Re: File Read Returns Non-existent Null Bytes

[Simo Sorce <simo@redhat.com>]

On Thu, 2015-02-26 at 10:34 -0500, Trond Myklebust wrote:
> On Thu, 2015-02-26 at 10:22 -0500, Simo Sorce wrote:
> > On Thu, 2015-02-26 at 09:10 -0500, Chris Perl wrote:
> > > > However if you are asking us for an extensive list of "this is what I
> > > > can expect if I ignore these rules", then I don't think you will find
> > > > much traction. Such a list would be committing us to defining a model
> > > > for "non-close-to-open" semantics, which isn't of interest to me at
> > > > least, and I doubt anyone else is interested in committing to
> > > > maintaining that.
> > > 
> > > One more point on this.  I wasn't really asking for a list of what I
> > > can expect if I ignore the rules (although I think pointing out that
> > > reading corrupt data from the cache is worth mentioning), I was asking
> > > what the rules for close-to-open consistency were so I can follow
> > > them.  I now know one of them is that if a file is open for writing on
> > > one client then it can't be read on another.  Are there others?
> > 
> > Is this a rule or a bug ?
> > How does an application know that the file is open elsewhere for
> > writing ?
> 
> It is up to you to ensure that you don't set up such a situation, just
> like it is also your responsibility to ensure that you don't run 2
> applications that do read-modify-writes to the same file on a regular
> POSIX filesystem.
> 
> This is a rule that has worked just fine for the NFS community for more
> than 30 years. It isn't anything new that we're only adding to Linux.

Ok, I assume proper use of locking will address the situation and allow
readers to get consistent data and not null bytes ?

Simo.


-- 
Simo Sorce * Red Hat, Inc * New York