OSDN Git Service

drm: Fixup racy refcounting in plane_force_disable
authorDaniel Vetter <daniel.vetter@ffwll.ch>
Fri, 27 Feb 2015 11:58:13 +0000 (12:58 +0100)
committerDave Airlie <airlied@redhat.com>
Tue, 24 Mar 2015 02:04:35 +0000 (12:04 +1000)
commit8218c3f4df3bb1c637c17552405039a6dd3c1ee1
tree982545a11600b7a261574e0327d6c0d69d8e215c
parent90a5a895cc8b284ac522757a01de15e36710c2b9
drm: Fixup racy refcounting in plane_force_disable

Originally it was impossible to be dropping the last refcount in this
function since there was always one around still from the idr. But in

commit 83f45fc360c8e16a330474860ebda872d1384c8c
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Wed Aug 6 09:10:18 2014 +0200

    drm: Don't grab an fb reference for the idr

we've switched to weak references, broke that assumption but forgot to
fix it up.

Since we still force-disable planes it's only possible to hit this
when racing multiple rmfb with fbdev restoring or similar evil things.
As long as userspace is nice it's impossible to hit the BUG_ON.

But the BUG_ON would most likely be hit from fbdev code, which usually
invovles the console_lock besides all modeset locks. So very likely
we'd never get the bug reports if this was hit in the wild, hence
better be safe than sorry and backport.

Spotted by Matt Roper while reviewing other patches.

[airlied: pull this back into 4.0 - the oops happens there]

Cc: stable@vger.kernel.org
Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
drivers/gpu/drm/drm_crtc.c