[Buildroot] GRUB2 2.04 (buildroot 2022.02.1) failure reading sector errors

[Buildroot] GRUB2 2.04 (buildroot 2022.02.1) failure reading sector errors

[Stefan Agner]

Hi,

On Home Assistant OS we are moving from Barebox to GRUB2 as the
bootloader for UEFI based systems. The switch is currently on our beta
channel with several hundert of users. I have now 4 independent reports
where the system stopped booting after upgrading to the new OS version
with GRUB2. The systems show the same error: "error: failure reading
sector 0x9c900 from `hd0`." The sector number is not always the same,
but it seems to be always within the partition holding the kernel. We
are using compressed squashfs which might be a factor influencing the
problem.

Despite testing on several actual and virtual x86-64 systems I am unable
to reproduce so far.

At least one of the systems has been successfully tested with Debian 11
bullseye (which is using GRUB2 2.04 as well). That boot flow doesn't
make use of squashfs of course.

Anyone seen similar issues? Ideas how to get to the bottom of this?

I am considering moving to 2.06 to see if that fixes the problem, is
anyone working on 2.06 already? I see that buildroot is carring 150
patches on-top of vanilla GRUB2, so that requires some work.

Best regards,
Stefan
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] GRUB2 2.04 (buildroot 2022.02.1) failure reading sector errors

[Arnout Vandecappelle]

On 29/04/2022 16:47, Stefan Agner wrote:
> Hi,
> 
> On Home Assistant OS we are moving from Barebox to GRUB2 as the
> bootloader for UEFI based systems. The switch is currently on our beta
> channel with several hundert of users. I have now 4 independent reports
> where the system stopped booting after upgrading to the new OS version
> with GRUB2. The systems show the same error: "error: failure reading
> sector 0x9c900 from `hd0`." The sector number is not always the same,
> but it seems to be always within the partition holding the kernel. We
> are using compressed squashfs which might be a factor influencing the
> problem.
> 
> Despite testing on several actual and virtual x86-64 systems I am unable
> to reproduce so far.
> 
> At least one of the systems has been successfully tested with Debian 11
> bullseye (which is using GRUB2 2.04 as well). That boot flow doesn't
> make use of squashfs of course.
> 
> Anyone seen similar issues? Ideas how to get to the bottom of this?
> 
> I am considering moving to 2.06 to see if that fixes the problem, is
> anyone working on 2.06 already? I see that buildroot is carring 150
> patches on-top of vanilla GRUB2, so that requires some work.

  (Almost?) all of those patches are backports of upstream, because there were a 
large number of CVEs beginning of 2021 and there was no release which included 
fixes for them.  So updating grub shouldn't be that much of a problem.

  Regards,
  Arnout


> 
> Best regards,
> Stefan
> _______________________________________________
> buildroot mailing list
> buildroot@buildroot.org
> https://lists.buildroot.org/mailman/listinfo/buildroot
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] GRUB2 2.04 (buildroot 2022.02.1) failure reading sector errors

[Stefan Agner]

On 2022-04-30 12:46, Arnout Vandecappelle wrote:
> On 29/04/2022 16:47, Stefan Agner wrote:
<snip>
>> I am considering moving to 2.06 to see if that fixes the problem, is
>> anyone working on 2.06 already? I see that buildroot is carring 150
>> patches on-top of vanilla GRUB2, so that requires some work.
> 
>  (Almost?) all of those patches are backports of upstream, because
> there were a large number of CVEs beginning of 2021 and there was no
> release which included fixes for them.  So updating grub shouldn't be
> that much of a problem.

I've noticed that some of the patches patch files only available in the
tarball release (where autoconf is already run). Not sure how those have
been generated?

In an experiment we used Debian sid 2.06 GRUB2 build on an affected
system, and it seems to suffer the same issue. Which would mean that its
not related to one of the patches, and the issue is also not fixed in
GRUB 2.06. Debian booted fine otherwise on that system, so it must be
related to the squashfs we are using.

It probably still makes sense to move GRUB 2.06, just to be on a release
closer to the current state of development.

Best regards,
Stefan

> 
>  Regards,
>  Arnout
> 
> 
>>
>> Best regards,
>> Stefan
>> _______________________________________________
>> buildroot mailing list
>> buildroot@buildroot.org
>> https://lists.buildroot.org/mailman/listinfo/buildroot
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot