream->timing.display_color_depth, requestColorDepth); pipe_ctx->stream->timing.display_color_depth = requestColorDepth; - dp_update_dsc_config(pipe_ctx); } + dp_update_dsc_config(pipe_ctx); + dc_link_dp_set_test_pattern( link, test_pattern, @@ -3182,7 +3257,7 @@ bool dc_link_handle_hpd_rx_irq(struct dc_link *link, union hpd_irq_data *out_hpd *out_link_loss = true; } - if (link->type == dc_connection_active_dongle && + if (link->type == dc_connection_sst_branch && hpd_irq_dpcd_data.bytes.sink_cnt.bits.SINK_COUNT != link->dpcd_sink_count) status = true; @@ -3233,6 +3308,12 @@ bool is_mst_supported(struct dc_link *link) bool is_dp_active_dongle(const struct dc_link *link) { + return (link->dpcd_caps.dongle_type >= DISPLAY_DONGLE_DP_VGA_CONVERTER) && + (link->dpcd_caps.dongle_type <= DISPLAY_DONGLE_DP_HDMI_CONVERTER); +} + +bool is_dp_branch_device(const struct dc_link *link) +{ return link->dpcd_caps.is_branch_dev; } @@ -3593,7 +3674,9 @@ static bool retrieve_link_cap(struct dc_link *link) lttpr_dpcd_data[DP_PHY_REPEATER_EXTENDED_WAIT_TIMEOUT - DP_LT_TUNABLE_PHY_REPEATER_FIELD_DATA_STRUCTURE_REV]; + /* Attempt to train in LTTPR transparent mode if repeater count exceeds 8. */ is_lttpr_present = (link->dpcd_caps.lttpr_caps.phy_repeater_cnt > 0 && + link->dpcd_caps.lttpr_caps.phy_repeater_cnt < 0xff && link->dpcd_caps.lttpr_caps.max_lane_count > 0 && link->dpcd_caps.lttpr_caps.max_lane_count <= 4 && link->dpcd_caps.lttpr_caps.revision.raw >= 0x14); diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_enc_cfg.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_enc_cfg.c index 1361b87d86d7..1a89d565c92e 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_enc_cfg.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_enc_cfg.c @@ -112,8 +112,8 @@ static void update_link_enc_assignment( /* Return first available DIG link encoder. */ static enum engine_id find_first_avail_link_enc( - struct dc_context *ctx, - struct dc_state *state) + const struct dc_context *ctx, + const struct dc_state *state) { enum engine_id eng_id = ENGINE_ID_UNKNOWN; int i; @@ -270,7 +270,7 @@ struct dc_link *link_enc_cfg_get_link_using_link_enc( struct link_encoder *link_enc_cfg_get_link_enc_used_by_link( struct dc_state *state, - struct dc_link *link) + const struct dc_link *link) { struct link_encoder *link_enc = NULL; struct display_endpoint_id ep_id; @@ -296,8 +296,20 @@ struct link_encoder *link_enc_cfg_get_link_enc_used_by_link( if (stream_idx != -1) link_enc = state->streams[stream_idx]->link_enc; - else - dm_output_to_console("%s: No link encoder used by link(%d).\n", __func__, link->link_index); + + return link_enc; +} + +struct link_encoder *link_enc_cfg_get_next_avail_link_enc( + const struct dc *dc, + const struct dc_state *state) +{ + struct link_encoder *link_enc = NULL; + enum engine_id eng_id = ENGINE_ID_UNKNOWN; + + eng_id = find_first_avail_link_enc(dc->ctx, state); + if (eng_id != ENGINE_ID_UNKNOWN) + link_enc = dc->res_pool->link_encoders[eng_id - ENGINE_ID_DIGA]; return link_enc; } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c index b426f878fb99..13c5c4a34a58 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c @@ -384,7 +384,8 @@ void dp_retrain_link_dp_test(struct dc_link *link, skip_video_pattern, LINK_TRAINING_ATTEMPTS, &pipes[i], - SIGNAL_TYPE_DISPLAY_PORT); + SIGNAL_TYPE_DISPLAY_PORT, + false); link->dc->hwss.enable_stream(&pipes[i]); 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 8cb937c046aa..f2bc630a0871 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c @@ -1706,12 +1706,6 @@ static bool is_timing_changed(struct dc_stream_state *cur_stream, if (cur_stream == NULL) return true; - /* If sink pointer changed, it means this is a hotplug, we should do - * full hw setting. - */ - if (cur_stream->sink != new_stream->sink) - return true; - /* If output color space is changed, need to reprogram info frames */ if (cur_stream->output_color_space != new_stream->output_color_space) return true; @@ -2679,6 +2673,7 @@ void dc_resource_state_destruct(struct dc_state *context) dc_stream_release(context->streams[i]); context->streams[i] = NULL; } + context->stream_count = 0; } void dc_resource_state_copy_construct( diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 100d434f7a03..d26153ab9d62 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -45,7 +45,7 @@ /* forward declaration */ struct aux_payload; -#define DC_VER "3.2.132" +#define DC_VER "3.2.135.1" #define MAX_SURFACES 3 #define MAX_PLANES 6 @@ -308,8 +308,6 @@ struct dc_config { #endif uint64_t vblank_alignment_dto_params; uint8_t vblank_alignment_max_frame_time_diff; - bool is_asymmetric_memory; - bool is_single_rank_dimm; }; enum visual_confirm { @@ -600,7 +598,6 @@ struct dc_bounding_box_overrides { int min_dcfclk_mhz; }; -struct dc_state; struct resource_pool; struct dce_hwseq; struct gpu_info_soc_bounding_box_v1_0; @@ -719,7 +716,6 @@ void dc_init_callbacks(struct dc *dc, void dc_deinit_callbacks(struct dc *dc); void dc_destroy(struct dc **dc); -void dc_wait_for_vblank(struct dc *dc, struct dc_stream_state *stream); /******************************************************************************* * Surface Interfaces ******************************************************************************/ diff --git a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c index 6b72af2b3f4c..c5dc3a947020 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c +++ b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c @@ -180,5 +180,5 @@ bool dc_dmub_srv_get_dmub_outbox0_msg(const struct dc *dc, struct dmcub_trace_bu void dc_dmub_trace_event_control(struct dc *dc, bool enable) { - dm_helpers_dmub_outbox0_interrupt_control(dc->ctx, enable); + dm_helpers_dmub_outbox_interrupt_control(dc->ctx, enable); } diff --git a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h index cc6fb838420e..fcfde2e04292 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h @@ -95,6 +95,7 @@ enum dc_dp_training_pattern { DP_TRAINING_PATTERN_SEQUENCE_2, DP_TRAINING_PATTERN_SEQUENCE_3, DP_TRAINING_PATTERN_SEQUENCE_4, + DP_TRAINING_PATTERN_VIDEOIDLE, }; struct dc_link_settings { diff --git a/drivers/gpu/drm/amd/display/dc/dc_dsc.h b/drivers/gpu/drm/amd/display/dc/dc_dsc.h index c51d2d961b7a..16cc76ce3739 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dsc.h +++ b/drivers/gpu/drm/amd/display/dc/dc_dsc.h @@ -78,7 +78,8 @@ bool dc_dsc_compute_config( const struct dc_crtc_timing *timing, struct dc_dsc_config *dsc_cfg); -uint32_t dc_dsc_stream_bandwidth_in_kbps(uint32_t pix_clk_100hz, uint32_t bpp_x16); +uint32_t dc_dsc_stream_bandwidth_in_kbps(const struct dc_crtc_timing *timing, + uint32_t bpp_x16, uint32_t num_slices_h, bool is_dp); void dc_dsc_get_policy_for_timing(const struct dc_crtc_timing *timing, uint32_t max_target_bpp_limit_override_x16, @@ -88,6 +89,6 @@ void dc_dsc_policy_set_max_target_bpp_limit(uint32_t limit); void dc_dsc_policy_set_enable_dsc_when_not_needed(bool enable); -uint32_t dc_dsc_stream_bandwidth_in_kbps(uint32_t pix_clk_100hz, uint32_t bpp_x16); +void dc_dsc_policy_set_disable_dsc_stream_overhead(bool disable); #endif 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 bcec019efa6f..04957a9efab2 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_hw_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_hw_types.h @@ -739,6 +739,7 @@ struct dc_dsc_config { uint32_t version_minor; /* DSC minor version. Full version is formed as 1.version_minor. */ bool ycbcr422_simple; /* Tell DSC engine to convert YCbCr 4:2:2 to 'YCbCr 4:2:2 simple'. */ int32_t rc_buffer_size; /* DSC RC buffer block size in bytes */ + bool is_dp; /* indicate if DSC is applied based on DP's capability */ }; struct dc_crtc_timing { uint32_t h_total; diff --git a/drivers/gpu/drm/amd/display/dc/dc_link.h b/drivers/gpu/drm/amd/display/dc/dc_link.h index fc5622ffec3d..c871923e7db0 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_link.h +++ b/drivers/gpu/drm/amd/display/dc/dc_link.h @@ -36,12 +36,6 @@ enum dc_link_fec_state { dc_link_fec_enabled }; -enum lttpr_mode { - LTTPR_MODE_NON_LTTPR, - LTTPR_MODE_TRANSPARENT, - LTTPR_MODE_NON_TRANSPARENT, -}; - struct dc_link_status { bool link_active; struct dpcd_caps *dpcd_caps; @@ -113,6 +107,7 @@ struct dc_link { /* TODO: Rename. Flag an endpoint as having a programmable mapping to a * DIG encoder. */ bool is_dig_mapping_flexible; + bool hpd_status; /* HPD status of link without physical HPD pin. */ bool edp_sink_present; @@ -363,9 +358,6 @@ bool dc_link_is_hdcp22(struct dc_link *link, enum signal_type signal); void dc_link_set_drive_settings(struct dc *dc, struct link_training_settings *lt_settings, const struct dc_link *link); -void dc_link_perform_link_training(struct dc *dc, - struct dc_link_settings *link_setting, - bool skip_video_pattern); void dc_link_set_preferred_link_settings(struct dc *dc, struct dc_link_settings *link_setting, struct dc_link *link); diff --git a/drivers/gpu/drm/amd/display/dc/dc_types.h b/drivers/gpu/drm/amd/display/dc/dc_types.h index 432754eaf10b..535da8db70b6 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_types.h @@ -404,7 +404,7 @@ enum dc_connection_type { dc_connection_none, dc_connection_single, dc_connection_mst_branch, - dc_connection_active_dongle + dc_connection_sst_branch }; struct dc_csc_adjustments { @@ -909,6 +909,7 @@ struct dsc_dec_dpcd_caps { uint32_t branch_overall_throughput_0_mps; /* In MPs */ uint32_t branch_overall_throughput_1_mps; /* In MPs */ uint32_t branch_max_line_width; + bool is_dp; }; struct dc_golden_table { diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c index 87d57e81de12..83d97dfe328f 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c @@ -595,6 +595,25 @@ int dce_aux_transfer_raw(struct ddc_service *ddc, return res; } +int dce_aux_transfer_dmub_raw(struct ddc_service *ddc, + struct aux_payload *payload, + enum aux_return_code_type *operation_result) +{ + struct ddc *ddc_pin = ddc->ddc_pin; + + if (ddc_pin != NULL) { + struct dce_aux *aux_engine = ddc->ctx->dc->res_pool->engines[ddc_pin->pin_data->en]; + /* XXX: Workaround to configure ddc channels for aux transactions */ + if (!acquire(aux_engine, ddc_pin)) { + *operation_result = AUX_RET_ERROR_ENGINE_ACQUIRE; + return -1; + } + release_engine(aux_engine); + } + + return dm_helper_dmub_aux_transfer_sync(ddc->ctx, ddc->link, payload, operation_result); +} + #define AUX_MAX_RETRIES 7 #define AUX_MAX_DEFER_RETRIES 7 #define AUX_MAX_I2C_DEFER_RETRIES 7 diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h index 566b1bddd8cc..e69f1899fbf0 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h @@ -304,6 +304,9 @@ int dce_aux_transfer_raw(struct ddc_service *ddc, struct aux_payload *cmd, enum aux_return_code_type *operation_result); +int dce_aux_transfer_dmub_raw(struct ddc_service *ddc, + struct aux_payload *payload, + enum aux_return_code_type *operation_result); bool dce_aux_transfer_with_retries(struct ddc_service *ddc, struct aux_payload *cmd); diff --git a/drivers/gpu/drm/amd/display/dc/dce/dmub_abm.c b/drivers/gpu/drm/amd/display/dc/dce/dmub_abm.c index eb1698d54a48..6939ca2e8212 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dmub_abm.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dmub_abm.c @@ -56,11 +56,19 @@ static void dmub_abm_enable_fractional_pwm(struct dc_context *dc) { union dmub_rb_cmd cmd; uint32_t fractional_pwm = (dc->dc->config.disable_fractional_pwm == false) ? 1 : 0; + uint32_t edp_id_count = dc->dc_edp_id_count; + int i; + uint8_t panel_mask = 0; + + for (i = 0; i < edp_id_count; i++) + panel_mask |= 0x01 << i; memset(&cmd, 0, sizeof(cmd)); cmd.abm_set_pwm_frac.header.type = DMUB_CMD__ABM; cmd.abm_set_pwm_frac.header.sub_type = DMUB_CMD__ABM_SET_PWM_FRAC; cmd.abm_set_pwm_frac.abm_set_pwm_frac_data.fractional_pwm = fractional_pwm; + cmd.abm_set_pwm_frac.abm_set_pwm_frac_data.version = DMUB_CMD_ABM_CONTROL_VERSION_1; + cmd.abm_set_pwm_frac.abm_set_pwm_frac_data.panel_mask = panel_mask; cmd.abm_set_pwm_frac.header.payload_bytes = sizeof(struct dmub_cmd_abm_set_pwm_frac_data); dc_dmub_srv_cmd_queue(dc->dmub_srv, &cmd); @@ -135,11 +143,24 @@ static bool dmub_abm_set_level(struct abm *abm, uint32_t level) { union dmub_rb_cmd cmd; struct dc_context *dc = abm->ctx; + struct dc_link *edp_links[MAX_NUM_EDP]; + int i; + int edp_num; + uint8_t panel_mask = 0; + + get_edp_links(dc->dc, edp_links, &edp_num); + + for (i = 0; i < edp_num; i++) { + if (edp_links[i]->link_status.link_active) + panel_mask |= (0x01 << i); + } memset(&cmd, 0, sizeof(cmd)); cmd.abm_set_level.header.type = DMUB_CMD__ABM; cmd.abm_set_level.header.sub_type = DMUB_CMD__ABM_SET_LEVEL; cmd.abm_set_level.abm_set_level_data.level = level; + cmd.abm_set_level.abm_set_level_data.version = DMUB_CMD_ABM_CONTROL_VERSION_1; + cmd.abm_set_level.abm_set_level_data.panel_mask = panel_mask; cmd.abm_set_level.header.payload_bytes = sizeof(struct dmub_cmd_abm_set_level_data); dc_dmub_srv_cmd_queue(dc->dmub_srv, &cmd); @@ -155,6 +176,12 @@ static bool dmub_abm_init_config(struct abm *abm, { union dmub_rb_cmd cmd; struct dc_context *dc = abm->ctx; + uint32_t edp_id_count = dc->dc_edp_id_count; + int i; + uint8_t panel_mask = 0; + + for (i = 0; i < edp_id_count; i++) + panel_mask |= 0x01 << i; // TODO: Optimize by only reading back final 4 bytes dmub_flush_buffer_mem(&dc->dmub_srv->dmub->scratch_mem_fb); @@ -168,6 +195,9 @@ static bool dmub_abm_init_config(struct abm *abm, cmd.abm_init_config.header.sub_type = DMUB_CMD__ABM_INIT_CONFIG; cmd.abm_init_config.abm_init_config_data.src.quad_part = dc->dmub_srv->dmub->scratch_mem_fb.gpu_addr; cmd.abm_init_config.abm_init_config_data.bytes = bytes; + cmd.abm_init_config.abm_init_config_data.version = DMUB_CMD_ABM_CONTROL_VERSION_1; + cmd.abm_init_config.abm_init_config_data.panel_mask = panel_mask; + cmd.abm_init_config.header.payload_bytes = sizeof(struct dmub_cmd_abm_init_config_data); dc_dmub_srv_cmd_queue(dc->dmub_srv, &cmd); diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c index efa86d5c6847..98ab4b776924 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c @@ -496,10 +496,13 @@ static enum lb_memory_config dpp1_dscl_find_lb_memory_config(struct dcn10_dpp *d int vtaps_c = scl_data->taps.v_taps_c; int ceil_vratio = dc_fixpt_ceil(scl_data->ratios.vert); int ceil_vratio_c = dc_fixpt_ceil(scl_data->ratios.vert_c); - enum lb_memory_config mem_cfg = LB_MEMORY_CONFIG_0; - if (dpp->base.ctx->dc->debug.use_max_lb) - return mem_cfg; + if (dpp->base.ctx->dc->debug.use_max_lb) { + if (scl_data->format == PIXEL_FORMAT_420BPP8 + || scl_data->format == PIXEL_FORMAT_420BPP10) + return LB_MEMORY_CONFIG_3; + return LB_MEMORY_CONFIG_0; + } dpp->base.caps->dscl_calc_lb_num_partitions( scl_data, LB_MEMORY_CONFIG_1, &num_part_y, &num_part_c); diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c index 7c939c0a977b..8c2d3d42d9da 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c @@ -2956,35 +2956,6 @@ void dcn10_program_pipe( { struct dce_hwseq *hws = dc->hwseq; - if (pipe_ctx->plane_state->update_flags.bits.full_update) - dcn10_enable_plane(dc, pipe_ctx, context); - - dcn10_update_dchubp_dpp(dc, pipe_ctx, context); - - hws->funcs.set_hdr_multiplier(pipe_ctx); - - if (pipe_ctx->plane_state->update_flags.bits.full_update || - pipe_ctx->plane_state->update_flags.bits.in_transfer_func_change || - pipe_ctx->plane_state->update_flags.bits.gamma_change) - hws->funcs.set_input_transfer_func(dc, pipe_ctx, pipe_ctx->plane_state); - - /* dcn10_translate_regamma_to_hw_format takes 750us to finish - * only do gamma programming for full update. - * TODO: This can be further optimized/cleaned up - * Always call this for now since it does memcmp inside before - * doing heavy calculation and programming - */ - if (pipe_ctx->plane_state->update_flags.bits.full_update) - hws->funcs.set_output_transfer_func(dc, pipe_ctx, pipe_ctx->stream); -} - -static void dcn10_program_all_pipe_in_tree( - struct dc *dc, - struct pipe_ctx *pipe_ctx, - struct dc_state *context) -{ - struct dce_hwseq *hws = dc->hwseq; - if (pipe_ctx->top_pipe == NULL) { bool blank = !is_pipe_tree_visible(pipe_ctx); @@ -3004,35 +2975,26 @@ static void dcn10_program_all_pipe_in_tree( hws->funcs.blank_pixel_data(dc, pipe_ctx, blank); } - if (pipe_ctx->plane_state != NULL) - hws->funcs.program_pipe(dc, pipe_ctx, context); - - if (pipe_ctx->bottom_pipe != NULL && pipe_ctx->bottom_pipe != pipe_ctx) - dcn10_program_all_pipe_in_tree(dc, pipe_ctx->bottom_pipe, context); -} - -static struct pipe_ctx *dcn10_find_top_pipe_for_stream( - struct dc *dc, - struct dc_state *context, - const struct dc_stream_state *stream) -{ - int i; + if (pipe_ctx->plane_state->update_flags.bits.full_update) + dcn10_enable_plane(dc, pipe_ctx, context); - 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]; + dcn10_update_dchubp_dpp(dc, pipe_ctx, context); - if (!pipe_ctx->plane_state && !old_pipe_ctx->plane_state) - continue; + hws->funcs.set_hdr_multiplier(pipe_ctx); - if (pipe_ctx->stream != stream) - continue; + if (pipe_ctx->plane_state->update_flags.bits.full_update || + pipe_ctx->plane_state->update_flags.bits.in_transfer_func_change || + pipe_ctx->plane_state->update_flags.bits.gamma_change) + hws->funcs.set_input_transfer_func(dc, pipe_ctx, pipe_ctx->plane_state); - if (!pipe_ctx->top_pipe && !pipe_ctx->prev_odm_pipe) - return pipe_ctx; - } - return NULL; + /* dcn10_translate_regamma_to_hw_format takes 750us to finish + * only do gamma programming for full update. + * TODO: This can be further optimized/cleaned up + * Always call this for now since it does memcmp inside before + * doing heavy calculation and programming + */ + if (pipe_ctx->plane_state->update_flags.bits.full_update) + hws->funcs.set_output_transfer_func(dc, pipe_ctx, pipe_ctx->stream); } void dcn10_wait_for_pending_cleared(struct dc *dc, @@ -3067,91 +3029,6 @@ void dcn10_wait_for_pending_cleared(struct dc *dc, } } -void dcn10_apply_ctx_for_surface( - struct dc *dc, - const struct dc_stream_state *stream, - int num_planes, - struct dc_state *context) -{ - struct dce_hwseq *hws = dc->hwseq; - int i; - struct timing_generator *tg; - uint32_t underflow_check_delay_us; - bool interdependent_update = false; - struct pipe_ctx *top_pipe_to_program = - dcn10_find_top_pipe_for_stream(dc, context, stream); - DC_LOGGER_INIT(dc->ctx->logger); - - // Clear pipe_ctx flag - for (i = 0; i < dc->res_pool->pipe_count; i++) { - struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; - pipe_ctx->update_flags.raw = 0; - } - - if (!top_pipe_to_program) - return; - - tg = top_pipe_to_program->stream_res.tg; - - interdependent_update = top_pipe_to_program->plane_state && - top_pipe_to_program->plane_state->update_flags.bits.full_update; - - underflow_check_delay_us = dc->debug.underflow_assert_delay_us; - - if (underflow_check_delay_us != 0xFFFFFFFF && hws->funcs.did_underflow_occur) - ASSERT(hws->funcs.did_underflow_occur(dc, top_pipe_to_program)); - - if (underflow_check_delay_us != 0xFFFFFFFF) - udelay(underflow_check_delay_us); - - if (underflow_check_delay_us != 0xFFFFFFFF && hws->funcs.did_underflow_occur) - ASSERT(hws->funcs.did_underflow_occur(dc, top_pipe_to_program)); - - if (num_planes == 0) { - /* OTG blank before remove all front end */ - hws->funcs.blank_pixel_data(dc, top_pipe_to_program, true); - } - - /* Disconnect unused mpcc */ - 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]; - - if ((!pipe_ctx->plane_state || - pipe_ctx->stream_res.tg != old_pipe_ctx->stream_res.tg) && - old_pipe_ctx->plane_state && - old_pipe_ctx->stream_res.tg == tg) { - - hws->funcs.plane_atomic_disconnect(dc, old_pipe_ctx); - pipe_ctx->update_flags.bits.disable = 1; - - DC_LOG_DC("Reset mpcc for pipe %d\n", - old_pipe_ctx->pipe_idx); - } - } - - if (num_planes > 0) - dcn10_program_all_pipe_in_tree(dc, top_pipe_to_program, context); - - /* Program secondary blending tree and writeback pipes */ - if ((stream->num_wb_info > 0) && (hws->funcs.program_all_writeback_pipes_in_tree)) - hws->funcs.program_all_writeback_pipes_in_tree(dc, stream, context); - if (interdependent_update) - for (i = 0; i < dc->res_pool->pipe_count; i++) { - struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; - /* Skip inactive pipes and ones already updated */ - if (!pipe_ctx->stream || pipe_ctx->stream == stream || - !pipe_ctx->plane_state || !tg->funcs->is_tg_enabled(tg)) - continue; - - pipe_ctx->plane_res.hubp->funcs->hubp_setup_interdependent( - pipe_ctx->plane_res.hubp, - &pipe_ctx->dlg_regs, - &pipe_ctx->ttu_regs); - } -} - void dcn10_post_unlock_program_front_end( struct dc *dc, struct dc_state *context) diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h index 37bec421fde8..c9bdffe5989b 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h @@ -80,11 +80,6 @@ void dcn10_lock_all_pipes( struct dc *dc, struct dc_state *context, bool lock); -void dcn10_apply_ctx_for_surface( - struct dc *dc, - const struct dc_stream_state *stream, - int num_planes, - struct dc_state *context); void dcn10_post_unlock_program_front_end( struct dc *dc, struct dc_state *context); diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_init.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_init.c index d532c78ee764..680ca53455a2 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_init.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_init.c @@ -26,13 +26,15 @@ #include "hw_sequencer_private.h" #include "dce110/dce110_hw_sequencer.h" #include "dcn10_hw_sequencer.h" +#include "dcn20/dcn20_hwseq.h" static const struct hw_sequencer_funcs dcn10_funcs = { .program_gamut_remap = dcn10_program_gamut_remap, .init_hw = dcn10_init_hw, .power_down_on_boot = dcn10_power_down_on_boot, .apply_ctx_to_hw = dce110_apply_ctx_to_hw, - .apply_ctx_for_surface = dcn10_apply_ctx_for_surface, + .apply_ctx_for_surface = NULL, + .program_front_end_for_ctx = dcn20_program_front_end_for_ctx, .post_unlock_program_front_end = dcn10_post_unlock_program_front_end, .wait_for_pending_cleared = dcn10_wait_for_pending_cleared, .update_plane_addr = dcn10_update_plane_addr, diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_dpp_cm.c b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_dpp_cm.c index 8dc3d1f73984..2feb051a2002 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_dpp_cm.c +++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_dpp_cm.c @@ -482,7 +482,7 @@ bool dpp20_program_blnd_lut( next_mode = LUT_RAM_A; dpp20_power_on_blnd_lut(dpp_base, true); - dpp20_configure_blnd_lut(dpp_base, next_mode == LUT_RAM_A ? true:false); + dpp20_configure_blnd_lut(dpp_base, next_mode == LUT_RAM_A); if (next_mode == LUT_RAM_A) dpp20_program_blnd_luta_settings(dpp_base, params); @@ -893,7 +893,7 @@ bool dpp20_program_shaper( else next_mode = LUT_RAM_A; - dpp20_configure_shaper_lut(dpp_base, next_mode == LUT_RAM_A ? true:false); + dpp20_configure_shaper_lut(dpp_base, next_mode == LUT_RAM_A); if (next_mode == LUT_RAM_A) dpp20_program_shaper_luta_settings(dpp_base, params); diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c index 6a10daec15cc..484a30592987 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c @@ -1700,7 +1700,11 @@ void dcn20_program_front_end_for_ctx( if (pipe->plane_state && !pipe->top_pipe) { while (pipe) { - dcn20_program_pipe(dc, pipe, context); + if (hws->funcs.program_pipe) + hws->funcs.program_pipe(dc, pipe, context); + else + dcn20_program_pipe(dc, pipe, context); + pipe = pipe->bottom_pipe; } /* Program secondary blending tree and writeback pipes */ diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c index 527e56c353cb..6a56a03cfba3 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c @@ -3667,7 +3667,7 @@ static bool dcn20_resource_construct( int i; struct dc_context *ctx = dc->ctx; struct irq_service_init_data init_data; - struct ddc_service_init_data ddc_init_data; + struct ddc_service_init_data ddc_init_data = {0}; struct _vcs_dpi_soc_bounding_box_st *loaded_bb = get_asic_rev_soc_bb(ctx->asic_id.hw_internal_rev); struct _vcs_dpi_ip_params_st *loaded_ip = diff --git a/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_hwseq.c b/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_hwseq.c index 8fccee5a3036..69cc192a7e71 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_hwseq.c @@ -218,7 +218,7 @@ bool dcn21_set_backlight_level(struct pipe_ctx *pipe_ctx, cmd.abm_set_backlight.header.sub_type = DMUB_CMD__ABM_SET_BACKLIGHT; cmd.abm_set_backlight.abm_set_backlight_data.frame_ramp = frame_ramp; cmd.abm_set_backlight.abm_set_backlight_data.backlight_user_level = backlight_pwm_u16_16; - cmd.abm_set_backlight.abm_set_backlight_data.version = DMUB_CMD_ABM_SET_BACKLIGHT_VERSION_1; + cmd.abm_set_backlight.abm_set_backlight_data.version = DMUB_CMD_ABM_CONTROL_VERSION_1; cmd.abm_set_backlight.abm_set_backlight_data.panel_mask = (0x01 << panel_cntl->inst); cmd.abm_set_backlight.header.payload_bytes = sizeof(struct dmub_cmd_abm_set_backlight_data); diff --git a/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_resource.c b/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_resource.c index 8e3f1d0b4cc3..38a2aa87f5f5 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_resource.c @@ -1575,10 +1575,12 @@ static struct _vcs_dpi_voltage_scaling_st construct_low_pstate_lvl(struct clk_li low_pstate_lvl.phyclk_d18_mhz = dcn2_1_soc.clock_limits[high_voltage_lvl].phyclk_d18_mhz; low_pstate_lvl.phyclk_mhz = dcn2_1_soc.clock_limits[high_voltage_lvl].phyclk_mhz; - for (i = clk_table->num_entries; i > 1; i--) - clk_table->entries[i] = clk_table->entries[i-1]; - clk_table->entries[1] = clk_table->entries[0]; - clk_table->num_entries++; + if (clk_table->num_entries < MAX_NUM_DPM_LVL) { + for (i = clk_table->num_entries; i > 1; i--) + clk_table->entries[i] = clk_table->entries[i-1]; + clk_table->entries[1] = clk_table->entries[0]; + clk_table->num_entries++; + } return low_pstate_lvl; } @@ -1610,10 +1612,6 @@ static void update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw_param } } - /* clk_table[1] is reserved for min DF PState. skip here to fill in later. */ - if (i == 1) - k++; - clock_limits[k].state = k; clock_limits[k].dcfclk_mhz = clk_table->entries[i].dcfclk_mhz; clock_limits[k].fabricclk_mhz = clk_table->entries[i].fclk_mhz; @@ -1630,14 +1628,25 @@ static void update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw_param k++; } - for (i = 0; i < clk_table->num_entries + 1; i++) - dcn2_1_soc.clock_limits[i] = clock_limits[i]; + + if (clk_table->num_entries >= MAX_NUM_DPM_LVL) { + for (i = 0; i < clk_table->num_entries + 1; i++) + dcn2_1_soc.clock_limits[i] = clock_limits[i]; + } else { + dcn2_1_soc.clock_limits[0] = clock_limits[0]; + for (i = 2; i < clk_table->num_entries + 1; i++) { + dcn2_1_soc.clock_limits[i] = clock_limits[i - 1]; + dcn2_1_soc.clock_limits[i].state = i; + } + } + if (clk_table->num_entries) { - dcn2_1_soc.num_states = clk_table->num_entries + 1; /* fill in min DF PState */ dcn2_1_soc.clock_limits[1] = construct_low_pstate_lvl(clk_table, closest_clk_lvl); + dcn2_1_soc.num_states = clk_table->num_entries; /* duplicate last level */ - dcn2_1_soc.clock_limits[dcn2_1_soc.num_states] = dcn2_1_soc.clock_limits[dcn2_1_soc.num_states - 1]; + dcn2_1_soc.clock_limits[dcn2_1_soc.num_states] = + dcn2_1_soc.clock_limits[dcn2_1_soc.num_states - 1]; dcn2_1_soc.clock_limits[dcn2_1_soc.num_states].state = dcn2_1_soc.num_states; } diff --git a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_dio_stream_encoder.h b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_dio_stream_encoder.h index 0b1755f1dea8..9566b9037458 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_dio_stream_encoder.h +++ b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_dio_stream_encoder.h @@ -85,7 +85,9 @@ SRI(DP_MSE_RATE_UPDATE, DP, id), \ SRI(DP_PIXEL_FORMAT, DP, id), \ SRI(DP_SEC_CNTL, DP, id), \ + SRI(DP_SEC_CNTL1, DP, id), \ SRI(DP_SEC_CNTL2, DP, id), \ + SRI(DP_SEC_CNTL5, DP, id), \ SRI(DP_SEC_CNTL6, DP, id), \ SRI(DP_STEER_FIFO, DP, id), \ SRI(DP_VID_M, DP, id), \ @@ -169,7 +171,9 @@ SE_SF(DP0_DP_SEC_CNTL, DP_SEC_GSP5_ENABLE, mask_sh),\ SE_SF(DP0_DP_SEC_CNTL, DP_SEC_GSP6_ENABLE, mask_sh),\ SE_SF(DP0_DP_SEC_CNTL, DP_SEC_GSP7_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL1, DP_SEC_GSP5_LINE_REFERENCE, mask_sh),\ SE_SF(DP0_DP_SEC_CNTL2, DP_SEC_GSP7_SEND, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL5, DP_SEC_GSP5_LINE_NUM, mask_sh),\ SE_SF(DP0_DP_SEC_CNTL6, DP_SEC_GSP7_LINE_NUM, mask_sh),\ SE_SF(DP0_DP_SEC_CNTL2, DP_SEC_GSP11_PPS, mask_sh),\ SE_SF(DP0_DP_GSP11_CNTL, DP_SEC_GSP11_ENABLE, mask_sh),\ diff --git a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_mpc.c b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_mpc.c index 910c17fd4278..950c9bfd53de 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_mpc.c +++ b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_mpc.c @@ -874,7 +874,7 @@ bool mpc3_program_shaper( else next_mode = LUT_RAM_A; - mpc3_configure_shaper_lut(mpc, next_mode == LUT_RAM_A ? true:false, rmu_idx); + mpc3_configure_shaper_lut(mpc, next_mode == LUT_RAM_A, rmu_idx); if (next_mode == LUT_RAM_A) mpc3_program_shaper_luta_settings(mpc, params, rmu_idx); diff --git a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_optc.c b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_optc.c index 8980c90b2277..ac478bdcfb2a 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_optc.c +++ b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_optc.c @@ -97,7 +97,7 @@ void optc3_lock_doublebuffer_disable(struct timing_generator *optc) MASTER_UPDATE_LOCK_DB_END_Y, 0); REG_UPDATE(OTG_GLOBAL_CONTROL2, GLOBAL_UPDATE_LOCK_EN, 0); - REG_UPDATE(OTG_GLOBAL_CONTROL0, MASTER_UPDATE_LOCK_DB_EN, 1); + REG_UPDATE(OTG_GLOBAL_CONTROL0, MASTER_UPDATE_LOCK_DB_EN, 0); } void optc3_lock(struct timing_generator *optc) diff --git a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_resource.c b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_resource.c index 4a5fa23d8e7b..45f96221a094 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_resource.c @@ -2538,7 +2538,7 @@ static bool dcn30_resource_construct( int i; struct dc_context *ctx = dc->ctx; struct irq_service_init_data init_data; - struct ddc_service_init_data ddc_init_data; + struct ddc_service_init_data ddc_init_data = {0}; uint32_t pipe_fuses = read_pipe_fuses(ctx); uint32_t num_pipes = 0; diff --git a/drivers/gpu/drm/amd/display/dc/dm_helpers.h b/drivers/gpu/drm/amd/display/dc/dm_helpers.h index 7617fab9e1f9..304d50d16d01 100644 --- a/drivers/gpu/drm/amd/display/dc/dm_helpers.h +++ b/drivers/gpu/drm/amd/display/dc/dm_helpers.h @@ -34,6 +34,8 @@ #include "dc.h" struct dp_mst_stream_allocation_table; +struct aux_payload; +enum aux_return_code_type; /* * Allocate memory accessible by the GPU @@ -158,6 +160,11 @@ void dm_set_dcn_clocks( struct dc_context *ctx, struct dc_clocks *clks); -bool dm_helpers_dmub_outbox0_interrupt_control(struct dc_context *ctx, bool enable); +bool dm_helpers_dmub_outbox_interrupt_control(struct dc_context *ctx, bool enable); +int dm_helper_dmub_aux_transfer_sync( + struct dc_context *ctx, + const struct dc_link *link, + struct aux_payload *payload, + enum aux_return_code_type *operation_result); #endif /* __DM_HELPERS__ */ diff --git a/drivers/gpu/drm/amd/display/dc/dml/dcn20/display_mode_vba_20.c b/drivers/gpu/drm/amd/display/dc/dml/dcn20/display_mode_vba_20.c index 9729cf292e84..d3b5b6fedf04 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/dcn20/display_mode_vba_20.c +++ b/drivers/gpu/drm/amd/display/dc/dml/dcn20/display_mode_vba_20.c @@ -2895,7 +2895,7 @@ static void dml20_DisplayPipeConfiguration(struct display_mode_lib *mode_lib) RoundedUpMaxSwathSizeBytesC = 0.0; if (RoundedUpMaxSwathSizeBytesY + RoundedUpMaxSwathSizeBytesC - <= mode_lib->vba.DETBufferSizeInKByte * 1024.0 / 2.0) { + <= mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 2.0) { mode_lib->vba.SwathHeightY[k] = MaximumSwathHeightY; mode_lib->vba.SwathHeightC[k] = MaximumSwathHeightC; } else { @@ -2904,17 +2904,17 @@ static void dml20_DisplayPipeConfiguration(struct display_mode_lib *mode_lib) } if (mode_lib->vba.SwathHeightC[k] == 0) { - mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte * 1024; + mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte[0] * 1024; mode_lib->vba.DETBufferSizeC[k] = 0; } else if (mode_lib->vba.SwathHeightY[k] <= mode_lib->vba.SwathHeightC[k]) { - mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte + mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 2; - mode_lib->vba.DETBufferSizeC[k] = mode_lib->vba.DETBufferSizeInKByte + mode_lib->vba.DETBufferSizeC[k] = mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 2; } else { - mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte + mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 * 2 / 3; - mode_lib->vba.DETBufferSizeC[k] = mode_lib->vba.DETBufferSizeInKByte + mode_lib->vba.DETBufferSizeC[k] = mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 3; } } @@ -3819,7 +3819,7 @@ void dml20_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l mode_lib->vba.MaximumSwathWidthInDETBuffer = dml_min( mode_lib->vba.MaximumSwathWidthSupport, - mode_lib->vba.DETBufferSizeInKByte * 1024.0 / 2.0 + mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 2.0 / (locals->BytePerPixelInDETY[k] * locals->MinSwathHeightY[k] + locals->BytePerPixelInDETC[k] @@ -4322,7 +4322,7 @@ void dml20_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l locals->RoundedUpMaxSwathSizeBytesC = 0; } - if (locals->RoundedUpMaxSwathSizeBytesY + locals->RoundedUpMaxSwathSizeBytesC <= locals->DETBufferSizeInKByte * 1024 / 2) { + if (locals->RoundedUpMaxSwathSizeBytesY + locals->RoundedUpMaxSwathSizeBytesC <= locals->DETBufferSizeInKByte[0] * 1024 / 2) { locals->SwathHeightYPerState[i][j][k] = locals->MaxSwathHeightY[k]; locals->SwathHeightCPerState[i][j][k] = locals->MaxSwathHeightC[k]; } else { @@ -4331,15 +4331,15 @@ void dml20_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l } if (locals->BytePerPixelInDETC[k] == 0) { - locals->LinesInDETLuma = locals->DETBufferSizeInKByte * 1024 / locals->BytePerPixelInDETY[k] / locals->SwathWidthYPerState[i][j][k]; + locals->LinesInDETLuma = locals->DETBufferSizeInKByte[0] * 1024 / locals->BytePerPixelInDETY[k] / locals->SwathWidthYPerState[i][j][k]; locals->LinesInDETChroma = 0; } else if (locals->SwathHeightYPerState[i][j][k] <= locals->SwathHeightCPerState[i][j][k]) { - locals->LinesInDETLuma = locals->DETBufferSizeInKByte * 1024 / 2 / locals->BytePerPixelInDETY[k] / + locals->LinesInDETLuma = locals->DETBufferSizeInKByte[0] * 1024 / 2 / locals->BytePerPixelInDETY[k] / locals->SwathWidthYPerState[i][j][k]; - locals->LinesInDETChroma = locals->DETBufferSizeInKByte * 1024 / 2 / locals->BytePerPixelInDETC[k] / (locals->SwathWidthYPerState[i][j][k] / 2); + locals->LinesInDETChroma = locals->DETBufferSizeInKByte[0] * 1024 / 2 / locals->BytePerPixelInDETC[k] / (locals->SwathWidthYPerState[i][j][k] / 2); } else { - locals->LinesInDETLuma = locals->DETBufferSizeInKByte * 1024 * 2 / 3 / locals->BytePerPixelInDETY[k] / locals->SwathWidthYPerState[i][j][k]; - locals->LinesInDETChroma = locals->DETBufferSizeInKByte * 1024 / 3 / locals->BytePerPixelInDETY[k] / (locals->SwathWidthYPerState[i][j][k] / 2); + locals->LinesInDETLuma = locals->DETBufferSizeInKByte[0] * 1024 * 2 / 3 / locals->BytePerPixelInDETY[k] / locals->SwathWidthYPerState[i][j][k]; + locals->LinesInDETChroma = locals->DETBufferSizeInKByte[0] * 1024 / 3 / locals->BytePerPixelInDETY[k] / (locals->SwathWidthYPerState[i][j][k] / 2); } locals->EffectiveLBLatencyHidingSourceLinesLuma = dml_min(locals->MaxLineBufferLines, diff --git a/drivers/gpu/drm/amd/display/dc/dml/dcn20/display_mode_vba_20v2.c b/drivers/gpu/drm/amd/display/dc/dml/dcn20/display_mode_vba_20v2.c index 51098c2c9854..fbed5304692d 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/dcn20/display_mode_vba_20v2.c +++ b/drivers/gpu/drm/amd/display/dc/dml/dcn20/display_mode_vba_20v2.c @@ -2968,7 +2968,7 @@ static void dml20v2_DisplayPipeConfiguration(struct display_mode_lib *mode_lib) RoundedUpMaxSwathSizeBytesC = 0.0; if (RoundedUpMaxSwathSizeBytesY + RoundedUpMaxSwathSizeBytesC - <= mode_lib->vba.DETBufferSizeInKByte * 1024.0 / 2.0) { + <= mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 2.0) { mode_lib->vba.SwathHeightY[k] = MaximumSwathHeightY; mode_lib->vba.SwathHeightC[k] = MaximumSwathHeightC; } else { @@ -2977,17 +2977,17 @@ static void dml20v2_DisplayPipeConfiguration(struct display_mode_lib *mode_lib) } if (mode_lib->vba.SwathHeightC[k] == 0) { - mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte * 1024; + mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte[0] * 1024; mode_lib->vba.DETBufferSizeC[k] = 0; } else if (mode_lib->vba.SwathHeightY[k] <= mode_lib->vba.SwathHeightC[k]) { - mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte + mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 2; - mode_lib->vba.DETBufferSizeC[k] = mode_lib->vba.DETBufferSizeInKByte + mode_lib->vba.DETBufferSizeC[k] = mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 2; } else { - mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte + mode_lib->vba.DETBufferSizeY[k] = mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 * 2 / 3; - mode_lib->vba.DETBufferSizeC[k] = mode_lib->vba.DETBufferSizeInKByte + mode_lib->vba.DETBufferSizeC[k] = mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 3; } } @@ -3926,7 +3926,7 @@ void dml20v2_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode mode_lib->vba.MaximumSwathWidthInDETBuffer = dml_min( mode_lib->vba.MaximumSwathWidthSupport, - mode_lib->vba.DETBufferSizeInKByte * 1024.0 / 2.0 + mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 2.0 / (locals->BytePerPixelInDETY[k] * locals->MinSwathHeightY[k] + locals->BytePerPixelInDETC[k] @@ -4443,7 +4443,7 @@ void dml20v2_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode locals->RoundedUpMaxSwathSizeBytesC = 0; } - if (locals->RoundedUpMaxSwathSizeBytesY + locals->RoundedUpMaxSwathSizeBytesC <= locals->DETBufferSizeInKByte * 1024 / 2) { + if (locals->RoundedUpMaxSwathSizeBytesY + locals->RoundedUpMaxSwathSizeBytesC <= locals->DETBufferSizeInKByte[0] * 1024 / 2) { locals->SwathHeightYPerState[i][j][k] = locals->MaxSwathHeightY[k]; locals->SwathHeightCPerState[i][j][k] = locals->MaxSwathHeightC[k]; } else { @@ -4452,15 +4452,15 @@ void dml20v2_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode } if (locals->BytePerPixelInDETC[k] == 0) { - locals->LinesInDETLuma = locals->DETBufferSizeInKByte * 1024 / locals->BytePerPixelInDETY[k] / locals->SwathWidthYPerState[i][j][k]; + locals->LinesInDETLuma = locals->DETBufferSizeInKByte[0] * 1024 / locals->BytePerPixelInDETY[k] / locals->SwathWidthYPerState[i][j][k]; locals->LinesInDETChroma = 0; } else if (locals->SwathHeightYPerState[i][j][k] <= locals->SwathHeightCPerState[i][j][k]) { - locals->LinesInDETLuma = locals->DETBufferSizeInKByte * 1024 / 2 / locals->BytePerPixelInDETY[k] / + locals->LinesInDETLuma = locals->DETBufferSizeInKByte[0] * 1024 / 2 / locals->BytePerPixelInDETY[k] / locals->SwathWidthYPerState[i][j][k]; - locals->LinesInDETChroma = locals->DETBufferSizeInKByte * 1024 / 2 / locals->BytePerPixelInDETC[k] / (locals->SwathWidthYPerState[i][j][k] / 2); + locals->LinesInDETChroma = locals->DETBufferSizeInKByte[0] * 1024 / 2 / locals->BytePerPixelInDETC[k] / (locals->SwathWidthYPerState[i][j][k] / 2); } else { - locals->LinesInDETLuma = locals->DETBufferSizeInKByte * 1024 * 2 / 3 / locals->BytePerPixelInDETY[k] / locals->SwathWidthYPerState[i][j][k]; - locals->LinesInDETChroma = locals->DETBufferSizeInKByte * 1024 / 3 / locals->BytePerPixelInDETY[k] / (locals->SwathWidthYPerState[i][j][k] / 2); + locals->LinesInDETLuma = locals->DETBufferSizeInKByte[0] * 1024 * 2 / 3 / locals->BytePerPixelInDETY[k] / locals->SwathWidthYPerState[i][j][k]; + locals->LinesInDETChroma = locals->DETBufferSizeInKByte[0] * 1024 / 3 / locals->BytePerPixelInDETY[k] / (locals->SwathWidthYPerState[i][j][k] / 2); } locals->EffectiveLBLatencyHidingSourceLinesLuma = dml_min(locals->MaxLineBufferLines, diff --git a/drivers/gpu/drm/amd/display/dc/dml/dcn21/display_mode_vba_21.c b/drivers/gpu/drm/amd/display/dc/dml/dcn21/display_mode_vba_21.c index 398210d1af34..c26e742e8137 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/dcn21/display_mode_vba_21.c +++ b/drivers/gpu/drm/amd/display/dc/dml/dcn21/display_mode_vba_21.c @@ -148,7 +148,7 @@ static double CalculateDCCConfiguration( bool DCCProgrammingAssumesScanDirectionUnknown, unsigned int ViewportWidth, unsigned int ViewportHeight, - double DETBufferSize, + unsigned int DETBufferSize, unsigned int RequestHeight256Byte, unsigned int SwathHeight, enum dm_swizzle_mode TilingFormat, @@ -289,7 +289,7 @@ static void CalculateWatermarksAndDRAMSpeedChangeSupport( unsigned int MaxLineBufferLines, unsigned int LineBufferSize, unsigned int DPPOutputBufferPixels, - double DETBufferSizeInKByte, + unsigned int DETBufferSizeInKByte, unsigned int WritebackInterfaceLumaBufferSize, unsigned int WritebackInterfaceChromaBufferSize, double DCFCLK, @@ -354,11 +354,11 @@ static void CalculateDCFCLKDeepSleep( double DPPCLK[], double *DCFCLKDeepSleep); static void CalculateDETBufferSize( - double DETBufferSizeInKByte, + unsigned int DETBufferSizeInKByte, unsigned int SwathHeightY, unsigned int SwathHeightC, - double *DETBufferSizeY, - double *DETBufferSizeC); + unsigned int *DETBufferSizeY, + unsigned int *DETBufferSizeC); static void CalculateUrgentBurstFactor( unsigned int DETBufferSizeInKByte, unsigned int SwathHeightY, @@ -1074,7 +1074,7 @@ static double CalculateDCCConfiguration( bool DCCProgrammingAssumesScanDirectionUnknown, unsigned int ViewportWidth, unsigned int ViewportHeight, - double DETBufferSize, + unsigned int DETBufferSize, unsigned int RequestHeight256Byte, unsigned int SwathHeight, enum dm_swizzle_mode TilingFormat, @@ -2246,7 +2246,7 @@ static void DISPCLKDPPCLKDCFCLKDeepSleepPrefetchParametersWatermarksAndPerforman } CalculateUrgentBurstFactor( - mode_lib->vba.DETBufferSizeInKByte, + mode_lib->vba.DETBufferSizeInKByte[0], mode_lib->vba.SwathHeightY[k], mode_lib->vba.SwathHeightC[k], locals->SwathWidthY[k], @@ -2415,7 +2415,7 @@ static void DISPCLKDPPCLKDCFCLKDeepSleepPrefetchParametersWatermarksAndPerforman mode_lib->vba.MaxLineBufferLines, mode_lib->vba.LineBufferSize, mode_lib->vba.DPPOutputBufferPixels, - mode_lib->vba.DETBufferSizeInKByte, + mode_lib->vba.DETBufferSizeInKByte[0], mode_lib->vba.WritebackInterfaceLumaBufferSize, mode_lib->vba.WritebackInterfaceChromaBufferSize, mode_lib->vba.DCFCLK, @@ -2588,7 +2588,7 @@ static void DISPCLKDPPCLKDCFCLKDeepSleepPrefetchParametersWatermarksAndPerforman false, // We should always know the direction DCCProgrammingAssumesScanDirectionUnknown, mode_lib->vba.ViewportWidth[k], mode_lib->vba.ViewportHeight[k], - mode_lib->vba.DETBufferSizeInKByte * 1024, + mode_lib->vba.DETBufferSizeInKByte[0] * 1024, locals->BlockHeight256BytesY[k], mode_lib->vba.SwathHeightY[k], mode_lib->vba.SurfaceTiling[k], @@ -2689,13 +2689,13 @@ static void DISPCLKDPPCLKDCFCLKDeepSleepPrefetchParametersWatermarksAndPerforman // Stutter Efficiency for (k = 0; k < mode_lib->vba.NumberOfActivePlanes; ++k) { CalculateDETBufferSize( - mode_lib->vba.DETBufferSizeInKByte, + mode_lib->vba.DETBufferSizeInKByte[0], mode_lib->vba.SwathHeightY[k], mode_lib->vba.SwathHeightC[k], &locals->DETBufferSizeY[k], &locals->DETBufferSizeC[k]); - locals->LinesInDETY[k] = locals->DETBufferSizeY[k] + locals->LinesInDETY[k] = (double)locals->DETBufferSizeY[k] / locals->BytePerPixelDETY[k] / locals->SwathWidthY[k]; locals->LinesInDETYRoundedDownToSwath[k] = dml_floor( locals->LinesInDETY[k], @@ -2984,7 +2984,7 @@ static void DisplayPipeConfiguration(struct display_mode_lib *mode_lib) RoundedUpMaxSwathSizeBytesC = 0.0; if (RoundedUpMaxSwathSizeBytesY + RoundedUpMaxSwathSizeBytesC - <= mode_lib->vba.DETBufferSizeInKByte * 1024.0 / 2.0) { + <= mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 2.0) { mode_lib->vba.SwathHeightY[k] = MaximumSwathHeightY; mode_lib->vba.SwathHeightC[k] = MaximumSwathHeightC; } else { @@ -2993,7 +2993,7 @@ static void DisplayPipeConfiguration(struct display_mode_lib *mode_lib) } CalculateDETBufferSize( - mode_lib->vba.DETBufferSizeInKByte, + mode_lib->vba.DETBufferSizeInKByte[0], mode_lib->vba.SwathHeightY[k], mode_lib->vba.SwathHeightC[k], &mode_lib->vba.DETBufferSizeY[k], @@ -3888,7 +3888,7 @@ void dml21_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l mode_lib->vba.MaximumSwathWidthInDETBuffer = dml_min( mode_lib->vba.MaximumSwathWidthSupport, - mode_lib->vba.DETBufferSizeInKByte * 1024.0 / 2.0 + mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 2.0 / (locals->BytePerPixelInDETY[k] * locals->MinSwathHeightY[k] + locals->BytePerPixelInDETC[k] @@ -4437,7 +4437,7 @@ void dml21_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l mode_lib->vba.RoundedUpMaxSwathSizeBytesC = 0.0; } if (mode_lib->vba.RoundedUpMaxSwathSizeBytesY + mode_lib->vba.RoundedUpMaxSwathSizeBytesC - <= mode_lib->vba.DETBufferSizeInKByte * 1024.0 / 2.0) { + <= mode_lib->vba.DETBufferSizeInKByte[0] * 1024.0 / 2.0) { locals->SwathHeightYThisState[k] = locals->MaxSwathHeightY[k]; locals->SwathHeightCThisState[k] = locals->MaxSwathHeightC[k]; } else { @@ -4801,7 +4801,7 @@ void dml21_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l } CalculateUrgentBurstFactor( - mode_lib->vba.DETBufferSizeInKByte, + mode_lib->vba.DETBufferSizeInKByte[0], locals->SwathHeightYThisState[k], locals->SwathHeightCThisState[k], locals->SwathWidthYThisState[k], @@ -4975,7 +4975,7 @@ void dml21_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l mode_lib->vba.MaxLineBufferLines, mode_lib->vba.LineBufferSize, mode_lib->vba.DPPOutputBufferPixels, - mode_lib->vba.DETBufferSizeInKByte, + mode_lib->vba.DETBufferSizeInKByte[0], mode_lib->vba.WritebackInterfaceLumaBufferSize, mode_lib->vba.WritebackInterfaceChromaBufferSize, mode_lib->vba.DCFCLKPerState[i], @@ -5230,7 +5230,7 @@ static void CalculateWatermarksAndDRAMSpeedChangeSupport( unsigned int MaxLineBufferLines, unsigned int LineBufferSize, unsigned int DPPOutputBufferPixels, - double DETBufferSizeInKByte, + unsigned int DETBufferSizeInKByte, unsigned int WritebackInterfaceLumaBufferSize, unsigned int WritebackInterfaceChromaBufferSize, double DCFCLK, @@ -5285,8 +5285,8 @@ static void CalculateWatermarksAndDRAMSpeedChangeSupport( double EffectiveLBLatencyHidingC; double DPPOutputBufferLinesY; double DPPOutputBufferLinesC; - double DETBufferSizeY; - double DETBufferSizeC; + unsigned int DETBufferSizeY; + unsigned int DETBufferSizeC; double LinesInDETY[DC__NUM_DPP__MAX]; double LinesInDETC; unsigned int LinesInDETYRoundedDownToSwath[DC__NUM_DPP__MAX]; @@ -5382,12 +5382,12 @@ static void CalculateWatermarksAndDRAMSpeedChangeSupport( &DETBufferSizeY, &DETBufferSizeC); - LinesInDETY[k] = DETBufferSizeY / BytePerPixelDETY[k] / SwathWidthY[k]; + LinesInDETY[k] = (double)DETBufferSizeY / BytePerPixelDETY[k] / SwathWidthY[k]; LinesInDETYRoundedDownToSwath[k] = dml_floor(LinesInDETY[k], SwathHeightY[k]); FullDETBufferingTimeY[k] = LinesInDETYRoundedDownToSwath[k] * (HTotal[k] / PixelClock[k]) / VRatio[k]; if (BytePerPixelDETC[k] > 0) { - LinesInDETC = DETBufferSizeC / BytePerPixelDETC[k] / (SwathWidthY[k] / 2.0); + LinesInDETC = (double)DETBufferSizeC / BytePerPixelDETC[k] / (SwathWidthY[k] / 2.0); LinesInDETCRoundedDownToSwath = dml_floor(LinesInDETC, SwathHeightC[k]); FullDETBufferingTimeC = LinesInDETCRoundedDownToSwath * (HTotal[k] / PixelClock[k]) / (VRatio[k] / 2); @@ -5574,11 +5574,11 @@ static void CalculateDCFCLKDeepSleep( } static void CalculateDETBufferSize( - double DETBufferSizeInKByte, + unsigned int DETBufferSizeInKByte, unsigned int SwathHeightY, unsigned int SwathHeightC, - double *DETBufferSizeY, - double *DETBufferSizeC) + unsigned int *DETBufferSizeY, + unsigned int *DETBufferSizeC) { if (SwathHeightC == 0) { *DETBufferSizeY = DETBufferSizeInKByte * 1024; @@ -5625,8 +5625,8 @@ static void CalculateUrgentBurstFactor( double DETBufferSizeInTimeLumaPre; double DETBufferSizeInTimeChroma; double DETBufferSizeInTimeChromaPre; - double DETBufferSizeY; - double DETBufferSizeC; + unsigned int DETBufferSizeY; + unsigned int DETBufferSizeC; *NotEnoughUrgentLatencyHiding = 0; *NotEnoughUrgentLatencyHidingPre = 0; @@ -5663,7 +5663,7 @@ static void CalculateUrgentBurstFactor( &DETBufferSizeY, &DETBufferSizeC); - LinesInDETLuma = DETBufferSizeY / BytePerPixelInDETY / SwathWidthY; + LinesInDETLuma = (double)DETBufferSizeY / BytePerPixelInDETY / SwathWidthY; DETBufferSizeInTimeLuma = dml_floor(LinesInDETLuma, SwathHeightY) * LineTime / VRatio; if (DETBufferSizeInTimeLuma - UrgentLatency <= 0) { *NotEnoughUrgentLatencyHiding = 1; @@ -5687,7 +5687,7 @@ static void CalculateUrgentBurstFactor( } if (BytePerPixelInDETC > 0) { - LinesInDETChroma = DETBufferSizeC / BytePerPixelInDETC / (SwathWidthY / 2); + LinesInDETChroma = (double)DETBufferSizeC / BytePerPixelInDETC / (SwathWidthY / 2); DETBufferSizeInTimeChroma = dml_floor(LinesInDETChroma, SwathHeightC) * LineTime / (VRatio / 2); if (DETBufferSizeInTimeChroma - UrgentLatency <= 0) { diff --git a/drivers/gpu/drm/amd/display/dc/dml/dcn30/display_mode_vba_30.c b/drivers/gpu/drm/amd/display/dc/dml/dcn30/display_mode_vba_30.c index cb3f70a71b51..ec56210b6180 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/dcn30/display_mode_vba_30.c +++ b/drivers/gpu/drm/amd/display/dc/dml/dcn30/display_mode_vba_30.c @@ -299,7 +299,7 @@ static void CalculateWatermarksAndDRAMSpeedChangeSupport( unsigned int MaxLineBufferLines, unsigned int LineBufferSize, unsigned int DPPOutputBufferPixels, - double DETBufferSizeInKByte, + unsigned int DETBufferSizeInKByte, unsigned int WritebackInterfaceBufferSize, double DCFCLK, double ReturnBW, @@ -318,8 +318,8 @@ static void CalculateWatermarksAndDRAMSpeedChangeSupport( unsigned int DPPPerPlane[], bool DCCEnable[], double DPPCLK[], - double DETBufferSizeY[], - double DETBufferSizeC[], + unsigned int DETBufferSizeY[], + unsigned int DETBufferSizeC[], unsigned int SwathHeightY[], unsigned int SwathHeightC[], unsigned int LBBitPerPixel[], @@ -570,7 +570,7 @@ static void CalculateStutterEfficiency( double SRExitTime, bool SynchronizedVBlank, int DPPPerPlane[], - double DETBufferSizeY[], + unsigned int DETBufferSizeY[], int BytePerPixelY[], double BytePerPixelDETY[], double SwathWidthY[], @@ -603,7 +603,7 @@ static void CalculateStutterEfficiency( static void CalculateSwathAndDETConfiguration( bool ForceSingleDPP, int NumberOfActivePlanes, - long DETBufferSizeInKByte, + unsigned int DETBufferSizeInKByte, double MaximumSwathWidthLuma[], double MaximumSwathWidthChroma[], enum scan_direction_class SourceScan[], @@ -635,8 +635,8 @@ static void CalculateSwathAndDETConfiguration( double SwathWidthChroma[], int SwathHeightY[], int SwathHeightC[], - double DETBufferSizeY[], - double DETBufferSizeC[], + unsigned int DETBufferSizeY[], + unsigned int DETBufferSizeC[], bool ViewportSizeSupportPerPlane[], bool *ViewportSizeSupport); static void CalculateSwathWidth( @@ -2613,7 +2613,7 @@ static void DISPCLKDPPCLKDCFCLKDeepSleepPrefetchParametersWatermarksAndPerforman CalculateUrgentBurstFactor( v->swath_width_luma_ub[k], v->swath_width_chroma_ub[k], - v->DETBufferSizeInKByte, + v->DETBufferSizeInKByte[0], v->SwathHeightY[k], v->SwathHeightC[k], v->HTotal[k] / v->PixelClock[k], @@ -2635,7 +2635,7 @@ static void DISPCLKDPPCLKDCFCLKDeepSleepPrefetchParametersWatermarksAndPerforman CalculateUrgentBurstFactor( v->swath_width_luma_ub[k], v->swath_width_chroma_ub[k], - v->DETBufferSizeInKByte, + v->DETBufferSizeInKByte[0], v->SwathHeightY[k], v->SwathHeightC[k], v->HTotal[k] / v->PixelClock[k], @@ -2808,7 +2808,7 @@ static void DISPCLKDPPCLKDCFCLKDeepSleepPrefetchParametersWatermarksAndPerforman v->MaxLineBufferLines, v->LineBufferSize, v->DPPOutputBufferPixels, - v->DETBufferSizeInKByte, + v->DETBufferSizeInKByte[0], v->WritebackInterfaceBufferSize, v->DCFCLK, v->ReturnBW, @@ -3027,7 +3027,7 @@ static void DISPCLKDPPCLKDCFCLKDeepSleepPrefetchParametersWatermarksAndPerforman v->SurfaceWidthC[k], v->SurfaceHeightY[k], v->SurfaceHeightC[k], - v->DETBufferSizeInKByte * 1024, + v->DETBufferSizeInKByte[0] * 1024, v->BlockHeight256BytesY[k], v->BlockHeight256BytesC[k], v->SurfaceTiling[k], @@ -3177,7 +3177,7 @@ static void DisplayPipeConfiguration(struct display_mode_lib *mode_lib) CalculateSwathAndDETConfiguration( false, mode_lib->vba.NumberOfActivePlanes, - mode_lib->vba.DETBufferSizeInKByte, + mode_lib->vba.DETBufferSizeInKByte[0], dummy1, dummy2, mode_lib->vba.SourceScan, @@ -3911,7 +3911,7 @@ void dml30_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l CalculateSwathAndDETConfiguration( true, v->NumberOfActivePlanes, - v->DETBufferSizeInKByte, + v->DETBufferSizeInKByte[0], v->MaximumSwathWidthLuma, v->MaximumSwathWidthChroma, v->SourceScan, @@ -4399,7 +4399,7 @@ void dml30_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l CalculateSwathAndDETConfiguration( false, v->NumberOfActivePlanes, - v->DETBufferSizeInKByte, + v->DETBufferSizeInKByte[0], v->MaximumSwathWidthLuma, v->MaximumSwathWidthChroma, v->SourceScan, @@ -4622,7 +4622,7 @@ void dml30_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l CalculateUrgentBurstFactor( v->swath_width_luma_ub_this_state[k], v->swath_width_chroma_ub_this_state[k], - v->DETBufferSizeInKByte, + v->DETBufferSizeInKByte[0], v->SwathHeightYThisState[k], v->SwathHeightCThisState[k], v->HTotal[k] / v->PixelClock[k], @@ -5025,7 +5025,7 @@ void dml30_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l CalculateUrgentBurstFactor( v->swath_width_luma_ub_this_state[k], v->swath_width_chroma_ub_this_state[k], - v->DETBufferSizeInKByte, + v->DETBufferSizeInKByte[0], v->SwathHeightYThisState[k], v->SwathHeightCThisState[k], v->HTotal[k] / v->PixelClock[k], @@ -5197,7 +5197,7 @@ void dml30_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l v->MaxLineBufferLines, v->LineBufferSize, v->DPPOutputBufferPixels, - v->DETBufferSizeInKByte, + v->DETBufferSizeInKByte[0], v->WritebackInterfaceBufferSize, v->DCFCLKState[i][j], v->ReturnBWPerState[i][j], @@ -5369,7 +5369,7 @@ static void CalculateWatermarksAndDRAMSpeedChangeSupport( unsigned int MaxLineBufferLines, unsigned int LineBufferSize, unsigned int DPPOutputBufferPixels, - double DETBufferSizeInKByte, + unsigned int DETBufferSizeInKByte, unsigned int WritebackInterfaceBufferSize, double DCFCLK, double ReturnBW, @@ -5388,8 +5388,8 @@ static void CalculateWatermarksAndDRAMSpeedChangeSupport( unsigned int DPPPerPlane[], bool DCCEnable[], double DPPCLK[], - double DETBufferSizeY[], - double DETBufferSizeC[], + unsigned int DETBufferSizeY[], + unsigned int DETBufferSizeC[], unsigned int SwathHeightY[], unsigned int SwathHeightC[], unsigned int LBBitPerPixel[], @@ -6126,7 +6126,7 @@ static void CalculateStutterEfficiency( double SRExitTime, bool SynchronizedVBlank, int DPPPerPlane[], - double DETBufferSizeY[], + unsigned int DETBufferSizeY[], int BytePerPixelY[], double BytePerPixelDETY[], double SwathWidthY[], @@ -6273,7 +6273,7 @@ static void CalculateStutterEfficiency( static void CalculateSwathAndDETConfiguration( bool ForceSingleDPP, int NumberOfActivePlanes, - long DETBufferSizeInKByte, + unsigned int DETBufferSizeInKByte, double MaximumSwathWidthLuma[], double MaximumSwathWidthChroma[], enum scan_direction_class SourceScan[], @@ -6305,8 +6305,8 @@ static void CalculateSwathAndDETConfiguration( double SwathWidthChroma[], int SwathHeightY[], int SwathHeightC[], - double DETBufferSizeY[], - double DETBufferSizeC[], + unsigned int DETBufferSizeY[], + unsigned int DETBufferSizeC[], bool ViewportSizeSupportPerPlane[], bool *ViewportSizeSupport) { diff --git a/drivers/gpu/drm/amd/display/dc/dml/display_mode_vba.c b/drivers/gpu/drm/amd/display/dc/dml/display_mode_vba.c index 2a967458065b..d764d784e279 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/display_mode_vba.c +++ b/drivers/gpu/drm/amd/display/dc/dml/display_mode_vba.c @@ -299,7 +299,7 @@ static void fetch_ip_params(struct display_mode_lib *mode_lib) mode_lib->vba.MaxDCHUBToPSCLThroughput = ip->max_dchub_pscl_bw_pix_per_clk; mode_lib->vba.MaxPSCLToLBThroughput = ip->max_pscl_lb_bw_pix_per_clk; mode_lib->vba.ROBBufferSizeInKByte = ip->rob_buffer_size_kbytes; - mode_lib->vba.DETBufferSizeInKByte = ip->det_buffer_size_kbytes; + mode_lib->vba.DETBufferSizeInKByte[0] = ip->det_buffer_size_kbytes; mode_lib->vba.PixelChunkSizeInKByte = ip->pixel_chunk_size_kbytes; mode_lib->vba.MetaChunkSize = ip->meta_chunk_size_kbytes; diff --git a/drivers/gpu/drm/amd/display/dc/dml/display_mode_vba.h b/drivers/gpu/drm/amd/display/dc/dml/display_mode_vba.h index 025aa5bd8ea0..86db86b7153e 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/display_mode_vba.h +++ b/drivers/gpu/drm/amd/display/dc/dml/display_mode_vba.h @@ -233,7 +233,7 @@ struct vba_vars_st { // IP Parameters // unsigned int ROBBufferSizeInKByte; - double DETBufferSizeInKByte; + unsigned int DETBufferSizeInKByte[DC__NUM_DPP__MAX]; double DETBufferSizeInTime; unsigned int DPPOutputBufferPixels; unsigned int OPPOutputBufferLines; @@ -351,8 +351,8 @@ struct vba_vars_st { // Intermediates/Informational bool ImmediateFlipSupport; - double DETBufferSizeY[DC__NUM_DPP__MAX]; - double DETBufferSizeC[DC__NUM_DPP__MAX]; + unsigned int DETBufferSizeY[DC__NUM_DPP__MAX]; + unsigned int DETBufferSizeC[DC__NUM_DPP__MAX]; unsigned int SwathHeightY[DC__NUM_DPP__MAX]; unsigned int SwathHeightC[DC__NUM_DPP__MAX]; unsigned int LBBitPerPixel[DC__NUM_DPP__MAX]; @@ -631,8 +631,8 @@ struct vba_vars_st { enum odm_combine_mode odm_combine_dummy[DC__NUM_DPP__MAX]; double dummy1[DC__NUM_DPP__MAX]; double dummy2[DC__NUM_DPP__MAX]; - double dummy3[DC__NUM_DPP__MAX]; - double dummy4[DC__NUM_DPP__MAX]; + unsigned int dummy3[DC__NUM_DPP__MAX]; + unsigned int dummy4[DC__NUM_DPP__MAX]; double dummy5; double dummy6; double dummy7[DC__NUM_DPP__MAX]; @@ -872,8 +872,8 @@ struct vba_vars_st { int PercentMarginOverMinimumRequiredDCFCLK; bool DynamicMetadataSupported[DC__VOLTAGE_STATES][2]; enum immediate_flip_requirement ImmediateFlipRequirement; - double DETBufferSizeYThisState[DC__NUM_DPP__MAX]; - double DETBufferSizeCThisState[DC__NUM_DPP__MAX]; + unsigned int DETBufferSizeYThisState[DC__NUM_DPP__MAX]; + unsigned int DETBufferSizeCThisState[DC__NUM_DPP__MAX]; bool NoUrgentLatencyHiding[DC__NUM_DPP__MAX]; bool NoUrgentLatencyHidingPre[DC__NUM_DPP__MAX]; int swath_width_luma_ub_this_state[DC__NUM_DPP__MAX]; diff --git a/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c b/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c index be57088d185d..f403d8e84a8c 100644 --- a/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c +++ b/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c @@ -37,6 +37,8 @@ static uint32_t dsc_policy_max_target_bpp_limit = 16; /* default DSC policy enables DSC only when needed */ static bool dsc_policy_enable_dsc_when_not_needed; +static bool dsc_policy_disable_dsc_stream_overhead; + static bool dsc_buff_block_size_from_dpcd(int dpcd_buff_block_size, int *buff_block_size) { @@ -250,6 +252,7 @@ static bool intersect_dsc_caps( if (pixel_encoding == PIXEL_ENCODING_YCBCR422 || pixel_encoding == PIXEL_ENCODING_YCBCR420) dsc_common_caps->bpp_increment_div = min(dsc_common_caps->bpp_increment_div, (uint32_t)8); + dsc_common_caps->is_dp = dsc_sink_caps->is_dp; return true; } @@ -258,12 +261,63 @@ static inline uint32_t dsc_div_by_10_round_up(uint32_t value) return (value + 9) / 10; } +static struct fixed31_32 compute_dsc_max_bandwidth_overhead( + const struct dc_crtc_timing *timing, + const int num_slices_h, + const bool is_dp) +{ + struct fixed31_32 max_dsc_overhead; + struct fixed31_32 refresh_rate; + + if (dsc_policy_disable_dsc_stream_overhead || !is_dp) + return dc_fixpt_from_int(0); + + /* use target bpp that can take entire target bandwidth */ + refresh_rate = dc_fixpt_from_int(timing->pix_clk_100hz); + refresh_rate = dc_fixpt_div_int(refresh_rate, timing->h_total); + refresh_rate = dc_fixpt_div_int(refresh_rate, timing->v_total); + refresh_rate = dc_fixpt_mul_int(refresh_rate, 100); + + max_dsc_overhead = dc_fixpt_from_int(num_slices_h); + max_dsc_overhead = dc_fixpt_mul_int(max_dsc_overhead, timing->v_total); + max_dsc_overhead = dc_fixpt_mul_int(max_dsc_overhead, 256); + max_dsc_overhead = dc_fixpt_div_int(max_dsc_overhead, 1000); + max_dsc_overhead = dc_fixpt_mul(max_dsc_overhead, refresh_rate); + + return max_dsc_overhead; +} + +static uint32_t compute_bpp_x16_from_target_bandwidth( + const uint32_t bandwidth_in_kbps, + const struct dc_crtc_timing *timing, + const uint32_t num_slices_h, + const uint32_t bpp_increment_div, + const bool is_dp) +{ + struct fixed31_32 overhead_in_kbps; + struct fixed31_32 effective_bandwidth_in_kbps; + struct fixed31_32 bpp_x16; + + overhead_in_kbps = compute_dsc_max_bandwidth_overhead( + timing, num_slices_h, is_dp); + effective_bandwidth_in_kbps = dc_fixpt_from_int(bandwidth_in_kbps); + effective_bandwidth_in_kbps = dc_fixpt_sub(effective_bandwidth_in_kbps, + overhead_in_kbps); + bpp_x16 = dc_fixpt_mul_int(effective_bandwidth_in_kbps, 10); + bpp_x16 = dc_fixpt_div_int(bpp_x16, timing->pix_clk_100hz); + bpp_x16 = dc_fixpt_from_int(dc_fixpt_floor(dc_fixpt_mul_int(bpp_x16, bpp_increment_div))); + bpp_x16 = dc_fixpt_div_int(bpp_x16, bpp_increment_div); + bpp_x16 = dc_fixpt_mul_int(bpp_x16, 16); + return dc_fixpt_floor(bpp_x16); +} + /* Get DSC bandwidth range based on [min_bpp, max_bpp] target bitrate range, and timing's pixel clock * and uncompressed bandwidth. */ static void get_dsc_bandwidth_range( const uint32_t min_bpp_x16, const uint32_t max_bpp_x16, + const uint32_t num_slices_h, const struct dsc_enc_caps *dsc_caps, const struct dc_crtc_timing *timing, struct dc_dsc_bw_range *range) @@ -272,16 +326,21 @@ static void get_dsc_bandwidth_range( range->stream_kbps = dc_bandwidth_in_kbps_from_timing(timing); /* max dsc target bpp */ - range->max_kbps = dc_dsc_stream_bandwidth_in_kbps(timing->pix_clk_100hz, max_bpp_x16); + range->max_kbps = dc_dsc_stream_bandwidth_in_kbps(timing, + max_bpp_x16, num_slices_h, dsc_caps->is_dp); range->max_target_bpp_x16 = max_bpp_x16; if (range->max_kbps > range->stream_kbps) { /* max dsc target bpp is capped to native bandwidth */ range->max_kbps = range->stream_kbps; - range->max_target_bpp_x16 = calc_dsc_bpp_x16(range->stream_kbps, timing->pix_clk_100hz, dsc_caps->bpp_increment_div); + range->max_target_bpp_x16 = compute_bpp_x16_from_target_bandwidth( + range->max_kbps, timing, num_slices_h, + dsc_caps->bpp_increment_div, + dsc_caps->is_dp); } /* min dsc target bpp */ - range->min_kbps = dc_dsc_stream_bandwidth_in_kbps(timing->pix_clk_100hz, min_bpp_x16); + range->min_kbps = dc_dsc_stream_bandwidth_in_kbps(timing, + min_bpp_x16, num_slices_h, dsc_caps->is_dp); range->min_target_bpp_x16 = min_bpp_x16; if (range->min_kbps > range->max_kbps) { /* min dsc target bpp is capped to max dsc bandwidth*/ @@ -290,7 +349,6 @@ static void get_dsc_bandwidth_range( } } - /* Decides if DSC should be used and calculates target bpp if it should, applying DSC policy. * * Returns: @@ -303,6 +361,7 @@ static bool decide_dsc_target_bpp_x16( const struct dsc_enc_caps *dsc_common_caps, const int target_bandwidth_kbps, const struct dc_crtc_timing *timing, + const int num_slices_h, int *target_bpp_x16) { bool should_use_dsc = false; @@ -311,7 +370,7 @@ static bool decide_dsc_target_bpp_x16( memset(&range, 0, sizeof(range)); get_dsc_bandwidth_range(policy->min_target_bpp * 16, policy->max_target_bpp * 16, - dsc_common_caps, timing, &range); + num_slices_h, dsc_common_caps, timing, &range); if (!policy->enable_dsc_when_not_needed && target_bandwidth_kbps >= range.stream_kbps) { /* enough bandwidth without dsc */ *target_bpp_x16 = 0; @@ -327,7 +386,10 @@ static bool decide_dsc_target_bpp_x16( should_use_dsc = true; } else if (target_bandwidth_kbps >= range.min_kbps) { /* use target bpp that can take entire target bandwidth */ - *target_bpp_x16 = calc_dsc_bpp_x16(target_bandwidth_kbps, timing->pix_clk_100hz, dsc_common_caps->bpp_increment_div); + *target_bpp_x16 = compute_bpp_x16_from_target_bandwidth( + target_bandwidth_kbps, timing, num_slices_h, + dsc_common_caps->bpp_increment_div, + dsc_common_caps->is_dp); should_use_dsc = true; } else { /* not enough bandwidth to fulfill minimum requirement */ @@ -531,18 +593,6 @@ static bool setup_dsc_config( if (!is_dsc_possible) goto done; - if (target_bandwidth_kbps > 0) { - is_dsc_possible = decide_dsc_target_bpp_x16( - &policy, - &dsc_common_caps, - target_bandwidth_kbps, - timing, - &target_bpp); - dsc_cfg->bits_per_pixel = target_bpp; - } - if (!is_dsc_possible) - goto done; - sink_per_slice_throughput_mps = 0; // Validate available DSC settings against the mode timing @@ -690,12 +740,26 @@ static bool setup_dsc_config( dsc_cfg->num_slices_v = pic_height/slice_height; + if (target_bandwidth_kbps > 0) { + is_dsc_possible = decide_dsc_target_bpp_x16( + &policy, + &dsc_common_caps, + target_bandwidth_kbps, + timing, + num_slices_h, + &target_bpp); + dsc_cfg->bits_per_pixel = target_bpp; + } + if (!is_dsc_possible) + goto done; + // Final decission: can we do DSC or not? if (is_dsc_possible) { // Fill out the rest of DSC settings dsc_cfg->block_pred_enable = dsc_common_caps.is_block_pred_supported; dsc_cfg->linebuf_depth = dsc_common_caps.lb_bit_depth; dsc_cfg->version_minor = (dsc_common_caps.dsc_version & 0xf0) >> 4; + dsc_cfg->is_dp = dsc_sink_caps->is_dp; } done: @@ -806,6 +870,7 @@ bool dc_dsc_parse_dsc_dpcd(const struct dc *dc, const uint8_t *dpcd_dsc_basic_da dsc_sink_caps->branch_max_line_width = dpcd_dsc_branch_decoder_caps[DP_DSC_BRANCH_MAX_LINE_WIDTH - DP_DSC_BRANCH_OVERALL_THROUGHPUT_0] * 320; ASSERT(dsc_sink_caps->branch_max_line_width == 0 || dsc_sink_caps->branch_max_line_width >= 5120); + dsc_sink_caps->is_dp = true; return true; } @@ -838,7 +903,8 @@ bool dc_dsc_compute_bandwidth_range( dsc_min_slice_height_override, max_bpp_x16, &config); if (is_dsc_possible) - get_dsc_bandwidth_range(min_bpp_x16, max_bpp_x16, &dsc_common_caps, timing, range); + get_dsc_bandwidth_range(min_bpp_x16, max_bpp_x16, + config.num_slices_h, &dsc_common_caps, timing, range); return is_dsc_possible; } @@ -864,13 +930,20 @@ bool dc_dsc_compute_config( return is_dsc_possible; } -uint32_t dc_dsc_stream_bandwidth_in_kbps(uint32_t pix_clk_100hz, uint32_t bpp_x16) +uint32_t dc_dsc_stream_bandwidth_in_kbps(const struct dc_crtc_timing *timing, + uint32_t bpp_x16, uint32_t num_slices_h, bool is_dp) { - struct fixed31_32 link_bw_kbps; - link_bw_kbps = dc_fixpt_from_int(pix_clk_100hz); - link_bw_kbps = dc_fixpt_div_int(link_bw_kbps, 160); - link_bw_kbps = dc_fixpt_mul_int(link_bw_kbps, bpp_x16); - return dc_fixpt_ceil(link_bw_kbps); + struct fixed31_32 overhead_in_kbps; + struct fixed31_32 bpp; + struct fixed31_32 actual_bandwidth_in_kbps; + + overhead_in_kbps = compute_dsc_max_bandwidth_overhead( + timing, num_slices_h, is_dp); + bpp = dc_fixpt_from_fraction(bpp_x16, 16); + actual_bandwidth_in_kbps = dc_fixpt_from_fraction(timing->pix_clk_100hz, 10); + actual_bandwidth_in_kbps = dc_fixpt_mul(actual_bandwidth_in_kbps, bpp); + actual_bandwidth_in_kbps = dc_fixpt_add(actual_bandwidth_in_kbps, overhead_in_kbps); + return dc_fixpt_ceil(actual_bandwidth_in_kbps); } void dc_dsc_get_policy_for_timing(const struct dc_crtc_timing *timing, uint32_t max_target_bpp_limit_override_x16, struct dc_dsc_policy *policy) @@ -954,3 +1027,8 @@ void dc_dsc_policy_set_enable_dsc_when_not_needed(bool enable) { dsc_policy_enable_dsc_when_not_needed = enable; } + +void dc_dsc_policy_set_disable_dsc_stream_overhead(bool disable) +{ + dsc_policy_disable_dsc_stream_overhead = disable; +} diff --git a/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.c b/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.c index c6a1cd80aeae..7b294f637881 100644 --- a/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.c +++ b/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.c @@ -284,26 +284,6 @@ static u32 _do_bytes_per_pixel_calc(int slice_width, u16 drm_bpp, return bytes_per_pixel; } -static u32 _do_calc_dsc_bpp_x16(u32 stream_bandwidth_kbps, u32 pix_clk_100hz, - u32 bpp_increment_div) -{ - u32 dsc_target_bpp_x16; - float f_dsc_target_bpp; - float f_stream_bandwidth_100bps; - // bpp_increment_div is actually precision - u32 precision = bpp_increment_div; - - f_stream_bandwidth_100bps = stream_bandwidth_kbps * 10.0f; - f_dsc_target_bpp = f_stream_bandwidth_100bps / pix_clk_100hz; - - // Round down to the nearest precision stop to bring it into DSC spec - // range - dsc_target_bpp_x16 = (u32)(f_dsc_target_bpp * precision); - dsc_target_bpp_x16 = (dsc_target_bpp_x16 * 16) / precision; - - return dsc_target_bpp_x16; -} - /** * calc_rc_params - reads the user's cmdline mode * @rc: DC internal DSC parameters @@ -367,26 +347,3 @@ u32 calc_dsc_bytes_per_pixel(const struct drm_dsc_config *pps) DC_FP_END(); return ret; } - -/** - * calc_dsc_bpp_x16 - retrieve the dsc bits per pixel - * @stream_bandwidth_kbps: - * @pix_clk_100hz: - * @bpp_increment_div: - * - * Calculate the total of bits per pixel for DSC configuration. - * - * @note This calculation requires float point operation, most of it executes - * under kernel_fpu_{begin,end}. - */ -u32 calc_dsc_bpp_x16(u32 stream_bandwidth_kbps, u32 pix_clk_100hz, - u32 bpp_increment_div) -{ - u32 dsc_bpp; - - DC_FP_START(); - dsc_bpp = _do_calc_dsc_bpp_x16(stream_bandwidth_kbps, pix_clk_100hz, - bpp_increment_div); - DC_FP_END(); - return dsc_bpp; -} diff --git a/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.h b/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.h index 8123827840c5..262f06afcbf9 100644 --- a/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.h +++ b/drivers/gpu/drm/amd/display/dc/dsc/rc_calc.h @@ -79,8 +79,6 @@ typedef struct qp_entry qp_table[]; void calc_rc_params(struct rc_params *rc, const struct drm_dsc_config *pps); u32 calc_dsc_bytes_per_pixel(const struct drm_dsc_config *pps); -u32 calc_dsc_bpp_x16(u32 stream_bandwidth_kbps, u32 pix_clk_100hz, - u32 bpp_increment_div); #endif diff --git a/drivers/gpu/drm/amd/display/dc/inc/dc_link_dp.h b/drivers/gpu/drm/amd/display/dc/inc/dc_link_dp.h index 3ae05c96d557..428842511c03 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/dc_link_dp.h +++ b/drivers/gpu/drm/amd/display/dc/inc/dc_link_dp.h @@ -65,7 +65,8 @@ bool perform_link_training_with_retries( bool skip_video_pattern, int attempts, struct pipe_ctx *pipe_ctx, - enum signal_type signal); + enum signal_type signal, + bool do_fallback); bool is_mst_supported(struct dc_link *link); @@ -75,6 +76,8 @@ void detect_edp_sink_caps(struct dc_link *link); bool is_dp_active_dongle(const struct dc_link *link); +bool is_dp_branch_device(const struct dc_link *link); + bool is_edp_ilr_optimization_required(struct dc_link *link, struct dc_crtc_timing *crtc_timing); void dp_enable_mst_on_sink(struct dc_link *link, bool enable); @@ -94,5 +97,12 @@ void dp_set_dsc_on_stream(struct pipe_ctx *pipe_ctx, bool enable); bool dp_update_dsc_config(struct pipe_ctx *pipe_ctx); bool dp_set_dsc_on_rx(struct pipe_ctx *pipe_ctx, bool enable); +/* Convert PHY repeater count read from DPCD uint8_t. */ +uint8_t dp_convert_to_count(uint8_t lttpr_repeater_count); + +/* Check DPCD training status registers to detect link loss. */ +enum link_training_result dp_check_link_loss_status( + struct dc_link *link, + const struct link_training_settings *link_training_setting); #endif /* __DC_LINK_DP_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/dsc.h b/drivers/gpu/drm/amd/display/dc/inc/hw/dsc.h index f520e13aee4c..f94135c6e3c2 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/dsc.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/dsc.h @@ -88,6 +88,7 @@ struct dsc_enc_caps { int32_t max_total_throughput_mps; /* Maximum total throughput with all the slices combined */ int32_t max_slice_width; uint32_t bpp_increment_div; /* bpp increment divisor, e.g. if 16, it's 1/16th of a bit */ + bool is_dp; }; struct dsc_funcs { diff --git a/drivers/gpu/drm/amd/display/dc/inc/link_enc_cfg.h b/drivers/gpu/drm/amd/display/dc/inc/link_enc_cfg.h index 7d36e55f3097..883dd8733ea4 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/link_enc_cfg.h +++ b/drivers/gpu/drm/amd/display/dc/inc/link_enc_cfg.h @@ -81,6 +81,11 @@ struct dc_link *link_enc_cfg_get_link_using_link_enc( /* Return DIG link encoder used by link. NULL if unused. */ struct link_encoder *link_enc_cfg_get_link_enc_used_by_link( struct dc_state *state, - struct dc_link *link); + const struct dc_link *link); + +/* Return next available DIG link encoder. NULL if none available. */ +struct link_encoder *link_enc_cfg_get_next_avail_link_enc( + const struct dc *dc, + const struct dc_state *state); #endif /* DC_INC_LINK_ENC_CFG_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/irq/dcn21/irq_service_dcn21.c b/drivers/gpu/drm/amd/display/dc/irq/dcn21/irq_service_dcn21.c index 1a5be2792055..ed54e1c819be 100644 --- a/drivers/gpu/drm/amd/display/dc/irq/dcn21/irq_service_dcn21.c +++ b/drivers/gpu/drm/amd/display/dc/irq/dcn21/irq_service_dcn21.c @@ -58,8 +58,8 @@ enum dc_irq_source to_dal_irq_source_dcn21( return DC_IRQ_SOURCE_VBLANK5; case DCN_1_0__SRCID__DC_D6_OTG_VSTARTUP: return DC_IRQ_SOURCE_VBLANK6; - case DCN_1_0__SRCID__DMCUB_OUTBOX_HIGH_PRIORITY_READY_INT: - return DC_IRQ_SOURCE_DMCUB_OUTBOX0; + case DCN_1_0__SRCID__DMCUB_OUTBOX_LOW_PRIORITY_READY_INT: + return DC_IRQ_SOURCE_DMCUB_OUTBOX; case DCN_1_0__SRCID__OTG1_VERTICAL_INTERRUPT0_CONTROL: return DC_IRQ_SOURCE_DC1_VLINE0; case DCN_1_0__SRCID__OTG2_VERTICAL_INTERRUPT0_CONTROL: @@ -187,7 +187,7 @@ static const struct irq_source_info_funcs vupdate_no_lock_irq_info_funcs = { .ack = NULL }; -static const struct irq_source_info_funcs dmub_trace_irq_info_funcs = { +static const struct irq_source_info_funcs dmub_outbox_irq_info_funcs = { .set = NULL, .ack = NULL }; @@ -301,11 +301,11 @@ static const struct irq_source_info_funcs vline0_irq_info_funcs = { .funcs = &vline0_irq_info_funcs\ } -#define dmub_trace_int_entry()\ - [DC_IRQ_SOURCE_DMCUB_OUTBOX0] = {\ - IRQ_REG_ENTRY_DMUB(DMCUB_INTERRUPT_ENABLE, DMCUB_OUTBOX0_READY_INT_EN,\ - DMCUB_INTERRUPT_ACK, DMCUB_OUTBOX0_READY_INT_ACK),\ - .funcs = &dmub_trace_irq_info_funcs\ +#define dmub_outbox_int_entry()\ + [DC_IRQ_SOURCE_DMCUB_OUTBOX] = {\ + IRQ_REG_ENTRY_DMUB(DMCUB_INTERRUPT_ENABLE, DMCUB_OUTBOX1_READY_INT_EN,\ + DMCUB_INTERRUPT_ACK, DMCUB_OUTBOX1_READY_INT_ACK),\ + .funcs = &dmub_outbox_irq_info_funcs\ } #define dummy_irq_entry() \ @@ -426,7 +426,7 @@ irq_source_info_dcn21[DAL_IRQ_SOURCES_NUMBER] = { vline0_int_entry(3), vline0_int_entry(4), vline0_int_entry(5), - dmub_trace_int_entry(), + dmub_outbox_int_entry(), }; static const struct irq_service_funcs irq_service_funcs_dcn21 = { diff --git a/drivers/gpu/drm/amd/display/dc/irq_types.h b/drivers/gpu/drm/amd/display/dc/irq_types.h index ae8f47ec0f8c..5f9346622301 100644 --- a/drivers/gpu/drm/amd/display/dc/irq_types.h +++ b/drivers/gpu/drm/amd/display/dc/irq_types.h @@ -150,7 +150,7 @@ enum dc_irq_source { DC_IRQ_SOURCE_DC4_VLINE1, DC_IRQ_SOURCE_DC5_VLINE1, DC_IRQ_SOURCE_DC6_VLINE1, - DC_IRQ_DMCUB_OUTBOX1, + DC_IRQ_SOURCE_DMCUB_OUTBOX, DC_IRQ_SOURCE_DMCUB_OUTBOX0, DAL_IRQ_SOURCES_NUMBER diff --git a/drivers/gpu/drm/amd/display/dmub/dmub_srv.h b/drivers/gpu/drm/amd/display/dmub/dmub_srv.h index b4e14960b164..3ef6b536ccea 100644 --- a/drivers/gpu/drm/amd/display/dmub/dmub_srv.h +++ b/drivers/gpu/drm/amd/display/dmub/dmub_srv.h @@ -216,6 +216,23 @@ struct dmub_srv_fb_info { struct dmub_fb fb[DMUB_WINDOW_TOTAL]; }; +/* + * struct dmub_srv_hw_params - params for dmub hardware initialization + * @fb: framebuffer info for each region + * @fb_base: base of the framebuffer aperture + * @fb_offset: offset of the framebuffer aperture + * @psp_version: psp version to pass for DMCU init + * @load_inst_const: true if DMUB should load inst const fw + */ +struct dmub_srv_hw_params { + struct dmub_fb *fb[DMUB_WINDOW_TOTAL]; + uint64_t fb_base; + uint64_t fb_offset; + uint32_t psp_version; + bool load_inst_const; + bool skip_panel_power_sequence; +}; + /** * struct dmub_srv_base_funcs - Driver specific base callbacks */ @@ -290,7 +307,8 @@ struct dmub_srv_hw_funcs { bool (*is_hw_init)(struct dmub_srv *dmub); bool (*is_phy_init)(struct dmub_srv *dmub); - void (*enable_dmub_boot_options)(struct dmub_srv *dmub); + void (*enable_dmub_boot_options)(struct dmub_srv *dmub, + const struct dmub_srv_hw_params *params); void (*skip_dmub_panel_power_sequence)(struct dmub_srv *dmub, bool skip); @@ -325,23 +343,6 @@ struct dmub_srv_create_params { bool is_virtual; }; -/* - * struct dmub_srv_hw_params - params for dmub hardware initialization - * @fb: framebuffer info for each region - * @fb_base: base of the framebuffer aperture - * @fb_offset: offset of the framebuffer aperture - * @psp_version: psp version to pass for DMCU init - * @load_inst_const: true if DMUB should load inst const fw - */ -struct dmub_srv_hw_params { - struct dmub_fb *fb[DMUB_WINDOW_TOTAL]; - uint64_t fb_base; - uint64_t fb_offset; - uint32_t psp_version; - bool load_inst_const; - bool skip_panel_power_sequence; -}; - /** * struct dmub_srv - software state for dmcub * @asic: dmub asic identifier diff --git a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h index 4195ff10c514..40ce15eb934c 100644 --- a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h +++ b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h @@ -47,10 +47,10 @@ /* Firmware versioning. */ #ifdef DMUB_EXPOSE_VERSION -#define DMUB_FW_VERSION_GIT_HASH 0x23db9b126 +#define DMUB_FW_VERSION_GIT_HASH 0x2cab49dfb #define DMUB_FW_VERSION_MAJOR 0 #define DMUB_FW_VERSION_MINOR 0 -#define DMUB_FW_VERSION_REVISION 62 +#define DMUB_FW_VERSION_REVISION 65 #define DMUB_FW_VERSION_TEST 0 #define DMUB_FW_VERSION_VBIOS 0 #define DMUB_FW_VERSION_HOTFIX 0 @@ -117,18 +117,20 @@ /* Maximum number of planes on any ASIC. */ #define DMUB_MAX_PLANES 6 +#define DMUB_MAX_SUBVP_STREAMS 2 + /* Trace buffer offset for entry */ #define TRACE_BUFFER_ENTRY_OFFSET 16 /** - * ABM backlight control version legacy + * ABM control version legacy */ -#define DMUB_CMD_ABM_SET_BACKLIGHT_VERSION_UNKNOWN 0x0 +#define DMUB_CMD_ABM_CONTROL_VERSION_UNKNOWN 0x0 /** - * ABM backlight control version with multi edp support + * ABM control version with multi edp support */ -#define DMUB_CMD_ABM_SET_BACKLIGHT_VERSION_1 0x1 +#define DMUB_CMD_ABM_CONTROL_VERSION_1 0x1 /** * Physical framebuffer address location, 64-bit. @@ -163,6 +165,13 @@ extern "C" { #endif /** + * Number of nanoseconds per DMUB tick. + * DMCUB_TIMER_CURRENT increments in DMUB ticks, which are 10ns by default. + * If DMCUB_TIMER_WINDOW is non-zero this will no longer be true. + */ +#define NS_PER_DMUB_TICK 10 + +/** * union dmub_addr - DMUB physical/virtual 64-bit address. */ union dmub_addr { @@ -328,7 +337,8 @@ union dmub_fw_boot_options { uint32_t skip_phy_access : 1; /**< 1 if PHY access should be skipped */ uint32_t disable_clk_gate: 1; /**< 1 if clock gating should be disabled */ uint32_t skip_phy_init_panel_sequence: 1; /**< 1 to skip panel init seq */ - uint32_t reserved : 26; /**< reserved */ + uint32_t reserved_unreleased: 1; /**< reserved for an unreleased feature */ + uint32_t reserved : 25; /**< reserved */ } bits; /**< boot bits */ uint32_t all; /**< 32-bit access to bits */ }; @@ -452,6 +462,61 @@ enum dmub_gpint_command { DMUB_GPINT__PSR_RESIDENCY = 9, }; +/** + * INBOX0 generic command definition + */ +union dmub_inbox0_cmd_common { + struct { + uint32_t command_code: 8; /**< INBOX0 command code */ + uint32_t param: 24; /**< 24-bit parameter */ + } bits; + uint32_t all; +}; + +/** + * INBOX0 hw_lock command definition + */ +union dmub_inbox0_cmd_lock_hw { + struct { + uint32_t command_code: 8; + + /* NOTE: Must be have enough bits to match: enum hw_lock_client */ + uint32_t hw_lock_client: 1; + + /* NOTE: Below fields must match with: struct dmub_hw_lock_inst_flags */ + uint32_t otg_inst: 3; + uint32_t opp_inst: 3; + uint32_t dig_inst: 3; + + /* NOTE: Below fields must match with: union dmub_hw_lock_flags */ + uint32_t lock_pipe: 1; + uint32_t lock_cursor: 1; + uint32_t lock_dig: 1; + uint32_t triple_buffer_lock: 1; + + uint32_t lock: 1; /**< Lock */ + uint32_t should_release: 1; /**< Release */ + uint32_t reserved: 8; /**< Reserved for extending more clients, HW, etc. */ + } bits; + uint32_t all; +}; + +union dmub_inbox0_data_register { + union dmub_inbox0_cmd_common inbox0_cmd_common; + union dmub_inbox0_cmd_lock_hw inbox0_cmd_lock_hw; +}; + +enum dmub_inbox0_command { + /** + * DESC: Invalid command, ignored. + */ + DMUB_INBOX0_CMD__INVALID_COMMAND = 0, + /** + * DESC: Notification to acquire/release HW lock + * ARGS: + */ + DMUB_INBOX0_CMD__HW_LOCK = 1, +}; //============================================================================== //</DMUB_GPINT>================================================================= //============================================================================== @@ -570,7 +635,8 @@ struct dmub_cmd_header { unsigned int type : 8; /**< command type */ unsigned int sub_type : 8; /**< command sub type */ unsigned int ret_status : 1; /**< 1 if returned data, 0 otherwise */ - unsigned int reserved0 : 7; /**< reserved bits */ + unsigned int multi_cmd_pending : 1; /**< 1 if multiple commands chained together */ + unsigned int reserved0 : 6; /**< reserved bits */ unsigned int payload_bytes : 6; /* payload excluding header - up to 60 bytes */ unsigned int reserved1 : 2; /**< reserved bits */ }; @@ -1343,6 +1409,9 @@ struct dmub_rb_cmd_psr_force_static { /** * Set of HW components that can be locked. + * + * Note: If updating with more HW components, fields + * in dmub_inbox0_cmd_lock_hw must be updated to match. */ union dmub_hw_lock_flags { /** @@ -1375,6 +1444,9 @@ union dmub_hw_lock_flags { /** * Instances of HW to be locked. + * + * Note: If updating with more HW components, fields + * in dmub_inbox0_cmd_lock_hw must be updated to match. */ struct dmub_hw_lock_inst_flags { /** @@ -1398,16 +1470,16 @@ struct dmub_hw_lock_inst_flags { /** * Clients that can acquire the HW Lock Manager. + * + * Note: If updating with more clients, fields in + * dmub_inbox0_cmd_lock_hw must be updated to match. */ enum hw_lock_client { /** * Driver is the client of HW Lock Manager. */ HW_LOCK_CLIENT_DRIVER = 0, - /** - * FW is the client of HW Lock Manager. - */ - HW_LOCK_CLIENT_FW, + HW_LOCK_CLIENT_SUBVP = 3, /** * Invalid client. */ @@ -1637,7 +1709,7 @@ struct dmub_cmd_abm_set_backlight_data { uint32_t backlight_user_level; /** - * Backlight data version. + * ABM control version. */ uint8_t version; @@ -1677,6 +1749,23 @@ struct dmub_cmd_abm_set_level_data { * Set current ABM operating/aggression level. */ uint32_t level; + + /** + * ABM control version. + */ + uint8_t version; + + /** + * Panel Control HW instance mask. + * Bit 0 is Panel Control HW instance 0. + * Bit 1 is Panel Control HW instance 1. + */ + uint8_t panel_mask; + + /** + * Explicit padding to 4 byte boundary. + */ + uint8_t pad[2]; }; /** @@ -1702,6 +1791,23 @@ struct dmub_cmd_abm_set_ambient_level_data { * Ambient light sensor reading from OS. */ uint32_t ambient_lux; + + /** + * ABM control version. + */ + uint8_t version; + + /** + * Panel Control HW instance mask. + * Bit 0 is Panel Control HW instance 0. + * Bit 1 is Panel Control HW instance 1. + */ + uint8_t panel_mask; + + /** + * Explicit padding to 4 byte boundary. + */ + uint8_t pad[2]; }; /** @@ -1728,6 +1834,23 @@ struct dmub_cmd_abm_set_pwm_frac_data { * TODO: Convert to uint8_t. */ uint32_t fractional_pwm; + + /** + * ABM control version. + */ + uint8_t version; + + /** + * Panel Control HW instance mask. + * Bit 0 is Panel Control HW instance 0. + * Bit 1 is Panel Control HW instance 1. + */ + uint8_t panel_mask; + + /** + * Explicit padding to 4 byte boundary. + */ + uint8_t pad[2]; }; /** @@ -1758,6 +1881,24 @@ struct dmub_cmd_abm_init_config_data { * Indirect buffer length. */ uint16_t bytes; + + + /** + * ABM control version. + */ + uint8_t version; + + /** + * Panel Control HW instance mask. + * Bit 0 is Panel Control HW instance 0. + * Bit 1 is Panel Control HW instance 1. + */ + uint8_t panel_mask; + + /** + * Explicit padding to 4 byte boundary. + */ + uint8_t pad[2]; }; /** @@ -2127,6 +2268,46 @@ static inline bool dmub_rb_front(struct dmub_rb *rb, } /** + * @brief Determines the next ringbuffer offset. + * + * @param rb DMUB inbox ringbuffer + * @param num_cmds Number of commands + * @param next_rptr The next offset in the ringbuffer + */ +static inline void dmub_rb_get_rptr_with_offset(struct dmub_rb *rb, + uint32_t num_cmds, + uint32_t *next_rptr) +{ + *next_rptr = rb->rptr + DMUB_RB_CMD_SIZE * num_cmds; + + if (*next_rptr >= rb->capacity) + *next_rptr %= rb->capacity; +} + +/** + * @brief Returns a pointer to a command in the inbox. + * + * @param rb DMUB inbox ringbuffer + * @param cmd The inbox command to return + * @param rptr The ringbuffer offset + * @return true if not empty + * @return false otherwise + */ +static inline bool dmub_rb_peek_offset(struct dmub_rb *rb, + union dmub_rb_cmd **cmd, + uint32_t rptr) +{ + uint8_t *rb_cmd = (uint8_t *)(rb->base_address) + rptr; + + if (dmub_rb_empty(rb)) + return false; + + *cmd = (union dmub_rb_cmd *)rb_cmd; + + return true; +} + +/** * @brief Returns the next unprocessed command in the outbox. * * @param rb DMUB outbox ringbuffer diff --git a/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn20.c b/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn20.c index 6934906c665e..b11f530f682c 100644 --- a/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn20.c +++ b/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn20.c @@ -385,7 +385,7 @@ union dmub_fw_boot_status dmub_dcn20_get_fw_boot_status(struct dmub_srv *dmub) return status; } -void dmub_dcn20_enable_dmub_boot_options(struct dmub_srv *dmub) +void dmub_dcn20_enable_dmub_boot_options(struct dmub_srv *dmub, const struct dmub_srv_hw_params *params) { union dmub_fw_boot_options boot_options = {0}; diff --git a/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn20.h b/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn20.h index de5351cd5abc..42d610a260ef 100644 --- a/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn20.h +++ b/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn20.h @@ -221,7 +221,7 @@ bool dmub_dcn20_is_gpint_acked(struct dmub_srv *dmub, uint32_t dmub_dcn20_get_gpint_response(struct dmub_srv *dmub); -void dmub_dcn20_enable_dmub_boot_options(struct dmub_srv *dmub); +void dmub_dcn20_enable_dmub_boot_options(struct dmub_srv *dmub, const struct dmub_srv_hw_params *params); void dmub_dcn20_skip_dmub_panel_power_sequence(struct dmub_srv *dmub, bool skip); diff --git a/drivers/gpu/drm/amd/display/dmub/src/dmub_srv.c b/drivers/gpu/drm/amd/display/dmub/src/dmub_srv.c index 1cbb125b4063..ed9fa6138aa6 100644 --- a/drivers/gpu/drm/amd/display/dmub/src/dmub_srv.c +++ b/drivers/gpu/drm/amd/display/dmub/src/dmub_srv.c @@ -514,6 +514,10 @@ enum dmub_status dmub_srv_hw_init(struct dmub_srv *dmub, outbox0_rb_params.capacity = tracebuff_fb->size - dmub_align(TRACE_BUFFER_ENTRY_OFFSET, 64); dmub_rb_init(&dmub->outbox0_rb, &outbox0_rb_params); + /* Report to DMUB what features are supported by current driver */ + if (dmub->hw_funcs.enable_dmub_boot_options) + dmub->hw_funcs.enable_dmub_boot_options(dmub, params); + if (dmub->hw_funcs.reset_release) dmub->hw_funcs.reset_release(dmub); diff --git a/drivers/gpu/drm/amd/display/include/link_service_types.h b/drivers/gpu/drm/amd/display/include/link_service_types.h index 7392a89e771f..73ada16658aa 100644 --- a/drivers/gpu/drm/amd/display/include/link_service_types.h +++ b/drivers/gpu/drm/amd/display/include/link_service_types.h @@ -68,6 +68,14 @@ enum link_training_result { LINK_TRAINING_LQA_FAIL, /* one of the CR,EQ or symbol lock is dropped */ LINK_TRAINING_LINK_LOSS, + /* Abort link training (because sink unplugged) */ + LINK_TRAINING_ABORT, +}; + +enum lttpr_mode { + LTTPR_MODE_NON_LTTPR, + LTTPR_MODE_TRANSPARENT, + LTTPR_MODE_NON_TRANSPARENT, }; struct link_training_settings { diff --git a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp.c b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp.c index 68a6481d7f8f..b963226e8af4 100644 --- a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp.c +++ b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp.c @@ -260,7 +260,6 @@ enum mod_hdcp_status mod_hdcp_setup(struct mod_hdcp *hdcp, struct mod_hdcp_output output; enum mod_hdcp_status status = MOD_HDCP_STATUS_SUCCESS; - memset(hdcp, 0, sizeof(struct mod_hdcp)); memset(&output, 0, sizeof(output)); hdcp->config = *config; HDCP_TOP_INTERFACE_TRACE(hdcp); diff --git a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp1_execution.c b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp1_execution.c index 2cbd931363bd..43e6f8b17e79 100644 --- a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp1_execution.c +++ b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp1_execution.c @@ -128,6 +128,11 @@ static inline uint8_t get_device_count(struct mod_hdcp *hdcp) static inline enum mod_hdcp_status check_device_count(struct mod_hdcp *hdcp) { + /* Avoid device count == 0 to do authentication */ + if (0 == get_device_count(hdcp)) { + return MOD_HDCP_STATUS_HDCP1_DEVICE_COUNT_MISMATCH_FAILURE; + } + /* Some MST display may choose to report the internal panel as an HDCP RX. * To update this condition with 1(because the immediate repeater's internal * panel is possibly not included in DEVICE_COUNT) + get_device_count(hdcp). diff --git a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp2_execution.c b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp2_execution.c index c1331facdcb4..117c6b45f718 100644 --- a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp2_execution.c +++ b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp2_execution.c @@ -207,6 +207,11 @@ static inline uint8_t get_device_count(struct mod_hdcp *hdcp) static enum mod_hdcp_status check_device_count(struct mod_hdcp *hdcp) { + /* Avoid device count == 0 to do authentication */ + if (0 == get_device_count(hdcp)) { + return MOD_HDCP_STATUS_HDCP1_DEVICE_COUNT_MISMATCH_FAILURE; + } + /* Some MST display may choose to report the internal panel as an HDCP RX. */ /* To update this condition with 1(because the immediate repeater's internal */ /* panel is possibly not included in DEVICE_COUNT) + get_device_count(hdcp). */ diff --git a/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_sh_mask.h index 7afa87c7ff54..f804e13b002e 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_sh_mask.h +++ b/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_sh_mask.h @@ -50,6 +50,7 @@ #define DF_CS_UMC_AON0_DramBaseAddress0__AddrRngVal_MASK 0x00000001L #define DF_CS_UMC_AON0_DramBaseAddress0__LgcyMmioHoleEn_MASK 0x00000002L #define DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan_MASK 0x0000003CL +#define ALDEBARAN_DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan_MASK 0x0000007CL #define DF_CS_UMC_AON0_DramBaseAddress0__IntLvAddrSel_MASK 0x00000E00L #define DF_CS_UMC_AON0_DramBaseAddress0__DramBaseAddr_MASK 0xFFFFF000L diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index 353468667036..e2d13131a432 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -350,6 +350,7 @@ struct amd_pm_funcs { unsigned int *num_states); int (*get_dpm_clock_table)(void *handle, struct dpm_clocks *clock_table); + int (*get_smu_prv_buf_details)(void *handle, void **addr, size_t *size); }; struct metrics_table_header { @@ -473,6 +474,68 @@ struct gpu_metrics_v1_1 { uint16_t temperature_hbm[NUM_HBM_INSTANCES]; }; +struct gpu_metrics_v1_2 { + struct metrics_table_header common_header; + + /* Temperature */ + uint16_t temperature_edge; + uint16_t temperature_hotspot; + uint16_t temperature_mem; + uint16_t temperature_vrgfx; + uint16_t temperature_vrsoc; + uint16_t temperature_vrmem; + + /* Utilization */ + uint16_t average_gfx_activity; + uint16_t average_umc_activity; // memory controller + uint16_t average_mm_activity; // UVD or VCN + + /* Power/Energy */ + uint16_t average_socket_power; + uint64_t energy_accumulator; + + /* Driver attached timestamp (in ns) */ + uint64_t system_clock_counter; + + /* Average clocks */ + uint16_t average_gfxclk_frequency; + uint16_t average_socclk_frequency; + uint16_t average_uclk_frequency; + uint16_t average_vclk0_frequency; + uint16_t average_dclk0_frequency; + uint16_t average_vclk1_frequency; + uint16_t average_dclk1_frequency; + + /* Current clocks */ + uint16_t current_gfxclk; + uint16_t current_socclk; + uint16_t current_uclk; + uint16_t current_vclk0; + uint16_t current_dclk0; + uint16_t current_vclk1; + uint16_t current_dclk1; + + /* Throttle status */ + uint32_t throttle_status; + + /* Fans */ + uint16_t current_fan_speed; + + /* Link width/speed */ + uint16_t pcie_link_width; + uint16_t pcie_link_speed; // in 0.1 GT/s + + uint16_t padding; + + uint32_t gfx_activity_acc; + uint32_t mem_activity_acc; + + uint16_t temperature_hbm[NUM_HBM_INSTANCES]; + + /* PMFW attached timestamp (10ns resolution) */ + uint64_t firmware_timestamp; +}; + /* * gpu_metrics_v2_0 is not recommended as it's not naturally aligned. * Use gpu_metrics_v2_1 or later instead. diff --git a/drivers/gpu/drm/amd/pm/amdgpu_pm.c b/drivers/gpu/drm/amd/pm/amdgpu_pm.c index 9a54066ec0af..13da377888d2 100644 --- a/drivers/gpu/drm/amd/pm/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/pm/amdgpu_pm.c @@ -735,6 +735,23 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, * - a list of valid ranges for sclk, mclk, and voltage curve points * labeled OD_RANGE * + * < For APUs > + * + * Reading the file will display: + * + * - minimum and maximum engine clock labeled OD_SCLK + * + * - a list of valid ranges for sclk labeled OD_RANGE + * + * < For VanGogh > + * + * Reading the file will display: + * + * - minimum and maximum engine clock labeled OD_SCLK + * - minimum and maximum core clocks labeled OD_CCLK + * + * - a list of valid ranges for sclk and cclk labeled OD_RANGE + * * To manually adjust these settings: * * - First select manual using power_dpm_force_performance_level @@ -743,7 +760,10 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, * string that contains "s/m index clock" to the file. The index * should be 0 if to set minimum clock. And 1 if to set maximum * clock. E.g., "s 0 500" will update minimum sclk to be 500 MHz. - * "m 1 800" will update maximum mclk to be 800Mhz. + * "m 1 800" will update maximum mclk to be 800Mhz. For core + * clocks on VanGogh, the string contains "p core index clock". + * E.g., "p 2 0 800" would set the minimum core clock on core + * 2 to 800Mhz. * * For sclk voltage curve, enter the new values by writing a * string that contains "vc point clock voltage" to the file. The @@ -3534,6 +3554,45 @@ out: DEFINE_SHOW_ATTRIBUTE(amdgpu_debugfs_pm_info); +/* + * amdgpu_pm_priv_buffer_read - Read memory region allocated to FW + * + * Reads debug memory region allocated to PMFW + */ +static ssize_t amdgpu_pm_prv_buffer_read(struct file *f, char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs; + void *pp_handle = adev->powerplay.pp_handle; + size_t smu_prv_buf_size; + void *smu_prv_buf; + + if (amdgpu_in_reset(adev)) + return -EPERM; + if (adev->in_suspend && !adev->in_runpm) + return -EPERM; + + if (pp_funcs && pp_funcs->get_smu_prv_buf_details) + pp_funcs->get_smu_prv_buf_details(pp_handle, &smu_prv_buf, + &smu_prv_buf_size); + else + return -ENOSYS; + + if (!smu_prv_buf || !smu_prv_buf_size) + return -EINVAL; + + return simple_read_from_buffer(buf, size, pos, smu_prv_buf, + smu_prv_buf_size); +} + +static const struct file_operations amdgpu_debugfs_pm_prv_buffer_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .read = amdgpu_pm_prv_buffer_read, + .llseek = default_llseek, +}; + #endif void amdgpu_debugfs_pm_init(struct amdgpu_device *adev) @@ -3545,5 +3604,10 @@ void amdgpu_debugfs_pm_init(struct amdgpu_device *adev) debugfs_create_file("amdgpu_pm_info", 0444, root, adev, &amdgpu_debugfs_pm_info_fops); + if (adev->pm.smu_prv_buffer_size > 0) + debugfs_create_file_size("amdgpu_pm_prv_buffer", 0444, root, + adev, + &amdgpu_debugfs_pm_prv_buffer_fops, + adev->pm.smu_prv_buffer_size); #endif } diff --git a/drivers/gpu/drm/amd/pm/inc/smu13_driver_if_aldebaran.h b/drivers/gpu/drm/amd/pm/inc/smu13_driver_if_aldebaran.h index d23533bda002..a017983ff1fa 100644 --- a/drivers/gpu/drm/amd/pm/inc/smu13_driver_if_aldebaran.h +++ b/drivers/gpu/drm/amd/pm/inc/smu13_driver_if_aldebaran.h @@ -64,7 +64,7 @@ #define FEATURE_SMUIO_CG_BIT 28 #define FEATURE_THM_CG_BIT 29 #define FEATURE_CLK_CG_BIT 30 -#define FEATURE_SPARE_31_BIT 31 +#define FEATURE_EDC_BIT 31 #define FEATURE_SPARE_32_BIT 32 #define FEATURE_SPARE_33_BIT 33 #define FEATURE_SPARE_34_BIT 34 @@ -439,8 +439,11 @@ typedef struct { int8_t XgmiOffset; // in Amps uint8_t Padding_TelemetryXgmi; + uint16_t EdcPowerLimit; + uint16_t spare6; + //reserved - uint32_t reserved[15]; + uint32_t reserved[14]; } PPTable_t; diff --git a/drivers/gpu/drm/amd/pm/inc/smu_v13_0.h b/drivers/gpu/drm/amd/pm/inc/smu_v13_0.h index 8145e1cbf181..1687709507b3 100644 --- a/drivers/gpu/drm/amd/pm/inc/smu_v13_0.h +++ b/drivers/gpu/drm/amd/pm/inc/smu_v13_0.h @@ -26,7 +26,7 @@ #include "amdgpu_smu.h" #define SMU13_DRIVER_IF_VERSION_INV 0xFFFFFFFF -#define SMU13_DRIVER_IF_VERSION_ALDE 0x6 +#define SMU13_DRIVER_IF_VERSION_ALDE 0x07 /* MP Apertures */ #define MP0_Public 0x03800000 diff --git a/drivers/gpu/drm/amd/pm/powerplay/amd_powerplay.c b/drivers/gpu/drm/amd/pm/powerplay/amd_powerplay.c index ee6340c6f921..c73504e998e5 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/amd_powerplay.c +++ b/drivers/gpu/drm/amd/pm/powerplay/amd_powerplay.c @@ -1651,6 +1651,26 @@ static int pp_gfx_state_change_set(void *handle, uint32_t state) return 0; } +static int pp_get_prv_buffer_details(void *handle, void **addr, size_t *size) +{ + struct pp_hwmgr *hwmgr = handle; + struct amdgpu_device *adev = hwmgr->adev; + + if (!addr || !size) + return -EINVAL; + + *addr = NULL; + *size = 0; + mutex_lock(&hwmgr->smu_lock); + if (adev->pm.smu_prv_buffer) { + amdgpu_bo_kmap(adev->pm.smu_prv_buffer, addr); + *size = adev->pm.smu_prv_buffer_size; + } + mutex_unlock(&hwmgr->smu_lock); + + return 0; +} + 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, @@ -1714,4 +1734,5 @@ static const struct amd_pm_funcs pp_dpm_funcs = { .set_xgmi_pstate = pp_set_xgmi_pstate, .get_gpu_metrics = pp_get_gpu_metrics, .gfx_state_change_set = pp_gfx_state_change_set, + .get_smu_prv_buf_details = pp_get_prv_buffer_details, }; diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/ppatomctrl.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/ppatomctrl.c index b1038d30c8dc..f503e61faa60 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/ppatomctrl.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/ppatomctrl.c @@ -275,7 +275,7 @@ static const ATOM_VOLTAGE_OBJECT_V3 *atomctrl_lookup_voltage_type_v3( } /** - * atomctrl_get_memory_pll_dividers_si(). + * atomctrl_get_memory_pll_dividers_si * * @hwmgr: input parameter: pointer to HwMgr * @clock_value: input parameter: memory clock @@ -328,7 +328,7 @@ int atomctrl_get_memory_pll_dividers_si( } /** - * atomctrl_get_memory_pll_dividers_vi(). + * atomctrl_get_memory_pll_dividers_vi * * @hwmgr: input parameter: pointer to HwMgr * @clock_value: input parameter: memory clock @@ -1104,7 +1104,7 @@ int atomctrl_calculate_voltage_evv_on_sclk( } /** - * atomctrl_get_voltage_evv_on_sclk gets voltage via call to ATOM COMMAND table. + * atomctrl_get_voltage_evv_on_sclk: gets voltage via call to ATOM COMMAND table. * @hwmgr: input: pointer to hwManager * @voltage_type: input: type of EVV voltage VDDC or VDDGFX * @sclk: input: in 10Khz unit. DPM state SCLK frequency @@ -1144,7 +1144,7 @@ int atomctrl_get_voltage_evv_on_sclk( } /** - * atomctrl_get_voltage_evv gets voltage via call to ATOM COMMAND table. + * atomctrl_get_voltage_evv: gets voltage via call to ATOM COMMAND table. * @hwmgr: input: pointer to hwManager * @virtual_voltage_id: input: voltage id which match per voltage DPM state: 0xff01, 0xff02.. 0xff08 * @voltage: output: real voltage level in unit of mv diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.c index 2a28c9df15a0..8d99c7a5abf8 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega20_baco.c @@ -85,7 +85,7 @@ int vega20_baco_set_state(struct pp_hwmgr *hwmgr, enum BACO_STATE state) return 0; if (state == BACO_STATE_IN) { - if (!ras || !ras->supported) { + if (!ras || !adev->ras_enabled) { data = RREG32_SOC15(THM, 0, mmTHM_BACO_CNTL); data |= 0x80000000; WREG32_SOC15(THM, 0, mmTHM_BACO_CNTL, data); diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index c29d8b3131b7..32dadeee9514 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -2933,6 +2933,26 @@ int smu_set_light_sbr(struct smu_context *smu, bool enable) return ret; } +static int smu_get_prv_buffer_details(void *handle, void **addr, size_t *size) +{ + struct smu_context *smu = handle; + struct smu_table_context *smu_table = &smu->smu_table; + struct smu_table *memory_pool = &smu_table->memory_pool; + + if (!addr || !size) + return -EINVAL; + + *addr = NULL; + *size = 0; + mutex_lock(&smu->mutex); + if (memory_pool->bo) { + *addr = memory_pool->cpu_addr; + *size = memory_pool->size; + } + mutex_unlock(&smu->mutex); + + return 0; +} static const struct amd_pm_funcs swsmu_pm_funcs = { /* export for sysfs */ @@ -2984,6 +3004,7 @@ static const struct amd_pm_funcs swsmu_pm_funcs = { .get_max_sustainable_clocks_by_dc = smu_get_max_sustainable_clocks_by_dc, .load_firmware = smu_load_microcode, .gfx_state_change_set = smu_gfx_state_change_set, + .get_smu_prv_buf_details = smu_get_prv_buffer_details, }; int smu_wait_for_event(struct amdgpu_device *adev, enum smu_event_type event, diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c index d2fd44b903ca..270b2b0b8e8a 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c @@ -302,7 +302,7 @@ sienna_cichlid_get_allowed_feature_mask(struct smu_context *smu, if (smu->dc_controlled_by_gpio) *(uint64_t *)feature_mask |= FEATURE_MASK(FEATURE_ACDC_BIT); - if (amdgpu_aspm == 1) + if (amdgpu_aspm) *(uint64_t *)feature_mask |= FEATURE_MASK(FEATURE_DS_LCLK_BIT); return 0; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c index 6274cae4a065..a06e6865507d 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c @@ -1531,7 +1531,8 @@ int smu_v11_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state) NULL); break; default: - if (!ras || !ras->supported || adev->gmc.xgmi.pending_reset) { + if (!ras || !adev->ras_enabled || + adev->gmc.xgmi.pending_reset) { if (adev->asic_type == CHIP_ARCTURUS) { data = RREG32_SOC15(THM, 0, mmTHM_BACO_CNTL_ARCT); data |= 0x80000000; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c index dcbe3a72da09..5d04a1dfdfd8 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c @@ -209,7 +209,7 @@ static int aldebaran_tables_init(struct smu_context *smu) return -ENOMEM; smu_table->metrics_time = 0; - smu_table->gpu_metrics_table_size = sizeof(struct gpu_metrics_v1_1); + smu_table->gpu_metrics_table_size = sizeof(struct gpu_metrics_v1_2); smu_table->gpu_metrics_table = kzalloc(smu_table->gpu_metrics_table_size, GFP_KERNEL); if (!smu_table->gpu_metrics_table) { kfree(smu_table->metrics_table); @@ -1632,8 +1632,8 @@ static ssize_t aldebaran_get_gpu_metrics(struct smu_context *smu, void **table) { struct smu_table_context *smu_table = &smu->smu_table; - struct gpu_metrics_v1_1 *gpu_metrics = - (struct gpu_metrics_v1_1 *)smu_table->gpu_metrics_table; + struct gpu_metrics_v1_2 *gpu_metrics = + (struct gpu_metrics_v1_2 *)smu_table->gpu_metrics_table; SmuMetrics_t metrics; int i, ret = 0; @@ -1643,7 +1643,7 @@ static ssize_t aldebaran_get_gpu_metrics(struct smu_context *smu, if (ret) return ret; - smu_cmn_init_soft_gpu_metrics(gpu_metrics, 1, 1); + smu_cmn_init_soft_gpu_metrics(gpu_metrics, 1, 2); gpu_metrics->temperature_edge = metrics.TemperatureEdge; gpu_metrics->temperature_hotspot = metrics.TemperatureHotspot; @@ -1657,7 +1657,9 @@ static ssize_t aldebaran_get_gpu_metrics(struct smu_context *smu, gpu_metrics->average_mm_activity = 0; gpu_metrics->average_socket_power = metrics.AverageSocketPower; - gpu_metrics->energy_accumulator = 0; + gpu_metrics->energy_accumulator = + (uint64_t)metrics.EnergyAcc64bitHigh << 32 | + metrics.EnergyAcc64bitLow; gpu_metrics->average_gfxclk_frequency = metrics.AverageGfxclkFrequency; gpu_metrics->average_socclk_frequency = metrics.AverageSocclkFrequency; @@ -1688,9 +1690,12 @@ static ssize_t aldebaran_get_gpu_metrics(struct smu_context *smu, for (i = 0; i < NUM_HBM_INSTANCES; i++) gpu_metrics->temperature_hbm[i] = metrics.TemperatureAllHBM[i]; + gpu_metrics->firmware_timestamp = ((uint64_t)metrics.TimeStampHigh << 32) | + metrics.TimeStampLow; + *table = (void *)gpu_metrics; - return sizeof(struct gpu_metrics_v1_1); + return sizeof(struct gpu_metrics_v1_2); } static int aldebaran_mode2_reset(struct smu_context *smu) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c index dc7d2e71aa6f..0934e5b3aa17 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c @@ -104,8 +104,8 @@ int smu_cmn_send_msg_without_waiting(struct smu_context *smu, ret = smu_cmn_wait_for_response(smu); if (ret != 0x1) { - dev_err(adev->dev, "Msg issuing pre-check failed and " - "SMU may be not in the right state!\n"); + dev_err(adev->dev, "Msg issuing pre-check failed(0x%x) and " + "SMU may be not in the right state!\n", ret); if (ret != -ETIME) ret = -EIO; return ret; @@ -761,6 +761,9 @@ void smu_cmn_init_soft_gpu_metrics(void *table, uint8_t frev, uint8_t crev) case METRICS_VERSION(1, 1): structure_size = sizeof(struct gpu_metrics_v1_1); break; + case METRICS_VERSION(1, 2): + structure_size = sizeof(struct gpu_metrics_v1_2); + break; case METRICS_VERSION(2, 0): structure_size = sizeof(struct gpu_metrics_v2_0); break; diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index cc445c4cba2e..46eea01950cb 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -1098,7 +1098,8 @@ static bool radeon_check_pot_argument(int arg) } /** - * Determine a sensible default GART size according to ASIC family. + * radeon_gart_size_auto - Determine a sensible default GART size + * according to ASIC family. * * @family: ASIC family name */ |