linux-kernel - Re: Linux 4.15-rc2: Regression in resume from ACPI S3

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: <20171210213516.GA5096@amd>
Date:   Sun, 10 Dec 2017 22:35:18 +0100
From:   Pavel Machek <pavel@....cz>
To:     Linus Torvalds <torvalds@...ux-foundation.org>
Cc:     Zhang Rui <rui.zhang@...el.com>,
        Andrew Lutomirski <luto@...nel.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Jarkko Nikula <jarkko.nikula@...ux.intel.com>,
        "Rafael J. Wysocki" <rjw@...ysocki.net>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        the arch/x86 maintainers <x86@...nel.org>
Subject: Re: Linux 4.15-rc2: Regression in resume from ACPI S3

On Sun 2017-12-10 13:28:50, Linus Torvalds wrote:
> On Sun, Dec 10, 2017 at 12:43 PM, Pavel Machek <pavel@....cz> wrote:
> >
> > For the record... this should fix it. Tested on x60. More tests pending.
> 
> This can't be right.
> 
> At the very least, now the comment is wrong. And the comment does seem
> relevant for 32-bit too:

Well, take a look at orignal patch. I'm reverting 32-bit code to
v4.15-rc1 version, while keeping 64-bit code at v4.15-rc3
version. Yes, my brain hurts from looking at the code :-(.

In the meantime, I did short test on 64-bit machine. No ill effect observed.

Hmm. Aha. Yes, the comment is wrong... as it was in wrong in -rc1.

> > -       fix_processor_context();
> > -
> >         /*
> >          * Restore segment registers.  This happens after restoring the GDT
> >          * and LDT, which happen in fix_processor_context().
> 
> Notice? You've moved down the 32-bit fix_processor_context() call to
> after the loadsegment() calls, which smells wrong.

Yeah, I did. There's where it was in v4.15-rc1, and that's what ws
working for me. 

> That said, this *all* smells wrong. Why is there a special
> fix_processor_context() function at all with different 32-bit and
> 64-bit behavior? This code is all written to be maximally confusing.
> 
> I think this could do with some re-org to make it more logical. That
> "some random things done in fix_processor_context(), other random
> things done directly in __restore_processor_state()" makes no sense at
> all to me. There's no logic to what is done where.

I have to agree.
									Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Download attachment "signature.asc" of type "application/pgp-signature" (182 bytes)