Re: Linux 7.1-rc1

[Hans de Goede <hansg@kernel.org>]

Hi,

On 29-Apr-26 08:12, Andy Shevchenko wrote:
> On Tue, Apr 28, 2026 at 09:37:31PM +0200, Rafael J. Wysocki wrote:
>> On Tue, Apr 28, 2026 at 8:46 PM Linus Torvalds
>> <torvalds@linux-foundation.org> wrote:
>>> On Tue, 28 Apr 2026 at 11:11, Luna Jernberg <droidbittin@gmail.com> wrote:
>>>>
>>>> have checked dmesg now, seems to be a module that's not loading as it should
>>>
>>> Hmm. Interesting, but I wouldn't expect that LPSS driver to matter.  I
>>> think it's mainly used for some random SoC boards, not by the display
>>> backlight code.
>>>
>>> And in the full dmesg I then see
>>>
>>>   intel-lpss 0000:00:15.2: enabling device (0000 -> 0002)
>>>   intel-lpss 0000:00:15.3: enabling device (0000 -> 0002)
>>>
>>> so I think it all should work regardless.
>>>
>>> Of course, maybe it does mess up i2c or something, which then messes
>>> up EDID or whatever, but I would expect the backlight to most likely
>>> be controlled by aome ACPI thing.
>>>
>>> So that module failure sounds to me like it should be unrelated - and
>>> might happen even on a working setup?
>>>
>>> I'm adding Rafael to the cc anyway, because he'd likely know about
>>> both the OpRegion conflict thing, but also maybe if something has
>>> changed in the ACPI backlight situation that might be the cause of
>>> your backlight issues...
> 
> Thanks for Cc'ing me.
> 
> The OpRegion conflict is known issue from the past. The background is that.
> The GEXP device described in the DSDT (main ACPI table) is a PCA953x compatible
> GPIO I²C expander which has a bare minimum driver written in ASL! We do not
> support such configuration in the kernel (we have C-based kernel driver for
> such hardware). The question is, what the configuration made that appears
> (I bet this conflict was on the working case). Because we have workarounds
> for PCI enumerated devices, but here it seems to come from ACPI-enumerated
> one?! Luna, please, share the output of the following run by root user (via
> some pastebin or alike service to avoid spamming too much the mailing lists):
> 
> - `acpidump -o dell-7390-skl.dat` # The *.dat file
> - `lspci -vv -nk`
> - `grep -H 15 /sys/bus/acpi/devices/*/status`
> - `cat /proc/iomem`
> 
> In either case, to avoid that OpRegion warning message we should add your
> platform to the quirk list (either to intel-lpss-pci.c or, surprisingly to
> me, to intel-lpss-acpi.c). One may see some in the PCI case already there.

We actually already have a patch pending upstream adding a similar quirk
mechanism to intel-lpss-acpi.c:

https://lore.kernel.org/platform-driver-x86/20260320000937.9177-2-tchatard@gmail.com/

Extending this with the DMI info from Luna's machine should fix
the resource conflict. But as Andy said this resource conflict is
a well known issue on this generation of laptops when using IPU3
MIPI cameras instead of a UVC camera.

Adding the quirk quirk will make the LPSS i2c-controller used
for the camera sensor but that should be completely unrelated to
the backlight problem.

Also note that just getting the i2c-controller resource conflict
resolved is NOT enough to get the cameras to work on these devices.

I expect that the working 7.0.2 dmesg will have the same resource-conflict.

Regards,

Hans



> 
>>> (Rafael, see
>>>
>>>   https://lore.kernel.org/all/CADo9pHg8c=rouBQAd0o0TzK4iZc88dUAhSfV7q2T3GCf2qjuiQ@mail.gmail.com/
>>>
>>> for dmesg etc information)
>>
>> Thanks!
>>
>> It would be good to know what the problem with the display is, specifically.
>>
>> Luna, can you send dmesg output from 7.0.2 for comparison?
>>
>> There are a few changes in 7.1-rc1 that touch the ACPI video bus driver:
>>
>> e18947038bf4 ACPI: driver: Do not set acpi_device_class() unnecessarily
>> 69652f32c9ac ACPI: event: Redefine acpi_notifier_call_chain()
>> 97892d5f0690 ACPI: driver: Do not set acpi_device_name() unnecessarily
>> 6a8e793ca8db ACPI: video: Consolidate pnp.bus_id workarounds handling
>> 9dc11faca245 ACPI: video: Rework checking for duplicate video bus devices
>>
>> but this one seems to be working:
>>
>> [    2.111324] ACPI: video: Video Device [GFX0] (multi-head: yes  rom:
>> no  post: no)
>> [    2.112585] input: Video Bus as
>> /devices/pci0000:00/acpi.video_bus.0/input/input5
>>
>> As for the OpRegion conflict message, it appears to come from
>> mfd_add_device() that invokes acpi_check_resource_conflict() and
>> passes its return value to the caller.  That caller is most likely
>> intel_lpss_probe() which then fails, so the probe fails for device
>> INT3446:00 which is an i2c controller AFAICS.
>>
>> I'm wondering if this also happens in 7.0.2.
>>
>> Adding Andy (for intel-lpss) and Hans who has more experience with the
>> ACPI video bus driver than I do.
>