This is temporary until the fence framework can be used. With the lru/reservation atomicity removal it is possible to see your old sequence number and the buffer being reserved, leading to erroneously reporting -EDEADLK.
Workaround it by bumping the sequence number every retry.
Signed-off-by: Maarten Lankhorst maarten.lankhorst@canonical.com --- drivers/gpu/drm/nouveau/nouveau_gem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c index 7b9364b..f2bd59e 100644 --- a/drivers/gpu/drm/nouveau/nouveau_gem.c +++ b/drivers/gpu/drm/nouveau/nouveau_gem.c @@ -322,8 +322,8 @@ validate_init(struct nouveau_channel *chan, struct drm_file *file_priv, int trycnt = 0; int ret, i;
- sequence = atomic_add_return(1, &drm->ttm.validate_sequence); retry: + sequence = atomic_add_return(1, &drm->ttm.validate_sequence); if (++trycnt > 100000) { NV_ERROR(drm, "%s failed and gave up.\n", __func__); return -EINVAL;