drm: Don't pass negative delta to ktime_sub_ns()

commit e91abf80a0998f326107874c88d549f94839f13c upstream.

It takes an unsigned value. This happens not to blow up on 64-bit
architectures, but it does on 32-bit, causing
drm_calc_vbltimestamp_from_scanoutpos() to calculate totally bogus
timestamps for vblank events. Which in turn causes e.g. gnome-shell to
hang after a DPMS off cycle with current xf86-video-ati Git.

[airlied: regression introduced in drm: use monotonic time in drm_calc_vbltimestamp_from_scanoutpos]

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=59339
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=59836
Tested-by: shui yangwei <yangweix.shui@intel.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c
index 8bcce7866d368e12df7c86edb17ff235ba58f4a7..f92da0a32f0d30d02fe8afeb09f5272a16b8a060 100644 (file)
--- a/drivers/gpu/drm/drm_irq.c
+++ b/drivers/gpu/drm/drm_irq.c
@@ -708,7 +708,10 @@ int drm_calc_vbltimestamp_from_scanoutpos(struct drm_device *dev, int crtc,
        /* Subtract time delta from raw timestamp to get final
         * vblank_time timestamp for end of vblank.
         */
-       etime = ktime_sub_ns(etime, delta_ns);
+       if (delta_ns < 0)
+               etime = ktime_add_ns(etime, -delta_ns);
+       else
+               etime = ktime_sub_ns(etime, delta_ns);
        *vblank_time = ktime_to_timeval(etime);
 
        DRM_DEBUG("crtc %d : v %d p(%d,%d)@ %ld.%ld -> %ld.%ld [e %d us, %d rep]\n",