[Issue] EFI runtime-wrapper is broken

[Ard Biesheuvel]

On Tue, 8 Nov 2022 at 13:26, Thorsten Leemhuis
<regressions at leemhuis.info> wrote:
>
> Hi, this is your Linux kernel regression tracker. CCing the regression
> mailing list, as it should be in the loop for all regressions, as
> explained here:
> https://www.kernel.org/doc/html/latest/admin-guide/reporting-issues.html
>
> On 28.10.22 11:39, Ard Biesheuvel wrote:
> > On Fri, 28 Oct 2022 at 11:38, Gavin Shan <gshan at redhat.com> wrote:
> >> On 10/28/22 5:21 PM, Ard Biesheuvel wrote:
> >>> On Fri, 28 Oct 2022 at 11:10, Gavin Shan <gshan at redhat.com> wrote:
> >>>> I ran into the following crash with v6.1.rc2. I'm not sure if it's known
> >>>> issue. Last good version is v6.0.rc6/rc7. The kernel configuration is attached.
> >>>
> >>> This is probably related to the patch below. It means the firmware is
> >>> buggy, and does not work as it should if the firmware never calls
> >>> SetVirtualAddressMap() [and doing so is explicitly described as
> >>> OPTIONAL by the EFI specification].
> >>>
> >>> The problem here is that other EFI implementations (primarily the
> >>> Qualcomm ones used on Windows-on-ARM laptops) are buggy too, and don't
> >>> tolerate SetVirtualAddressMap().
> >>>
> >>> Can you try whether reverting the patch below helps?
> >>
> >> I ran into the issue on server, equipped with Ampere CPUs. With d3549a938b73
> >> reverted, the crash disappeared.
> >
> > Thanks for confirming
> >
> > Can you please share the complete boot log and the DMI/SMBIOS data? Thanks.
>
> Gavin, did you ever send this?
>
> Or Ard, was there any progress to get the issue fixed?
>
> I ask because I have this on the list of tracked regression -- and I
> noticed there is another report about problems caused by d3549a938b73
> now from Alexandru Elisei:
> https://lore.kernel.org/lkml/Y2lAB508TrrjpDPi@monolith.localdoman/
>

Yes, this is essentially the exact same problem, only on a different
type of Ampere hardware

I have an idea how to fix this more comprehensively - I'll work with
Alexandru to converge on a fix and get it queued up.