On Nov 2, 2013, at 10:51, Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > On Sat, Nov 02, 2013 at 10:44:31AM -0400, J. Bruce Fields wrote: >> On Sat, Nov 02, 2013 at 06:54:09AM -0700, Christoph Hellwig wrote: >>> On Tue, Oct 29, 2013 at 09:06:49AM -0400, J. Bruce Fields wrote: >>>> What I meant was--why are we doing a sync at all, instead of returning >>>> NFS_UNSTABLE and making the client commit? >>> >>> Did NFSv4.2 introduce a concept of unstable metadata operations? >> >> No, but I think WRITE_PLUS does have a stable/unstable bit so I think we >> could choose not to do the sync if that'd make sense. > > Both operations are idempotent, so supporting it shouldn't be a major > obstactle. But suddenly having some metadata operations that can be > unstable seems like a major wart in the spec. COMMIT has always had a metadata part to it, though. It guarantees stability of the ctime/mime and change attributes in addition to the file size and data. IOW: it really is more akin to fsync() than to fdatasync(). Cheers Trond-- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html