On Sun, Dec 25, 2016 at 2:31 AM, xxhdx1985126 <xxhdx1985126@xxxxxxx> wrote: > Hi, everyone. > > I'm reading the source code of Pipe in ceph, and I'm confused about the code for connection race processing. Suppose that the policy the two Pipe involved is lossless_peer, one of them wins and send the other pipe a CEPH_MSGR_TAG_WAIT message which makes the other Pipe's writer to cond.wait(), and then the winning pipe goes back to waiting for a ceph_msg_connect message to come. But, under this circumstance, the writer of the other pipe is waiting, and as far as I know, unless something happens it will never send ceph_msg_connect messages, so they will wait for each other for ever........ There must be some misunderstanding here, but I don't know where, please help me. Thank you:-) The point is that we have racing connect attempts here — the "winner" has already sent out a connection attempt which the loser will respond to. So neither of them needs to do anything; when the loser processes the winner's connection attempt, it will shut down its losing Pipe, start up a Pipe with the incoming attempt, and send back an acknowledgement to the winner. -Greg -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
