Re: [RFC] mm/migrate: Add new migration reason MR_HUGETLB

[Andrew Morton]

On Wed, 31 Jan 2018 08:58:52 +0100 Michal Hocko <mhocko@xxxxxxxxxx> wrote:

> On Wed 31-01-18 07:55:05, Anshuman Khandual wrote:
> > On 01/30/2018 01:29 PM, Michal Hocko wrote:
> > > On Tue 30-01-18 08:37:14, Anshuman Khandual wrote:
> > >> alloc_contig_range() initiates compaction and eventual migration for
> > >> the purpose of either CMA or HugeTLB allocation. At present, reason
> > >> code remains the same MR_CMA for either of those cases. Lets add a
> > >> new reason code which will differentiate the purpose of migration
> > >> as HugeTLB allocation instead.
> > > Why do we need it?
> > 
> > The same reason why we have MR_CMA (maybe some other ones as well) at
> > present, for reporting purpose through traces at the least. It just
> > seemed like same reason code is being used for two different purpose
> > of migration.
> 
> But do we have any real user asking for this kind of information?

It seems a reasonable cleanup: reusing MR_CMA for hugetlb just because
it happens to do the right thing is a bit hacky - the two things aren't
particularly related and a reader could be excused for feeling
confusion.

But the change seems incomplete:

> +		if (migratetype == MIGRATE_CMA)
> +			migrate_reason = MR_CMA;
> +		else
> +			migrate_reason = MR_HUGETLB;

If we're going to do this cleanup then shouldn't we go all the way and
add MIGRATE_HUGETLB?


Alternatively...  instead of adding MR_HUGETLB (and perhaps
MIGRATE_HUGETLB), can we identify what characteristics these two things
have in common and invent a new, more generic identifier?  So that both
migrate-for-CMA and migrate-for-HUGETLB will use MIGRATE_NEWNAME and
MR_NEWNAME?