Re: [RFC] atomic create+open

[Trond Myklebust]

to den 06.10.2005 Klokka 19:51 (+0200) skreiv Miklos Szeredi:
> > > Say you have NFS mount on /mnt and a bind mount over the regular file
> > > /mnt/foo.  You do open("/mnt/foo", O_RDWR | O_CREAT, 0644).  How do
> > > you solve the atomic open case.
> > 
> > > If you open in ->lookup("foo") you will be opening the wrong file,
> > > unless you want to follow mounts inside ->lookup.
> > 
> > Firstly, if that is the case, then you will have dentries for both the
> > covered and the covering copies of /mnt/foo. A simple test of
> > have_submounts() on the dentry will suffice to tell the filesystem.
> > whether or not it should open the file.
> 
> And if the bind is umounted after NFS determined not to open the file,
> and at the same time it changes to a symlink on the server (not very
> likely I agree, but possible nonetheless), then shit happens.

Please elaborate. What would you expect to see happen in this situation?

> > Secondly, Linux doesn't actually allow bind mounts on top of regular
> > files.
> 
> It does.  Try it.

Nice. I never noticed that.

Cheers,
  Trond

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/