From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from Chamillionaire.breakpoint.cc (Chamillionaire.breakpoint.cc [193.142.43.52]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id DD94F3FC1 for ; Mon, 6 Sep 2021 11:35:50 +0000 (UTC) Received: from fw by Chamillionaire.breakpoint.cc with local (Exim 4.92) (envelope-from ) id 1mNCuh-0001L4-Ix; Mon, 06 Sep 2021 13:35:47 +0200 Date: Mon, 6 Sep 2021 13:35:47 +0200 From: Florian Westphal To: Paolo Abeni Cc: Florian Westphal , mptcp@lists.linux.dev, Matthieu Baerts Subject: Re: [mptcp-next 2/2] mptcp: re-set push-pending bit on retransmit failure Message-ID: <20210906113547.GE23554@breakpoint.cc> References: <20210906060614.25217-1-fw@strlen.de> <20210906060614.25217-3-fw@strlen.de> <20210906075531.GD23554@breakpoint.cc> Precedence: bulk X-Mailing-List: mptcp@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.10.1 (2018-07-13) Paolo Abeni wrote: > On Mon, 2021-09-06 at 09:55 +0200, Florian Westphal wrote: > > Because I don't think it makes sense to re-try ASAP if we could > > not find a ssk, and instead defer until next firing of rtx timer. > > Ok > > > > > Do we need something similar for __mptcp_subflow_push_pending(), too? > > > > I don't think so, its thats not invoked from the mptcp-release cb; > > yup, right. To be clear, with "something similar" I also referred to > ensuring the rtx timer is armed. __mptcp_subflow_push_pending() is only called when we already have pending xmit, i.e. the timer is already running and with the change in this patch the timer is guaranteed to re-arm itself unless there is nothing to transmit anymore. __mptcp_push_pending is different because that is called when new data is passed in via sendmsg, i.e. timer might not be running (yet). Makes sense to you? > > > I'm wondering if this will cause a sort of 'spinning'/busy loop on > > > above condition inside mptcp_release_cb(). > > > > Hmm, right, that might indeed happen. > > It should be sae to remove the above clause and unconditioally wait > > for the reworked rtx timer to assert pending bit again. > > +1 :) (assuming s/sae/safe/ :) Yep, safe. Test has been running for a few hours now with no failures, will send v2 shortly.