Re: [PATCH v2] mm/madvise: add vmstat statistics for madvise_[cold|pageout]

[Michal Hocko]

On Thu 26-01-23 09:10:46, Minchan Kim wrote:
> On Thu, Jan 26, 2023 at 09:50:37AM +0100, Michal Hocko wrote:
[...]
> > I suspect you try to mimic pgscan/pgsteal effectivness metric on the
> > address space but that is a fundamentally different thing.
> 
> I don't see anything different, fundamentally.

OK, this really explains our disconnect here. Your metric reports
nr_page_tables (nr_scanned) and number of aged and potentially reclaimed
pages. You do not know whether that reclaim was successful. So you
effectively learn how many pages have already been unmapped before your
call. Can this be sometimes useful? Probably yes. Does it say anything
about the reclaim efficiency? I do not think so. You could have hit
pinned pages or countless other conditions why those pages couldn't have
been reclaimed and they have stayed mapped after madvise call.

pgsteal tells you how many pages from those scanned have been reclaimed.
See the difference?

Also I do not find information about how many non-present ptes have
been scann super interesting. Sure that is a burnt time as well but to
me it would be much more valuable information to see how many of those
resident could have been actually reclaimed. Because that tells whether
your reclaim target was a good choice and IMHO that is a valuable
information for user space memory reclaim agent.

Again consider a large sparsely mapped memory but mostly inactive memory
and a condensed active one with the same rss. The reclaim could have
been successful for the former while not on the latter. Your matric
would give a rather misleading numbers, don't you think?
-- 
Michal Hocko
SUSE Labs