RE: [External] [RFC PATCH v1 3/6] mm, zone_type: create ZONE_NVM and fill into GFP_ZONE_TABLE

From: Huaisheng HS1 Ye
Date: Wed May 09 2018 - 23:54:04 EST


>
> On Wed 09-05-18 14:04:21, Huaisheng HS1 Ye wrote:
> > > From: owner-linux-mm@xxxxxxxxx [mailto:owner-linux-mm@xxxxxxxxx] On Behalf Of
> Michal Hocko
> > >
> > > On Wed 09-05-18 04:22:10, Huaisheng HS1 Ye wrote:
> [...]
> > > > Current mm treats all memory regions equally, it divides zones just by size,
> like
> > > 16M for DMA, 4G for DMA32, and others above for Normal.
> > > > The spanned range of all zones couldn't be overlapped.
> > >
> > > No, this is not correct. Zones can overlap.
> >
> > Hi Michal,
> >
> > Thanks for pointing it out.
> > But function zone_sizes_init decides
> > arch_zone_lowest/highest_possible_pfn's size by max_low_pfn, then
> > free_area_init_nodes/node are responsible for calculating the spanned
> > size of zones from memblock memory regions. So, ZONE_DMA and
> > ZONE_DMA32 and ZONE_NORMAL have separate address scope. How can they
> > be overlapped with each other?
>
> Sorry, I could have been a bit more specific. DMA, DMA32 and Normal
> zones are exclusive. They are mapped to a specific physical range of
> memory so they cannot overlap. I was referring to a general property
> that zones might interleave. Especially zone Normal, Movable and Device.

Exactly, here ZONE_NVM is a real physical range same as ZONE_DMA, ZONE_DMA32 and ZONE_Normal. So, it couldn't overlap with other zones.
Just like you mentioned, ZONE_MOVABLE is virtual zone, which comes ZONE_Normal.
The way of virtual zone is another implementation compared with current patch for ZONE_NVM.
It has advantages but also disadvantages, which need to be clarified and discussed.

Sincerely,
Huaisheng Ye