[OpenAFS-devel] openafs hangs on shutdown with selinux (caused by callback expiration via umount?)

[Christopher Allen Wing]

Jeff:

Thanks for writing back.


On Wed, 2 Jan 2008, Jeffrey Altman wrote:

> If you could capture the network traffic you could determine exactly
> which RPC is being sent.

The problem is that I only see this occasionally (out of the 1000+ 
machines running RHEL5 here), and only during a reboot.  I wasn't able to 
trigger the condition on demand, so I thought I'd be lazy and email 
openafs-devel in case someone could think of what happens in the openafs 
module that:

 	(a) tries to send packets when /afs is unmounted, presumably
 	    via afs_shutdown().

and

 	(b) does not happen every time /afs is unmounted, but only
 	    sometimes.


I was thinking it might be callback de-registration.  I tried pulling the 
network cable with open files in /afs and live callbacks (according to 
cmdebug), then closing the files and unmounting /afs but this didn't do it 
either - the system rebooted fine.  Maybe there is some other procedure 
I'm not aware of which would cause AFS to attempt some RPC at unmount 
time?  The only thing I'm sure of is that there aren't any open files in 
/afs when this happens; else the umount wouldn't get that far at all.


I think I fixed the problem here by hacking around it in any case, but I'd 
like to know how to reproduce this on demand, and ideally I'd come up with 
a solution for the main openafs distribution.

If I have a moment I'll try rebooting a machine here with a packet capture 
running, in case I can get the problem to recur.



Thanks,

Chris
wingc@umich.edu