| Message ID | 1436477570-4936-12-git-send-email-daniel.vetter@ffwll.ch (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Thu, Jul 09, 2015 at 11:32:43PM +0200, Daniel Vetter wrote: > Since David Herrmann's mmap vma manager rework we don't need to grab > dev->struct_mutex any more to prevent races when looking up the mmap > offset. Drop it and instead don't forget to use the unref_unlocked > variant (since the drm core still cares). > > While at it also fix a leak when this ioctl is called on an imported > buffer. > > Cc: Russell King <rmk+kernel@arm.linux.org.uk> > Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> > --- > drivers/gpu/drm/armada/armada_gem.c | 11 ++++------- > 1 file changed, 4 insertions(+), 7 deletions(-) Reviewed-by: Thierry Reding <treding@nvidia.com>
On Thu, Jul 09, 2015 at 11:32:43PM +0200, Daniel Vetter wrote: > Since David Herrmann's mmap vma manager rework we don't need to grab > dev->struct_mutex any more to prevent races when looking up the mmap > offset. Drop it and instead don't forget to use the unref_unlocked > variant (since the drm core still cares). > > While at it also fix a leak when this ioctl is called on an imported > buffer. Good catch, thanks Daniel. I'd prefer these to be two different commits though, so that the leak can be easily backported to stable kernels if needed. I suspect this should go through the same tree that David's work is, otherwise we end up with random dependencies between trees. With the bug fix separated out: Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
diff --git a/drivers/gpu/drm/armada/armada_gem.c b/drivers/gpu/drm/armada/armada_gem.c index 580e10acaa3a..95df74227ec0 100644 --- a/drivers/gpu/drm/armada/armada_gem.c +++ b/drivers/gpu/drm/armada/armada_gem.c @@ -273,18 +273,16 @@ int armada_gem_dumb_map_offset(struct drm_file *file, struct drm_device *dev, struct armada_gem_object *obj; int ret = 0; - mutex_lock(&dev->struct_mutex); obj = armada_gem_object_lookup(dev, file, handle); if (!obj) { DRM_ERROR("failed to lookup gem object\n"); - ret = -EINVAL; - goto err_unlock; + return -EINVAL; } /* Don't allow imported objects to be mapped */ if (obj->obj.import_attach) { ret = -EINVAL; - goto err_unlock; + goto err_unref; } ret = drm_gem_create_mmap_offset(&obj->obj); @@ -293,9 +291,8 @@ int armada_gem_dumb_map_offset(struct drm_file *file, struct drm_device *dev, DRM_DEBUG_DRIVER("handle %#x offset %llx\n", handle, *offset); } - drm_gem_object_unreference(&obj->obj); - err_unlock: - mutex_unlock(&dev->struct_mutex); +err_unref: + drm_gem_object_unreference_unlocked(&obj->obj); return ret; }
Since David Herrmann's mmap vma manager rework we don't need to grab dev->struct_mutex any more to prevent races when looking up the mmap offset. Drop it and instead don't forget to use the unref_unlocked variant (since the drm core still cares). While at it also fix a leak when this ioctl is called on an imported buffer. Cc: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> --- drivers/gpu/drm/armada/armada_gem.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-)