From: Thierry Reding treding@nvidia.com
Turn nouveau_fence_sync() into a low-level helper that adds fence waits to the channel command stream. The new nouveau_bo_sync() helper replaces the previous nouveau_fence_sync() implementation. It passes each of the buffer object's fences to nouveau_fence_sync() in turn.
This provides more fine-grained control over fences which is needed by subsequent patches for sync fd support.
Heavily based on work by Lauri Peltonen lpeltonen@nvidia.com.
Signed-off-by: Thierry Reding treding@nvidia.com --- drivers/gpu/drm/nouveau/dispnv04/crtc.c | 4 +- drivers/gpu/drm/nouveau/nouveau_bo.c | 38 +++++++++++++- drivers/gpu/drm/nouveau/nouveau_bo.h | 2 + drivers/gpu/drm/nouveau/nouveau_fence.c | 68 +++++-------------------- drivers/gpu/drm/nouveau/nouveau_fence.h | 2 +- drivers/gpu/drm/nouveau/nouveau_gem.c | 2 +- 6 files changed, 57 insertions(+), 59 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/dispnv04/crtc.c b/drivers/gpu/drm/nouveau/dispnv04/crtc.c index 6416b6907aeb..4af702d0d6bf 100644 --- a/drivers/gpu/drm/nouveau/dispnv04/crtc.c +++ b/drivers/gpu/drm/nouveau/dispnv04/crtc.c @@ -1117,7 +1117,7 @@ nv04_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, chan, false, false); + ret = nouveau_bo_sync(old_bo, chan, false, false); if (ret) goto fail;
@@ -1183,7 +1183,7 @@ nv04_crtc_page_flip(struct drm_crtc *crtc, struct drm_framebuffer *fb, goto fail_unpin;
/* synchronise rendering channel with the kernel's channel */ - ret = nouveau_fence_sync(new_bo, chan, false, true); + ret = nouveau_bo_sync(new_bo, chan, false, true); if (ret) { ttm_bo_unreserve(&new_bo->bo); goto fail_unpin; diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c index 9140387f30dc..25ceabfa741c 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bo.c +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c @@ -574,6 +574,42 @@ nouveau_bo_sync_for_cpu(struct nouveau_bo *nvbo) PAGE_SIZE, DMA_FROM_DEVICE); }
+int +nouveau_bo_sync(struct nouveau_bo *nvbo, struct nouveau_channel *chan, + bool exclusive, bool intr) +{ + struct dma_resv *resv = nvbo->bo.base.resv; + struct dma_resv_list *fobj; + struct dma_fence *fence; + int ret = 0, i; + + if (!exclusive) { + ret = dma_resv_reserve_shared(resv, 1); + if (ret < 0) + return ret; + } + + fobj = dma_resv_get_list(resv); + fence = dma_resv_get_excl(resv); + + if (fence && (!exclusive || !fobj || !fobj->shared_count)) + return nouveau_fence_sync(fence, chan, intr); + + if (!exclusive || !fobj) + return ret; + + for (i = 0; i < fobj->shared_count && !ret; ++i) { + fence = rcu_dereference_protected(fobj->shared[i], + dma_resv_held(resv)); + + ret = nouveau_fence_sync(fence, chan, intr); + if (ret < 0) + break; + } + + return ret; +} + int nouveau_bo_validate(struct nouveau_bo *nvbo, bool interruptible, bool no_wait_gpu) @@ -717,7 +753,7 @@ 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(nouveau_bo(bo), chan, true, intr); + ret = nouveau_bo_sync(nouveau_bo(bo), chan, true, intr); if (ret == 0) { ret = drm->ttm.move(chan, bo, &bo->mem, new_reg); if (ret == 0) { diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.h b/drivers/gpu/drm/nouveau/nouveau_bo.h index aecb7481df0d..93d1706619a1 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bo.h +++ b/drivers/gpu/drm/nouveau/nouveau_bo.h @@ -96,6 +96,8 @@ int nouveau_bo_validate(struct nouveau_bo *, bool interruptible, bool no_wait_gpu); void nouveau_bo_sync_for_device(struct nouveau_bo *nvbo); void nouveau_bo_sync_for_cpu(struct nouveau_bo *nvbo); +int nouveau_bo_sync(struct nouveau_bo *nvbo, struct nouveau_channel *channel, + bool exclusive, bool intr);
/* TODO: submit equivalent to TTM generic API upstream? */ static inline void __iomem * diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c b/drivers/gpu/drm/nouveau/nouveau_fence.c index e5dcbf67de7e..8e7550553584 100644 --- a/drivers/gpu/drm/nouveau/nouveau_fence.c +++ b/drivers/gpu/drm/nouveau/nouveau_fence.c @@ -339,66 +339,26 @@ nouveau_fence_wait(struct nouveau_fence *fence, bool lazy, bool intr) }
int -nouveau_fence_sync(struct nouveau_bo *nvbo, struct nouveau_channel *chan, bool exclusive, bool intr) +nouveau_fence_sync(struct dma_fence *fence, struct nouveau_channel *chan, + bool intr) { struct nouveau_fence_chan *fctx = chan->fence; - struct dma_fence *fence; - struct dma_resv *resv = nvbo->bo.base.resv; - struct dma_resv_list *fobj; + struct nouveau_channel *prev = NULL; struct nouveau_fence *f; - int ret = 0, i; - - if (!exclusive) { - ret = dma_resv_reserve_shared(resv, 1); + bool must_wait = true; + int ret = 0;
- if (ret) - return ret; + f = nouveau_local_fence(fence, chan->drm); + if (f) { + rcu_read_lock(); + prev = rcu_dereference(f->channel); + if (prev && (prev == chan || fctx->sync(f, prev, chan) == 0)) + must_wait = false; + rcu_read_unlock(); }
- fobj = dma_resv_get_list(resv); - fence = dma_resv_get_excl(resv); - - if (fence && (!exclusive || !fobj || !fobj->shared_count)) { - struct nouveau_channel *prev = NULL; - bool must_wait = true; - - f = nouveau_local_fence(fence, chan->drm); - if (f) { - rcu_read_lock(); - prev = rcu_dereference(f->channel); - if (prev && (prev == chan || fctx->sync(f, prev, chan) == 0)) - must_wait = false; - rcu_read_unlock(); - } - - if (must_wait) - ret = dma_fence_wait(fence, intr); - - return ret; - } - - if (!exclusive || !fobj) - return ret; - - for (i = 0; i < fobj->shared_count && !ret; ++i) { - struct nouveau_channel *prev = NULL; - bool must_wait = true; - - fence = rcu_dereference_protected(fobj->shared[i], - dma_resv_held(resv)); - - f = nouveau_local_fence(fence, chan->drm); - if (f) { - rcu_read_lock(); - prev = rcu_dereference(f->channel); - if (prev && (prev == chan || fctx->sync(f, prev, chan) == 0)) - must_wait = false; - rcu_read_unlock(); - } - - if (must_wait) - ret = dma_fence_wait(fence, intr); - } + if (must_wait) + ret = dma_fence_wait(fence, intr);
return ret; } diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.h b/drivers/gpu/drm/nouveau/nouveau_fence.h index 4887caa69c65..76cbf0c27a30 100644 --- a/drivers/gpu/drm/nouveau/nouveau_fence.h +++ b/drivers/gpu/drm/nouveau/nouveau_fence.h @@ -24,7 +24,7 @@ void nouveau_fence_unref(struct nouveau_fence **); int nouveau_fence_emit(struct nouveau_fence *, struct nouveau_channel *); bool nouveau_fence_done(struct nouveau_fence *); int nouveau_fence_wait(struct nouveau_fence *, bool lazy, bool intr); -int nouveau_fence_sync(struct nouveau_bo *, struct nouveau_channel *, bool exclusive, bool intr); +int nouveau_fence_sync(struct dma_fence *, struct nouveau_channel *, bool intr);
struct nouveau_fence_chan { spinlock_t lock; diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c index 81f111ad3f4f..590e4c1d2e8a 100644 --- a/drivers/gpu/drm/nouveau/nouveau_gem.c +++ b/drivers/gpu/drm/nouveau/nouveau_gem.c @@ -513,7 +513,7 @@ validate_list(struct nouveau_channel *chan, struct nouveau_cli *cli, return ret; }
- ret = nouveau_fence_sync(nvbo, chan, !!b->write_domains, true); + ret = nouveau_bo_sync(nvbo, chan, !!b->write_domains, true); if (unlikely(ret)) { if (ret != -ERESTARTSYS) NV_PRINTK(err, cli, "fail post-validate sync\n");