The patch works OK, I actually tried something like that already. It solves the loop-problem, I can remove the 1 microsecond sleep and exit gracefully with ctrl-c. But it has not solved the problem with BU's. The MN is still not sending any BU's at all. I tried with MIPL-2.0-rc3+the UMIP patch and this works well running as MN and with MIPL-2.0.1 running as HA. But it would be nice if I could use the same version on all systems :)
Leif > -----Original Message----- > From: Ville Nuorvala [mailto:[EMAIL PROTECTED] > Sent: Wednesday, 19 April, 2006 12:35 > To: Buch-Pedersen, Leif > Cc: [email protected] > Subject: Re: [mipl] MIPL 2.0.1 problems with MN > > Buch-Pedersen, Leif wrote: > > Hello, > > Hello Leif! > > > I am returning to the problem I experienced one month ago (and still > > have) with MIPL 2.0.1 running as MN, where BU's is never > sent, mip6d > > appears to be hanging and can not be stopped by ctrl-c. > > > > I think I have found the place where the problem occurs: > > > > The routine rtnl_ext_listen in rtnl.c makes a call to recvmsg. This > > call returns a status of -1 and an error EINVAL that indicates a > > parameter problem. This does not cause rtnl_ext_listen to exit. It > > continues and enters the loop starting with "for (h =" > where it goes > > into an endless loop, that is the reason why ctrl-c does > not stop the > > program. By inserting in the loop a 1 microsecond sleep it > gets slowed > > down so it can be aborted and ctrl-c works again. This > obviously does > > not solve the problem of BU's not being sent and I am assuming the > > parameter problem is the reason. Any suggestions on solving the > > problem would be highly appreciated. > > Thanks for tracking this down. AFAICS there is really no reason for > rtnl_ext_listen() *not* to exit if recvmsg() fails. Would you > try the attached patch to see if it works? > > Regards, > Ville > > _______________________________________________ mipl mailing list [email protected] http://www.mobile-ipv6.org/cgi-bin/mailman/listinfo/mipl
