drm: Drop dev->event_lock spinlock around faulting copy_to_user()
authorChris Wilson <chris@chris-wilson.co.uk>
Wed, 25 Nov 2015 14:39:02 +0000 (14:39 +0000)
committerDaniel Vetter <daniel.vetter@ffwll.ch>
Thu, 26 Nov 2015 14:21:06 +0000 (15:21 +0100)
commit83eb64c85b80959549c114365016276f318afeb2
tree61115345021697f6b6dad4121cf0171b07b3f833
parent87069f4493b2101a71a92b7b9565f488a605a88f
drm: Drop dev->event_lock spinlock around faulting copy_to_user()

In

commit cdd1cf799bd24ac0a4184549601ae302267301c5
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Dec 4 21:03:25 2014 +0000

    drm: Make drm_read() more robust against multithreaded races

I fixed the races by serialising the use of the event by extending the
dev->event_lock. However, as Thomas pointed out, the copy_to_user() may
fault (even the __copy_to_user_inatomic() variant used here) and calling
into the driver backend with the spinlock held is bad news. Therefore we
have to drop the spinlock before the copy, but that exposes us to the
old race whereby a second reader could see an out-of-order event (as the
first reader may claim the first request but fail to copy it back to
userspace and so on returning it to the event list it will be behind the
current event being copied by the second reader).

Reported-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Cc: Takashi Iwai <tiwai@suse.de>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1448462343-2072-1-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
drivers/gpu/drm/drm_fops.c