diff mbox

drm/i915: Unref context on failed eb_create

Message ID 1396676467-21368-1-git-send-email-benjamin.widawsky@intel.com (mailing list archive)
State New, archived
Headers show

Commit Message

Ben Widawsky April 5, 2014, 5:41 a.m. UTC
I opted to do this instead of grabbing the context reference after
eb_create since eb_create can potentially call the shrinker, and that
makes things very complicated. This simple patch balances the ref count
without requiring a great deal of review to make sure the shrinker path
is safe.

Theoretically (by design) the shrinker can end up destroying a context,
which enforces the reasoning for doing the fix this way instead of
moving the reference to later in the function.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
---
 drivers/gpu/drm/i915/i915_gem_execbuffer.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Comments

Chris Wilson April 5, 2014, 6:24 a.m. UTC | #1
On Fri, Apr 04, 2014 at 10:41:07PM -0700, Ben Widawsky wrote:
> I opted to do this instead of grabbing the context reference after
> eb_create since eb_create can potentially call the shrinker, and that
> makes things very complicated. This simple patch balances the ref count
> without requiring a great deal of review to make sure the shrinker path
> is safe.
> 
> Theoretically (by design) the shrinker can end up destroying a context,
> which enforces the reasoning for doing the fix this way instead of
> moving the reference to later in the function.
> 
> Cc: Chris Wilson <chris@chris-wilson.co.uk>
> Signed-off-by: Ben Widawsky <ben@bwidawsk.net>

Even the whitespace fix is correct.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
-Chris
Daniel Vetter April 5, 2014, 3:24 p.m. UTC | #2
On Sat, Apr 05, 2014 at 07:24:01AM +0100, Chris Wilson wrote:
> On Fri, Apr 04, 2014 at 10:41:07PM -0700, Ben Widawsky wrote:
> > I opted to do this instead of grabbing the context reference after
> > eb_create since eb_create can potentially call the shrinker, and that
> > makes things very complicated. This simple patch balances the ref count
> > without requiring a great deal of review to make sure the shrinker path
> > is safe.
> > 
> > Theoretically (by design) the shrinker can end up destroying a context,
> > which enforces the reasoning for doing the fix this way instead of
> > moving the reference to later in the function.
> > 
> > Cc: Chris Wilson <chris@chris-wilson.co.uk>
> > Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
> 
> Even the whitespace fix is correct.
> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

Queued for -next, thanks for the patch.
-Daniel
diff mbox

Patch

diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index e4f0e2c..b785459 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -1133,7 +1133,7 @@  i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 		mutex_unlock(&dev->struct_mutex);
 		ret = PTR_ERR(ctx);
 		goto pre_mutex_err;
-	} 
+	}
 
 	i915_gem_context_reference(ctx);
 
@@ -1143,6 +1143,7 @@  i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 
 	eb = eb_create(args);
 	if (eb == NULL) {
+		i915_gem_context_unreference(ctx);
 		mutex_unlock(&dev->struct_mutex);
 		ret = -ENOMEM;
 		goto pre_mutex_err;