On Tue, Apr 14, 2009 at 01:48:55PM -0700, Ben Greear wrote: > ======================================================= > [ INFO: possible circular locking dependency detected ] > 2.6.29.1c3 #20 > ------------------------------------------------------- > ip/3220 is trying to acquire lock: > (&ifsta->work){--..}, at: [<c0135c9f>] __cancel_work_timer+0x3f/0x190 > > but task is already holding lock: > (rtnl_mutex){--..}, at: [<c032d71f>] rtnl_lock+0xf/0x20
Lock order between cfg80211 mutex and rtnl it seems... Hmm, was that before or after the change to cfg80211_get_dev_from_ifindex()? > ath5k phy0: noise floor calibration failed (2462MHz) > BUG: soft lockup - CPU#0 stuck for 61s! [iwconfig:24158] > Modules linked in: michael_mic ath5k mac80211 cfg80211 nfs lockd auth_rpcgss > 8021q garp stp llc nf_d] I guess this is the noise floor calibration loop called from the reset workqueue? I had a patch a while ago to spinlock reset calls but I think the lock was too big of a hammer. Having a spinlock around ath5k_hw_* may be a good idea though, especially with lots of vifs. >>> modprobe ath5k with nohwaccel=1 (I think that's the name...test system is >> >> nohwcrypt=1... but it would be nice if we swapped in/out keys into the >> hw crypto key cache based on vif. > > Seems the HW hashes on the AP's MAC, and so if you have multiple VIFS > associated with the same AP, it breaks. This is from memory, > so I could be wrong here. Hmm, good point. Just curious, what's the use case for multiple virtual stations associated with the same AP? > Yes, I put it all in one patch since we don't have things sorted out > quite yet. Okay, cool. Looks promising though. -- Bob Copeland %% www.bobcopeland.com _______________________________________________ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel