Hi, >>> So I think we're probably better off accepting the set_key but not >>> actually using it, and instead disconnecting... even if that's awkward >>> and should come with a big comment :-) >> >> Instead of returning an error I'll change the code to accept the rekey >> but do nothing with it. (Basically delete the new key and keep the old >> active). >> And of course calling ieee80211_set_disassoc() prior to return "success". > > Right. Did you handle/consider modes other than BSS/P2P client btw? I've tested that on a client only and it's already not working. Problem is, that with ieee80211_set_disassoc() we just dump the association in the kernel and are not informing the userspace, so the state machine is stuck in "connected" but the STA is unable to communicate. I also tried ieee80211_mgd_deauth(): While better this is basically the same behaviour as returning an error on key replace. By setting the error code to inactivity at least wpa_supplicant was actually starting to reconnect, unfortunatelly set_key then failes since we purged the assosication in the kernel. (Or that's my best interpretation from the logs). Networkmanager then again prompted for the password... I started experimenting with just signals to the userspace, but then paused... Trying to control the state machine with spoofed errors trying to trigger an "desireable" action can't be the way forward, can it? Even if we get that working changes or different implementations in userspace may well break it. I now think we only have two reasonable ways forward: 1) The user friendly one: If the userspace does not know about the new API just print a error message and do the insecure key replace. With potential clear text packet leaks and connection freezes. 2) The secure way: Report an error on key install and accept that users will encounter new problems when they are using rekeys with the old API with "old" userspace software. Since we have this issue in the kernel at least as long as we have mac80211 I would vote for 1) here. Fix mac80211 and add a new way for the userspace to to securely replace the keys and detect when this is not possible. Then get the userspace software updated to act accordingly, finally preventing the clear text packet leaks. After some time we can then decide to reject insecure rekeys, burning only those who use kernels much newer than the userspace. Does this sound like an reasonable approch or should I go back figuring out how to trick the userspace to reconnect without user notification/intervention? Alexander
Attachment:
pEpkey.asc
Description: application/pgp-keys
