Re: [GIT PULL afs: Development for 5.4

[Ilya Dryomov <idryomov@gmail.com>]

On Thu, Sep 19, 2019 at 3:55 PM Matthew Wilcox <willy@infradead.org> wrote:
>
> On Thu, Sep 19, 2019 at 10:49:22AM +0100, David Howells wrote:
> > David Howells <dhowells@redhat.com> wrote:
> >
> > > > However, I was close to unpulling it again. It has a merge commit with
> > > > this merge message:
> > > >
> > > >     Merge remote-tracking branch 'net/master' into afs-next
> > > >
> > > > and that simply is not acceptable.
> > >
> > > Apologies - I meant to rebase that away.  There was a bug fix to rxrpc in
> > > net/master that didn't get pulled into your tree until Saturday.
> >
> > Actually, waiting for all outstanding fixes to get merged and then rebasing
> > might not be the right thing here.  The problem is that there are fixes in
> > both trees: afs fixes go directly into yours whereas rxrpc fixes go via
> > networking and I would prefer to base my patches on both of them for testing
> > purposes.  What's the preferred method for dealing with that?  Base on a merge
> > of the lastest of those fixes in each tree?
>
> Why is it organised this way?  I mean, yes, technically, rxrpc is a
> generic layer-6 protocol that any blah blah blah, but in practice no
> other user has come up in the last 37 years, so why bother pretending
> one is going to?  Just git mv net/rxrpc fs/afs/ and merge everything
> through your tree.
>
> I feel similarly about net/9p, net/sunrpc and net/ceph.  Every filesystem
> comes with its own presentation layer; nobody reuses an existing one.
> Just stop pretending they're separate components.

net/ceph is also being used by drivers/block/rbd.c.  net/ceph was split
out of fs/ceph when rbd was introduced.  We continued to manage them in
a single ceph-client.git tree though.

Thanks,

                Ilya