I gratefully appreciate your help and sincere trust. I'll try the approach you suggested. After I downloaded the network SCTP test suite from git, I found it's difficult for me to handle. Because I am not understand perl script language even I have no idea of running it. However, it has really good design and is gracefully implemented. It must be nutritious to me, and I would learn plentiful knowledge from it, I think. Thanks a lot! Richard -----Original Message----- From: Marcelo Ricardo Leitner [mailto:marcelo.leitner@xxxxxxxxx] Sent: 2015年11月10日 1:07 To: YUAN Jia; linux-sctp@xxxxxxxxxxxxxxx Subject: Re: About SCTP restart? Em 09-11-2015 00:17, YUAN Jia escreveu: > Indeed, RFC4960 has been already supported. I was misled. Okay. Please don't top post. It makes reading the emails harder. > Another question: > I am making a suite of SCTP testing tools. The paragraph 5.2.4.1 of RFC 4960 describes an example of an association restart. But I don't know how to trig or simulate a procedure of endpoints' restart. I find there are only functions to handle restart initiated by peer, but I've not found functions to initiate restart from local. Does lksctp lib have APIs or functions to initiate a restart? Or some of code fractions in kernel can do so? > > Handling restart initiated by peer: > sctp_do_sm > => sctp_sf_do_5_2_4_dupcook > => sctp_sf_do_dupcook_a // Section 5.2.4, case 'A' This is used to handle an error situation on which the peer was restarted, and not as a normal procedure. Please don't confuse it with stream reset. That said, I'm not aware of any mechanism in lksctp to trigger this willingly, but you may do it in your application. You can do it by: 1.bind your client socket to a known local port before connecting to the server 2.connect to remote server 3.add a firewall rule on client blocking outputs to that server (port) 4.kill your client application (4.b.ensure association is really down) 5.remove firewall rule 6.restart your application (which should bind to the same local port) The firewall rule will block the shutdown/abort from client to server when you close/kill your client app, thus simulating a crash from the server point of view. Btw, you may want to take a look on TAHI then. It already supports quite some SCTP tests. http://networktest.sourceforge.net/ maybe you can help us improve it? :-) Regards, Marcelo ?韬{.n?????%??檩??w?{.n???{炳??Ф?塄}?财??j:+v?? ? ???2??璀??摺?囤??z夸z罐?+?????w棹f
