> On Apr 12, 2016, at 1:01 PM, Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > > On Tue, Apr 12, 2016 at 10:49:24AM -0400, Chuck Lever wrote: >>>> No, pre-4.2 Linux servers have a bug that can be fixed by >>>> applying the above commit, which shouldn't be difficult to >>>> backport. >>>> >>>> Clients have always been allowed to send NFS READ this way. >>> >>> But the clients might be out in a the wild. I think we'll need at >>> least a mount option to work around them. >> >> Huh? >> >> I'm talking about non-Linux clients. RFC 5666 allows >> this behavior, it always has. The Linux server before >> 9d11b51ce7c1 is broken, full stop. > > You are deliverately breaking exsisting and fairly recent clients. How am I breaking existing clients? Before 4.7, our client works with a broken Linux NFS server and it works with a fixed one. _New_ clients won't work with broken servers. > I see why we want to do this, but there needs to be an options > that things keep working. The idea of upgrading your client from > 4.6 to 4.7 and NFS simply not working anymore with the developer > knowing it and not providing a workaround is simply not an option. I don't understand why this approach was reasonable in the past, but is now suddenly not workable. We have a workaround: Use NFS/TCP on IPoIB. -- Chuck Lever -- 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
