linux-kernel - Re: [PATCH] drm/atmel-hlcdc: prevent divide by zero

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20190109125833.0af5dc52@bbrezillon>
Date:   Wed, 9 Jan 2019 13:00:08 +0100
From:   Boris Brezillon <bbrezillon@...nel.org>
To:     Peter Rosin <peda@...ntia.se>
Cc:     "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Alexandre Belloni <alexandre.belloni@...tlin.com>,
        David Airlie <airlied@...ux.ie>,
        Nicolas Ferre <nicolas.ferre@...rochip.com>,
        "dri-devel@...ts.freedesktop.org" <dri-devel@...ts.freedesktop.org>,
        Boris Brezillon <boris.brezillon@...tlin.com>,
        "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>
Subject: Re: [PATCH] drm/atmel-hlcdc: prevent divide by zero

On Wed, 9 Jan 2019 11:37:19 +0000
Peter Rosin <peda@...ntia.se> wrote:

> On 2019-01-09 11:12, Daniel Vetter wrote:
> > On Tue, Jan 08, 2019 at 12:31:36PM +0000, Peter Rosin wrote:  
> >> While trying to temporarily hide a plane, one thing that was attempted
> >> was to call (from libdrm)
> >>
> >> 	drmModeSetPlane(fd, plane_id, crtc_id, fb_id, 0,
> >> 			0, 0, 0, 0, 0, 0, 0, 0);
> >>
> >> This call causes a pair of "Division by zero in kernel." messages. Kill
> >> those messages, but preserve the current behaviour that also happen to
> >> make the plane disappear with the above call.
> >>
> >> Signed-off-by: Peter Rosin <peda@...ntia.se>
> >> ---
> >>  drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_plane.c | 14 ++++++++++----
> >>  1 file changed, 10 insertions(+), 4 deletions(-)
> >>
> >> Side note, when comparing with drm_atomic_helper_check_plane_state(), I get
> >> the feeling that the src rect should be clipped together with the crtc rect
> >> if/when clipping is needed. That function calls drm_rect_clip_scaled(), and
> >> the equivalent does not seem to happen here. Should clipping be performed
> >> on the src rect?  
> > 
> > Any reasons atmel can't switch over to that helper? Would compute a nice  
> > ->visible state variable for you ...  
> > -Daniel  
> 
> I have not researched that, and as stated above, I was not sure if that was the
> correct approach to begin with. Boris, any insights?

You can look at vc4_plane.c if you want an example of how this helper
can be used to get clipped dimensions of the plane.

> Does this code predate the
> helper or something?

Yes, and I must admit I'm not actively maintaining the driver, so there
are probably a few other things we could simplify by using generic
helpers.

> 
> Maybe there's some register bit that hides a plane explicitly, matching the
> ->visible state variable? I haven't looked at that either...

I don't think so, but you can simply disable the plane when ->visible
is false.