linux-kernel - Re: try_to_freeze() called with IRQs disabled on ARM

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: <20110825121416.GB8883@n2100.arm.linux.org.uk>
Date:	Thu, 25 Aug 2011 13:14:16 +0100
From:	Russell King - ARM Linux <linux@....linux.org.uk>
To:	Tejun Heo <tj@...nel.org>
Cc:	"Rafael J. Wysocki" <rjw@...k.pl>,
	Mark Brown <broonie@...nsource.wolfsonmicro.com>,
	linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org
Subject: Re: try_to_freeze() called with IRQs disabled on ARM

On Wed, Aug 24, 2011 at 12:17:03AM +0200, Tejun Heo wrote:
> On Wed, Aug 24, 2011 at 12:13 AM, Russell King - ARM Linux
> <linux@....linux.org.uk> wrote:
> > Thereby entirely preventing threads from being frozen?  You're asking
> > me to effectively disable suspend/resume on an architecture where it's
> > heavily used.  That's not a good idea, and would be an out-right
> > regression.
> 
> Eh? So, it's supposed to enter refrigerator with IRQ disabled? Then,
> moving might_sleep() inside refrigerator() doesn't help either, does
> it? Then we should be doing is,
> 
> if (freezing() && IRQ disabled) {
>       bust on IRQ;
>       try_to_freeze();
>       replug IRQ;
> }
> 
> But, that can't be right. The current code isn't triggering warning
> from scheduler code, right? If the above is the case, it should be
> triggering that. What am I missing?

The scheduler code does not check for CPU IRQs being masked.  It just
checks the preempt count, nothing more.

might_sleep() on the other hand checks the preempt count _and_ CPU IRQ
mask state.

Note that your 'IRQ disabled' will always be true for ARM at this point
at the moment - and conditionalizing this won't help (see below).

Given that get_signal_to_deliver() already forcefully enables IRQs,
I think our syscall restarting is already horribly racy here in the
way I've previously described.

It also looks to me like get_signal_to_deliver() already handles the
freezing stuff, so does ARM even need this call here?  Maybe when:

commit fc558a7496bfab3d29a68953b07a95883fdcfbb1
Author: Rafael J. Wysocki <rjw@...k.pl>
Date:   Thu Mar 23 03:00:05 2006 -0800

    [PATCH] swsusp: finally solve mysqld problem

was introduced, every other architecture should have been updated for
that change...  So this call in the ARM signal handling code to
try_to_freeze() should just be deleted as it should've been done five
years ago.
--
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/