linux-kernel - Re: [PATCH] avoid race condition in pick_next_task_fair in kernel/sched

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: <20101222151434.GW10557@genesis.frugalware.org>
Date:	Wed, 22 Dec 2010 16:14:34 +0100
From:	Miklos Vajna <vmiklos@...galware.org>
To:	Peter Zijlstra <peterz@...radead.org>
Cc:	Mike Galbraith <efault@....de>,
	shenghui <crosslonelyover@...il.com>,
	kernel-janitors@...r.kernel.org, linux-kernel@...r.kernel.org,
	mingo@...e.hu, Greg KH <greg@...ah.com>
Subject: Re: [PATCH] avoid race condition in pick_next_task_fair in
 kernel/sched_fair.c

On Wed, Dec 22, 2010 at 03:11:52PM +0100, Peter Zijlstra <peterz@...radead.org> wrote:
> > I started it with something like: 
> >  qemu -kernel foo-build/arch/x86/boot/bzImage -append "root=/dev/sda1
> > debug sched_debug ignore_loglevel sysrq_always_enabled console=ttyS0"
> > -hda systemd.img -serial stdio -m 1G
> > 
> > Where foo-build/ contains a kernel build using your .config.
> > 
> > I'll have a poke at it..
> 
> Hrm,. its not really wanting to start properly..
> 
> ---
> EXT4-fs (sda1): mounted filesystem with ordered data mode. Opts: (null)
> VFS: Mounted root (ext4 filesystem) readonly on device 8:1.
> Freeing unused kernel memory: 520k freed
> EXT4-fs (sda1): re-mounted. Opts: (null)
> md: stopping all md devices.
> Restarting system.
> machine restart
> ---
> 
> Does it _require_ initrd like muck? Or are there some modules that need
> to get built-in in order for the thing to boot?

Nope.

> This is an utter lack of error reporting here, no idea what's wrong.

I tried to do something really similar to your commandline:

qemu -enable-kvm -kernel kernel-build/arch/x86/boot/bzImage -append "root=/dev/sda1 debug sched_debug ignore_loglevel sysrq_always_enabled console=ttyS0" -hda systemd.img -serial stdio -m 1G -vnc :0

This boots up properly here, I can login using root/root from vnc.

qemu -enable-kvm -kernel kernel-build/arch/x86/boot/bzImage -append "root=/dev/sda1 debug sched_debug ignore_loglevel sysrq_always_enabled console=ttyS0 init=/bin/systemd" -hda systemd.img -serial stdio -m 1G -vnc :0

^ Only init=/bin/systemd added, and this results in a panic in most
cases. I'm attaching the stdout of qemu, showing the fail in
put_prev_task_fair.

$ qemu -version
QEMU emulator version 0.13.0, Copyright (c) 2003-2008 Fabrice Bellard

kernel-build is a git build using the config I already sent and after a
'git checkout v2.6.36'. I can try to build master as well, so far what I
saw is that the bug occurs there as well, but less frequently, so maybe
that's a bit harder to debug.

I also want to note that - at least on my machine - if I drop
-enable-kvm the bug is hard to reproduce, maybe that's because that way
it does not trigger a race condition or my machine is just too slow
without kvm and it triggers some timeout, changing the behaviour; I'm
not exactly sure.  (But again, I can reproduce the bug on real hardware,
so I don't think we have a kvm bug here.)

Thanks,

Miklos

View attachment "qemu.log" of type "text/plain" (24012 bytes)

Content of type "application/pgp-signature" skipped