From mboxrd@z Thu Jan 1 00:00:00 1970 From: Wei Yongjun Date: Thu, 03 Mar 2011 09:27:12 +0000 Subject: Re: Gap is only retransmitted once Message-Id: <4D6F5EF0.50301@cn.fujitsu.com> List-Id: References: In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: linux-sctp@vger.kernel.org > More read about the ethreal captures. > > The peer shutdown association 6 seconds after it missed the > retransmittion of sctp.data_tsn = 447375618. During the 6 seconds, > the peer sent a lot of SACKs but Linux did not retransmit the chunk. > > Could someone look at it? I will post a patch about this, can you try it and tell us whether the patch can fix your problem? > Great thanks! > > 2011/3/3 Mingyuan Zhu : >> I attached more capture packets. >> From the ethreal file, I found some chunks are retransmitted the >> second time after 3 seconds. The peer does not shutdown the >> association. But sometimes the peer shutdown the association. >> For example: >> sctp.data_tsn = 447375491 or sctp.data_tsn = 447375492 are >> retransmitted 3 seconds after the first retransmittion. >> >> And sctp.data_tsn = 447375618, the peer shutdown association before >> the 2nd retransmittion. >> >> So comes the question, why doesn't sctp retransmit immediately when it >> receives SACK, instead it retransmit after 3 seconds? >> >> 2011/3/1 Mingyuan Zhu : >>> My test is working on kernel 2.6.34-4. >>> >>> 2011/3/1 Wei Yongjun : >>>>> I searched the src code. We did not use sinfo_timetolive. >>>> I guess there were some retransmit issues in 2.6.31 and >>>> latest kernel hit the rwnd limit, I will try to reproduce it >>>> and fix it. wait some time since I have other work to do now. >>>> >>>>> 2011/3/1 Wei Yongjun : >>>>>> Did you using sinfo_timetolive when you send sctp message? >>>>>> There is a bug when we handle sinfo_timetolive, I will fix this later. >>>>>> >>>>>>> Mingyuan Zhu wrote, at 02/22/2011 09:14 PM: >>>>>>>> What kind of test tool you used for the testing? >>>>>>> No test tool can directly test your case. >>>>>>> But SCTP Conformance Test Suite Project can help you. >>>>>>> See: http://networktest.sourceforge.net/ >>>>>>> >>>>>>> Your case is not in the tool, so you need to write your case by yourself. >>>>>>> data#22 has similar test procedure, see >>>>>>> http://networktest.sourceforge.net/log/2.6.29-rc4/IPv4/data/22.html >>>>>>> >>>>> -- >>>>> To unsubscribe from this list: send the line "unsubscribe linux-sctp" in >>>>> the body of a message to majordomo@vger.kernel.org >>>>> More majordomo info at http://vger.kernel.org/majordomo-info.html >>>>>