Skip to contentSkip to conversation

d80211: fix wpa_supplicant reassoc problem

PATCH Applied Hong Liu 2 participants 4 messages Aug 25, 2006 – Sep 21, 2006

Subsystems (2)
Networking Network drivers
lore
Share & tools
lore

Hong Liu

author
#0 lore raw

PATCH 1/2 d80211: fix wpa_supplicant reassoc problem

After key negotiation completed using wpa_supplicant, wpa_supplicant can't reassoc with the AP if we reboot the AP. It always fails at the 4-way handshake.
The problem is the key info is not cleared correctly. Thus when wpa_supplicant send the EAPOL-KEY packet, the d80211 stack finds the old key and uses it to encrypt the packet.

The patch removes the sta_info when we disassociate with AP.

Thanks,
Hong

net/d80211/ieee80211_sta.c +8 -0
index 8caf352..2144b34 100644
@@ -739,6 +739,14 @@ static void ieee80211_associated(struct 
 		wireless_send_event(dev, SIOCGIWAP, &wrqu, NULL);
 		mod_timer(&ifsta->timer,
 			  jiffies + IEEE80211_MONITORING_INTERVAL + 30 * HZ);
+
+		sta = sta_info_get(local, ifsta->bssid);
+		if (sta) {
+			sta_info_free(sta, 0);
+			sta_info_put(sta);
+		}
+
+		ifsta->probereq_poll = 0;
 	} else {
 		mod_timer(&ifsta->timer,
 			  jiffies + IEEE80211_MONITORING_INTERVAL);
net/d80211/sta_info.c +6 -2
index 7f5febe..8902816 100644
@@ -197,6 +197,12 @@ #ifdef CONFIG_D80211_VERBOSE_DEBUG
 	       local->mdev->name, MAC_ARG(sta->addr));
 #endif /* CONFIG_D80211_VERBOSE_DEBUG */
 
+	if (sta->key) {
+		ieee80211_key_sysfs_remove(sta->key);
+		ieee80211_key_free(sta->key);
+		sta->key = NULL;
+	}
+
 	rate_control_remove_sta_attrs(local, sta->rate_ctrl_priv, &sta->kobj);
 	ieee80211_sta_sysfs_remove(sta);
 

@@ -244,8 +250,6 @@ void sta_info_free(struct sta_info *sta,
 				kfree(key);
 			}
 		}
-		ieee80211_key_free(sta->key);
-		sta->key = NULL;
 	} else if (sta->key_idx_compression != HW_KEY_IDX_INVALID) {
 		struct ieee80211_key_conf conf;
 		memset(&conf, 0, sizeof(conf));
#1 lore raw
Quoted 8 lines · Hong Liu

After key negotiation completed using wpa_supplicant, wpa_supplicant can't reassoc with the AP if we reboot the AP. It always fails at the 4-way handshake.
The problem is the key info is not cleared correctly. Thus when wpa_supplicant send the EAPOL-KEY packet, the d80211 stack finds the old key and uses it to encrypt the packet.

The patch removes the sta_info when we disassociate with AP.

Please:
- add proper Signed-off-by line
- do not send patches as an attachment
- Cc: John Linville <[email protected]> as well as me on d80211 patches

Quoted 6 lines
net/d80211/ieee80211_sta.c +0 -0
index 8caf352..2144b34 100644
@@ -739,6 +739,14 @@ static void ieee80211_associated(struct 
 		wireless_send_event(dev, SIOCGIWAP, &wrqu, NULL);

net/d80211/ieee80211_sta.c static void ieee80211_associated(struct · line 749

Quoted 10 lines
net/d80211/ieee80211_sta.c +8 -0
 		mod_timer(&ifsta->timer,
 			  jiffies + IEEE80211_MONITORING_INTERVAL + 30 * HZ);
+
+		sta = sta_info_get(local, ifsta->bssid);
+		if (sta) {
+			sta_info_free(sta, 0);
+			sta_info_put(sta);
+		}
+
+		ifsta->probereq_poll = 0;

Wouldn't it be better to move this into the 'else' branch of previous 'if' statement (i.e. just before sta_info_put call in the ieee80211_associated function)?

Thanks,

Jiri

Signature · 2 lines
Jiri Benc
SUSE Labs
author
#2 lore raw
Quoted 28 lines · Jiri Benc
Quoted 8 lines · Hong Liu

After key negotiation completed using wpa_supplicant, wpa_supplicant can't reassoc with the AP if we reboot the AP. It always fails at the 4-way handshake.
The problem is the key info is not cleared correctly. Thus when wpa_supplicant send the EAPOL-KEY packet, the d80211 stack finds the old key and uses it to encrypt the packet.

The patch removes the sta_info when we disassociate with AP.

Please:
- add proper Signed-off-by line
- do not send patches as an attachment
- Cc: John Linville <[email protected]> as well as me on d80211 patches

Quoted 12 lines
net/d80211/ieee80211_sta.c +4 -0
index 8caf352..2144b34 100644
@@ -739,6 +739,14 @@ static void ieee80211_associated(struct
              wireless_send_event(dev, SIOCGIWAP, &wrqu, NULL);
              mod_timer(&ifsta->timer,
                        jiffies + IEEE80211_MONITORING_INTERVAL + 30 * HZ);
+
+             sta = sta_info_get(local, ifsta->bssid);
+             if (sta) {
+                     sta_info_free(sta, 0);
Quoted 9 lines

net/d80211/ieee80211_sta.c static void ieee80211_associated(struct · line 749

Quoted 4 lines
net/d80211/ieee80211_sta.c +4 -0
+                     sta_info_put(sta);
+             }
+
+             ifsta->probereq_poll = 0;

Wouldn't it be better to move this into the 'else' branch of previous 'if' statement (i.e. just before sta_info_put call in the ieee80211_associated function)?

Rework the patch according to Jiri's comments.

Signed-off-by Hong Liu <[email protected]>
net/d80211/ieee80211_sta.c +2 -0
index a933d92..557ba17 100644
@@ -710,6 +710,8 @@ static void ieee80211_associated(struct 
 				       "range\n",
 				       dev->name, MAC_ARG(ifsta->bssid));
 				disassoc = 1;
+				sta_info_free(sta, 0);
+				ifsta->probereq_poll = 0;
 			} else {
 				ieee80211_send_probe_req(dev, ifsta->bssid,
 							 local->scan_ssid,
net/d80211/sta_info.c +6 -2
index 622b234..5edbd0c 100644
@@ -197,6 +197,12 @@ #ifdef CONFIG_D80211_VERBOSE_DEBUG
 	       local->mdev->name, MAC_ARG(sta->addr));
 #endif /* CONFIG_D80211_VERBOSE_DEBUG */
 
+	if (sta->key) {
+		ieee80211_key_sysfs_remove(sta->key);
+		ieee80211_key_free(sta->key);
+		sta->key = NULL;
+	}
+
 	rate_control_remove_sta_attrs(local, sta->rate_ctrl_priv, &sta->kobj);
 	ieee80211_sta_sysfs_remove(sta);
 

@@ -244,8 +250,6 @@ void sta_info_free(struct sta_info *sta,
 				kfree(key);
 			}
 		}
-		ieee80211_key_free(sta->key);
-		sta->key = NULL;
 	} else if (sta->key_idx_compression != HW_KEY_IDX_INVALID) {
 		struct ieee80211_key_conf conf;
 		memset(&conf, 0, sizeof(conf));
Applied
#3 lore raw
Quoted 1 line · Hong Liu

Rework the patch according to Jiri's comments.

Applied to my tree. Thanks for the patch!

Jiri

Signature · 2 lines
Jiri Benc
SUSE Labs