drm/i915: Allow nonblocking update of pageflips.
authorMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Tue, 17 May 2016 13:08:04 +0000 (15:08 +0200)
committerMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Thu, 19 May 2016 12:39:40 +0000 (14:39 +0200)
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463490484-19540-22-git-send-email-maarten.lankhorst@linux.intel.com
Reviewed-by: Patrik Jakobsson <patrik.jakobsson@linux.intel.com>
drivers/gpu/drm/i915/intel_display.c

index 4001838..60ffbfd 100644 (file)
@@ -108,8 +108,6 @@ static void vlv_prepare_pll(struct intel_crtc *crtc,
                            const struct intel_crtc_state *pipe_config);
 static void chv_prepare_pll(struct intel_crtc *crtc,
                            const struct intel_crtc_state *pipe_config);
-static void intel_begin_crtc_commit(struct drm_crtc *, struct drm_crtc_state *);
-static void intel_finish_crtc_commit(struct drm_crtc *, struct drm_crtc_state *);
 static void skl_init_scalers(struct drm_device *dev, struct intel_crtc *intel_crtc,
        struct intel_crtc_state *crtc_state);
 static void skylake_pfit_enable(struct intel_crtc *crtc);
@@ -10981,7 +10979,7 @@ static void intel_mmio_flip_work_func(struct work_struct *w)
        struct drm_device *dev = crtc->dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_i915_gem_request *req;
-       int i;
+       int i, ret;
 
        if (!needs_modeset(&crtc_state->base) && crtc_state->update_pipe) {
                work->put_power_domains =
@@ -11003,7 +11001,14 @@ static void intel_mmio_flip_work_func(struct work_struct *w)
                                            &dev_priv->rps.mmioflips));
        }
 
-       intel_frontbuffer_flip_prepare(dev, crtc_state->fb_bits);
+       ret = drm_crtc_vblank_get(crtc);
+       I915_STATE_WARN(ret < 0, "enabling vblank failed with %i\n", ret);
+
+       if (work->num_planes &&
+           work->old_plane_state[0]->base.plane == crtc->primary)
+               intel_fbc_enable(intel_crtc, work->new_crtc_state, work->new_plane_state[0]);
+
+       intel_frontbuffer_flip_prepare(dev, work->fb_bits);
 
        intel_pipe_update_start(intel_crtc);
        if (!needs_modeset(&crtc_state->base)) {
@@ -11022,206 +11027,15 @@ static void intel_mmio_flip_work_func(struct work_struct *w)
                struct intel_plane_state *new_plane_state = work->new_plane_state[i];
                struct intel_plane *plane = to_intel_plane(new_plane_state->base.plane);
 
-               plane->update_plane(&plane->base, crtc_state, new_plane_state);
+               if (new_plane_state->visible)
+                       plane->update_plane(&plane->base, crtc_state, new_plane_state);
+               else
+                       plane->disable_plane(&plane->base, crtc);
        }
 
        intel_pipe_update_end(intel_crtc, work);
 }
 
-static struct fence *intel_get_excl_fence(struct drm_i915_gem_object *obj)
-{
-       struct reservation_object *resv;
-
-
-       if (!obj->base.dma_buf)
-               return NULL;
-
-       resv = obj->base.dma_buf->resv;
-
-       /* For framebuffer backed by dmabuf, wait for fence */
-       while (1) {
-               struct fence *fence_excl, *ret = NULL;
-
-               rcu_read_lock();
-
-               fence_excl = rcu_dereference(resv->fence_excl);
-               if (fence_excl)
-                       ret = fence_get_rcu(fence_excl);
-
-               rcu_read_unlock();
-
-               if (ret == fence_excl)
-                       return ret;
-       }
-}
-
-static int intel_crtc_page_flip(struct drm_crtc *crtc,
-                               struct drm_framebuffer *fb,
-                               struct drm_pending_vblank_event *event,
-                               uint32_t page_flip_flags)
-{
-       struct drm_device *dev = crtc->dev;
-       struct drm_i915_private *dev_priv = dev->dev_private;
-       struct drm_plane_state *old_state, *new_state = NULL;
-       struct drm_crtc_state *new_crtc_state = NULL;
-       struct drm_framebuffer *old_fb = crtc->primary->state->fb;
-       struct drm_i915_gem_object *obj = intel_fb_obj(fb);
-       struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
-       struct drm_plane *primary = crtc->primary;
-       struct intel_flip_work *work;
-       int ret;
-
-       old_state = crtc->primary->state;
-
-       if (!crtc->state->active)
-               return -EINVAL;
-
-       /*
-        * drm_mode_page_flip_ioctl() should already catch this, but double
-        * check to be safe.  In the future we may enable pageflipping from
-        * a disabled primary plane.
-        */
-       if (WARN_ON(intel_fb_obj(old_fb) == NULL))
-               return -EBUSY;
-
-       /* Can't change pixel format via MI display flips. */
-       if (fb->pixel_format != old_fb->pixel_format)
-               return -EINVAL;
-
-       /*
-        * TILEOFF/LINOFF registers can't be changed via MI display flips.
-        * Note that pitch changes could also affect these register.
-        */
-       if (INTEL_INFO(dev)->gen > 3 &&
-           (fb->offsets[0] != old_fb->offsets[0] ||
-            fb->pitches[0] != old_fb->pitches[0]))
-               return -EINVAL;
-
-       work = kzalloc(sizeof(*work), GFP_KERNEL);
-       new_crtc_state = intel_crtc_duplicate_state(crtc);
-       new_state = intel_plane_duplicate_state(primary);
-
-       if (!work || !new_crtc_state || !new_state) {
-               ret = -ENOMEM;
-               goto cleanup;
-       }
-
-       drm_framebuffer_unreference(new_state->fb);
-       drm_framebuffer_reference(fb);
-       new_state->fb = fb;
-
-       work->event = event;
-       INIT_WORK(&work->unpin_work, intel_unpin_work_fn);
-       INIT_WORK(&work->mmio_work, intel_mmio_flip_work_func);
-
-       work->new_crtc_state = to_intel_crtc_state(new_crtc_state);
-       work->old_crtc_state = intel_crtc->config;
-
-       work->fb_bits = to_intel_plane(primary)->frontbuffer_bit;
-       work->new_crtc_state->fb_bits = work->fb_bits;
-
-       work->can_async_unpin = true;
-       work->num_planes = 1;
-       work->old_plane_state[0] = to_intel_plane_state(old_state);
-       work->new_plane_state[0] = to_intel_plane_state(new_state);
-
-       /* Step 1: vblank waiting and workqueue throttling,
-        * similar to intel_atomic_prepare_commit
-        */
-       ret = drm_crtc_vblank_get(crtc);
-       if (ret)
-               goto cleanup;
-
-       /* We borrow the event spin lock for protecting flip_work */
-       spin_lock_irq(&dev->event_lock);
-       if (!list_empty(&intel_crtc->flip_work)) {
-               struct intel_flip_work *old_work;
-
-               old_work = list_last_entry(&intel_crtc->flip_work,
-                                          struct intel_flip_work, head);
-
-               /* Before declaring the flip queue wedged, check if
-                * the hardware completed the operation behind our backs.
-                */
-               if (pageflip_finished(intel_crtc, old_work)) {
-                       DRM_DEBUG_DRIVER("flip queue: previous flip completed, continuing\n");
-                       page_flip_completed(intel_crtc, old_work);
-               } else {
-                       DRM_DEBUG_DRIVER("flip queue: crtc already busy\n");
-                       spin_unlock_irq(&dev->event_lock);
-
-                       ret = -EBUSY;
-                       goto cleanup_vblank;
-               }
-       }
-       list_add_tail(&work->head, &intel_crtc->flip_work);
-       spin_unlock_irq(&dev->event_lock);
-
-       if (atomic_read(&intel_crtc->unpin_work_count) >= 2)
-               flush_workqueue(dev_priv->wq);
-
-       /* step 2, similar to intel_prepare_plane_fb */
-       ret = mutex_lock_interruptible(&dev->struct_mutex);
-       if (ret)
-               goto cleanup_work;
-
-       ret = intel_pin_and_fence_fb_obj(fb, new_state->rotation);
-       if (ret)
-               goto cleanup_unlock;
-
-       i915_gem_track_fb(intel_fb_obj(old_fb), obj,
-                         to_intel_plane(primary)->frontbuffer_bit);
-
-       /* point of no return, swap state */
-       primary->state = new_state;
-       crtc->state = new_crtc_state;
-       intel_crtc->config = to_intel_crtc_state(new_crtc_state);
-       primary->fb = fb;
-
-       /* scheduling flip work */
-       atomic_inc(&intel_crtc->unpin_work_count);
-
-       if (obj->last_write_req &&
-           !i915_gem_request_completed(obj->last_write_req, true))
-               i915_gem_request_assign(&work->old_plane_state[0]->wait_req,
-                                       obj->last_write_req);
-
-       if (obj->base.dma_buf)
-               work->old_plane_state[0]->base.fence = intel_get_excl_fence(obj);
-
-       intel_fbc_pre_update(intel_crtc,
-                            to_intel_crtc_state(new_crtc_state),
-                            to_intel_plane_state(new_state));
-
-       schedule_work(&work->mmio_work);
-
-       mutex_unlock(&dev->struct_mutex);
-
-       trace_i915_flip_request(intel_crtc->plane, obj);
-
-       return 0;
-
-cleanup_unlock:
-       mutex_unlock(&dev->struct_mutex);
-cleanup_work:
-       spin_lock_irq(&dev->event_lock);
-       list_del(&work->head);
-       spin_unlock_irq(&dev->event_lock);
-
-cleanup_vblank:
-       drm_crtc_vblank_put(crtc);
-cleanup:
-       if (new_state)
-               intel_plane_destroy_state(primary, new_state);
-
-       if (new_crtc_state)
-               intel_crtc_destroy_state(crtc, new_crtc_state);
-
-       intel_free_flip_work(work);
-       return ret;
-}
-
-
 /**
  * intel_wm_need_update - Check whether watermarks need updating
  * @plane: drm plane
@@ -11498,8 +11312,6 @@ static int intel_crtc_atomic_check(struct drm_crtc *crtc,
 
 static const struct drm_crtc_helper_funcs intel_helper_funcs = {
        .mode_set_base_atomic = intel_pipe_set_base_atomic,
-       .atomic_begin = intel_begin_crtc_commit,
-       .atomic_flush = intel_finish_crtc_commit,
        .atomic_check = intel_crtc_atomic_check,
 };
 
@@ -12919,11 +12731,6 @@ static int intel_atomic_prepare_commit(struct drm_device *dev,
        struct drm_crtc *crtc;
        int i, ret;
 
-       if (nonblock) {
-               DRM_DEBUG_KMS("i915 does not yet support nonblocking commit\n");
-               return -EINVAL;
-       }
-
        for_each_crtc_in_state(state, crtc, crtc_state, i) {
                struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
                struct intel_flip_work *work;
@@ -12965,6 +12772,11 @@ static int intel_atomic_prepare_commit(struct drm_device *dev,
                }
        }
 
+       if (intel_state->modeset && nonblock) {
+               DRM_DEBUG_ATOMIC("Nonblock modesets are not yet supported!\n");
+               return -EINVAL;
+       }
+
        ret = mutex_lock_interruptible(&dev->struct_mutex);
        if (ret)
                return ret;
@@ -13118,13 +12930,41 @@ static void intel_schedule_unpin(struct drm_crtc *crtc,
        queue_work(dev_priv->wq, &work->unpin_work);
 }
 
+static void intel_schedule_flip(struct drm_crtc *crtc,
+                               struct intel_atomic_state *state,
+                               struct intel_flip_work *work,
+                               bool nonblock)
+{
+       struct intel_crtc_state *crtc_state = work->new_crtc_state;
+
+       if (crtc_state->base.planes_changed ||
+           needs_modeset(&crtc_state->base) ||
+           crtc_state->update_pipe) {
+               if (nonblock)
+                       schedule_work(&work->mmio_work);
+               else
+                       intel_mmio_flip_work_func(&work->mmio_work);
+       } else {
+               int ret;
+
+               ret = drm_crtc_vblank_get(crtc);
+               I915_STATE_WARN(ret < 0, "enabling vblank failed with %i\n", ret);
+
+               work->flip_queued_vblank = intel_crtc_get_vblank_counter(to_intel_crtc(crtc));
+               smp_mb__before_atomic();
+               atomic_set(&work->pending, 1);
+       }
+}
+
 static void intel_schedule_update(struct drm_crtc *crtc,
                                  struct intel_atomic_state *state,
-                                 struct intel_flip_work *work)
+                                 struct intel_flip_work *work,
+                                 bool nonblock)
 {
        struct drm_device *dev = crtc->dev;
+       struct intel_crtc_state *pipe_config = work->new_crtc_state;
 
-       if (work->can_async_unpin) {
+       if (!pipe_config->base.active && work->can_async_unpin) {
                INIT_LIST_HEAD(&work->head);
                intel_schedule_unpin(crtc, state, work);
                return;
@@ -13134,7 +12974,10 @@ static void intel_schedule_update(struct drm_crtc *crtc,
        list_add_tail(&work->head, &to_intel_crtc(crtc)->flip_work);
        spin_unlock_irq(&dev->event_lock);
 
-       intel_schedule_unpin(crtc, state, work);
+       if (!pipe_config->base.active)
+               intel_schedule_unpin(crtc, state, work);
+       else
+               intel_schedule_flip(crtc, state, work, nonblock);
 }
 
 /**
@@ -13228,11 +13071,9 @@ static int intel_atomic_commit(struct drm_device *dev,
 
        /* Now enable the clocks, plane, pipe, and connectors that we set up. */
        for_each_crtc_in_state(state, crtc, old_crtc_state, i) {
+               struct intel_flip_work *work = intel_state->work[i];
                struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
                bool modeset = needs_modeset(crtc->state);
-               struct intel_crtc_state *pipe_config =
-                       to_intel_crtc_state(crtc->state);
-               bool update_pipe = !modeset && pipe_config->update_pipe;
 
                if (modeset && crtc->state->active) {
                        update_scanline_offset(to_intel_crtc(crtc));
@@ -13242,22 +13083,6 @@ static int intel_atomic_commit(struct drm_device *dev,
                if (!modeset)
                        intel_pre_plane_update(to_intel_crtc_state(old_crtc_state));
 
-               if (crtc->state->active &&
-                   drm_atomic_get_existing_plane_state(state, crtc->primary))
-                       intel_fbc_enable(intel_crtc, pipe_config, to_intel_plane_state(crtc->primary->state));
-
-               if (crtc->state->active &&
-                   (crtc->state->planes_changed || update_pipe))
-                       drm_atomic_helper_commit_planes_on_crtc(old_crtc_state);
-       }
-
-       /* FIXME: add subpixel order */
-
-       for_each_crtc_in_state(state, crtc, old_crtc_state, i) {
-               struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
-               struct intel_flip_work *work =
-                       intel_state->work[i];
-
                if (!work) {
                        if (!list_empty_careful(&intel_crtc->flip_work)) {
                                spin_lock_irq(&dev->event_lock);
@@ -13277,9 +13102,11 @@ static int intel_atomic_commit(struct drm_device *dev,
 
                intel_state->work[i] = NULL;
                intel_prepare_work(crtc, work, state, old_crtc_state);
-               intel_schedule_update(crtc, intel_state, work);
+               intel_schedule_update(crtc, intel_state, work, nonblock);
        }
 
+       /* FIXME: add subpixel order */
+
        drm_atomic_state_free(state);
 
        /* As one of the primary mmio accessors, KMS has a high likelihood
@@ -13343,11 +13170,38 @@ static const struct drm_crtc_funcs intel_crtc_funcs = {
        .set_config = drm_atomic_helper_set_config,
        .set_property = drm_atomic_helper_crtc_set_property,
        .destroy = intel_crtc_destroy,
-       .page_flip = intel_crtc_page_flip,
+       .page_flip = drm_atomic_helper_page_flip,
        .atomic_duplicate_state = intel_crtc_duplicate_state,
        .atomic_destroy_state = intel_crtc_destroy_state,
 };
 
+static struct fence *intel_get_excl_fence(struct drm_i915_gem_object *obj)
+{
+       struct reservation_object *resv;
+
+
+       if (!obj->base.dma_buf)
+               return NULL;
+
+       resv = obj->base.dma_buf->resv;
+
+       /* For framebuffer backed by dmabuf, wait for fence */
+       while (1) {
+               struct fence *fence_excl, *ret = NULL;
+
+               rcu_read_lock();
+
+               fence_excl = rcu_dereference(resv->fence_excl);
+               if (fence_excl)
+                       ret = fence_get_rcu(fence_excl);
+
+               rcu_read_unlock();
+
+               if (ret == fence_excl)
+                       return ret;
+       }
+}
+
 /**
  * intel_prepare_plane_fb - Prepare fb for usage on plane
  * @plane: drm plane to prepare for
@@ -13534,42 +13388,6 @@ intel_check_primary_plane(struct drm_plane *plane,
                                             &state->visible);
 }
 
-static void intel_begin_crtc_commit(struct drm_crtc *crtc,
-                                   struct drm_crtc_state *old_crtc_state)
-{
-       struct drm_device *dev = crtc->dev;
-       struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
-       struct intel_crtc_state *old_intel_state =
-               to_intel_crtc_state(old_crtc_state);
-       bool modeset = needs_modeset(crtc->state);
-
-       intel_frontbuffer_flip_prepare(dev, to_intel_crtc_state(crtc->state)->fb_bits);
-
-       /* Perform vblank evasion around commit operation */
-       intel_pipe_update_start(intel_crtc);
-
-       if (modeset)
-               return;
-
-       if (crtc->state->color_mgmt_changed || to_intel_crtc_state(crtc->state)->update_pipe) {
-               intel_color_set_csc(crtc->state);
-               intel_color_load_luts(crtc->state);
-       }
-
-       if (to_intel_crtc_state(crtc->state)->update_pipe)
-               intel_update_pipe_config(intel_crtc, old_intel_state);
-       else if (INTEL_INFO(dev)->gen >= 9)
-               skl_detach_scalers(intel_crtc);
-}
-
-static void intel_finish_crtc_commit(struct drm_crtc *crtc,
-                                    struct drm_crtc_state *old_crtc_state)
-{
-       struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
-
-       intel_pipe_update_end(intel_crtc, NULL);
-}
-
 /**
  * intel_plane_destroy - destroy a plane
  * @plane: plane to destroy