On 21/10/2025 17:10, Adrian Chadd wrote:
hi,
Ok, let's file a PR with what you have and then dig into it a bit more!
Okay, I can / should do that.
But just now I've got a little bit lucky and got some trace.
It contains only state transitions and only for wlan0, so a lot of the picture
is missing, but maybe there is something in it.
wlan0: start running, 0 vaps running
wlan0: ieee80211_start_locked: up parent ath0
wlan0: start running, 1 vaps running
^^^^^ these look like competing calls to ieee80211_start_locked
wlan0: ieee80211_new_state_locked:2751: starting state update INIT -> INIT
(SCAN)
wlan0: ieee80211_new_state_locked: INIT -> SCAN (arg 0) (nrunning 0 nscanning 0)
wlan0: ieee80211_newstate_cb:2567: running state update INIT -> SCAN (1)
wlan0: ieee80211_newstate_cb: INIT -> INIT arg 0 -> SCAN arg 0
wlan0: hostap_newstate: INIT -> INIT (0)
wlan0: ieee80211_swscan_cancel_scan: called; F_SCAN=0, vap=nomatch, signal=0
wlan0: ieee80211_new_state_locked:2751: starting state update INIT -> INIT
(SCAN)
wlan0: ieee80211_new_state_locked: INIT -> SCAN (arg 0) (nrunning 0 nscanning 0)
wlan0: ieee80211_newstate_cb:2567: running state update INIT -> SCAN (1)
wlan0: ieee80211_newstate_cb: INIT -> SCAN arg 0
wlan0: hostap_newstate: INIT -> SCAN (0)
wlan0: ieee80211_swscan_cancel_scan: called; F_SCAN=0, vap=nomatch, signal=0
wlan0: ieee80211_create_ibss: creating HOSTAP on channel 11n flags 0x00010480
wlan0: ieee80211_new_state_locked:2751: starting state update SCAN -> SCAN (RUN)
wlan0: ieee80211_new_state_locked: SCAN -> RUN (arg -1) (nrunning 0 nscanning 0)
wlan0: ieee80211_init
wlan0: start running, 1 vaps running
^^^^^ ieee80211_start_locked is called yet again
wlan0: ieee80211_new_state_locked:2751: starting state update SCAN -> SCAN
(SCAN)
wlan0: ieee80211_new_state_locked: RUN -> SCAN (arg 0) (nrunning 0 nscanning 0)
wlan0: stop running, 1 vaps running
^^^^ ieee80211_stop_locked is called, it would clear IFF_DRV_RUNNING
wlan0: ieee80211_new_state_locked:2751: starting state update SCAN -> SCAN
(INIT)
wlan0: ieee80211_new_state_locked: SCAN -> INIT (arg -1) (nrunning 0 nscanning
0)
wlan0: ieee80211_swscan_cancel_scan: called; F_SCAN=0, vap=nomatch, signal=0
wlan0: down parent ath0
^^^^ there was only one running VAP so the driver is stopped
vvvv state transitions continue, though
wlan0: ieee80211_newstate_cb:2567: running state update SCAN -> RUN (1)
wlan0: ieee80211_newstate_cb: SCAN -> INIT arg 0 -> RUN arg -1
wlan0: hostap_newstate: SCAN -> INIT (0)
wlan0: ieee80211_swscan_cancel_scan: called; F_SCAN=0, vap=nomatch, signal=0
wlan0: ieee80211_new_state_locked:2751: starting state update INIT -> INIT (RUN)
wlan0: ieee80211_new_state_locked: INIT -> RUN (arg -1) (nrunning 0 nscanning 0)
wlan0: ieee80211_newstate_cb:2567: running state update INIT -> SCAN (1)
wlan0: ieee80211_newstate_cb: INIT -> SCAN arg 0
wlan0: hostap_newstate: INIT -> SCAN (0)
wlan0: ieee80211_swscan_cancel_scan: called; F_SCAN=0, vap=nomatch, signal=0
wlan0: ieee80211_create_ibss: creating HOSTAP on channel 11n flags 0x00010480
wlan0: ieee80211_new_state_locked:2751: starting state update SCAN -> SCAN (RUN)
wlan0: ieee80211_new_state_locked: RUN -> RUN (arg -1) (nrunning 0 nscanning 0)
wlan0: ieee80211_newstate_cb:2567: running state update SCAN -> INIT (1)
wlan0: ieee80211_newstate_cb: SCAN -> INIT arg -1
wlan0: hostap_newstate: SCAN -> INIT (-1)
wlan0: ieee80211_swscan_cancel_scan: called; F_SCAN=0, vap=nomatch, signal=0
wlan0: ieee80211_newstate_cb:2567: running state update INIT -> RUN (1)
wlan0: ieee80211_newstate_cb: INIT -> RUN arg -1
wlan0: hostap_newstate: INIT -> RUN (-1)
wlan0: ieee80211_swscan_cancel_scan: called; F_SCAN=0, vap=nomatch, signal=0
wlan0: ieee80211_create_ibss: creating HOSTAP on channel 11n flags 0x00010480
wlan0: ieee80211_new_state_locked:2751: starting state update RUN -> RUN (RUN)
wlan0: ieee80211_new_state_locked: RUN -> RUN (arg -1) (nrunning 0 nscanning 0)
wlan0: ieee80211_newstate_cb:2567: running state update RUN -> RUN (1)
wlan0: ieee80211_newstate_cb: RUN -> RUN arg -1
wlan0: hostap_newstate: RUN -> RUN (-1)
wlan0: ieee80211_swscan_cancel_scan: called; F_SCAN=0, vap=nomatch, signal=0
wlan0: ieee80211_newstate_cb:2567: running state update RUN -> RUN (1)
wlan0: ieee80211_newstate_cb: RUN -> RUN arg -1
wlan0: hostap_newstate: RUN -> RUN (-1)
wlan0: ieee80211_swscan_cancel_scan: called; F_SCAN=0, vap=nomatch, signal=0
The VAP ends up in RUN state, but there hasn't been another
ieee80211_start_locked call since the ieee80211_stop_locked call, so the
interface is not RUNNING.
Now, the only thing is to figure out what hostapd and network.subr were doing
with wlan0... :-)
On Tue, 21 Oct 2025 at 06:30, Andriy Gapon <[email protected]
<mailto:[email protected]>> wrote:
On 21/10/2025 16:05, Andriy Gapon wrote:
> It seems that the problem happens because the VAP is already in
IEEE80211_S_RUN
> state somehow and ieee80211_start_locked is not called, obviously, in
> SIOCSIFFLAGS handling code.
I have this DTrace output to confirm that:
ieee80211_ioctl:entry wlan0 cmd 0x80206910 flags 0x8902 drv flags 0 state 5
ieee80211_ioctl:entry wlan0 cmd 0x80206910 flags 0x8903 drv flags 0 state 5
So, the VAP state is indeed IEEE80211_S_RUN.
But IFF_RUNNING / IFF_DRV_RUNNING is not set in either if_flags or
if_drv_flags.
Looks like something set the VAP state bypassing the interface state
machinery.
I tried to catch the state transitions with wlandebug_wlan0 in rc.conf but
apparently that changes timing enough to mask the issue.
--
Andriy Gapon
--
Andriy Gapon
