Re: [PATCH v6 00/26] fs/dax: Fix ZONE_DEVICE page reference counts

[Alistair Popple]

On Fri, Jan 10, 2025 at 07:35:57PM -0800, Dan Williams wrote:
> Andrew Morton wrote:
> > On Thu, 9 Jan 2025 23:05:56 -0800 Dan Williams <dan.j.williams@xxxxxxxxx> wrote:
> > 
> > > >  - Remove PTE_DEVMAP definitions from Loongarch which were added since
> > > >    this series was initially written.
> > > [..]
> > > > 
> > > > base-commit: e25c8d66f6786300b680866c0e0139981273feba
> > > 
> > > If this is going to go through nvdimm.git I will need it against a
> > > mainline tag baseline. Linus will want to see the merge conflicts.
> > > 
> > > Otherwise if that merge commit is too messy, or you would rather not
> > > rebase, then it either needs to go one of two options:
> > > 
> > > - Andrew's tree which is the only tree I know of that can carry
> > >   patches relative to linux-next.
> > 
> > I used to be able to do that but haven't got around to setting up such
> > a thing with mm.git.  This is the first time the need has arisen,
> > really.
> 
> Oh, good to know.
> 
> > 
> > > - Wait for v6.14-rc1 
> > 
> > I'm thinking so.  Darrick's review comments indicate that we'll be seeing a v7.

I'm ok with that. It could do with a decent soak in linux-next anyway given it
touches a lot of mm and fs.

Once v6.14-rc1 is released I will do a rebase on top of that.

> > > and get this into nvdimm.git early in the cycle
> > >   when the conflict storm will be low.
> > 
> > erk.  This patchset hits mm/ a lot, and nvdimm hardly at all.  Is it
> > not practical to carry this in mm.git?
> 
> I'm totally fine with it going through mm.git. nvdimm.git is just the
> historical path for touches to fs/dax.c, and git blame points mostly to
> me for the issues Alistair is fixing. I am happy to review and ack and
> watch this go through mm.git.