netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Ian Campbell <Ian.Campbell@citrix.com>
To: Eric Dumazet <eric.dumazet@gmail.com>
Cc: "netdev@vger.kernel.org" <netdev@vger.kernel.org>,
	David Miller <davem@davemloft.net>
Subject: Re: [PATCH 2/6] net: pad skb data and shinfo as a whole rather than individually
Date: Thu, 5 Jan 2012 19:19:24 +0000	[thread overview]
Message-ID: <1325791164.29084.8.camel@dagon.hellion.org.uk> (raw)
In-Reply-To: <1325785607.4759.12.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC>

On Thu, 2012-01-05 at 17:46 +0000, Eric Dumazet wrote:
> Le jeudi 05 janvier 2012 à 17:13 +0000, Ian Campbell a écrit :
> > diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> > index da0c97f..b6de604 100644
> > --- a/net/core/skbuff.c
> > +++ b/net/core/skbuff.c
> > @@ -189,8 +189,7 @@ struct sk_buff *__alloc_skb(unsigned int size, gfp_t gfp_mask,
> >  	 * aligned memory blocks, unless SLUB/SLAB debug is enabled.
> >  	 * Both skb->head and skb_shared_info are cache line aligned.
> >  	 */
> > -	size = SKB_DATA_ALIGN(size);
> > -	size += SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
> > +	size = SKB_DATA_ALIGN(size + sizeof(struct skb_shared_info));
> >  	data = kmalloc_node_track_caller(size, gfp_mask, node);
> >  	if (!data)
> >  		goto nodata;
> 
> Unfortunately this makes the frequently use part of skb_shared_info not
> any more in a single cache line.
> 
> This will slow down many operations like kfree_skb() of cold skbs (TX
> completion for example)

I've been toying with the following which moves the infrequently used
destructor_arg field to the front. With 32 or 64 byte cache lines this
is then the only field which ends up on a different cache line. With 128
byte cache lines it is all already on the same line.

Does that alleviate your concern? If so I can clean it up and include it
in the series.

Cheers,
Ian.

diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index cbc815c..533c2ea 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -249,6 +249,11 @@ struct ubuf_info {
  * the end of the header data, ie. at skb->end.
  */
 struct skb_shared_info {
+	/* Intermediate layers must ensure that destructor_arg
+	 * remains valid until skb destructor */
+	void *		destructor_arg;
+
+	/* Warning all fields from here until dataref are cleared in __alloc_skb() */
 	unsigned char	nr_frags;
 	__u8		tx_flags;
 	unsigned short	gso_size;
@@ -264,9 +269,6 @@ struct skb_shared_info {
 	 */
 	atomic_t	dataref;
 
-	/* Intermediate layers must ensure that destructor_arg
-	 * remains valid until skb destructor */
-	void *		destructor_arg;
 
 	/* must be last field, see pskb_expand_head() */
 	skb_frag_t	frags[MAX_SKB_FRAGS];
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index e942a86..592b8a8 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -219,7 +219,10 @@ struct sk_buff *__alloc_skb(unsigned int size, gfp_t gfp_mask,
 
 	/* make sure we initialize shinfo sequentially */
 	shinfo = skb_shinfo(skb);
-	memset(shinfo, 0, offsetof(struct skb_shared_info, dataref));
+
+	memset(&shinfo->nr_frags, 0,
+	       offsetof(struct skb_shared_info, dataref)
+	       - offsetof(struct skb_shared_info, nr_frags));
 	atomic_set(&shinfo->dataref, 1);
 	kmemcheck_annotate_variable(shinfo->destructor_arg);
 

  parent reply	other threads:[~2012-01-05 19:19 UTC|newest]

Thread overview: 27+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-01-05 17:09 [PATCH 0/6 v2] skb paged fragment destructors Ian Campbell
2012-01-05 17:13 ` [PATCH 1/6] net: pack skb_shared_info more efficiently Ian Campbell
2012-01-05 18:30   ` Eric Dumazet
2012-01-05 19:09     ` David Miller
2012-01-05 17:13 ` [PATCH 2/6] net: pad skb data and shinfo as a whole rather than individually Ian Campbell
2012-01-05 17:46   ` Eric Dumazet
2012-01-05 19:13     ` David Miller
2012-01-05 20:00       ` Ian Campbell
2012-01-07 18:18         ` David Miller
2012-01-05 19:19     ` Ian Campbell [this message]
2012-01-05 20:22       ` Eric Dumazet
2012-01-06 11:20         ` Ian Campbell
2012-01-06 12:33           ` Eric Dumazet
2012-01-06 13:20             ` Ian Campbell
2012-01-06 13:37               ` Eric Dumazet
2012-01-06 13:49                 ` Ian Campbell
2012-01-06 13:29             ` Ian Campbell
2012-01-05 17:13 ` [PATCH 3/6] net: add support for per-paged-fragment destructors Ian Campbell
2012-01-06  1:15   ` Michał Mirosław
2012-01-06  8:50     ` Ian Campbell
2012-01-05 17:13 ` [PATCH 4/6] net: only allow paged fragments with the same destructor to be coalesced Ian Campbell
2012-01-05 17:13 ` [PATCH 5/6] net: add paged frag destructor support to kernel_sendpage Ian Campbell
2012-01-05 19:15   ` David Miller
2012-01-05 20:04     ` Ian Campbell
     [not found] ` <1325783399.25206.413.camel-o4Be2W7LfRlXesXXhkcM7miJhflN2719@public.gmane.org>
2012-01-05 17:13   ` [PATCH 6/6] sunrpc: use SKB fragment destructors to delay completion until page is released by network stack Ian Campbell
2012-01-05 17:27 ` [PATCH 0/6 v2] skb paged fragment destructors David Laight
2012-01-05 17:34   ` David Miller

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1325791164.29084.8.camel@dagon.hellion.org.uk \
    --to=ian.campbell@citrix.com \
    --cc=davem@davemloft.net \
    --cc=eric.dumazet@gmail.com \
    --cc=netdev@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).