[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1426088560.11398.44.camel@edumazet-glaptop2.roam.corp.google.com>
Date: Wed, 11 Mar 2015 08:42:40 -0700
From: Eric Dumazet <eric.dumazet@...il.com>
To: Alexander Duyck <alexander.duyck@...il.com>
Cc: Govindarajulu Varadarajan <_govind@....com>,
Alexander Duyck <alexander.h.duyck@...hat.com>,
davem@...emloft.net, netdev@...r.kernel.org, ssujith@...co.com,
benve@...co.com
Subject: Re: [PATCH net-next v3 1/2] net: implement dma cache skb allocator
On Wed, 2015-03-11 at 06:55 -0700, Alexander Duyck wrote:
> That definitely helps but still leaves you open since a 257 byte packet
> would be consuming 9K in that case. Also standard flows with frames of
> 1514 still take a hard hit with a receive buffer size of 9K.
One possible way to deal with this would be to try to use order-2 pages,
(but not compound pages) but break them.
split_page(page, 2);
struct page *page0 = page;
struct page *page1 = page + 1;
struct page *page2 = page + 2;
struct page *page3 = page + 3;
if frame length <= 4096, attach page0 as skb first frag, and free
page[1-3]
If frame length <= 8192 bytes, driver can attach page0 & page1 to the
skb, and free page[2-3]
Otherwise, attach page[0-2] to skb and free page3
Not sure if buddy allocator will be pleased, but worth trying ?
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists