Bug#1036530: linux-signed-amd64: Hard lock up of system

[Salvatore Bonaccorso <carnil@debian.org>]

Hi Nick,

On Thu, May 25, 2023 at 08:23:15AM +0900, Nick Hastings wrote:
> Hi,
> 
> * Salvatore Bonaccorso <carnil@debian.org> [230524 19:26]:
> >
> > Given you were able to bisect it so far, can you try to isolate the
> > commit from the merge commit causing it?
> 
> I guess I can try. The commit message states:
> 
> Merge: c77f54a9bcec a1cf1fd62ae7 562163595a91 018d6711c26e 6cc401be1648
> 
> Is there a way extract out each of those?

Th way i usuually get all commits from a merge request is

git log --oneline $mergecommit^$mergecommit^2

though here we have three merge commits, merged with one merge commit
on top, so you would go down the merges of the acpi-properties,
acpi-tables, acpi-x86 and acpi-soc branches. Those are those:

    * acpi-properties:
      ACPI: property: Silence missing-declarations warning in apple.c

    * acpi-tables:
      ACPI: HMAT: Drop unused dev_fmt() and redundant 'HMAT' prefix
      ACPI: tables: FPDT: Don't call acpi_os_map_memory() on invalid phys address

    * acpi-x86:
      ACPI: x86: Add a quirk for Dell Inspiron 14 2-in-1 for StorageD3Enable

    * acpi-soc:
      ACPI: LPSS: Deduplicate skipping device in acpi_lpss_create_device()
      ACPI: LPSS: Replace loop with first entry retrieval

> > One remotely related might be "ACPI: x86: Add a quirk for Dell
> > Inspiron 14 2-in-1 for StorageD3Enable".
> 
> Manually looking at the diff with
> git diff e996c7e01892ac18ec0db447294d4f591c325efe~  e996c7e01892ac18ec0db447294d4f591c325efe 
> I guess that means the following:
> 
> --- a/drivers/acpi/x86/utils.c
> +++ b/drivers/acpi/x86/utils.c
> @@ -207,9 +207,26 @@ static const struct x86_cpu_id storage_d3_cpu_ids[] = {
>         {}
>  };
>  
> +static const struct dmi_system_id force_storage_d3_dmi[] = {
> +       {
> +               /*
> +                * _ADR is ambiguous between GPP1.DEV0 and GPP1.NVME
> +                * but .NVME is needed to get StorageD3Enable node
> +                * https://bugzilla.kernel.org/show_bug.cgi?id=216440
> +                */
> +               .matches = {
> +                       DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
> +                       DMI_MATCH(DMI_PRODUCT_NAME, "Inspiron 14 7425 2-in-1"),
> +               }
> +       },
> +       {}
> +};
> +
>  bool force_storage_d3(void)
>  {
> -       return x86_match_cpu(storage_d3_cpu_ids);
> +       const struct dmi_system_id *dmi_id = dmi_first_match(force_storage_d3_dmi);
> +
> +       return dmi_id || x86_match_cpu(storage_d3_cpu_ids);
>  }

That probably won't work actually as the code has been refactored
substantiantly after the commit. 

In the ideal case we could confirm the quirk change is the responsable
commit, so we can make upstream aware.

Regards,
Salvatore