On Sun, Jul 05, 2015 at 08:21:15PM -0500, Walter Daugherity wrote:
> >Synopsis: NMI crash in ieee80211_recv_auth called from malo driver
> 
> 
> >Category: kernel
> 
> 
> >Environment:
> 
>         System      : OpenBSD CVS + debug flags
> 
>         Details     : OpenBSD 5.8-beta (GENERIC.DEBUG) #1: Fri Jul  3
> 17:24:02 CDT 2015 [email protected]
> :/usr/src/sys/arch/i386/compile/GENERIC.DEBUG
> 
>         Architecture: OpenBSD.i386
> 
>         Machine     : i386
> 
> 
> >Description: This is a long-standing problem which I have now narrowed
> down to a 100% reproducible NMI crash in function ieee80211_recv_auth in
> source file ieee80211_input.c, called from malo.c.

Did you see my reply back in April?
http://marc.info/?l=openbsd-bugs&m=142849629211619&w=2
 
As stated there, I very much doubt this problem is rooted anywhere near the
malo driver and net80211. I've tested malo on both amd64 and i386 machines
based on your prior and I could not reproduce the issue.

An NMI indicates the chipset on the motherboard is detecting a non-recoverable
error of some sort and can't continue. I don't know what could be causing this.

Perhaps it's worth noting that your machine has several address conflicts:

> 0:0:0: mem address conflict 0xffffff0/0x10
> 0:0:0: io address conflict 0x10000e10/0x10
> 0:0:0: io address conflict 0x62000146/0x2

One interesting date point would be if your hardware setup works as
expected with OpenBSD 4.1 (the release where malo first appeared).

> I cranked up several debug flag levels for ieee80211 and malo (which
> exposed a syntax error: in malo.c the line
> 
>      sc->sc_txring.stat, sc->sc_dev.dv_xname);

Thanks, I'll fix this.

Reply via email to