Bug#1000481: Bullseye kernel hangs while initialising i915 gpu driver on old intel graphic chip

[Bogdan Veringioiu <bogdan.veringioiu@amano.eu>]

Hi

this issue affecting the LVDS connector was reported here:

https://gitlab.freedesktop.org/drm/intel/-/issues/7301

and fixed.

Any chance it will be backported in 5.10?

Thanks

On 10.06.22 15:01, Bogdan Veringioiu wrote:
> Hi Diederik,
>
> thanks for answering.
>
> I am attaching dmesg, Xorg, xrandr output as well as a listing of 
> /sys/class/drm directory of:
>
> - a buster install running kernel 4.19.98 which is working fine
>
> - a bullseye install running kernel 5.10.106 which is not working
>
> As mentioned intel_iommu parameter does not help me.
>
> After more tests, I understood what is happening under bullseye with 
> 5.10 kernel. The built-in monitor of the POS PC is connected via LVDS. 
> This LVDS connection is not detected/listed at all by the i915 driver 
> upon initialization on boot and the screen goes blank. See the Xorg + 
> xrand outputs. Under buster + kernel 4.19 it works perfectly.
>
> Additionaly (bullseye): If I hook an external VGA monitor using the 
> VGA connector in the back of the POS PC, that monitor is detected and 
> it is working fine, Xorg is using it as the main display. The built-in 
> LVDS monitor remains blank, the LVDS connection is not listed.
>
> Regards
>
> Bogdan
>
>
> On 10.06.22 12:35, Diederik de Haas wrote:
> > Control: found -1 linux/5.10.106-1
> > Control: found -1 linux/5.16.12-1~bpo11+1
> >
> > On Friday, 10 June 2022 07:57:12 CEST Bogdan Veringioiu wrote:
> > > the bug is still exists in kernel version 5.10.106 (32 bit).
> > >
> > > Also tried a newer kernel linux-image-5.16.0-0.bpo.4 from
> > > bullseye-backports, with the same result.
> > Thanks for that. Metadata updated accordingly.
> >
> > > Symptoms:
> > > The monitor  gets blank when the i915 changes resolution on boot.
> > > The PC remains accessible (ssh).
> > That seems different from what Ben reported, which may be useful info.
> > @Ben: can you tell whether you can ssh into your device or not?
> >
> > > The intel_iommu=off does not work.
> > >
> > > Some details about the system below:
> > >
> > > $ dmesg | grep 915
> > >
> > > [    8.816308] i915 0000:00:02.0: vgaarb: deactivate vga console
> > > [    8.846333] i915 0000:00:02.0: [drm] Failed to find VBIOS tables 
> > > (VBT)
> > When putting "i915 [drm] Failed to find VBIOS tables (VBT)" into a 
> > search
> > engine, most results were about GPU passthrough (VMs).
> > I don't see a (direct) connection with this bug, but thought I'd 
> > share this
> > observation in case it might give a clue to others.
> >
> > > [    8.846882] i915 0000:00:02.0: vgaarb: changed VGA decodes:
> > > olddecodes=io+mem,decodes=io+mem:owns=io+mem
> > > [    9.116269] i915 0000:00:02.0: [drm] Initialized overlay support.
> > > [    9.118527] [drm] Initialized i915 1.6.0 20200917 for 
> > > 0000:00:02.0 on
> > > minor 0
> > > [    9.163178] fbcon: i915drmfb (fb0) is primary device
> > > [    9.203620] i915 0000:00:02.0: [drm] fb0: i915drmfb frame buffer 
> > > device
> > Request to both Bogdan and Ben:
> > Can you share the full dmesg output of the 'failing' system and in 
> > case of a
> > working workaround (like "intel_iommu=off"), also the dmesg output of 
> > that?
> >
> > The full dmesg output may contain clues as to why this issue occurs.