lkml.org 
[lkml]   [1998]   [Jan]   [22]   [last100]   RSS Feed
Views: [more markup]  [less markup]  [headers]  [forward] 
 
Messages in this thread
/
FromNigel Metheringham <>
Subject2.0.x bug/patch - NFS - RPC: rpc_send sending evil packet:
DateThu, 22 Jan 1998 11:34:56 +0000
[Could the LMP people look at taking this up for 2.0.34]

We have seen a number of NFS related crashes in 2.0.x (seen in 2.0.30 RH 
patches, rather worse for some reason in 2.0.33).  In each case a number 
of these messages are seen:-

Jan 12 19:28:21 svr-a-02 kernel: RPC: rpc_doio sending evil packet:
Jan 12 19:28:21 svr-a-02 kernel:      19cca34a 00000001 00000000 00000000 
				      00000000 00000000 00000000 00200002
Jan 12 19:28:21 svr-a-02 kernel: RPC: rpc_send sending evil packet:
Jan 12 19:28:21 svr-a-02 kernel:      19cca34a 00000001 00000000 00000000 
				      00000000 00000000 00000000 00200002
Then more "kernel: NFS server server_mail not responding", followed by 
lots of fork failed and the system becoming pretty solidly locked up.

We finally found the following message on linux-kernel from last July 
which appears to be the only mention of this problem, from Olaf Kirch:-
[ see http://linuxwww.db.erau.edu/mail_archives/linux-kernel/Jul_97/0470.ht
ml ]

Olaf Kirch <okir@monad.swb.de> said:
} This is definitely a bug in the NFS client. Let me explain what's
} happening and why the packet is evil:
} The client sees a timeout (or some other problem, e.g. garbage reply).
} While it correctly figures that it should retransmit the packet, it
} does not rebuild the packet but resends the current network buffer
} which, for some reason, already contains an RPC *reply* (you can see
} that from the second long word - a call has a 0 there, while replies
} have a 1). The first long word is the transmission id (XID).
} It appears there's a race condition somewhere in the RPC socket
} handling where the client doesn't notice that some packet has already
} been received when it starts waiting for the reply.
} Can you please apply the enclosed patch and check what happens?
} May I also request that people put `NFS' somewhere in the subject when
} they report a problem to linux-kernel that they suspect is
} nfs-related? Quite frequently, I scan the lists only by subject, so
} threads like this one usually escape me.
} Cheers Olaf
We used the patch given in that message (attached below) and it appears to 
have fixed the problem for us.

Could this patch be evaluated for taking forward to the next 2.0.x 
release...

	Nigel.

[unhandled content-type:application/x-patch ][ Nigel.Metheringham@theplanet.net   -  Systems Software Engineer ]
[ Tel : +44 113 251 6012                   Fax : +44 113 234 6065 ]
[      Real life is but a pale imitation of a Dilbert strip       ]
\
 
 \ /
  Last update: 2005-03-22 12:41    [from the cache]
©2003-2008