drm-next 98faa78ce7f1f986e11e7805d31b409782a6d2d4

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

13 files changed, 547 insertions, 396 deletions

diff --git a/drm/nouveau_bo.c b/drm/nouveau_bo.c
index 01da50862..eea74b127 100644
--- a/drm/nouveau_bo.c
+++ b/drm/nouveau_bo.c
@@ -88,13 +88,13 @@ nv10_bo_get_tile_region(struct drm_device *dev, int i)
 
 static void
 nv10_bo_put_tile_region(struct drm_device *dev, struct nouveau_drm_tile *tile,
-			struct nouveau_fence *fence)
+			struct fence *fence)
 {
 	struct nouveau_drm *drm = nouveau_drm(dev);
 
 	if (tile) {
 		spin_lock(&drm->tile.lock);
-		tile->fence = nouveau_fence_ref(fence);
+		tile->fence = (struct nouveau_fence *)fence_get(fence);
 		tile->used = false;
 		spin_unlock(&drm->tile.lock);
 	}
@@ -241,16 +241,16 @@ nouveau_bo_new(struct drm_device *dev, int size, int align,
 }
 
 static void
-set_placement_list(uint32_t *pl, unsigned *n, uint32_t type, uint32_t flags)
+set_placement_list(struct ttm_place *pl, unsigned *n, uint32_t type, uint32_t flags)
 {
 	*n = 0;
 
 	if (type & TTM_PL_FLAG_VRAM)
-		pl[(*n)++] = TTM_PL_FLAG_VRAM | flags;
+		pl[(*n)++].flags = TTM_PL_FLAG_VRAM | flags;
 	if (type & TTM_PL_FLAG_TT)
-		pl[(*n)++] = TTM_PL_FLAG_TT | flags;
+		pl[(*n)++].flags = TTM_PL_FLAG_TT | flags;
 	if (type & TTM_PL_FLAG_SYSTEM)
-		pl[(*n)++] = TTM_PL_FLAG_SYSTEM | flags;
+		pl[(*n)++].flags = TTM_PL_FLAG_SYSTEM | flags;
 }
 
 static void
@@ -258,6 +258,7 @@ set_placement_range(struct nouveau_bo *nvbo, uint32_t type)
 {
 	struct nouveau_drm *drm = nouveau_bdev(nvbo->bo.bdev);
 	u32 vram_pages = drm->device.info.ram_size >> PAGE_SHIFT;
+	unsigned i, fpfn, lpfn;
 
 	if (drm->device.info.family == NV_DEVICE_INFO_V0_CELSIUS &&
 	    nvbo->tile_mode && (type & TTM_PL_FLAG_VRAM) &&
@@ -269,11 +270,19 @@ set_placement_range(struct nouveau_bo *nvbo, uint32_t type)
 		 * at the same time.
 		 */
 		if (nvbo->tile_flags & NOUVEAU_GEM_TILE_ZETA) {
-			nvbo->placement.fpfn = vram_pages / 2;
-			nvbo->placement.lpfn = ~0;
+			fpfn = vram_pages / 2;
+			lpfn = ~0;
 		} else {
-			nvbo->placement.fpfn = 0;
-			nvbo->placement.lpfn = vram_pages / 2;
+			fpfn = 0;
+			lpfn = vram_pages / 2;
+		}
+		for (i = 0; i < nvbo->placement.num_placement; ++i) {
+			nvbo->placements[i].fpfn = fpfn;
+			nvbo->placements[i].lpfn = lpfn;
+		}
+		for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
+			nvbo->busy_placements[i].fpfn = fpfn;
+			nvbo->busy_placements[i].lpfn = lpfn;
 		}
 	}
 }
@@ -961,13 +970,14 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict, bool intr,
 	}
 
 	mutex_lock_nested(&cli->mutex, SINGLE_DEPTH_NESTING);
-	ret = nouveau_fence_sync(bo->sync_obj, chan);
+	ret = nouveau_fence_sync(nouveau_bo(bo), chan, true);
 	if (ret == 0) {
 		ret = drm->ttm.move(chan, bo, &bo->mem, new_mem);
 		if (ret == 0) {
 			ret = nouveau_fence_new(chan, false, &fence);
 			if (ret == 0) {
-				ret = ttm_bo_move_accel_cleanup(bo, fence,
+				ret = ttm_bo_move_accel_cleanup(bo,
+								&fence->base,
 								evict,
 								no_wait_gpu,
 								new_mem);
@@ -1041,12 +1051,15 @@ static int
 nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict, bool intr,
 		      bool no_wait_gpu, struct ttm_mem_reg *new_mem)
 {
-	u32 placement_memtype = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING;
+	struct ttm_place placement_memtype = {
+		.fpfn = 0,
+		.lpfn = 0,
+		.flags = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING
+	};
 	struct ttm_placement placement;
 	struct ttm_mem_reg tmp_mem;
 	int ret;
 
-	placement.fpfn = placement.lpfn = 0;
 	placement.num_placement = placement.num_busy_placement = 1;
 	placement.placement = placement.busy_placement = &placement_memtype;
 
@@ -1074,12 +1087,15 @@ static int
 nouveau_bo_move_flips(struct ttm_buffer_object *bo, bool evict, bool intr,
 		      bool no_wait_gpu, struct ttm_mem_reg *new_mem)
 {
-	u32 placement_memtype = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING;
+	struct ttm_place placement_memtype = {
+		.fpfn = 0,
+		.lpfn = 0,
+		.flags = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING
+	};
 	struct ttm_placement placement;
 	struct ttm_mem_reg tmp_mem;
 	int ret;
 
-	placement.fpfn = placement.lpfn = 0;
 	placement.num_placement = placement.num_busy_placement = 1;
 	placement.placement = placement.busy_placement = &placement_memtype;
 
@@ -1152,8 +1168,9 @@ nouveau_bo_vm_cleanup(struct ttm_buffer_object *bo,
 {
 	struct nouveau_drm *drm = nouveau_bdev(bo->bdev);
 	struct drm_device *dev = drm->dev;
+	struct fence *fence = reservation_object_get_excl(bo->resv);
 
-	nv10_bo_put_tile_region(dev, *old_tile, bo->sync_obj);
+	nv10_bo_put_tile_region(dev, *old_tile, fence);
 	*old_tile = new_tile;
 }
 
@@ -1197,9 +1214,7 @@ nouveau_bo_move(struct ttm_buffer_object *bo, bool evict, bool intr,
 	}
 
 	/* Fallback to software copy. */
-	spin_lock(&bo->bdev->fence_lock);
 	ret = ttm_bo_wait(bo, true, intr, no_wait_gpu);
-	spin_unlock(&bo->bdev->fence_lock);
 	if (ret == 0)
 		ret = ttm_bo_move_memcpy(bo, evict, no_wait_gpu, new_mem);
 
@@ -1294,7 +1309,7 @@ nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
 	struct nouveau_bo *nvbo = nouveau_bo(bo);
 	struct nvif_device *device = &drm->device;
 	u32 mappable = nv_device_resource_len(nvkm_device(device), 1) >> PAGE_SHIFT;
-	int ret;
+	int i, ret;
 
 	/* as long as the bo isn't in vram, and isn't tiled, we've got
 	 * nothing to do here.
@@ -1319,9 +1334,16 @@ nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
 	    bo->mem.start + bo->mem.num_pages < mappable)
 		return 0;
 
+	for (i = 0; i < nvbo->placement.num_placement; ++i) {
+		nvbo->placements[i].fpfn = 0;
+		nvbo->placements[i].lpfn = mappable;
+	}
+
+	for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
+		nvbo->busy_placements[i].fpfn = 0;
+		nvbo->busy_placements[i].lpfn = mappable;
+	}
 
-	nvbo->placement.fpfn = 0;
-	nvbo->placement.lpfn = mappable;
 	nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_VRAM, 0);
 	return nouveau_bo_validate(nvbo, false, false);
 }
@@ -1436,47 +1458,14 @@ nouveau_ttm_tt_unpopulate(struct ttm_tt *ttm)
 }
 
 void
-nouveau_bo_fence(struct nouveau_bo *nvbo, struct nouveau_fence *fence)
-{
-	struct nouveau_fence *new_fence = nouveau_fence_ref(fence);
-	struct nouveau_fence *old_fence = NULL;
-
-	spin_lock(&nvbo->bo.bdev->fence_lock);
-	old_fence = nvbo->bo.sync_obj;
-	nvbo->bo.sync_obj = new_fence;
-	spin_unlock(&nvbo->bo.bdev->fence_lock);
-
-	nouveau_fence_unref(&old_fence);
-}
-
-static void
-nouveau_bo_fence_unref(void **sync_obj)
-{
-	nouveau_fence_unref((struct nouveau_fence **)sync_obj);
-}
-
-static void *
-nouveau_bo_fence_ref(void *sync_obj)
+nouveau_bo_fence(struct nouveau_bo *nvbo, struct nouveau_fence *fence, bool exclusive)
 {
-	return nouveau_fence_ref(sync_obj);
-}
+	struct reservation_object *resv = nvbo->bo.resv;
 
-static bool
-nouveau_bo_fence_signalled(void *sync_obj)
-{
-	return nouveau_fence_done(sync_obj);
-}
-
-static int
-nouveau_bo_fence_wait(void *sync_obj, bool lazy, bool intr)
-{
-	return nouveau_fence_wait(sync_obj, lazy, intr);
-}
-
-static int
-nouveau_bo_fence_flush(void *sync_obj)
-{
-	return 0;
+	if (exclusive)
+		reservation_object_add_excl_fence(resv, &fence->base);
+	else if (fence)
+		reservation_object_add_shared_fence(resv, &fence->base);
 }
 
 struct ttm_bo_driver nouveau_bo_driver = {
@@ -1489,11 +1478,6 @@ struct ttm_bo_driver nouveau_bo_driver = {
 	.move_notify = nouveau_bo_move_ntfy,
 	.move = nouveau_bo_move,
 	.verify_access = nouveau_bo_verify_access,
-	.sync_obj_signaled = nouveau_bo_fence_signalled,
-	.sync_obj_wait = nouveau_bo_fence_wait,
-	.sync_obj_flush = nouveau_bo_fence_flush,
-	.sync_obj_unref = nouveau_bo_fence_unref,
-	.sync_obj_ref = nouveau_bo_fence_ref,
 	.fault_reserve_notify = &nouveau_ttm_fault_reserve_notify,
 	.io_mem_reserve = &nouveau_ttm_io_mem_reserve,
 	.io_mem_free = &nouveau_ttm_io_mem_free,
diff --git a/drm/nouveau_bo.h b/drm/nouveau_bo.h
index ff17c1f43..ae95b2d43 100644
--- a/drm/nouveau_bo.h
+++ b/drm/nouveau_bo.h
@@ -9,8 +9,8 @@ struct nouveau_bo {
 	struct ttm_buffer_object bo;
 	struct ttm_placement placement;
 	u32 valid_domains;
-	u32 placements[3];
-	u32 busy_placements[3];
+	struct ttm_place placements[3];
+	struct ttm_place busy_placements[3];
 	struct ttm_bo_kmap_obj kmap;
 	struct list_head head;
 
@@ -78,7 +78,7 @@ u16  nouveau_bo_rd16(struct nouveau_bo *, unsigned index);
 void nouveau_bo_wr16(struct nouveau_bo *, unsigned index, u16 val);
 u32  nouveau_bo_rd32(struct nouveau_bo *, unsigned index);
 void nouveau_bo_wr32(struct nouveau_bo *, unsigned index, u32 val);
-void nouveau_bo_fence(struct nouveau_bo *, struct nouveau_fence *);
+void nouveau_bo_fence(struct nouveau_bo *, struct nouveau_fence *, bool exclusive);
 int  nouveau_bo_validate(struct nouveau_bo *, bool interruptible,
 			 bool no_wait_gpu);
 
diff --git a/drm/nouveau_display.c b/drm/nouveau_display.c
index 4a21b2b06..2bd5ebe73 100644
--- a/drm/nouveau_display.c
+++ b/drm/nouveau_display.c
@@ -126,7 +126,7 @@ nouveau_display_scanoutpos_head(struct drm_crtc *crtc, int *vpos, int *hpos,
 	if (etime) *etime = ns_to_ktime(args.scan.time[1]);
 
 	if (*vpos < 0)
-		ret |= DRM_SCANOUTPOS_INVBL;
+		ret |= DRM_SCANOUTPOS_IN_VBLANK;
 	return ret;
 }
 
@@ -657,7 +657,7 @@ nouveau_page_flip_emit(struct nouveau_channel *chan,
 	spin_unlock_irqrestore(&dev->event_lock, flags);
 
 	/* Synchronize with the old framebuffer */
-	ret = nouveau_fence_sync(old_bo->bo.sync_obj, chan);
+	ret = nouveau_fence_sync(old_bo, chan, false);
 	if (ret)
 		goto fail;
 
@@ -716,19 +716,24 @@ nouveau_crtc_page_flip(struct drm_crtc *crtc, struct drm_framebuffer *fb,
 	}
 
 	mutex_lock(&cli->mutex);
-
-	/* synchronise rendering channel with the kernel's channel */
-	spin_lock(&new_bo->bo.bdev->fence_lock);
-	fence = nouveau_fence_ref(new_bo->bo.sync_obj);
-	spin_unlock(&new_bo->bo.bdev->fence_lock);
-	ret = nouveau_fence_sync(fence, chan);
-	nouveau_fence_unref(&fence);
+	ret = ttm_bo_reserve(&new_bo->bo, true, false, false, NULL);
 	if (ret)
 		goto fail_unpin;
 
-	ret = ttm_bo_reserve(&old_bo->bo, true, false, false, NULL);
-	if (ret)
+	/* synchronise rendering channel with the kernel's channel */
+	ret = nouveau_fence_sync(new_bo, chan, false);
+	if (ret) {
+		ttm_bo_unreserve(&new_bo->bo);
 		goto fail_unpin;
+	}
+
+	if (new_bo != old_bo) {
+		ttm_bo_unreserve(&new_bo->bo);
+
+		ret = ttm_bo_reserve(&old_bo->bo, true, false, false, NULL);
+		if (ret)
+			goto fail_unpin;
+	}
 
 	/* Initialize a page flip struct */
 	*s = (struct nouveau_page_flip_state)
@@ -774,7 +779,7 @@ nouveau_crtc_page_flip(struct drm_crtc *crtc, struct drm_framebuffer *fb,
 	/* Update the crtc struct and cleanup */
 	crtc->primary->fb = fb;
 
-	nouveau_bo_fence(old_bo, fence);
+	nouveau_bo_fence(old_bo, fence, false);
 	ttm_bo_unreserve(&old_bo->bo);
 	if (old_bo != new_bo)
 		nouveau_bo_unpin(old_bo);
diff --git a/drm/nouveau_drm.c b/drm/nouveau_drm.c
index 3ed32dd90..64405d79c 100644
--- a/drm/nouveau_drm.c
+++ b/drm/nouveau_drm.c
@@ -73,7 +73,9 @@ MODULE_PARM_DESC(runpm, "disable (0), force enable (1), optimus only default (-1
 int nouveau_runtime_pm = -1;
 module_param_named(runpm, nouveau_runtime_pm, int, 0400);
 
-static struct drm_driver driver;
+static struct drm_driver driver_stub;
+static struct drm_driver driver_pci;
+static struct drm_driver driver_platform;
 
 static u64
 nouveau_pci_name(struct pci_dev *pdev)
@@ -322,7 +324,7 @@ static int nouveau_drm_probe(struct pci_dev *pdev,
 
 	pci_set_master(pdev);
 
-	ret = drm_get_pci_dev(pdev, pent, &driver);
+	ret = drm_get_pci_dev(pdev, pent, &driver_pci);
 	if (ret) {
 		nouveau_object_ref(NULL, (struct nouveau_object **)&device);
 		return ret;
@@ -831,7 +833,7 @@ nouveau_driver_fops = {
 };
 
 static struct drm_driver
-driver = {
+driver_stub = {
 	.driver_features =
 		DRIVER_USE_AGP |
 		DRIVER_GEM | DRIVER_MODESET | DRIVER_PRIME | DRIVER_RENDER,
@@ -1037,7 +1039,7 @@ nouveau_platform_device_create_(struct platform_device *pdev, int size,
 	if (err)
 		return ERR_PTR(err);
 
-	drm = drm_dev_alloc(&driver, &pdev->dev);
+	drm = drm_dev_alloc(&driver_platform, &pdev->dev);
 	if (!drm) {
 		err = -ENOMEM;
 		goto err_free;
@@ -1062,6 +1064,11 @@ EXPORT_SYMBOL(nouveau_platform_device_create_);
 static int __init
 nouveau_drm_init(void)
 {
+	driver_pci = driver_stub;
+	driver_pci.set_busid = drm_pci_set_busid;
+	driver_platform = driver_stub;
+	driver_platform.set_busid = drm_platform_set_busid;
+
 	if (nouveau_modeset == -1) {
 #ifdef CONFIG_VGA_CONSOLE
 		if (vgacon_text_force())
@@ -1073,7 +1080,7 @@ nouveau_drm_init(void)
 		return 0;
 
 	nouveau_register_dsm_handler();
-	return drm_pci_init(&driver, &nouveau_drm_pci_driver);
+	return drm_pci_init(&driver_pci, &nouveau_drm_pci_driver);
 }
 
 static void __exit
@@ -1082,7 +1089,7 @@ nouveau_drm_exit(void)
 	if (!nouveau_modeset)
 		return;
 
-	drm_pci_exit(&driver, &nouveau_drm_pci_driver);
+	drm_pci_exit(&driver_pci, &nouveau_drm_pci_driver);
 	nouveau_unregister_dsm_handler();
 }
 
diff --git a/drm/nouveau_fence.c b/drm/nouveau_fence.c
index 0a9311415..decfe6c4a 100644
--- a/drm/nouveau_fence.c
+++ b/drm/nouveau_fence.c
@@ -28,6 +28,7 @@
 
 #include <linux/ktime.h>
 #include <linux/hrtimer.h>
+#include <trace/events/fence.h>
 
 #include <nvif/notify.h>
 #include <nvif/event.h>
@@ -36,123 +37,210 @@
 #include "nouveau_dma.h"
 #include "nouveau_fence.h"
 
-struct fence_work {
-	struct work_struct base;
-	struct list_head head;
-	void (*func)(void *);
-	void *data;
-};
+static const struct fence_ops nouveau_fence_ops_uevent;
+static const struct fence_ops nouveau_fence_ops_legacy;
+
+static inline struct nouveau_fence *
+from_fence(struct fence *fence)
+{
+	return container_of(fence, struct nouveau_fence, base);
+}
+
+static inline struct nouveau_fence_chan *
+nouveau_fctx(struct nouveau_fence *fence)
+{
+	return container_of(fence->base.lock, struct nouveau_fence_chan, lock);
+}
 
 static void
 nouveau_fence_signal(struct nouveau_fence *fence)
 {
-	struct fence_work *work, *temp;
+	fence_signal_locked(&fence->base);
+	list_del(&fence->head);
+
+	if (test_bit(FENCE_FLAG_USER_BITS, &fence->base.flags)) {
+		struct nouveau_fence_chan *fctx = nouveau_fctx(fence);
 
-	list_for_each_entry_safe(work, temp, &fence->work, head) {
-		schedule_work(&work->base);
-		list_del(&work->head);
+		if (!--fctx->notify_ref)
+			nvif_notify_put(&fctx->notify);
 	}
 
-	fence->channel = NULL;
-	list_del(&fence->head);
+	fence_put(&fence->base);
+}
+
+static struct nouveau_fence *
+nouveau_local_fence(struct fence *fence, struct nouveau_drm *drm) {
+	struct nouveau_fence_priv *priv = (void*)drm->fence;
+
+	if (fence->ops != &nouveau_fence_ops_legacy &&
+	    fence->ops != &nouveau_fence_ops_uevent)
+		return NULL;
+
+	if (fence->context < priv->context_base ||
+	    fence->context >= priv->context_base + priv->contexts)
+		return NULL;
+
+	return from_fence(fence);
 }
 
 void
 nouveau_fence_context_del(struct nouveau_fence_chan *fctx)
 {
-	struct nouveau_fence *fence, *fnext;
-	spin_lock(&fctx->lock);
-	list_for_each_entry_safe(fence, fnext, &fctx->pending, head) {
+	struct nouveau_fence *fence;
+
+	nvif_notify_fini(&fctx->notify);
+
+	spin_lock_irq(&fctx->lock);
+	while (!list_empty(&fctx->pending)) {
+		fence = list_entry(fctx->pending.next, typeof(*fence), head);
+
+		nouveau_fence_signal(fence);
+		fence->channel = NULL;
+	}
+	spin_unlock_irq(&fctx->lock);
+}
+
+static void
+nouveau_fence_update(struct nouveau_channel *chan, struct nouveau_fence_chan *fctx)
+{
+	struct nouveau_fence *fence;
+
+	u32 seq = fctx->read(chan);
+
+	while (!list_empty(&fctx->pending)) {
+		fence = list_entry(fctx->pending.next, typeof(*fence), head);
+
+		if ((int)(seq - fence->base.seqno) < 0)
+			return;
+
 		nouveau_fence_signal(fence);
 	}
-	spin_unlock(&fctx->lock);
+}
+
+static int
+nouveau_fence_wait_uevent_handler(struct nvif_notify *notify)
+{
+	struct nouveau_fence_chan *fctx =
+		container_of(notify, typeof(*fctx), notify);
+	unsigned long flags;
+
+	spin_lock_irqsave(&fctx->lock, flags);
+	if (!list_empty(&fctx->pending)) {
+		struct nouveau_fence *fence;
+
+		fence = list_entry(fctx->pending.next, typeof(*fence), head);
+		nouveau_fence_update(fence->channel, fctx);
+	}
+	spin_unlock_irqrestore(&fctx->lock, flags);
+
+	/* Always return keep here. NVIF refcount is handled with nouveau_fence_update */
+	return NVIF_NOTIFY_KEEP;
 }
 
 void
-nouveau_fence_context_new(struct nouveau_fence_chan *fctx)
+nouveau_fence_context_new(struct nouveau_channel *chan, struct nouveau_fence_chan *fctx)
 {
+	struct nouveau_fence_priv *priv = (void*)chan->drm->fence;
+	int ret;
+
 	INIT_LIST_HEAD(&fctx->flip);
 	INIT_LIST_HEAD(&fctx->pending);
 	spin_lock_init(&fctx->lock);
+	fctx->context = priv->context_base + chan->chid;
+
+	if (!priv->uevent)
+		return;
+
+	ret = nvif_notify_init(chan->object, NULL,
+			 nouveau_fence_wait_uevent_handler, false,
+			 G82_CHANNEL_DMA_V0_NTFY_UEVENT,
+			 &(struct nvif_notify_uevent_req) { },
+			 sizeof(struct nvif_notify_uevent_req),
+			 sizeof(struct nvif_notify_uevent_rep),
+			 &fctx->notify);
+
+	WARN_ON(ret);
 }
 
+struct nouveau_fence_work {
+	struct work_struct work;
+	struct fence_cb cb;
+	void (*func)(void *);
+	void *data;
+};
+
 static void
 nouveau_fence_work_handler(struct work_struct *kwork)
 {
-	struct fence_work *work = container_of(kwork, typeof(*work), base);
+	struct nouveau_fence_work *work = container_of(kwork, typeof(*work), work);
 	work->func(work->data);
 	kfree(work);
 }
 
+static void nouveau_fence_work_cb(struct fence *fence, struct fence_cb *cb)
+{
+	struct nouveau_fence_work *work = container_of(cb, typeof(*work), cb);
+
+	schedule_work(&work->work);
+}
+
 void
-nouveau_fence_work(struct nouveau_fence *fence,
+nouveau_fence_work(struct fence *fence,
 		   void (*func)(void *), void *data)
 {
-	struct nouveau_channel *chan = fence->channel;
-	struct nouveau_fence_chan *fctx;
-	struct fence_work *work = NULL;
+	struct nouveau_fence_work *work;
 
-	if (nouveau_fence_done(fence)) {
-		func(data);
-		return;
-	}
+	if (fence_is_signaled(fence))
+		goto err;
 
-	fctx = chan->fence;
 	work = kmalloc(sizeof(*work), GFP_KERNEL);
 	if (!work) {
-		WARN_ON(nouveau_fence_wait(fence, false, false));
-		func(data);
-		return;
+		WARN_ON(nouveau_fence_wait((struct nouveau_fence *)fence,
+					   false, false));
+		goto err;
 	}
 
-	spin_lock(&fctx->lock);
-	if (!fence->channel) {
-		spin_unlock(&fctx->lock);
-		kfree(work);
-		func(data);
-		return;
-	}
-
-	INIT_WORK(&work->base, nouveau_fence_work_handler);
+	INIT_WORK(&work->work, nouveau_fence_work_handler);
 	work->func = func;
 	work->data = data;
-	list_add(&work->head, &fence->work);
-	spin_unlock(&fctx->lock);
-}
-
-static void
-nouveau_fence_update(struct nouveau_channel *chan)
-{
-	struct nouveau_fence_chan *fctx = chan->fence;
-	struct nouveau_fence *fence, *fnext;
 
-	spin_lock(&fctx->lock);
-	list_for_each_entry_safe(fence, fnext, &fctx->pending, head) {
-		if (fctx->read(chan) < fence->sequence)
-			break;
+	if (fence_add_callback(fence, &work->cb, nouveau_fence_work_cb) < 0)
+		goto err_free;
+	return;
 
-		nouveau_fence_signal(fence);
-		nouveau_fence_unref(&fence);
-	}
-	spin_unlock(&fctx->lock);
+err_free:
+	kfree(work);
+err:
+	func(data);
 }
 
 int
 nouveau_fence_emit(struct nouveau_fence *fence, struct nouveau_channel *chan)
 {
 	struct nouveau_fence_chan *fctx = chan->fence;
+	struct nouveau_fence_priv *priv = (void*)chan->drm->fence;
 	int ret;
 
 	fence->channel  = chan;
 	fence->timeout  = jiffies + (15 * HZ);
-	fence->sequence = ++fctx->sequence;
 
+	if (priv->uevent)
+		fence_init(&fence->base, &nouveau_fence_ops_uevent,
+			   &fctx->lock,
+			   priv->context_base + chan->chid, ++fctx->sequence);
+	else
+		fence_init(&fence->base, &nouveau_fence_ops_legacy,
+			   &fctx->lock,
+			   priv->context_base + chan->chid, ++fctx->sequence);
+
+	trace_fence_emit(&fence->base);
 	ret = fctx->emit(fence);
 	if (!ret) {
-		kref_get(&fence->kref);
-		spin_lock(&fctx->lock);
+		fence_get(&fence->base);
+		spin_lock_irq(&fctx->lock);
+		nouveau_fence_update(chan, fctx);
 		list_add_tail(&fence->head, &fctx->pending);
-		spin_unlock(&fctx->lock);
+		spin_unlock_irq(&fctx->lock);
 	}
 
 	return ret;
@@ -161,114 +249,70 @@ nouveau_fence_emit(struct nouveau_fence *fence, struct nouveau_channel *chan)
 bool
 nouveau_fence_done(struct nouveau_fence *fence)
 {
-	if (fence->channel)
-		nouveau_fence_update(fence->channel);
-	return !fence->channel;
-}
+	if (fence->base.ops == &nouveau_fence_ops_legacy ||
+	    fence->base.ops == &nouveau_fence_ops_uevent) {
+		struct nouveau_fence_chan *fctx = nouveau_fctx(fence);
+		unsigned long flags;
 
-struct nouveau_fence_wait {
-	struct nouveau_fence_priv *priv;
-	struct nvif_notify notify;
-};
+		if (test_bit(FENCE_FLAG_SIGNALED_BIT, &fence->base.flags))
+			return true;
 
-static int
-nouveau_fence_wait_uevent_handler(struct nvif_notify *notify)
-{
-	struct nouveau_fence_wait *wait =
-		container_of(notify, typeof(*wait), notify);
-	wake_up_all(&wait->priv->waiting);
-	return NVIF_NOTIFY_KEEP;
+		spin_lock_irqsave(&fctx->lock, flags);
+		nouveau_fence_update(fence->channel, fctx);
+		spin_unlock_irqrestore(&fctx->lock, flags);
+	}
+	return fence_is_signaled(&fence->base);
 }
 
-static int
-nouveau_fence_wait_uevent(struct nouveau_fence *fence, bool intr)
-
+static long
+nouveau_fence_wait_legacy(struct fence *f, bool intr, long wait)
 {
-	struct nouveau_channel *chan = fence->channel;
-	struct nouveau_fence_priv *priv = chan->drm->fence;
-	struct nouveau_fence_wait wait = { .priv = priv };
-	int ret = 0;
+	struct nouveau_fence *fence = from_fence(f);
+	unsigned long sleep_time = NSEC_PER_MSEC / 1000;
+	unsigned long t = jiffies, timeout = t + wait;
 
-	ret = nvif_notify_init(chan->object, NULL,
-			       nouveau_fence_wait_uevent_handler, false,
-			       G82_CHANNEL_DMA_V0_NTFY_UEVENT,
-			       &(struct nvif_notify_uevent_req) {
-			       },
-			       sizeof(struct nvif_notify_uevent_req),
-			       sizeof(struct nvif_notify_uevent_rep),
-			       &wait.notify);
-	if (ret)
-		return ret;
+	while (!nouveau_fence_done(fence)) {
+		ktime_t kt;
 
-	nvif_notify_get(&wait.notify);
-
-	if (fence->timeout) {
-		unsigned long timeout = fence->timeout - jiffies;
-
-		if (time_before(jiffies, fence->timeout)) {
-			if (intr) {
-				ret = wait_event_interruptible_timeout(
-						priv->waiting,
-						nouveau_fence_done(fence),
-						timeout);
-			} else {
-				ret = wait_event_timeout(priv->waiting,
-						nouveau_fence_done(fence),
-						timeout);
-			}
-		}
+		t = jiffies;
 
-		if (ret >= 0) {
-			fence->timeout = jiffies + ret;
-			if (time_after_eq(jiffies, fence->timeout))
-				ret = -EBUSY;
-		}
-	} else {
-		if (intr) {
-			ret = wait_event_interruptible(priv->waiting,
-					nouveau_fence_done(fence));
-		} else {
-			wait_event(priv->waiting, nouveau_fence_done(fence));
+		if (wait != MAX_SCHEDULE_TIMEOUT && time_after_eq(t, timeout)) {
+			__set_current_state(TASK_RUNNING);
+			return 0;
 		}
+
+		__set_current_state(intr ? TASK_INTERRUPTIBLE :
+					   TASK_UNINTERRUPTIBLE);
+
+		kt = ktime_set(0, sleep_time);
+		schedule_hrtimeout(&kt, HRTIMER_MODE_REL);
+		sleep_time *= 2;
+		if (sleep_time > NSEC_PER_MSEC)
+			sleep_time = NSEC_PER_MSEC;
+
+		if (intr && signal_pending(current))
+			return -ERESTARTSYS;
 	}
 
-	nvif_notify_fini(&wait.notify);
-	if (unlikely(ret < 0))
-		return ret;
+	__set_current_state(TASK_RUNNING);
 
-	return 0;
+	return timeout - t;
 }
 
-int
-nouveau_fence_wait(struct nouveau_fence *fence, bool lazy, bool intr)
+static int
+nouveau_fence_wait_busy(struct nouveau_fence *fence, bool intr)
 {
-	struct nouveau_channel *chan = fence->channel;
-	struct nouveau_fence_priv *priv = chan ? chan->drm->fence : NULL;
-	unsigned long sleep_time = NSEC_PER_MSEC / 1000;
-	ktime_t t;
 	int ret = 0;
 
-	while (priv && priv->uevent && lazy && !nouveau_fence_done(fence)) {
-		ret = nouveau_fence_wait_uevent(fence, intr);
-		if (ret < 0)
-			return ret;
-	}
-
 	while (!nouveau_fence_done(fence)) {
-		if (fence->timeout && time_after_eq(jiffies, fence->timeout)) {
+		if (time_after_eq(jiffies, fence->timeout)) {
 			ret = -EBUSY;
 			break;
 		}
 
-		__set_current_state(intr ? TASK_INTERRUPTIBLE :
-					   TASK_UNINTERRUPTIBLE);
-		if (lazy) {
-			t = ktime_set(0, sleep_time);
-			schedule_hrtimeout(&t, HRTIMER_MODE_REL);
-			sleep_time *= 2;
-			if (sleep_time > NSEC_PER_MSEC)
-				sleep_time = NSEC_PER_MSEC;
-		}
+		__set_current_state(intr ?
+				    TASK_INTERRUPTIBLE :
+				    TASK_UNINTERRUPTIBLE);
 
 		if (intr && signal_pending(current)) {
 			ret = -ERESTARTSYS;
@@ -281,47 +325,86 @@ nouveau_fence_wait(struct nouveau_fence *fence, bool lazy, bool intr)
 }
 
 int
-nouveau_fence_sync(struct nouveau_fence *fence, struct nouveau_channel *chan)
+nouveau_fence_wait(struct nouveau_fence *fence, bool lazy, bool intr)
 {
-	struct nouveau_fence_chan *fctx = chan->fence;
-	struct nouveau_channel *prev;
-	int ret = 0;
+	long ret;
 
-	prev = fence ? fence->channel : NULL;
-	if (prev) {
-		if (unlikely(prev != chan && !nouveau_fence_done(fence))) {
-			ret = fctx->sync(fence, prev, chan);
-			if (unlikely(ret))
-				ret = nouveau_fence_wait(fence, true, false);
-		}
-	}
+	if (!lazy)
+		return nouveau_fence_wait_busy(fence, intr);
 
-	return ret;
+	ret = fence_wait_timeout(&fence->base, intr, 15 * HZ);
+	if (ret < 0)
+		return ret;
+	else if (!ret)
+		return -EBUSY;
+	else
+		return 0;
 }
 
-static void
-nouveau_fence_del(struct kref *kref)
+int
+nouveau_fence_sync(struct nouveau_bo *nvbo, struct nouveau_channel *chan, bool exclusive)
 {
-	struct nouveau_fence *fence = container_of(kref, typeof(*fence), kref);
-	kfree(fence);
+	struct nouveau_fence_chan *fctx = chan->fence;
+	struct fence *fence;
+	struct reservation_object *resv = nvbo->bo.resv;
+	struct reservation_object_list *fobj;
+	struct nouveau_fence *f;
+	int ret = 0, i;
+
+	if (!exclusive) {
+		ret = reservation_object_reserve_shared(resv);
+
+		if (ret)
+			return ret;
+	}
+
+	fobj = reservation_object_get_list(resv);
+	fence = reservation_object_get_excl(resv);
+
+	if (fence && (!exclusive || !fobj || !fobj->shared_count)) {
+		struct nouveau_channel *prev = NULL;
+
+		f = nouveau_local_fence(fence, chan->drm);
+		if (f)
+			prev = f->channel;
+
+		if (!prev || (prev != chan && (ret = fctx->sync(f, prev, chan))))
+			ret = fence_wait(fence, true);
+
+		return ret;
+	}
+
+	if (!exclusive || !fobj)
+		return ret;
+
+	for (i = 0; i < fobj->shared_count && !ret; ++i) {
+		struct nouveau_channel *prev = NULL;
+
+		fence = rcu_dereference_protected(fobj->shared[i],
+						reservation_object_held(resv));
+
+		f = nouveau_local_fence(fence, chan->drm);
+		if (f)
+			prev = f->channel;
+
+		if (!prev || (ret = fctx->sync(f, prev, chan)))
+			ret = fence_wait(fence, true);
+
+		if (ret)
+			break;
+	}
+
+	return ret;
 }
 
 void
 nouveau_fence_unref(struct nouveau_fence **pfence)
 {
 	if (*pfence)
-		kref_put(&(*pfence)->kref, nouveau_fence_del);
+		fence_put(&(*pfence)->base);
 	*pfence = NULL;
 }
 
-struct nouveau_fence *
-nouveau_fence_ref(struct nouveau_fence *fence)
-{
-	if (fence)
-		kref_get(&fence->kref);
-	return fence;
-}
-
 int
 nouveau_fence_new(struct nouveau_channel *chan, bool sysmem,
 		  struct nouveau_fence **pfence)
@@ -336,9 +419,7 @@ nouveau_fence_new(struct nouveau_channel *chan, bool sysmem,
 	if (!fence)
 		return -ENOMEM;
 
-	INIT_LIST_HEAD(&fence->work);
 	fence->sysmem = sysmem;
-	kref_init(&fence->kref);
 
 	ret = nouveau_fence_emit(fence, chan);
 	if (ret)
@@ -347,3 +428,92 @@ nouveau_fence_new(struct nouveau_channel *chan, bool sysmem,
 	*pfence = fence;
 	return ret;
 }
+
+static const char *nouveau_fence_get_get_driver_name(struct fence *fence)
+{
+	return "nouveau";
+}
+
+static const char *nouveau_fence_get_timeline_name(struct fence *f)
+{
+	struct nouveau_fence *fence = from_fence(f);
+	struct nouveau_fence_chan *fctx = nouveau_fctx(fence);
+
+	return fence->channel ? fctx->name : "dead channel";
+}
+
+/*
+ * In an ideal world, read would not assume the channel context is still alive.
+ * This function may be called from another device, running into free memory as a
+ * result. The drm node should still be there, so we can derive the index from
+ * the fence context.
+ */
+static bool nouveau_fence_is_signaled(struct fence *f)
+{
+	struct nouveau_fence *fence = from_fence(f);
+	struct nouveau_fence_chan *fctx = nouveau_fctx(fence);
+	struct nouveau_channel *chan = fence->channel;
+
+	return (int)(fctx->read(chan) - fence->base.seqno) >= 0;
+}
+
+static bool nouveau_fence_no_signaling(struct fence *f)
+{
+	struct nouveau_fence *fence = from_fence(f);
+
+	/*
+	 * caller should have a reference on the fence,
+	 * else fence could get freed here
+	 */
+	WARN_ON(atomic_read(&fence->base.refcount.refcount) <= 1);
+
+	/*
+	 * This needs uevents to work correctly, but fence_add_callback relies on
+	 * being able to enable signaling. It will still get signaled eventually,
+	 * just not right away.
+	 */
+	if (nouveau_fence_is_signaled(f)) {
+		list_del(&fence->head);
+
+		fence_put(&fence->base);
+		return false;
+	}
+
+	return true;
+}
+
+static const struct fence_ops nouveau_fence_ops_legacy = {
+	.get_driver_name = nouveau_fence_get_get_driver_name,
+	.get_timeline_name = nouveau_fence_get_timeline_name,
+	.enable_signaling = nouveau_fence_no_signaling,
+	.signaled = nouveau_fence_is_signaled,
+	.wait = nouveau_fence_wait_legacy,
+	.release = NULL
+};
+
+static bool nouveau_fence_enable_signaling(struct fence *f)
+{
+	struct nouveau_fence *fence = from_fence(f);
+	struct nouveau_fence_chan *fctx = nouveau_fctx(fence);
+	bool ret;
+
+	if (!fctx->notify_ref++)
+		nvif_notify_get(&fctx->notify);
+
+	ret = nouveau_fence_no_signaling(f);
+	if (ret)
+		set_bit(FENCE_FLAG_USER_BITS, &fence->base.flags);
+	else if (!--fctx->notify_ref)
+		nvif_notify_put(&fctx->notify);
+
+	return ret;
+}
+
+static const struct fence_ops nouveau_fence_ops_uevent = {
+	.get_driver_name = nouveau_fence_get_get_driver_name,
+	.get_timeline_name = nouveau_fence_get_timeline_name,
+	.enable_signaling = nouveau_fence_enable_signaling,
+	.signaled = nouveau_fence_is_signaled,
+	.wait = fence_default_wait,
+	.release = NULL
+};
diff --git a/drm/nouveau_fence.h b/drm/nouveau_fence.h
index c57bb61da..986c8135e 100644
--- a/drm/nouveau_fence.h
+++ b/drm/nouveau_fence.h
@@ -1,33 +1,35 @@
 #ifndef __NOUVEAU_FENCE_H__
 #define __NOUVEAU_FENCE_H__
 
+#include <linux/fence.h>
+#include <nvif/notify.h>
+
 struct nouveau_drm;
+struct nouveau_bo;
 
 struct nouveau_fence {
+	struct fence base;
+
 	struct list_head head;
-	struct list_head work;
-	struct kref kref;
 
 	bool sysmem;
 
 	struct nouveau_channel *channel;
 	unsigned long timeout;
-	u32 sequence;
 };
 
 int  nouveau_fence_new(struct nouveau_channel *, bool sysmem,
 		       struct nouveau_fence **);
-struct nouveau_fence *
-nouveau_fence_ref(struct nouveau_fence *);
 void nouveau_fence_unref(struct nouveau_fence **);
 
 int  nouveau_fence_emit(struct nouveau_fence *, struct nouveau_channel *);
 bool nouveau_fence_done(struct nouveau_fence *);
-void nouveau_fence_work(struct nouveau_fence *, void (*)(void *), void *);
+void nouveau_fence_work(struct fence *, void (*)(void *), void *);
 int  nouveau_fence_wait(struct nouveau_fence *, bool lazy, bool intr);
-int  nouveau_fence_sync(struct nouveau_fence *, struct nouveau_channel *);
+int  nouveau_fence_sync(struct nouveau_bo *, struct nouveau_channel *, bool exclusive);
 
 struct nouveau_fence_chan {
+	spinlock_t lock;
 	struct list_head pending;
 	struct list_head flip;
 
@@ -38,8 +40,12 @@ struct nouveau_fence_chan {
 	int  (*emit32)(struct nouveau_channel *, u64, u32);
 	int  (*sync32)(struct nouveau_channel *, u64, u32);
 
-	spinlock_t lock;
 	u32 sequence;
+	u32 context;
+	char name[24];
+
+	struct nvif_notify notify;
+	int notify_ref;
 };
 
 struct nouveau_fence_priv {
@@ -49,13 +55,13 @@ struct nouveau_fence_priv {
 	int  (*context_new)(struct nouveau_channel *);
 	void (*context_del)(struct nouveau_channel *);
 
-	wait_queue_head_t waiting;
+	u32 contexts, context_base;
 	bool uevent;
 };
 
 #define nouveau_fence(drm) ((struct nouveau_fence_priv *)(drm)->fence)
 
-void nouveau_fence_context_new(struct nouveau_fence_chan *);
+void nouveau_fence_context_new(struct nouveau_channel *, struct nouveau_fence_chan *);
 void nouveau_fence_context_del(struct nouveau_fence_chan *);
 
 int nv04_fence_create(struct nouveau_drm *);
diff --git a/drm/nouveau_gem.c b/drm/nouveau_gem.c
index 292a677bf..b7dbd1690 100644
--- a/drm/nouveau_gem.c
+++ b/drm/nouveau_gem.c
@@ -98,17 +98,23 @@ static void
 nouveau_gem_object_unmap(struct nouveau_bo *nvbo, struct nouveau_vma *vma)
 {
 	const bool mapped = nvbo->bo.mem.mem_type != TTM_PL_SYSTEM;
-	struct nouveau_fence *fence = NULL;
+	struct reservation_object *resv = nvbo->bo.resv;
+	struct reservation_object_list *fobj;
+	struct fence *fence = NULL;
+
+	fobj = reservation_object_get_list(resv);
 
 	list_del(&vma->head);
 
-	if (mapped) {
-		spin_lock(&nvbo->bo.bdev->fence_lock);
-		fence = nouveau_fence_ref(nvbo->bo.sync_obj);
-		spin_unlock(&nvbo->bo.bdev->fence_lock);
-	}
+	if (fobj && fobj->shared_count > 1)
+		ttm_bo_wait(&nvbo->bo, true, false, false);
+	else if (fobj && fobj->shared_count == 1)
+		fence = rcu_dereference_protected(fobj->shared[0],
+						reservation_object_held(resv));
+	else
+		fence = reservation_object_get_excl(nvbo->bo.resv);
 
-	if (fence) {
+	if (fence && mapped) {
 		nouveau_fence_work(fence, nouveau_gem_object_delete, vma);
 	} else {
 		if (mapped)
@@ -116,7 +122,6 @@ nouveau_gem_object_unmap(struct nouveau_bo *nvbo, struct nouveau_vma *vma)
 		nouveau_vm_put(vma);
 		kfree(vma);
 	}
-	nouveau_fence_unref(&fence);
 }
 
 void
@@ -288,24 +293,23 @@ nouveau_gem_set_domain(struct drm_gem_object *gem, uint32_t read_domains,
 }
 
 struct validate_op {
-	struct list_head vram_list;
-	struct list_head gart_list;
-	struct list_head both_list;
+	struct list_head list;
 	struct ww_acquire_ctx ticket;
 };
 
 static void
-validate_fini_list(struct list_head *list, struct nouveau_fence *fence,
-		   struct ww_acquire_ctx *ticket)
+validate_fini_no_ticket(struct validate_op *op, struct nouveau_fence *fence,
+			struct drm_nouveau_gem_pushbuf_bo *pbbo)
 {
-	struct list_head *entry, *tmp;
 	struct nouveau_bo *nvbo;
+	struct drm_nouveau_gem_pushbuf_bo *b;
 
-	list_for_each_safe(entry, tmp, list) {
-		nvbo = list_entry(entry, struct nouveau_bo, entry);
+	while (!list_empty(&op->list)) {
+		nvbo = list_entry(op->list.next, struct nouveau_bo, entry);
+		b = &pbbo[nvbo->pbbo_index];
 
 		if (likely(fence))
-			nouveau_bo_fence(nvbo, fence);
+			nouveau_bo_fence(nvbo, fence, !!b->write_domains);
 
 		if (unlikely(nvbo->validate_mapped)) {
 			ttm_bo_kunmap(&nvbo->kmap);
@@ -314,23 +318,16 @@ validate_fini_list(struct list_head *list, struct nouveau_fence *fence,
 
 		list_del(&nvbo->entry);
 		nvbo->reserved_by = NULL;
-		ttm_bo_unreserve_ticket(&nvbo->bo, ticket);
+		ttm_bo_unreserve_ticket(&nvbo->bo, &op->ticket);
 		drm_gem_object_unreference_unlocked(&nvbo->gem);
 	}
 }
 
 static void
-validate_fini_no_ticket(struct validate_op *op, struct nouveau_fence *fence)
+validate_fini(struct validate_op *op, struct nouveau_fence *fence,
+	      struct drm_nouveau_gem_pushbuf_bo *pbbo)
 {
-	validate_fini_list(&op->vram_list, fence, &op->ticket);
-	validate_fini_list(&op->gart_list, fence, &op->ticket);
-	validate_fini_list(&op->both_list, fence, &op->ticket);
-}
-
-static void
-validate_fini(struct validate_op *op, struct nouveau_fence *fence)
-{
-	validate_fini_no_ticket(op, fence);
+	validate_fini_no_ticket(op, fence, pbbo);
 	ww_acquire_fini(&op->ticket);
 }
 
@@ -344,6 +341,9 @@ validate_init(struct nouveau_channel *chan, struct drm_file *file_priv,
 	int trycnt = 0;
 	int ret, i;
 	struct nouveau_bo *res_bo = NULL;
+	LIST_HEAD(gart_list);
+	LIST_HEAD(vram_list);
+	LIST_HEAD(both_list);
 
 	ww_acquire_init(&op->ticket, &reservation_ww_class);
 retry:
@@ -360,9 +360,8 @@ retry:
 		gem = drm_gem_object_lookup(dev, file_priv, b->handle);
 		if (!gem) {
 			NV_PRINTK(error, cli, "Unknown handle 0x%08x\n", b->handle);
-			ww_acquire_done(&op->ticket);
-			validate_fini(op, NULL);
-			return -ENOENT;
+			ret = -ENOENT;
+			break;
 		}
 		nvbo = nouveau_gem_object(gem);
 		if (nvbo == res_bo) {
@@ -375,14 +374,16 @@ retry:
 			NV_PRINTK(error, cli, "multiple instances of buffer %d on "
 				      "validation list\n", b->handle);
 			drm_gem_object_unreference_unlocked(gem);
-			ww_acquire_done(&op->ticket);
-			validate_fini(op, NULL);
-			return -EINVAL;
+			ret = -EINVAL;
+			break;
 		}
 
 		ret = ttm_bo_reserve(&nvbo->bo, true, false, true, &op->ticket);
 		if (ret) {
-			validate_fini_no_ticket(op, NULL);
+			list_splice_tail_init(&vram_list, &op->list);
+			list_splice_tail_init(&gart_list, &op->list);
+			list_splice_tail_init(&both_list, &op->list);
+			validate_fini_no_ticket(op, NULL, NULL);
 			if (unlikely(ret == -EDEADLK)) {
 				ret = ttm_bo_reserve_slowpath(&nvbo->bo, true,
 							      &op->ticket);
@@ -390,12 +391,9 @@ retry:
 					res_bo = nvbo;
 			}
 			if (unlikely(ret)) {
-				ww_acquire_done(&op->ticket);
-				ww_acquire_fini(&op->ticket);
-				drm_gem_object_unreference_unlocked(gem);
 				if (ret != -ERESTARTSYS)
 					NV_PRINTK(error, cli, "fail reserve\n");
-				return ret;
+				break;
 			}
 		}
 
@@ -404,45 +402,32 @@ retry:
 		nvbo->pbbo_index = i;
 		if ((b->valid_domains & NOUVEAU_GEM_DOMAIN_VRAM) &&
 		    (b->valid_domains & NOUVEAU_GEM_DOMAIN_GART))
-			list_add_tail(&nvbo->entry, &op->both_list);
+			list_add_tail(&nvbo->entry, &both_list);
 		else
 		if (b->valid_domains & NOUVEAU_GEM_DOMAIN_VRAM)
-			list_add_tail(&nvbo->entry, &op->vram_list);
+			list_add_tail(&nvbo->entry, &vram_list);
 		else
 		if (b->valid_domains & NOUVEAU_GEM_DOMAIN_GART)
-			list_add_tail(&nvbo->entry, &op->gart_list);
+			list_add_tail(&nvbo->entry, &gart_list);
 		else {
 			NV_PRINTK(error, cli, "invalid valid domains: 0x%08x\n",
 				 b->valid_domains);
-			list_add_tail(&nvbo->entry, &op->both_list);
-			ww_acquire_done(&op->ticket);
-			validate_fini(op, NULL);
-			return -EINVAL;
+			list_add_tail(&nvbo->entry, &both_list);
+			ret = -EINVAL;
+			break;
 		}
 		if (nvbo == res_bo)
 			goto retry;
 	}
 
 	ww_acquire_done(&op->ticket);
-	return 0;
-}
-
-static int
-validate_sync(struct nouveau_channel *chan, struct nouveau_bo *nvbo)
-{
-	struct nouveau_fence *fence = NULL;
-	int ret = 0;
-
-	spin_lock(&nvbo->bo.bdev->fence_lock);
-	fence = nouveau_fence_ref(nvbo->bo.sync_obj);
-	spin_unlock(&nvbo->bo.bdev->fence_lock);
-
-	if (fence) {
-		ret = nouveau_fence_sync(fence, chan);
-		nouveau_fence_unref(&fence);
-	}
-
+	list_splice_tail(&vram_list, &op->list);
+	list_splice_tail(&gart_list, &op->list);
+	list_splice_tail(&both_list, &op->list);
+	if (ret)
+		validate_fini(op, NULL, NULL);
 	return ret;
+
 }
 
 static int
@@ -474,9 +459,10 @@ validate_list(struct nouveau_channel *chan, struct nouveau_cli *cli,
 			return ret;
 		}
 
-		ret = validate_sync(chan, nvbo);
+		ret = nouveau_fence_sync(nvbo, chan, !!b->write_domains);
 		if (unlikely(ret)) {
-			NV_PRINTK(error, cli, "fail post-validate sync\n");
+			if (ret != -ERESTARTSYS)
+				NV_PRINTK(error, cli, "fail post-validate sync\n");
 			return ret;
 		}
 
@@ -513,11 +499,9 @@ nouveau_gem_pushbuf_validate(struct nouveau_channel *chan,
 			     struct validate_op *op, int *apply_relocs)
 {
 	struct nouveau_cli *cli = nouveau_cli(file_priv);
-	int ret, relocs = 0;
+	int ret;
 
-	INIT_LIST_HEAD(&op->vram_list);
-	INIT_LIST_HEAD(&op->gart_list);
-	INIT_LIST_HEAD(&op->both_list);
+	INIT_LIST_HEAD(&op->list);
 
 	if (nr_buffers == 0)
 		return 0;
@@ -529,34 +513,14 @@ nouveau_gem_pushbuf_validate(struct nouveau_channel *chan,
 		return ret;
 	}
 
-	ret = validate_list(chan, cli, &op->vram_list, pbbo, user_buffers);
-	if (unlikely(ret < 0)) {
-		if (ret != -ERESTARTSYS)
-			NV_PRINTK(error, cli, "validate vram_list\n");
-		validate_fini(op, NULL);
-		return ret;
-	}
-	relocs += ret;
-
-	ret = validate_list(chan, cli, &op->gart_list, pbbo, user_buffers);
-	if (unlikely(ret < 0)) {
-		if (ret != -ERESTARTSYS)
-			NV_PRINTK(error, cli, "validate gart_list\n");
-		validate_fini(op, NULL);
-		return ret;
-	}
-	relocs += ret;
-
-	ret = validate_list(chan, cli, &op->both_list, pbbo, user_buffers);
+	ret = validate_list(chan, cli, &op->list, pbbo, user_buffers);
 	if (unlikely(ret < 0)) {
 		if (ret != -ERESTARTSYS)
-			NV_PRINTK(error, cli, "validate both_list\n");
-		validate_fini(op, NULL);
+			NV_PRINTK(error, cli, "validating bo list\n");
+		validate_fini(op, NULL, NULL);
 		return ret;
 	}
-	relocs += ret;
-
-	*apply_relocs = relocs;
+	*apply_relocs = ret;
 	return 0;
 }
 
@@ -659,9 +623,7 @@ nouveau_gem_pushbuf_reloc_apply(struct nouveau_cli *cli,
 				data |= r->vor;
 		}
 
-		spin_lock(&nvbo->bo.bdev->fence_lock);
-		ret = ttm_bo_wait(&nvbo->bo, false, false, false);
-		spin_unlock(&nvbo->bo.bdev->fence_lock);
+		ret = ttm_bo_wait(&nvbo->bo, true, false, false);
 		if (ret) {
 			NV_PRINTK(error, cli, "reloc wait_idle failed: %d\n", ret);
 			break;
@@ -839,7 +801,7 @@ nouveau_gem_ioctl_pushbuf(struct drm_device *dev, void *data,
 	}
 
 out:
-	validate_fini(&op, fence);
+	validate_fini(&op, fence, bo);
 	nouveau_fence_unref(&fence);
 
 out_prevalid:
@@ -884,17 +846,29 @@ nouveau_gem_ioctl_cpu_prep(struct drm_device *dev, void *data,
 	struct drm_gem_object *gem;
 	struct nouveau_bo *nvbo;
 	bool no_wait = !!(req->flags & NOUVEAU_GEM_CPU_PREP_NOWAIT);
-	int ret = -EINVAL;
+	bool write = !!(req->flags & NOUVEAU_GEM_CPU_PREP_WRITE);
+	int ret;
 
 	gem = drm_gem_object_lookup(dev, file_priv, req->handle);
 	if (!gem)
 		return -ENOENT;
 	nvbo = nouveau_gem_object(gem);
 
-	spin_lock(&nvbo->bo.bdev->fence_lock);
-	ret = ttm_bo_wait(&nvbo->bo, true, true, no_wait);
-	spin_unlock(&nvbo->bo.bdev->fence_lock);
+	if (no_wait)
+		ret = reservation_object_test_signaled_rcu(nvbo->bo.resv, write) ? 0 : -EBUSY;
+	else {
+		long lret;
+
+		lret = reservation_object_wait_timeout_rcu(nvbo->bo.resv, write, true, 30 * HZ);
+		if (!lret)
+			ret = -EBUSY;
+		else if (lret > 0)
+			ret = 0;
+		else
+			ret = lret;
+	}
 	drm_gem_object_unreference_unlocked(gem);
+
 	return ret;
 }
 
diff --git a/drm/nouveau_ttm.c b/drm/nouveau_ttm.c
index 53874b76b..e81d08657 100644
--- a/drm/nouveau_ttm.c
+++ b/drm/nouveau_ttm.c
@@ -71,8 +71,7 @@ nouveau_vram_manager_del(struct ttm_mem_type_manager *man,
 static int
 nouveau_vram_manager_new(struct ttm_mem_type_manager *man,
 			 struct ttm_buffer_object *bo,
-			 struct ttm_placement *placement,
-			 uint32_t flags,
+			 const struct ttm_place *place,
 			 struct ttm_mem_reg *mem)
 {
 	struct nouveau_drm *drm = nouveau_bdev(man->bdev);
@@ -158,8 +157,7 @@ nouveau_gart_manager_del(struct ttm_mem_type_manager *man,
 static int
 nouveau_gart_manager_new(struct ttm_mem_type_manager *man,
 			 struct ttm_buffer_object *bo,
-			 struct ttm_placement *placement,
-			 uint32_t flags,
+			 const struct ttm_place *place,
 			 struct ttm_mem_reg *mem)
 {
 	struct nouveau_drm *drm = nouveau_bdev(bo->bdev);
@@ -239,8 +237,7 @@ nv04_gart_manager_del(struct ttm_mem_type_manager *man, struct ttm_mem_reg *mem)
 static int
 nv04_gart_manager_new(struct ttm_mem_type_manager *man,
 		      struct ttm_buffer_object *bo,
-		      struct ttm_placement *placement,
-		      uint32_t flags,
+		      const struct ttm_place *place,
 		      struct ttm_mem_reg *mem)
 {
 	struct nouveau_mem *node;
diff --git a/drm/nv04_fence.c b/drm/nv04_fence.c
index 239c2c5a9..4484131d8 100644
--- a/drm/nv04_fence.c
+++ b/drm/nv04_fence.c
@@ -41,7 +41,7 @@ nv04_fence_emit(struct nouveau_fence *fence)
 	int ret = RING_SPACE(chan, 2);
 	if (ret == 0) {
 		BEGIN_NV04(chan, NvSubSw, 0x0150, 1);
-		OUT_RING  (chan, fence->sequence);
+		OUT_RING  (chan, fence->base.seqno);
 		FIRE_RING (chan);
 	}
 	return ret;
@@ -75,7 +75,7 @@ nv04_fence_context_new(struct nouveau_channel *chan)
 {
 	struct nv04_fence_chan *fctx = kzalloc(sizeof(*fctx), GFP_KERNEL);
 	if (fctx) {
-		nouveau_fence_context_new(&fctx->base);
+		nouveau_fence_context_new(chan, &fctx->base);
 		fctx->base.emit = nv04_fence_emit;
 		fctx->base.sync = nv04_fence_sync;
 		fctx->base.read = nv04_fence_read;
@@ -105,5 +105,7 @@ nv04_fence_create(struct nouveau_drm *drm)
 	priv->base.dtor = nv04_fence_destroy;
 	priv->base.context_new = nv04_fence_context_new;
 	priv->base.context_del = nv04_fence_context_del;
+	priv->base.contexts = 15;
+	priv->base.context_base = fence_context_alloc(priv->base.contexts);
 	return 0;
 }
diff --git a/drm/nv10_fence.c b/drm/nv10_fence.c
index 4faaf0acf..737d066ff 100644
--- a/drm/nv10_fence.c
+++ b/drm/nv10_fence.c
@@ -33,7 +33,7 @@ nv10_fence_emit(struct nouveau_fence *fence)
 	int ret = RING_SPACE(chan, 2);
 	if (ret == 0) {
 		BEGIN_NV04(chan, 0, NV10_SUBCHAN_REF_CNT, 1);
-		OUT_RING  (chan, fence->sequence);
+		OUT_RING  (chan, fence->base.seqno);
 		FIRE_RING (chan);
 	}
 	return ret;
@@ -75,7 +75,7 @@ nv10_fence_context_new(struct nouveau_channel *chan)
 	if (!fctx)
 		return -ENOMEM;
 
-	nouveau_fence_context_new(&fctx->base);
+	nouveau_fence_context_new(chan, &fctx->base);
 	fctx->base.emit = nv10_fence_emit;
 	fctx->base.read = nv10_fence_read;
 	fctx->base.sync = nv10_fence_sync;
@@ -106,6 +106,8 @@ nv10_fence_create(struct nouveau_drm *drm)
 	priv->base.dtor = nv10_fence_destroy;
 	priv->base.context_new = nv10_fence_context_new;
 	priv->base.context_del = nv10_fence_context_del;
+	priv->base.contexts = 31;
+	priv->base.context_base = fence_context_alloc(priv->base.contexts);
 	spin_lock_init(&priv->lock);
 	return 0;
 }
diff --git a/drm/nv17_fence.c b/drm/nv17_fence.c
index ca907479f..6f9a1f8e2 100644
--- a/drm/nv17_fence.c
+++ b/drm/nv17_fence.c
@@ -84,7 +84,7 @@ nv17_fence_context_new(struct nouveau_channel *chan)
 	if (!fctx)
 		return -ENOMEM;
 
-	nouveau_fence_context_new(&fctx->base);
+	nouveau_fence_context_new(chan, &fctx->base);
 	fctx->base.emit = nv10_fence_emit;
 	fctx->base.read = nv10_fence_read;
 	fctx->base.sync = nv17_fence_sync;
@@ -124,6 +124,8 @@ nv17_fence_create(struct nouveau_drm *drm)
 	priv->base.resume = nv17_fence_resume;
 	priv->base.context_new = nv17_fence_context_new;
 	priv->base.context_del = nv10_fence_context_del;
+	priv->base.contexts = 31;
+	priv->base.context_base = fence_context_alloc(priv->base.contexts);
 	spin_lock_init(&priv->lock);
 
 	ret = nouveau_bo_new(drm->dev, 4096, 0x1000, TTM_PL_FLAG_VRAM,
diff --git a/drm/nv50_fence.c b/drm/nv50_fence.c
index 195cf51a7..08fad3668 100644
--- a/drm/nv50_fence.c
+++ b/drm/nv50_fence.c
@@ -46,7 +46,7 @@ nv50_fence_context_new(struct nouveau_channel *chan)
 	if (!fctx)
 		return -ENOMEM;
 
-	nouveau_fence_context_new(&fctx->base);
+	nouveau_fence_context_new(chan, &fctx->base);
 	fctx->base.emit = nv10_fence_emit;
 	fctx->base.read = nv10_fence_read;
 	fctx->base.sync = nv17_fence_sync;
@@ -95,6 +95,8 @@ nv50_fence_create(struct nouveau_drm *drm)
 	priv->base.resume = nv17_fence_resume;
 	priv->base.context_new = nv50_fence_context_new;
 	priv->base.context_del = nv10_fence_context_del;
+	priv->base.contexts = 127;
+	priv->base.context_base = fence_context_alloc(priv->base.contexts);
 	spin_lock_init(&priv->lock);
 
 	ret = nouveau_bo_new(drm->dev, 4096, 0x1000, TTM_PL_FLAG_VRAM,
diff --git a/drm/nv84_fence.c b/drm/nv84_fence.c
index 933a779c9..a2f28082c 100644
--- a/drm/nv84_fence.c
+++ b/drm/nv84_fence.c
@@ -82,7 +82,7 @@ nv84_fence_emit(struct nouveau_fence *fence)
 	else
 		addr += fctx->vma.offset;
 
-	return fctx->base.emit32(chan, addr, fence->sequence);
+	return fctx->base.emit32(chan, addr, fence->base.seqno);
 }
 
 static int
@@ -97,7 +97,7 @@ nv84_fence_sync(struct nouveau_fence *fence,
 	else
 		addr += fctx->vma.offset;
 
-	return fctx->base.sync32(chan, addr, fence->sequence);
+	return fctx->base.sync32(chan, addr, fence->base.seqno);
 }
 
 static u32
@@ -139,12 +139,13 @@ nv84_fence_context_new(struct nouveau_channel *chan)
 	if (!fctx)
 		return -ENOMEM;
 
-	nouveau_fence_context_new(&fctx->base);
+	nouveau_fence_context_new(chan, &fctx->base);
 	fctx->base.emit = nv84_fence_emit;
 	fctx->base.sync = nv84_fence_sync;
 	fctx->base.read = nv84_fence_read;
 	fctx->base.emit32 = nv84_fence_emit32;
 	fctx->base.sync32 = nv84_fence_sync32;
+	fctx->base.sequence = nv84_fence_read(chan);
 
 	ret = nouveau_bo_vma_add(priv->bo, cli->vm, &fctx->vma);
 	if (ret == 0) {
@@ -168,13 +169,12 @@ nv84_fence_context_new(struct nouveau_channel *chan)
 static bool
 nv84_fence_suspend(struct nouveau_drm *drm)
 {
-	struct nouveau_fifo *pfifo = nvkm_fifo(&drm->device);
 	struct nv84_fence_priv *priv = drm->fence;
 	int i;
 
-	priv->suspend = vmalloc((pfifo->max + 1) * sizeof(u32));
+	priv->suspend = vmalloc(priv->base.contexts * sizeof(u32));
 	if (priv->suspend) {
-		for (i = 0; i <= pfifo->max; i++)
+		for (i = 0; i < priv->base.contexts; i++)
 			priv->suspend[i] = nouveau_bo_rd32(priv->bo, i*4);
 	}
 
@@ -184,12 +184,11 @@ nv84_fence_suspend(struct nouveau_drm *drm)
 static void
 nv84_fence_resume(struct nouveau_drm *drm)
 {
-	struct nouveau_fifo *pfifo = nvkm_fifo(&drm->device);
 	struct nv84_fence_priv *priv = drm->fence;
 	int i;
 
 	if (priv->suspend) {
-		for (i = 0; i <= pfifo->max; i++)
+		for (i = 0; i < priv->base.contexts; i++)
 			nouveau_bo_wr32(priv->bo, i*4, priv->suspend[i]);
 		vfree(priv->suspend);
 		priv->suspend = NULL;
@@ -229,10 +228,11 @@ nv84_fence_create(struct nouveau_drm *drm)
 	priv->base.context_new = nv84_fence_context_new;
 	priv->base.context_del = nv84_fence_context_del;
 
-	init_waitqueue_head(&priv->base.waiting);
+	priv->base.contexts = pfifo->max + 1;
+	priv->base.context_base = fence_context_alloc(priv->base.contexts);
 	priv->base.uevent = true;
 
-	ret = nouveau_bo_new(drm->dev, 16 * (pfifo->max + 1), 0,
+	ret = nouveau_bo_new(drm->dev, 16 * priv->base.contexts, 0,
 			     TTM_PL_FLAG_VRAM, 0, 0, NULL, &priv->bo);
 	if (ret == 0) {
 		ret = nouveau_bo_pin(priv->bo, TTM_PL_FLAG_VRAM);
@@ -246,7 +246,7 @@ nv84_fence_create(struct nouveau_drm *drm)
 	}
 
 	if (ret == 0)
-		ret = nouveau_bo_new(drm->dev, 16 * (pfifo->max + 1), 0,
+		ret = nouveau_bo_new(drm->dev, 16 * priv->base.contexts, 0,
 				     TTM_PL_FLAG_TT, 0, 0, NULL,
 				     &priv->bo_gart);
 	if (ret == 0) {