Skip to contentSkip to conversation

mac80211: handle failed scan requests in STA mode

PATCH In review Helmut Schaa 2 participants 5 messages Mar 9, 2009 mac80211

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mac80211: handle failed scan requests in STA mode

If cfg80211 requests a scan it awaits either a return code != 0 from the scan function or the cfg80211_scan_done to be called. In case of a STA mac80211's scan function ever returns 0 and queues the scan request. If ieee80211_sta_work is executed and ieee80211_start_scan fails for some reason cfg80211_scan_done will never be called but cfg80211 still thinks the scan was triggered successfully and will refuse any future scan requests due to drv->scan_req not being cleaned up.

If a scan is triggered from within the MLME a similar problem appears. If ieee80211_start_scan returns an error, local->scan_req will not be reset and mac80211 will refuse any future scan requests.

Hence, in both cases call ieee80211_scan_failed (which notifies cfg80211 and resets local->scan_req) if ieee80211_start_scan returns an error.

---

It just happend to me that after rf-killing iwlwifi wpa_supplicant triggered a scan (which was accepted by mac80211 as it is not rfkill aware yet) which failed in iwlwifi and left cfg80211 in a state where it won't accept a new scan request even after unkilling the radio again.

Johannes, do you think it is worth to add something like a scan watchdog that triggers after 30 seconds to cfg80211 and prints "scan request failed", cancels the scan request and notifies user space?

net/mac80211/ieee80211_i.h +1 -0
index ecbc8e0..fbb91f1 100644
@@ -972,6 +972,7 @@ int ieee80211_sta_set_extra_ie(struct ieee80211_sub_if_data *sdata,
 			       char *ie, size_t len);
 
 void ieee80211_mlme_notify_scan_completed(struct ieee80211_local *local);
+void ieee80211_scan_failed(struct ieee80211_local *local);
 int ieee80211_start_scan(struct ieee80211_sub_if_data *scan_sdata,
 			 struct cfg80211_scan_request *req);
 struct ieee80211_bss *
net/mac80211/mlme.c +12 -2
index eeb6da8..6e92674 100644
@@ -1720,7 +1720,10 @@ static int ieee80211_sta_config_auth(struct ieee80211_sub_if_data *sdata)
 				local->int_scan_req.ssids[0].ssid_len = 0;
 			else
 				local->int_scan_req.ssids[0].ssid_len = ifmgd->ssid_len;
-			ieee80211_start_scan(sdata, &local->int_scan_req);
+
+			if (ieee80211_start_scan(sdata, &local->int_scan_req))
+				ieee80211_scan_failed(local);
+
 			ifmgd->state = IEEE80211_STA_MLME_AUTHENTICATE;
 			set_bit(IEEE80211_STA_REQ_AUTH, &ifmgd->request);
 		} else {
@@ -1757,7 +1760,14 @@ static void ieee80211_sta_work(struct work_struct *work)
 	    ifmgd->state != IEEE80211_STA_MLME_AUTHENTICATE &&
 	    ifmgd->state != IEEE80211_STA_MLME_ASSOCIATE &&
 	    test_and_clear_bit(IEEE80211_STA_REQ_SCAN, &ifmgd->request)) {
-		ieee80211_start_scan(sdata, local->scan_req);
+		/*
+		 * The call to ieee80211_start_scan can fail but ieee80211_request_scan 
+		 * (which queued ieee80211_sta_work) did not return an error. Thus, call
+		 * ieee80211_scan_failed here if ieee80211_start_scan fails in order to
+		 * notify the scan requester.
+		 */
+		if (ieee80211_start_scan(sdata, local->scan_req))
+			ieee80211_scan_failed(local);
 		return;
 	}
 
net/mac80211/scan.c +12 -0
index 0e81e16..5030a3c 100644
@@ -202,6 +202,18 @@ ieee80211_scan_rx(struct ieee80211_sub_if_data *sdata, struct sk_buff *skb,
 	return RX_QUEUED;
 }
 
+void ieee80211_scan_failed(struct ieee80211_local *local)
+{
+	if (WARN_ON(!local->scan_req))
+		return;
+
+	/* notify cfg80211 about the failed scan */
+	if (local->scan_req != &local->int_scan_req)
+		cfg80211_scan_done(local->scan_req, true);
+
+	local->scan_req = NULL;
+}
+
 void ieee80211_scan_completed(struct ieee80211_hw *hw, bool aborted)
 {
 	struct ieee80211_local *local = hw_to_local(hw);
maintainer
#1 lore raw
Quoted 8 lines · Helmut Schaa

If cfg80211 requests a scan it awaits either a return code != 0 from the scan function or the cfg80211_scan_done to be called. In case of a STA mac80211's scan function ever returns 0 and queues the scan request. If ieee80211_sta_work is executed and ieee80211_start_scan fails for some reason cfg80211_scan_done will never be called but cfg80211 still thinks the scan was triggered successfully and will refuse any future scan requests due to drv->scan_req not being cleaned up.

Quoted 6 lines

If a scan is triggered from within the MLME a similar problem appears. If ieee80211_start_scan returns an error, local->scan_req will not be reset and mac80211 will refuse any future scan requests.

Hence, in both cases call ieee80211_scan_failed (which notifies cfg80211 and resets local->scan_req) if ieee80211_start_scan returns an error.

This patch looks good.

Quoted 11 lines

---

It just happend to me that after rf-killing iwlwifi wpa_supplicant triggered a scan (which was accepted by mac80211 as it is not rfkill aware yet) which failed in iwlwifi and left cfg80211 in a state where it won't accept a new scan request even after unkilling the radio again.

Johannes, do you think it is worth to add something like a scan watchdog that triggers after 30 seconds to cfg80211 and prints "scan request failed", cancels the scan request and notifies user space?

But I don't really think it's worth doing that.

johannes

#2 lore raw

Am Montag, 9. M=E4rz 2009 schrieb Johannes Berg:

Quoted 2 lines · Helmut Schaa
Quoted 1 line

Johannes, do you think it is worth to add something like a scan wat=

chdog that

Quoted 1 line
Quoted 1 line

triggers after 30 seconds to cfg80211 and prints "scan request fail=

ed", cancels

Quoted 3 lines
Quoted 1 line

the scan request and notifies user space?

=20
But I don't really think it's worth doing that.

=46ine with me. This was just a quick idea on how we could easily spot = such=20
bugs in the future (for example when fullmac drivers start using cfg802= 11).

Helmut
--
To unsubscribe from this list: send the line "unsubscribe linux-wireles= s" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

maintainer
#3 lore raw
Quoted 10 lines · Helmut Schaa

Am Montag, 9. März 2009 schrieb Johannes Berg:

Quoted 6 lines · Helmut Schaa
Quoted 3 lines

Johannes, do you think it is worth to add something like a scan watchdog that triggers after 30 seconds to cfg80211 and prints "scan request failed", cancels the scan request and notifies user space?

But I don't really think it's worth doing that.

Fine with me. This was just a quick idea on how we could easily spot such bugs in the future (for example when fullmac drivers start using cfg80211).

Yeah, might be worth it later, but I suspect that other drivers will have state machines that are easier to verify.

johannes

#4 lore raw
In reply to Helmut Schaa

Am Montag, 9. M=E4rz 2009 schrieb Helmut Schaa:

Quoted 1 line

If cfg80211 requests a scan it awaits either a return code !=3D 0 fro=

m

Quoted 2 lines

the scan function or the cfg80211_scan_done to be called. In case of a STA mac80211's scan function ever returns 0 and queues the scan req=

uest.

Quoted 2 lines

If ieee80211_sta_work is executed and ieee80211_start_scan fails for some reason cfg80211_scan_done will never be called but cfg80211 stil=

l

Quoted 4 lines

thinks the scan was triggered successfully and will refuse any future scan requests due to drv->scan_req not being cleaned up. =20
If a scan is triggered from within the MLME a similar problem appears=

=2E If

Quoted 1 line

ieee80211_start_scan returns an error, local->scan_req will not be re=

set

Quoted 3 lines

and mac80211 will refuse any future scan requests.
=20
Hence, in both cases call ieee80211_scan_failed (which notifies cfg80=

211

Quoted 3 lines

and resets local->scan_req) if ieee80211_start_scan returns an error. =20

Btw, John, I guess this should go into linux-next as the cfg80211-scan = code
is there already and otherwise cfg80211 and mac80211 can get stuck (ref= uses
any scan requests) when the iwlwifi killswitch is used (only tested wit= h
NetworkManager).

I'm not sure if anything apart from the killswitch also triggers this i= ssue.

Thanks,
Helmut
--
To unsubscribe from this list: send the line "unsubscribe linux-wireles= s" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html