diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
28 files changed, 678 insertions, 339 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 0283e2b3c851..447c4c7a36d6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -74,6 +74,7 @@ #include "amdgpu_gart.h" #include "amdgpu_debugfs.h" #include "amdgpu_job.h" +#include "amdgpu_bo_list.h" /* * Modules parameters. @@ -690,45 +691,6 @@ struct amdgpu_fpriv { }; /* - * residency list - */ -struct amdgpu_bo_list_entry { - struct amdgpu_bo *robj; - struct ttm_validate_buffer tv; - struct amdgpu_bo_va *bo_va; - uint32_t priority; - struct page **user_pages; - int user_invalidated; -}; - -struct amdgpu_bo_list { - struct mutex lock; - struct rcu_head rhead; - struct kref refcount; - struct amdgpu_bo *gds_obj; - struct amdgpu_bo *gws_obj; - struct amdgpu_bo *oa_obj; - unsigned first_userptr; - unsigned num_entries; - struct amdgpu_bo_list_entry *array; -}; - -struct amdgpu_bo_list * -amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id); -void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, - struct list_head *validated); -void amdgpu_bo_list_put(struct amdgpu_bo_list *list); -void amdgpu_bo_list_free(struct amdgpu_bo_list *list); -int amdgpu_bo_create_list_entry_array(struct drm_amdgpu_bo_list_in *in, - struct drm_amdgpu_bo_list_entry **info_param); - -int amdgpu_bo_list_create(struct amdgpu_device *adev, - struct drm_file *filp, - struct drm_amdgpu_bo_list_entry *info, - unsigned num_entries, - struct amdgpu_bo_list **list); - -/* * GFX stuff */ #include "clearstate_defs.h" @@ -1748,6 +1710,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #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))) #define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib))) +#define amdgpu_ring_patch_cs_in_place(r, p, ib) ((r)->funcs->patch_cs_in_place((p), (ib))) #define amdgpu_ring_test_ring(r) (r)->funcs->test_ring((r)) #define amdgpu_ring_test_ib(r, t) (r)->funcs->test_ib((r), (t)) #define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c index f4c474a95875..71efcf38f11b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c @@ -57,6 +57,10 @@ #define ACP_I2S_COMP2_CAP_REG_OFFSET 0xa8 #define ACP_I2S_COMP1_PLAY_REG_OFFSET 0x6c #define ACP_I2S_COMP2_PLAY_REG_OFFSET 0x68 +#define ACP_BT_PLAY_REGS_START 0x14970 +#define ACP_BT_PLAY_REGS_END 0x14a24 +#define ACP_BT_COMP1_REG_OFFSET 0xac +#define ACP_BT_COMP2_REG_OFFSET 0xa8 #define mmACP_PGFSM_RETAIN_REG 0x51c9 #define mmACP_PGFSM_CONFIG_REG 0x51ca @@ -77,7 +81,7 @@ #define ACP_SOFT_RESET_DONE_TIME_OUT_VALUE 0x000000FF #define ACP_TIMEOUT_LOOP 0x000000FF -#define ACP_DEVS 3 +#define ACP_DEVS 4 #define ACP_SRC_ID 162 enum { @@ -316,14 +320,13 @@ static int acp_hw_init(void *handle) if (adev->acp.acp_cell == NULL) return -ENOMEM; - adev->acp.acp_res = kcalloc(4, sizeof(struct resource), GFP_KERNEL); - + adev->acp.acp_res = kcalloc(5, sizeof(struct resource), GFP_KERNEL); if (adev->acp.acp_res == NULL) { kfree(adev->acp.acp_cell); return -ENOMEM; } - i2s_pdata = kcalloc(2, sizeof(struct i2s_platform_data), GFP_KERNEL); + i2s_pdata = kcalloc(3, sizeof(struct i2s_platform_data), GFP_KERNEL); if (i2s_pdata == NULL) { kfree(adev->acp.acp_res); kfree(adev->acp.acp_cell); @@ -358,6 +361,20 @@ static int acp_hw_init(void *handle) i2s_pdata[1].i2s_reg_comp1 = ACP_I2S_COMP1_CAP_REG_OFFSET; i2s_pdata[1].i2s_reg_comp2 = ACP_I2S_COMP2_CAP_REG_OFFSET; + i2s_pdata[2].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET; + switch (adev->asic_type) { + case CHIP_STONEY: + i2s_pdata[2].quirks |= DW_I2S_QUIRK_16BIT_IDX_OVERRIDE; + break; + default: + break; + } + + i2s_pdata[2].cap = DWC_I2S_PLAY | DWC_I2S_RECORD; + i2s_pdata[2].snd_rates = SNDRV_PCM_RATE_8000_96000; + i2s_pdata[2].i2s_reg_comp1 = ACP_BT_COMP1_REG_OFFSET; + i2s_pdata[2].i2s_reg_comp2 = ACP_BT_COMP2_REG_OFFSET; + adev->acp.acp_res[0].name = "acp2x_dma"; adev->acp.acp_res[0].flags = IORESOURCE_MEM; adev->acp.acp_res[0].start = acp_base; @@ -373,13 +390,18 @@ static int acp_hw_init(void *handle) adev->acp.acp_res[2].start = acp_base + ACP_I2S_CAP_REGS_START; adev->acp.acp_res[2].end = acp_base + ACP_I2S_CAP_REGS_END; - adev->acp.acp_res[3].name = "acp2x_dma_irq"; - adev->acp.acp_res[3].flags = IORESOURCE_IRQ; - adev->acp.acp_res[3].start = amdgpu_irq_create_mapping(adev, 162); - adev->acp.acp_res[3].end = adev->acp.acp_res[3].start; + adev->acp.acp_res[3].name = "acp2x_dw_bt_i2s_play_cap"; + adev->acp.acp_res[3].flags = IORESOURCE_MEM; + adev->acp.acp_res[3].start = acp_base + ACP_BT_PLAY_REGS_START; + adev->acp.acp_res[3].end = acp_base + ACP_BT_PLAY_REGS_END; + + adev->acp.acp_res[4].name = "acp2x_dma_irq"; + adev->acp.acp_res[4].flags = IORESOURCE_IRQ; + adev->acp.acp_res[4].start = amdgpu_irq_create_mapping(adev, 162); + adev->acp.acp_res[4].end = adev->acp.acp_res[4].start; adev->acp.acp_cell[0].name = "acp_audio_dma"; - adev->acp.acp_cell[0].num_resources = 4; + adev->acp.acp_cell[0].num_resources = 5; adev->acp.acp_cell[0].resources = &adev->acp.acp_res[0]; adev->acp.acp_cell[0].platform_data = &adev->asic_type; adev->acp.acp_cell[0].pdata_size = sizeof(adev->asic_type); @@ -396,6 +418,12 @@ static int acp_hw_init(void *handle) adev->acp.acp_cell[2].platform_data = &i2s_pdata[1]; adev->acp.acp_cell[2].pdata_size = sizeof(struct i2s_platform_data); + adev->acp.acp_cell[3].name = "designware-i2s"; + adev->acp.acp_cell[3].num_resources = 1; + adev->acp.acp_cell[3].resources = &adev->acp.acp_res[3]; + adev->acp.acp_cell[3].platform_data = &i2s_pdata[2]; + adev->acp.acp_cell[3].pdata_size = sizeof(struct i2s_platform_data); + r = mfd_add_hotplug_devices(adev->acp.parent, adev->acp.acp_cell, ACP_DEVS); if (r) @@ -451,7 +479,6 @@ static int acp_hw_init(void *handle) val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET); val &= ~ACP_SOFT_RESET__SoftResetAud_MASK; cgs_write_register(adev->acp.cgs_device, mmACP_SOFT_RESET, val); - return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index e3ed08dca7b7..f8bbbb3a9504 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -243,6 +243,33 @@ int amdgpu_amdkfd_resume(struct amdgpu_device *adev) return r; } +int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev) +{ + int r = 0; + + if (adev->kfd) + r = kgd2kfd->pre_reset(adev->kfd); + + return r; +} + +int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev) +{ + int r = 0; + + if (adev->kfd) + r = kgd2kfd->post_reset(adev->kfd); + + return r; +} + +void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + + amdgpu_device_gpu_recover(adev, NULL, false); +} + int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, void **mem_obj, uint64_t *gpu_addr, void **cpu_ptr) @@ -461,6 +488,14 @@ err: return ret; } +void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + + amdgpu_dpm_switch_power_profile(adev, + PP_SMC_POWER_PROFILE_COMPUTE, !idle); +} + bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid) { if (adev->kfd) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index a8418a3f4e9d..2f379c183ed2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -119,6 +119,7 @@ int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm); int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine, uint32_t vmid, uint64_t gpu_addr, uint32_t *ib_cmd, uint32_t ib_len); +void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle); struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void); struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void); @@ -126,6 +127,12 @@ struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void); bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid); +int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev); + +int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev); + +void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd); + /* Shared API */ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, void **mem_obj, uint64_t *gpu_addr, @@ -183,6 +190,9 @@ int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd, int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info, struct dma_fence **ef); +int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, + struct kfd_vm_fault_info *info); + void amdgpu_amdkfd_gpuvm_init_mem_limits(void); void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index ea79908dac4c..ea3f698aef5e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -145,6 +145,7 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, uint32_t page_table_base); static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid); static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid); +static uint32_t read_vmid_from_vmfault_reg(struct kgd_dev *kgd); /* Because of REG_GET_FIELD() being used, we put this function in the * asic specific file. @@ -216,6 +217,10 @@ static const struct kfd2kgd_calls kfd2kgd = { .invalidate_tlbs = invalidate_tlbs, .invalidate_tlbs_vmid = invalidate_tlbs_vmid, .submit_ib = amdgpu_amdkfd_submit_ib, + .get_vm_fault_info = amdgpu_amdkfd_gpuvm_get_vm_fault_info, + .read_vmid_from_vmfault_reg = read_vmid_from_vmfault_reg, + .gpu_recover = amdgpu_amdkfd_gpu_reset, + .set_compute_idle = amdgpu_amdkfd_set_compute_idle }; struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void) @@ -571,6 +576,9 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, unsigned long flags, end_jiffies; int retry; + if (adev->in_gpu_reset) + return -EIO; + acquire_queue(kgd, pipe_id, queue_id); WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, 0); @@ -882,6 +890,9 @@ static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid) int vmid; unsigned int tmp; + if (adev->in_gpu_reset) + return -EIO; + for (vmid = 0; vmid < 16; vmid++) { if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) continue; @@ -911,3 +922,19 @@ static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid) RREG32(mmVM_INVALIDATE_RESPONSE); return 0; } + + /** + * read_vmid_from_vmfault_reg - read vmid from register + * + * adev: amdgpu_device pointer + * @vmid: vmid pointer + * read vmid from register (CIK). + */ +static uint32_t read_vmid_from_vmfault_reg(struct kgd_dev *kgd) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + uint32_t status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS); + + return REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index 19dd665e7307..f6e53e9352bd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -176,6 +176,9 @@ static const struct kfd2kgd_calls kfd2kgd = { .invalidate_tlbs = invalidate_tlbs, .invalidate_tlbs_vmid = invalidate_tlbs_vmid, .submit_ib = amdgpu_amdkfd_submit_ib, + .get_vm_fault_info = amdgpu_amdkfd_gpuvm_get_vm_fault_info, + .gpu_recover = amdgpu_amdkfd_gpu_reset, + .set_compute_idle = amdgpu_amdkfd_set_compute_idle }; struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void) @@ -568,6 +571,9 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, int retry; struct vi_mqd *m = get_mqd(mqd); + if (adev->in_gpu_reset) + return -EIO; + acquire_queue(kgd, pipe_id, queue_id); if (m->cp_hqd_vmid == 0) @@ -844,6 +850,9 @@ static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid) int vmid; unsigned int tmp; + if (adev->in_gpu_reset) + return -EIO; + for (vmid = 0; vmid < 16; vmid++) { if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) continue; 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 1db60aa5b7f0..8efedfcb9dfc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -213,6 +213,8 @@ static const struct kfd2kgd_calls kfd2kgd = { .invalidate_tlbs = invalidate_tlbs, .invalidate_tlbs_vmid = invalidate_tlbs_vmid, .submit_ib = amdgpu_amdkfd_submit_ib, + .gpu_recover = amdgpu_amdkfd_gpu_reset, + .set_compute_idle = amdgpu_amdkfd_set_compute_idle }; struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void) @@ -679,6 +681,9 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, uint32_t temp; struct v9_mqd *m = get_mqd(mqd); + if (adev->in_gpu_reset) + return -EIO; + acquire_queue(kgd, pipe_id, queue_id); if (m->cp_hqd_vmid == 0) @@ -866,6 +871,9 @@ static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid) int vmid; struct amdgpu_ring *ring = &adev->gfx.kiq.ring; + if (adev->in_gpu_reset) + return -EIO; + if (ring->ready) return invalidate_tlbs_with_kiq(adev, pasid); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index fa38a960ce00..8a707d8bbb1c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -1621,6 +1621,20 @@ bo_reserve_failed: return ret; } +int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, + struct kfd_vm_fault_info *mem) +{ + struct amdgpu_device *adev; + + adev = (struct amdgpu_device *)kgd; + if (atomic_read(&adev->gmc.vm_fault_info_updated) == 1) { + *mem = *adev->gmc.vm_fault_info; + mb(); + atomic_set(&adev->gmc.vm_fault_info_updated, 0); + } + return 0; +} + /* Evict a userptr BO by stopping the queues if necessary * * Runs in MMU notifier, may be in RECLAIM_FS context. This means it diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c index b33f1680c9a3..a028661d9e20 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c @@ -575,6 +575,7 @@ static const struct amdgpu_px_quirk amdgpu_px_quirk_list[] = { { 0x1002, 0x6900, 0x1002, 0x0124, AMDGPU_PX_QUIRK_FORCE_ATPX }, { 0x1002, 0x6900, 0x1028, 0x0812, AMDGPU_PX_QUIRK_FORCE_ATPX }, { 0x1002, 0x6900, 0x1028, 0x0813, AMDGPU_PX_QUIRK_FORCE_ATPX }, + { 0x1002, 0x6900, 0x1025, 0x125A, AMDGPU_PX_QUIRK_FORCE_ATPX }, { 0, 0, 0, 0, 0 }, }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c index 7679c068c89a..d472a2c8399f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c @@ -35,83 +35,53 @@ #define AMDGPU_BO_LIST_MAX_PRIORITY 32u #define AMDGPU_BO_LIST_NUM_BUCKETS (AMDGPU_BO_LIST_MAX_PRIORITY + 1) -static int amdgpu_bo_list_set(struct amdgpu_device *adev, - struct drm_file *filp, - struct amdgpu_bo_list *list, - struct drm_amdgpu_bo_list_entry *info, - unsigned num_entries); +static void amdgpu_bo_list_free_rcu(struct rcu_head *rcu) +{ + struct amdgpu_bo_list *list = container_of(rcu, struct amdgpu_bo_list, + rhead); + + kvfree(list); +} -static void amdgpu_bo_list_release_rcu(struct kref *ref) +static void amdgpu_bo_list_free(struct kref *ref) { - unsigned i; struct amdgpu_bo_list *list = container_of(ref, struct amdgpu_bo_list, refcount); + struct amdgpu_bo_list_entry *e; - for (i = 0; i < list->num_entries; ++i) - amdgpu_bo_unref(&list->array[i].robj); + amdgpu_bo_list_for_each_entry(e, list) + amdgpu_bo_unref(&e->robj); - mutex_destroy(&list->lock); - kvfree(list->array); - kfree_rcu(list, rhead); + call_rcu(&list->rhead, amdgpu_bo_list_free_rcu); } -int amdgpu_bo_list_create(struct amdgpu_device *adev, - struct drm_file *filp, - struct drm_amdgpu_bo_list_entry *info, - unsigned num_entries, - struct amdgpu_bo_list **list_out) +int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp, + struct drm_amdgpu_bo_list_entry *info, + unsigned num_entries, struct amdgpu_bo_list **result) { + unsigned last_entry = 0, first_userptr = num_entries; + struct amdgpu_bo_list_entry *array; struct amdgpu_bo_list *list; + uint64_t total_size = 0; + size_t size; + unsigned i; int r; + if (num_entries > SIZE_MAX / sizeof(struct amdgpu_bo_list_entry)) + return -EINVAL; - list = kzalloc(sizeof(struct amdgpu_bo_list), GFP_KERNEL); + size = sizeof(struct amdgpu_bo_list); + size += num_entries * sizeof(struct amdgpu_bo_list_entry); + list = kvmalloc(size, GFP_KERNEL); if (!list) return -ENOMEM; - /* initialize bo list*/ - mutex_init(&list->lock); kref_init(&list->refcount); - r = amdgpu_bo_list_set(adev, filp, list, info, num_entries); - if (r) { - kfree(list); - return r; - } - - *list_out = list; - return 0; -} - -static void amdgpu_bo_list_destroy(struct amdgpu_fpriv *fpriv, int id) -{ - struct amdgpu_bo_list *list; + list->gds_obj = adev->gds.gds_gfx_bo; + list->gws_obj = adev->gds.gws_gfx_bo; + list->oa_obj = adev->gds.oa_gfx_bo; - mutex_lock(&fpriv->bo_list_lock); - list = idr_remove(&fpriv->bo_list_handles, id); - mutex_unlock(&fpriv->bo_list_lock); - if (list) - kref_put(&list->refcount, amdgpu_bo_list_release_rcu); -} - -static int amdgpu_bo_list_set(struct amdgpu_device *adev, - struct drm_file *filp, - struct amdgpu_bo_list *list, - struct drm_amdgpu_bo_list_entry *info, - unsigned num_entries) -{ - struct amdgpu_bo_list_entry *array; - struct amdgpu_bo *gds_obj = adev->gds.gds_gfx_bo; - struct amdgpu_bo *gws_obj = adev->gds.gws_gfx_bo; - struct amdgpu_bo *oa_obj = adev->gds.oa_gfx_bo; - - unsigned last_entry = 0, first_userptr = num_entries; - unsigned i; - int r; - unsigned long total_size = 0; - - array = kvmalloc_array(num_entries, sizeof(struct amdgpu_bo_list_entry), GFP_KERNEL); - if (!array) - return -ENOMEM; + array = amdgpu_bo_list_array_entry(list, 0); memset(array, 0, num_entries * sizeof(struct amdgpu_bo_list_entry)); for (i = 0; i < num_entries; ++i) { @@ -148,59 +118,56 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev, entry->tv.shared = !entry->robj->prime_shared_count; if (entry->robj->preferred_domains == AMDGPU_GEM_DOMAIN_GDS) - gds_obj = entry->robj; + list->gds_obj = entry->robj; if (entry->robj->preferred_domains == AMDGPU_GEM_DOMAIN_GWS) - gws_obj = entry->robj; + list->gws_obj = entry->robj; if (entry->robj->preferred_domains == AMDGPU_GEM_DOMAIN_OA) - oa_obj = entry->robj; + list->oa_obj = entry->robj; total_size += amdgpu_bo_size(entry->robj); trace_amdgpu_bo_list_set(list, entry->robj); } - for (i = 0; i < list->num_entries; ++i) - amdgpu_bo_unref(&list->array[i].robj); - - kvfree(list->array); - - list->gds_obj = gds_obj; - list->gws_obj = gws_obj; - list->oa_obj = oa_obj; list->first_userptr = first_userptr; - list->array = array; list->num_entries = num_entries; trace_amdgpu_cs_bo_status(list->num_entries, total_size); + + *result = list; return 0; error_free: while (i--) amdgpu_bo_unref(&array[i].robj); - kvfree(array); + kvfree(list); return r; + } -struct amdgpu_bo_list * -amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id) +static void amdgpu_bo_list_destroy(struct amdgpu_fpriv *fpriv, int id) { - struct amdgpu_bo_list *result; + struct amdgpu_bo_list *list; + mutex_lock(&fpriv->bo_list_lock); + list = idr_remove(&fpriv->bo_list_handles, id); + mutex_unlock(&fpriv->bo_list_lock); + if (list) + kref_put(&list->refcount, amdgpu_bo_list_free); +} + +int amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id, + struct amdgpu_bo_list **result) +{ rcu_read_lock(); - result = idr_find(&fpriv->bo_list_handles, id); + *result = idr_find(&fpriv->bo_list_handles, id); - if (result) { - if (kref_get_unless_zero(&result->refcount)) { - rcu_read_unlock(); - mutex_lock(&result->lock); - } else { - rcu_read_unlock(); - result = NULL; - } - } else { + if (*result && kref_get_unless_zero(&(*result)->refcount)) { rcu_read_unlock(); + return 0; } - return result; + rcu_read_unlock(); + return -ENOENT; } void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, @@ -211,6 +178,7 @@ void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, * concatenated in descending order. */ struct list_head bucket[AMDGPU_BO_LIST_NUM_BUCKETS]; + struct amdgpu_bo_list_entry *e; unsigned i; for (i = 0; i < AMDGPU_BO_LIST_NUM_BUCKETS; i++) @@ -221,14 +189,13 @@ void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, * in the list, the sort mustn't change the ordering of buffers * with the same priority, i.e. it must be stable. */ - for (i = 0; i < list->num_entries; i++) { - unsigned priority = list->array[i].priority; + amdgpu_bo_list_for_each_entry(e, list) { + unsigned priority = e->priority; - if (!list->array[i].robj->parent) - list_add_tail(&list->array[i].tv.head, - &bucket[priority]); + if (!e->robj->parent) + list_add_tail(&e->tv.head, &bucket[priority]); - list->array[i].user_pages = NULL; + e->user_pages = NULL; } /* Connect the sorted buckets in the output list. */ @@ -238,20 +205,7 @@ void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, void amdgpu_bo_list_put(struct amdgpu_bo_list *list) { - mutex_unlock(&list->lock); - kref_put(&list->refcount, amdgpu_bo_list_release_rcu); -} - -void amdgpu_bo_list_free(struct amdgpu_bo_list *list) -{ - unsigned i; - - for (i = 0; i < list->num_entries; ++i) - amdgpu_bo_unref(&list->array[i].robj); - - mutex_destroy(&list->lock); - kvfree(list->array); - kfree(list); + kref_put(&list->refcount, amdgpu_bo_list_free); } int amdgpu_bo_create_list_entry_array(struct drm_amdgpu_bo_list_in *in, @@ -304,7 +258,7 @@ int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, union drm_amdgpu_bo_list *args = data; uint32_t handle = args->in.list_handle; struct drm_amdgpu_bo_list_entry *info = NULL; - struct amdgpu_bo_list *list; + struct amdgpu_bo_list *list, *old; int r; r = amdgpu_bo_create_list_entry_array(&args->in, &info); @@ -322,7 +276,7 @@ int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, r = idr_alloc(&fpriv->bo_list_handles, list, 1, 0, GFP_KERNEL); mutex_unlock(&fpriv->bo_list_lock); if (r < 0) { - amdgpu_bo_list_free(list); + amdgpu_bo_list_put(list); return r; } @@ -335,17 +289,22 @@ int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, break; case AMDGPU_BO_LIST_OP_UPDATE: - r = -ENOENT; - list = amdgpu_bo_list_get(fpriv, handle); - if (!list) + r = amdgpu_bo_list_create(adev, filp, info, args->in.bo_number, + &list); + if (r) goto error_free; - r = amdgpu_bo_list_set(adev, filp, list, info, - args->in.bo_number); - amdgpu_bo_list_put(list); - if (r) + mutex_lock(&fpriv->bo_list_lock); + old = idr_replace(&fpriv->bo_list_handles, list, handle); + mutex_unlock(&fpriv->bo_list_lock); + + if (IS_ERR(old)) { + amdgpu_bo_list_put(list); + r = PTR_ERR(old); goto error_free; + } + amdgpu_bo_list_put(old); break; default: diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h new file mode 100644 index 000000000000..61b089768e1c --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h @@ -0,0 +1,85 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __AMDGPU_BO_LIST_H__ +#define __AMDGPU_BO_LIST_H__ + +#include <drm/ttm/ttm_execbuf_util.h> +#include <drm/amdgpu_drm.h> + +struct amdgpu_device; +struct amdgpu_bo; +struct amdgpu_bo_va; +struct amdgpu_fpriv; + +struct amdgpu_bo_list_entry { + struct amdgpu_bo *robj; + struct ttm_validate_buffer tv; + struct amdgpu_bo_va *bo_va; + uint32_t priority; + struct page **user_pages; + int user_invalidated; +}; + +struct amdgpu_bo_list { + struct rcu_head rhead; + struct kref refcount; + struct amdgpu_bo *gds_obj; + struct amdgpu_bo *gws_obj; + struct amdgpu_bo *oa_obj; + unsigned first_userptr; + unsigned num_entries; +}; + +int amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id, + struct amdgpu_bo_list **result); +void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, + struct list_head *validated); +void amdgpu_bo_list_put(struct amdgpu_bo_list *list); +int amdgpu_bo_create_list_entry_array(struct drm_amdgpu_bo_list_in *in, + struct drm_amdgpu_bo_list_entry **info_param); + +int amdgpu_bo_list_create(struct amdgpu_device *adev, + struct drm_file *filp, + struct drm_amdgpu_bo_list_entry *info, + unsigned num_entries, + struct amdgpu_bo_list **list); + +static inline struct amdgpu_bo_list_entry * +amdgpu_bo_list_array_entry(struct amdgpu_bo_list *list, unsigned index) +{ + struct amdgpu_bo_list_entry *array = (void *)&list[1]; + + return &array[index]; +} + +#define amdgpu_bo_list_for_each_entry(e, list) \ + for (e = amdgpu_bo_list_array_entry(list, 0); \ + e != amdgpu_bo_list_array_entry(list, (list)->num_entries); \ + ++e) + +#define amdgpu_bo_list_for_each_userptr_entry(e, list) \ + for (e = amdgpu_bo_list_array_entry(list, (list)->first_userptr); \ + e != amdgpu_bo_list_array_entry(list, (list)->num_entries); \ + ++e) + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 178d9ce4eba1..502b94fb116a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -561,28 +561,38 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, union drm_amdgpu_cs *cs) { struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_bo_list_entry *e; struct list_head duplicates; - unsigned i, tries = 10; struct amdgpu_bo *gds; struct amdgpu_bo *gws; struct amdgpu_bo *oa; + unsigned tries = 10; int r; INIT_LIST_HEAD(&p->validated); /* p->bo_list could already be assigned if AMDGPU_CHUNK_ID_BO_HANDLES is present */ - if (!p->bo_list) - p->bo_list = amdgpu_bo_list_get(fpriv, cs->in.bo_list_handle); - else - mutex_lock(&p->bo_list->lock); + if (cs->in.bo_list_handle) { + if (p->bo_list) + return -EINVAL; - if (p->bo_list) { - amdgpu_bo_list_get_list(p->bo_list, &p->validated); - if (p->bo_list->first_userptr != p->bo_list->num_entries) - p->mn = amdgpu_mn_get(p->adev, AMDGPU_MN_TYPE_GFX); + r = amdgpu_bo_list_get(fpriv, cs->in.bo_list_handle, + &p->bo_list); + if (r) + return r; + } else if (!p->bo_list) { + /* Create a empty bo_list when no handle is provided */ + r = amdgpu_bo_list_create(p->adev, p->filp, NULL, 0, + &p->bo_list); + if (r) + return r; } + amdgpu_bo_list_get_list(p->bo_list, &p->validated); + if (p->bo_list->first_userptr != p->bo_list->num_entries) + p->mn = amdgpu_mn_get(p->adev, AMDGPU_MN_TYPE_GFX); + INIT_LIST_HEAD(&duplicates); amdgpu_vm_get_pd_bo(&fpriv->vm, &p->validated, &p->vm_pd); @@ -591,7 +601,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, while (1) { struct list_head need_pages; - unsigned i; r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true, &duplicates); @@ -601,17 +610,9 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, goto error_free_pages; } - /* Without a BO list we don't have userptr BOs */ - if (!p->bo_list) - break; - INIT_LIST_HEAD(&need_pages); - for (i = p->bo_list->first_userptr; - i < p->bo_list->num_entries; ++i) { - struct amdgpu_bo *bo; - - e = &p->bo_list->array[i]; - bo = e->robj; + amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) { + struct amdgpu_bo *bo = e->robj; if (amdgpu_ttm_tt_userptr_invalidated(bo->tbo.ttm, &e->user_invalidated) && e->user_pages) { @@ -703,23 +704,12 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, amdgpu_cs_report_moved_bytes(p->adev, p->bytes_moved, p->bytes_moved_vis); - if (p->bo_list) { - struct amdgpu_vm *vm = &fpriv->vm; - unsigned i; - - gds = p->bo_list->gds_obj; - gws = p->bo_list->gws_obj; - oa = p->bo_list->oa_obj; - for (i = 0; i < p->bo_list->num_entries; i++) { - struct amdgpu_bo *bo = p->bo_list->array[i].robj; + gds = p->bo_list->gds_obj; + gws = p->bo_list->gws_obj; + oa = p->bo_list->oa_obj; - p->bo_list->array[i].bo_va = amdgpu_vm_bo_find(vm, bo); - } - } else { - gds = p->adev->gds.gds_gfx_bo; - gws = p->adev->gds.gws_gfx_bo; - oa = p->adev->gds.oa_gfx_bo; - } + amdgpu_bo_list_for_each_entry(e, p->bo_list) + e->bo_va = amdgpu_vm_bo_find(vm, e->robj); if (gds) { p->job->gds_base = amdgpu_bo_gpu_offset(gds); @@ -747,18 +737,13 @@ error_validate: error_free_pages: - if (p->bo_list) { - for (i = p->bo_list->first_userptr; - i < p->bo_list->num_entries; ++i) { - e = &p->bo_list->array[i]; - - if (!e->user_pages) - continue; + amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) { + if (!e->user_pages) + continue; - release_pages(e->user_pages, - e->robj->tbo.ttm->num_pages); - kvfree(e->user_pages); - } + release_pages(e->user_pages, + e->robj->tbo.ttm->num_pages); + kvfree(e->user_pages); } return r; @@ -820,12 +805,13 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error, static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) { - struct amdgpu_device *adev = p->adev; struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + struct amdgpu_device *adev = p->adev; struct amdgpu_vm *vm = &fpriv->vm; + struct amdgpu_bo_list_entry *e; struct amdgpu_bo_va *bo_va; struct amdgpu_bo *bo; - int i, r; + int r; r = amdgpu_vm_clear_freed(adev, vm, NULL); if (r) @@ -855,29 +841,26 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) return r; } - if (p->bo_list) { - for (i = 0; i < p->bo_list->num_entries; i++) { - struct dma_fence *f; - - /* ignore duplicates */ - bo = p->bo_list->array[i].robj; - if (!bo) - continue; + amdgpu_bo_list_for_each_entry(e, p->bo_list) { + struct dma_fence *f; - bo_va = p->bo_list->array[i].bo_va; - if (bo_va == NULL) - continue; + /* ignore duplicates */ + bo = e->robj; + if (!bo) + continue; - r = amdgpu_vm_bo_update(adev, bo_va, false); - if (r) - return r; + bo_va = e->bo_va; + if (bo_va == NULL) + continue; - f = bo_va->last_pt_update; - r = amdgpu_sync_fence(adev, &p->job->sync, f, false); - if (r) - return r; - } + r = amdgpu_vm_bo_update(adev, bo_va, false); + if (r) + return r; + f = bo_va->last_pt_update; + r = amdgpu_sync_fence(adev, &p->job->sync, f, false); + if (r) + return r; } r = amdgpu_vm_handle_moved(adev, vm); @@ -892,15 +875,14 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) if (r) return r; - if (amdgpu_vm_debug && p->bo_list) { + if (amdgpu_vm_debug) { /* Invalidate all BOs to test for userspace bugs */ - for (i = 0; i < p->bo_list->num_entries; i++) { + amdgpu_bo_list_for_each_entry(e, p->bo_list) { /* ignore duplicates */ - bo = p->bo_list->array[i].robj; - if (!bo) + if (!e->robj) continue; - amdgpu_vm_bo_invalidate(adev, bo, false); + amdgpu_vm_bo_invalidate(adev, e->robj, false); } } @@ -916,7 +898,7 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, int r; /* Only for UVD/VCE VM emulation */ - if (p->ring->funcs->parse_cs) { + if (p->ring->funcs->parse_cs || p->ring->funcs->patch_cs_in_place) { unsigned i, j; for (i = 0, j = 0; i < p->nchunks && j < p->job->num_ibs; i++) { @@ -957,12 +939,20 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, offset = m->start * AMDGPU_GPU_PAGE_SIZE; kptr += va_start - offset; - memcpy(ib->ptr, kptr, chunk_ib->ib_bytes); - amdgpu_bo_kunmap(aobj); - - r = amdgpu_ring_parse_cs(ring, p, j); - if (r) - return r; + if (p->ring->funcs->parse_cs) { + memcpy(ib->ptr, kptr, chunk_ib->ib_bytes); + amdgpu_bo_kunmap(aobj); + + r = amdgpu_ring_parse_cs(ring, p, j); + if (r) + return r; + } else { + ib->ptr = (uint32_t *)kptr; + r = amdgpu_ring_patch_cs_in_place(ring, p, j); + amdgpu_bo_kunmap(aobj); + if (r) + return r; + } j++; } @@ -1207,25 +1197,23 @@ static void amdgpu_cs_post_dependencies(struct amdgpu_cs_parser *p) static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, union drm_amdgpu_cs *cs) { + struct amdgpu_fpriv *fpriv = p->filp->driver_priv; struct amdgpu_ring *ring = p->ring; struct drm_sched_entity *entity = &p->ctx->rings[ring->idx].entity; enum drm_sched_priority priority; + struct amdgpu_bo_list_entry *e; struct amdgpu_job *job; - unsigned i; uint64_t seq; int r; amdgpu_mn_lock(p->mn); - if (p->bo_list) { - for (i = p->bo_list->first_userptr; - i < p->bo_list->num_entries; ++i) { - struct amdgpu_bo *bo = p->bo_list->array[i].robj; - - if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm)) { - amdgpu_mn_unlock(p->mn); - return -ERESTARTSYS; - } + amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) { + struct amdgpu_bo *bo = e->robj; + + if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm)) { + amdgpu_mn_unlock(p->mn); + return -ERESTARTSYS; } } @@ -1259,6 +1247,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, amdgpu_job_free_resources(job); trace_amdgpu_cs_ioctl(job); + amdgpu_vm_bo_trace_cs(&fpriv->vm, &p->ticket); priority = job->base.s_priority; drm_sched_entity_push_job(&job->base, entity); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index ec53d8f96d06..e839470880d7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2843,6 +2843,9 @@ int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon) if (r) return r; + /* Make sure IB tests flushed */ + flush_delayed_work(&adev->late_init_work); + /* blat the mode back in */ if (fbcon) { if (!amdgpu_device_has_dc_support(adev)) { @@ -3307,6 +3310,9 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, atomic_inc(&adev->gpu_reset_counter); adev->in_gpu_reset = 1; + /* Block kfd */ + amdgpu_amdkfd_pre_reset(adev); + /* block TTM */ resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev); @@ -3322,7 +3328,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, if (job && job->base.sched == &ring->sched) continue; - drm_sched_hw_job_reset(&ring->sched, &job->base); + drm_sched_hw_job_reset(&ring->sched, job ? &job->base : NULL); /* after all hw jobs are reset, hw fence is meaningless, so force_completion */ amdgpu_fence_driver_force_completion(ring); @@ -3363,6 +3369,8 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, dev_info(adev->dev, "GPU reset(%d) succeeded!\n",atomic_read(&adev->gpu_reset_counter)); } + /*unlock kfd */ + amdgpu_amdkfd_post_reset(adev); amdgpu_vf_error_trans_all(adev); adev->in_gpu_reset = 0; mutex_unlock(&adev->lock_reset); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index 6cb4948233cb..bb5a47a45790 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -105,6 +105,8 @@ struct amdgpu_gmc { /* protects concurrent invalidation */ spinlock_t invalidate_lock; bool translate_further; + struct kfd_vm_fault_info *vm_fault_info; + atomic_t vm_fault_info_updated; const struct amdgpu_gmc_funcs *gmc_funcs; }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index c7dce14fd47d..bd98cc5fb97b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -286,7 +286,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file struct drm_crtc *crtc; uint32_t ui32 = 0; uint64_t ui64 = 0; - int i, found; + int i, j, found; int ui32_size = sizeof(ui32); if (!info->return_size || !info->return_pointer) @@ -348,7 +348,11 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file break; case AMDGPU_HW_IP_UVD: type = AMD_IP_BLOCK_TYPE_UVD; - ring_mask |= adev->uvd.inst[0].ring.ready; + for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (adev->uvd.harvest_config & (1 << i)) + continue; + ring_mask |= adev->uvd.inst[i].ring.ready; + } ib_start_alignment = 64; ib_size_alignment = 64; break; @@ -361,9 +365,12 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file break; case AMDGPU_HW_IP_UVD_ENC: type = AMD_IP_BLOCK_TYPE_UVD; - for (i = 0; i < adev->uvd.num_enc_rings; i++) - ring_mask |= - adev->uvd.inst[0].ring_enc[i].ready << i; + for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (adev->uvd.harvest_config & (1 << i)) + continue; + for (j = 0; j < adev->uvd.num_enc_rings; j++) + ring_mask |= adev->uvd.inst[i].ring_enc[j].ready << j; + } ib_start_alignment = 64; ib_size_alignment = 64; break; @@ -960,7 +967,7 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, amdgpu_bo_unref(&pd); idr_for_each_entry(&fpriv->bo_list_handles, list, handle) - amdgpu_bo_list_free(list); + amdgpu_bo_list_put(list); idr_destroy(&fpriv->bo_list_handles); mutex_destroy(&fpriv->bo_list_lock); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 21bfa2d8039e..b0e14a3d54ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -825,7 +825,7 @@ struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo) if (bo == NULL) return NULL; - ttm_bo_reference(&bo->tbo); + ttm_bo_get(&bo->tbo); return bo; } @@ -843,9 +843,8 @@ void amdgpu_bo_unref(struct amdgpu_bo **bo) return; tbo = &((*bo)->tbo); - ttm_bo_unref(&tbo); - if (tbo == NULL) - *bo = NULL; + ttm_bo_put(tbo); + *bo = NULL; } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c index 15a1192c1ec5..8f98629fbe59 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c @@ -31,7 +31,7 @@ #include <linux/power_supply.h> #include <linux/hwmon.h> #include <linux/hwmon-sysfs.h> - +#include <linux/nospec.h> static int amdgpu_debugfs_pm_init(struct amdgpu_device *adev); @@ -403,6 +403,7 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev, count = -EINVAL; goto fail; } + idx = array_index_nospec(idx, ARRAY_SIZE(data.states)); amdgpu_dpm_get_pp_num_states(adev, &data); state = data.states[idx]; @@ -1185,7 +1186,7 @@ static ssize_t amdgpu_hwmon_show_vddnb(struct device *dev, int r, size = sizeof(vddnb); /* only APUs have vddnb */ - if (adev->flags & AMD_IS_APU) + if (!(adev->flags & AMD_IS_APU)) return -EINVAL; /* Can't get voltage when the card is off */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c index d8357290ad09..a172bba32b45 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c @@ -214,7 +214,7 @@ int amdgpu_queue_mgr_map(struct amdgpu_device *adev, u32 hw_ip, u32 instance, u32 ring, struct amdgpu_ring **out_ring) { - int r, ip_num_rings; + int i, r, ip_num_rings = 0; struct amdgpu_queue_mapper *mapper = &mgr->mapper[hw_ip]; if (!adev || !mgr || !out_ring) @@ -243,14 +243,21 @@ int amdgpu_queue_mgr_map(struct amdgpu_device *adev, ip_num_rings = adev->sdma.num_instances; break; case AMDGPU_HW_IP_UVD: - ip_num_rings = adev->uvd.num_uvd_inst; + for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (!(adev->uvd.harvest_config & (1 << i))) + ip_num_rings++; + } break; case AMDGPU_HW_IP_VCE: ip_num_rings = adev->vce.num_rings; break; case AMDGPU_HW_IP_UVD_ENC: + for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (!(adev->uvd.harvest_config & (1 << i))) + ip_num_rings++; + } ip_num_rings = - adev->uvd.num_enc_rings * adev->uvd.num_uvd_inst; + adev->uvd.num_enc_rings * ip_num_rings; break; case AMDGPU_HW_IP_VCN_DEC: ip_num_rings = 1; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 5018c0b6bf1a..d242b9a51e90 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -123,6 +123,7 @@ struct amdgpu_ring_funcs { void (*set_wptr)(struct amdgpu_ring *ring); /* validating and patching of IBs */ int (*parse_cs)(struct amdgpu_cs_parser *p, uint32_t ib_idx); + int (*patch_cs_in_place)(struct amdgpu_cs_parser *p, uint32_t ib_idx); /* constants to calculate how many DW are needed for an emit */ unsigned emit_frame_size; unsigned emit_ib_size; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index 11f262f15200..7206a0025b17 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -314,6 +314,11 @@ DEFINE_EVENT(amdgpu_vm_mapping, amdgpu_vm_bo_mapping, TP_ARGS(mapping) ); +DEFINE_EVENT(amdgpu_vm_mapping, amdgpu_vm_bo_cs, + TP_PROTO(struct amdgpu_bo_va_mapping *mapping), + TP_ARGS(mapping) +); + TRACE_EVENT(amdgpu_vm_set_ptes, TP_PROTO(uint64_t pe, uint64_t addr, unsigned count, uint32_t incr, uint64_t flags), diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 8c4358e36c87..fcf421263fd9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -92,11 +92,9 @@ static void amdgpu_ttm_mem_global_release(struct drm_global_reference *ref) } /** - * amdgpu_ttm_global_init - Initialize global TTM memory reference - * structures. + * amdgpu_ttm_global_init - Initialize global TTM memory reference structures. * - * @adev: AMDGPU device for which the global structures need to be - * registered. + * @adev: AMDGPU device for which the global structures need to be registered. * * This is called as part of the AMDGPU ttm init from amdgpu_ttm_init() * during bring up. @@ -162,13 +160,12 @@ static int amdgpu_invalidate_caches(struct ttm_bo_device *bdev, uint32_t flags) } /** - * amdgpu_init_mem_type - Initialize a memory manager for a specific - * type of memory request. + * amdgpu_init_mem_type - Initialize a memory manager for a specific type of + * memory request. * - * @bdev: The TTM BO device object (contains a reference to - * amdgpu_device) - * @type: The type of memory requested - * @man: + * @bdev: The TTM BO device object (contains a reference to amdgpu_device) + * @type: The type of memory requested + * @man: The memory type manager for each domain * * This is called by ttm_bo_init_mm() when a buffer object is being * initialized. @@ -292,8 +289,8 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, /** * amdgpu_verify_access - Verify access for a mmap call * - * @bo: The buffer object to map - * @filp: The file pointer from the process performing the mmap + * @bo: The buffer object to map + * @filp: The file pointer from the process performing the mmap * * This is called by ttm_bo_mmap() to verify whether a process * has the right to mmap a BO to their process space. @@ -318,11 +315,10 @@ static int amdgpu_verify_access(struct ttm_buffer_object *bo, struct file *filp) /** * amdgpu_move_null - Register memory for a buffer object * - * @bo: The bo to assign the memory to - * @new_mem: The memory to be assigned. + * @bo: The bo to assign the memory to + * @new_mem: The memory to be assigned. * - * Assign the memory from new_mem to the memory of the buffer object - * bo. + * Assign the memory from new_mem to the memory of the buffer object bo. */ static void amdgpu_move_null(struct ttm_buffer_object *bo, struct ttm_mem_reg *new_mem) @@ -335,8 +331,12 @@ static void amdgpu_move_null(struct ttm_buffer_object *bo, } /** - * amdgpu_mm_node_addr - Compute the GPU relative offset of a GTT - * buffer. + * amdgpu_mm_node_addr - Compute the GPU relative offset of a GTT buffer. + * + * @bo: The bo to assign the memory to. + * @mm_node: Memory manager node for drm allocator. + * @mem: The region where the bo resides. + * */ static uint64_t amdgpu_mm_node_addr(struct ttm_buffer_object *bo, struct drm_mm_node *mm_node, @@ -352,10 +352,12 @@ static uint64_t amdgpu_mm_node_addr(struct ttm_buffer_object *bo, } /** - * amdgpu_find_mm_node - Helper function finds the drm_mm_node - * corresponding to @offset. It also modifies - * the offset to be within the drm_mm_node - * returned + * amdgpu_find_mm_node - Helper function finds the drm_mm_node corresponding to + * @offset. It also modifies the offset to be within the drm_mm_node returned + * + * @mem: The region where the bo resides. + * @offset: The offset that drm_mm_node is used for finding. + * */ static struct drm_mm_node *amdgpu_find_mm_node(struct ttm_mem_reg *mem, unsigned long *offset) @@ -497,8 +499,8 @@ error: /** * amdgpu_move_blit - Copy an entire buffer to another buffer * - * This is a helper called by amdgpu_bo_move() and - * amdgpu_move_vram_ram() to help move buffers to and from VRAM. + * This is a helper called by amdgpu_bo_move() and amdgpu_move_vram_ram() to + * help move buffers to and from VRAM. */ static int amdgpu_move_blit(struct ttm_buffer_object *bo, bool evict, bool no_wait_gpu, @@ -580,7 +582,7 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, bool evict, } /* blit VRAM to GTT */ - r = amdgpu_move_blit(bo, true, ctx->no_wait_gpu, &tmp_mem, old_mem); + r = amdgpu_move_blit(bo, evict, ctx->no_wait_gpu, &tmp_mem, old_mem); if (unlikely(r)) { goto out_cleanup; } @@ -632,7 +634,7 @@ static int amdgpu_move_ram_vram(struct ttm_buffer_object *bo, bool evict, } /* copy to VRAM */ - r = amdgpu_move_blit(bo, true, ctx->no_wait_gpu, new_mem, old_mem); + r = amdgpu_move_blit(bo, evict, ctx->no_wait_gpu, new_mem, old_mem); if (unlikely(r)) { goto out_cleanup; } @@ -794,8 +796,8 @@ struct amdgpu_ttm_tt { }; /** - * amdgpu_ttm_tt_get_user_pages - Pin pages of memory pointed to - * by a USERPTR pointer to memory + * amdgpu_ttm_tt_get_user_pages - Pin pages of memory pointed to by a USERPTR + * pointer to memory * * Called by amdgpu_gem_userptr_ioctl() and amdgpu_cs_parser_bos(). * This provides a wrapper around the get_user_pages() call to provide @@ -818,8 +820,10 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) down_read(&mm->mmap_sem); if (gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) { - /* check that we only use anonymous memory - to prevent problems with writeback */ + /* + * check that we only use anonymous memory to prevent problems + * with writeback + */ unsigned long end = gtt->userptr + ttm->num_pages * PAGE_SIZE; struct vm_area_struct *vma; @@ -870,10 +874,9 @@ release_pages: } /** - * amdgpu_ttm_tt_set_user_pages - Copy pages in, putting old pages - * as necessary. + * amdgpu_ttm_tt_set_user_pages - Copy pages in, putting old pages as necessary. * - * Called by amdgpu_cs_list_validate(). This creates the page list + * Called by amdgpu_cs_list_validate(). This creates the page list * that backs user memory and will ultimately be mapped into the device * address space. */ @@ -915,8 +918,7 @@ void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm) } /** - * amdgpu_ttm_tt_pin_userptr - prepare the sg table with the - * user pages + * amdgpu_ttm_tt_pin_userptr - prepare the sg table with the user pages * * Called by amdgpu_ttm_backend_bind() **/ @@ -1295,8 +1297,8 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm) } /** - * amdgpu_ttm_tt_set_userptr - Initialize userptr GTT ttm_tt - * for the current task + * amdgpu_ttm_tt_set_userptr - Initialize userptr GTT ttm_tt for the current + * task * * @ttm: The ttm_tt object to bind this userptr object to * @addr: The address in the current tasks VM space to use @@ -1346,9 +1348,8 @@ struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm) } /** - * amdgpu_ttm_tt_affect_userptr - Determine if a ttm_tt object lays - * inside an address range for the - * current task. + * amdgpu_ttm_tt_affect_userptr - Determine if a ttm_tt object lays inside an + * address range for the current task. * */ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, @@ -1386,8 +1387,7 @@ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, } /** - * amdgpu_ttm_tt_userptr_invalidated - Has the ttm_tt object been - * invalidated? + * amdgpu_ttm_tt_userptr_invalidated - Has the ttm_tt object been invalidated? */ bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, int *last_invalidated) @@ -1400,10 +1400,8 @@ bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, } /** - * amdgpu_ttm_tt_userptr_needs_pages - Have the pages backing this - * ttm_tt object been invalidated - * since the last time they've - * been set? + * amdgpu_ttm_tt_userptr_needs_pages - Have the pages backing this ttm_tt object + * been invalidated since the last time they've been set? */ bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm) { @@ -1459,13 +1457,12 @@ uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, } /** - * amdgpu_ttm_bo_eviction_valuable - Check to see if we can evict - * a buffer object. + * amdgpu_ttm_bo_eviction_valuable - Check to see if we can evict a buffer + * object. * - * Return true if eviction is sensible. Called by - * ttm_mem_evict_first() on behalf of ttm_bo_mem_force_space() - * which tries to evict buffer objects until it can find space - * for a new object and by ttm_bo_force_list_clean() which is + * Return true if eviction is sensible. Called by ttm_mem_evict_first() on + * behalf of ttm_bo_mem_force_space() which tries to evict buffer objects until + * it can find space for a new object and by ttm_bo_force_list_clean() which is * used to clean out a memory space. */ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, @@ -1515,8 +1512,7 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, } /** - * amdgpu_ttm_access_memory - Read or Write memory that backs a - * buffer object. + * amdgpu_ttm_access_memory - Read or Write memory that backs a buffer object. * * @bo: The buffer object to read/write * @offset: Offset into buffer object @@ -1704,8 +1700,8 @@ error_create: return r; } /** - * amdgpu_ttm_init - Init the memory management (ttm) as well as - * various gtt/vram related fields. + * amdgpu_ttm_init - Init the memory management (ttm) as well as various + * gtt/vram related fields. * * This initializes all of the memory space pools that the TTM layer * will need such as the GTT space (system memory mapped to the device), @@ -1856,8 +1852,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) } /** - * amdgpu_ttm_late_init - Handle any late initialization for - * amdgpu_ttm + * amdgpu_ttm_late_init - Handle any late initialization for amdgpu_ttm */ void amdgpu_ttm_late_init(struct amdgpu_device *adev) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index fca86d71fafc..632fa5980ff4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -255,7 +255,8 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) bo_size += AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8); for (j = 0; j < adev->uvd.num_uvd_inst; j++) { - + if (adev->uvd.harvest_config & (1 << j)) + continue; r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, &adev->uvd.inst[j].vcpu_bo, &adev->uvd.inst[j].gpu_addr, &adev->uvd.inst[j].cpu_addr); @@ -308,6 +309,8 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev) drm_sched_entity_destroy(&adev->uvd.entity); for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { + if (adev->uvd.harvest_config & (1 << j)) + continue; kfree(adev->uvd.inst[j].saved_bo); amdgpu_bo_free_kernel(&adev->uvd.inst[j].vcpu_bo, @@ -343,6 +346,8 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev) } for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { + if (adev->uvd.harvest_config & (1 << j)) + continue; if (adev->uvd.inst[j].vcpu_bo == NULL) continue; @@ -365,6 +370,8 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev) int i; for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (adev->uvd.harvest_config & (1 << i)) + continue; if (adev->uvd.inst[i].vcpu_bo == NULL) return -EINVAL; @@ -1159,6 +1166,8 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work) unsigned fences = 0, i, j; for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring); for (j = 0; j < adev->uvd.num_enc_rings; ++j) { fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring_enc[j]); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h index 66872286ab12..33c5f806f925 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h @@ -48,6 +48,9 @@ struct amdgpu_uvd_inst { uint32_t srbm_soft_reset; }; +#define AMDGPU_UVD_HARVEST_UVD0 (1 << 0) +#define AMDGPU_UVD_HARVEST_UVD1 (1 << 1) + struct amdgpu_uvd { const struct firmware *fw; /* UVD firmware */ unsigned fw_version; @@ -61,6 +64,7 @@ struct amdgpu_uvd { atomic_t handles[AMDGPU_MAX_UVD_HANDLES]; struct drm_sched_entity entity; struct delayed_work idle_work; + unsigned harvest_config; }; int amdgpu_uvd_sw_init(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 5d7d7900ccab..ece0ac703e27 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -495,11 +495,12 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, eaddr = eaddr & ((1 << shift) - 1); flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; + if (vm->root.base.bo->shadow) + flags |= AMDGPU_GEM_CREATE_SHADOW; if (vm->use_cpu_for_update) flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; else - flags |= (AMDGPU_GEM_CREATE_NO_CPU_ACCESS | - AMDGPU_GEM_CREATE_SHADOW); + flags |= AMDGPU_GEM_CREATE_NO_CPU_ACCESS; /* walk over the address space and allocate the page tables */ for (pt_idx = from; pt_idx <= to; ++pt_idx) { @@ -2344,6 +2345,35 @@ struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm, } /** + * amdgpu_vm_bo_trace_cs - trace all reserved mappings + * + * @vm: the requested vm + * @ticket: CS ticket + * + * Trace all mappings of BOs reserved during a command submission. + */ +void amdgpu_vm_bo_trace_cs(struct amdgpu_vm *vm, struct ww_acquire_ctx *ticket) +{ + struct amdgpu_bo_va_mapping *mapping; + + if (!trace_amdgpu_vm_bo_cs_enabled()) + return; + + for (mapping = amdgpu_vm_it_iter_first(&vm->va, 0, U64_MAX); mapping; + mapping = amdgpu_vm_it_iter_next(mapping, 0, U64_MAX)) { + if (mapping->bo_va && mapping->bo_va->base.bo) { + struct amdgpu_bo *bo; + + bo = mapping->bo_va->base.bo; + if (READ_ONCE(bo->tbo.resv->lock.ctx) != ticket) + continue; + } + + trace_amdgpu_vm_bo_cs(mapping); + } +} + +/** * amdgpu_vm_bo_rmv - remove a bo to a specific vm * * @adev: amdgpu_device pointer @@ -2587,7 +2617,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; if (vm->use_cpu_for_update) flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; - else + else if (vm_context != AMDGPU_VM_CONTEXT_COMPUTE) flags |= AMDGPU_GEM_CREATE_SHADOW; size = amdgpu_vm_bo_size(adev, adev->vm_manager.root_level); @@ -2662,8 +2692,7 @@ error_free_sched_entity: * - pasid (old PASID is released, because compute manages its own PASIDs) * * Reinitializes the page directory to reflect the changed ATS - * setting. May leave behind an unused shadow BO for the page - * directory when switching from SDMA updates to CPU updates. + * setting. * * Returns: * 0 for success, -errno for errors. @@ -2713,6 +2742,9 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm) vm->pasid = 0; } + /* Free the shadow bo for compute VM */ + amdgpu_bo_unref(&vm->root.base.bo->shadow); + error: amdgpu_bo_unreserve(vm->root.base.bo); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index d416f895233d..67a15d439ac0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -318,6 +318,7 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, uint64_t saddr, uint64_t size); struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm, uint64_t addr); +void amdgpu_vm_bo_trace_cs(struct amdgpu_vm *vm, struct ww_acquire_ctx *ticket); void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va); void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t vm_size, diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index 10920f0bd85f..36dc367c4b45 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -28,6 +28,7 @@ #include "cik.h" #include "gmc_v7_0.h" #include "amdgpu_ucode.h" +#include "amdgpu_amdkfd.h" #include "bif/bif_4_1_d.h" #include "bif/bif_4_1_sh_mask.h" @@ -1078,6 +1079,12 @@ static int gmc_v7_0_sw_init(void *handle) adev->vm_manager.vram_base_offset = 0; } + adev->gmc.vm_fault_info = kmalloc(sizeof(struct kfd_vm_fault_info), + GFP_KERNEL); + if (!adev->gmc.vm_fault_info) + return -ENOMEM; + atomic_set(&adev->gmc.vm_fault_info_updated, 0); + return 0; } @@ -1087,6 +1094,7 @@ static int gmc_v7_0_sw_fini(void *handle) amdgpu_gem_force_release(adev); amdgpu_vm_manager_fini(adev); + kfree(adev->gmc.vm_fault_info); gmc_v7_0_gart_fini(adev); amdgpu_bo_fini(adev); release_firmware(adev->gmc.fw); @@ -1276,7 +1284,7 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - u32 addr, status, mc_client; + u32 addr, status, mc_client, vmid; addr = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR); status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS); @@ -1301,6 +1309,29 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev, entry->pasid); } + vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, + VMID); + if (amdgpu_amdkfd_is_kfd_vmid(adev, vmid) + && !atomic_read(&adev->gmc.vm_fault_info_updated)) { + struct kfd_vm_fault_info *info = adev->gmc.vm_fault_info; + u32 protections = REG_GET_FIELD(status, + VM_CONTEXT1_PROTECTION_FAULT_STATUS, + PROTECTIONS); + + info->vmid = vmid; + info->mc_id = REG_GET_FIELD(status, + VM_CONTEXT1_PROTECTION_FAULT_STATUS, + MEMORY_CLIENT_ID); + info->status = status; + info->page_addr = addr; + info->prot_valid = protections & 0x7 ? true : false; + info->prot_read = protections & 0x8 ? true : false; + info->prot_write = protections & 0x10 ? true : false; + info->prot_exec = protections & 0x20 ? true : false; + mb(); + atomic_set(&adev->gmc.vm_fault_info_updated, 1); + } + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index 75f3ffb2891e..70fc97b59b4f 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -26,6 +26,7 @@ #include "amdgpu.h" #include "gmc_v8_0.h" #include "amdgpu_ucode.h" +#include "amdgpu_amdkfd.h" #include "gmc/gmc_8_1_d.h" #include "gmc/gmc_8_1_sh_mask.h" @@ -1182,6 +1183,12 @@ static int gmc_v8_0_sw_init(void *handle) adev->vm_manager.vram_base_offset = 0; } + adev->gmc.vm_fault_info = kmalloc(sizeof(struct kfd_vm_fault_info), + GFP_KERNEL); + if (!adev->gmc.vm_fault_info) + return -ENOMEM; + atomic_set(&adev->gmc.vm_fault_info_updated, 0); + return 0; } @@ -1191,6 +1198,7 @@ static int gmc_v8_0_sw_fini(void *handle) amdgpu_gem_force_release(adev); amdgpu_vm_manager_fini(adev); + kfree(adev->gmc.vm_fault_info); gmc_v8_0_gart_fini(adev); amdgpu_bo_fini(adev); release_firmware(adev->gmc.fw); @@ -1426,7 +1434,7 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - u32 addr, status, mc_client; + u32 addr, status, mc_client, vmid; if (amdgpu_sriov_vf(adev)) { dev_err(adev->dev, "GPU fault detected: %d 0x%08x\n", @@ -1463,6 +1471,29 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev, entry->pasid); } + vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, + VMID); + if (amdgpu_amdkfd_is_kfd_vmid(adev, vmid) + && !atomic_read(&adev->gmc.vm_fault_info_updated)) { + struct kfd_vm_fault_info *info = adev->gmc.vm_fault_info; + u32 protections = REG_GET_FIELD(status, + VM_CONTEXT1_PROTECTION_FAULT_STATUS, + PROTECTIONS); + + info->vmid = vmid; + info->mc_id = REG_GET_FIELD(status, + VM_CONTEXT1_PROTECTION_FAULT_STATUS, + MEMORY_CLIENT_ID); + info->status = status; + info->page_addr = addr; + info->prot_valid = protections & 0x7 ? true : false; + info->prot_read = protections & 0x8 ? true : false; + info->prot_write = protections & 0x10 ? true : false; + info->prot_exec = protections & 0x20 ? true : false; + mb(); + atomic_set(&adev->gmc.vm_fault_info_updated, 1); + } + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index db5f3d78ab12..5fab3560a71d 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -41,6 +41,12 @@ #include "mmhub/mmhub_1_0_sh_mask.h" #include "ivsrcid/uvd/irqsrcs_uvd_7_0.h" +#define mmUVD_PG0_CC_UVD_HARVESTING 0x00c7 +#define mmUVD_PG0_CC_UVD_HARVESTING_BASE_IDX 1 +//UVD_PG0_CC_UVD_HARVESTING +#define UVD_PG0_CC_UVD_HARVESTING__UVD_DISABLE__SHIFT 0x1 +#define UVD_PG0_CC_UVD_HARVESTING__UVD_DISABLE_MASK 0x00000002L + #define UVD7_MAX_HW_INSTANCES_VEGA20 2 static void uvd_v7_0_set_ring_funcs(struct amdgpu_device *adev); @@ -370,10 +376,25 @@ error: static int uvd_v7_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->asic_type == CHIP_VEGA20) + + if (adev->asic_type == CHIP_VEGA20) { + u32 harvest; + int i; + adev->uvd.num_uvd_inst = UVD7_MAX_HW_INSTANCES_VEGA20; - else + for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + harvest = RREG32_SOC15(UVD, i, mmUVD_PG0_CC_UVD_HARVESTING); + if (harvest & UVD_PG0_CC_UVD_HARVESTING__UVD_DISABLE_MASK) { + adev->uvd.harvest_config |= 1 << i; + } + } + if (adev->uvd.harvest_config == (AMDGPU_UVD_HARVEST_UVD0 | + AMDGPU_UVD_HARVEST_UVD1)) + /* both instances are harvested, disable the block */ + return -ENOENT; + } else { adev->uvd.num_uvd_inst = 1; + } if (amdgpu_sriov_vf(adev)) adev->uvd.num_enc_rings = 1; @@ -393,6 +414,8 @@ static int uvd_v7_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (j = 0; j < adev->uvd.num_uvd_inst; j++) { + if (adev->uvd.harvest_config & (1 << j)) + continue; /* UVD TRAP */ r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_uvds[j], UVD_7_0__SRCID__UVD_SYSTEM_MESSAGE_INTERRUPT, &adev->uvd.inst[j].irq); if (r) @@ -425,6 +448,8 @@ static int uvd_v7_0_sw_init(void *handle) return r; for (j = 0; j < adev->uvd.num_uvd_inst; j++) { + if (adev->uvd.harvest_config & (1 << j)) + continue; if (!amdgpu_sriov_vf(adev)) { ring = &adev->uvd.inst[j].ring; sprintf(ring->name, "uvd<%d>", j); @@ -472,6 +497,8 @@ static int uvd_v7_0_sw_fini(void *handle) return r; for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { + if (adev->uvd.harvest_config & (1 << j)) + continue; for (i = 0; i < adev->uvd.num_enc_rings; ++i) amdgpu_ring_fini(&adev->uvd.inst[j].ring_enc[i]); } @@ -500,6 +527,8 @@ static int uvd_v7_0_hw_init(void *handle) goto done; for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { + if (adev->uvd.harvest_config & (1 << j)) + continue; ring = &adev->uvd.inst[j].ring; if (!amdgpu_sriov_vf(adev)) { @@ -579,8 +608,11 @@ static int uvd_v7_0_hw_fini(void *handle) DRM_DEBUG("For SRIOV client, shouldn't do anything.\n"); } - for (i = 0; i < adev->uvd.num_uvd_inst; ++i) + for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; adev->uvd.inst[i].ring.ready = false; + } return 0; } @@ -623,6 +655,8 @@ static void uvd_v7_0_mc_resume(struct amdgpu_device *adev) int i; for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, lower_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr)); @@ -695,6 +729,8 @@ static int uvd_v7_0_mmsch_start(struct amdgpu_device *adev, WREG32_SOC15(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP, 0); for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; WDOORBELL32(adev->uvd.inst[i].ring_enc[0].doorbell_index, 0); adev->wb.wb[adev->uvd.inst[i].ring_enc[0].wptr_offs] = 0; adev->uvd.inst[i].ring_enc[0].wptr = 0; @@ -751,6 +787,8 @@ static int uvd_v7_0_sriov_start(struct amdgpu_device *adev) init_table += header->uvd_table_offset; for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; ring = &adev->uvd.inst[i].ring; ring->wptr = 0; size = AMDGPU_GPU_PAGE_ALIGN(adev->uvd.fw->size + 4); @@ -890,6 +928,8 @@ static int uvd_v7_0_start(struct amdgpu_device *adev) int i, j, k, r; for (k = 0; k < adev->uvd.num_uvd_inst; ++k) { + if (adev->uvd.harvest_config & (1 << k)) + continue; /* disable DPG */ WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); @@ -902,6 +942,8 @@ static int uvd_v7_0_start(struct amdgpu_device *adev) uvd_v7_0_mc_resume(adev); for (k = 0; k < adev->uvd.num_uvd_inst; ++k) { + if (adev->uvd.harvest_config & (1 << k)) + continue; ring = &adev->uvd.inst[k].ring; /* disable clock gating */ WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_CGC_CTRL), 0, @@ -1069,6 +1111,8 @@ static void uvd_v7_0_stop(struct amdgpu_device *adev) uint8_t i = 0; for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; /* force RBC into idle state */ WREG32_SOC15(UVD, i, mmUVD_RBC_RB_CNTL, 0x11010101); @@ -1206,6 +1250,34 @@ static int uvd_v7_0_ring_test_ring(struct amdgpu_ring *ring) } /** + * uvd_v7_0_ring_patch_cs_in_place - Patch the IB for command submission. + * + * @p: the CS parser with the IBs + * @ib_idx: which IB to patch + * + */ +static int uvd_v7_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, + uint32_t ib_idx) +{ + struct amdgpu_ib *ib = &p->job->ibs[ib_idx]; + unsigned i; + + /* No patching necessary for the first instance */ + if (!p->ring->me) + return 0; + + for (i = 0; i < ib->length_dw; i += 2) { + uint32_t reg = amdgpu_get_ib_value(p, ib_idx, i); + + reg -= p->adev->reg_offset[UVD_HWIP][0][1]; + reg += p->adev->reg_offset[UVD_HWIP][1][1]; + + amdgpu_set_ib_value(p, ib_idx, i, reg); + } + return 0; +} + +/** * uvd_v7_0_ring_emit_ib - execute indirect buffer * * @ring: amdgpu_ring pointer @@ -1697,6 +1769,7 @@ static const struct amdgpu_ring_funcs uvd_v7_0_ring_vm_funcs = { .get_rptr = uvd_v7_0_ring_get_rptr, .get_wptr = uvd_v7_0_ring_get_wptr, .set_wptr = uvd_v7_0_ring_set_wptr, + .patch_cs_in_place = uvd_v7_0_ring_patch_cs_in_place, .emit_frame_size = 6 + /* hdp invalidate */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + @@ -1756,6 +1829,8 @@ static void uvd_v7_0_set_ring_funcs(struct amdgpu_device *adev) int i; for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (adev->uvd.harvest_config & (1 << i)) + continue; adev->uvd.inst[i].ring.funcs = &uvd_v7_0_ring_vm_funcs; adev->uvd.inst[i].ring.me = i; DRM_INFO("UVD(%d) is enabled in VM mode\n", i); @@ -1767,6 +1842,8 @@ static void uvd_v7_0_set_enc_ring_funcs(struct amdgpu_device *adev) int i, j; for (j = 0; j < adev->uvd.num_uvd_inst; j++) { + if (adev->uvd.harvest_config & (1 << j)) + continue; for (i = 0; i < adev->uvd.num_enc_rings; ++i) { adev->uvd.inst[j].ring_enc[i].funcs = &uvd_v7_0_enc_ring_vm_funcs; adev->uvd.inst[j].ring_enc[i].me = j; @@ -1786,6 +1863,8 @@ static void uvd_v7_0_set_irq_funcs(struct amdgpu_device *adev) int i; for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (adev->uvd.harvest_config & (1 << i)) + continue; adev->uvd.inst[i].irq.num_types = adev->uvd.num_enc_rings + 1; adev->uvd.inst[i].irq.funcs = &uvd_v7_0_irq_funcs; } |
