X86 parallel bring up regression

X86 parallel bring up regression

[Mario Limonciello]

Hi,

We've recently found that there was a regression from 6.5 that broke 
resume from suspend-to-ram on some AMD platforms.

We bisected it and confirmed it's introduced by 0c7ffa32dbd6 
("x86/smpboot/64: Implement arch_cpuhp_init_parallel_bringup() and 
enable it").

Furthermore; we confirmed that on 6.6-rc6 adding cpuhp.parallel=0 to the 
kernel command line avoids the issue.

Here are bootup kernel logs from an affected system at 7e75178 (doesn't 
reproduce) and 0c7ffa32 (does reproduce).
https://gist.github.com/superm1/c251c0849956b8389309f03871fba091

Is it possible that this is caused by the system entering S3 with x2apic 
enabled but leaving with it disabled?

I notice in the resume path on a boot with cpuhp.parallel=0 that x2apic 
"explicitly" gets turned on again.

<snip>
smpboot: CPU 127 is now offline
ACPI: PM: Low-level resume complete
ACPI: PM: Restoring platform NVS memory
x2apic enabled
AMD-Vi: Virtual APIC enabled
AMD-Vi: Virtual APIC enabled
LVT offset 0 assigned for vector 0x400
Enabling non-boot CPUs ...
smpboot: Booting Node 0 Processor 1 APIC 0x2
masked ExtINT on CPU#1
ACPI: \_SB_.PLTF.C002: Found 2 idle states
CPU1 is up
smpboot: Booting Node 0 Processor 2 APIC 0x4
</snip>

I hypothesize this could be caused by .Lread_apicid finding x2apic 
disabled but then trying to read from apic_mmio_base which isn't 
initialized because x2apic was used (AFAICT apic_set_fixmap() never gets 
called in this case).

Thanks,

Re: X86 parallel bring up regression

[Linux regression tracking #adding (Thorsten Leemhuis)]

[TLDR: I'm adding this report to the list of tracked Linux kernel
regressions; the text you find below is based on a few templates
paragraphs you might have encountered already in similar form.
See link in footer if these mails annoy you.]

On 19.10.23 22:26, Mario Limonciello wrote:
> 
> We've recently found that there was a regression from 6.5 that broke
> resume from suspend-to-ram on some AMD platforms.
> 
> We bisected it and confirmed it's introduced by 0c7ffa32dbd6
> ("x86/smpboot/64: Implement arch_cpuhp_init_parallel_bringup() and
> enable it").
> 
> Furthermore; we confirmed that on 6.6-rc6 adding cpuhp.parallel=0 to the
> kernel command line avoids the issue.
> 
> Here are bootup kernel logs from an affected system at 7e75178 (doesn't
> reproduce) and 0c7ffa32 (does reproduce).
> https://gist.github.com/superm1/c251c0849956b8389309f03871fba091
> 
> Is it possible that this is caused by the system entering S3 with x2apic
> enabled but leaving with it disabled?
> 
> I notice in the resume path on a boot with cpuhp.parallel=0 that x2apic
> "explicitly" gets turned on again.
> 
> <snip>
> smpboot: CPU 127 is now offline
> ACPI: PM: Low-level resume complete
> ACPI: PM: Restoring platform NVS memory
> x2apic enabled
> AMD-Vi: Virtual APIC enabled
> AMD-Vi: Virtual APIC enabled
> LVT offset 0 assigned for vector 0x400
> Enabling non-boot CPUs ...
> smpboot: Booting Node 0 Processor 1 APIC 0x2
> masked ExtINT on CPU#1
> ACPI: \_SB_.PLTF.C002: Found 2 idle states
> CPU1 is up
> smpboot: Booting Node 0 Processor 2 APIC 0x4
> </snip>
> 
> I hypothesize this could be caused by .Lread_apicid finding x2apic
> disabled but then trying to read from apic_mmio_base which isn't
> initialized because x2apic was used (AFAICT apic_set_fixmap() never gets
> called in this case).

Thanks for the report. To be sure the issue doesn't fall through the
cracks unnoticed, I'm adding it to regzbot, the Linux kernel regression
tracking bot:

#regzbot ^introduced 0c7ffa32dbd6
#regzbot title x86: resume from suspend-to-ram broken on some AMD platforms
#regzbot ignore-activity

This isn't a regression? This issue or a fix for it are already
discussed somewhere else? It was fixed already? You want to clarify when
the regression started to happen? Or point out I got the title or
something else totally wrong? Then just reply and tell me -- ideally
while also telling regzbot about it, as explained by the page listed in
the footer of this mail.

Developers: When fixing the issue, remember to add 'Link:' tags pointing
to the report (the parent of this mail). See page linked in footer for
details.

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat)
--
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
That page also explains what to do if mails like this annoy you.

Re: X86 parallel bring up regression

[Mario Limonciello]

On 10/19/2023 15:26, Mario Limonciello wrote:
> Hi,
> 
> We've recently found that there was a regression from 6.5 that broke 
> resume from suspend-to-ram on some AMD platforms.
> 
> We bisected it and confirmed it's introduced by 0c7ffa32dbd6 
> ("x86/smpboot/64: Implement arch_cpuhp_init_parallel_bringup() and 
> enable it").
> 
> Furthermore; we confirmed that on 6.6-rc6 adding cpuhp.parallel=0 to the 
> kernel command line avoids the issue.
> 
> Here are bootup kernel logs from an affected system at 7e75178 (doesn't 
> reproduce) and 0c7ffa32 (does reproduce).
> https://gist.github.com/superm1/c251c0849956b8389309f03871fba091
> 
> Is it possible that this is caused by the system entering S3 with x2apic 
> enabled but leaving with it disabled?
> 
> I notice in the resume path on a boot with cpuhp.parallel=0 that x2apic 
> "explicitly" gets turned on again.
> 
> <snip>
> smpboot: CPU 127 is now offline
> ACPI: PM: Low-level resume complete
> ACPI: PM: Restoring platform NVS memory
> x2apic enabled
> AMD-Vi: Virtual APIC enabled
> AMD-Vi: Virtual APIC enabled
> LVT offset 0 assigned for vector 0x400
> Enabling non-boot CPUs ...
> smpboot: Booting Node 0 Processor 1 APIC 0x2
> masked ExtINT on CPU#1
> ACPI: \_SB_.PLTF.C002: Found 2 idle states
> CPU1 is up
> smpboot: Booting Node 0 Processor 2 APIC 0x4
> </snip>
> 
> I hypothesize this could be caused by .Lread_apicid finding x2apic 
> disabled but then trying to read from apic_mmio_base which isn't 
> initialized because x2apic was used (AFAICT apic_set_fixmap() never gets 
> called in this case).
> 
> Thanks,

AFAICT the issue is indeed that x2apic isn't re-enabled during s3
resume.

I've posted a series that I confirmed resolves the issue here:

https://lore.kernel.org/all/20231023160018.164054-1-mario.limonciello@amd.com/T/#t

Re: X86 parallel bring up regression

[Linux regression tracking #update (Thorsten Leemhuis)]

[TLDR: This mail in primarily relevant for Linux kernel regression
tracking. See link in footer if these mails annoy you.]

On 19.10.23 22:26, Mario Limonciello wrote:
> 
> We've recently found that there was a regression from 6.5 that broke
> resume from suspend-to-ram on some AMD platforms.

#regzbot monitor
https://lore.kernel.org/all/20231023160018.164054-1-mario.limonciello@amd.com/
#regzbot monitor
https://lore.kernel.org/all/20231026170330.4657-1-mario.limonciello@amd.com/
#regzbot ignore-activity
#regzbot resolve: turned out to be a BIOS bug

For details see https://lore.kernel.org/all/878r7imlb0.ffs@tglx/

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat)
--
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
That page also explains what to do if mails like this annoy you.