Re: Unhelpful caching decisions, possibly related to active/inactive sizing

[Johannes Weiner <hannes@cmpxchg.org>]

On Fri, Feb 12, 2016 at 08:35:53PM +0100, Andres Freund wrote:
> To make an actually usable patch out of this it seems we'd have to add a
> 'partial' argument to grab_cache_page_write_begin(), so writes to parts
> of a page still cause the pages to be marked active.  Is it preferrable
> to change all callers of grab_cache_page_write_begin and
> add_to_page_cache_lru or make them into wrapper functions, and call the
> real deal when it matters?

Personally, I'd prefer explicit arguments over another layer of
wrappers, especially in the add_to_page_cache family. But it's
possible others will disagree and only voice their opinion once you
went through the hassle and sent a patch.

> I do think that that's a reasonable algorithmic change, but nonetheless
> its obviously possible that such changes regress some workloads. What's
> the policy around testing such things?

How about a FGP_WRITE that only sets the page's referenced bit, but
doesn't activate or refault-activate the page?

That way, pages that are only ever written would never get activated,
but a single read mixed in would activate the page straightaway;
either in mark_page_accessed() or through refault-activation.