[PATCH] copy-chunk fails if src and dst file are on different tcons

[David Disseldorp]

On Mon, 3 Jul 2017 15:26:24 +0200, Ralph Böhme wrote:

> > > see one regression when running the fs_specific dup_extents_src_lock and
> > > dup_extents_dest_lock tests atop btrfs (which supports dup-extents):
...
> woo, thanks a lot for spotting this!
> 
> Turns out there was a simple logic flaw in the offload token DB: trying to
> insert a token that was already stored was not allowed. That of course breaks
> the permissible SMB2 behaviour of calling the fetch-resume-key ioctl twice for
> the same filehandle, or, in the failing btrfs test case, calling the dup-extents
> ioctl twice on the same file-handle.
> 
> Attached is an updated patchset, I'm also attaching the diff-diff. This one
> passes the dup-extents smbtorture tests on a btrfs fs.

Thanks for the follow-up! The diff-diff looks fine. I'll leave the final
ack up to Metze.

Cheers, David