Re: [HMM 01/16] mm/memory/hotplug: add memory type parameter to arch_add/remove_memory

From: Jerome Glisse
Date: Fri Apr 07 2017 - 14:28:08 EST


On Fri, Apr 07, 2017 at 07:59:12PM +0200, Michal Hocko wrote:
> On Fri 07-04-17 13:10:59, Jerome Glisse wrote:
> > On Fri, Apr 07, 2017 at 06:37:37PM +0200, Michal Hocko wrote:
> > > On Fri 07-04-17 12:10:00, Jerome Glisse wrote:
> [...]
> > > > No guaranteed so yes i somewhat care about max_pfn, i do not care about
> > > > any of its existing user last time i check but it might matter for some
> > > > new user.
> > >
> > > OK, then we can add add_pages() which would do __add_pages by default
> > > (#ifndef ARCH_HAS_ADD_PAGES) and x86 would override it do also call
> > > update_end_of_memory_vars. This sounds easier to me than updating all
> > > the archs and add something that most of them do not really care about.
> > >
> > > But I will not insist. If you think that your approach is better I will
> > > not object.
> >
> > Something like attached patch ?
>
> No I meant something like the diff below but maybe even that is too
> excessive.

No looks good to me at least. But i am no authority there.


> > > Btw. is your series reviewed and ready to be applied to the mm tree? I
> > > planed to post mine on Monday so I would like to know how do we
> > > coordinate. I rebase on topo of yours or vice versa.
> >
> > Well v18 core patches were review by Mel, i did include all of his comment
> > in v19 (i don't think i did miss any). I think Dan still want to look at
> > patch 1 and 3 for ZONE_DEVICE.
> >
> > But i always welcome more review. I know Anshuman replied to this patch
> > to improve a comments. Balbir had issue on powerpc because iomem_resource.end
> > isn't clamped to MAX_PHYSMEM_BITS But that is all review i got so far on v19.
> >
> > I don't mind rebasing on top of your patchset. What ever is easier for
> > Andrew i guess.
>
> Well, considering that my patchset is changing the behavior of the core
> of the memory hotplug I would prefer if it could go first and add new
> user on top. But I realize that you are maintaining your series for a
> _long_ time so I would completely understand if you wouldn't be
> impressed by another rebase...
>
> If you are OK with rebasing and I will help you with that as much as I
> can I would be really grateful.


I don't mind rebasing on top of your patchset after you post. This is minor
change for me.

Cheers,
Jérôme