From a2bf92e8cc16f6bf33bc8053c776630c79ad0325 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 25 Sep 2018 09:31:59 +0100 Subject: drm/i915/execlists: Avoid kicking priority on the current context If the request is currently on the HW (in port 0), then we do not need to kick the submission tasklet to evaluate whether we should be preempting itself in order to execute it again. In the case that was annoying me: execlists_schedule: rq(18:211173).prio=0 -> 2 need_preempt: last(18:211174).prio=0, queue.prio=2 We are bumping the priority of the first of a pair of requests running in the current context. Then when evaluating preempt, we would see that that our priority request is higher than the last executing request in ELSP0 and so trigger preemption, not realising that our intended request was already executing. v2: As we assume state of the execlists->port[] that is only valid while we hold the timeline lock we have to repeat some earlier tests that on the validity of the node. v3: Wrap guc submission under the timeline.lock as is now the way of all things. Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Reviewed-by: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20180925083205.2229-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_guc_submission.c | 18 ++++++------------ 1 file changed, 6 insertions(+), 12 deletions(-) (limited to 'drivers/gpu/drm/i915/intel_guc_submission.c') diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index a81f04d46e87..4874a212754c 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -791,19 +791,8 @@ done: static void guc_dequeue(struct intel_engine_cs *engine) { - unsigned long flags; - bool submit; - - local_irq_save(flags); - - spin_lock(&engine->timeline.lock); - submit = __guc_dequeue(engine); - spin_unlock(&engine->timeline.lock); - - if (submit) + if (__guc_dequeue(engine)) guc_submit(engine); - - local_irq_restore(flags); } static void guc_submission_tasklet(unsigned long data) @@ -812,6 +801,9 @@ static void guc_submission_tasklet(unsigned long data) struct intel_engine_execlists * const execlists = &engine->execlists; struct execlist_port *port = execlists->port; struct i915_request *rq; + unsigned long flags; + + spin_lock_irqsave(&engine->timeline.lock, flags); rq = port_request(port); while (rq && i915_request_completed(rq)) { @@ -835,6 +827,8 @@ static void guc_submission_tasklet(unsigned long data) if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_PREEMPT)) guc_dequeue(engine); + + spin_unlock_irqrestore(&engine->timeline.lock, flags); } static struct i915_request * -- cgit v1.2.3 From 85f5e1f385b7643ee31e0530a1daa2438ac27aaf Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 1 Oct 2018 13:32:04 +0100 Subject: drm/i915: Combine multiple internal plists into the same i915_priolist bucket As we are about to allow ourselves to slightly bump the user priority into a few different sublevels, packthose internal priority lists into the same i915_priolist to keep the rbtree compact and avoid having to allocate the default user priority even after the internal bumping. The downside to having an requests[] rather than a node per active list, is that we then have to walk over the empty higher priority lists. To compensate, we track the active buckets and use a small bitmap to skip over any inactive ones. v2: Use MASK of internal levels to simplify our usage. v3: Prevent overflow when SHIFT is zero. Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Reviewed-by: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20181001123204.23982-4-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_engine_cs.c | 6 +- drivers/gpu/drm/i915/intel_guc_submission.c | 12 ++-- drivers/gpu/drm/i915/intel_lrc.c | 87 ++++++++++++++++++++--------- drivers/gpu/drm/i915/intel_ringbuffer.h | 13 ++++- 4 files changed, 80 insertions(+), 38 deletions(-) (limited to 'drivers/gpu/drm/i915/intel_guc_submission.c') diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 6726d57f018f..1c6143bdf5a4 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1534,10 +1534,10 @@ void intel_engine_dump(struct intel_engine_cs *engine, count = 0; drm_printf(m, "\t\tQueue priority: %d\n", execlists->queue_priority); for (rb = rb_first_cached(&execlists->queue); rb; rb = rb_next(rb)) { - struct i915_priolist *p = - rb_entry(rb, typeof(*p), node); + struct i915_priolist *p = rb_entry(rb, typeof(*p), node); + int i; - list_for_each_entry(rq, &p->requests, sched.link) { + priolist_for_each_request(rq, p, i) { if (count++ < MAX_REQUESTS_TO_SHOW - 1) print_request(m, rq, "\t\tQ "); else diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index 4874a212754c..ac862b42f6a1 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -746,30 +746,28 @@ static bool __guc_dequeue(struct intel_engine_cs *engine) while ((rb = rb_first_cached(&execlists->queue))) { struct i915_priolist *p = to_priolist(rb); struct i915_request *rq, *rn; + int i; - list_for_each_entry_safe(rq, rn, &p->requests, sched.link) { + priolist_for_each_request_consume(rq, rn, p, i) { if (last && rq->hw_context != last->hw_context) { - if (port == last_port) { - __list_del_many(&p->requests, - &rq->sched.link); + if (port == last_port) goto done; - } if (submit) port_assign(port, last); port++; } - INIT_LIST_HEAD(&rq->sched.link); + list_del_init(&rq->sched.link); __i915_request_submit(rq); trace_i915_request_in(rq, port_index(port, execlists)); + last = rq; submit = true; } rb_erase_cached(&p->node, &execlists->queue); - INIT_LIST_HEAD(&p->requests); if (p->priority != I915_PRIORITY_NORMAL) kmem_cache_free(engine->i915->priorities, p); } diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 15345e74d8ce..4ee00f531153 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -259,14 +259,49 @@ intel_lr_context_descriptor_update(struct i915_gem_context *ctx, ce->lrc_desc = desc; } -static struct i915_priolist * +static void assert_priolists(struct intel_engine_execlists * const execlists, + long queue_priority) +{ + struct rb_node *rb; + long last_prio, i; + + if (!IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) + return; + + GEM_BUG_ON(rb_first_cached(&execlists->queue) != + rb_first(&execlists->queue.rb_root)); + + last_prio = (queue_priority >> I915_USER_PRIORITY_SHIFT) + 1; + for (rb = rb_first_cached(&execlists->queue); rb; rb = rb_next(rb)) { + struct i915_priolist *p = to_priolist(rb); + + GEM_BUG_ON(p->priority >= last_prio); + last_prio = p->priority; + + GEM_BUG_ON(!p->used); + for (i = 0; i < ARRAY_SIZE(p->requests); i++) { + if (list_empty(&p->requests[i])) + continue; + + GEM_BUG_ON(!(p->used & BIT(i))); + } + } +} + +static struct list_head * lookup_priolist(struct intel_engine_cs *engine, int prio) { struct intel_engine_execlists * const execlists = &engine->execlists; struct i915_priolist *p; struct rb_node **parent, *rb; bool first = true; + int idx, i; + + assert_priolists(execlists, INT_MAX); + /* buckets sorted from highest [in slot 0] to lowest priority */ + idx = I915_PRIORITY_COUNT - (prio & I915_PRIORITY_MASK) - 1; + prio >>= I915_USER_PRIORITY_SHIFT; if (unlikely(execlists->no_priolist)) prio = I915_PRIORITY_NORMAL; @@ -283,7 +318,7 @@ find_priolist: parent = &rb->rb_right; first = false; } else { - return p; + goto out; } } @@ -309,11 +344,15 @@ find_priolist: } p->priority = prio; - INIT_LIST_HEAD(&p->requests); + for (i = 0; i < ARRAY_SIZE(p->requests); i++) + INIT_LIST_HEAD(&p->requests[i]); rb_link_node(&p->node, rb, parent); rb_insert_color_cached(&p->node, &execlists->queue, first); + p->used = 0; - return p; +out: + p->used |= BIT(idx); + return &p->requests[idx]; } static void unwind_wa_tail(struct i915_request *rq) @@ -325,7 +364,7 @@ static void unwind_wa_tail(struct i915_request *rq) static void __unwind_incomplete_requests(struct intel_engine_cs *engine) { struct i915_request *rq, *rn; - struct i915_priolist *uninitialized_var(p); + struct list_head *uninitialized_var(pl); int last_prio = I915_PRIORITY_INVALID; lockdep_assert_held(&engine->timeline.lock); @@ -342,12 +381,11 @@ static void __unwind_incomplete_requests(struct intel_engine_cs *engine) GEM_BUG_ON(rq_prio(rq) == I915_PRIORITY_INVALID); if (rq_prio(rq) != last_prio) { last_prio = rq_prio(rq); - p = lookup_priolist(engine, last_prio); + pl = lookup_priolist(engine, last_prio); } GEM_BUG_ON(RB_EMPTY_ROOT(&engine->execlists.queue.rb_root)); - GEM_BUG_ON(p->priority != rq_prio(rq)); - list_add(&rq->sched.link, &p->requests); + list_add(&rq->sched.link, pl); } } @@ -664,8 +702,9 @@ static void execlists_dequeue(struct intel_engine_cs *engine) while ((rb = rb_first_cached(&execlists->queue))) { struct i915_priolist *p = to_priolist(rb); struct i915_request *rq, *rn; + int i; - list_for_each_entry_safe(rq, rn, &p->requests, sched.link) { + priolist_for_each_request_consume(rq, rn, p, i) { /* * Can we combine this request with the current port? * It has to be the same context/ringbuffer and not @@ -684,11 +723,8 @@ static void execlists_dequeue(struct intel_engine_cs *engine) * combine this request with the last, then we * are done. */ - if (port == last_port) { - __list_del_many(&p->requests, - &rq->sched.link); + if (port == last_port) goto done; - } /* * If GVT overrides us we only ever submit @@ -698,11 +734,8 @@ static void execlists_dequeue(struct intel_engine_cs *engine) * request) to the second port. */ if (ctx_single_port_submission(last->hw_context) || - ctx_single_port_submission(rq->hw_context)) { - __list_del_many(&p->requests, - &rq->sched.link); + ctx_single_port_submission(rq->hw_context)) goto done; - } GEM_BUG_ON(last->hw_context == rq->hw_context); @@ -713,15 +746,16 @@ static void execlists_dequeue(struct intel_engine_cs *engine) GEM_BUG_ON(port_isset(port)); } - INIT_LIST_HEAD(&rq->sched.link); + list_del_init(&rq->sched.link); + __i915_request_submit(rq); trace_i915_request_in(rq, port_index(port, execlists)); + last = rq; submit = true; } rb_erase_cached(&p->node, &execlists->queue); - INIT_LIST_HEAD(&p->requests); if (p->priority != I915_PRIORITY_NORMAL) kmem_cache_free(engine->i915->priorities, p); } @@ -745,6 +779,7 @@ done: */ execlists->queue_priority = port != execlists->port ? rq_prio(last) : INT_MIN; + assert_priolists(execlists, execlists->queue_priority); if (submit) { port_assign(port, last); @@ -856,16 +891,16 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine) /* Flush the queued requests to the timeline list (for retiring). */ while ((rb = rb_first_cached(&execlists->queue))) { struct i915_priolist *p = to_priolist(rb); + int i; - list_for_each_entry_safe(rq, rn, &p->requests, sched.link) { - INIT_LIST_HEAD(&rq->sched.link); + priolist_for_each_request_consume(rq, rn, p, i) { + list_del_init(&rq->sched.link); dma_fence_set_error(&rq->fence, -EIO); __i915_request_submit(rq); } rb_erase_cached(&p->node, &execlists->queue); - INIT_LIST_HEAD(&p->requests); if (p->priority != I915_PRIORITY_NORMAL) kmem_cache_free(engine->i915->priorities, p); } @@ -1071,8 +1106,7 @@ static void queue_request(struct intel_engine_cs *engine, struct i915_sched_node *node, int prio) { - list_add_tail(&node->link, - &lookup_priolist(engine, prio)->requests); + list_add_tail(&node->link, lookup_priolist(engine, prio)); } static void __update_queue(struct intel_engine_cs *engine, int prio) @@ -1142,7 +1176,7 @@ sched_lock_engine(struct i915_sched_node *node, struct intel_engine_cs *locked) static void execlists_schedule(struct i915_request *request, const struct i915_sched_attr *attr) { - struct i915_priolist *uninitialized_var(pl); + struct list_head *uninitialized_var(pl); struct intel_engine_cs *engine, *last; struct i915_dependency *dep, *p; struct i915_dependency stack; @@ -1241,8 +1275,7 @@ static void execlists_schedule(struct i915_request *request, pl = lookup_priolist(engine, prio); last = engine; } - GEM_BUG_ON(pl->priority != prio); - list_move_tail(&node->link, &pl->requests); + list_move_tail(&node->link, pl); } else { /* * If the request is not in the priolist queue because diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 2dfa585712c2..1534de5bb852 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -190,11 +190,22 @@ enum intel_engine_id { }; struct i915_priolist { + struct list_head requests[I915_PRIORITY_COUNT]; struct rb_node node; - struct list_head requests; + unsigned long used; int priority; }; +#define priolist_for_each_request(it, plist, idx) \ + for (idx = 0; idx < ARRAY_SIZE((plist)->requests); idx++) \ + list_for_each_entry(it, &(plist)->requests[idx], sched.link) + +#define priolist_for_each_request_consume(it, n, plist, idx) \ + for (; (idx = ffs((plist)->used)); (plist)->used &= ~BIT(idx - 1)) \ + list_for_each_entry_safe(it, n, \ + &(plist)->requests[idx - 1], \ + sched.link) + struct st_preempt_hang { struct completion completion; bool inject_hang; -- cgit v1.2.3 From 48b426a9b9ab93481a2c5b913c2c6add5fb1001a Mon Sep 17 00:00:00 2001 From: Daniele Ceraolo Spurio Date: Tue, 2 Oct 2018 14:54:28 -0700 Subject: drm/i915/guc: init GuC descriptors after GuC load GuC stores some data in there, which might be stale after a reset. We already reset the WQ head and tail, but more things are being moved to the descriptor with the interface updates. Instead of trying to track them one by one, always memset and init the descriptors from scratch after GuC is loaded. The code is also reorganized so that the above operations and the doorbell creation are grouped as "client enabling" v2: add proc_desc_fini for symmetry (Daniele), remove unneeded var init, add guc_is_alive() (Michal) Cc: Michal Wajdeczko Signed-off-by: Daniele Ceraolo Spurio Reviewed-by: Michal Wajdeczko Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20181002215430.15049-1-daniele.ceraolospurio@intel.com --- drivers/gpu/drm/i915/intel_guc.h | 5 + drivers/gpu/drm/i915/intel_guc_submission.c | 140 +++++++++++++++------------- drivers/gpu/drm/i915/intel_uc_fw.h | 7 +- drivers/gpu/drm/i915/selftests/intel_guc.c | 17 ++-- 4 files changed, 96 insertions(+), 73 deletions(-) (limited to 'drivers/gpu/drm/i915/intel_guc_submission.c') diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h index ad42faf48c46..0f1c4f9ebfd8 100644 --- a/drivers/gpu/drm/i915/intel_guc.h +++ b/drivers/gpu/drm/i915/intel_guc.h @@ -95,6 +95,11 @@ struct intel_guc { void (*notify)(struct intel_guc *guc); }; +static inline bool intel_guc_is_alive(struct intel_guc *guc) +{ + return intel_uc_fw_is_loaded(&guc->fw); +} + static inline int intel_guc_send(struct intel_guc *guc, const u32 *action, u32 len) { diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index ac862b42f6a1..aa4d6bbdd1e9 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -282,8 +282,7 @@ __get_process_desc(struct intel_guc_client *client) /* * Initialise the process descriptor shared with the GuC firmware. */ -static void guc_proc_desc_init(struct intel_guc *guc, - struct intel_guc_client *client) +static void guc_proc_desc_init(struct intel_guc_client *client) { struct guc_process_desc *desc; @@ -304,6 +303,14 @@ static void guc_proc_desc_init(struct intel_guc *guc, desc->priority = client->priority; } +static void guc_proc_desc_fini(struct intel_guc_client *client) +{ + struct guc_process_desc *desc; + + desc = __get_process_desc(client); + memset(desc, 0, sizeof(*desc)); +} + static int guc_stage_desc_pool_create(struct intel_guc *guc) { struct i915_vma *vma; @@ -341,9 +348,9 @@ static void guc_stage_desc_pool_destroy(struct intel_guc *guc) * data structures relating to this client (doorbell, process descriptor, * write queue, etc). */ -static void guc_stage_desc_init(struct intel_guc *guc, - struct intel_guc_client *client) +static void guc_stage_desc_init(struct intel_guc_client *client) { + struct intel_guc *guc = client->guc; struct drm_i915_private *dev_priv = guc_to_i915(guc); struct intel_engine_cs *engine; struct i915_gem_context *ctx = client->owner; @@ -424,8 +431,7 @@ static void guc_stage_desc_init(struct intel_guc *guc, desc->desc_private = ptr_to_u64(client); } -static void guc_stage_desc_fini(struct intel_guc *guc, - struct intel_guc_client *client) +static void guc_stage_desc_fini(struct intel_guc_client *client) { struct guc_stage_desc *desc; @@ -486,14 +492,6 @@ static void guc_wq_item_append(struct intel_guc_client *client, WRITE_ONCE(desc->tail, (wq_off + wqi_size) & (GUC_WQ_SIZE - 1)); } -static void guc_reset_wq(struct intel_guc_client *client) -{ - struct guc_process_desc *desc = __get_process_desc(client); - - desc->head = 0; - desc->tail = 0; -} - static void guc_ring_doorbell(struct intel_guc_client *client) { struct guc_doorbell_info *db; @@ -898,45 +896,6 @@ static bool guc_verify_doorbells(struct intel_guc *guc) return true; } -static int guc_clients_doorbell_init(struct intel_guc *guc) -{ - int ret; - - ret = create_doorbell(guc->execbuf_client); - if (ret) - return ret; - - if (guc->preempt_client) { - ret = create_doorbell(guc->preempt_client); - if (ret) { - destroy_doorbell(guc->execbuf_client); - return ret; - } - } - - return 0; -} - -static void guc_clients_doorbell_fini(struct intel_guc *guc) -{ - /* - * By the time we're here, GuC has already been reset. - * Instead of trying (in vain) to communicate with it, let's just - * cleanup the doorbell HW and our internal state. - */ - if (guc->preempt_client) { - __destroy_doorbell(guc->preempt_client); - __update_doorbell_desc(guc->preempt_client, - GUC_DOORBELL_INVALID); - } - - if (guc->execbuf_client) { - __destroy_doorbell(guc->execbuf_client); - __update_doorbell_desc(guc->execbuf_client, - GUC_DOORBELL_INVALID); - } -} - /** * guc_client_alloc() - Allocate an intel_guc_client * @dev_priv: driver private data structure @@ -1009,9 +968,6 @@ guc_client_alloc(struct drm_i915_private *dev_priv, else client->proc_desc_offset = (GUC_DB_SIZE / 2); - guc_proc_desc_init(guc, client); - guc_stage_desc_init(guc, client); - ret = reserve_doorbell(client); if (ret) goto err_vaddr; @@ -1037,7 +993,6 @@ err_client: static void guc_client_free(struct intel_guc_client *client) { unreserve_doorbell(client); - guc_stage_desc_fini(client->guc, client); i915_vma_unpin_and_release(&client->vma, I915_VMA_RELEASE_MAP); ida_simple_remove(&client->guc->stage_ids, client->stage_id); kfree(client); @@ -1104,6 +1059,69 @@ static void guc_clients_destroy(struct intel_guc *guc) guc_client_free(client); } +static int __guc_client_enable(struct intel_guc_client *client) +{ + int ret; + + guc_proc_desc_init(client); + guc_stage_desc_init(client); + + ret = create_doorbell(client); + if (ret) + goto fail; + + return 0; + +fail: + guc_stage_desc_fini(client); + guc_proc_desc_fini(client); + return ret; +} + +static void __guc_client_disable(struct intel_guc_client *client) +{ + /* + * By the time we're here, GuC may have already been reset. if that is + * the case, instead of trying (in vain) to communicate with it, let's + * just cleanup the doorbell HW and our internal state. + */ + if (intel_guc_is_alive(client->guc)) + destroy_doorbell(client); + else + __destroy_doorbell(client); + + guc_stage_desc_fini(client); + guc_proc_desc_fini(client); +} + +static int guc_clients_enable(struct intel_guc *guc) +{ + int ret; + + ret = __guc_client_enable(guc->execbuf_client); + if (ret) + return ret; + + if (guc->preempt_client) { + ret = __guc_client_enable(guc->preempt_client); + if (ret) { + __guc_client_disable(guc->execbuf_client); + return ret; + } + } + + return 0; +} + +static void guc_clients_disable(struct intel_guc *guc) +{ + if (guc->preempt_client) + __guc_client_disable(guc->preempt_client); + + if (guc->execbuf_client) + __guc_client_disable(guc->execbuf_client); +} + /* * Set up the memory resources to be shared with the GuC (via the GGTT) * at firmware loading time. @@ -1287,15 +1305,11 @@ int intel_guc_submission_enable(struct intel_guc *guc) GEM_BUG_ON(!guc->execbuf_client); - guc_reset_wq(guc->execbuf_client); - if (guc->preempt_client) - guc_reset_wq(guc->preempt_client); - err = intel_guc_sample_forcewake(guc); if (err) return err; - err = guc_clients_doorbell_init(guc); + err = guc_clients_enable(guc); if (err) return err; @@ -1317,7 +1331,7 @@ void intel_guc_submission_disable(struct intel_guc *guc) GEM_BUG_ON(dev_priv->gt.awake); /* GT should be parked first */ guc_interrupts_release(dev_priv); - guc_clients_doorbell_fini(guc); + guc_clients_disable(guc); } #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) diff --git a/drivers/gpu/drm/i915/intel_uc_fw.h b/drivers/gpu/drm/i915/intel_uc_fw.h index 87910aa83267..0e3bd580e267 100644 --- a/drivers/gpu/drm/i915/intel_uc_fw.h +++ b/drivers/gpu/drm/i915/intel_uc_fw.h @@ -115,9 +115,14 @@ static inline bool intel_uc_fw_is_selected(struct intel_uc_fw *uc_fw) return uc_fw->path != NULL; } +static inline bool intel_uc_fw_is_loaded(struct intel_uc_fw *uc_fw) +{ + return uc_fw->load_status == INTEL_UC_FIRMWARE_SUCCESS; +} + static inline void intel_uc_fw_sanitize(struct intel_uc_fw *uc_fw) { - if (uc_fw->load_status == INTEL_UC_FIRMWARE_SUCCESS) + if (intel_uc_fw_is_loaded(uc_fw)) uc_fw->load_status = INTEL_UC_FIRMWARE_PENDING; } diff --git a/drivers/gpu/drm/i915/selftests/intel_guc.c b/drivers/gpu/drm/i915/selftests/intel_guc.c index 0c0ab82b6228..bf27162fb327 100644 --- a/drivers/gpu/drm/i915/selftests/intel_guc.c +++ b/drivers/gpu/drm/i915/selftests/intel_guc.c @@ -159,6 +159,7 @@ static int igt_guc_clients(void *args) * Get rid of clients created during driver load because the test will * recreate them. */ + guc_clients_disable(guc); guc_clients_destroy(guc); if (guc->execbuf_client || guc->preempt_client) { pr_err("guc_clients_destroy lied!\n"); @@ -197,8 +198,8 @@ static int igt_guc_clients(void *args) goto out; } - /* Now create the doorbells */ - guc_clients_doorbell_init(guc); + /* Now enable the clients */ + guc_clients_enable(guc); /* each client should now have received a doorbell */ if (!client_doorbell_in_sync(guc->execbuf_client) || @@ -212,7 +213,7 @@ static int igt_guc_clients(void *args) * Basic test - an attempt to reallocate a valid doorbell to the * client it is currently assigned should not cause a failure. */ - err = guc_clients_doorbell_init(guc); + err = create_doorbell(guc->execbuf_client); if (err) goto out; @@ -263,12 +264,10 @@ out: * Leave clean state for other test, plus the driver always destroy the * clients during unload. */ - destroy_doorbell(guc->execbuf_client); - if (guc->preempt_client) - destroy_doorbell(guc->preempt_client); + guc_clients_disable(guc); guc_clients_destroy(guc); guc_clients_create(guc); - guc_clients_doorbell_init(guc); + guc_clients_enable(guc); unlock: intel_runtime_pm_put(dev_priv); mutex_unlock(&dev_priv->drm.struct_mutex); @@ -352,7 +351,7 @@ static int igt_guc_doorbells(void *arg) db_id = clients[i]->doorbell_id; - err = create_doorbell(clients[i]); + err = __guc_client_enable(clients[i]); if (err) { pr_err("[%d] Failed to create a doorbell\n", i); goto out; @@ -378,7 +377,7 @@ static int igt_guc_doorbells(void *arg) out: for (i = 0; i < ATTEMPTS; i++) if (!IS_ERR_OR_NULL(clients[i])) { - destroy_doorbell(clients[i]); + __guc_client_disable(clients[i]); guc_client_free(clients[i]); } unlock: -- cgit v1.2.3 From 0fc645f338e3e1a4ee05e5529d86ac596c5921b6 Mon Sep 17 00:00:00 2001 From: Daniele Ceraolo Spurio Date: Tue, 2 Oct 2018 14:54:29 -0700 Subject: drm/i915/guc: Don't clear the cookie on doorbell destroy If the HW has not processed the db invalidation request yet, clearing the cookie can generate a db ring. We clear the cookie when we (re-)allocate the doorbell so no need to do it on destroy as well as no one is going to look at it while the doorbell is inactive v2: fix typo in patch title (Michal) Cc: Michal Wajdeczko Signed-off-by: Daniele Ceraolo Spurio Reviewed-by: Michal Wajdeczko Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20181002215430.15049-2-daniele.ceraolospurio@intel.com --- drivers/gpu/drm/i915/intel_guc_submission.c | 1 - 1 file changed, 1 deletion(-) (limited to 'drivers/gpu/drm/i915/intel_guc_submission.c') diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index aa4d6bbdd1e9..eae668442ebe 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -209,7 +209,6 @@ static void __destroy_doorbell(struct intel_guc_client *client) doorbell = __get_doorbell(client); doorbell->db_status = GUC_DOORBELL_DISABLED; - doorbell->cookie = 0; /* Doorbell release flow requires that we wait for GEN8_DRB_VALID bit * to go to zero after updating db_status before we call the GuC to -- cgit v1.2.3 From 6f1312e6e046de8a77aa5e657e269f1968eca00c Mon Sep 17 00:00:00 2001 From: Daniele Ceraolo Spurio Date: Mon, 22 Oct 2018 16:04:22 -0700 Subject: drm/i915/guc: rename __create/destroy_doorbell The 2 functions don't create or destroy anything, they just update the doorbell state in memory. Use init and fini instead for clarity. Cc: Michal Wajdeczko Reviewed-by: Michal Wajdeczko Signed-off-by: Daniele Ceraolo Spurio Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20181022230427.5616-1-daniele.ceraolospurio@intel.com --- drivers/gpu/drm/i915/intel_guc_submission.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/i915/intel_guc_submission.c') diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index eae668442ebe..b089e5283307 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -192,7 +192,7 @@ static struct guc_doorbell_info *__get_doorbell(struct intel_guc_client *client) return client->vaddr + client->doorbell_offset; } -static void __create_doorbell(struct intel_guc_client *client) +static void __init_doorbell(struct intel_guc_client *client) { struct guc_doorbell_info *doorbell; @@ -201,7 +201,7 @@ static void __create_doorbell(struct intel_guc_client *client) doorbell->cookie = 0; } -static void __destroy_doorbell(struct intel_guc_client *client) +static void __fini_doorbell(struct intel_guc_client *client) { struct drm_i915_private *dev_priv = guc_to_i915(client->guc); struct guc_doorbell_info *doorbell; @@ -226,11 +226,11 @@ static int create_doorbell(struct intel_guc_client *client) return -ENODEV; /* internal setup error, should never happen */ __update_doorbell_desc(client, client->doorbell_id); - __create_doorbell(client); + __init_doorbell(client); ret = __guc_allocate_doorbell(client->guc, client->stage_id); if (ret) { - __destroy_doorbell(client); + __fini_doorbell(client); __update_doorbell_desc(client, GUC_DOORBELL_INVALID); DRM_DEBUG_DRIVER("Couldn't create client %u doorbell: %d\n", client->stage_id, ret); @@ -246,7 +246,7 @@ static int destroy_doorbell(struct intel_guc_client *client) GEM_BUG_ON(!has_doorbell(client)); - __destroy_doorbell(client); + __fini_doorbell(client); ret = __guc_deallocate_doorbell(client->guc, client->stage_id); if (ret) DRM_ERROR("Couldn't destroy client %u doorbell: %d\n", @@ -1087,7 +1087,7 @@ static void __guc_client_disable(struct intel_guc_client *client) if (intel_guc_is_alive(client->guc)) destroy_doorbell(client); else - __destroy_doorbell(client); + __fini_doorbell(client); guc_stage_desc_fini(client); guc_proc_desc_fini(client); -- cgit v1.2.3 From bfeabcc87ae0f2dd75bc64f3fedbb6b762b4e5c6 Mon Sep 17 00:00:00 2001 From: Daniele Ceraolo Spurio Date: Mon, 22 Oct 2018 16:04:23 -0700 Subject: drm/i915/guc: reserve the doorbell before selecting the cacheline Cacheline selection is only needed if we actually manage to reserve a doorbell. Cc: Michal Wajdeczko Reviewed-by: Michal Wajdeczko Signed-off-by: Daniele Ceraolo Spurio Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20181022230427.5616-2-daniele.ceraolospurio@intel.com --- drivers/gpu/drm/i915/intel_guc_submission.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/i915/intel_guc_submission.c') diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index b089e5283307..8c3b5a9facee 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -955,6 +955,10 @@ guc_client_alloc(struct drm_i915_private *dev_priv, } client->vaddr = vaddr; + ret = reserve_doorbell(client); + if (ret) + goto err_vaddr; + client->doorbell_offset = __select_cacheline(guc); /* @@ -967,10 +971,6 @@ guc_client_alloc(struct drm_i915_private *dev_priv, else client->proc_desc_offset = (GUC_DB_SIZE / 2); - ret = reserve_doorbell(client); - if (ret) - goto err_vaddr; - DRM_DEBUG_DRIVER("new priority %u client %p for engine(s) 0x%x: stage_id %u\n", priority, client, client->engines, client->stage_id); DRM_DEBUG_DRIVER("doorbell id %u, cacheline offset 0x%lx\n", -- cgit v1.2.3 From fb0c37f67010cbafd284b2db351a332d9084c86c Mon Sep 17 00:00:00 2001 From: Daniele Ceraolo Spurio Date: Mon, 22 Oct 2018 16:04:24 -0700 Subject: drm/i915/guc: doorbell checking cleanup A collection of very small cleanups/improvements around doorbell checking that do not deserve their own patch: - Move doorbell-related HW defs to intel_guc_reg.h - use GUC_NUM_DOORBELLS instead of GUC_DOORBELL_INVALID where appropriate - do not stop on error in guc_verify_doorbells - do not print drbreg on error: the only content of the register apart from the valid bit is the lower part of the physical memory address, which we can't use even if valid because we don't know which descriptor it came from (since the doorbell is in an unexpected state) - Move the checking of doorbell valid bit to a common helper. v2: add more cleanups (move defs, use GUC_NUM_DOORBELLS, don't stop in guc_verify_doorbells) (Michal) v3: move more things to intel_guc_reg, redefine GUC_DOORBELL_INVALID (Michal), drop guc_doorbell_qw since it just duplicates guc_doorbell_info Cc: Michal Wajdeczko Signed-off-by: Daniele Ceraolo Spurio Reviewed-by: Michal Wajdeczko Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20181022230427.5616-3-daniele.ceraolospurio@intel.com --- drivers/gpu/drm/i915/intel_guc_fwif.h | 28 +++++----------------------- drivers/gpu/drm/i915/intel_guc_reg.h | 12 ++++++++++++ drivers/gpu/drm/i915/intel_guc_submission.c | 27 ++++++++++++++++----------- 3 files changed, 33 insertions(+), 34 deletions(-) (limited to 'drivers/gpu/drm/i915/intel_guc_submission.c') diff --git a/drivers/gpu/drm/i915/intel_guc_fwif.h b/drivers/gpu/drm/i915/intel_guc_fwif.h index ef79e0f144d7..b2f5148f4f17 100644 --- a/drivers/gpu/drm/i915/intel_guc_fwif.h +++ b/drivers/gpu/drm/i915/intel_guc_fwif.h @@ -39,6 +39,11 @@ #define GUC_VIDEO_ENGINE2 4 #define GUC_MAX_ENGINES_NUM (GUC_VIDEO_ENGINE2 + 1) +#define GUC_DOORBELL_INVALID 256 + +#define GUC_DB_SIZE (PAGE_SIZE) +#define GUC_WQ_SIZE (PAGE_SIZE * 2) + /* Work queue item header definitions */ #define WQ_STATUS_ACTIVE 1 #define WQ_STATUS_SUSPENDED 2 @@ -59,9 +64,6 @@ #define WQ_RING_TAIL_MAX 0x7FF /* 2^11 QWords */ #define WQ_RING_TAIL_MASK (WQ_RING_TAIL_MAX << WQ_RING_TAIL_SHIFT) -#define GUC_DOORBELL_ENABLED 1 -#define GUC_DOORBELL_DISABLED 0 - #define GUC_STAGE_DESC_ATTR_ACTIVE BIT(0) #define GUC_STAGE_DESC_ATTR_PENDING_DB BIT(1) #define GUC_STAGE_DESC_ATTR_KERNEL BIT(2) @@ -219,26 +221,6 @@ struct uc_css_header { u32 header_info; } __packed; -struct guc_doorbell_info { - u32 db_status; - u32 cookie; - u32 reserved[14]; -} __packed; - -union guc_doorbell_qw { - struct { - u32 db_status; - u32 cookie; - }; - u64 value_qw; -} __packed; - -#define GUC_NUM_DOORBELLS 256 -#define GUC_DOORBELL_INVALID (GUC_NUM_DOORBELLS) - -#define GUC_DB_SIZE (PAGE_SIZE) -#define GUC_WQ_SIZE (PAGE_SIZE * 2) - /* Work item for submitting workloads into work queue of GuC. */ struct guc_wq_item { u32 header; diff --git a/drivers/gpu/drm/i915/intel_guc_reg.h b/drivers/gpu/drm/i915/intel_guc_reg.h index d86084742a4a..57e7ad522c2f 100644 --- a/drivers/gpu/drm/i915/intel_guc_reg.h +++ b/drivers/gpu/drm/i915/intel_guc_reg.h @@ -104,6 +104,18 @@ #define GUC_SEND_INTERRUPT _MMIO(0xc4c8) #define GUC_SEND_TRIGGER (1<<0) +#define GUC_NUM_DOORBELLS 256 + +/* format of the HW-monitored doorbell cacheline */ +struct guc_doorbell_info { + u32 db_status; +#define GUC_DOORBELL_DISABLED 0 +#define GUC_DOORBELL_ENABLED 1 + + u32 cookie; + u32 reserved[14]; +} __packed; + #define GEN8_DRBREGL(x) _MMIO(0x1000 + (x) * 8) #define GEN8_DRB_VALID (1<<0) #define GEN8_DRBREGU(x) _MMIO(0x1000 + (x) * 8 + 4) diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index 8c3b5a9facee..1570dcbe249c 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -192,6 +192,14 @@ static struct guc_doorbell_info *__get_doorbell(struct intel_guc_client *client) return client->vaddr + client->doorbell_offset; } +static bool __doorbell_valid(struct intel_guc *guc, u16 db_id) +{ + struct drm_i915_private *dev_priv = guc_to_i915(guc); + + GEM_BUG_ON(db_id >= GUC_NUM_DOORBELLS); + return I915_READ(GEN8_DRBREGL(db_id)) & GEN8_DRB_VALID; +} + static void __init_doorbell(struct intel_guc_client *client) { struct guc_doorbell_info *doorbell; @@ -203,7 +211,6 @@ static void __init_doorbell(struct intel_guc_client *client) static void __fini_doorbell(struct intel_guc_client *client) { - struct drm_i915_private *dev_priv = guc_to_i915(client->guc); struct guc_doorbell_info *doorbell; u16 db_id = client->doorbell_id; @@ -214,7 +221,7 @@ static void __fini_doorbell(struct intel_guc_client *client) * to go to zero after updating db_status before we call the GuC to * release the doorbell */ - if (wait_for_us(!(I915_READ(GEN8_DRBREGL(db_id)) & GEN8_DRB_VALID), 10)) + if (wait_for_us(!__doorbell_valid(client->guc, db_id), 10)) WARN_ONCE(true, "Doorbell never became invalid after disable\n"); } @@ -866,33 +873,31 @@ guc_reset_prepare(struct intel_engine_cs *engine) /* Check that a doorbell register is in the expected state */ static bool doorbell_ok(struct intel_guc *guc, u16 db_id) { - struct drm_i915_private *dev_priv = guc_to_i915(guc); - u32 drbregl; bool valid; - GEM_BUG_ON(db_id >= GUC_DOORBELL_INVALID); + GEM_BUG_ON(db_id >= GUC_NUM_DOORBELLS); - drbregl = I915_READ(GEN8_DRBREGL(db_id)); - valid = drbregl & GEN8_DRB_VALID; + valid = __doorbell_valid(guc, db_id); if (test_bit(db_id, guc->doorbell_bitmap) == valid) return true; - DRM_DEBUG_DRIVER("Doorbell %d has unexpected state (0x%x): valid=%s\n", - db_id, drbregl, yesno(valid)); + DRM_DEBUG_DRIVER("Doorbell %u has unexpected state: valid=%s\n", + db_id, yesno(valid)); return false; } static bool guc_verify_doorbells(struct intel_guc *guc) { + bool doorbells_ok = true; u16 db_id; for (db_id = 0; db_id < GUC_NUM_DOORBELLS; ++db_id) if (!doorbell_ok(guc, db_id)) - return false; + doorbells_ok = false; - return true; + return doorbells_ok; } /** -- cgit v1.2.3