Stefan Sperling <[email protected]> writes:
> On Tue, Dec 12, 2017 at 02:45:51PM +0100, Stefan Sperling wrote:
>> On Tue, Dec 12, 2017 at 02:38:14PM +0100, Stefan Sperling wrote:
>> > On Tue, Dec 12, 2017 at 02:17:03PM +0100, Grégoire Jadi wrote:
>> > > I tried a simple reboot, and I also repeated the upgrade but it didn't
>> > > crash.
>> > >
>> > > Is there anything else I can try?
>> >
>> > Nothing so far, apart from trying over and over again.
>> >
>> > This looks like a race where the device receives a frame before things
>> > are fully initialized. Still trying to put my finger on it though, it's
>> > still unclear how this can happen exactly.
>> >
>>
>> Grégoire, what does your hostname.iwn0 file look like (without any wifi
>> passwords of course)? I'd like to know if there's something about it that
>> would cause the kernel to put the driver up/down multiple times, which
>> could increase the likelyhood of such races.
>>
>
> Nevermind. I have figured it out.
I've applied/compiled & rebooted 10 times without crashes (during the
boot).
Thank you!
> The stupid nasty 'any channel' token bites again:
>
> #define IEEE80211_CHAN_MAX 255
> #define IEEE80211_CHAN_ANY 0xffff /* token for ``any
> channel'' */
> #define IEEE80211_CHAN_ANYC \
> ((struct ieee80211_channel *) IEEE80211_CHAN_ANY)
>
> Whoever thought that was a good idea already owes me some of their lifetime...
>
> Index: ieee80211_node.c
> ===================================================================
> RCS file: /cvs/src/sys/net80211/ieee80211_node.c,v
> retrieving revision 1.123
> diff -u -p -r1.123 ieee80211_node.c
> --- ieee80211_node.c 12 Dec 2017 00:24:21 -0000 1.123
> +++ ieee80211_node.c 12 Dec 2017 13:51:19 -0000
> @@ -994,6 +994,9 @@ ieee80211_node_checkrssi(struct ieee8021
> {
> uint8_t thres;
>
> + if (ni->ni_chan == IEEE80211_CHAN_ANYC)
> + return 0;
> +
> if (ic->ic_max_rssi) {
> thres = (IEEE80211_IS_CHAN_2GHZ(ni->ni_chan)) ?
> IEEE80211_RSSI_THRES_RATIO_2GHZ :
