linux-kernel - Re: [3.4-rc3] Thread overran stack, or stack corrupted

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]

Date:	Wed, 18 Apr 2012 10:02:06 -0700
From:	Linus Torvalds <torvalds@...ux-foundation.org>
To:	Dave Jones <davej@...hat.com>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	Linux Kernel <linux-kernel@...r.kernel.org>
Subject: Re: [3.4-rc3] Thread overran stack, or stack corrupted

On Tue, Apr 17, 2012 at 8:19 PM, Dave Jones <davej@...hat.com> wrote:
>
> So now that I'm instrumenting it, it's taking a lot longer to trigger
> (how typical). Almost 6 hours in though, it's down to 72 bytes, and spewed
> the traces below, which look pretty.. deep.

Yeah. Sadly, they are less useful than I was hoping for. It's not some
single deep call-chain, it's almost all debug stuff and the "did we
release the RCU lock" or preemption checks, which I guess makes sense.
You have tons of options enabled in your kernel that makes for deeper
stack traces, and then all the interesting stuff gets overwritten by
what happened later.

For example, it looks like you have the USB serial console enabled,
and some of those deep stack traces are about that - and largely got
overwritten by the "dump_trace()" logic itself. So dump_trace()
printing stuff out also ended up overwriting the stack trace that we
were interested in.

I assume you have USB serial console on for a reason (ie: great for
catching oopses before the machine dies), but in this case it hurts.

Could you try just adding a

   console_lock();
   ...
   console_unlock();

around the show_trace() call. That will force the code to not actually
call down to the console layer until after the console_unlock(), so
the printing of the stack trace won't affect the stack *too* much.

That said, if you get the function tracer thing working, that will
give much nicer backtraces.

                   Linus
--
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/