Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 24 Nov 2004 18:07:43 +0000
From:      Gavin Atkinson <gavin.atkinson@ury.york.ac.uk>
To:        Nate Lawson <nate@root.org>
Cc:        freebsd-current@freebsd.org
Subject:   Re: Memory modified after free: Most recently used by acpitask
Message-ID:  <1101319662.56574.141.camel@buffy.york.ac.uk>
In-Reply-To: <41A4BB82.2010406@root.org>
References:  <1101312453.56574.122.camel@buffy.york.ac.uk> <41A4BB82.2010406@root.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 2004-11-24 at 16:49, Nate Lawson wrote:
> Gavin Atkinson wrote:
> > Hi,
> > 
> > Just got a panic on a 6-CURRENT (Thu Nov 18 16:36:35 GMT 2004) machine,
> > while copying a large amount of data around.
> > 
> > Seems to be an ACPI related reuse-after-free.  As far as I can tell, 20
> > bytes into the acpi_task structure is (int)ta_flags within the embedded
> > struct task, but I can't see use of this field in the ACPI code so ACPI
> > may be a red herring.
> > 
> > 
> > # cp -Rp /usr/* /var/usr
> > [about 10 minutes later]
> > Memory modified after free 0xc44a8420(28) val=0 @ 0xc44a8434
> > panic: Most recently used by acpitask
> 
> Unfortunately, the panic message doesn't tell you who modified it since 
> someone with a stray pointer (say, who allocated/freed it before acpi) 
> could overwrite it and it was only detected on the next malloc.  The way 
> I've found these is to boot -d (into ddb) and type "watch 0xc44a8420". 
> Then hit "c" to continue the boot.  Dump a "tr" any time the watchpoint 
> triggers and look for suspicious callers.

Sadly, I suspect it's not going to be that easy.  I've just had another
panic, same trigger and symptoms but different memory address.

Memory modified after free 0xc50441c0(28) val=0 @ 0xc50441d4
panic: Most recently used by acpitask

cpuid = 0
KDB: enter: panic
[thread 100111]
Stopped at      kdb_enter+0x2c: leave

I'll try taking the box to top-of-tree current in case it has already
been fixed - however that will probably have to wait until tomorrow now
as this machine cannot reboot without physical help.  Surely it seems
like quite a coincidence that both times it was 20 bytes into memory
once owned by acpitask, though?

Gavin



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1101319662.56574.141.camel>