[releng_6] mpt(4) Memory modified after free panic

[Scott]

On Tue, 27 Sep 2005, Pawel Worach wrote:

> On 9/27/05, Kris Kennaway <kris at obsecurity.org> wrote:
>> On Mon, Sep 26, 2005 at 11:55:53PM +0200, Pawel Worach wrote:
>>> Trying to use a mpt controller with only one disk attached so it's not
>>> possible to configure a RAID-1 volume.
>>> Trying to boot 6.0-BETA1 install cd results in this panic. Should it
>>> possible to use a single disk behind an mpt(4) with the updated driver? This
>>> configuration works fine on 5.4.
>>
>> You should retry with 6.0-BETA5 - dozens of bugs have been fixed since
>> then.  If it persists, you should set up DEBUG_MEMGUARD to watch the
>> M_BUS malloc type..this requires some minor source file editing and a
>> kernel recompile (you can build it on e.g. 5.4).
>
> False alarm, BETA5 with WITNESS and INVARIANTS re-enabled works just fine.
> BETA1 was the only thing I had handy the first time and since mpt(4)
> was unchanged I thought it would still be broken, I'll try to think
> less next time :)
> For future reference, how do I get hold of the M_BUS malloc type in
> kern_malloc.c ?
>
> --
> Pawel
>
mpt likely is not the problem.  It's the one that catches the misbehavior.

Scott