Re: [BUG] SPD5118 Intermittent MR11 Corruption During Suspend/Resume

[Guenter Roeck]

Hi Matthew,

On Thu, Sep 03, 2026 at 07:24:12PM -0500, Matthew Bettencourt wrote:
> Hello Guenter,
> 
> >> - Do you have the latest BIOS installed on your system ?
> >>   Some statements online seem to suggest that the problem
> >>   _may_ be fixed with the latest BIOS on at least some
> >>   of the affected systems.
> 
> Yup as of Sept 3 2026 I am running the latest BIOS, 4.43. I think a new one
> is coming soonish as Asrock has started rolling out 4.50 on other x870
> motherboards
> 
> >> - Please boot the affected system with "usbcore.autosuspend=-1"
> >>   kernel command line parameter.
> >>   There are some claims online that this might resolve the problem.
> >>   I have no idea why that would be the case, but I guess it does
> >>   not hurt trying. Of course, if that _does_ work, we have yet another
> >>   mystery to solve.
> 
> I think we have another mystery to solve. I enabled "usbcore.autosuspend=-1"
> as a boot parameter and the errors have cleared up when I suspend/resume. I
> doubled checked this by removing "usbcore.autosuspend=-1" from my boot
> parameters and the errors came back.
> 

Outch. The mystery (or misery) deepens. So we kind of know that this has
something to do with ACPI _and_ with USB, such as ACPI accessing (or trying
to access) a device on the I2C bus when suspending (or trying to suspend)
a USB port / bus.

As Armin has pointed out, it isn't safe for the piix4 controller driver to
access the hardware when its memory space is reserved by ACPI, even if that
ACPI access turns out to be plain wrong. My ACPI / DSL knowledge isn't good
enough to figure out what exactly is happening, but even if we do that it
won't help: The PIIX4 driver must not load if its memory space is reserved
by ACPI, or it must access devices on the I2C bus through ACPI.

Thanks,
Guenter