Re: [PATCH 4.4 00/37] 4.4.110-stable review

[Pavel Tatashin]

[    6.159992] Code: 89 83 78 06 01 00 b8 01 00 00 00 5b 41 5c 41 5d
5d c3 0f 1f 80 00 00 00 00 0f 1f 44 00 00 55 31 d2 48 8b 87 c8 00 00
00 48 89 e5 <f0> 0f c1 50 0c 89 97 d0 00 00 00 83 e2 01 b8 01 00 00 00
74 1d

Also, attached is the full console output.

Thank you,
Pavel

On Thu, Jan 4, 2018 at 4:37 PM, Hugh Dickins <hughd@xxxxxxxxxx> wrote:
> On Thu, Jan 4, 2018 at 1:23 PM, Pavel Tatashin
> <pasha.tatashin@xxxxxxxxxx> wrote:
>> I tried cherry picking
>>  435086b36f62 x86/vsyscall/64: Explicitly set _PAGE_USER in the
>> pagetable hierarchy
>>
>> on top of 4.4.110-rc1, (needed to resolve a small 5level table to
>> 4level page table conflict). Unfortunately, this does not solve the
>> panic/hanging problem I reported. For some reason I do not see the
>> panic message anymore. Machine hangs here:
>>
>> [    5.023052] zswap: loaded using pool lzo/zbud
>> [    5.023063] page_owner is disabled
>> [    5.026492] Key type trusted registered
>> [    5.029325] Key type encrypted registered
>> [    5.029330] ima: No TPM chip found, activating TPM-bypass!
>> [    5.029365] evm: HMAC attrs: 0x1
>> [    5.034696] rtc_cmos 00:00: setting system clock to 2018-01-04
>> 21:20:34 UTC (1515100834)
>> [    5.216862] Freeing unused kernel memory: 1856K
>> <hang>
>>
>> And reboots after about half a minute.
>
> Thanks for trying, but yes, I wouldn't expect a straight cherry-pick
> of that to work in the context of 4.4.110: it needs to be
> cherry-picked "in principle".  Which Borislav has done, and I'll
> forward you his (not yet reviewed) patch too, but frankly I've much
> less hope that it will help your crash than Thomas's.
>
> So please revert that cherry-pick; and if Borislav's patch doesn't
> help, if you can send us a "Code:" line from the crash, that may still
> give us more to go on.
>
> As Linus remarked earlier, "The PTI patches obviously change percpu
> stuff, but this looks like an odd place for that to manifest".
> Exactly: segfault and panic when starting init is a "normal" symptom
> when we get something wrong with Kaiser/PTI, but a kthread crashing in
> dyntick_save_progress_counter is something new to me.
>
> Hugh

Attachment: console_panic.output.gz
Description: GNU Zip compressed data