diff options
-rw-r--r-- | drivers/accel/ivpu/ivpu_drv.c | 5 | ||||
-rw-r--r-- | drivers/accel/ivpu/ivpu_fw.c | 1 | ||||
-rw-r--r-- | drivers/accel/ivpu/ivpu_hw_37xx.c | 2 | ||||
-rw-r--r-- | drivers/accel/ivpu/ivpu_hw_40xx.c | 7 | ||||
-rw-r--r-- | drivers/accel/ivpu/ivpu_job.c | 4 | ||||
-rw-r--r-- | drivers/accel/ivpu/ivpu_mmu.c | 36 | ||||
-rw-r--r-- | drivers/gpu/drm/nouveau/include/nvkm/subdev/gsp.h | 2 | ||||
-rw-r--r-- | drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c | 77 | ||||
-rw-r--r-- | drivers/gpu/drm/scheduler/sched_main.c | 15 | ||||
-rw-r--r-- | drivers/gpu/drm/xe/xe_display.c | 6 | ||||
-rw-r--r-- | drivers/gpu/drm/xe/xe_exec_queue.c | 8 | ||||
-rw-r--r-- | drivers/gpu/drm/xe/xe_gt.c | 5 | ||||
-rw-r--r-- | drivers/gpu/drm/xe/xe_gt_pagefault.c | 2 | ||||
-rw-r--r-- | drivers/gpu/drm/xe/xe_migrate.c | 28 | ||||
-rw-r--r-- | drivers/gpu/drm/xe/xe_sched_job.c | 1 | ||||
-rw-r--r-- | drivers/gpu/drm/xe/xe_sync.c | 2 | ||||
-rw-r--r-- | drivers/gpu/drm/xe/xe_vm.c | 62 | ||||
-rw-r--r-- | drivers/gpu/drm/xe/xe_vm_types.h | 8 | ||||
-rw-r--r-- | include/uapi/drm/ivpu_accel.h | 1 |
19 files changed, 143 insertions, 129 deletions
diff --git a/drivers/accel/ivpu/ivpu_drv.c b/drivers/accel/ivpu/ivpu_drv.c index 9418c73ee8ef..4b0640226986 100644 --- a/drivers/accel/ivpu/ivpu_drv.c +++ b/drivers/accel/ivpu/ivpu_drv.c @@ -480,9 +480,8 @@ static int ivpu_pci_init(struct ivpu_device *vdev) /* Clear any pending errors */ pcie_capability_clear_word(pdev, PCI_EXP_DEVSTA, 0x3f); - /* VPU 37XX does not require 10m D3hot delay */ - if (ivpu_hw_gen(vdev) == IVPU_HW_37XX) - pdev->d3hot_delay = 0; + /* NPU does not require 10m D3hot delay */ + pdev->d3hot_delay = 0; ret = pcim_enable_device(pdev); if (ret) { diff --git a/drivers/accel/ivpu/ivpu_fw.c b/drivers/accel/ivpu/ivpu_fw.c index 6576232f3e67..5fa8bd4603d5 100644 --- a/drivers/accel/ivpu/ivpu_fw.c +++ b/drivers/accel/ivpu/ivpu_fw.c @@ -222,7 +222,6 @@ ivpu_fw_init_wa(struct ivpu_device *vdev) const struct vpu_firmware_header *fw_hdr = (const void *)vdev->fw->file->data; if (IVPU_FW_CHECK_API_VER_LT(vdev, fw_hdr, BOOT, 3, 17) || - (ivpu_hw_gen(vdev) > IVPU_HW_37XX) || (ivpu_test_mode & IVPU_TEST_MODE_D0I3_MSG_DISABLE)) vdev->wa.disable_d0i3_msg = true; diff --git a/drivers/accel/ivpu/ivpu_hw_37xx.c b/drivers/accel/ivpu/ivpu_hw_37xx.c index f15a93d83057..77accd029c4a 100644 --- a/drivers/accel/ivpu/ivpu_hw_37xx.c +++ b/drivers/accel/ivpu/ivpu_hw_37xx.c @@ -525,7 +525,7 @@ static void ivpu_boot_no_snoop_enable(struct ivpu_device *vdev) u32 val = REGV_RD32(VPU_37XX_HOST_IF_TCU_PTW_OVERRIDES); val = REG_SET_FLD(VPU_37XX_HOST_IF_TCU_PTW_OVERRIDES, NOSNOOP_OVERRIDE_EN, val); - val = REG_SET_FLD(VPU_37XX_HOST_IF_TCU_PTW_OVERRIDES, AW_NOSNOOP_OVERRIDE, val); + val = REG_CLR_FLD(VPU_37XX_HOST_IF_TCU_PTW_OVERRIDES, AW_NOSNOOP_OVERRIDE, val); val = REG_SET_FLD(VPU_37XX_HOST_IF_TCU_PTW_OVERRIDES, AR_NOSNOOP_OVERRIDE, val); REGV_WR32(VPU_37XX_HOST_IF_TCU_PTW_OVERRIDES, val); diff --git a/drivers/accel/ivpu/ivpu_hw_40xx.c b/drivers/accel/ivpu/ivpu_hw_40xx.c index 704288084f37..1c995307c113 100644 --- a/drivers/accel/ivpu/ivpu_hw_40xx.c +++ b/drivers/accel/ivpu/ivpu_hw_40xx.c @@ -530,7 +530,7 @@ static void ivpu_boot_no_snoop_enable(struct ivpu_device *vdev) u32 val = REGV_RD32(VPU_40XX_HOST_IF_TCU_PTW_OVERRIDES); val = REG_SET_FLD(VPU_40XX_HOST_IF_TCU_PTW_OVERRIDES, SNOOP_OVERRIDE_EN, val); - val = REG_CLR_FLD(VPU_40XX_HOST_IF_TCU_PTW_OVERRIDES, AW_SNOOP_OVERRIDE, val); + val = REG_SET_FLD(VPU_40XX_HOST_IF_TCU_PTW_OVERRIDES, AW_SNOOP_OVERRIDE, val); val = REG_CLR_FLD(VPU_40XX_HOST_IF_TCU_PTW_OVERRIDES, AR_SNOOP_OVERRIDE, val); REGV_WR32(VPU_40XX_HOST_IF_TCU_PTW_OVERRIDES, val); @@ -704,7 +704,6 @@ static int ivpu_hw_40xx_info_init(struct ivpu_device *vdev) { struct ivpu_hw_info *hw = vdev->hw; u32 tile_disable; - u32 tile_enable; u32 fuse; fuse = REGB_RD32(VPU_40XX_BUTTRESS_TILE_FUSE); @@ -725,10 +724,6 @@ static int ivpu_hw_40xx_info_init(struct ivpu_device *vdev) else ivpu_dbg(vdev, MISC, "Fuse: All %d tiles enabled\n", TILE_MAX_NUM); - tile_enable = (~tile_disable) & TILE_MAX_MASK; - - hw->sku = REG_SET_FLD_NUM(SKU, HW_ID, LNL_HW_ID, hw->sku); - hw->sku = REG_SET_FLD_NUM(SKU, TILE, tile_enable, hw->sku); hw->tile_fuse = tile_disable; hw->pll.profiling_freq = PLL_PROFILING_FREQ_DEFAULT; diff --git a/drivers/accel/ivpu/ivpu_job.c b/drivers/accel/ivpu/ivpu_job.c index 0440bee3ecaf..e70cfb859339 100644 --- a/drivers/accel/ivpu/ivpu_job.c +++ b/drivers/accel/ivpu/ivpu_job.c @@ -294,7 +294,7 @@ static int ivpu_job_signal_and_destroy(struct ivpu_device *vdev, u32 job_id, u32 return -ENOENT; if (job->file_priv->has_mmu_faults) - job_status = VPU_JSM_STATUS_ABORTED; + job_status = DRM_IVPU_JOB_STATUS_ABORTED; job->bos[CMD_BUF_IDX]->job_status = job_status; dma_fence_signal(job->done_fence); @@ -315,7 +315,7 @@ void ivpu_jobs_abort_all(struct ivpu_device *vdev) unsigned long id; xa_for_each(&vdev->submitted_jobs_xa, id, job) - ivpu_job_signal_and_destroy(vdev, id, VPU_JSM_STATUS_ABORTED); + ivpu_job_signal_and_destroy(vdev, id, DRM_IVPU_JOB_STATUS_ABORTED); } static int ivpu_job_submit(struct ivpu_job *job) diff --git a/drivers/accel/ivpu/ivpu_mmu.c b/drivers/accel/ivpu/ivpu_mmu.c index 9a3122ffce03..91bd640655ab 100644 --- a/drivers/accel/ivpu/ivpu_mmu.c +++ b/drivers/accel/ivpu/ivpu_mmu.c @@ -72,10 +72,10 @@ #define IVPU_MMU_Q_COUNT_LOG2 4 /* 16 entries */ #define IVPU_MMU_Q_COUNT ((u32)1 << IVPU_MMU_Q_COUNT_LOG2) -#define IVPU_MMU_Q_WRAP_BIT (IVPU_MMU_Q_COUNT << 1) -#define IVPU_MMU_Q_WRAP_MASK (IVPU_MMU_Q_WRAP_BIT - 1) -#define IVPU_MMU_Q_IDX_MASK (IVPU_MMU_Q_COUNT - 1) +#define IVPU_MMU_Q_WRAP_MASK GENMASK(IVPU_MMU_Q_COUNT_LOG2, 0) +#define IVPU_MMU_Q_IDX_MASK (IVPU_MMU_Q_COUNT - 1) #define IVPU_MMU_Q_IDX(val) ((val) & IVPU_MMU_Q_IDX_MASK) +#define IVPU_MMU_Q_WRP(val) ((val) & IVPU_MMU_Q_COUNT) #define IVPU_MMU_CMDQ_CMD_SIZE 16 #define IVPU_MMU_CMDQ_SIZE (IVPU_MMU_Q_COUNT * IVPU_MMU_CMDQ_CMD_SIZE) @@ -475,20 +475,32 @@ static int ivpu_mmu_cmdq_wait_for_cons(struct ivpu_device *vdev) return 0; } +static bool ivpu_mmu_queue_is_full(struct ivpu_mmu_queue *q) +{ + return ((IVPU_MMU_Q_IDX(q->prod) == IVPU_MMU_Q_IDX(q->cons)) && + (IVPU_MMU_Q_WRP(q->prod) != IVPU_MMU_Q_WRP(q->cons))); +} + +static bool ivpu_mmu_queue_is_empty(struct ivpu_mmu_queue *q) +{ + return ((IVPU_MMU_Q_IDX(q->prod) == IVPU_MMU_Q_IDX(q->cons)) && + (IVPU_MMU_Q_WRP(q->prod) == IVPU_MMU_Q_WRP(q->cons))); +} + static int ivpu_mmu_cmdq_cmd_write(struct ivpu_device *vdev, const char *name, u64 data0, u64 data1) { - struct ivpu_mmu_queue *q = &vdev->mmu->cmdq; - u64 *queue_buffer = q->base; - int idx = IVPU_MMU_Q_IDX(q->prod) * (IVPU_MMU_CMDQ_CMD_SIZE / sizeof(*queue_buffer)); + struct ivpu_mmu_queue *cmdq = &vdev->mmu->cmdq; + u64 *queue_buffer = cmdq->base; + int idx = IVPU_MMU_Q_IDX(cmdq->prod) * (IVPU_MMU_CMDQ_CMD_SIZE / sizeof(*queue_buffer)); - if (!CIRC_SPACE(IVPU_MMU_Q_IDX(q->prod), IVPU_MMU_Q_IDX(q->cons), IVPU_MMU_Q_COUNT)) { + if (ivpu_mmu_queue_is_full(cmdq)) { ivpu_err(vdev, "Failed to write MMU CMD %s\n", name); return -EBUSY; } queue_buffer[idx] = data0; queue_buffer[idx + 1] = data1; - q->prod = (q->prod + 1) & IVPU_MMU_Q_WRAP_MASK; + cmdq->prod = (cmdq->prod + 1) & IVPU_MMU_Q_WRAP_MASK; ivpu_dbg(vdev, MMU, "CMD write: %s data: 0x%llx 0x%llx\n", name, data0, data1); @@ -560,7 +572,6 @@ static int ivpu_mmu_reset(struct ivpu_device *vdev) mmu->cmdq.cons = 0; memset(mmu->evtq.base, 0, IVPU_MMU_EVTQ_SIZE); - clflush_cache_range(mmu->evtq.base, IVPU_MMU_EVTQ_SIZE); mmu->evtq.prod = 0; mmu->evtq.cons = 0; @@ -874,14 +885,10 @@ static u32 *ivpu_mmu_get_event(struct ivpu_device *vdev) u32 *evt = evtq->base + (idx * IVPU_MMU_EVTQ_CMD_SIZE); evtq->prod = REGV_RD32(IVPU_MMU_REG_EVTQ_PROD_SEC); - if (!CIRC_CNT(IVPU_MMU_Q_IDX(evtq->prod), IVPU_MMU_Q_IDX(evtq->cons), IVPU_MMU_Q_COUNT)) + if (ivpu_mmu_queue_is_empty(evtq)) return NULL; - clflush_cache_range(evt, IVPU_MMU_EVTQ_CMD_SIZE); - evtq->cons = (evtq->cons + 1) & IVPU_MMU_Q_WRAP_MASK; - REGV_WR32(IVPU_MMU_REG_EVTQ_CONS_SEC, evtq->cons); - return evt; } @@ -902,6 +909,7 @@ void ivpu_mmu_irq_evtq_handler(struct ivpu_device *vdev) } ivpu_mmu_user_context_mark_invalid(vdev, ssid); + REGV_WR32(IVPU_MMU_REG_EVTQ_CONS_SEC, vdev->mmu->evtq.cons); } } diff --git a/drivers/gpu/drm/nouveau/include/nvkm/subdev/gsp.h b/drivers/gpu/drm/nouveau/include/nvkm/subdev/gsp.h index d1437c08645f..6f5d376d8fcc 100644 --- a/drivers/gpu/drm/nouveau/include/nvkm/subdev/gsp.h +++ b/drivers/gpu/drm/nouveau/include/nvkm/subdev/gsp.h @@ -9,7 +9,7 @@ #define GSP_PAGE_SIZE BIT(GSP_PAGE_SHIFT) struct nvkm_gsp_mem { - u32 size; + size_t size; void *data; dma_addr_t addr; }; diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c index 5e1fa176aac4..a41735ab6068 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c @@ -997,6 +997,32 @@ r535_gsp_rpc_get_gsp_static_info(struct nvkm_gsp *gsp) return 0; } +static void +nvkm_gsp_mem_dtor(struct nvkm_gsp *gsp, struct nvkm_gsp_mem *mem) +{ + if (mem->data) { + /* + * Poison the buffer to catch any unexpected access from + * GSP-RM if the buffer was prematurely freed. + */ + memset(mem->data, 0xFF, mem->size); + + dma_free_coherent(gsp->subdev.device->dev, mem->size, mem->data, mem->addr); + memset(mem, 0, sizeof(*mem)); + } +} + +static int +nvkm_gsp_mem_ctor(struct nvkm_gsp *gsp, size_t size, struct nvkm_gsp_mem *mem) +{ + mem->size = size; + mem->data = dma_alloc_coherent(gsp->subdev.device->dev, size, &mem->addr, GFP_KERNEL); + if (WARN_ON(!mem->data)) + return -ENOMEM; + + return 0; +} + static int r535_gsp_postinit(struct nvkm_gsp *gsp) { @@ -1024,6 +1050,13 @@ r535_gsp_postinit(struct nvkm_gsp *gsp) nvkm_inth_allow(&gsp->subdev.inth); nvkm_wr32(device, 0x110004, 0x00000040); + + /* Release the DMA buffers that were needed only for boot and init */ + nvkm_gsp_mem_dtor(gsp, &gsp->boot.fw); + nvkm_gsp_mem_dtor(gsp, &gsp->libos); + nvkm_gsp_mem_dtor(gsp, &gsp->rmargs); + nvkm_gsp_mem_dtor(gsp, &gsp->wpr_meta); + return ret; } @@ -1532,27 +1565,6 @@ r535_gsp_msg_run_cpu_sequencer(void *priv, u32 fn, void *repv, u32 repc) return 0; } -static void -nvkm_gsp_mem_dtor(struct nvkm_gsp *gsp, struct nvkm_gsp_mem *mem) -{ - if (mem->data) { - dma_free_coherent(gsp->subdev.device->dev, mem->size, mem->data, mem->addr); - mem->data = NULL; - } -} - -static int -nvkm_gsp_mem_ctor(struct nvkm_gsp *gsp, u32 size, struct nvkm_gsp_mem *mem) -{ - mem->size = size; - mem->data = dma_alloc_coherent(gsp->subdev.device->dev, size, &mem->addr, GFP_KERNEL); - if (WARN_ON(!mem->data)) - return -ENOMEM; - - return 0; -} - - static int r535_gsp_booter_unload(struct nvkm_gsp *gsp, u32 mbox0, u32 mbox1) { @@ -1938,20 +1950,20 @@ nvkm_gsp_radix3_dtor(struct nvkm_gsp *gsp, struct nvkm_gsp_radix3 *rx3) * See kgspCreateRadix3_IMPL */ static int -nvkm_gsp_radix3_sg(struct nvkm_device *device, struct sg_table *sgt, u64 size, +nvkm_gsp_radix3_sg(struct nvkm_gsp *gsp, struct sg_table *sgt, u64 size, struct nvkm_gsp_radix3 *rx3) { u64 addr; for (int i = ARRAY_SIZE(rx3->mem) - 1; i >= 0; i--) { u64 *ptes; - int idx; + size_t bufsize; + int ret, idx; - rx3->mem[i].size = ALIGN((size / GSP_PAGE_SIZE) * sizeof(u64), GSP_PAGE_SIZE); - rx3->mem[i].data = dma_alloc_coherent(device->dev, rx3->mem[i].size, - &rx3->mem[i].addr, GFP_KERNEL); - if (WARN_ON(!rx3->mem[i].data)) - return -ENOMEM; + bufsize = ALIGN((size / GSP_PAGE_SIZE) * sizeof(u64), GSP_PAGE_SIZE); + ret = nvkm_gsp_mem_ctor(gsp, bufsize, &rx3->mem[i]); + if (ret) + return ret; ptes = rx3->mem[i].data; if (i == 2) { @@ -1991,7 +2003,7 @@ r535_gsp_fini(struct nvkm_gsp *gsp, bool suspend) if (ret) return ret; - ret = nvkm_gsp_radix3_sg(gsp->subdev.device, &gsp->sr.sgt, len, &gsp->sr.radix3); + ret = nvkm_gsp_radix3_sg(gsp, &gsp->sr.sgt, len, &gsp->sr.radix3); if (ret) return ret; @@ -2150,6 +2162,11 @@ r535_gsp_dtor(struct nvkm_gsp *gsp) mutex_destroy(&gsp->cmdq.mutex); r535_gsp_dtor_fws(gsp); + + nvkm_gsp_mem_dtor(gsp, &gsp->shm.mem); + nvkm_gsp_mem_dtor(gsp, &gsp->loginit); + nvkm_gsp_mem_dtor(gsp, &gsp->logintr); + nvkm_gsp_mem_dtor(gsp, &gsp->logrm); } int @@ -2194,7 +2211,7 @@ r535_gsp_oneinit(struct nvkm_gsp *gsp) memcpy(gsp->sig.data, data, size); /* Build radix3 page table for ELF image. */ - ret = nvkm_gsp_radix3_sg(device, &gsp->fw.mem.sgt, gsp->fw.len, &gsp->radix3); + ret = nvkm_gsp_radix3_sg(gsp, &gsp->fw.mem.sgt, gsp->fw.len, &gsp->radix3); if (ret) return ret; diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c index 85f082396d42..d442b893275b 100644 --- a/drivers/gpu/drm/scheduler/sched_main.c +++ b/drivers/gpu/drm/scheduler/sched_main.c @@ -1178,21 +1178,24 @@ static void drm_sched_run_job_work(struct work_struct *w) struct drm_sched_entity *entity; struct dma_fence *fence; struct drm_sched_fence *s_fence; - struct drm_sched_job *sched_job = NULL; + struct drm_sched_job *sched_job; int r; if (READ_ONCE(sched->pause_submit)) return; /* Find entity with a ready job */ - while (!sched_job && (entity = drm_sched_select_entity(sched))) { - sched_job = drm_sched_entity_pop_job(entity); - if (!sched_job) - complete_all(&entity->entity_idle); - } + entity = drm_sched_select_entity(sched); if (!entity) return; /* No more work */ + sched_job = drm_sched_entity_pop_job(entity); + if (!sched_job) { + complete_all(&entity->entity_idle); + drm_sched_run_job_queue(sched); + return; + } + s_fence = sched_job->s_fence; atomic_add(sched_job->credits, &sched->credit_count); diff --git a/drivers/gpu/drm/xe/xe_display.c b/drivers/gpu/drm/xe/xe_display.c index 74391d9b11ae..e4db069f0db3 100644 --- a/drivers/gpu/drm/xe/xe_display.c +++ b/drivers/gpu/drm/xe/xe_display.c @@ -134,8 +134,6 @@ static void xe_display_fini_nommio(struct drm_device *dev, void *dummy) int xe_display_init_nommio(struct xe_device *xe) { - int err; - if (!xe->info.enable_display) return 0; @@ -145,10 +143,6 @@ int xe_display_init_nommio(struct xe_device *xe) /* This must be called before any calls to HAS_PCH_* */ intel_detect_pch(xe); - err = intel_power_domains_init(xe); - if (err) - return err; - return drmm_add_action_or_reset(&xe->drm, xe_display_fini_nommio, xe); } diff --git a/drivers/gpu/drm/xe/xe_exec_queue.c b/drivers/gpu/drm/xe/xe_exec_queue.c index bcfc4127c7c5..254b1d3af4cb 100644 --- a/drivers/gpu/drm/xe/xe_exec_queue.c +++ b/drivers/gpu/drm/xe/xe_exec_queue.c @@ -926,20 +926,24 @@ void xe_exec_queue_last_fence_put_unlocked(struct xe_exec_queue *q) * @q: The exec queue * @vm: The VM the engine does a bind or exec for * - * Get last fence, does not take a ref + * Get last fence, takes a ref * * Returns: last fence if not signaled, dma fence stub if signaled */ struct dma_fence *xe_exec_queue_last_fence_get(struct xe_exec_queue *q, struct xe_vm *vm) { + struct dma_fence *fence; + xe_exec_queue_last_fence_lockdep_assert(q, vm); if (q->last_fence && test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &q->last_fence->flags)) xe_exec_queue_last_fence_put(q, vm); - return q->last_fence ? q->last_fence : dma_fence_get_stub(); + fence = q->last_fence ? q->last_fence : dma_fence_get_stub(); + dma_fence_get(fence); + return fence; } /** diff --git a/drivers/gpu/drm/xe/xe_gt.c b/drivers/gpu/drm/xe/xe_gt.c index 3af2adec1295..35474ddbaf97 100644 --- a/drivers/gpu/drm/xe/xe_gt.c +++ b/drivers/gpu/drm/xe/xe_gt.c @@ -437,7 +437,10 @@ static int all_fw_domain_init(struct xe_gt *gt) * USM has its only SA pool to non-block behind user operations */ if (gt_to_xe(gt)->info.has_usm) { - gt->usm.bb_pool = xe_sa_bo_manager_init(gt_to_tile(gt), SZ_1M, 16); + struct xe_device *xe = gt_to_xe(gt); + + gt->usm.bb_pool = xe_sa_bo_manager_init(gt_to_tile(gt), + IS_DGFX(xe) ? SZ_1M : SZ_512K, 16); if (IS_ERR(gt->usm.bb_pool)) { err = PTR_ERR(gt->usm.bb_pool); goto err_force_wake; diff --git a/drivers/gpu/drm/xe/xe_gt_pagefault.c b/drivers/gpu/drm/xe/xe_gt_pagefault.c index 9c2fe1697d6e..73f08f1924df 100644 --- a/drivers/gpu/drm/xe/xe_gt_pagefault.c +++ b/drivers/gpu/drm/xe/xe_gt_pagefault.c @@ -335,7 +335,7 @@ int xe_guc_pagefault_handler(struct xe_guc *guc, u32 *msg, u32 len) return -EPROTO; asid = FIELD_GET(PFD_ASID, msg[1]); - pf_queue = >->usm.pf_queue[asid % NUM_PF_QUEUE]; + pf_queue = gt->usm.pf_queue + (asid % NUM_PF_QUEUE); spin_lock_irqsave(&pf_queue->lock, flags); full = pf_queue_full(pf_queue); diff --git a/drivers/gpu/drm/xe/xe_migrate.c b/drivers/gpu/drm/xe/xe_migrate.c index 5c6c54624252..70480c305602 100644 --- a/drivers/gpu/drm/xe/xe_migrate.c +++ b/drivers/gpu/drm/xe/xe_migrate.c @@ -170,11 +170,6 @@ static int xe_migrate_prepare_vm(struct xe_tile *tile, struct xe_migrate *m, if (!IS_DGFX(xe)) { /* Write out batch too */ m->batch_base_ofs = NUM_PT_SLOTS * XE_PAGE_SIZE; - if (xe->info.has_usm) { - batch = tile->primary_gt->usm.bb_pool->bo; - m->usm_batch_base_ofs = m->batch_base_ofs; - } - for (i = 0; i < batch->size; i += vm->flags & XE_VM_FLAG_64K ? XE_64K_PAGE_SIZE : XE_PAGE_SIZE) { @@ -185,6 +180,24 @@ static int xe_migrate_prepare_vm(struct xe_tile *tile, struct xe_migrate *m, entry); level++; } + if (xe->info.has_usm) { + xe_tile_assert(tile, batch->size == SZ_1M); + + batch = tile->primary_gt->usm.bb_pool->bo; + m->usm_batch_base_ofs = m->batch_base_ofs + SZ_1M; + xe_tile_assert(tile, batch->size == SZ_512K); + + for (i = 0; i < batch->size; + i += vm->flags & XE_VM_FLAG_64K ? XE_64K_PAGE_SIZE : + XE_PAGE_SIZE) { + entry = vm->pt_ops->pte_encode_bo(batch, i, + pat_index, 0); + + xe_map_wr(xe, &bo->vmap, map_ofs + level * 8, u64, + entry); + level++; + } + } } else { u64 batch_addr = xe_bo_addr(batch, 0, XE_PAGE_SIZE); @@ -1204,8 +1217,11 @@ static bool no_in_syncs(struct xe_vm *vm, struct xe_exec_queue *q, } if (q) { fence = xe_exec_queue_last_fence_get(q, vm); - if (!test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) + if (!test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) { + dma_fence_put(fence); return false; + } + dma_fence_put(fence); } return true; diff --git a/drivers/gpu/drm/xe/xe_sched_job.c b/drivers/gpu/drm/xe/xe_sched_job.c index 01106a1156ad..4e2ccad0e52f 100644 --- a/drivers/gpu/drm/xe/xe_sched_job.c +++ b/drivers/gpu/drm/xe/xe_sched_job.c @@ -274,7 +274,6 @@ int xe_sched_job_last_fence_add_dep(struct xe_sched_job *job, struct xe_vm *vm) struct dma_fence *fence; fence = xe_exec_queue_last_fence_get(job->q, vm); - dma_fence_get(fence); return drm_sched_job_add_dependency(&job->drm, fence); } diff --git a/drivers/gpu/drm/xe/xe_sync.c b/drivers/gpu/drm/xe/xe_sync.c index e4c220cf9115..aab92bee1d7c 100644 --- a/drivers/gpu/drm/xe/xe_sync.c +++ b/drivers/gpu/drm/xe/xe_sync.c @@ -307,7 +307,6 @@ xe_sync_in_fence_get(struct xe_sync_entry *sync, int num_sync, /* Easy case... */ if (!num_in_fence) { fence = xe_exec_queue_last_fence_get(q, vm); - dma_fence_get(fence); return fence; } @@ -322,7 +321,6 @@ xe_sync_in_fence_get(struct xe_sync_entry *sync, int num_sync, } } fences[current_fence++] = xe_exec_queue_last_fence_get(q, vm); - dma_fence_get(fences[current_fence - 1]); cf = dma_fence_array_create(num_in_fence, fences, vm->composite_fence_ctx, vm->composite_fence_seqno++, diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index 30db264d34a3..865e10d0a06a 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -37,8 +37,6 @@ #include "generated/xe_wa_oob.h" #include "xe_wa.h" -#define TEST_VM_ASYNC_OPS_ERROR - static struct drm_gem_object *xe_vm_obj(struct xe_vm *vm) { return vm->gpuvm.r_obj; @@ -114,11 +112,8 @@ retry: num_pages - pinned, read_only ? 0 : FOLL_WRITE, &pages[pinned]); - if (ret < 0) { - if (in_kthread) - ret = 0; + if (ret < 0) break; - } pinned += ret; ret = 0; @@ -1984,6 +1979,7 @@ static int xe_vm_prefetch(struct xe_vm *vm, struct xe_vma *vma, xe_exec_queue_last_fence_get(wait_exec_queue, vm); xe_sync_entry_signal(&syncs[i], NULL, fence); + dma_fence_put(fence); } } @@ -2064,7 +2060,6 @@ vm_bind_ioctl_ops_create(struct xe_vm *vm, struct xe_bo *bo, struct drm_gem_object *obj = bo ? &bo->ttm.base : NULL; struct drm_gpuva_ops *ops; struct drm_gpuva_op *__op; - struct xe_vma_op *op; struct drm_gpuvm_bo *vm_bo; int err; @@ -2111,15 +2106,6 @@ vm_bind_ioctl_ops_create(struct xe_vm *vm, struct xe_bo *bo, if (IS_ERR(ops)) return ops; -#ifdef TEST_VM_ASYNC_OPS_ERROR - if (operation & FORCE_ASYNC_OP_ERROR) { - op = list_first_entry_or_null(&ops->list, struct xe_vma_op, - base.entry); - if (op) - op->inject_error = true; - } -#endif - drm_gpuva_for_each_op(__op, ops) { struct xe_vma_op *op = gpuva_op_to_vma_op(__op); @@ -2199,8 +2185,10 @@ static u64 xe_vma_max_pte_size(struct xe_vma *vma) return SZ_1G; else if (vma->gpuva.flags & XE_VMA_PTE_2M) return SZ_2M; + else if (vma->gpuva.flags & XE_VMA_PTE_4K) + return SZ_4K; - return SZ_4K; + return SZ_1G; /* Uninitialized, used max size */ } static u64 xe_vma_set_pte_size(struct xe_vma *vma, u64 size) @@ -2530,13 +2518,25 @@ retry_userptr: } drm_exec_fini(&exec); - if (err == -EAGAIN && xe_vma_is_userptr(vma)) { + if (err == -EAGAIN) { lockdep_assert_held_write(&vm->lock); - err = xe_vma_userptr_pin_pages(to_userptr_vma(vma)); - if (!err) - goto retry_userptr; - trace_xe_vma_fail(vma); + if (op->base.op == DRM_GPUVA_OP_REMAP) { + if (!op->remap.unmap_done) + vma = gpuva_to_vma(op->base.remap.unmap->va); + else if (op->remap.prev) + vma = op->remap.prev; + else + vma = op->remap.next; + } + + if (xe_vma_is_userptr(vma)) { + err = xe_vma_userptr_pin_pages(to_userptr_vma(vma)); + if (!err) + goto retry_userptr; + + trace_xe_vma_fail(vma); + } } return err; @@ -2548,13 +2548,6 @@ static int xe_vma_op_execute(struct xe_vm *vm, struct xe_vma_op *op) lockdep_assert_held_write(&vm->lock); -#ifdef TEST_VM_ASYNC_OPS_ERROR - if (op->inject_error) { - op->inject_error = false; - return -ENOMEM; - } -#endif - switch (op->base.op) { case DRM_GPUVA_OP_MAP: ret = __xe_vma_op_execute(vm, op->map.vma, op); @@ -2669,7 +2662,7 @@ static void vm_bind_ioctl_ops_unwind(struct xe_vm *vm, { int i; - for (i = num_ops_list - 1; i; ++i) { + for (i = num_ops_list - 1; i >= 0; --i) { struct drm_gpuva_ops *__ops = ops[i]; struct drm_gpuva_op *__op; @@ -2714,16 +2707,9 @@ static int vm_bind_ioctl_ops_execute(struct xe_vm *vm, return 0; } -#ifdef TEST_VM_ASYNC_OPS_ERROR -#define SUPPORTED_FLAGS \ - (FORCE_ASYNC_OP_ERROR | DRM_XE_VM_BIND_FLAG_READONLY | \ - DRM_XE_VM_BIND_FLAG_IMMEDIATE | DRM_XE_VM_BIND_FLAG_NULL | 0xffff) -#else #define SUPPORTED_FLAGS \ (DRM_XE_VM_BIND_FLAG_READONLY | \ - DRM_XE_VM_BIND_FLAG_IMMEDIATE | DRM_XE_VM_BIND_FLAG_NULL | \ - 0xffff) -#endif + DRM_XE_VM_BIND_FLAG_IMMEDIATE | DRM_XE_VM_BIND_FLAG_NULL) #define XE_64K_PAGE_MASK 0xffffull #define ALL_DRM_XE_SYNCS_FLAGS (DRM_XE_SYNCS_FLAG_WAIT_FOR_OP) diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h index 1fec66ae2eb2..5ac9c5bebabc 100644 --- a/drivers/gpu/drm/xe/xe_vm_types.h +++ b/drivers/gpu/drm/xe/xe_vm_types.h @@ -21,9 +21,6 @@ struct xe_bo; struct xe_sync_entry; struct xe_vm; -#define TEST_VM_ASYNC_OPS_ERROR -#define FORCE_ASYNC_OP_ERROR BIT(31) - #define XE_VMA_READ_ONLY DRM_GPUVA_USERBITS #define XE_VMA_DESTROYED (DRM_GPUVA_USERBITS << 1) #define XE_VMA_ATOMIC_PTE_BIT (DRM_GPUVA_USERBITS << 2) @@ -360,11 +357,6 @@ struct xe_vma_op { /** @flags: operation flags */ enum xe_vma_op_flags flags; -#ifdef TEST_VM_ASYNC_OPS_ERROR - /** @inject_error: inject error to test async op error handling */ - bool inject_error; -#endif - union { /** @map: VMA map operation specific data */ struct xe_vma_op_map map; diff --git a/include/uapi/drm/ivpu_accel.h b/include/uapi/drm/ivpu_accel.h index 63c49318a863..19a13468eca5 100644 --- a/include/uapi/drm/ivpu_accel.h +++ b/include/uapi/drm/ivpu_accel.h @@ -305,6 +305,7 @@ struct drm_ivpu_submit { /* drm_ivpu_bo_wait job status codes */ #define DRM_IVPU_JOB_STATUS_SUCCESS 0 +#define DRM_IVPU_JOB_STATUS_ABORTED 256 /** * struct drm_ivpu_bo_wait - Wait for BO to become inactive |