A Gentoo user at http://bugs.gentoo.org/124884 reports that the following
message appears in the logs over 650 times every day:
svc: unknown version (0)
The system is a NFS server with many active clients.
Given that this #define only controls printk output, does it make sense to
disable it by default?
Signed-off-by: Daniel Drake <[email protected]>
--- linux/net/sunrpc/svc.c.orig 2006-03-05 00:51:10.000000000 +0000
+++ linux/net/sunrpc/svc.c 2006-03-05 00:52:40.000000000 +0000
@@ -20,7 +20,7 @@
#include <linux/sunrpc/clnt.h>
#define RPCDBG_FACILITY RPCDBG_SVCDSP
-#define RPC_PARANOIA 1
+#undef RPC_PARANOIA
/*
* Create an RPC service
Trond,
Trond Myklebust wrote:
> They are probably trying to ping the server with a NULL procedure call
> to test for service availability. We should allow that particular
> usage...
Thanks, that sounds likely. Can you give some hints as to how a NULL
procedure call might appear? Would testing for prog==0 and/or proc==0 be
appropriate?
Daniel
On Tue, 2006-03-14 at 21:21 +0000, Daniel Drake wrote:
> Trond,
>
> Trond Myklebust wrote:
> > They are probably trying to ping the server with a NULL procedure call
> > to test for service availability. We should allow that particular
> > usage...
>
> Thanks, that sounds likely. Can you give some hints as to how a NULL
> procedure call might appear? Would testing for prog==0 and/or proc==0 be
> appropriate?
I can't see that authorising calls to prog==0 could ever be useful (what
would that mean?), but proc==0 is another matter: that is precisely the
NULL procedure call that I mentioned above.
Cheers,
Trond
>
>A Gentoo user at http://bugs.gentoo.org/124884 reports that the following
>message appears in the logs over 650 times every day:
>
> svc: unknown version (0)
>
Should not the clients be fixed?
Jan Engelhardt
--
On Sun, 2006-03-05 at 14:52 +0100, Jan Engelhardt wrote:
> >
> >A Gentoo user at http://bugs.gentoo.org/124884 reports that the following
> >message appears in the logs over 650 times every day:
> >
> > svc: unknown version (0)
> >
> Should not the clients be fixed?
>
They are probably trying to ping the server with a NULL procedure call
to test for service availability. We should allow that particular
usage...
Cheers,
Trond
Trond Myklebust wrote:
> I can't see that authorising calls to prog==0 could ever be useful (what
> would that mean?), but proc==0 is another matter: that is precisely the
> NULL procedure call that I mentioned above.
Just to let you know, I did cook up a patch but the user has gone
unresponsive (it's an important production server so finding downtime is
hard).
Here is the patch if you are interested:
http://bugs.gentoo.org/attachment.cgi?id=82195&action=view
If/when he responds, I'll remove the 2nd hunk and submit the patch properly.
Thanks,
Daniel