drm/vmwgfx: fix lock breakage
authorRob Clark <robdclark@gmail.com>
Thu, 30 Oct 2014 17:39:04 +0000 (13:39 -0400)
committerThomas Hellstrom <thellstrom@vmware.com>
Fri, 31 Oct 2014 08:44:35 +0000 (09:44 +0100)
commit21e88620aa21b48d4f62d29275e3e2944a5ea2b5
tree4ea16fa234bc54398da1518c16028189176d4995
parentd34d4d8af8d24afe4e16128e4e7f5c5967c89027
drm/vmwgfx: fix lock breakage

After:

commit d059f652e73c35678d28d4cd09ab2cec89696af9
Author:     Daniel Vetter <daniel.vetter@ffwll.ch>
AuthorDate: Fri Jul 25 18:07:40 2014 +0200

    drm: Handle legacy per-crtc locking with full acquire ctx

drm_mode_cursor_common() was switched to use drm_modeset_(un)lock_crtc()
which uses full aquire ctx.  So dropping/reaquiring the lock via
drm_modeset_(un)lock() directly isn't the right thing to do, as lockdep
kindly points out.

The 'FIXME's about sorting out whether vmwgfx *really* needs to lock-all
for cursor updates still apply.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Tested-by: Thomas Hellstrom <thellstrom@vmware.com>
drivers/gpu/drm/vmwgfx/vmwgfx_kms.c