Re: nfs + Reiser4

[Chuck Lever <chuck.lever@oracle.com>]

On 04/07/2010 03:20 PM, J. Bruce Fields wrote:
> On Wed, Apr 07, 2010 at 02:59:36PM -0400, Chuck Lever wrote:
>> On 04/07/2010 02:51 PM, J. Bruce Fields wrote:
>>> On Wed, Apr 07, 2010 at 02:44:01PM -0400, Chuck Lever wrote:
>>>> On 04/07/2010 01:34 PM, J. Bruce Fields wrote:
>>>>> On Tue, Apr 06, 2010 at 07:52:21PM +0200, gg-B3jsHfKwJfLR7s880joybQ@public.gmane.org wrote:
>>>>>> I am having serious headaches using nfs between a reiser4 server and arm
>>>>>> client.
>>>>>> Both on 2.6.29 vintage kernels.
>>>>>>
>>>>>> Files are constantly getting out of sync.
>>>>>>
>>>>>> Example :
>>>>>>
>>>>>> boot ARM via nfs
>>>>>> edit lighttpd.conf on ARM
>>>>>> check edit is visible on server. OK
>>>>>>
>>>>>> reboot ARM
>>>>>> check file : reverted to an earlier state.
>>>>>> check server: edited version still showing.
>>>>>
>>>>> So, on a freshly booted NFS client, you're opening and reading a file
>>>>> and seeing file data that isn't even on the NFS server any more?
>>>>>
>>>>> That's beyond bizarre.  Do you have a reliable way to reproduce the
>>>>> problem?
>>>>
>>>> Could be XID replay.
>>>
>>> I'm not following you.  You're thinking of a read request after the
>>> reboot that unluckily reuses an old XID and gets stale data from the
>>> servers reply cache?  Or something else?
>>
>> Nothing unlucky about it.  Just after a boot, if the client
>> implementation isn't careful about choosing an initial XID, (eg it
>> always starts with a psuedorandom number but uses the same seed every
>> time), it will hit the server's replay cache.
>
> Hm, OK.
>
>> This can be quite reproducible for NFSROOT and a quiescent server.
>
> The Linux server doesn't cache READ results as far as I can tell.

OK.  Some servers do.

Sounds like a server side network trace is in order.

-- 
chuck[dot]lever[at]oracle[dot]com

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs
_______________________________________________
Please note that nfs@lists.sourceforge.net is being discontinued.
Please subscribe to linux-nfs@vger.kernel.org instead.
    http://vger.kernel.org/vger-lists.html#linux-nfs