Merge tag 'drm-misc-next-2017-05-16' of git://anongit.freedesktop.org/git/drm-misc into drm-next

UAPI Changes:
- Return -ENODEV instead of -ENXIO when creating cma fb w/o valid gem (Daniel)
- Add aspect ratio and custom scaling propertis to connector state (Maarten)

Cross-subsystem Changes:
- None

Core Changes:
- Add Laurent as bridge reviewer and Andrzej as bridge maintainer (Archit)
- Maintain new STM driver through -misc (Yannick)
- Misc doc improvements (as is tradition) (Daniel)
- Add driver-private objects to atomic state (Dhinakaran)
- Deprecate preclose hook in modern drivers (use postclose) (Daniel)
- Add hwmode to vblank struct. This fixes mode access in irq context and reduced
  a bunch of boilerplate (Daniel)

Driver Changes:
- vc4: Add out-fence support to vc4 V3D rendering (Eric)
- stm: Add stm32f429 display hw and am-480272h3tmqw-t01h panel support (Yannick)
- vc4: Remove 256MB cma limit from vc4 (Eric)
- dw-hdmi: Disable audio when inactive, instead of always enabled (Romain)
- zte: Add support for VGA to the ZTE driver (Shawn)
- i915: Track DP MST bandwidth and check it in atomic_check (Dhinakaran)
- vgem: Enable gem dmabuf import iface to facilitate ion testing (Laura)
- vc4: Add support for Cygnus (new dt compat string + couple bug fixes) (Eric)
- pl111: Add driver for pl111 CLCD display controller (Eric/Tom)
- vgem: Subclass drm_device instead of standalone platform device (Chris)

Cc: Archit Taneja <architt@codeaurora.org>
Cc: Eric Anholt <eric@anholt.net>
Cc: Yannick Fertre <yannick.fertre@st.com>
Cc: Romain Perier <romain.perier@collabora.com>
Cc: Navare, Manasi D <manasi.d.navare@intel.com>
Cc: Shawn Guo <shawn.guo@linaro.org>
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Tom Cooksey <tom.cooksey@arm.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>

* tag 'drm-misc-next-2017-05-16' of git://anongit.freedesktop.org/git/drm-misc: (72 commits)
  drm: add missing declaration to drm_blend.h
  drm/dp: Wait up all outstanding tx waiters
  drm/dp: Read the tx msg state once after checking for an event
  drm/prime: Forward declare struct device
  drm/vblank: Lock down vblank->hwmode more
  drm/vblank: drop the mode argument from drm_calc_vbltimestamp_from_scanoutpos
  drm/vblank: Add FIXME comments about moving the vblank ts hooks
  drm/vblank: Switch to bool in_vblank_irq in get_vblank_timestamp
  drm/vblank: Switch drm_driver->get_vblank_timestamp to return a bool
  drm/vgem: Convert to a struct drm_device subclass
  gpu: drm: gma500: remove dead code
  drm/sti: Adjust two checks for null pointers in sti_hqvdp_probe()
  drm/sti: Fix typos in a comment line
  drm/sti: Fix a typo in a comment line
  drm/sti: Replace 17 seq_puts() calls by seq_putc()
  drm/sti: Reduce function calls for sequence output at five places
  drm/sti: use seq_puts to display a string
  drm: Nerf the preclose callback for modern drivers
  drm/exynos: Merge pre/postclose hooks
  drm/tegra: switch to postclose
  ...

1 files changed, 150 insertions, 11 deletions

diff --git a/drivers/gpu/drm/vc4/vc4_gem.c b/drivers/gpu/drm/vc4/vc4_gem.c
index e9c381c42139..735412e3725a 100644
--- a/drivers/gpu/drm/vc4/vc4_gem.c
+++ b/drivers/gpu/drm/vc4/vc4_gem.c
@@ -463,6 +463,8 @@ vc4_update_bo_seqnos(struct vc4_exec_info *exec, uint64_t seqno)
 	for (i = 0; i < exec->bo_count; i++) {
 		bo = to_vc4_bo(&exec->bo[i]->base);
 		bo->seqno = seqno;
+
+		reservation_object_add_shared_fence(bo->resv, exec->fence);
 	}
 
 	list_for_each_entry(bo, &exec->unref_list, unref_head) {
@@ -472,7 +474,103 @@ vc4_update_bo_seqnos(struct vc4_exec_info *exec, uint64_t seqno)
 	for (i = 0; i < exec->rcl_write_bo_count; i++) {
 		bo = to_vc4_bo(&exec->rcl_write_bo[i]->base);
 		bo->write_seqno = seqno;
+
+		reservation_object_add_excl_fence(bo->resv, exec->fence);
+	}
+}
+
+static void
+vc4_unlock_bo_reservations(struct drm_device *dev,
+			   struct vc4_exec_info *exec,
+			   struct ww_acquire_ctx *acquire_ctx)
+{
+	int i;
+
+	for (i = 0; i < exec->bo_count; i++) {
+		struct vc4_bo *bo = to_vc4_bo(&exec->bo[i]->base);
+
+		ww_mutex_unlock(&bo->resv->lock);
+	}
+
+	ww_acquire_fini(acquire_ctx);
+}
+
+/* Takes the reservation lock on all the BOs being referenced, so that
+ * at queue submit time we can update the reservations.
+ *
+ * We don't lock the RCL the tile alloc/state BOs, or overflow memory
+ * (all of which are on exec->unref_list).  They're entirely private
+ * to vc4, so we don't attach dma-buf fences to them.
+ */
+static int
+vc4_lock_bo_reservations(struct drm_device *dev,
+			 struct vc4_exec_info *exec,
+			 struct ww_acquire_ctx *acquire_ctx)
+{
+	int contended_lock = -1;
+	int i, ret;
+	struct vc4_bo *bo;
+
+	ww_acquire_init(acquire_ctx, &reservation_ww_class);
+
+retry:
+	if (contended_lock != -1) {
+		bo = to_vc4_bo(&exec->bo[contended_lock]->base);
+		ret = ww_mutex_lock_slow_interruptible(&bo->resv->lock,
+						       acquire_ctx);
+		if (ret) {
+			ww_acquire_done(acquire_ctx);
+			return ret;
+		}
+	}
+
+	for (i = 0; i < exec->bo_count; i++) {
+		if (i == contended_lock)
+			continue;
+
+		bo = to_vc4_bo(&exec->bo[i]->base);
+
+		ret = ww_mutex_lock_interruptible(&bo->resv->lock, acquire_ctx);
+		if (ret) {
+			int j;
+
+			for (j = 0; j < i; j++) {
+				bo = to_vc4_bo(&exec->bo[j]->base);
+				ww_mutex_unlock(&bo->resv->lock);
+			}
+
+			if (contended_lock != -1 && contended_lock >= i) {
+				bo = to_vc4_bo(&exec->bo[contended_lock]->base);
+
+				ww_mutex_unlock(&bo->resv->lock);
+			}
+
+			if (ret == -EDEADLK) {
+				contended_lock = i;
+				goto retry;
+			}
+
+			ww_acquire_done(acquire_ctx);
+			return ret;
+		}
 	}
+
+	ww_acquire_done(acquire_ctx);
+
+	/* Reserve space for our shared (read-only) fence references,
+	 * before we commit the CL to the hardware.
+	 */
+	for (i = 0; i < exec->bo_count; i++) {
+		bo = to_vc4_bo(&exec->bo[i]->base);
+
+		ret = reservation_object_reserve_shared(bo->resv);
+		if (ret) {
+			vc4_unlock_bo_reservations(dev, exec, acquire_ctx);
+			return ret;
+		}
+	}
+
+	return 0;
 }
 
 /* Queues a struct vc4_exec_info for execution.  If no job is
@@ -484,19 +582,34 @@ vc4_update_bo_seqnos(struct vc4_exec_info *exec, uint64_t seqno)
  * then bump the end address.  That's a change for a later date,
  * though.
  */
-static void
-vc4_queue_submit(struct drm_device *dev, struct vc4_exec_info *exec)
+static int
+vc4_queue_submit(struct drm_device *dev, struct vc4_exec_info *exec,
+		 struct ww_acquire_ctx *acquire_ctx)
 {
 	struct vc4_dev *vc4 = to_vc4_dev(dev);
 	uint64_t seqno;
 	unsigned long irqflags;
+	struct vc4_fence *fence;
+
+	fence = kzalloc(sizeof(*fence), GFP_KERNEL);
+	if (!fence)
+		return -ENOMEM;
+	fence->dev = dev;
 
 	spin_lock_irqsave(&vc4->job_lock, irqflags);
 
 	seqno = ++vc4->emit_seqno;
 	exec->seqno = seqno;
+
+	dma_fence_init(&fence->base, &vc4_fence_ops, &vc4->job_lock,
+		       vc4->dma_fence_context, exec->seqno);
+	fence->seqno = exec->seqno;
+	exec->fence = &fence->base;
+
 	vc4_update_bo_seqnos(exec, seqno);
 
+	vc4_unlock_bo_reservations(dev, exec, acquire_ctx);
+
 	list_add_tail(&exec->head, &vc4->bin_job_list);
 
 	/* If no job was executing, kick ours off.  Otherwise, it'll
@@ -509,6 +622,8 @@ vc4_queue_submit(struct drm_device *dev, struct vc4_exec_info *exec)
 	}
 
 	spin_unlock_irqrestore(&vc4->job_lock, irqflags);
+
+	return 0;
 }
 
 /**
@@ -705,8 +820,15 @@ static void
 vc4_complete_exec(struct drm_device *dev, struct vc4_exec_info *exec)
 {
 	struct vc4_dev *vc4 = to_vc4_dev(dev);
+	unsigned long irqflags;
 	unsigned i;
 
+	/* If we got force-completed because of GPU reset rather than
+	 * through our IRQ handler, signal the fence now.
+	 */
+	if (exec->fence)
+		dma_fence_signal(exec->fence);
+
 	if (exec->bo) {
 		for (i = 0; i < exec->bo_count; i++)
 			drm_gem_object_unreference_unlocked(&exec->bo[i]->base);
@@ -720,6 +842,11 @@ vc4_complete_exec(struct drm_device *dev, struct vc4_exec_info *exec)
 		drm_gem_object_unreference_unlocked(&bo->base.base);
 	}
 
+	/* Free up the allocation of any bin slots we used. */
+	spin_lock_irqsave(&vc4->job_lock, irqflags);
+	vc4->bin_alloc_used &= ~exec->bin_slots;
+	spin_unlock_irqrestore(&vc4->job_lock, irqflags);
+
 	mutex_lock(&vc4->power_lock);
 	if (--vc4->power_refcount == 0) {
 		pm_runtime_mark_last_busy(&vc4->v3d->pdev->dev);
@@ -874,6 +1001,7 @@ vc4_submit_cl_ioctl(struct drm_device *dev, void *data,
 	struct vc4_dev *vc4 = to_vc4_dev(dev);
 	struct drm_vc4_submit_cl *args = data;
 	struct vc4_exec_info *exec;
+	struct ww_acquire_ctx acquire_ctx;
 	int ret = 0;
 
 	if ((args->flags & ~VC4_SUBMIT_CL_USE_CLEAR_COLOR) != 0) {
@@ -888,13 +1016,16 @@ vc4_submit_cl_ioctl(struct drm_device *dev, void *data,
 	}
 
 	mutex_lock(&vc4->power_lock);
-	if (vc4->power_refcount++ == 0)
+	if (vc4->power_refcount++ == 0) {
 		ret = pm_runtime_get_sync(&vc4->v3d->pdev->dev);
-	mutex_unlock(&vc4->power_lock);
-	if (ret < 0) {
-		kfree(exec);
-		return ret;
+		if (ret < 0) {
+			mutex_unlock(&vc4->power_lock);
+			vc4->power_refcount--;
+			kfree(exec);
+			return ret;
+		}
 	}
+	mutex_unlock(&vc4->power_lock);
 
 	exec->args = args;
 	INIT_LIST_HEAD(&exec->unref_list);
@@ -916,12 +1047,18 @@ vc4_submit_cl_ioctl(struct drm_device *dev, void *data,
 	if (ret)
 		goto fail;
 
+	ret = vc4_lock_bo_reservations(dev, exec, &acquire_ctx);
+	if (ret)
+		goto fail;
+
 	/* Clear this out of the struct we'll be putting in the queue,
 	 * since it's part of our stack.
 	 */
 	exec->args = NULL;
 
-	vc4_queue_submit(dev, exec);
+	ret = vc4_queue_submit(dev, exec, &acquire_ctx);
+	if (ret)
+		goto fail;
 
 	/* Return the seqno for our job. */
 	args->seqno = vc4->emit_seqno;
@@ -939,6 +1076,8 @@ vc4_gem_init(struct drm_device *dev)
 {
 	struct vc4_dev *vc4 = to_vc4_dev(dev);
 
+	vc4->dma_fence_context = dma_fence_context_alloc(1);
+
 	INIT_LIST_HEAD(&vc4->bin_job_list);
 	INIT_LIST_HEAD(&vc4->render_job_list);
 	INIT_LIST_HEAD(&vc4->job_done_list);
@@ -968,9 +1107,9 @@ vc4_gem_destroy(struct drm_device *dev)
 	/* V3D should already have disabled its interrupt and cleared
 	 * the overflow allocation registers.  Now free the object.
 	 */
-	if (vc4->overflow_mem) {
-		drm_gem_object_unreference_unlocked(&vc4->overflow_mem->base.base);
-		vc4->overflow_mem = NULL;
+	if (vc4->bin_bo) {
+		drm_gem_object_put_unlocked(&vc4->bin_bo->base.base);
+		vc4->bin_bo = NULL;
 	}
 
 	if (vc4->hang_state)