On Wed, Feb 24, 2021 at 04:50:39PM -0700, Yu Zhao wrote:

On Wed, Feb 24, 2021 at 10:48:46PM +0000, Matthew Wilcox wrote:

quoted

On Wed, Feb 24, 2021 at 03:34:16PM -0700, Yu Zhao wrote:

quoted

quoted

If only somebody were working on a patch series to get rid of
all those calls to compound_head()!  Some reviews on
https://lore.kernel.org/linux-mm/20210128070404.1922318-2-willy@infradead.org/ (local)
would be nice.

I'm on board with the idea and have done some research in this
direction. We've found that the ideal *anon* page size for Chrome OS
is not 4KB or 2MB, but 32KB. I hope we could leverage the folio to
support flexible anon page size to reduce the number of page faults
(vs 4KB) or internal fragmentation (vs 2MB).

That being said, it seems to me this is a long term plan and right
now we need something smaller. So if you don't mind, I'll just go
ahead and remove compound_head() from Page{LRU,Active,Unevictable,
SwapBacked} first?

It's really not a big change I'm suggesting here.  You need
https://lore.kernel.org/linux-mm/20210128070404.1922318-2-willy@infradead.org/ (local)
https://lore.kernel.org/linux-mm/20210128070404.1922318-5-willy@infradead.org/ (local)
https://lore.kernel.org/linux-mm/20210128070404.1922318-8-willy@infradead.org/ (local)
and then the patch I sent above to create folio_lru().

Then any changes you want to make to use folios more broadly will
incrementally move us towards your goal of 32kB anon pages.

Well, these patches introduce a new concept which I'm on board with.

It's not really a new concept ... it's a new type for an existing concept
(a head page).

Assume everybody else is too, it still seems to me it's an overkill
to employee folio to just get rid of unnecessary compound_head()
in page_lru() -- this is not a criticism but a compliment.

It's not overkill, that really is the point of a folio!  If you
think about it, only head pages can be on the LRU list (because the
compound_head is in the union with the lru list_head).  So it
always makes sense to talk about folios on the LRU list.

Let me work out something *conceptually* smaller first, and if you
think folio is absolutely more suitable even for this specific issue,
I'll go review and test the four patches you listed. Sounds good?

Umm.  It seems to me that no matter what you do, it'll be equivalent to
this, only without the type-safety?