OpLock+flat DB corruption (Was: How Samba let us down)

Chris de Vidal cdevidal at yahoo.com
Mon Oct 28 21:55:37 GMT 2002


--- jra at dp.samba.org wrote:
> On Thu, Oct 24, 2002 at 04:43:53AM -0700, Chris de
> Vidal wrote:
> 
> > OpLocks were indeed causing corruption; we only
> turned
> > them off, made no other changes, and have no more
> > corruption, as I reported yesterday.  Wouldn't
> that be
> > a priority 1, "drop everything" bug?  Other
> experience
> > was confirmed by doing a Google, by 2 Samba
> authors,
> > and by the results of our one simple change.
> > 
> > If you'd like, I can submit an official bug
> report.
> 
> Is it completely reproducible ? Problems cuased by
> clients
> not responding to oplock breaks are notoriously
> dependent
> on network hardware and client issues (network
> drivers etc).

Sorry for the late reply; I was out all weekend.

We can't reproduce the problem because we don't have
another Oce' and Opus setup ($$) to test it on, but in
essence:
No corruption on Netware, no oplocks.
Corruption on Samba with oplocks.
No corruption on Samba, no oplocks.

Yes, turning off all oplocks was the only change made.

2 Samba authors and a Google search confirmed this
kind of corruption.

> We drive the client differently than a Windows TCP
> stack,
> and remember Microsoft don't test with anything than
> their own stack.
> 
> Problems like this come under the "oplock break"
> problems, not generic corruption.

Still, it's corruption, and the user ought to be made
aware.  I would like (at least) to submit a
documentation patch.  It might have saved us grief.

/dev/idal

__________________________________________________
Do you Yahoo!?
Y! Web Hosting - Let the expert host your web site
http://webhosting.yahoo.com/



More information about the samba-technical mailing list