drm: fix a memleak on mutex failure path
authorOleg Drokin <green@linuxhacker.ru>
Mon, 27 Apr 2015 15:36:46 +0000 (11:36 -0400)
committerDaniel Vetter <daniel.vetter@ffwll.ch>
Tue, 5 May 2015 07:26:43 +0000 (09:26 +0200)
Need to free just allocated ctx allocation if we cannot
get our config mutex.

This one has been flagged by kbuild bot all the way back in August,
but somehow nobody picked it up:
https://lists.01.org/pipermail/kbuild/2014-August/001691.html

In addition there is another failure path that leaks the same
ctx reference that is fixed.

Found with smatch.

Signed-off-by: Oleg Drokin <green@linuxhacker.ru>
CC: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
drivers/gpu/drm/drm_modeset_lock.c

index 51cc47d827d82f09f1ed9d0e4200e299b4fa628c..c0a5cd8c52621301d0cb848e485bbe26cb1190bc 100644 (file)
@@ -80,8 +80,10 @@ int __drm_modeset_lock_all(struct drm_device *dev,
                return -ENOMEM;
 
        if (trylock) {
-               if (!mutex_trylock(&config->mutex))
-                       return -EBUSY;
+               if (!mutex_trylock(&config->mutex)) {
+                       ret = -EBUSY;
+                       goto out;
+               }
        } else {
                mutex_lock(&config->mutex);
        }
@@ -114,6 +116,8 @@ fail:
                goto retry;
        }
 
+out:
+       kfree(ctx);
        return ret;
 }
 EXPORT_SYMBOL(__drm_modeset_lock_all);