Re: [RFC 0/3] mm: Discard lazily freed pages when migrating

[Huang\, Ying]

Michal Hocko <mhocko@xxxxxxxxxx> writes:

> On Mon 02-03-20 22:12:53, Huang, Ying wrote:
>> Michal Hocko <mhocko@xxxxxxxxxx> writes:
> [...]
>> > And MADV_FREE pages are a kind of cache as well. If the target node is
>> > short on memory then those will be reclaimed as a cache so a
>> > pro-active freeing sounds counter productive as you do not have any
>> > idea whether that cache is going to be used in future. In other words
>> > you are not going to free a clean page cache if you want to use that
>> > memory as a migration target right? So you should make a clear case
>> > about why MADV_FREE cache is less important than the clean page cache
>> > and ideally have a good justification backed by real workloads.
>> 
>> Clean page cache still have valid contents, while clean MADV_FREE pages
>> has no valid contents.  So penalty of discarding the clean page cache is
>> reading from disk, while the penalty of discarding clean MADV_FREE pages
>> is just page allocation and zeroing.
>
> And "just page allocation and zeroing" overhead is the primary
> motivation to keep the page in memory. It is a decision of the workload
> to use MADV_FREE because chances are that this will speed things up. All
> that with a contract that the memory goes away under memory pressure so
> with a good workload/memory sizing you do not really lose that
> optimization. Now you want to make decision on behalf of the consumer of
> the MADV_FREE memory.

I understand that MADV_FREE helps in some situations.  And if the
application want to keep the "contract" after migration, they should
have a way to do that.

>> I understand that MADV_FREE is another kind of cache and has its value.
>> But in the original implementation, during migration, we have already
>> freed the original "cache", then reallocate the cache elsewhere and
>> copy.  This appears more like all pages are populated in mmap() always.
>> I know there's value to populate all pages in mmap(), but does that need
>> to be done always by default?
>
> It is not. You have to explicitly request MAP_POPULATE to initialize
> mmap.

Yes.  mmap() can control whether to populate the underlying physical
pages.  But for migrating MADV_FREE pages, there's no control, all pages
will be populated again always by default.  Maybe we should avoid to do
that in some situations too.

Best Regards,
Huang, Ying