drm/vgem: Close use-after-free race in vgem_gem_create
authorDaniel Vetter <daniel.vetter@ffwll.ch>
Sun, 2 Feb 2020 13:21:33 +0000 (14:21 +0100)
committerDaniel Vetter <daniel.vetter@ffwll.ch>
Thu, 6 Feb 2020 18:04:41 +0000 (19:04 +0100)
commit4b848f20eda5974020f043ca14bacf7a7e634fc8
treecb8133c9dd80bcb546a363c455bd59e46fdea65b
parentdb735fc4036bbe1fbe606819b5f0ff26cc76cdff
drm/vgem: Close use-after-free race in vgem_gem_create

There's two references floating around here (for the object reference,
not the handle_count reference, that's a different thing):

- The temporary reference held by vgem_gem_create, acquired by
  creating the object and released by calling
  drm_gem_object_put_unlocked.

- The reference held by the object handle, created by
  drm_gem_handle_create. This one generally outlives the function,
  except if a 2nd thread races with a GEM_CLOSE ioctl call.

So usually everything is correct, except in that race case, where the
access to gem_object->size could be looking at freed data already.
Which again isn't a real problem (userspace shot its feet off already
with the race, we could return garbage), but maybe someone can exploit
this as an information leak.

Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Hillf Danton <hdanton@sina.com>
Reported-by: syzbot+0dc4444774d419e916c8@syzkaller.appspotmail.com
Cc: stable@vger.kernel.org
Cc: Emil Velikov <emil.velikov@collabora.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Sean Paul <seanpaul@chromium.org>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Eric Anholt <eric@anholt.net>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Rob Clark <robdclark@chromium.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200202132133.1891846-1-daniel.vetter@ffwll.ch
drivers/gpu/drm/vgem/vgem_drv.c