On Wed, Jan 13, 2016 at 03:29:24PM +0200, Timo Myyrä wrote:
> Stefan Sperling <s...@stsp.name> writes:
>
> > On Wed, Jan 13, 2016 at 01:30:39PM +0200, timo.my...@wickedbsd.net wrote:
> >
> >> >Synopsis: kernel panic when booting MP kernel, SP works
> >> >Category: kernel
> >> >Environment:
> >> System : OpenBSD 5.9
> >> Details : OpenBSD 5.9-beta (GENERIC) #1679: Fri Jan 8 23:44:41 MST
> >> 2016
> >>
> >> dera...@amd64.openbsd.org:/usr/src/sys/arch/amd64/compile/GENERIC
> >>
> >> Architecture: OpenBSD.amd64
> >> Machine : amd64
> >> >Description:
> >> Updated to latest snapshot and I get kernel panic when booting the
> >> GENERIC.MP kernel.
> >> SP kernel works.
> >>
> >> kernel: integer divide fault trap, code=0
> >> Stopped at ar5008_set_delta_slope+0x40: idivl %ecx,%eax
> >
> > It's probably crashing because c->ic_freq is zero here:
> >
> > /* Set Delta Slope (exponent and mantissa). */
> > coeff = (100 << 24) / c->ic_freq;
> >
> > This smells like the iwn(4) RXON SYSASSERT issue fixed this morning.
> > Given that the driver starts out using a channel that hasn't been
> > initialized, I don't understand how this code ever worked.
> >
> > Does this diff help?
> > If it does, we should check all other drivers for the same problem.
> >
> > Index: athn.c
> > ===================================================================
> > RCS file: /cvs/src/sys/dev/ic/athn.c,v
> > retrieving revision 1.92
> > diff -u -p -r1.92 athn.c
> > --- athn.c 5 Jan 2016 18:41:15 -0000 1.92
> > +++ athn.c 13 Jan 2016 12:07:32 -0000
> > @@ -333,8 +333,8 @@ athn_attach(struct athn_softc *sc)
> > /* Get the list of authorized/supported channels. */
> > athn_get_chanlist(sc);
> >
> > - /* IBSS channel undefined for now. */
> > - ic->ic_ibss_chan = &ic->ic_channels[0];
> > + /* Use channel 1 as initial channel. */
> > + ic->ic_ibss_chan = &ic->ic_channels[1];
> >
> > ifp->if_softc = sc;
> > ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
>
> Built an new MP kernel with above patch and it boots now. Thanks for quick
> fix.
>
> Timo
This is the real fix, for all drivers.
I messed this up while removing turbo mode.
Index: ieee80211.c
===================================================================
RCS file: /cvs/src/sys/net80211/ieee80211.c,v
retrieving revision 1.57
diff -u -p -r1.57 ieee80211.c
--- ieee80211.c 12 Jan 2016 09:28:09 -0000 1.57
+++ ieee80211.c 13 Jan 2016 14:19:26 -0000
@@ -749,8 +749,10 @@ ieee80211_setmode(struct ieee80211com *i
modeflags = chanflags[mode];
for (i = 0; i <= IEEE80211_CHAN_MAX; i++) {
c = &ic->ic_channels[i];
- if (mode == IEEE80211_MODE_AUTO ||
- (c->ic_flags & modeflags) == modeflags)
+ if (mode == IEEE80211_MODE_AUTO) {
+ if (c->ic_flags != 0)
+ break;
+ } else if ((c->ic_flags & modeflags) == modeflags)
break;
}
if (i > IEEE80211_CHAN_MAX) {
@@ -764,8 +766,10 @@ ieee80211_setmode(struct ieee80211com *i
memset(ic->ic_chan_active, 0, sizeof(ic->ic_chan_active));
for (i = 0; i <= IEEE80211_CHAN_MAX; i++) {
c = &ic->ic_channels[i];
- if (mode == IEEE80211_MODE_AUTO ||
- (c->ic_flags & modeflags) == modeflags)
+ if (mode == IEEE80211_MODE_AUTO) {
+ if (c->ic_flags != 0)
+ setbit(ic->ic_chan_active, i);
+ } else if ((c->ic_flags & modeflags) == modeflags)
setbit(ic->ic_chan_active, i);
}
/*
