[OpenAFS] Misconfigured Windows client hangs Linux server

Rainer Toebbicke rtb@pclella.cern.ch
Thu, 23 Jun 2005 11:58:39 +0200


Jeffrey Altman wrote:

>>Having a fileserver dump with a similar case on my desk I can confirm
>>that a windows client can indeed "hang" an AFS server (although I don't
>>know if that had anything to do with digital signatures).
>>
>>The scenario: all threads blocked in h_Lock_r on the same host,
>>triggered by a SRXAFS_GetTime call. Except the one holding the lock who
>>is waiting for RXAFSCB_WhoAreYou (to that host). Sure, it will
>>eventually time out if there is no response but if the GetTime RPCs are
>>sufficiently frequent this will do the job.
> 
> 
> What version of the Windows client?
> 

rxdebug: AFS version: OpenAFS1.3.7300

cmdebug: error checking locks: RPC interface mismatch (-455)
cmdebug: failed to get cache entry 0 (RPC interface mismatch (-455))


> Is the client reachable from the server?
> 
> WhoAreYou is not implemented in clients earlier than 1.3.8000.
> If the client is 1.3.8000 or higher, what do rxdebug and cmdebug
> report for the client?
> 

If it's not implemented the server should get RXGEN_OPCODE and go do 
something else, which, as far as I can judge from the dump, is not what 
happened.

The point about hardening the file server is that the overall quality of 
the AFS service must not rely on every [Windows or not] user upgrading 
whenever there is a new version.


-- 
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Rainer Toebbicke
European Laboratory for Particle Physics(CERN) - Geneva, Switzerland
Phone: +41 22 767 8985       Fax: +41 22 767 7155