Re: Libertas related kernel crash

[Dan Williams <dcbw@redhat.com>]

On Tue, 2009-11-10 at 07:54 +0100, Daniel Mack wrote:
> On Mon, Nov 09, 2009 at 10:51:33PM -0800, Dan Williams wrote:
> > On Mon, 2009-11-09 at 16:53 +0100, Daniel Mack wrote:
> > > On Thu, Nov 05, 2009 at 01:05:49PM +0100, Daniel Mack wrote:
> > > > On an ARM (PXA300) embdedded platform with a libertas chip connected via
> > > > SDIO, we happen to see the kernel Ooops below once in a while.
> > > > 
> > > > Any pointer on where to dig?
> > > 
> > > Some more input on this. Oopses similar to the one below are likely
> > > triggered when switching from Ad-hoc to managed mode multiple times in a
> > > row, and something seems corrupt the memory badly. I've searched for the
> > > obvious (double frees, out-of-bound writes etc), but I couldn't find
> > > anything yet. It is, however, related to the wireless core and/or the
> > > libertas driver.
> > 
> > Probably just libertas.  Any chance you can enable debugging options
> > (either in libertas, or in the kernel allocator) to help narrow down the
> > issue?
> 
> Sure. I already enabled DEBUG_VM, but that didn't spit out anything
> before it crashes. Which other debug flags would you recommend to set?

Offhand I don't actually know; but ISTR there used to be slab debugging
options in the kernel that could help find memory corruption issues.  I
can't think of anything in libertas offhand that would trigger this, but
maybe something is not getting properly cleaned up when switching modes?

Dan