On Tue, 2009-01-27 at 11:41 +0200, Tomas Winkler wrote: > On Mon, Jan 26, 2009 at 8:47 PM, Johannes Berg > <johannes@xxxxxxxxxxxxxxxx> wrote: > > As far as I can tell, there are possible lockups because both the RX > > session_timer and TX addba_resp_timer are del_timer_sync'ed under > > the sta spinlock which both timer functions take. Additionally, the > > TX agg code seems to leak memory when TX aggregation is not disabled > > before the sta_info is freed. > > Funny, it was your deed to replaced the two locks with one. Woudn;t > be better just put it back? Not sure. The problem here seems to be that ever since the agg structures are dynamically allocated (my pushing, I know) we need to protect sta->ampdu_mlme.tid_tx[i]/sta->ampdu_mlme.tid_rx[i] with the corresponding spinlock. Hence, if you look at my earlier patch closely, it didn't actually introduce this problem, it just shifted it from one lock to another. johannes
Attachment:
signature.asc
Description: This is a digitally signed message part
