Re: [PATCH 1/3] mm: Add __page_cache_alloc_order

From: Matthew Wilcox <willy@infradead.org>
To: Song Liu <songliubraving@fb.com>
Cc: Linux MM <linux-mm@kvack.org>,
	"linux-fsdevel@vger.kernel.org" <linux-fsdevel@vger.kernel.org>,
	Kirill Shutemov <kirill@shutemov.name>,
	William Kucharski <william.kucharski@oracle.com>,
	Johannes Weiner <jweiner@fb.com>
Subject: Re: [PATCH 1/3] mm: Add __page_cache_alloc_order
Date: Thu, 5 Sep 2019 12:02:38 -0700	[thread overview]
Message-ID: <20190905190238.GT29434@bombadil.infradead.org> (raw)
In-Reply-To: <75104154-A1A4-4FE3-920C-0069E1B5848D@fb.com>

On Thu, Sep 05, 2019 at 06:58:53PM +0000, Song Liu wrote:
> > On Sep 5, 2019, at 11:23 AM, Matthew Wilcox <willy@infradead.org> wrote:
> > This new function allows page cache pages to be allocated that are
> > larger than an order-0 page.
> > 
> > Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> > ---
> > include/linux/pagemap.h | 14 +++++++++++---
> > mm/filemap.c            | 11 +++++++----
> > 2 files changed, 18 insertions(+), 7 deletions(-)
> > 
> > diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
> > index 103205494ea0..d2147215d415 100644
> > --- a/include/linux/pagemap.h
> > +++ b/include/linux/pagemap.h
> > @@ -208,14 +208,22 @@ static inline int page_cache_add_speculative(struct page *page, int count)
> > }
> > 
> > #ifdef CONFIG_NUMA
> > -extern struct page *__page_cache_alloc(gfp_t gfp);
> > +extern struct page *__page_cache_alloc_order(gfp_t gfp, unsigned int order);
> 
> I guess we need __page_cache_alloc(gfp_t gfp) here for CONFIG_NUMA. 

... no?  The __page_cache_alloc() below is outside the ifdef/else/endif, so
it's the same for both NUMA and non-NUMA.

> > #else
> > -static inline struct page *__page_cache_alloc(gfp_t gfp)
> > +static inline
> > +struct page *__page_cache_alloc_order(gfp_t gfp, unsigned int order)
> > {
> > -	return alloc_pages(gfp, 0);
> > +	if (order > 0)
> > +		gfp |= __GFP_COMP;
> > +	return alloc_pages(gfp, order);
> > }
> > #endif
> > 
> > +static inline struct page *__page_cache_alloc(gfp_t gfp)
> > +{
> > +	return __page_cache_alloc_order(gfp, 0);
> 
> Maybe "return alloc_pages(gfp, 0);" here to avoid checking "order > 0"?

For non-NUMA cases, the __page_cache_alloc_order() will be inlined into
__page_cache_alloc() and the copiler will eliminate the test.  Or you
need a better compiler ;-)

> > -struct page *__page_cache_alloc(gfp_t gfp)
> > +struct page *__page_cache_alloc_order(gfp_t gfp, unsigned int order)
> > {
> > 	int n;
> > 	struct page *page;
> > 
> > +	if (order > 0)
> > +		gfp |= __GFP_COMP;
> > +
> 
> I think it will be good to have separate __page_cache_alloc() for order 0, 
> so that we avoid checking "order > 0", but that may require too much 
> duplication. So I am on the fence for this one. 

We're about to dive into the page allocator ... two extra instructions
here aren't going to be noticable.