NFS 4 client callback port

[Guillaume Rousse]

Talpey, Thomas a écrit :
> At 09:14 AM 6/12/2008, Trond Myklebust wrote:
>> On Thu, 2008-06-12 at 08:52 -0400, Talpey, Thomas wrote:
>>> At 08:47 AM 6/12/2008, Guillaume Rousse wrote:
>>>> I got only one, that's correct:
>>>> [root at chatelet ~]# ps ax | grep nfsv4-svc
>>>>  7146 ?        S      0:00 [nfsv4-svc]
>>> I think this is sufficient evidence that the ports you are seeing are not
>>> NFSv4 callback ports. Do you see the port count rise as you do nfs4
>>> mounts manually?
>> The following patch was merged into 2.6.25 to fix a callback server
>> leak:
>>
>>  
>> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git&a=com
>> mitdiff&h=8e60029f403781b8a63b7ffdb7dc1faff6ca651e
>>
>> If Guillaume is running a 2.6.24 based kernel, then this might well be
>> the source of his extra sockets.
> 
> Aha - shame on me for not checking the .25 logs. Yes, this would
> exactly explain the single thread, but multiple sockets.
> 
> It's just a resource issue though, shouldn't affect operation. But,
> if the tcp_callback_tcpport is set, it might become an issue because
> the client will attempt to re-bind the port, and fail. So Guillaume
> should definitely merge to this patch in order to control the port.
OK, yet another patch I need to convince mandriva kernel maintainers to 
merge if don't want to maintain my own kernels indefinitely...

I'm trying this immediately.
-- 
Guillaume Rousse
Moyens Informatiques - INRIA Futurs
Tel: 01 69 35 69 62