Bob Copeland wrote:
> On Sun, Mar 15, 2009 at 2:02 PM, Johannes Berg
> <johannes@xxxxxxxxxxxxxxxx> wrote:
> > On Sun, 2009-03-15 at 12:13 +0530, Sujith wrote:
> >> When the driver has been notified with a STA_REMOVE, it tears down
> >> the internal ADDBA state. On resume, trying to initiate aggregation would
> >> fail because mac80211 has not cleared the operational state for that <TID,STA>.
> >> This can be fixed by tearing down the existing sessions on a suspend.
> >
> > Agreed, but Bob said there were some deadlocks with this?
>
> I think the deadlock went away, not sure what fixed it but I guess
> the HT rework. I tested it a couple of weeks ago on ath5k and Luis
> did on ath9k so I say let's go ahead and apply the patch... in the
> meantime I'll try again on ath5k just to be sure.
There is a window for a race. Something like this:
>From mac80211:
__ieee80211_suspend()
tear_down_BA_sessions(TX, RX)
ampdu_action(STOP)
remove_vifs()
At this point, the driver executes its remove_interface routine.
While we are doing this, a TX completion interrupt could be raised,
(HW hasn't been stopped yet) and nothing stops the driver from calling
ieee80211_start_tx_ba_session().
So the question is: should mac80211 deny ADDBA requests in this case ?
Sujith
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
