Date: Sat, 12 Sep 2009 09:41:13 -0500 From: "Doug Poland" <doug@polands.org> To: "Rick Macklem" <rmacklem@uoguelph.ca> Cc: "Li, Qing" <qing.li@bluecoat.com>, freebsd-stable@freebsd.org, FreeBSD Current <current@freebsd.org>, Andrzej Tobola <ato@iem.pw.edu.pl> Subject: RE: NFS issues on 8.0-BETA4 Message-ID: <19c32ef4263ed3c2ab1624a43ae86c78.squirrel@email.polands.org> In-Reply-To: <Pine.GSO.4.63.0909111621420.6191@muncher.cs.uoguelph.ca> References: <9ef3bf09fa0e081eca3965e3f0e84f82.squirrel@email.polands.org> <B583FBF374231F4A89607B4D08578A4305454467@bcs-mail03.internal.cacheflow.com> <Pine.GSO.4.63.0909111621420.6191@muncher.cs.uoguelph.ca>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, September 11, 2009 15:37, Rick Macklem wrote: > >>> >>> I cannot sucessfully mount exports from the NFSv3 server on the >>> 8.0-BETA4 client. All works well with 7.2 clients. >>> >>> The strange thing is, the directory in which I mount the nfs >>> filesystem disappears, and I get an error when I attempt to >>> access the directory. >>> > > I went and looked at the message over in stable@ (I guess I have to > join that mailing list, too:-). > > I think that the second mount attempt, which failed with errno==64 > EHOSTDOWN probably gives you a better indication of what is broken > and hints at a networking problem, which hopefully others can help > with? > > mount_nfs currently has the quirk that, if you don't specify either > "udp" or "tcp", it will use UDP for the mount protocol and then > switch to TCP for NFS. (I didn't make it that way, but noticed it > when I was adding changes for the experimental client.) So, you might > want to try adding "udp" or "tcp" to the mount options for your > "simplified mount" and see what happens? (I suggest this, since it > seems to have been able to talk to mountd on the server, but not NFS > on the server and the only explanation I can think of is the switch > to TCP for that.) > > I think the first case failed after the retrycnt, due to the "soft" > option, but hadn't succeeded in doing any NFS Getattr, so the mount > point's type wasn't VDIR--> returning errno==1 EPERM for the "ls -l". > > > If the mount attempts with "udp" or "tcp" specified give you the same > behaviour, you could try using wireshark to capture the traffic. > (Wireshark is pretty good at decoding NFS traffic.) If you don't have > wireshark, you can use "tcpdump -w <file> -s 0 <host>" to capture > the traffic in a file that can be fed to wireshark later. (I'd be > happy to look at the traffic, if you were to email me the above > <file>.) > > Good luck with it, rick ps: I'll assume that the client can talk to > the server for other stuff ok. pss: A big change between 7 and 8 is > use of a new kernel RPC layer that handles the RPC transport (again, > I wasn't the author, but am somewhat familiar with it). > > Simply adding -o udp to my mount command made the NFS mount work correctly. Qing, would it be beneficial to attempt the patch in light of these findings? Thanks all for the help, sorry for the noise. -- Regards, Doug
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19c32ef4263ed3c2ab1624a43ae86c78.squirrel>