Skip to contentSkip to conversation

wifi: mac80211: clear beaconing state in ieee80211_do_stop()

PATCH In review Hojun Choi 2 participants 3 messages Jun 28 – Jul 6 mac80211

lore
Share & tools
lore

Hojun Choi

author
#0 lore raw

PATCH 1/1 wifi: mac80211: clear beaconing state in ieee80211_do_stop()

ieee80211_do_stop() zeroes vif.bss_conf.beacon_int but, unlike ieee80211_stop_ap(), ieee80211_ibss_disconnect() and ieee80211_stop_mesh(), leaves enable_beacon set and the SDATA_STATE_OFFCHANNEL_BEACON_STOPPED bit pending. cfg80211 rejects a beacon interval below 10, so a zero interval comes from this teardown, not from userspace.

do_stop() clears SDATA_STATE_RUNNING first, so nothing acts on the stale state while the interface is down. But do_open() re-sets RUNNING without restoring beacon_int or clearing enable_beacon and the bit, so the state survives a down/up cycle. A later ieee80211_offchannel_return() on scan completion then consumes the stale bit and re-enables beaconing while beacon_int is still 0, and drivers arming a beacon timer from the interval divide by zero. syzbot reported this as a divide error in mac80211_hwsim_link_info_changed(); the exact sequence that leaves do_stop() with beaconing still enabled was not reproduced.

Clear enable_beacon and the bit in ieee80211_do_stop() as well, mirroring the other teardown paths, so the state cannot outlive the interface. do_stop() warns on a vif that still has valid links, so MLD links are already torn down here and only the deflink beacon_int is zeroed; clearing the deflink enable_beacon is sufficient.

---
 net/mac80211/iface.c | 4 ++++
 1 file changed, 4 insertions(+)
net/mac80211/iface.c +4 -0
index 086272c3ec08..c070d47cb631 100644
@@ -719,6 +719,10 @@ static void ieee80211_do_stop(struct ieee80211_sub_if_data *sdata, bool going_do
 
 	sdata->vif.bss_conf.beacon_int = 0;
 
+	/* mirror stop_ap()/IBSS/mesh: also clear the beaconing state */
+	sdata->vif.bss_conf.enable_beacon = false;
+	clear_bit(SDATA_STATE_OFFCHANNEL_BEACON_STOPPED, &sdata->state);
+
 	/*
 	 * If the interface goes down while suspended, presumably because
 	 * the device was unplugged and that happens before our resume,
Signature · 1 line
2.54.0
author
#1 lore raw

Following up with the concrete path, since the question is whether cfg80211_leave() already covers this. It does not for AP start failure:
ieee80211_start_ap() sets enable_beacon before its fail-able steps, but its error path (cfg.c:1930) only releases the channel, never clearing it (unlike the IBSS/mesh paths), while cfg80211 leaves ap.beacon_interval at
0. A later interface-down then makes ___cfg80211_stop_ap() short-circuit (-ENOENT, net/wireless/ap.c:30) and skip the stop_ap clear, so ieee80211_do_stop() runs with enable_beacon still set. I could not get a driver to fail start_ap() there, so this is by inspection, not a reproducer.

I can fold this into the commit message in a v2, or instead fix the producer by clearing enable_beacon on the start_ap() error path (mirroring IBSS/mesh) - whichever you prefer. I used do_stop() since you suggested it might need to clear that bit.

Thanks,
Hojun

maintainer
#2 lore raw
Quoted 10 lines · Hojun Choi

Following up with the concrete path, since the question is whether cfg80211_leave() already covers this. It does not for AP start failure:
ieee80211_start_ap() sets enable_beacon before its fail-able steps, but its error path (cfg.c:1930) only releases the channel, never clearing it (unlike the IBSS/mesh paths), while cfg80211 leaves ap.beacon_interval at
0. A later interface-down then makes ___cfg80211_stop_ap() short-circuit (-ENOENT, net/wireless/ap.c:30) and skip the stop_ap clear, so ieee80211_do_stop() runs with enable_beacon still set. I could not get a driver to fail start_ap() there, so this is by inspection, not a reproducer.

How did syzbot hit it then?

Quoted 4 lines

I can fold this into the commit message in a v2, or instead fix the producer by clearing enable_beacon on the start_ap() error path (mirroring IBSS/mesh) - whichever you prefer. I used do_stop() since you suggested it might need to clear that bit.

Who is "you" here? I guess not me?

It seems to me the cleanup paths make more sense though? This looks really quite out-of-place in do_stop() I think, especially the offchannel bit? How would that even happen?

johannes