Re: problems with nfsd (due to RPCSEC_GSS changes?)

On 12 Nov 2008, at 19:36, Navdeep Parhar wrote:

> On Wed, Nov 12, 2008 at 1:20 AM, Doug Rabson <dfr_at_rabson.org> wrote:
>>
>> On 12 Nov 2008, at 01:30, Navdeep Parhar wrote:
>>
>>> I had a FreeBSD NFS server running a month+ old current (from Oct  
>>> 2 or
>>> so).  I upgraded to a current current (Nov 11) and nfsd stopped  
>>> working.
>>> I was able to mount the exported filesystem but anything else would
>>> yield an "Input/output error." nfsstat -s showed "Server Ret-Failed"
>>> going up everytime I tried a 'cd', 'ls', etc. from the client.  (I  
>>> tried
>>> both FreeBSD and Solaris clients).
>>>
>>> Ultimately, I had to add NFS_LEGACYRPC in order to get a working  
>>> nfsd.
>>> Looks like there may be a problem with the new code that was added  
>>> as
>>> part of RPCSEC_GSS support.  Note that I did not enable KGSSAPI in  
>>> my
>>> kernel as I have no need for it.
>>>
>>> Are there any knows issues with the new code?  Feel free to ask if  
>>> you
>>> need any more information about my setup.
>>
>> I don't know of anything specific. If I could see a packet trace  
>> including
>> both the mount request and at least one failed access attempt, it  
>> would help
>> to understand what is happening here.
>>
>
> I saw a handful of commits from you last night so I updated +  
> rebuilt the
> server's kernel to include them.  These traces are with today's code  
> (Nov12
> 11AM Pacific) on the server and yesterday's code on the client.
>
> The server is .2 and the client is .1, the trace is using tcpdump -s
> 256 -vvn on the server.
>
> # mount /usr/obj  (and then wait a couple of seconds.  The mount  
> succeeds)

Could you try this again with 'tcpdump -w foo.pcap ...' and then send  
me the resulting file - its very hard to see what is really happening  
from a simple tcpdump output.