Re: [PATCH 3/4] NFSD: Add WRITE_PLUS support for hole punches

[Dr Fields James Bruce <bfields@fieldses.org>]

On Tue, Oct 29, 2013 at 01:23:19PM +0000, Myklebust, Trond wrote:
> 
> On Oct 29, 2013, at 9:06 AM, J. Bruce Fields <bfields@fieldses.org> wrote:
> 
> > On Tue, Oct 29, 2013 at 08:50:56AM -0400, Anna Schumaker wrote:
> >> On Mon 28 Oct 2013 05:40:30 PM EDT, J. Bruce Fields wrote:
> >>> On Mon, Oct 28, 2013 at 10:57:25AM -0400, Anna Schumaker wrote:
> >>>> diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> >>>> index 419572f..3210c6f 100644
> >>>> --- a/fs/nfsd/nfs4proc.c
> >>>> +++ b/fs/nfsd/nfs4proc.c
> >>>> @@ -1028,6 +1028,42 @@ nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
> >>>> 	return status;
> >>>> }
> >>>> 
> >>>> +static __be32
> >>>> +nfsd4_write_plus_hole(struct file *file, struct nfsd4_write_plus *writeplus,
> >>>> +		      struct net *net)
> >>>> +{
> >>>> +	__be32 status;
> >>>> +
> >>>> +	status = nfsd4_vfs_fallocate(file, writeplus->wp_allocated,
> >>>> +				writeplus->wp_offset, writeplus->wp_length);
> >>>> +	if (status == nfs_ok) {
> >>>> +		writeplus->wp_res.wr_stid = NULL;
> >>>> +		writeplus->wp_res.wr_bytes_written = writeplus->wp_length;
> >>>> +		writeplus->wp_res.wr_stable_how = NFS_FILE_SYNC;
> >>> 
> >>> Do we need to sync?
> >> 
> >> I did the sync in nfsd4_vfs_fallocate (below), but I can move it if 
> >> that would make more sense.
> > 
> > What I meant was--why are we doing a sync at all, instead of returning
> > NFS_UNSTABLE and making the client commit?
> 
> What if the client specifies FILE_SYNC? The spec doesn't allow the server to return UNSTABLE in that situation.

Of course, sorry for being unclear.  I was just wondering if there's
some particular reason we're not using NFS_UNSTABLE when the client
allows it.

But "because it's easier" is a fine answer.  We can always add that
later if it turns out to be useful.

--b.