On Wed, Jun 21, 2000 at 01:08:31PM +0900, Stephen J. Turnbull wrote:
> I haven't checked this carefully to see exactly what's going on, but
> with venus/vice 5.3.7 the sequence
> 
> clog
> ifconfig eth0 down
> (assorted manipulations of cfs)
> 
> leaves me unable to write any files or even list acls (I had just
> started venus and had not done any writes), while continuing
> 
> ifconfig eth0 up
> cfs la .
> cfs disconnect
> ifconfig eth0 down
> 
> works fine with full access (per ACLs of course) to writing files.

Venus not only caches files, but also caches `trust', clog discards all
cached access permissions for the user because it has to re-authenticate
with the servers using the new token.

If you do a 'hoard walk' between the clog and the ifconfig eth0 down, it
should work as intended. The cfs disconnect actually doesn't do anything
inside of venus, but just blocks the incoming/outgoing rpc2 packets.

> This doesn't seem very good to me, nor do the docs mention it (in
> particular, the Coda-HOWTO seems to indicate that disconnecting the
> network is a substitute for cfs disconnect).

Yes, cfs la . forced a reconnection and cache revalidation which gave
you the access rights on the objects back.

Jan