Hello, </snip> > > > > would it make sense to have let's say > > > > mq_move2mlist(struct mbuf_queue *, struct mbuf_list *) > > This already exists, it's called mq_delist() >
thanks, I've missed that when skimming through sources. > > > > This would allow as to move whole globally visible la->la_mq into > > into mbuf list, which will be a local variable. This way we won't > > need to jump on la->la_mq's mutex with every loop iteration. > > > > I it makes sense, we can do it as a follow up change. > > We'd need some other way to do the 'mbuf is back in queue' detection, > but I agree this seems like a sensible thing to do. > I agree, I think this should be discussed in new email thread, because I don't want to hijack bluhm's diff for that. as I've said earlier I'm OK if bluhm's will commit his diff. thanks and regards sashan