linux/drivers/gpu/drm/nouveau/nouveau_fence.c - nest-guard/1.0/linux - Git at Google

 /*
  * Copyright (C) 2007 Ben Skeggs.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining
  * a copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
  * without limitation the rights to use, copy, modify, merge, publish,
  * distribute, sublicense, and/or sell copies of the Software, and to
  * permit persons to whom the Software is furnished to do so, subject to
  * the following conditions:
  *
  * The above copyright notice and this permission notice (including the
  * next paragraph) shall be included in all copies or substantial
  * portions of the Software.
  *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
  * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
  * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
  * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
  * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  *
  */

 #include "drmP.h"
 #include "drm.h"

 #include "nouveau_drv.h"
 #include "nouveau_ramht.h"
 #include "nouveau_dma.h"

 #define USE_REFCNT(dev) (nouveau_private(dev)->chipset >= 0x10)
 #define USE_SEMA(dev) (nouveau_private(dev)->chipset >= 0x17)

 struct nouveau_fence {
 	struct nouveau_channel *channel;
 	struct kref refcount;
 	struct list_head entry;

 	uint32_t sequence;
 	bool signalled;

 	void (*work)(void *priv, bool signalled);
 	void *priv;
 };

 struct nouveau_semaphore {
 	struct kref ref;
 	struct drm_device *dev;
 	struct drm_mm_node *mem;
 };

 static inline struct nouveau_fence *
 nouveau_fence(void *sync_obj)
 {
 	return (struct nouveau_fence *)sync_obj;
 }

 static void
 nouveau_fence_del(struct kref *ref)
 {
 	struct nouveau_fence *fence =
 		container_of(ref, struct nouveau_fence, refcount);

 	kfree(fence);
 }

 void
 nouveau_fence_update(struct nouveau_channel *chan)
 {
 	struct drm_device *dev = chan->dev;
 	struct nouveau_fence *tmp, *fence;
 	uint32_t sequence;

 	spin_lock(&chan->fence.lock);

 	if (USE_REFCNT(dev))
 		sequence = nvchan_rd32(chan, 0x48);
 	else
 		sequence = atomic_read(&chan->fence.last_sequence_irq);

 	if (chan->fence.sequence_ack == sequence)
 		goto out;
 	chan->fence.sequence_ack = sequence;

 	list_for_each_entry_safe(fence, tmp, &chan->fence.pending, entry) {
 		sequence = fence->sequence;
 		fence->signalled = true;
 		list_del(&fence->entry);

 		if (unlikely(fence->work))
 			fence->work(fence->priv, true);

 		kref_put(&fence->refcount, nouveau_fence_del);

 		if (sequence == chan->fence.sequence_ack)
 			break;
 	}
 out:
 	spin_unlock(&chan->fence.lock);
 }

 int
 nouveau_fence_new(struct nouveau_channel *chan, struct nouveau_fence **pfence,
 		  bool emit)
 {
 	struct nouveau_fence *fence;
 	int ret = 0;

 	fence = kzalloc(sizeof(*fence), GFP_KERNEL);
 	if (!fence)
 		return -ENOMEM;
 	kref_init(&fence->refcount);
 	fence->channel = chan;

 	if (emit)
 		ret = nouveau_fence_emit(fence);

 	if (ret)
 		nouveau_fence_unref((void *)&fence);
 	*pfence = fence;
 	return ret;
 }

 struct nouveau_channel *
 nouveau_fence_channel(struct nouveau_fence *fence)
 {
 	return fence ? fence->channel : NULL;
 }

 int
 nouveau_fence_emit(struct nouveau_fence *fence)
 {
 	struct nouveau_channel *chan = fence->channel;
 	struct drm_device *dev = chan->dev;
 	int ret;

 	ret = RING_SPACE(chan, 2);
 	if (ret)
 		return ret;

 	if (unlikely(chan->fence.sequence == chan->fence.sequence_ack - 1)) {
 		nouveau_fence_update(chan);

 		BUG_ON(chan->fence.sequence ==
 		       chan->fence.sequence_ack - 1);
 	}

 	fence->sequence = ++chan->fence.sequence;

 	kref_get(&fence->refcount);
 	spin_lock(&chan->fence.lock);
 	list_add_tail(&fence->entry, &chan->fence.pending);
 	spin_unlock(&chan->fence.lock);

 	BEGIN_RING(chan, NvSubSw, USE_REFCNT(dev) ? 0x0050 : 0x0150, 1);
 	OUT_RING(chan, fence->sequence);
 	FIRE_RING(chan);

 	return 0;
 }

 void
 nouveau_fence_work(struct nouveau_fence *fence,
 		   void (*work)(void *priv, bool signalled),
 		   void *priv)
 {
 	BUG_ON(fence->work);

 	spin_lock(&fence->channel->fence.lock);

 	if (fence->signalled) {
 		work(priv, true);
 	} else {
 		fence->work = work;
 		fence->priv = priv;
 	}

 	spin_unlock(&fence->channel->fence.lock);
 }

 void
 nouveau_fence_unref(void **sync_obj)
 {
 	struct nouveau_fence *fence = nouveau_fence(*sync_obj);

 	if (fence)
 		kref_put(&fence->refcount, nouveau_fence_del);
 	*sync_obj = NULL;
 }

 void *
 nouveau_fence_ref(void *sync_obj)
 {
 	struct nouveau_fence *fence = nouveau_fence(sync_obj);

 	kref_get(&fence->refcount);
 	return sync_obj;
 }

 bool
 nouveau_fence_signalled(void *sync_obj, void *sync_arg)
 {
 	struct nouveau_fence *fence = nouveau_fence(sync_obj);
 	struct nouveau_channel *chan = fence->channel;

 	if (fence->signalled)
 		return true;

 	nouveau_fence_update(chan);
 	return fence->signalled;
 }

 int
 nouveau_fence_wait(void *sync_obj, void *sync_arg, bool lazy, bool intr)
 {
 	unsigned long timeout = jiffies + (3 * DRM_HZ);
 	int ret = 0;

 	while (1) {
 		if (nouveau_fence_signalled(sync_obj, sync_arg))
 			break;

 		if (time_after_eq(jiffies, timeout)) {
 			ret = -EBUSY;
 			break;
 		}

 		__set_current_state(intr ? TASK_INTERRUPTIBLE
 			: TASK_UNINTERRUPTIBLE);
 		if (lazy)
 			schedule_timeout(1);

 		if (intr && signal_pending(current)) {
 			ret = -ERESTARTSYS;
 			break;
 		}
 	}

 	__set_current_state(TASK_RUNNING);

 	return ret;
 }

 static struct nouveau_semaphore *
 alloc_semaphore(struct drm_device *dev)
 {
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_semaphore *sema;
 	int ret;

 	if (!USE_SEMA(dev))
 		return NULL;

 	sema = kmalloc(sizeof(*sema), GFP_KERNEL);
 	if (!sema)
 		goto fail;

 	ret = drm_mm_pre_get(&dev_priv->fence.heap);
 	if (ret)
 		goto fail;

 	spin_lock(&dev_priv->fence.lock);
 	sema->mem = drm_mm_search_free(&dev_priv->fence.heap, 4, 0, 0);
 	if (sema->mem)
 		sema->mem = drm_mm_get_block_atomic(sema->mem, 4, 0);
 	spin_unlock(&dev_priv->fence.lock);

 	if (!sema->mem)
 		goto fail;

 	kref_init(&sema->ref);
 	sema->dev = dev;
 	nouveau_bo_wr32(dev_priv->fence.bo, sema->mem->start / 4, 0);

 	return sema;
 fail:
 	kfree(sema);
 	return NULL;
 }

 static void
 free_semaphore(struct kref *ref)
 {
 	struct nouveau_semaphore *sema =
 		container_of(ref, struct nouveau_semaphore, ref);
 	struct drm_nouveau_private *dev_priv = sema->dev->dev_private;

 	spin_lock(&dev_priv->fence.lock);
 	drm_mm_put_block(sema->mem);
 	spin_unlock(&dev_priv->fence.lock);

 	kfree(sema);
 }

 static void
 semaphore_work(void *priv, bool signalled)
 {
 	struct nouveau_semaphore *sema = priv;
 	struct drm_nouveau_private *dev_priv = sema->dev->dev_private;

 	if (unlikely(!signalled))
 		nouveau_bo_wr32(dev_priv->fence.bo, sema->mem->start / 4, 1);

 	kref_put(&sema->ref, free_semaphore);
 }

 static int
 emit_semaphore(struct nouveau_channel *chan, int method,
 	       struct nouveau_semaphore *sema)
 {
 	struct drm_nouveau_private *dev_priv = sema->dev->dev_private;
 	struct nouveau_fence *fence;
 	bool smart = (dev_priv->card_type >= NV_50);
 	int ret;

 	ret = RING_SPACE(chan, smart ? 8 : 4);
 	if (ret)
 		return ret;

 	if (smart) {
 		BEGIN_RING(chan, NvSubSw, NV_SW_DMA_SEMAPHORE, 1);
 		OUT_RING(chan, NvSema);
 	}
 	BEGIN_RING(chan, NvSubSw, NV_SW_SEMAPHORE_OFFSET, 1);
 	OUT_RING(chan, sema->mem->start);

 	if (smart && method == NV_SW_SEMAPHORE_ACQUIRE) {
 		/*
 		 * NV50 tries to be too smart and context-switch
 		 * between semaphores instead of doing a "first come,
 		 * first served" strategy like previous cards
 		 * do.
 		 *
 		 * That's bad because the ACQUIRE latency can get as
 		 * large as the PFIFO context time slice in the
 		 * typical DRI2 case where you have several
 		 * outstanding semaphores at the same moment.
 		 *
 		 * If we're going to ACQUIRE, force the card to
 		 * context switch before, just in case the matching
 		 * RELEASE is already scheduled to be executed in
 		 * another channel.
 		 */
 		BEGIN_RING(chan, NvSubSw, NV_SW_YIELD, 1);
 		OUT_RING(chan, 0);
 	}

 	BEGIN_RING(chan, NvSubSw, method, 1);
 	OUT_RING(chan, 1);

 	if (smart && method == NV_SW_SEMAPHORE_RELEASE) {
 		/*
 		 * Force the card to context switch, there may be
 		 * another channel waiting for the semaphore we just
 		 * released.
 		 */
 		BEGIN_RING(chan, NvSubSw, NV_SW_YIELD, 1);
 		OUT_RING(chan, 0);
 	}

 	/* Delay semaphore destruction until its work is done */
 	ret = nouveau_fence_new(chan, &fence, true);
 	if (ret)
 		return ret;

 	kref_get(&sema->ref);
 	nouveau_fence_work(fence, semaphore_work, sema);
 	nouveau_fence_unref((void *)&fence);

 	return 0;
 }

 int
 nouveau_fence_sync(struct nouveau_fence *fence,
 		   struct nouveau_channel *wchan)
 {
 	struct nouveau_channel *chan = nouveau_fence_channel(fence);
 	struct drm_device *dev = wchan->dev;
 	struct nouveau_semaphore *sema;
 	int ret;

 	if (likely(!fence || chan == wchan ||
 		   nouveau_fence_signalled(fence, NULL)))
 		return 0;

 	sema = alloc_semaphore(dev);
 	if (!sema) {
 		/* Early card or broken userspace, fall back to
 		 * software sync. */
 		return nouveau_fence_wait(fence, NULL, false, false);
 	}

 	/* Make wchan wait until it gets signalled */
 	ret = emit_semaphore(wchan, NV_SW_SEMAPHORE_ACQUIRE, sema);
 	if (ret)
 		goto out;

 	/* Signal the semaphore from chan */
 	ret = emit_semaphore(chan, NV_SW_SEMAPHORE_RELEASE, sema);
 out:
 	kref_put(&sema->ref, free_semaphore);
 	return ret;
 }

 int
 nouveau_fence_flush(void *sync_obj, void *sync_arg)
 {
 	return 0;
 }

 int
 nouveau_fence_channel_init(struct nouveau_channel *chan)
 {
 	struct drm_device *dev = chan->dev;
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_gpuobj *obj = NULL;
 	int ret;

 	/* Create an NV_SW object for various sync purposes */
 	ret = nouveau_gpuobj_sw_new(chan, NV_SW, &obj);
 	if (ret)
 		return ret;

 	ret = nouveau_ramht_insert(chan, NvSw, obj);
 	nouveau_gpuobj_ref(NULL, &obj);
 	if (ret)
 		return ret;

 	ret = RING_SPACE(chan, 2);
 	if (ret)
 		return ret;
 	BEGIN_RING(chan, NvSubSw, 0, 1);
 	OUT_RING(chan, NvSw);

 	/* Create a DMA object for the shared cross-channel sync area. */
 	if (USE_SEMA(dev)) {
 		struct drm_mm_node *mem = dev_priv->fence.bo->bo.mem.mm_node;

 		ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY,
 					     mem->start << PAGE_SHIFT,
 					     mem->size << PAGE_SHIFT,
 					     NV_DMA_ACCESS_RW,
 					     NV_DMA_TARGET_VIDMEM, &obj);
 		if (ret)
 			return ret;

 		ret = nouveau_ramht_insert(chan, NvSema, obj);
 		nouveau_gpuobj_ref(NULL, &obj);
 		if (ret)
 			return ret;

 		ret = RING_SPACE(chan, 2);
 		if (ret)
 			return ret;
 		BEGIN_RING(chan, NvSubSw, NV_SW_DMA_SEMAPHORE, 1);
 		OUT_RING(chan, NvSema);
 	}

 	FIRE_RING(chan);

 	INIT_LIST_HEAD(&chan->fence.pending);
 	spin_lock_init(&chan->fence.lock);
 	atomic_set(&chan->fence.last_sequence_irq, 0);

 	return 0;
 }

 void
 nouveau_fence_channel_fini(struct nouveau_channel *chan)
 {
 	struct nouveau_fence *tmp, *fence;

 	list_for_each_entry_safe(fence, tmp, &chan->fence.pending, entry) {
 		fence->signalled = true;
 		list_del(&fence->entry);

 		if (unlikely(fence->work))
 			fence->work(fence->priv, false);

 		kref_put(&fence->refcount, nouveau_fence_del);
 	}
 }

 int
 nouveau_fence_init(struct drm_device *dev)
 {
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	int ret;

 	/* Create a shared VRAM heap for cross-channel sync. */
 	if (USE_SEMA(dev)) {
 		ret = nouveau_bo_new(dev, NULL, 4096, 0, TTM_PL_FLAG_VRAM,
 				     0, 0, false, true, &dev_priv->fence.bo);
 		if (ret)
 			return ret;

 		ret = nouveau_bo_pin(dev_priv->fence.bo, TTM_PL_FLAG_VRAM);
 		if (ret)
 			goto fail;

 		ret = nouveau_bo_map(dev_priv->fence.bo);
 		if (ret)
 			goto fail;

 		ret = drm_mm_init(&dev_priv->fence.heap, 0,
 				  dev_priv->fence.bo->bo.mem.size);
 		if (ret)
 			goto fail;

 		spin_lock_init(&dev_priv->fence.lock);
 	}

 	return 0;
 fail:
 	nouveau_bo_unmap(dev_priv->fence.bo);
 	nouveau_bo_ref(NULL, &dev_priv->fence.bo);
 	return ret;
 }

 void
 nouveau_fence_fini(struct drm_device *dev)
 {
 	struct drm_nouveau_private *dev_priv = dev->dev_private;

 	if (USE_SEMA(dev)) {
 		drm_mm_takedown(&dev_priv->fence.heap);
 		nouveau_bo_unmap(dev_priv->fence.bo);
 		nouveau_bo_unpin(dev_priv->fence.bo);
 		nouveau_bo_ref(NULL, &dev_priv->fence.bo);
 	}
 }
	/*
	* Copyright (C) 2007 Ben Skeggs.
	* All Rights Reserved.
	*
	* Permission is hereby granted, free of charge, to any person obtaining
	* a copy of this software and associated documentation files (the
	* "Software"), to deal in the Software without restriction, including
	* without limitation the rights to use, copy, modify, merge, publish,
	* distribute, sublicense, and/or sell copies of the Software, and to
	* permit persons to whom the Software is furnished to do so, subject to
	* the following conditions:
	*
	* The above copyright notice and this permission notice (including the
	* next paragraph) shall be included in all copies or substantial
	* portions of the Software.
	*
	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
	* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
	* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
	* IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
	* LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
	* OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
	* WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
	*
	*/

	#include "drmP.h"
	#include "drm.h"

	#include "nouveau_drv.h"
	#include "nouveau_ramht.h"
	#include "nouveau_dma.h"

	#define USE_REFCNT(dev) (nouveau_private(dev)->chipset >= 0x10)
	#define USE_SEMA(dev) (nouveau_private(dev)->chipset >= 0x17)

	struct nouveau_fence {
	struct nouveau_channel *channel;
	struct kref refcount;
	struct list_head entry;

	uint32_t sequence;
	bool signalled;

	void (work)(void priv, bool signalled);
	void *priv;
	};

	struct nouveau_semaphore {
	struct kref ref;
	struct drm_device *dev;
	struct drm_mm_node *mem;
	};

	static inline struct nouveau_fence *
	nouveau_fence(void *sync_obj)
	{
	return (struct nouveau_fence *)sync_obj;
	}

	static void
	nouveau_fence_del(struct kref *ref)
	{
	struct nouveau_fence *fence =
	container_of(ref, struct nouveau_fence, refcount);

	kfree(fence);
	}

	void
	nouveau_fence_update(struct nouveau_channel *chan)
	{
	struct drm_device *dev = chan->dev;
	struct nouveau_fence tmp, fence;
	uint32_t sequence;

	spin_lock(&chan->fence.lock);

	if (USE_REFCNT(dev))
	sequence = nvchan_rd32(chan, 0x48);
	else
	sequence = atomic_read(&chan->fence.last_sequence_irq);

	if (chan->fence.sequence_ack == sequence)
	goto out;
	chan->fence.sequence_ack = sequence;

	list_for_each_entry_safe(fence, tmp, &chan->fence.pending, entry) {
	sequence = fence->sequence;
	fence->signalled = true;
	list_del(&fence->entry);

	if (unlikely(fence->work))
	fence->work(fence->priv, true);

	kref_put(&fence->refcount, nouveau_fence_del);

	if (sequence == chan->fence.sequence_ack)
	break;
	}
	out:
	spin_unlock(&chan->fence.lock);
	}

	int
	nouveau_fence_new(struct nouveau_channel chan, struct nouveau_fence *pfence,
	bool emit)
	{
	struct nouveau_fence *fence;
	int ret = 0;

	fence = kzalloc(sizeof(*fence), GFP_KERNEL);
	if (!fence)
	return -ENOMEM;
	kref_init(&fence->refcount);
	fence->channel = chan;

	if (emit)
	ret = nouveau_fence_emit(fence);

	if (ret)
	nouveau_fence_unref((void *)&fence);
	*pfence = fence;
	return ret;
	}

	struct nouveau_channel *
	nouveau_fence_channel(struct nouveau_fence *fence)
	{
	return fence ? fence->channel : NULL;
	}

	int
	nouveau_fence_emit(struct nouveau_fence *fence)
	{
	struct nouveau_channel *chan = fence->channel;
	struct drm_device *dev = chan->dev;
	int ret;

	ret = RING_SPACE(chan, 2);
	if (ret)
	return ret;

	if (unlikely(chan->fence.sequence == chan->fence.sequence_ack - 1)) {
	nouveau_fence_update(chan);

	BUG_ON(chan->fence.sequence ==
	chan->fence.sequence_ack - 1);
	}

	fence->sequence = ++chan->fence.sequence;

	kref_get(&fence->refcount);
	spin_lock(&chan->fence.lock);
	list_add_tail(&fence->entry, &chan->fence.pending);
	spin_unlock(&chan->fence.lock);

	BEGIN_RING(chan, NvSubSw, USE_REFCNT(dev) ? 0x0050 : 0x0150, 1);
	OUT_RING(chan, fence->sequence);
	FIRE_RING(chan);

	return 0;
	}

	void
	nouveau_fence_work(struct nouveau_fence *fence,
	void (work)(void priv, bool signalled),
	void *priv)
	{
	BUG_ON(fence->work);

	spin_lock(&fence->channel->fence.lock);

	if (fence->signalled) {
	work(priv, true);
	} else {
	fence->work = work;
	fence->priv = priv;
	}

	spin_unlock(&fence->channel->fence.lock);
	}

	void
	nouveau_fence_unref(void **sync_obj)
	{
	struct nouveau_fence fence = nouveau_fence(sync_obj);

	if (fence)
	kref_put(&fence->refcount, nouveau_fence_del);
	*sync_obj = NULL;
	}

	void *
	nouveau_fence_ref(void *sync_obj)
	{
	struct nouveau_fence *fence = nouveau_fence(sync_obj);

	kref_get(&fence->refcount);
	return sync_obj;
	}

	bool
	nouveau_fence_signalled(void sync_obj, void sync_arg)
	{
	struct nouveau_fence *fence = nouveau_fence(sync_obj);
	struct nouveau_channel *chan = fence->channel;

	if (fence->signalled)
	return true;

	nouveau_fence_update(chan);
	return fence->signalled;
	}

	int
	nouveau_fence_wait(void sync_obj, void sync_arg, bool lazy, bool intr)
	{
	unsigned long timeout = jiffies + (3 * DRM_HZ);
	int ret = 0;

	while (1) {
	if (nouveau_fence_signalled(sync_obj, sync_arg))
	break;

	if (time_after_eq(jiffies, timeout)) {
	ret = -EBUSY;
	break;
	}

	__set_current_state(intr ? TASK_INTERRUPTIBLE
	: TASK_UNINTERRUPTIBLE);
	if (lazy)
	schedule_timeout(1);

	if (intr && signal_pending(current)) {
	ret = -ERESTARTSYS;
	break;
	}
	}

	__set_current_state(TASK_RUNNING);

	return ret;
	}

	static struct nouveau_semaphore *
	alloc_semaphore(struct drm_device *dev)
	{
	struct drm_nouveau_private *dev_priv = dev->dev_private;
	struct nouveau_semaphore *sema;
	int ret;

	if (!USE_SEMA(dev))
	return NULL;

	sema = kmalloc(sizeof(*sema), GFP_KERNEL);
	if (!sema)
	goto fail;

	ret = drm_mm_pre_get(&dev_priv->fence.heap);
	if (ret)
	goto fail;

	spin_lock(&dev_priv->fence.lock);
	sema->mem = drm_mm_search_free(&dev_priv->fence.heap, 4, 0, 0);
	if (sema->mem)
	sema->mem = drm_mm_get_block_atomic(sema->mem, 4, 0);
	spin_unlock(&dev_priv->fence.lock);

	if (!sema->mem)
	goto fail;

	kref_init(&sema->ref);
	sema->dev = dev;
	nouveau_bo_wr32(dev_priv->fence.bo, sema->mem->start / 4, 0);

	return sema;
	fail:
	kfree(sema);
	return NULL;
	}

	static void
	free_semaphore(struct kref *ref)
	{
	struct nouveau_semaphore *sema =
	container_of(ref, struct nouveau_semaphore, ref);
	struct drm_nouveau_private *dev_priv = sema->dev->dev_private;

	spin_lock(&dev_priv->fence.lock);
	drm_mm_put_block(sema->mem);
	spin_unlock(&dev_priv->fence.lock);

	kfree(sema);
	}

	static void
	semaphore_work(void *priv, bool signalled)
	{
	struct nouveau_semaphore *sema = priv;
	struct drm_nouveau_private *dev_priv = sema->dev->dev_private;

	if (unlikely(!signalled))
	nouveau_bo_wr32(dev_priv->fence.bo, sema->mem->start / 4, 1);

	kref_put(&sema->ref, free_semaphore);
	}

	static int
	emit_semaphore(struct nouveau_channel *chan, int method,
	struct nouveau_semaphore *sema)
	{
	struct drm_nouveau_private *dev_priv = sema->dev->dev_private;
	struct nouveau_fence *fence;
	bool smart = (dev_priv->card_type >= NV_50);
	int ret;

	ret = RING_SPACE(chan, smart ? 8 : 4);
	if (ret)
	return ret;

	if (smart) {
	BEGIN_RING(chan, NvSubSw, NV_SW_DMA_SEMAPHORE, 1);
	OUT_RING(chan, NvSema);
	}
	BEGIN_RING(chan, NvSubSw, NV_SW_SEMAPHORE_OFFSET, 1);
	OUT_RING(chan, sema->mem->start);

	if (smart && method == NV_SW_SEMAPHORE_ACQUIRE) {
	/*
	* NV50 tries to be too smart and context-switch
	* between semaphores instead of doing a "first come,
	* first served" strategy like previous cards
	* do.
	*
	* That's bad because the ACQUIRE latency can get as
	* large as the PFIFO context time slice in the
	* typical DRI2 case where you have several
	* outstanding semaphores at the same moment.
	*
	* If we're going to ACQUIRE, force the card to
	* context switch before, just in case the matching
	* RELEASE is already scheduled to be executed in
	* another channel.
	*/
	BEGIN_RING(chan, NvSubSw, NV_SW_YIELD, 1);
	OUT_RING(chan, 0);
	}

	BEGIN_RING(chan, NvSubSw, method, 1);
	OUT_RING(chan, 1);

	if (smart && method == NV_SW_SEMAPHORE_RELEASE) {
	/*
	* Force the card to context switch, there may be
	* another channel waiting for the semaphore we just
	* released.
	*/
	BEGIN_RING(chan, NvSubSw, NV_SW_YIELD, 1);
	OUT_RING(chan, 0);
	}

	/* Delay semaphore destruction until its work is done */
	ret = nouveau_fence_new(chan, &fence, true);
	if (ret)
	return ret;

	kref_get(&sema->ref);
	nouveau_fence_work(fence, semaphore_work, sema);
	nouveau_fence_unref((void *)&fence);

	return 0;
	}

	int
	nouveau_fence_sync(struct nouveau_fence *fence,
	struct nouveau_channel *wchan)
	{
	struct nouveau_channel *chan = nouveau_fence_channel(fence);
	struct drm_device *dev = wchan->dev;
	struct nouveau_semaphore *sema;
	int ret;

	if (likely(!fence \|\| chan == wchan \|\|
	nouveau_fence_signalled(fence, NULL)))
	return 0;

	sema = alloc_semaphore(dev);
	if (!sema) {
	/* Early card or broken userspace, fall back to
	* software sync. */
	return nouveau_fence_wait(fence, NULL, false, false);
	}

	/* Make wchan wait until it gets signalled */
	ret = emit_semaphore(wchan, NV_SW_SEMAPHORE_ACQUIRE, sema);
	if (ret)
	goto out;

	/* Signal the semaphore from chan */
	ret = emit_semaphore(chan, NV_SW_SEMAPHORE_RELEASE, sema);
	out:
	kref_put(&sema->ref, free_semaphore);
	return ret;
	}

	int
	nouveau_fence_flush(void sync_obj, void sync_arg)
	{
	return 0;
	}

	int
	nouveau_fence_channel_init(struct nouveau_channel *chan)
	{
	struct drm_device *dev = chan->dev;
	struct drm_nouveau_private *dev_priv = dev->dev_private;
	struct nouveau_gpuobj *obj = NULL;
	int ret;

	/* Create an NV_SW object for various sync purposes */
	ret = nouveau_gpuobj_sw_new(chan, NV_SW, &obj);
	if (ret)
	return ret;

	ret = nouveau_ramht_insert(chan, NvSw, obj);
	nouveau_gpuobj_ref(NULL, &obj);
	if (ret)
	return ret;

	ret = RING_SPACE(chan, 2);
	if (ret)
	return ret;
	BEGIN_RING(chan, NvSubSw, 0, 1);
	OUT_RING(chan, NvSw);

	/* Create a DMA object for the shared cross-channel sync area. */
	if (USE_SEMA(dev)) {
	struct drm_mm_node *mem = dev_priv->fence.bo->bo.mem.mm_node;

	ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY,
	mem->start << PAGE_SHIFT,
	mem->size << PAGE_SHIFT,
	NV_DMA_ACCESS_RW,
	NV_DMA_TARGET_VIDMEM, &obj);
	if (ret)
	return ret;

	ret = nouveau_ramht_insert(chan, NvSema, obj);
	nouveau_gpuobj_ref(NULL, &obj);
	if (ret)
	return ret;

	ret = RING_SPACE(chan, 2);
	if (ret)
	return ret;
	BEGIN_RING(chan, NvSubSw, NV_SW_DMA_SEMAPHORE, 1);
	OUT_RING(chan, NvSema);
	}

	FIRE_RING(chan);

	INIT_LIST_HEAD(&chan->fence.pending);
	spin_lock_init(&chan->fence.lock);
	atomic_set(&chan->fence.last_sequence_irq, 0);

	return 0;
	}

	void
	nouveau_fence_channel_fini(struct nouveau_channel *chan)
	{
	struct nouveau_fence tmp, fence;

	list_for_each_entry_safe(fence, tmp, &chan->fence.pending, entry) {
	fence->signalled = true;
	list_del(&fence->entry);

	if (unlikely(fence->work))
	fence->work(fence->priv, false);

	kref_put(&fence->refcount, nouveau_fence_del);
	}
	}

	int
	nouveau_fence_init(struct drm_device *dev)
	{
	struct drm_nouveau_private *dev_priv = dev->dev_private;
	int ret;

	/* Create a shared VRAM heap for cross-channel sync. */
	if (USE_SEMA(dev)) {
	ret = nouveau_bo_new(dev, NULL, 4096, 0, TTM_PL_FLAG_VRAM,
	0, 0, false, true, &dev_priv->fence.bo);
	if (ret)
	return ret;

	ret = nouveau_bo_pin(dev_priv->fence.bo, TTM_PL_FLAG_VRAM);
	if (ret)
	goto fail;

	ret = nouveau_bo_map(dev_priv->fence.bo);
	if (ret)
	goto fail;

	ret = drm_mm_init(&dev_priv->fence.heap, 0,
	dev_priv->fence.bo->bo.mem.size);
	if (ret)
	goto fail;

	spin_lock_init(&dev_priv->fence.lock);
	}

	return 0;
	fail:
	nouveau_bo_unmap(dev_priv->fence.bo);
	nouveau_bo_ref(NULL, &dev_priv->fence.bo);
	return ret;
	}

	void
	nouveau_fence_fini(struct drm_device *dev)
	{
	struct drm_nouveau_private *dev_priv = dev->dev_private;

	if (USE_SEMA(dev)) {
	drm_mm_takedown(&dev_priv->fence.heap);
	nouveau_bo_unmap(dev_priv->fence.bo);
	nouveau_bo_unpin(dev_priv->fence.bo);
	nouveau_bo_ref(NULL, &dev_priv->fence.bo);
	}
	}