From: Walter Alejandro Iglesias <wai@roquesor.com>
Subject: Re: Intel GPU, ghosting after zzz or ZZZ
To: Vitaliy Makkoveev <mvs@openbsd.org>
Cc: tech@openbsd.org
Date: Tue, 26 May 2026 21:37:06 +0200

On Tue, May 26, 2026 at 09:53:58PM +0300, Vitaliy Makkoveev wrote:
> On Tue, May 26, 2026 at 08:43:48PM +0200, Walter Alejandro Iglesias wrote:
> > On Tue, May 26, 2026 at 06:29:32PM +0000, Vitaliy Makkoveev wrote:
> > > On Tue, May 26, 2026 at 08:00:58PM +0200, Walter Alejandro Iglesias wrote:
> > > > On Tue, May 26, 2026 at 08:28:42PM +0300, Vitaliy Makkoveev wrote:
> > > > > On Tue, May 26, 2026 at 10:12:31AM +0200, Walter Alejandro Iglesias wrote:
> > > > > > Vitaliy,
> > > > > > 
> > > > > > > On Mon, May 25, 2026 at 10:00:44PM +0300, Vitaliy Makkoveev wrote:
> > > > > > > > I also have the problem with wscons screen burner on ALDERLAKE. I don't
> > > > > > > > know how describe it, so the screeshot [1]. After screen unblan, the
> > > > > > > > output happens only in upper left corner, the upper half of the first
> > > > > > > > line. The other screen contains garbage.
> > > > > > > > 
> > > > > > 
> > > > > > Just to clear up a doubt.  Could you try the diff below, please?
> > > > > > 
> > > > > > This diff flips the order in which rasops_show_screen and
> > > > > > drm_client_dev_restore are called.  Notice inteldrm_doswitch is called
> > > > > > before inteldrm_burner_cb:
> > > > > > 
> > > > > > void
> > > > > > inteldrm_attachhook(struct device *self)
> > > > > > {
> > > > > > ...
> > > > > > 	task_set(&dev_priv->switchtask, inteldrm_doswitch, dev_priv);
> > > > > > 	task_set(&dev_priv->burner_task, inteldrm_burner_cb, dev_priv);
> > > > > > 
> > > > > > 
> > > > > 
> > > > > With this diff after screen on, my ALDERLAKE machine has no garbage on
> > > > > screen, screen is the same as it was before screen off.
> > > > 
> > > > So, the weird half line from the picture does not happen?
> > 
> > This happens without your diff?  I mean, this half line is a side effect
> > of your diff?
> > 
> 
> We have 3 cases:
> 
>  1. Without any diffs this half line always presents.

I wanted to be sure that was not consequence of calling
rapsops_show_screen.

The Ctl+Alt+F* switching was to verify if what you're dealing with is
what I described as ghosting.  Let me show you my case:

   https://en.roquesor.com/Downloads/ghosting/original.jpg

You'll see in the next image parts of the 'cal -y' output overlapping
valid output (the kernel messages are another unrelated issue):

   https://en.roquesor.com/Downloads/ghosting/after_resume.jpg

After switching to and back from another tty (Ctl+Alt+F*) the ghosting
goes away:

   https://en.roquesor.com/Downloads/ghosting/after_switching_tty.jpg

My suspicion is that the weird half line in your Alder Lake is a
different issue.  We have to find out what causes it.  Anyways, what
fixes one version of the Intel GPU can break another; what you are
telling me now, that moving rasops_show_screen after
drm_client_dev_restore leaves your console without output, makes me
doubt my original patch.



 
> 
>  2. With my diff this half line always presents.
> 
>  3. With your diff below this half line always absent.
> 
> > > 
> > > Yes.
> > > 
> > > > 
> > > > > It has no output
> > > > > on screen until I do tty switch with Ctl + Alt + F*. After tty switch
> > > > > everything works as before sceen burner.
> > > > 
> > > > Do you mean that the console is unresponsive?
> > > > 
> > > 
> > > No output until Ctl + Alt + F* switch. For example I type 'ls' and
> > > nothing happens. However, after Ctl + Alt + F* switch the 'ls' and
> > > it's output appears on console. After this Ctl + Alt + F* switch
> > > sequence console works as it does.
> > > 
> > > Screen off/on ttyC0 -> Console without garbage but no output -> 
> > > Ctl + Alt + F2 -> Ctl + Alt + F1 -> ttyC0 works fine.
> > > 
> > > The console could be ttyC1, ttyC2 - there behavior is the same.
> > > 
> > > > > 
> > > > > > 
> > > > > > Index: i915_driver.c
> > > > > > ===================================================================
> > > > > > RCS file: /cvs/src/sys/dev/pci/drm/i915/i915_driver.c,v
> > > > > > diff -u -p -u -p -r1.30 i915_driver.c
> > > > > > --- i915_driver.c	30 Apr 2026 04:33:06 -0000	1.30
> > > > > > +++ i915_driver.c	26 May 2026 07:57:22 -0000
> > > > > > @@ -2077,8 +2077,8 @@ inteldrm_doswitch(void *v)
> > > > > >  	struct drm_device *dev = &dev_priv->drm;
> > > > > >  	struct rasops_info *ri = &dev_priv->ro;
> > > > > >  
> > > > > > -	rasops_show_screen(ri, dev_priv->switchcookie, 0, NULL, NULL);
> > > > > >  	drm_client_dev_restore(dev);
> > > > > > +	rasops_show_screen(ri, dev_priv->switchcookie, 0, NULL, NULL);
> > > > > >  
> > > > > >  	if (dev_priv->switchcb)
> > > > > >  		(*dev_priv->switchcb)(dev_priv->switchcbarg, 0, 0);
> > > > > > @@ -2094,8 +2094,8 @@ inteldrm_enter_ddb(void *v, void *cookie
> > > > > >  	if (cookie == ri->ri_active)
> > > > > >  		return;
> > > > > >  
> > > > > > -	rasops_show_screen(ri, cookie, 0, NULL, NULL);
> > > > > >  	drm_client_dev_restore(dev);
> > > > > > +	rasops_show_screen(ri, cookie, 0, NULL, NULL);
> > > > > >  }
> > > > > >  
> > > > > >  int
> > > > > > 
> > > > > > 
> > > > > > -- 
> > > > > > Walter
> > > > > > 
> > > > 
> > > > -- 
> > > > Walter
> > 
> > -- 
> > Walter
> > 

-- 
Walter