diff mbox series

[net-next,1/4] mm: page_frag: Introduce page_frag_alloc_align()

Message ID 20210123115903.31302-2-haokexin@gmail.com (mailing list archive)
State Superseded
Delegated to: Netdev Maintainers
Headers show
Series net: Avoid the memory waste in some Ethernet drivers | expand

Checks

Context Check Description
netdev/apply fail Patch does not apply to net-next
netdev/tree_selection success Clearly marked for net-next

Commit Message

Kevin Hao Jan. 23, 2021, 11:59 a.m. UTC
In the current implementation of page_frag_alloc(), it doesn't have
any align guarantee for the returned buffer address. But for some
hardwares they do require the DMA buffer to be aligned correctly,
so we would have to use some workarounds like below if the buffers
allocated by the page_frag_alloc() are used by these hardwares for
DMA.
    buf = page_frag_alloc(really_needed_size + align);
    buf = PTR_ALIGN(buf, align);

These codes seems ugly and would waste a lot of memories if the buffers
are used in a network driver for the TX/RX. So introduce
page_frag_alloc_align() to make sure that an aligned buffer address is
returned.

Signed-off-by: Kevin Hao <haokexin@gmail.com>
---
 include/linux/gfp.h |  3 +++
 mm/page_alloc.c     | 12 ++++++++++--
 2 files changed, 13 insertions(+), 2 deletions(-)

Comments

Jakub Kicinski Jan. 23, 2021, 8:52 p.m. UTC | #1
On Sat, 23 Jan 2021 19:59:00 +0800 Kevin Hao wrote:
> +void *page_frag_alloc(struct page_frag_cache *nc,
> +		      unsigned int fragsz, gfp_t gfp_mask)
> +{
> +	return page_frag_alloc_align(nc, fragsz, gfp_mask, 0);
> +}
>  EXPORT_SYMBOL(page_frag_alloc);

Isn't it better to make this a static inline now?

Either way you'll need to repost after net is merged into net-next
(probably ~this Friday), please mark the posting as RFC before that.
Please make sure you CC the author of the code.
Kevin Hao Jan. 24, 2021, 3:55 a.m. UTC | #2
On Sat, Jan 23, 2021 at 12:52:21PM -0800, Jakub Kicinski wrote:
> On Sat, 23 Jan 2021 19:59:00 +0800 Kevin Hao wrote:
> > +void *page_frag_alloc(struct page_frag_cache *nc,
> > +		      unsigned int fragsz, gfp_t gfp_mask)
> > +{
> > +	return page_frag_alloc_align(nc, fragsz, gfp_mask, 0);
> > +}
> >  EXPORT_SYMBOL(page_frag_alloc);
> 
> Isn't it better to make this a static inline now?

Sure. I will also inline the {netdev,napi}_alloc_frag().

> 
> Either way you'll need to repost after net is merged into net-next
> (probably ~this Friday), please mark the posting as RFC before that.

Sorry, I missed that. I will repost after the net is merged into net-next.

> Please make sure you CC the author of the code.

Will do.

Thanks,
Kevin
Vlastimil Babka Jan. 26, 2021, 4:19 p.m. UTC | #3
On 1/23/21 12:59 PM, Kevin Hao wrote:
> In the current implementation of page_frag_alloc(), it doesn't have
> any align guarantee for the returned buffer address. But for some
> hardwares they do require the DMA buffer to be aligned correctly,
> so we would have to use some workarounds like below if the buffers
> allocated by the page_frag_alloc() are used by these hardwares for
> DMA.
>     buf = page_frag_alloc(really_needed_size + align);
>     buf = PTR_ALIGN(buf, align);
> 
> These codes seems ugly and would waste a lot of memories if the buffers
> are used in a network driver for the TX/RX. So introduce
> page_frag_alloc_align() to make sure that an aligned buffer address is
> returned.
> 
> Signed-off-by: Kevin Hao <haokexin@gmail.com>

Acked-by: Vlastimil Babka <vbabka@suse.cz>

Agree with Jakub about static inline.

> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -5135,8 +5135,8 @@ void __page_frag_cache_drain(struct page *page, unsigned int count)
>  }
>  EXPORT_SYMBOL(__page_frag_cache_drain);
>  
> -void *page_frag_alloc(struct page_frag_cache *nc,
> -		      unsigned int fragsz, gfp_t gfp_mask)
> +void *page_frag_alloc_align(struct page_frag_cache *nc,
> +		      unsigned int fragsz, gfp_t gfp_mask, int align)
>  {
>  	unsigned int size = PAGE_SIZE;
>  	struct page *page;
> @@ -5188,10 +5188,18 @@ void *page_frag_alloc(struct page_frag_cache *nc,
>  	}
>  
>  	nc->pagecnt_bias--;
> +	offset = align ? ALIGN_DOWN(offset, align) : offset;

We don't change offset if align == 0, so I'd go with simpler
if (align)
	offset = ...

>  	nc->offset = offset;
>  
>  	return nc->va + offset;
>  }
> +EXPORT_SYMBOL(page_frag_alloc_align);
> +
> +void *page_frag_alloc(struct page_frag_cache *nc,
> +		      unsigned int fragsz, gfp_t gfp_mask)
> +{
> +	return page_frag_alloc_align(nc, fragsz, gfp_mask, 0);
> +}
>  EXPORT_SYMBOL(page_frag_alloc);
>  
>  /*
>
diff mbox series

Patch

diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index 6e479e9c48ce..e76e8618e9d7 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -583,6 +583,9 @@  extern void free_pages(unsigned long addr, unsigned int order);
 
 struct page_frag_cache;
 extern void __page_frag_cache_drain(struct page *page, unsigned int count);
+extern void *page_frag_alloc_align(struct page_frag_cache *nc,
+				   unsigned int fragsz, gfp_t gfp_mask,
+				   int align);
 extern void *page_frag_alloc(struct page_frag_cache *nc,
 			     unsigned int fragsz, gfp_t gfp_mask);
 extern void page_frag_free(void *addr);
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 027f6481ba59..80f7c5f7d738 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -5135,8 +5135,8 @@  void __page_frag_cache_drain(struct page *page, unsigned int count)
 }
 EXPORT_SYMBOL(__page_frag_cache_drain);
 
-void *page_frag_alloc(struct page_frag_cache *nc,
-		      unsigned int fragsz, gfp_t gfp_mask)
+void *page_frag_alloc_align(struct page_frag_cache *nc,
+		      unsigned int fragsz, gfp_t gfp_mask, int align)
 {
 	unsigned int size = PAGE_SIZE;
 	struct page *page;
@@ -5188,10 +5188,18 @@  void *page_frag_alloc(struct page_frag_cache *nc,
 	}
 
 	nc->pagecnt_bias--;
+	offset = align ? ALIGN_DOWN(offset, align) : offset;
 	nc->offset = offset;
 
 	return nc->va + offset;
 }
+EXPORT_SYMBOL(page_frag_alloc_align);
+
+void *page_frag_alloc(struct page_frag_cache *nc,
+		      unsigned int fragsz, gfp_t gfp_mask)
+{
+	return page_frag_alloc_align(nc, fragsz, gfp_mask, 0);
+}
 EXPORT_SYMBOL(page_frag_alloc);
 
 /*