mac80211: fail back to use associate from reassociate

mac80211: fail back to use associate from reassociate failure

Some APs have strict checking between associate and reassociate. In a case when an AP is restarted during a connection, it denies the mac80211 reassoc request since this is a new association for the AP. To fix this problem, we need to check the status code against WLAN_STATUS_REASSOC_NO_ASSOC and clear ifsta->prev_bssid_set in handling the association failure response.

index b003912..dacd2d3 100644
@@ -1182,6 +1182,8 @@ static void ieee80211_rx_mgmt_assoc_resp(struct net_device *dev,
 	if (status_code != WLAN_STATUS_SUCCESS) {
 		printk(KERN_DEBUG "%s: AP denied association (code=%d)\n",
 		       dev->name, status_code);
+		if (status_code == WLAN_STATUS_REASSOC_NO_ASSOC)
+			ifsta->prev_bssid_set = 0;
 		return;
 	}
 

Can you please identify such an AP? This sounds like a workaround that makes the non-AP STA (client) behave in a way that does not match with IEEE 802.11 standard and in general, I'm against this change without a very good reason for it. This type of change would, e.g., break IEEE 802.11r fast transition which can only be done using reassociation.

Jouni Malinen                                            PGP id EFC895FA

I take it an error handling code. When an AP denies a STA reassociation request with the reason "there is no previous association, why do you send reassociation request to me?", this is an indication for the STA to switch to association request and retry. Currently, mac80211 just try reassociation over and over. The result is it can never associate with the AP.

I don't see how it breaks the 802.11 standard. If you do, please point it out and how are you going to fix this problem?

In .11r, can AP reject reassociation request if there is no previous association from the same STA?

Thanks,
-yi

I would like to offer my 2 cents here.

While the proposed fix for this problem seems reasonable, I think the AP behavior is definitely suspect.

It is certainly legal for a STA to send a reassoc request to an AP with which it has no existing association as long as the STA has an existing association with some other AP within the same ESS.

Therefore, the AP should not return the said error code (WLAN_STATUS_REASSOC_NO_ASSOC) unless it has checked (using some inter AP protocol) that no other AP in the ESS has an existing association with the requesting STA. If the AP is returning the error without doing this check, I would think that it is an AP bug.

It would be interesting to know which AP behaves in this way and whether is really does the check correctly or not. Note that the AP itself having rebooted shouldn't cause this error code to be returned, which is what I found strange.

Regards,
Sandesh

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Zhu Yi
Sent: Thursday, May 10, 2007 8:51 AM
To: Jouni Malinen
Cc: [email protected]; John W. Linville
Subject: Re: [PATCH] mac80211: fail back to use associate from reassociate

I take it an error handling code. When an AP denies a STA reassociation request with the reason "there is no previous association, why do you send reassociation request to me?", this is an indication for the STA to switch to association request and retry. Currently, mac80211 just try reassociation over and over. The result is it can never associate with the AP.

I don't see how it breaks the 802.11 standard. If you do, please point it out and how are you going to fix this problem?

In .11r, can AP reject reassociation request if there is no previous association from the same STA?

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

The AP was the only one in the ESS, so it should be OK.

Do you mean the AP should remember all its associated STAs before it rebooted?

Thanks,
-yi

-----Original Message-----
From: Zhu Yi [mailto:[email protected]]
Sent: Tuesday, May 15, 2007 12:23 PM
To: Sandesh Goel
Cc: Jouni Malinen; [email protected]; John W. Linville
Subject: RE: [PATCH] mac80211: fail back to use associate from reassociate

The AP was the only one in the ESS, so it should be OK. [Sandesh]
[Sandesh] How does the AP know that it is the only one in the ESS? It can't surely assume that. Is it configured with this information? Did it try to find out if there are other APs in the ESS? E.g. would it behave differently if you setup another AP in your ESS?

Do you mean the AP should remember all its associated STAs before it rebooted?
[Sandesh]
[Sandesh] I didn't say that AP needs to remember that. However, it should honor the reassoc request if it is not sure about whether the requesting STA was associated in the ESS or not.

Thanks,
-yi

OK. This sounds reasonable. I went through the IEEE 802.11 standard and it does not seem to describe this part of re-association process in any level of detail apart from allocation a status code for reporting failures to validate association status.

.11r does not use this status code. I was concerned about this in general, i.e., anything that converts re-association to association sounds somewhat undesired, but it looks like this particular status code could be the only exception for this. I still don't see how the AP in this case would have known about the association status for the full ESS (since I don't see how it would know that it is the only AP in the ESS). Anyway, non-AP STA does not have enough context to determine what exactly happen, so it cannot really do much more than either stop trying to re-associate or try to associate what is what your change is proposing ot do.

Jouni Malinen                                            PGP id EFC895FA