Re: [PATCH 0/2] vfs: move the clone/dedupe/remap helpers to a single file

[Darrick J. Wong]

On Thu, Oct 15, 2020 at 04:18:19AM +0100, Al Viro wrote:
> On Wed, Oct 14, 2020 at 05:31:14PM -0700, Darrick J. Wong wrote:
> 
> > AFAICT, nobody is attempting to land any major changes in any of the vfs
> > remap functions during the 5.10 window -- for-next showed conflicts only
> > in the Makefile, so it seems like a quiet enough time to do this.  There
> > are no functional changes here, it's just moving code blocks around.
> > 
> > So, I have a few questions, particularly for Al, Andrew, and Linus:
> > 
> > (1) Do you find this reorganizing acceptable?
> 
> No objections, assuming that it's really a move (it's surprisingly easy to
> screw that up - BTDT ;-/)
> 
> I have not done function-by-function comparison, but assuming it holds...
> no problem.

<nod> The only changes between before and after are that some of the
functions lose their static status, and some gain it; and a minor
indenting issue that I'll fix for the final patch series.

As far as makefiles go, both read_write.o and filemap.o are both obj-y
targets, so I think it's safe to make remap_range.o also an obj-y
target.  The fun part will be the careful Kconfig surgery to make
remap_range.o an optional build target, but that will come later.

> > (2) I was planning to rebase this series next Friday and try to throw it
> > in at the end of the merge window; is that ok?  (The current patches are
> > based on 5.9, and applying them manually to current master and for-next
> > didn't show any new conflicts.)
> 
> Up to Linus.  I don't have anything in vfs.git around that area; the
> only remaining stuff touching fs/read_write.c is nowhere near those,
> AFAICS.

<nod> Thanks. :)

--D