[REGRESSION] Platforms supporting SGX fail to kexec due to 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")

[REGRESSION] Platforms supporting SGX fail to kexec due to 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")

[Ashok Raj]

Hi 

96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")

Causes kexec failure. Backing out this change, kexec succeeds. Symptom is
it appears to hang, possibly hung at the panic. Although I have the
earlyprintk enabled, I don't see any console messages when new kernel
boots. 

Also tested turning off CONFIG_X86_SGX, the kernel with this commit
included also kexec's fine.

Booting from warm/cold reset has no issues. Only kexec to new kernel with
this change included and CONFIG_X86_SGX=y causes the kexec failure.

#regzbot introduced: 96c6b8f212a
-- 
Cheers,
Ashok

Re: [REGRESSION] Platforms supporting SGX fail to kexec due to 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")

[Borislav Petkov]

On Thu, Dec 28, 2023 at 04:33:49PM -0800, Ashok Raj wrote:
> Hi 
> 
> 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")
> 
> Causes kexec failure. Backing out this change, kexec succeeds. Symptom is
> it appears to hang, possibly hung at the panic. Although I have the
> earlyprintk enabled, I don't see any console messages when new kernel
> boots. 
> 
> Also tested turning off CONFIG_X86_SGX, the kernel with this commit
> included also kexec's fine.
> 
> Booting from warm/cold reset has no issues. Only kexec to new kernel with
> this change included and CONFIG_X86_SGX=y causes the kexec failure.
> 
> #regzbot introduced: 96c6b8f212a

Leaving in the whole mail for the people who you should have CCed. Lemme
CC them.

$ git show 96c6b8f212a
fatal: ambiguous argument '96c6b8f212a': unknown revision or path not in the working tree.

/me greps a bit.

Ah, you've dropped the first char of the sha1 - an 'e':

$ git show e96c6b8f212a
commit e96c6b8f212a510c9b22362de519f6e1d7920de5
Author: Song Shuai <songshuaishuai@tinylab.org>
Date:   Sat Jun 24 11:26:07 2023 +0800

    memblock: report failures when memblock_can_resize is not set

...

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette

Re: [REGRESSION] Platforms supporting SGX fail to kexec due to 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")

[Mike Rapoport]

Hi,

On Thu, Dec 28, 2023 at 04:33:49PM -0800, Ashok Raj wrote:
> Hi 
> 
> 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")
> 
> Causes kexec failure. Backing out this change, kexec succeeds. Symptom is
> it appears to hang, possibly hung at the panic. Although I have the
> earlyprintk enabled, I don't see any console messages when new kernel
> boots. 
> 
> Also tested turning off CONFIG_X86_SGX, the kernel with this commit
> included also kexec's fine.
> 
> Booting from warm/cold reset has no issues. Only kexec to new kernel with
> this change included and CONFIG_X86_SGX=y causes the kexec failure.

Can you add memblock=debug to the kernel command line and send logs for
normal boot and kexec with CONFIG_X86_SGX=y and e96c6b8f212a reverted?
 
> #regzbot introduced: 96c6b8f212a
> -- 
> Cheers,
> Ashok

-- 
Sincerely yours,
Mike.

Re: [REGRESSION] Platforms supporting SGX fail to kexec due to 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")

[Linux regression tracking (Thorsten Leemhuis)]

On 31.12.23 08:26, Mike Rapoport wrote:
> On Thu, Dec 28, 2023 at 04:33:49PM -0800, Ashok Raj wrote:
>> 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")
>>
>> Causes kexec failure. Backing out this change, kexec succeeds. Symptom is
>> it appears to hang, possibly hung at the panic. Although I have the
>> earlyprintk enabled, I don't see any console messages when new kernel
>> boots. 
>>
>> Also tested turning off CONFIG_X86_SGX, the kernel with this commit
>> included also kexec's fine.
>>
>> Booting from warm/cold reset has no issues. Only kexec to new kernel with
>> this change included and CONFIG_X86_SGX=y causes the kexec failure.
> 
> Can you add memblock=debug to the kernel command line and send logs for
> normal boot and kexec with CONFIG_X86_SGX=y and e96c6b8f212a reverted?

Ashok, you afaics never replied. Did you forget about it? Or was the
issue resolved later or never a regression in the first place? I for now
assume it's one of the latter and stop tracking this:

#regzbot inconclusive: radio silence from reporter
#regzbot ignore-activity

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

Re: [REGRESSION] Platforms supporting SGX fail to kexec due to 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")

[Ashok Raj]

Hi Mike,

On Thu, Feb 01, 2024 at 11:38:38AM +0100, Linux regression tracking (Thorsten Leemhuis) wrote:
> On 31.12.23 08:26, Mike Rapoport wrote:
> > On Thu, Dec 28, 2023 at 04:33:49PM -0800, Ashok Raj wrote:
> >> 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")
> >>
> >> Causes kexec failure. Backing out this change, kexec succeeds. Symptom is
> >> it appears to hang, possibly hung at the panic. Although I have the
> >> earlyprintk enabled, I don't see any console messages when new kernel
> >> boots. 
> >>
> >> Also tested turning off CONFIG_X86_SGX, the kernel with this commit
> >> included also kexec's fine.
> >>
> >> Booting from warm/cold reset has no issues. Only kexec to new kernel with
> >> this change included and CONFIG_X86_SGX=y causes the kexec failure.
> > 
> > Can you add memblock=debug to the kernel command line and send logs for
> > normal boot and kexec with CONFIG_X86_SGX=y and e96c6b8f212a reverted?
> 
> Ashok, you afaics never replied. Did you forget about it? Or was the
> issue resolved later or never a regression in the first place? I for now
> assume it's one of the latter and stop tracking this:

Sorry I went AWOL on this.. I did try the newer kernel's (6.8-rc1) and its not
happening any more, and kexec seems to work well. 

On the same problem kernel adding memblock=debug make it dissappear, and
even adding some code to that suspect area changed behavior.

I'm not able to reproduce it again. Sorry for the delay.

Cheers,
Ashok

Re: [REGRESSION] Platforms supporting SGX fail to kexec due to 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")

[Linux regression tracking #update (Thorsten Leemhuis)]

On 29.12.23 01:33, Ashok Raj wrote:
> 
> 96c6b8f212a ("memblock: report failures when memblock_can_resize is not set")
> 
> Causes kexec failure. Backing out this change, kexec succeeds. Symptom is
> it appears to hang, possibly hung at the panic. Although I have the
> earlyprintk enabled, I don't see any console messages when new kernel
> boots. 
> 
> Also tested turning off CONFIG_X86_SGX, the kernel with this commit
> included also kexec's fine.
> 
> Booting from warm/cold reset has no issues. Only kexec to new kernel with
> this change included and CONFIG_X86_SGX=y causes the kexec failure.
> 
> #regzbot introduced: 96c6b8f212a

Fix a typo and do a few other things while at it:

#regzbot introduced: e96c6b8f212a
#regzbot title: memblock: platforms supporting SGX fail to kexec
#regzbot ignore-activity

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.