On Tue, Jan 07 2020, J. Bruce Fields wrote: > On Fri, Dec 20, 2019 at 04:19:56PM +1100, NeilBrown wrote: >> I was a bit surprised that nfs4_map_atomic_open_claim() exists at all, >> but given that it did, I assumed it would be used more uniformly. >> >> So this all implies that Linux NFS server claimed to support NFSv4.1 >> before it actually did - which seems odd. This is just a bug (which are >> expected), but a clear ommission. > > For what it's worth, I did make some attempt to keep 4.1 by default > until 3.11 (see d109148111cd "nfsd4: support minorversion 1 by default") > but probably could have communicated that better. This isn't the only > blatant known issue in older code. Ahh... thanks for that. Looking more deeply, I see that we (SUSE) left it that way, but there is a sysconfig option to explicitly enable NFSv4.1 service, and the customer has explicitly enabled that. So it is sort-of there fault. Maybe we shouldn't have given them the option. Anyway, it is all clear now. Thanks. NeilBrown
Attachment:
signature.asc
Description: PGP signature
