From patchwork Thu Jul 25 22:38:43 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Chris Wilson X-Patchwork-Id: 11059913 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 4280A6C5 for ; Thu, 25 Jul 2019 22:38:51 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 2908828A6D for ; Thu, 25 Jul 2019 22:38:51 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 1D20E28A6F; Thu, 25 Jul 2019 22:38:51 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-5.2 required=2.0 tests=BAYES_00,MAILING_LIST_MULTI, RCVD_IN_DNSWL_MED autolearn=ham version=3.3.1 Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher DHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.wl.linuxfoundation.org (Postfix) with ESMTPS id 2916228A6E for ; Thu, 25 Jul 2019 22:38:50 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 729416E839; Thu, 25 Jul 2019 22:38:49 +0000 (UTC) X-Original-To: intel-gfx@lists.freedesktop.org Delivered-To: intel-gfx@lists.freedesktop.org Received: from fireflyinternet.com (mail.fireflyinternet.com [109.228.58.192]) by gabe.freedesktop.org (Postfix) with ESMTPS id 36C8C6E838 for ; Thu, 25 Jul 2019 22:38:47 +0000 (UTC) X-Default-Received-SPF: pass (skip=forwardok (res=PASS)) x-ip-name=78.156.65.138; Received: from haswell.alporthouse.com (unverified [78.156.65.138]) by fireflyinternet.com (Firefly Internet (M1)) with ESMTP id 17610856-1500050 for ; Thu, 25 Jul 2019 23:38:44 +0100 From: Chris Wilson To: intel-gfx@lists.freedesktop.org Date: Thu, 25 Jul 2019 23:38:43 +0100 Message-Id: <20190725223843.8971-1-chris@chris-wilson.co.uk> X-Mailer: git-send-email 2.22.0 MIME-Version: 1.0 Subject: [Intel-gfx] [CI] drm/i915: Capture vma contents outside of spinlock X-BeenThere: intel-gfx@lists.freedesktop.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: Intel graphics driver community testing & development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: intel-gfx-bounces@lists.freedesktop.org Sender: "Intel-gfx" X-Virus-Scanned: ClamAV using ClamSMTP Currently we use the engine->active.lock to ensure that the request is not retired as we capture the data. However, we only need to ensure that the vma are not removed prior to use acquiring their contents, and since we have already relinquished our stop-machine protection, we assume that the user will not be overwriting the contents before we are able to record them. In order to capture the vma outside of the spinlock, we acquire a reference and mark the vma as active to prevent it from being unbound. However, since it is tricky allocate an entry in the fence tree (doing so would require taking a mutex) while inside the engine spinlock, we use an atomic bit and special case the handling for i915_active_wait. The core benefit is that we can use some non-atomic methods for mapping the device pages, we can remove the slow compression phase out of atomic context (i.e. stop antagonising the nmi-watchdog), and no we longer need large reserves of atomic pages. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111215 Signed-off-by: Chris Wilson Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/i915_active.c | 34 ++++++- drivers/gpu/drm/i915/i915_active.h | 3 + drivers/gpu/drm/i915/i915_active_types.h | 3 + drivers/gpu/drm/i915/i915_gpu_error.c | 112 ++++++++++++++++------- 4 files changed, 117 insertions(+), 35 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c index 13f304a29fc8..22341c62c204 100644 --- a/drivers/gpu/drm/i915/i915_active.c +++ b/drivers/gpu/drm/i915/i915_active.c @@ -196,6 +196,7 @@ void __i915_active_init(struct drm_i915_private *i915, debug_active_init(ref); ref->i915 = i915; + ref->flags = 0; ref->active = active; ref->retire = retire; ref->tree = RB_ROOT; @@ -262,6 +263,34 @@ void i915_active_release(struct i915_active *ref) active_retire(ref); } +static void __active_ungrab(struct i915_active *ref) +{ + clear_and_wake_up_bit(I915_ACTIVE_GRAB_BIT, &ref->flags); +} + +bool i915_active_trygrab(struct i915_active *ref) +{ + debug_active_assert(ref); + + if (test_and_set_bit(I915_ACTIVE_GRAB_BIT, &ref->flags)) + return false; + + if (!atomic_add_unless(&ref->count, 1, 0)) { + __active_ungrab(ref); + return false; + } + + return true; +} + +void i915_active_ungrab(struct i915_active *ref) +{ + GEM_BUG_ON(!test_bit(I915_ACTIVE_GRAB_BIT, &ref->flags)); + + active_retire(ref); + __active_ungrab(ref); +} + int i915_active_wait(struct i915_active *ref) { struct active_node *it, *n; @@ -270,7 +299,7 @@ int i915_active_wait(struct i915_active *ref) might_sleep(); might_lock(&ref->mutex); - if (RB_EMPTY_ROOT(&ref->tree)) + if (i915_active_is_idle(ref)) return 0; err = mutex_lock_interruptible(&ref->mutex); @@ -292,6 +321,9 @@ int i915_active_wait(struct i915_active *ref) if (err) return err; + if (wait_on_bit(&ref->flags, I915_ACTIVE_GRAB_BIT, TASK_KILLABLE)) + return -EINTR; + if (!i915_active_is_idle(ref)) return -EBUSY; diff --git a/drivers/gpu/drm/i915/i915_active.h b/drivers/gpu/drm/i915/i915_active.h index 134166d31251..ba68b077ec6c 100644 --- a/drivers/gpu/drm/i915/i915_active.h +++ b/drivers/gpu/drm/i915/i915_active.h @@ -395,6 +395,9 @@ int i915_active_acquire(struct i915_active *ref); void i915_active_release(struct i915_active *ref); void __i915_active_release_nested(struct i915_active *ref, int subclass); +bool i915_active_trygrab(struct i915_active *ref); +void i915_active_ungrab(struct i915_active *ref); + static inline bool i915_active_is_idle(const struct i915_active *ref) { diff --git a/drivers/gpu/drm/i915/i915_active_types.h b/drivers/gpu/drm/i915/i915_active_types.h index 5b0a3024ce24..74743dd0d5f0 100644 --- a/drivers/gpu/drm/i915/i915_active_types.h +++ b/drivers/gpu/drm/i915/i915_active_types.h @@ -36,6 +36,9 @@ struct i915_active { struct mutex mutex; atomic_t count; + unsigned long flags; +#define I915_ACTIVE_GRAB_BIT 0 + int (*active)(struct i915_active *ref); void (*retire)(struct i915_active *ref); diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 56dfc2650836..674d341a23f6 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -298,7 +298,7 @@ static void *compress_next_page(struct compress *c, if (dst->page_count >= dst->num_pages) return ERR_PTR(-ENOSPC); - page = pool_alloc(&c->pool, ATOMIC_MAYFAIL); + page = pool_alloc(&c->pool, ALLOW_FAIL); if (!page) return ERR_PTR(-ENOMEM); @@ -327,8 +327,6 @@ static int compress_page(struct compress *c, if (zlib_deflate(zstream, Z_NO_FLUSH) != Z_OK) return -EIO; - - touch_nmi_watchdog(); } while (zstream->avail_in); /* Fallback to uncompressed if we increase size? */ @@ -407,7 +405,7 @@ static int compress_page(struct compress *c, { void *ptr; - ptr = pool_alloc(&c->pool, ATOMIC_MAYFAIL); + ptr = pool_alloc(&c->pool, ALLOW_FAIL); if (!ptr) return -ENOMEM; @@ -1001,12 +999,14 @@ i915_error_object_create(struct drm_i915_private *i915, dma_addr_t dma; int ret; + might_sleep(); + if (!vma || !vma->pages) return NULL; num_pages = min_t(u64, vma->size, vma->obj->base.size) >> PAGE_SHIFT; num_pages = DIV_ROUND_UP(10 * num_pages, 8); /* worstcase zlib growth */ - dst = kmalloc(sizeof(*dst) + num_pages * sizeof(u32 *), ATOMIC_MAYFAIL); + dst = kmalloc(sizeof(*dst) + num_pages * sizeof(u32 *), ALLOW_FAIL); if (!dst) return NULL; @@ -1027,9 +1027,9 @@ i915_error_object_create(struct drm_i915_private *i915, ggtt->vm.insert_page(&ggtt->vm, dma, slot, I915_CACHE_NONE, 0); - s = io_mapping_map_atomic_wc(&ggtt->iomap, slot); + s = io_mapping_map_wc(&ggtt->iomap, slot, PAGE_SIZE); ret = compress_page(compress, (void __force *)s, dst); - io_mapping_unmap_atomic(s); + io_mapping_unmap(s); if (ret) break; } @@ -1302,10 +1302,42 @@ static void record_context(struct drm_i915_error_context *e, e->active = atomic_read(&ctx->active_count); } -static void +struct capture_vma { + struct capture_vma *next; + void **slot; +}; + +static struct capture_vma * +capture_vma(struct capture_vma *next, + struct i915_vma *vma, + struct drm_i915_error_object **out) +{ + struct capture_vma *c; + + *out = NULL; + if (!vma) + return next; + + c = kmalloc(sizeof(*c), ATOMIC_MAYFAIL); + if (!c) + return next; + + if (!i915_active_trygrab(&vma->active)) { + kfree(c); + return next; + } + + c->slot = (void **)out; + *c->slot = i915_vma_get(vma); + + c->next = next; + return c; +} + +static struct capture_vma * request_record_user_bo(struct i915_request *request, struct drm_i915_error_engine *ee, - struct compress *compress) + struct capture_vma *capture) { struct i915_capture_list *c; struct drm_i915_error_object **bo; @@ -1315,7 +1347,7 @@ request_record_user_bo(struct i915_request *request, for (c = request->capture_list; c; c = c->next) max++; if (!max) - return; + return capture; bo = kmalloc_array(max, sizeof(*bo), ATOMIC_MAYFAIL); if (!bo) { @@ -1324,21 +1356,19 @@ request_record_user_bo(struct i915_request *request, bo = kmalloc_array(max, sizeof(*bo), ATOMIC_MAYFAIL); } if (!bo) - return; + return capture; count = 0; for (c = request->capture_list; c; c = c->next) { - bo[count] = i915_error_object_create(request->i915, - c->vma, - compress); - if (!bo[count]) - break; + capture = capture_vma(capture, c->vma, &bo[count]); if (++count == max) break; } ee->user_bo = bo; ee->user_bo_count = count; + + return capture; } static struct drm_i915_error_object * @@ -1369,6 +1399,7 @@ gem_record_rings(struct i915_gpu_state *error, struct compress *compress) for (i = 0; i < I915_NUM_ENGINES; i++) { struct intel_engine_cs *engine = i915->engine[i]; struct drm_i915_error_engine *ee = &error->engine[i]; + struct capture_vma *capture = NULL; struct i915_request *request; unsigned long flags; @@ -1393,26 +1424,29 @@ gem_record_rings(struct i915_gpu_state *error, struct compress *compress) record_context(&ee->context, ctx); - /* We need to copy these to an anonymous buffer + /* + * We need to copy these to an anonymous buffer * as the simplest method to avoid being overwritten * by userspace. */ - ee->batchbuffer = - i915_error_object_create(i915, - request->batch, - compress); + capture = capture_vma(capture, + request->batch, + &ee->batchbuffer); if (HAS_BROKEN_CS_TLB(i915)) - ee->wa_batchbuffer = - i915_error_object_create(i915, - engine->gt->scratch, - compress); - request_record_user_bo(request, ee, compress); + capture = capture_vma(capture, + engine->gt->scratch, + &ee->wa_batchbuffer); + + capture = request_record_user_bo(request, ee, capture); - ee->ctx = - i915_error_object_create(i915, - request->hw_context->state, - compress); + capture = capture_vma(capture, + request->hw_context->state, + &ee->ctx); + + capture = capture_vma(capture, + ring->vma, + &ee->ringbuffer); error->simulated |= i915_gem_context_no_error_capture(ctx); @@ -1423,15 +1457,25 @@ gem_record_rings(struct i915_gpu_state *error, struct compress *compress) ee->cpu_ring_head = ring->head; ee->cpu_ring_tail = ring->tail; - ee->ringbuffer = - i915_error_object_create(i915, - ring->vma, - compress); engine_record_requests(engine, request, ee); } spin_unlock_irqrestore(&engine->active.lock, flags); + while (capture) { + struct capture_vma *this = capture; + struct i915_vma *vma = *this->slot; + + *this->slot = + i915_error_object_create(i915, vma, compress); + + i915_active_ungrab(&vma->active); + i915_vma_put(vma); + + capture = this->next; + kfree(this); + } + ee->hws_page = i915_error_object_create(i915, engine->status_page.vma,