On Wed, 2010-06-09 at 10:54 +0200, ext Johannes Berg wrote: > On Wed, 2010-06-09 at 11:50 +0300, Juuso Oikarinen wrote: > > On Wed, 2010-06-09 at 09:32 +0200, ext Johannes Berg wrote: > > > On Wed, 2010-06-09 at 09:50 +0300, Juuso Oikarinen wrote: > > > > > > > v3: - queue reconfig filter work to kernel default wq instead of mac80211 wq > > > > > > I think it would make more sense to use a different lock instead, like > > > the iflist mutex? > > > > Yeah had that originally, but apparently the iflist mutex is acquired > > *after* the mgd mutex in some places of the code, especially all around > > mlme.c, and here I would have to acquire them in reverse order. So I > > changed to the rtnl lock. > > Ah, right. > > However I'm a bit worried that making this workqueue change will allow > the filter reconfigure to cross with suspend, which might mean it could > execute after the device is suspended? Ah yes. That's an obvious side-effect. I wonder if it would be solved by flushing this particular work along with the mac80211 wq in the suspend call, or would the implications then just move further? Apart from the above, I'm out of ideas. If this is getting too complex, I'll need to revert back to the patch that does not take into account the promisc mode. -Juuso -- 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
