linux-kernel - Re: kernel/printk/printk.c: Invalid access when buffer wraps around?

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: <14d57dba-e293-11c2-67b2-18313a3a9b34@cern.ch>
Date:	Sun, 31 Jul 2016 15:26:41 +0200
From:	Vincent Brillault <vincent.brillault@...n.ch>
To:	Sergey Senozhatsky <sergey.senozhatsky@...il.com>
CC:	Andrew Morton <akpm@...ux-foundation.org>,
	Petr Mladek <pmladek@...e.com>,
	Andrey Ryabinin <aryabinin@...tuozzo.com>,
	Kees Cook <keescook@...omium.org>,
	Thierry Reding <treding@...dia.com>,
	Geliang Tang <geliangtang@....com>, Tejun Heo <tj@...nel.org>,
	Ivan Delalande <colona@...sta.com>,
	<linux-kernel@...r.kernel.org>
Subject: Re: kernel/printk/printk.c: Invalid access when buffer wraps around?

Dear Sergey,

> sorry for long reply. do you see this in practice?

No, I've only thought of the bug will trying to adapt this code to build
a separate cyclic buffer in a dedicated kernel module.

> the first printk()->console_unlock() to notice `seen_seq != log_next_seq`
> will wakeup a task from log_wait, sleeping on
> 	wait_event_interruptible(seq != log_next_seq)

Yes, but a task could be not waiting to read reading while still having
open /dev/kmsg (e.g. after having read it in O_NONBLOCK)

> so I believe your assumption here is that we wrap around and then fill up
> the log_buf again without waking up the klogd even once, correct?
> 
> 	CPU0				CPU1
> 
> 	console_lock();
> 	printk();
> 	...				devkmsg_read();
> 	printk();
> 	console_unlock();
> 
> like the above?

Mmm, I did not think of such a case, which might be possible. I was more
thinking of a userland daemon reading the buffer (via /dev/kmsg) in
non-blocking mode and only pulling from time to time. I agree that this
is probably not seen often, which could explain why nobody can see it in
practice.

Thanks for your time,
Vincent Brillault

Download attachment "signature.asc" of type "application/pgp-signature" (820 bytes)