Re: [RFC PATCH 34/40] btrfs: allocate eb-attached btree pages as movable

[Matthew Wilcox <willy@infradead.org>]

On Sat, May 23, 2026 at 09:43:23PM -0400, Rik van Riel wrote:
> On Sat, 2026-05-23 at 17:58 +0200, David Sterba wrote:
> > On Wed, May 20, 2026 at 10:47:49AM -0700, Boris Burkov wrote:
> > > My gut says it is nicer to just transparently pass along a gfp flag
> > > rather than a carefully documented movable boolean and let the
> > > callers
> > > do the right thing. Just a nit, I don't care strongly about it.
> > 
> > Agreed, the GFP flags would be better and also a bit more self
> > documenting where the functions are called. The patch is not
> > standalone
> > so I don't mind having the booleans, we can clean it up later to let
> > the
> > 40-patch series proceed.
> 
> 
> I think this patch could go in by itself. It doesn't
> depend on anything else in the series.
> 
> I sent it with this series mostly because it's needed
> to make the series work, and to provide context on why
> it's needed.
> 
> I'm happy to resend it with a GFP mask passed in by
> each caller. That would look better, indeed!

If you're respinning anyway ...

Every use of GFP_NOFS is a bad code smell.  Filesystems should be using
memalloc_nofs_save/restore.  The problem is that it takes deep filesystem
knowledge to understand where the memalloc_nofs_save() should start.
It would be good to lift the GFP_NOFS to the callers of this function as
it gets us slightly closer to being able to replace it with GFP_KERNEL
in the places it's not actually needed.

ie what I'm asking for is, please don't do the 'extra_gfp' thing.
Just pass in gfp from the callers and make them all pass in GFP_NOFS |
__GFP_MOVABLE (or a plain GFP_NOFS) for now.  Someone can come along
later and change them to GFP_KERNEL later where warranted.