// SPDX-License-Identifier: GPL-2.0 /* * Copyright 2016-2019 HabanaLabs, Ltd. * All Rights Reserved. */ #include #include "habanalabs.h" #include #include #define HL_CS_FLAGS_TYPE_MASK (HL_CS_FLAGS_SIGNAL | HL_CS_FLAGS_WAIT | \ HL_CS_FLAGS_COLLECTIVE_WAIT) /** * enum hl_cs_wait_status - cs wait status * @CS_WAIT_STATUS_BUSY: cs was not completed yet * @CS_WAIT_STATUS_COMPLETED: cs completed * @CS_WAIT_STATUS_GONE: cs completed but fence is already gone */ enum hl_cs_wait_status { CS_WAIT_STATUS_BUSY, CS_WAIT_STATUS_COMPLETED, CS_WAIT_STATUS_GONE }; static void job_wq_completion(struct work_struct *work); static int _hl_cs_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx, u64 timeout_us, u64 seq, enum hl_cs_wait_status *status, s64 *timestamp); static void cs_do_release(struct kref *ref); static void hl_sob_reset(struct kref *ref) { struct hl_hw_sob *hw_sob = container_of(ref, struct hl_hw_sob, kref); struct hl_device *hdev = hw_sob->hdev; hdev->asic_funcs->reset_sob(hdev, hw_sob); } void hl_sob_reset_error(struct kref *ref) { struct hl_hw_sob *hw_sob = container_of(ref, struct hl_hw_sob, kref); struct hl_device *hdev = hw_sob->hdev; dev_crit(hdev->dev, "SOB release shouldn't be called here, q_idx: %d, sob_id: %d\n", hw_sob->q_idx, hw_sob->sob_id); } /** * hl_gen_sob_mask() - Generates a sob mask to be used in a monitor arm packet * @sob_base: sob base id * @sob_mask: sob user mask, each bit represents a sob offset from sob base * @mask: generated mask * * Return: 0 if given parameters are valid */ int hl_gen_sob_mask(u16 sob_base, u8 sob_mask, u8 *mask) { int i; if (sob_mask == 0) return -EINVAL; if (sob_mask == 0x1) { *mask = ~(1 << (sob_base & 0x7)); } else { /* find msb in order to verify sob range is valid */ for (i = BITS_PER_BYTE - 1 ; i >= 0 ; i--) if (BIT(i) & sob_mask) break; if (i > (HL_MAX_SOBS_PER_MONITOR - (sob_base & 0x7) - 1)) return -EINVAL; *mask = ~sob_mask; } return 0; } static void hl_fence_release(struct kref *kref) { struct hl_fence *fence = container_of(kref, struct hl_fence, refcount); struct hl_cs_compl *hl_cs_cmpl = container_of(fence, struct hl_cs_compl, base_fence); struct hl_device *hdev = hl_cs_cmpl->hdev; /* EBUSY means the CS was never submitted and hence we don't have * an attached hw_sob object that we should handle here */ if (fence->error == -EBUSY) goto free; if ((hl_cs_cmpl->type == CS_TYPE_SIGNAL) || (hl_cs_cmpl->type == CS_TYPE_WAIT) || (hl_cs_cmpl->type == CS_TYPE_COLLECTIVE_WAIT)) { dev_dbg(hdev->dev, "CS 0x%llx type %d finished, sob_id: %d, sob_val: 0x%x\n", hl_cs_cmpl->cs_seq, hl_cs_cmpl->type, hl_cs_cmpl->hw_sob->sob_id, hl_cs_cmpl->sob_val); /* * A signal CS can get completion while the corresponding wait * for signal CS is on its way to the PQ. The wait for signal CS * will get stuck if the signal CS incremented the SOB to its * max value and there are no pending (submitted) waits on this * SOB. * We do the following to void this situation: * 1. The wait for signal CS must get a ref for the signal CS as * soon as possible in cs_ioctl_signal_wait() and put it * before being submitted to the PQ but after it incremented * the SOB refcnt in init_signal_wait_cs(). * 2. Signal/Wait for signal CS will decrement the SOB refcnt * here. * These two measures guarantee that the wait for signal CS will * reset the SOB upon completion rather than the signal CS and * hence the above scenario is avoided. */ kref_put(&hl_cs_cmpl->hw_sob->kref, hl_sob_reset); if (hl_cs_cmpl->type == CS_TYPE_COLLECTIVE_WAIT) hdev->asic_funcs->reset_sob_group(hdev, hl_cs_cmpl->sob_group); } free: kfree(hl_cs_cmpl); } void hl_fence_put(struct hl_fence *fence) { if (fence) kref_put(&fence->refcount, hl_fence_release); } void hl_fence_get(struct hl_fence *fence) { if (fence) kref_get(&fence->refcount); } static void hl_fence_init(struct hl_fence *fence) { kref_init(&fence->refcount); fence->error = 0; fence->timestamp = ktime_set(0, 0); init_completion(&fence->completion); } void cs_get(struct hl_cs *cs) { kref_get(&cs->refcount); } static int cs_get_unless_zero(struct hl_cs *cs) { return kref_get_unless_zero(&cs->refcount); } static void cs_put(struct hl_cs *cs) { kref_put(&cs->refcount, cs_do_release); } static void cs_job_do_release(struct kref *ref) { struct hl_cs_job *job = container_of(ref, struct hl_cs_job, refcount); kfree(job); } static void cs_job_put(struct hl_cs_job *job) { kref_put(&job->refcount, cs_job_do_release); } static bool is_cb_patched(struct hl_device *hdev, struct hl_cs_job *job) { /* * Patched CB is created for external queues jobs, and for H/W queues * jobs if the user CB was allocated by driver and MMU is disabled. */ return (job->queue_type == QUEUE_TYPE_EXT || (job->queue_type == QUEUE_TYPE_HW && job->is_kernel_allocated_cb && !hdev->mmu_enable)); } /* * cs_parser - parse the user command submission * * @hpriv : pointer to the private data of the fd * @job : pointer to the job that holds the command submission info * * The function parses the command submission of the user. It calls the * ASIC specific parser, which returns a list of memory blocks to send * to the device as different command buffers * */ static int cs_parser(struct hl_fpriv *hpriv, struct hl_cs_job *job) { struct hl_device *hdev = hpriv->hdev; struct hl_cs_parser parser; int rc; parser.ctx_id = job->cs->ctx->asid; parser.cs_sequence = job->cs->sequence; parser.job_id = job->id; parser.hw_queue_id = job->hw_queue_id; parser.job_userptr_list = &job->userptr_list; parser.patched_cb = NULL; parser.user_cb = job->user_cb; parser.user_cb_size = job->user_cb_size; parser.queue_type = job->queue_type; parser.is_kernel_allocated_cb = job->is_kernel_allocated_cb; job->patched_cb = NULL; rc = hdev->asic_funcs->cs_parser(hdev, &parser); if (is_cb_patched(hdev, job)) { if (!rc) { job->patched_cb = parser.patched_cb; job->job_cb_size = parser.patched_cb_size; job->contains_dma_pkt = parser.contains_dma_pkt; atomic_inc(&job->patched_cb->cs_cnt); } /* * Whether the parsing worked or not, we don't need the * original CB anymore because it was already parsed and * won't be accessed again for this CS */ atomic_dec(&job->user_cb->cs_cnt); hl_cb_put(job->user_cb); job->user_cb = NULL; } else if (!rc) { job->job_cb_size = job->user_cb_size; } return rc; } static void complete_job(struct hl_device *hdev, struct hl_cs_job *job) { struct hl_cs *cs = job->cs; if (is_cb_patched(hdev, job)) { hl_userptr_delete_list(hdev, &job->userptr_list); /* * We might arrive here from rollback and patched CB wasn't * created, so we need to check it's not NULL */ if (job->patched_cb) { atomic_dec(&job->patched_cb->cs_cnt); hl_cb_put(job->patched_cb); } } /* For H/W queue jobs, if a user CB was allocated by driver and MMU is * enabled, the user CB isn't released in cs_parser() and thus should be * released here. * This is also true for INT queues jobs which were allocated by driver */ if (job->is_kernel_allocated_cb && ((job->queue_type == QUEUE_TYPE_HW && hdev->mmu_enable) || job->queue_type == QUEUE_TYPE_INT)) { atomic_dec(&job->user_cb->cs_cnt); hl_cb_put(job->user_cb); } /* * This is the only place where there can be multiple threads * modifying the list at the same time */ spin_lock(&cs->job_lock); list_del(&job->cs_node); spin_unlock(&cs->job_lock); hl_debugfs_remove_job(hdev, job); if (job->queue_type == QUEUE_TYPE_EXT || job->queue_type == QUEUE_TYPE_HW) cs_put(cs); cs_job_put(job); } static void cs_do_release(struct kref *ref) { struct hl_cs *cs = container_of(ref, struct hl_cs, refcount); struct hl_device *hdev = cs->ctx->hdev; struct hl_cs_job *job, *tmp; cs->completed = true; /* * Although if we reached here it means that all external jobs have * finished, because each one of them took refcnt to CS, we still * need to go over the internal jobs and complete them. Otherwise, we * will have leaked memory and what's worse, the CS object (and * potentially the CTX object) could be released, while the JOB * still holds a pointer to them (but no reference). */ list_for_each_entry_safe(job, tmp, &cs->job_list, cs_node) complete_job(hdev, job); if (!cs->submitted) { /* In case the wait for signal CS was submitted, the put occurs * in init_signal_wait_cs() or collective_wait_init_cs() * right before hanging on the PQ. */ if (cs->type == CS_TYPE_WAIT || cs->type == CS_TYPE_COLLECTIVE_WAIT) hl_fence_put(cs->signal_fence); goto out; } hdev->asic_funcs->hw_queues_lock(hdev); hdev->cs_active_cnt--; if (!hdev->cs_active_cnt) { struct hl_device_idle_busy_ts *ts; ts = &hdev->idle_busy_ts_arr[hdev->idle_busy_ts_idx++]; ts->busy_to_idle_ts = ktime_get(); if (hdev->idle_busy_ts_idx == HL_IDLE_BUSY_TS_ARR_SIZE) hdev->idle_busy_ts_idx = 0; } else if (hdev->cs_active_cnt < 0) { dev_crit(hdev->dev, "CS active cnt %d is negative\n", hdev->cs_active_cnt); } hdev->asic_funcs->hw_queues_unlock(hdev); /* Need to update CI for internal queues */ hl_int_hw_queue_update_ci(cs); /* remove CS from CS mirror list */ spin_lock(&hdev->cs_mirror_lock); list_del_init(&cs->mirror_node); spin_unlock(&hdev->cs_mirror_lock); /* Don't cancel TDR in case this CS was timedout because we might be * running from the TDR context */ if (!cs->timedout && hdev->timeout_jiffies != MAX_SCHEDULE_TIMEOUT) { struct hl_cs *next; if (cs->tdr_active) cancel_delayed_work_sync(&cs->work_tdr); spin_lock(&hdev->cs_mirror_lock); /* queue TDR for next CS */ next = list_first_entry_or_null(&hdev->cs_mirror_list, struct hl_cs, mirror_node); if (next && !next->tdr_active) { next->tdr_active = true; schedule_delayed_work(&next->work_tdr, hdev->timeout_jiffies); } spin_unlock(&hdev->cs_mirror_lock); } out: /* Must be called before hl_ctx_put because inside we use ctx to get * the device */ hl_debugfs_remove_cs(cs); hl_ctx_put(cs->ctx); /* We need to mark an error for not submitted because in that case * the hl fence release flow is different. Mainly, we don't need * to handle hw_sob for signal/wait */ if (cs->timedout) cs->fence->error = -ETIMEDOUT; else if (cs->aborted) cs->fence->error = -EIO; else if (!cs->submitted) cs->fence->error = -EBUSY; if (cs->timestamp) cs->fence->timestamp = ktime_get(); complete_all(&cs->fence->completion); hl_fence_put(cs->fence); kfree(cs->jobs_in_queue_cnt); kfree(cs); } static void cs_timedout(struct work_struct *work) { struct hl_device *hdev; int rc; struct hl_cs *cs = container_of(work, struct hl_cs, work_tdr.work); rc = cs_get_unless_zero(cs); if (!rc) return; if ((!cs->submitted) || (cs->completed)) { cs_put(cs); return; } /* Mark the CS is timed out so we won't try to cancel its TDR */ cs->timedout = true; hdev = cs->ctx->hdev; switch (cs->type) { case CS_TYPE_SIGNAL: dev_err(hdev->dev, "Signal command submission %llu has not finished in time!\n", cs->sequence); break; case CS_TYPE_WAIT: dev_err(hdev->dev, "Wait command submission %llu has not finished in time!\n", cs->sequence); break; case CS_TYPE_COLLECTIVE_WAIT: dev_err(hdev->dev, "Collective Wait command submission %llu has not finished in time!\n", cs->sequence); break; default: dev_err(hdev->dev, "Command submission %llu has not finished in time!\n", cs->sequence); break; } cs_put(cs); if (hdev->reset_on_lockup) hl_device_reset(hdev, false, false); else hdev->needs_reset = true; } static int allocate_cs(struct hl_device *hdev, struct hl_ctx *ctx, enum hl_cs_type cs_type, struct hl_cs **cs_new) { struct hl_cs_counters_atomic *cntr; struct hl_fence *other = NULL; struct hl_cs_compl *cs_cmpl; struct hl_cs *cs; int rc; cntr = &hdev->aggregated_cs_counters; cs = kzalloc(sizeof(*cs), GFP_ATOMIC); if (!cs) return -ENOMEM; cs->ctx = ctx; cs->submitted = false; cs->completed = false; cs->type = cs_type; INIT_LIST_HEAD(&cs->job_list); INIT_DELAYED_WORK(&cs->work_tdr, cs_timedout); kref_init(&cs->refcount); spin_lock_init(&cs->job_lock); cs_cmpl = kmalloc(sizeof(*cs_cmpl), GFP_ATOMIC); if (!cs_cmpl) { rc = -ENOMEM; goto free_cs; } cs_cmpl->hdev = hdev; cs_cmpl->type = cs->type; spin_lock_init(&cs_cmpl->lock); cs->fence = &cs_cmpl->base_fence; spin_lock(&ctx->cs_lock); cs_cmpl->cs_seq = ctx->cs_sequence; other = ctx->cs_pending[cs_cmpl->cs_seq & (hdev->asic_prop.max_pending_cs - 1)]; if (other && !completion_done(&other->completion)) { dev_dbg_ratelimited(hdev->dev, "Rejecting CS because of too many in-flights CS\n"); atomic64_inc(&ctx->cs_counters.max_cs_in_flight_drop_cnt); atomic64_inc(&cntr->max_cs_in_flight_drop_cnt); rc = -EAGAIN; goto free_fence; } cs->jobs_in_queue_cnt = kcalloc(hdev->asic_prop.max_queues, sizeof(*cs->jobs_in_queue_cnt), GFP_ATOMIC); if (!cs->jobs_in_queue_cnt) { rc = -ENOMEM; goto free_fence; } /* init hl_fence */ hl_fence_init(&cs_cmpl->base_fence); cs->sequence = cs_cmpl->cs_seq; ctx->cs_pending[cs_cmpl->cs_seq & (hdev->asic_prop.max_pending_cs - 1)] = &cs_cmpl->base_fence; ctx->cs_sequence++; hl_fence_get(&cs_cmpl->base_fence); hl_fence_put(other); spin_unlock(&ctx->cs_lock); *cs_new = cs; return 0; free_fence: spin_unlock(&ctx->cs_lock); kfree(cs_cmpl); free_cs: kfree(cs); return rc; } static void cs_rollback(struct hl_device *hdev, struct hl_cs *cs) { struct hl_cs_job *job, *tmp; list_for_each_entry_safe(job, tmp, &cs->job_list, cs_node) complete_job(hdev, job); } void hl_cs_rollback_all(struct hl_device *hdev) { int i; struct hl_cs *cs, *tmp; /* flush all completions */ for (i = 0 ; i < hdev->asic_prop.completion_queues_count ; i++) flush_workqueue(hdev->cq_wq[i]); /* Make sure we don't have leftovers in the H/W queues mirror list */ list_for_each_entry_safe(cs, tmp, &hdev->cs_mirror_list, mirror_node) { cs_get(cs); cs->aborted = true; dev_warn_ratelimited(hdev->dev, "Killing CS %d.%llu\n", cs->ctx->asid, cs->sequence); cs_rollback(hdev, cs); cs_put(cs); } } static void job_wq_completion(struct work_struct *work) { struct hl_cs_job *job = container_of(work, struct hl_cs_job, finish_work); struct hl_cs *cs = job->cs; struct hl_device *hdev = cs->ctx->hdev; /* job is no longer needed */ complete_job(hdev, job); } static int validate_queue_index(struct hl_device *hdev, struct hl_cs_chunk *chunk, enum hl_queue_type *queue_type, bool *is_kernel_allocated_cb) { struct asic_fixed_properties *asic = &hdev->asic_prop; struct hw_queue_properties *hw_queue_prop; /* This must be checked here to prevent out-of-bounds access to * hw_queues_props array */ if (chunk->queue_index >= asic->max_queues) { dev_err(hdev->dev, "Queue index %d is invalid\n", chunk->queue_index); return -EINVAL; } hw_queue_prop = &asic->hw_queues_props[chunk->queue_index]; if (hw_queue_prop->type == QUEUE_TYPE_NA) { dev_err(hdev->dev, "Queue index %d is invalid\n", chunk->queue_index); return -EINVAL; } if (hw_queue_prop->driver_only) { dev_err(hdev->dev, "Queue index %d is restricted for the kernel driver\n", chunk->queue_index); return -EINVAL; } /* When hw queue type isn't QUEUE_TYPE_HW, * USER_ALLOC_CB flag shall be referred as "don't care". */ if (hw_queue_prop->type == QUEUE_TYPE_HW) { if (chunk->cs_chunk_flags & HL_CS_CHUNK_FLAGS_USER_ALLOC_CB) { if (!(hw_queue_prop->cb_alloc_flags & CB_ALLOC_USER)) { dev_err(hdev->dev, "Queue index %d doesn't support user CB\n", chunk->queue_index); return -EINVAL; } *is_kernel_allocated_cb = false; } else { if (!(hw_queue_prop->cb_alloc_flags & CB_ALLOC_KERNEL)) { dev_err(hdev->dev, "Queue index %d doesn't support kernel CB\n", chunk->queue_index); return -EINVAL; } *is_kernel_allocated_cb = true; } } else { *is_kernel_allocated_cb = !!(hw_queue_prop->cb_alloc_flags & CB_ALLOC_KERNEL); } *queue_type = hw_queue_prop->type; return 0; } static struct hl_cb *get_cb_from_cs_chunk(struct hl_device *hdev, struct hl_cb_mgr *cb_mgr, struct hl_cs_chunk *chunk) { struct hl_cb *cb; u32 cb_handle; cb_handle = (u32) (chunk->cb_handle >> PAGE_SHIFT); cb = hl_cb_get(hdev, cb_mgr, cb_handle); if (!cb) { dev_err(hdev->dev, "CB handle 0x%x invalid\n", cb_handle); return NULL; } if ((chunk->cb_size < 8) || (chunk->cb_size > cb->size)) { dev_err(hdev->dev, "CB size %u invalid\n", chunk->cb_size); goto release_cb; } atomic_inc(&cb->cs_cnt); return cb; release_cb: hl_cb_put(cb); return NULL; } struct hl_cs_job *hl_cs_allocate_job(struct hl_device *hdev, enum hl_queue_type queue_type, bool is_kernel_allocated_cb) { struct hl_cs_job *job; job = kzalloc(sizeof(*job), GFP_ATOMIC); if (!job) return NULL; kref_init(&job->refcount); job->queue_type = queue_type; job->is_kernel_allocated_cb = is_kernel_allocated_cb; if (is_cb_patched(hdev, job)) INIT_LIST_HEAD(&job->userptr_list); if (job->queue_type == QUEUE_TYPE_EXT) INIT_WORK(&job->finish_work, job_wq_completion); return job; } static enum hl_cs_type hl_cs_get_cs_type(u32 cs_type_flags) { if (cs_type_flags & HL_CS_FLAGS_SIGNAL) return CS_TYPE_SIGNAL; else if (cs_type_flags & HL_CS_FLAGS_WAIT) return CS_TYPE_WAIT; else if (cs_type_flags & HL_CS_FLAGS_COLLECTIVE_WAIT) return CS_TYPE_COLLECTIVE_WAIT; else return CS_TYPE_DEFAULT; } static int hl_cs_sanity_checks(struct hl_fpriv *hpriv, union hl_cs_args *args) { struct hl_device *hdev = hpriv->hdev; struct hl_ctx *ctx = hpriv->ctx; u32 cs_type_flags, num_chunks; enum hl_device_status status; enum hl_cs_type cs_type; if (!hl_device_operational(hdev, &status)) { dev_warn_ratelimited(hdev->dev, "Device is %s. Can't submit new CS\n", hdev->status[status]); return -EBUSY; } cs_type_flags = args->in.cs_flags & HL_CS_FLAGS_TYPE_MASK; if (unlikely(cs_type_flags && !is_power_of_2(cs_type_flags))) { dev_err(hdev->dev, "CS type flags are mutually exclusive, context %d\n", ctx->asid); return -EINVAL; } cs_type = hl_cs_get_cs_type(cs_type_flags); num_chunks = args->in.num_chunks_execute; if (unlikely((cs_type != CS_TYPE_DEFAULT) && !hdev->supports_sync_stream)) { dev_err(hdev->dev, "Sync stream CS is not supported\n"); return -EINVAL; } if (cs_type == CS_TYPE_DEFAULT) { if (!num_chunks) { dev_err(hdev->dev, "Got execute CS with 0 chunks, context %d\n", ctx->asid); return -EINVAL; } } else if (num_chunks != 1) { dev_err(hdev->dev, "Sync stream CS mandates one chunk only, context %d\n", ctx->asid); return -EINVAL; } return 0; } static int hl_cs_copy_chunk_array(struct hl_device *hdev, struct hl_cs_chunk **cs_chunk_array, void __user *chunks, u32 num_chunks) { u32 size_to_copy; if (num_chunks > HL_MAX_JOBS_PER_CS) { dev_err(hdev->dev, "Number of chunks can NOT be larger than %d\n", HL_MAX_JOBS_PER_CS); return -EINVAL; } *cs_chunk_array = kmalloc_array(num_chunks, sizeof(**cs_chunk_array), GFP_ATOMIC); if (!*cs_chunk_array) return -ENOMEM; size_to_copy = num_chunks * sizeof(struct hl_cs_chunk); if (copy_from_user(*cs_chunk_array, chunks, size_to_copy)) { dev_err(hdev->dev, "Failed to copy cs chunk array from user\n"); kfree(*cs_chunk_array); return -EFAULT; } return 0; } static int cs_ioctl_default(struct hl_fpriv *hpriv, void __user *chunks, u32 num_chunks, u64 *cs_seq, bool timestamp) { bool int_queues_only = true; struct hl_device *hdev = hpriv->hdev; struct hl_cs_chunk *cs_chunk_array; struct hl_cs_counters_atomic *cntr; struct hl_cs_job *job; struct hl_cs *cs; struct hl_cb *cb; int rc, i; cntr = &hdev->aggregated_cs_counters; *cs_seq = ULLONG_MAX; rc = hl_cs_copy_chunk_array(hdev, &cs_chunk_array, chunks, num_chunks); if (rc) goto out; /* increment refcnt for context */ hl_ctx_get(hdev, hpriv->ctx); rc = allocate_cs(hdev, hpriv->ctx, CS_TYPE_DEFAULT, &cs); if (rc) { hl_ctx_put(hpriv->ctx); goto free_cs_chunk_array; } cs->timestamp = !!timestamp; *cs_seq = cs->sequence; hl_debugfs_add_cs(cs); /* Validate ALL the CS chunks before submitting the CS */ for (i = 0 ; i < num_chunks ; i++) { struct hl_cs_chunk *chunk = &cs_chunk_array[i]; enum hl_queue_type queue_type; bool is_kernel_allocated_cb; rc = validate_queue_index(hdev, chunk, &queue_type, &is_kernel_allocated_cb); if (rc) { atomic64_inc(&hpriv->ctx->cs_counters.parsing_drop_cnt); atomic64_inc(&cntr->parsing_drop_cnt); goto free_cs_object; } if (is_kernel_allocated_cb) { cb = get_cb_from_cs_chunk(hdev, &hpriv->cb_mgr, chunk); if (!cb) { atomic64_inc( &hpriv->ctx->cs_counters.parsing_drop_cnt); atomic64_inc(&cntr->parsing_drop_cnt); rc = -EINVAL; goto free_cs_object; } } else { cb = (struct hl_cb *) (uintptr_t) chunk->cb_handle; } if (queue_type == QUEUE_TYPE_EXT || queue_type == QUEUE_TYPE_HW) int_queues_only = false; job = hl_cs_allocate_job(hdev, queue_type, is_kernel_allocated_cb); if (!job) { atomic64_inc( &hpriv->ctx->cs_counters.out_of_mem_drop_cnt); atomic64_inc(&cntr->out_of_mem_drop_cnt); dev_err(hdev->dev, "Failed to allocate a new job\n"); rc = -ENOMEM; if (is_kernel_allocated_cb) goto release_cb; goto free_cs_object; } job->id = i + 1; job->cs = cs; job->user_cb = cb; job->user_cb_size = chunk->cb_size; job->hw_queue_id = chunk->queue_index; cs->jobs_in_queue_cnt[job->hw_queue_id]++; list_add_tail(&job->cs_node, &cs->job_list); /* * Increment CS reference. When CS reference is 0, CS is * done and can be signaled to user and free all its resources * Only increment for JOB on external or H/W queues, because * only for those JOBs we get completion */ if (job->queue_type == QUEUE_TYPE_EXT || job->queue_type == QUEUE_TYPE_HW) cs_get(cs); hl_debugfs_add_job(hdev, job); rc = cs_parser(hpriv, job); if (rc) { atomic64_inc(&hpriv->ctx->cs_counters.parsing_drop_cnt); atomic64_inc(&cntr->parsing_drop_cnt); dev_err(hdev->dev, "Failed to parse JOB %d.%llu.%d, err %d, rejecting the CS\n", cs->ctx->asid, cs->sequence, job->id, rc); goto free_cs_object; } } if (int_queues_only) { atomic64_inc(&hpriv->ctx->cs_counters.parsing_drop_cnt); atomic64_inc(&cntr->parsing_drop_cnt); dev_err(hdev->dev, "Reject CS %d.%llu because only internal queues jobs are present\n", cs->ctx->asid, cs->sequence); rc = -EINVAL; goto free_cs_object; } rc = hl_hw_queue_schedule_cs(cs); if (rc) { if (rc != -EAGAIN) dev_err(hdev->dev, "Failed to submit CS %d.%llu to H/W queues, error %d\n", cs->ctx->asid, cs->sequence, rc); goto free_cs_object; } rc = HL_CS_STATUS_SUCCESS; goto put_cs; release_cb: atomic_dec(&cb->cs_cnt); hl_cb_put(cb); free_cs_object: cs_rollback(hdev, cs); *cs_seq = ULLONG_MAX; /* The path below is both for good and erroneous exits */ put_cs: /* We finished with the CS in this function, so put the ref */ cs_put(cs); free_cs_chunk_array: kfree(cs_chunk_array); out: return rc; } static int hl_cs_ctx_switch(struct hl_fpriv *hpriv, union hl_cs_args *args, u64 *cs_seq) { struct hl_device *hdev = hpriv->hdev; struct hl_ctx *ctx = hpriv->ctx; bool need_soft_reset = false; int rc = 0, do_ctx_switch; void __user *chunks; u32 num_chunks, tmp; int ret; do_ctx_switch = atomic_cmpxchg(&ctx->thread_ctx_switch_token, 1, 0); if (do_ctx_switch || (args->in.cs_flags & HL_CS_FLAGS_FORCE_RESTORE)) { mutex_lock(&hpriv->restore_phase_mutex); if (do_ctx_switch) { rc = hdev->asic_funcs->context_switch(hdev, ctx->asid); if (rc) { dev_err_ratelimited(hdev->dev, "Failed to switch to context %d, rejecting CS! %d\n", ctx->asid, rc); /* * If we timedout, or if the device is not IDLE * while we want to do context-switch (-EBUSY), * we need to soft-reset because QMAN is * probably stuck. However, we can't call to * reset here directly because of deadlock, so * need to do it at the very end of this * function */ if ((rc == -ETIMEDOUT) || (rc == -EBUSY)) need_soft_reset = true; mutex_unlock(&hpriv->restore_phase_mutex); goto out; } } hdev->asic_funcs->restore_phase_topology(hdev); chunks = (void __user *) (uintptr_t) args->in.chunks_restore; num_chunks = args->in.num_chunks_restore; if (!num_chunks) { dev_dbg(hdev->dev, "Need to run restore phase but restore CS is empty\n"); rc = 0; } else { rc = cs_ioctl_default(hpriv, chunks, num_chunks, cs_seq, false); } mutex_unlock(&hpriv->restore_phase_mutex); if (rc) { dev_err(hdev->dev, "Failed to submit restore CS for context %d (%d)\n", ctx->asid, rc); goto out; } /* Need to wait for restore completion before execution phase */ if (num_chunks) { enum hl_cs_wait_status status; wait_again: ret = _hl_cs_wait_ioctl(hdev, ctx, jiffies_to_usecs(hdev->timeout_jiffies), *cs_seq, &status, NULL); if (ret) { if (ret == -ERESTARTSYS) { usleep_range(100, 200); goto wait_again; } dev_err(hdev->dev, "Restore CS for context %d failed to complete %d\n", ctx->asid, ret); rc = -ENOEXEC; goto out; } } ctx->thread_ctx_switch_wait_token = 1; } else if (!ctx->thread_ctx_switch_wait_token) { rc = hl_poll_timeout_memory(hdev, &ctx->thread_ctx_switch_wait_token, tmp, (tmp == 1), 100, jiffies_to_usecs(hdev->timeout_jiffies), false); if (rc == -ETIMEDOUT) { dev_err(hdev->dev, "context switch phase timeout (%d)\n", tmp); goto out; } } out: if ((rc == -ETIMEDOUT || rc == -EBUSY) && (need_soft_reset)) hl_device_reset(hdev, false, false); return rc; } static int cs_ioctl_extract_signal_seq(struct hl_device *hdev, struct hl_cs_chunk *chunk, u64 *signal_seq) { u64 *signal_seq_arr = NULL; u32 size_to_copy, signal_seq_arr_len; int rc = 0; signal_seq_arr_len = chunk->num_signal_seq_arr; /* currently only one signal seq is supported */ if (signal_seq_arr_len != 1) { dev_err(hdev->dev, "Wait for signal CS supports only one signal CS seq\n"); return -EINVAL; } signal_seq_arr = kmalloc_array(signal_seq_arr_len, sizeof(*signal_seq_arr), GFP_ATOMIC); if (!signal_seq_arr) return -ENOMEM; size_to_copy = chunk->num_signal_seq_arr * sizeof(*signal_seq_arr); if (copy_from_user(signal_seq_arr, u64_to_user_ptr(chunk->signal_seq_arr), size_to_copy)) { dev_err(hdev->dev, "Failed to copy signal seq array from user\n"); rc = -EFAULT; goto out; } /* currently it is guaranteed to have only one signal seq */ *signal_seq = signal_seq_arr[0]; out: kfree(signal_seq_arr); return rc; } static int cs_ioctl_signal_wait_create_jobs(struct hl_device *hdev, struct hl_ctx *ctx, struct hl_cs *cs, enum hl_queue_type q_type, u32 q_idx) { struct hl_cs_counters_atomic *cntr; struct hl_cs_job *job; struct hl_cb *cb; u32 cb_size; cntr = &hdev->aggregated_cs_counters; job = hl_cs_allocate_job(hdev, q_type, true); if (!job) { atomic64_inc(&ctx->cs_counters.out_of_mem_drop_cnt); atomic64_inc(&cntr->out_of_mem_drop_cnt); dev_err(hdev->dev, "Failed to allocate a new job\n"); return -ENOMEM; } if (cs->type == CS_TYPE_WAIT) cb_size = hdev->asic_funcs->get_wait_cb_size(hdev); else cb_size = hdev->asic_funcs->get_signal_cb_size(hdev); cb = hl_cb_kernel_create(hdev, cb_size, q_type == QUEUE_TYPE_HW && hdev->mmu_enable); if (!cb) { atomic64_inc(&ctx->cs_counters.out_of_mem_drop_cnt); atomic64_inc(&cntr->out_of_mem_drop_cnt); kfree(job); return -EFAULT; } job->id = 0; job->cs = cs; job->user_cb = cb; atomic_inc(&job->user_cb->cs_cnt); job->user_cb_size = cb_size; job->hw_queue_id = q_idx; /* * No need in parsing, user CB is the patched CB. * We call hl_cb_destroy() out of two reasons - we don't need the CB in * the CB idr anymore and to decrement its refcount as it was * incremented inside hl_cb_kernel_create(). */ job->patched_cb = job->user_cb; job->job_cb_size = job->user_cb_size; hl_cb_destroy(hdev, &hdev->kernel_cb_mgr, cb->id << PAGE_SHIFT); /* increment refcount as for external queues we get completion */ cs_get(cs); cs->jobs_in_queue_cnt[job->hw_queue_id]++; list_add_tail(&job->cs_node, &cs->job_list); hl_debugfs_add_job(hdev, job); return 0; } static int cs_ioctl_signal_wait(struct hl_fpriv *hpriv, enum hl_cs_type cs_type, void __user *chunks, u32 num_chunks, u64 *cs_seq, bool timestamp) { struct hl_cs_chunk *cs_chunk_array, *chunk; struct hw_queue_properties *hw_queue_prop; struct hl_device *hdev = hpriv->hdev; struct hl_cs_compl *sig_waitcs_cmpl; u32 q_idx, collective_engine_id = 0; struct hl_fence *sig_fence = NULL; struct hl_ctx *ctx = hpriv->ctx; enum hl_queue_type q_type; struct hl_cs *cs; u64 signal_seq; int rc; *cs_seq = ULLONG_MAX; rc = hl_cs_copy_chunk_array(hdev, &cs_chunk_array, chunks, num_chunks); if (rc) goto out; /* currently it is guaranteed to have only one chunk */ chunk = &cs_chunk_array[0]; if (chunk->queue_index >= hdev->asic_prop.max_queues) { dev_err(hdev->dev, "Queue index %d is invalid\n", chunk->queue_index); rc = -EINVAL; goto free_cs_chunk_array; } q_idx = chunk->queue_index; hw_queue_prop = &hdev->asic_prop.hw_queues_props[q_idx]; q_type = hw_queue_prop->type; if (!hw_queue_prop->supports_sync_stream) { dev_err(hdev->dev, "Queue index %d does not support sync stream operations\n", q_idx); rc = -EINVAL; goto free_cs_chunk_array; } if (cs_type == CS_TYPE_COLLECTIVE_WAIT) { if (!(hw_queue_prop->collective_mode == HL_COLLECTIVE_MASTER)) { dev_err(hdev->dev, "Queue index %d is invalid\n", q_idx); rc = -EINVAL; goto free_cs_chunk_array; } collective_engine_id = chunk->collective_engine_id; } if (cs_type == CS_TYPE_WAIT || cs_type == CS_TYPE_COLLECTIVE_WAIT) { rc = cs_ioctl_extract_signal_seq(hdev, chunk, &signal_seq); if (rc) goto free_cs_chunk_array; sig_fence = hl_ctx_get_fence(ctx, signal_seq); if (IS_ERR(sig_fence)) { dev_err(hdev->dev, "Failed to get signal CS with seq 0x%llx\n", signal_seq); rc = PTR_ERR(sig_fence); goto free_cs_chunk_array; } if (!sig_fence) { /* signal CS already finished */ rc = 0; goto free_cs_chunk_array; } sig_waitcs_cmpl = container_of(sig_fence, struct hl_cs_compl, base_fence); if (sig_waitcs_cmpl->type != CS_TYPE_SIGNAL) { dev_err(hdev->dev, "CS seq 0x%llx is not of a signal CS\n", signal_seq); hl_fence_put(sig_fence); rc = -EINVAL; goto free_cs_chunk_array; } if (completion_done(&sig_fence->completion)) { /* signal CS already finished */ hl_fence_put(sig_fence); rc = 0; goto free_cs_chunk_array; } } /* increment refcnt for context */ hl_ctx_get(hdev, ctx); rc = allocate_cs(hdev, ctx, cs_type, &cs); if (rc) { if (cs_type == CS_TYPE_WAIT || cs_type == CS_TYPE_COLLECTIVE_WAIT) hl_fence_put(sig_fence); hl_ctx_put(ctx); goto free_cs_chunk_array; } cs->timestamp = !!timestamp; /* * Save the signal CS fence for later initialization right before * hanging the wait CS on the queue. */ if (cs_type == CS_TYPE_WAIT || cs_type == CS_TYPE_COLLECTIVE_WAIT) cs->signal_fence = sig_fence; hl_debugfs_add_cs(cs); *cs_seq = cs->sequence; if (cs_type == CS_TYPE_WAIT || cs_type == CS_TYPE_SIGNAL) rc = cs_ioctl_signal_wait_create_jobs(hdev, ctx, cs, q_type, q_idx); else if (cs_type == CS_TYPE_COLLECTIVE_WAIT) rc = hdev->asic_funcs->collective_wait_create_jobs(hdev, ctx, cs, q_idx, collective_engine_id); else rc = -EINVAL; if (rc) goto free_cs_object; rc = hl_hw_queue_schedule_cs(cs); if (rc) { if (rc != -EAGAIN) dev_err(hdev->dev, "Failed to submit CS %d.%llu to H/W queues, error %d\n", ctx->asid, cs->sequence, rc); goto free_cs_object; } rc = HL_CS_STATUS_SUCCESS; goto put_cs; free_cs_object: cs_rollback(hdev, cs); *cs_seq = ULLONG_MAX; /* The path below is both for good and erroneous exits */ put_cs: /* We finished with the CS in this function, so put the ref */ cs_put(cs); free_cs_chunk_array: kfree(cs_chunk_array); out: return rc; } int hl_cs_ioctl(struct hl_fpriv *hpriv, void *data) { union hl_cs_args *args = data; enum hl_cs_type cs_type; u64 cs_seq = ULONG_MAX; void __user *chunks; u32 num_chunks; int rc; rc = hl_cs_sanity_checks(hpriv, args); if (rc) goto out; rc = hl_cs_ctx_switch(hpriv, args, &cs_seq); if (rc) goto out; cs_type = hl_cs_get_cs_type(args->in.cs_flags & ~HL_CS_FLAGS_FORCE_RESTORE); chunks = (void __user *) (uintptr_t) args->in.chunks_execute; num_chunks = args->in.num_chunks_execute; switch (cs_type) { case CS_TYPE_SIGNAL: case CS_TYPE_WAIT: case CS_TYPE_COLLECTIVE_WAIT: rc = cs_ioctl_signal_wait(hpriv, cs_type, chunks, num_chunks, &cs_seq, args->in.cs_flags & HL_CS_FLAGS_TIMESTAMP); break; default: rc = cs_ioctl_default(hpriv, chunks, num_chunks, &cs_seq, args->in.cs_flags & HL_CS_FLAGS_TIMESTAMP); break; } out: if (rc != -EAGAIN) { memset(args, 0, sizeof(*args)); args->out.status = rc; args->out.seq = cs_seq; } return rc; } static int _hl_cs_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx, u64 timeout_us, u64 seq, enum hl_cs_wait_status *status, s64 *timestamp) { struct hl_fence *fence; unsigned long timeout; int rc = 0; long completion_rc; if (timestamp) *timestamp = 0; if (timeout_us == MAX_SCHEDULE_TIMEOUT) timeout = timeout_us; else timeout = usecs_to_jiffies(timeout_us); hl_ctx_get(hdev, ctx); fence = hl_ctx_get_fence(ctx, seq); if (IS_ERR(fence)) { rc = PTR_ERR(fence); if (rc == -EINVAL) dev_notice_ratelimited(hdev->dev, "Can't wait on CS %llu because current CS is at seq %llu\n", seq, ctx->cs_sequence); } else if (fence) { if (!timeout_us) completion_rc = completion_done(&fence->completion); else completion_rc = wait_for_completion_interruptible_timeout( &fence->completion, timeout); if (completion_rc > 0) { *status = CS_WAIT_STATUS_COMPLETED; if (timestamp) *timestamp = ktime_to_ns(fence->timestamp); } else { *status = CS_WAIT_STATUS_BUSY; } if (fence->error == -ETIMEDOUT) rc = -ETIMEDOUT; else if (fence->error == -EIO) rc = -EIO; hl_fence_put(fence); } else { dev_dbg(hdev->dev, "Can't wait on seq %llu because current CS is at seq %llu (Fence is gone)\n", seq, ctx->cs_sequence); *status = CS_WAIT_STATUS_GONE; } hl_ctx_put(ctx); return rc; } int hl_cs_wait_ioctl(struct hl_fpriv *hpriv, void *data) { struct hl_device *hdev = hpriv->hdev; union hl_wait_cs_args *args = data; enum hl_cs_wait_status status; u64 seq = args->in.seq; s64 timestamp; int rc; rc = _hl_cs_wait_ioctl(hdev, hpriv->ctx, args->in.timeout_us, seq, &status, ×tamp); memset(args, 0, sizeof(*args)); if (rc) { if (rc == -ERESTARTSYS) { dev_err_ratelimited(hdev->dev, "user process got signal while waiting for CS handle %llu\n", seq); args->out.status = HL_WAIT_CS_STATUS_INTERRUPTED; rc = -EINTR; } else if (rc == -ETIMEDOUT) { dev_err_ratelimited(hdev->dev, "CS %llu has timed-out while user process is waiting for it\n", seq); args->out.status = HL_WAIT_CS_STATUS_TIMEDOUT; } else if (rc == -EIO) { dev_err_ratelimited(hdev->dev, "CS %llu has been aborted while user process is waiting for it\n", seq); args->out.status = HL_WAIT_CS_STATUS_ABORTED; } return rc; } if (timestamp) { args->out.flags |= HL_WAIT_CS_STATUS_FLAG_TIMESTAMP_VLD; args->out.timestamp_nsec = timestamp; } switch (status) { case CS_WAIT_STATUS_GONE: args->out.flags |= HL_WAIT_CS_STATUS_FLAG_GONE; fallthrough; case CS_WAIT_STATUS_COMPLETED: args->out.status = HL_WAIT_CS_STATUS_COMPLETED; break; case CS_WAIT_STATUS_BUSY: default: args->out.status = HL_WAIT_CS_STATUS_BUSY; break; } return 0; }