linux-kernel - Re: code sections beyond .text skipped from alternatives_smp_module

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <BANLkTi=DagQzcXQnHsVWjR0SeDDPnizVfQ@mail.gmail.com>
Date:	Wed, 22 Jun 2011 14:58:12 -0700
From:	Deep Debroy <ddebroy@...il.com>
To:	Konrad Rzeszutek Wilk <konrad.wilk@...cle.com>,
	linux-kernel@...r.kernel.org, kraxel@...hat.com
Subject: Re: code sections beyond .text skipped from alternatives_smp_module_add

On Wed, Jun 22, 2011 at 10:27 AM, Deep Debroy <ddebroy@...il.com> wrote:
> On Wed, Jun 22, 2011 at 6:21 AM, Konrad Rzeszutek Wilk
> <konrad.wilk@...cle.com> wrote:
>>> > Looking at the code, in module_finalize for x86, only .text seems to
>>> > be getting picked for the patching of lock prefixes while other
>>> > sections such as .exit.text or .init.text are not. Is there a reason
>>> > we skip the other *.text code sections from the lock patches? Would
>>> + Gerd Hoffmann who introduced the SMP patching code below back in Jan
>>> 2006 as part of 2.6.15.
>>
>> Whoa, long time ago.
>>
>>>
>>> Any comments on why patching of smp_lock prefixes should be restricted
>>> to .text and not other *.text code sections?
>>
>> It could be that at that time the .exit.text or .init.text did not exist.
>>
>> As in, the patching code just hasn't kept up. One way of checking that
>> is just finding the ancient 2.6.15 code and seeing if there is any
>> mention of those extra segments.
>>
>
> Thanks Konrad. One slight correction: after rechecking the kernel
> sources, it appears the smp lock prefix code first made it's
> appearance in the official trees during 2.6.18. In any case, going
> back even to 2.6.16 sources, layout_sections in module.c specially
> handled .init prefixed sections from the rest i.e. core sections.
> Further, the module struct in include/module/linux.h seems to have had
> members such as init_text_size which suggests atleast .init.text did
> exit back then as well. While I didn't find any crumbs in the code
> that point to the existence of a .exit.text (besides a function
> pointer called exit which most likely ended up in the .exit.text), the
> ELF headers for Centos 5.6 kernel objects (which uses the 2.6.18
> kernel) typically have a .exit.text.
>
>> Do you have a patch to fix this?
>>
>
> I can work on that. Just wanted to first make sure that there wasn't
> any specific reason to avoid patching non .text sections.
>
> Thanks,
> Deep
>

Some further digging through messages revealed a patch from Randy
Dunlap in June 2006: "[PATCH] ignore smp_locks section warnings from
init/exit code." Given this patch came in after the smp locking
hotpatching mechanism was introduced, there may have been an
assumption that instructions that results in entries in smp_locks
relocations in the object file should not exist in the init/exit.text
sections.

Thanks,
Deep
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/