Re: [PATCH] mm/page_alloc: don't reserve ZONE_HIGHMEM for ZONE_MOVABLE request
From: Joonsoo Kim
Date: Mon Aug 28 2017 - 20:36:22 EST
On Mon, Aug 28, 2017 at 08:45:07AM +0200, Vlastimil Babka wrote:
> +CC linux-api
>
> On 08/28/2017 02:28 AM, Joonsoo Kim wrote:
> > On Fri, Aug 25, 2017 at 09:56:10AM +0200, Vlastimil Babka wrote:
> >> On 08/25/2017 02:20 AM, Joonsoo Kim wrote:
> >>> On Thu, Aug 24, 2017 at 11:41:58AM +0200, Vlastimil Babka wrote:
> >>>
> >>> Hmm, this is already pointed by Minchan and I have answered that.
> >>>
> >>> lkml.kernel.org/r/<20170421013243.GA13966@js1304-desktop>
> >>>
> >>> If you have a better idea, please let me know.
> >>
> >> My idea is that size of sysctl_lowmem_reserve_ratio is ZONE_NORMAL+1 and
> >> it has no entries for zones > NORMAL. The
> >> setup_per_zone_lowmem_reserve() is adjusted to only set
> >> lower_zone->lowmem_reserve[j] for idx <= ZONE_NORMAL.
> >>
> >> I can't imagine somebody would want override the ratio for HIGHMEM or
> >> MOVABLE
> >> (where it has no effect anyway) so the simplest thing is not to expose
> >> it at all.
> >
> > Seems reasonable. However, if there is a user who checks
> > sysctl_lowmem_reserve_ratio entry for HIGHMEM and change it, suggested
> > interface will cause a problem since it doesn't expose ratio for
> > HIGHMEM. Am I missing something?
>
> As you explained, it makes little sense to change it for HIGHMEM which
> only affects MOVABLE allocations. Also I doubt there are many systems
> with both HIGHMEM (implies 32bit) *and* MOVABLE (implies NUMA, memory
> hotplug...) zones. So I would just remove it, and if somebody will
> really miss it, we can always add it back. In any case, please CC
> linux-api on the next version.
If we will accept a change that potentially breaks the user, I think
that making zero as a special value for sysctl_lowmem_reserve_ratio
is better solution. How about this way?
Thanks.