From 1df6a2ebd75067aefbdf07482bf8e3d0584e04ee Mon Sep 17 00:00:00 2001 From: Thomas Hellstrom Date: Thu, 30 Sep 2010 12:36:45 +0200 Subject: drm/ttm: Fix two race conditions + fix busy codepaths This fixes a race pointed out by Dave Airlie where we don't take a buffer object about to be destroyed off the LRU lists properly. It also fixes a rare case where a buffer object could be destroyed in the middle of an accelerated eviction. The patch also adds a utility function that can be used to prematurely release GPU memory space usage of an object waiting to be destroyed. For example during eviction or swapout. The above mentioned commit didn't queue the buffer on the delayed destroy list under some rare circumstances. It also didn't completely honor the remove_all parameter. Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=615505 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=591061 Signed-off-by: Thomas Hellstrom Signed-off-by: Dave Airlie --- drivers/gpu/drm/ttm/ttm_bo.c | 83 +++++++++++++++++++++++++++++++++++++------- 1 file changed, 71 insertions(+), 12 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index cb4cf7ef4d1e..db809e034cc4 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -441,6 +441,43 @@ out_err: return ret; } +/** + * Call bo::reserved and with the lru lock held. + * Will release GPU memory type usage on destruction. + * This is the place to put in driver specific hooks. + * Will release the bo::reserved lock and the + * lru lock on exit. + */ + +static void ttm_bo_cleanup_memtype_use(struct ttm_buffer_object *bo) +{ + struct ttm_bo_global *glob = bo->glob; + + if (bo->ttm) { + + /** + * Release the lru_lock, since we don't want to have + * an atomic requirement on ttm_tt[unbind|destroy]. + */ + + spin_unlock(&glob->lru_lock); + ttm_tt_unbind(bo->ttm); + ttm_tt_destroy(bo->ttm); + bo->ttm = NULL; + spin_lock(&glob->lru_lock); + } + + if (bo->mem.mm_node) { + drm_mm_put_block(bo->mem.mm_node); + bo->mem.mm_node = NULL; + } + + atomic_set(&bo->reserved, 0); + wake_up_all(&bo->event_queue); + spin_unlock(&glob->lru_lock); +} + + /** * If bo idle, remove from delayed- and lru lists, and unref. * If not idle, and already on delayed list, do nothing. @@ -456,6 +493,7 @@ static int ttm_bo_cleanup_refs(struct ttm_buffer_object *bo, bool remove_all) int ret; spin_lock(&bo->lock); +retry: (void) ttm_bo_wait(bo, false, false, !remove_all); if (!bo->sync_obj) { @@ -464,31 +502,52 @@ static int ttm_bo_cleanup_refs(struct ttm_buffer_object *bo, bool remove_all) spin_unlock(&bo->lock); spin_lock(&glob->lru_lock); - put_count = ttm_bo_del_from_lru(bo); + ret = ttm_bo_reserve_locked(bo, false, !remove_all, false, 0); + + /** + * Someone else has the object reserved. Bail and retry. + */ - ret = ttm_bo_reserve_locked(bo, false, false, false, 0); - BUG_ON(ret); - if (bo->ttm) - ttm_tt_unbind(bo->ttm); + if (unlikely(ret == -EBUSY)) { + spin_unlock(&glob->lru_lock); + spin_lock(&bo->lock); + goto requeue; + } + + /** + * We can re-check for sync object without taking + * the bo::lock since setting the sync object requires + * also bo::reserved. A busy object at this point may + * be caused by another thread starting an accelerated + * eviction. + */ + + if (unlikely(bo->sync_obj)) { + atomic_set(&bo->reserved, 0); + wake_up_all(&bo->event_queue); + spin_unlock(&glob->lru_lock); + spin_lock(&bo->lock); + if (remove_all) + goto retry; + else + goto requeue; + } + + put_count = ttm_bo_del_from_lru(bo); if (!list_empty(&bo->ddestroy)) { list_del_init(&bo->ddestroy); ++put_count; } - if (bo->mem.mm_node) { - drm_mm_put_block(bo->mem.mm_node); - bo->mem.mm_node = NULL; - } - spin_unlock(&glob->lru_lock); - atomic_set(&bo->reserved, 0); + ttm_bo_cleanup_memtype_use(bo); while (put_count--) kref_put(&bo->list_kref, ttm_bo_ref_bug); return 0; } - +requeue: spin_lock(&glob->lru_lock); if (list_empty(&bo->ddestroy)) { void *sync_obj = bo->sync_obj; -- cgit v1.2.3 From dab8dcfa3c8e3b021a138bee7c17791b4991ba55 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Thu, 7 Oct 2010 14:01:17 +1000 Subject: drm: don't drop handle reference on unload since the handle references are all tied to a file_priv, and when it disappears all the handle refs go with it. The fbcon ones we'd only notice on unload, but the nouveau notifier one would would happen on reboot. nouveau: Reported-by: Marc Dionne nouveau: Tested-by: Marc Dionne i915 unload: Reported-by: Keith Packard Acked-by: Ben Skeggs Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/intel_fb.c | 2 +- drivers/gpu/drm/nouveau/nouveau_fbcon.c | 1 - drivers/gpu/drm/nouveau/nouveau_notifier.c | 1 - drivers/gpu/drm/radeon/radeon_fb.c | 1 - 4 files changed, 1 insertion(+), 4 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/i915/intel_fb.c b/drivers/gpu/drm/i915/intel_fb.c index 56ad9df2ccb5..b61966c126d3 100644 --- a/drivers/gpu/drm/i915/intel_fb.c +++ b/drivers/gpu/drm/i915/intel_fb.c @@ -238,8 +238,8 @@ int intel_fbdev_destroy(struct drm_device *dev, drm_framebuffer_cleanup(&ifb->base); if (ifb->obj) { - drm_gem_object_handle_unreference(ifb->obj); drm_gem_object_unreference(ifb->obj); + ifb->obj = NULL; } return 0; diff --git a/drivers/gpu/drm/nouveau/nouveau_fbcon.c b/drivers/gpu/drm/nouveau/nouveau_fbcon.c index d2047713dc59..dbd30b2e43fd 100644 --- a/drivers/gpu/drm/nouveau/nouveau_fbcon.c +++ b/drivers/gpu/drm/nouveau/nouveau_fbcon.c @@ -352,7 +352,6 @@ nouveau_fbcon_destroy(struct drm_device *dev, struct nouveau_fbdev *nfbdev) if (nouveau_fb->nvbo) { nouveau_bo_unmap(nouveau_fb->nvbo); - drm_gem_object_handle_unreference_unlocked(nouveau_fb->nvbo->gem); drm_gem_object_unreference_unlocked(nouveau_fb->nvbo->gem); nouveau_fb->nvbo = NULL; } diff --git a/drivers/gpu/drm/nouveau/nouveau_notifier.c b/drivers/gpu/drm/nouveau/nouveau_notifier.c index 3c9964a8fbad..3ec181ff50ce 100644 --- a/drivers/gpu/drm/nouveau/nouveau_notifier.c +++ b/drivers/gpu/drm/nouveau/nouveau_notifier.c @@ -79,7 +79,6 @@ nouveau_notifier_takedown_channel(struct nouveau_channel *chan) mutex_lock(&dev->struct_mutex); nouveau_bo_unpin(chan->notifier_bo); mutex_unlock(&dev->struct_mutex); - drm_gem_object_handle_unreference_unlocked(chan->notifier_bo->gem); drm_gem_object_unreference_unlocked(chan->notifier_bo->gem); drm_mm_takedown(&chan->notifier_heap); } diff --git a/drivers/gpu/drm/radeon/radeon_fb.c b/drivers/gpu/drm/radeon/radeon_fb.c index 9cdf6a35bc2c..40b0c087b592 100644 --- a/drivers/gpu/drm/radeon/radeon_fb.c +++ b/drivers/gpu/drm/radeon/radeon_fb.c @@ -97,7 +97,6 @@ static void radeonfb_destroy_pinned_object(struct drm_gem_object *gobj) radeon_bo_unpin(rbo); radeon_bo_unreserve(rbo); } - drm_gem_object_handle_unreference(gobj); drm_gem_object_unreference_unlocked(gobj); } -- cgit v1.2.3