From: Linus Torvalds Date: Tue, 15 Sep 2009 09:30:54 -0700 (PDT) Subject: [Ocfs2-devel] [GIT PULL] ocfs2 changes for 2.6.32 In-Reply-To: <20090915040601.GE4507@mail.oracle.com> References: <20090911200458.GA15416@mail.oracle.com> <20090914221434.GA4507@mail.oracle.com> <20090915000417.GC4507@mail.oracle.com> <20090915005417.GD4507@mail.oracle.com> <20090915040601.GE4507@mail.oracle.com> Message-ID: List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Joel Becker Cc: Mark Fasheh , Andrew Morton , Linux Kernel Mailing List , ocfs2-devel@oss.oracle.com On Mon, 14 Sep 2009, Joel Becker wrote: > > > > If you're talking about falling back to manually just copying the data, > > then nobody is interested in that. User space can do that better with a > > simple read-write loop or with splice, or whatever. There's no reaason > > what-so-ever to do that. > > I'm talking about any facility for copying that isn't just a > userspace loop. Like your discussion of network filesystems. HOW? We need to have a per-filesystem interface to that. Having a '->copyfile()' function would be great. But don't you see how _idiotic_ it is to then also having a '->reflink()' function that does _conceptually_ the exact same thing, except it does it by incrementing a usage count instead? Do you see why I'm so unhappy to add a ->reflink() function? > Hence I brought this to the filesystem summit and then fsdevel > rather than just implementing it in ocfs2. I know NFS folks were in the > room in April, and they said the call definition was workable. Can't > remember if CIFS folks were there, but I think so. It's not workable if you define the 'reflink()' function to not use any disk space on the filesystem. Because SMB _will_ do a copy (and I presume the NFS thing will too). So it would not in general be what you call reflink, it will not be a "snapshot". So if you _define_ the semantics of "reflink" to be that it's atomic and doesn't use any new diskspace (apart from the new inode/directory entry, of course), then it will be almost totally useless to other filesystems. In fact, it's entirely possible to have filesystems that can avoid copying the _data_ blocks, but would need to copy the indirect blocks - maybe the data blocks are ref-counted, but the metadata needs to be per-file (I can see many reasons to do it that way, even if it's organized as a tree - it's how we do page table COW, for example, and it makes some things much simpler). Would that be a 'reflink()' or not? I have no way of knowing, because you have decided on reflink on a purely ocfs2-specific implementation basis. But I do know that such a filesystem would be perfectly happy to have a 'copyfile' function. This is why I want the VFS pointers to be about _semantics_, not about some random implementation detail. Linus