On Tue, Jul 07, 2015 at 04:21:50PM +0800, John Hunter wrote:

From: Zhao Junwang zhjwpku@gmail.com

legacy setcrtc ioctl does take a 32 bit value which might indeed overflow

the checks of crtc_req->x > INT_MAX and crtc_req->y > INT_MAX aren't needed any more with this

Cc: Daniel Vetter daniel.vetter@ffwll.ch Signed-off-by: Zhao Junwang zhjwpku@gmail.com

---

drivers/gpu/drm/drm_crtc.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c index b69ed97..c307435 100644 --- a/drivers/gpu/drm/drm_crtc.c +++ b/drivers/gpu/drm/drm_crtc.c @@ -2706,8 +2706,11 @@ int drm_mode_setcrtc(struct drm_device *dev, void *data, if (!drm_core_check_feature(dev, DRIVER_MODESET)) return -EINVAL;

• /* For some reason crtc x/y offsets are signed internally. */
• if (crtc_req->x > INT_MAX || crtc_req->y > INT_MAX)

• /*

• * Legacy setcrtc ioctl does take a 32 bit value which might
  

• * overflow
  

That legacy takes 32bit is clear, the important part to mention is that atomic only takes a 16.16 fixed point for this. And to avoid surprises in for atomic drivers it's better to check this here. Actually this even holds for just universal planes, no need for full-blown atomic. So what about instead

/* * Universal lane src offsets are only 16.16, prevent havoc for * drivers using universal plane code internally. */ -Daniel

• */
  

• if (crtc_req->x & 0xffff0000 || crtc_req->y & 0xffff0000) return -ERANGE;

  drm_modeset_lock_all(dev);

-- 1.7.10.4

---

dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel