From patchwork Wed Dec 12 16:16:31 2012 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: Ville Syrjala X-Patchwork-Id: 1869351 Return-Path: X-Original-To: patchwork-intel-gfx@patchwork.kernel.org Delivered-To: patchwork-process-083081@patchwork2.kernel.org Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) by patchwork2.kernel.org (Postfix) with ESMTP id F3FD4DF266 for ; Wed, 12 Dec 2012 17:56:19 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 03E1CE65AA for ; Wed, 12 Dec 2012 09:56:20 -0800 (PST) X-Original-To: intel-gfx@lists.freedesktop.org Delivered-To: intel-gfx@lists.freedesktop.org Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) by gabe.freedesktop.org (Postfix) with ESMTP id EBCC6E5C97; Wed, 12 Dec 2012 08:20:44 -0800 (PST) Received: from orsmga002.jf.intel.com ([10.7.209.21]) by orsmga102.jf.intel.com with ESMTP; 12 Dec 2012 08:19:51 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.84,267,1355126400"; d="scan'208";a="256478624" Received: from stinkbox.fi.intel.com (HELO stinkbox) ([10.237.72.168]) by orsmga002.jf.intel.com with SMTP; 12 Dec 2012 08:20:38 -0800 Received: by stinkbox (sSMTP sendmail emulation); Wed, 12 Dec 2012 18:20:38 +0200 From: ville.syrjala@linux.intel.com To: dri-devel@lists.freedesktop.org Date: Wed, 12 Dec 2012 18:16:31 +0200 Message-Id: <1355329008-31459-65-git-send-email-ville.syrjala@linux.intel.com> X-Mailer: git-send-email 1.7.8.6 In-Reply-To: <1355329008-31459-1-git-send-email-ville.syrjala@linux.intel.com> References: <1355329008-31459-1-git-send-email-ville.syrjala@linux.intel.com> MIME-Version: 1.0 Cc: intel-gfx@lists.freedesktop.org Subject: [Intel-gfx] [PATCH 64/81] drm/i915: Implement a non-blocking GPU synchronization mechanism for atomic page flips X-BeenThere: intel-gfx@lists.freedesktop.org X-Mailman-Version: 2.1.13 Precedence: list List-Id: Intel graphics driver community testing & development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: intel-gfx-bounces+patchwork-intel-gfx=patchwork.kernel.org@lists.freedesktop.org Errors-To: intel-gfx-bounces+patchwork-intel-gfx=patchwork.kernel.org@lists.freedesktop.org From: Ville Syrjälä After the atomic flip has been split up into individual flip requests for each scanout engine, put each such request into a FIFO. Then for each flip request add new request to the ring(s) in order to get an interrupt once the GPU has finished whatever it was doing with the new front buffer. Once the flip requests associated with the same atomic flip and the same pipe at the head of the queue are ready, pull them from the queue and issue the flips to the hardware. Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/i915_drv.h | 8 ++ drivers/gpu/drm/i915/i915_irq.c | 2 + drivers/gpu/drm/i915/intel_atomic.c | 215 ++++++++++++++++++++++++++++++++++- drivers/gpu/drm/i915/intel_drv.h | 1 + 4 files changed, 222 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 6bb8b3f..80b4b30 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -960,6 +960,14 @@ typedef struct drm_i915_private { struct i915_dri1_state dri1; struct drm_flip_driver flip_driver; + + struct { + struct list_head list; + spinlock_t lock; + struct work_struct work; + struct workqueue_struct *wq; + unsigned int next_flip_seq; + } flip; } drm_i915_private_t; /* Iterate over initialised rings */ diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 430f201..2f0dce2 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -349,6 +349,8 @@ static void notify_ring(struct drm_device *dev, { struct drm_i915_private *dev_priv = dev->dev_private; + intel_atomic_notify_ring(dev, ring); + if (ring->obj == NULL) return; diff --git a/drivers/gpu/drm/i915/intel_atomic.c b/drivers/gpu/drm/i915/intel_atomic.c index 33357ec..2ebceec 100644 --- a/drivers/gpu/drm/i915/intel_atomic.c +++ b/drivers/gpu/drm/i915/intel_atomic.c @@ -30,6 +30,8 @@ #include "intel_drv.h" +#define USE_WRITE_SEQNO + struct intel_flip { struct drm_flip base; u32 vbl_count; @@ -42,6 +44,9 @@ struct intel_flip { struct drm_pending_atomic_event *event; uint32_t old_fb_id; struct list_head pending_head; + struct intel_ring_buffer *ring; + u32 seqno; + unsigned int flip_seq; }; struct intel_plane_state { @@ -801,6 +806,7 @@ static int pin_fbs(struct drm_device *dev, struct intel_atomic_state *s) { int i, ret; + bool nonblock = s->flags & DRM_MODE_ATOMIC_NONBLOCK; for (i = 0; i < dev->mode_config.num_crtc; i++) { struct intel_crtc_state *st = &s->crtc[i]; @@ -816,7 +822,7 @@ static int pin_fbs(struct drm_device *dev, obj = to_intel_framebuffer(crtc->fb)->obj; mutex_lock(&dev->struct_mutex); - ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); + ret = intel_pin_and_fence_fb_obj(dev, obj, nonblock ? obj->ring : NULL); mutex_unlock(&dev->struct_mutex); if (ret) @@ -839,7 +845,7 @@ static int pin_fbs(struct drm_device *dev, obj = to_intel_framebuffer(plane->fb)->obj; mutex_lock(&dev->struct_mutex); - ret = intel_pin_and_fence_fb_obj(dev, obj, NULL); + ret = intel_pin_and_fence_fb_obj(dev, obj, nonblock ? obj->ring : NULL); mutex_unlock(&dev->struct_mutex); if (ret) @@ -2040,6 +2046,8 @@ static const struct drm_flip_driver_funcs intel_flip_driver_funcs = { .flush = intel_flip_driver_flush, }; +static void intel_atomic_process_flips_work(struct work_struct *work); + static void intel_flip_init(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -2058,6 +2066,11 @@ static void intel_flip_init(struct drm_device *dev) list_for_each_entry(intel_plane, &dev->mode_config.plane_list, base.head) drm_flip_helper_init(&intel_plane->flip_helper, &dev_priv->flip_driver, &intel_flip_funcs); + + INIT_LIST_HEAD(&dev_priv->flip.list); + spin_lock_init(&dev_priv->flip.lock); + INIT_WORK(&dev_priv->flip.work, intel_atomic_process_flips_work); + dev_priv->flip.wq = create_singlethread_workqueue("intel_flip"); } static void intel_flip_fini(struct drm_device *dev) @@ -2075,6 +2088,30 @@ static void intel_flip_fini(struct drm_device *dev) drm_flip_driver_fini(&dev_priv->flip_driver); } +static bool intel_atomic_postpone_flip(struct intel_flip *intel_flip) +{ + struct intel_ring_buffer *ring = intel_flip->ring; + int ret; + + ret = i915_gem_check_olr(ring, intel_flip->seqno); + if (WARN_ON(ret)) { + intel_flip->ring = NULL; + return false; + } + + if (i915_seqno_passed(ring->get_seqno(ring, true), intel_flip->seqno)) { + intel_flip->ring = NULL; + return false; + } + + if (WARN_ON(!ring->irq_get(ring))) { + intel_flip->ring = NULL; + return false; + } + + return true; +} + static void intel_atomic_schedule_flips(struct drm_i915_private *dev_priv, struct intel_crtc *intel_crtc, struct list_head *flips) @@ -2095,15 +2132,120 @@ static void intel_atomic_schedule_flips(struct drm_i915_private *dev_priv, local_irq_enable(); } +static bool intel_atomic_flips_ready(struct drm_device *dev, unsigned int flip_seq) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_flip *intel_flip; + + /* check if all flips w/ same flip_seq are ready */ + list_for_each_entry(intel_flip, &dev_priv->flip.list, base.list) { + if (intel_flip->flip_seq != flip_seq) + break; + + if (intel_flip->ring) + return false; + } + + return true; +} + +static void intel_atomic_process_flips_work(struct work_struct *work) +{ + struct drm_i915_private *dev_priv = container_of(work, struct drm_i915_private, flip.work); + struct drm_device *dev = dev_priv->dev; + + for (;;) { + struct intel_flip *intel_flip, *next; + unsigned int flip_seq; + struct intel_crtc *intel_crtc; + LIST_HEAD(flips); + unsigned long flags; + + if (list_empty(&dev_priv->flip.list)) + return; + + spin_lock_irqsave(&dev_priv->flip.lock, flags); + + intel_flip = list_first_entry(&dev_priv->flip.list, struct intel_flip, base.list); + flip_seq = intel_flip->flip_seq; + intel_crtc = to_intel_crtc(intel_flip->crtc); + + if (intel_atomic_flips_ready(dev, flip_seq)) { + list_for_each_entry_safe(intel_flip, next, &dev_priv->flip.list, base.list) { + if (intel_flip->flip_seq != flip_seq) + break; + list_move_tail(&intel_flip->base.list, &flips); + } + } + + spin_unlock_irqrestore(&dev_priv->flip.lock, flags); + + if (list_empty(&flips)) + return; + + mutex_lock(&dev->mode_config.mutex); + intel_atomic_schedule_flips(dev_priv, intel_crtc, &flips); + mutex_unlock(&dev->mode_config.mutex); + } +} + +static void intel_atomic_check_flips_ready(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_flip *intel_flip; + + if (list_empty(&dev_priv->flip.list)) + return; + + intel_flip = list_first_entry(&dev_priv->flip.list, struct intel_flip, base.list); + if (intel_atomic_flips_ready(dev, intel_flip->flip_seq)) + queue_work(dev_priv->flip.wq, &dev_priv->flip.work); +} + +void intel_atomic_notify_ring(struct drm_device *dev, + struct intel_ring_buffer *ring) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_flip *intel_flip; + unsigned long flags; + u32 seqno; + + if (list_empty(&dev_priv->flip.list)) + return; + + seqno = ring->get_seqno(ring, false); + + spin_lock_irqsave(&dev_priv->flip.lock, flags); + + list_for_each_entry(intel_flip, &dev_priv->flip.list, base.list) { + if (ring != intel_flip->ring) + continue; + + if (i915_seqno_passed(seqno, intel_flip->seqno)) { + intel_flip->ring = NULL; + ring->irq_put(ring); + } + } + + intel_atomic_check_flips_ready(dev); + + spin_unlock_irqrestore(&dev_priv->flip.lock, flags); +} + static void atomic_pipe_commit(struct drm_device *dev, struct intel_atomic_state *state, int pipe) { struct drm_i915_private *dev_priv = dev->dev_private; struct drm_i915_file_private *file_priv = state->file->driver_priv; - struct intel_crtc *intel_crtc = to_intel_crtc(intel_get_crtc_for_pipe(dev, pipe)); LIST_HEAD(flips); int i; + /* FIXME treat flips for all pipes as one set for GPU sync */ + unsigned int flip_seq = dev_priv->flip.next_flip_seq++; + struct intel_flip *intel_flip, *next; + unsigned long flags; + struct intel_ring_buffer *ring; + unsigned int rings_mask = 0; for (i = 0; i < dev->mode_config.num_crtc; i++) { struct intel_crtc_state *st = &state->crtc[i]; @@ -2122,6 +2264,8 @@ static void atomic_pipe_commit(struct drm_device *dev, drm_flip_init(&intel_flip->base, &intel_crtc->flip_helper); + intel_flip->flip_seq = flip_seq; + if (st->event) { intel_flip->event = st->event; st->event = NULL; @@ -2134,6 +2278,19 @@ static void atomic_pipe_commit(struct drm_device *dev, intel_flip->crtc = crtc; + if (crtc->fb) { + struct drm_i915_gem_object *obj = to_intel_framebuffer(crtc->fb)->obj; + + mutex_lock(&dev->struct_mutex); +#ifdef USE_WRITE_SEQNO + intel_flip->seqno = obj->last_write_seqno; +#else + intel_flip->seqno = obj->last_read_seqno; +#endif + intel_flip->ring = obj->ring; + mutex_unlock(&dev->struct_mutex); + } + /* update primary_disabled befoer calc_plane() */ intel_crtc->primary_disabled = st->primary_disabled; @@ -2177,6 +2334,8 @@ static void atomic_pipe_commit(struct drm_device *dev, drm_flip_init(&intel_flip->base, &intel_plane->flip_helper); + intel_flip->flip_seq = flip_seq; + if (st->event) { intel_flip->event = st->event; st->event = NULL; @@ -2190,6 +2349,19 @@ static void atomic_pipe_commit(struct drm_device *dev, intel_flip->crtc = intel_get_crtc_for_pipe(dev, pipe); intel_flip->plane = plane; + if (plane->fb) { + struct drm_i915_gem_object *obj = to_intel_framebuffer(plane->fb)->obj; + + mutex_lock(&dev->struct_mutex); +#ifdef USE_WRITE_SEQNO + intel_flip->seqno = obj->last_write_seqno; +#else + intel_flip->seqno = obj->last_read_seqno; +#endif + intel_flip->ring = obj->ring; + mutex_unlock(&dev->struct_mutex); + } + intel_plane->calc(plane, plane->fb, &st->coords); if (st->old.fb) @@ -2209,7 +2381,42 @@ static void atomic_pipe_commit(struct drm_device *dev, if (list_empty(&flips)) return; - intel_atomic_schedule_flips(dev_priv, intel_crtc, &flips); + mutex_lock(&dev->struct_mutex); + + list_for_each_entry(intel_flip, &flips, base.list) { + struct intel_ring_buffer *ring = intel_flip->ring; + + if (!ring) + continue; + + if (intel_atomic_postpone_flip(intel_flip)) + rings_mask |= intel_ring_flag(ring); + } + + spin_lock_irqsave(&dev_priv->flip.lock, flags); + + list_for_each_entry_safe(intel_flip, next, &flips, base.list) + list_move_tail(&intel_flip->base.list, &dev_priv->flip.list); + + /* if no rings are involved, we can avoid checking seqnos */ + if (rings_mask == 0) + intel_atomic_check_flips_ready(dev); + + spin_unlock_irqrestore(&dev_priv->flip.lock, flags); + + mutex_unlock(&dev->struct_mutex); + + if (rings_mask == 0) + return; + + /* + * Double check to catch cases where the irq + * fired before the flip was placed onto flip.list. + */ + for_each_ring(ring, dev_priv, i) { + if (rings_mask & intel_ring_flag(ring)) + intel_atomic_notify_ring(dev, ring); + } } void intel_atomic_handle_vblank(struct drm_device *dev, int pipe) diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 24dc192..446fa1c 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -730,6 +730,7 @@ extern void intel_atomic_fini(struct drm_device *dev); extern void intel_atomic_free_events(struct drm_device *dev, struct drm_file *file); extern void intel_atomic_handle_vblank(struct drm_device *dev, int pipe); extern void intel_atomic_clear_flips(struct drm_crtc *crtc); +extern void intel_atomic_notify_ring(struct drm_device *dev, struct intel_ring_buffer *ring); extern void intel_enable_primary(struct drm_crtc *crtc); extern void intel_disable_primary(struct drm_crtc *crtc);