linux-kernel - Re: [PATCH] drm: send vblank event with the attached sequence rather than current

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

Date:   Fri, 3 Dec 2021 09:25:20 +0200
From:   Ville Syrjälä <ville.syrjala@...ux.intel.com>
To:     Mark Yacoub <markyacoub@...omium.org>
Cc:     dri-devel@...ts.freedesktop.org, chunkuang.hu@...nel.org,
        Thomas Zimmermann <tzimmermann@...e.de>,
        David Airlie <airlied@...ux.ie>, jason-jh.lin@...iatek.com,
        linux-kernel@...r.kernel.org, tzungbi@...gle.com,
        seanpaul@...omium.org, matthias.bgg@...il.com,
        Mark Yacoub <markyacoub@...gle.com>
Subject: Re: [PATCH] drm: send vblank event with the attached sequence rather
 than current

On Thu, Dec 02, 2021 at 10:11:55AM -0500, Mark Yacoub wrote:
> From: Mark Yacoub <markyacoub@...gle.com>
> 
> [Why]
> drm_handle_vblank_events loops over vblank_event_list to send any event
> that is current or has passed.
> More than 1 event could be pending with past sequence time that need to
> be send. This can be a side effect of drivers without hardware vblank
> counter and they depend on the difference in the timestamps and the
> frame/field duration calculated in drm_update_vblank_count. This can
> lead to 1 vblirq being ignored due to very small diff, resulting in a
> subsequent vblank with 2 pending vblank events to be sent, each with a
> unique sequence expected by user space.
> 
> [How]
> Send each pending vblank event with the sequence it's waiting on instead
> of assigning the current sequence to all of them.
> 
> Fixes igt@..._flip "Unexpected frame sequence"
> Tested on Jacuzzi (MT8183)
> 
> Signed-off-by: Mark Yacoub <markyacoub@...omium.org>
> ---
>  drivers/gpu/drm/drm_vblank.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/drm_vblank.c b/drivers/gpu/drm/drm_vblank.c
> index 3417e1ac79185..47da8056abc14 100644
> --- a/drivers/gpu/drm/drm_vblank.c
> +++ b/drivers/gpu/drm/drm_vblank.c
> @@ -1902,7 +1902,7 @@ static void drm_handle_vblank_events(struct drm_device *dev, unsigned int pipe)
>  
>  		list_del(&e->base.link);
>  		drm_vblank_put(dev, pipe);
> -		send_vblank_event(dev, e, seq, now);
> +		send_vblank_event(dev, e, e->sequence, now);

This doesn't look right. The timestamp corresponds to 'seq' not
e->sequence (ie. whatever sequqnece number the user asked to wait
for).

>  	}
>  
>  	if (crtc && crtc->funcs->get_vblank_timestamp)
> -- 
> 2.34.0.rc2.393.gf8c9666880-goog

-- 
Ville Syrjälä
Intel