Hello Peter,
On 18/01/15(Sun) 13:45, [email protected] wrote:
> >Synopsis: Crash triggered by ifconfig up on urtwn interface with post
> >usb3 snapshots
> >Category: amd64 kernel
> >Environment:
> System : OpenBSD 5.7
> Details : OpenBSD 5.7-beta (GENERIC.MP) #788: Sat Jan 17 07:35:25
> MST 2015
>
> [email protected]:/usr/src/sys/arch/amd64/compile/GENERIC.MP
>
> Architecture: OpenBSD.amd64
> Machine : amd64
> >Description:
> After the recent USB changes, my urtwn would lose link significantly
> more often than previous. At most times
> netstart urtwn0 would produce desired effect, however every 5 or 6
> times the machine would crash.
> Photos of the desired post crash output are at
> http://www.bsdly.net/~peter/20150118_crash/
> >How-To-Repeat:
> amd64 machine with usb3 only ports, no bios option to downshift to
> older mode. use urtwn wifi dongle,
> when link drops, use /etc/netstart to re-establish link. every 5 to 6
> times, the system crashes.
Thanks for the bug report. Due to the existing problem with xhci(4) on
Intel hardware and the fact that urtwn(4) does not bring itself down
when the interface times out, you're exercising some unusual code path
in the network stack.
And you found a regression the hard way. Sorry for that, diff below
should fix it. Well it won't make your urtwn(4) work better, but at
least it should not panic(9) your kernel.
Index: route.c
===================================================================
RCS file: /cvs/src/sys/net/route.c,v
retrieving revision 1.199
diff -u -p -r1.199 route.c
--- route.c 13 Jan 2015 12:14:00 -0000 1.199
+++ route.c 18 Jan 2015 13:02:00 -0000
@@ -1045,8 +1045,10 @@ rt_checkgate(struct ifnet *ifp, struct r
if ((rt->rt_flags & RTF_UP) == 0) {
rt = rtalloc(dst, RT_REPORT|RT_RESOLVE, rtableid);
+ if (rt == NULL)
+ return (EHOSTUNREACH);
rt->rt_refcnt--;
- if (rt == NULL || rt->rt_ifp != ifp)
+ if (rt->rt_ifp != ifp)
return (EHOSTUNREACH);
}
