From: Kim Phillips <kim.phillips@freescale.com>
To: David Miller <davem@davemloft.net>
Cc: <stable@vger.kernel.org>, <eric.dumazet@gmail.com>,
<zheng.z.yan@intel.com>, <netdev@vger.kernel.org>,
David Ward <david.ward@ll.mit.edu>
Subject: [stable] net: Handle different key sizes between address families in flow cache
Date: Fri, 4 Nov 2011 14:46:59 -0500 [thread overview]
Message-ID: <20111104144659.3e0a30afffffd9dd73cf740e@freescale.com> (raw)
In-Reply-To: <20111103.224329.2276058062630695656.davem@davemloft.net>
commit aa1c366e4febc7f5c2b84958a2dd7cd70e28f9d0 upstream.
With the conversion of struct flowi to a union of AF-specific structs, some
operations on the flow cache need to account for the exact size of the key.
Signed-off-by: David Ward <david.ward@ll.mit.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: stable@vger.kernel.org # v3.0.x
---
This patch is the result of a clean cherry-pick onto v3.0.8.
It restores IPSec fwding performance from ~4kpps back to ~44kpps on
a P2020DS board.
include/net/flow.h | 19 +++++++++++++++++++
net/core/flow.c | 31 +++++++++++++++++--------------
2 files changed, 36 insertions(+), 14 deletions(-)
diff --git a/include/net/flow.h b/include/net/flow.h
index c6d5fe5..93a8785 100644
--- a/include/net/flow.h
+++ b/include/net/flow.h
@@ -7,6 +7,7 @@
#ifndef _NET_FLOW_H
#define _NET_FLOW_H
+#include <linux/socket.h>
#include <linux/in6.h>
#include <asm/atomic.h>
@@ -161,6 +162,24 @@ static inline struct flowi *flowidn_to_flowi(struct flowidn *fldn)
return container_of(fldn, struct flowi, u.dn);
}
+typedef unsigned long flow_compare_t;
+
+static inline size_t flow_key_size(u16 family)
+{
+ switch (family) {
+ case AF_INET:
+ BUILD_BUG_ON(sizeof(struct flowi4) % sizeof(flow_compare_t));
+ return sizeof(struct flowi4) / sizeof(flow_compare_t);
+ case AF_INET6:
+ BUILD_BUG_ON(sizeof(struct flowi6) % sizeof(flow_compare_t));
+ return sizeof(struct flowi6) / sizeof(flow_compare_t);
+ case AF_DECnet:
+ BUILD_BUG_ON(sizeof(struct flowidn) % sizeof(flow_compare_t));
+ return sizeof(struct flowidn) / sizeof(flow_compare_t);
+ }
+ return 0;
+}
+
#define FLOW_DIR_IN 0
#define FLOW_DIR_OUT 1
#define FLOW_DIR_FWD 2
diff --git a/net/core/flow.c b/net/core/flow.c
index 990703b..a6bda2a 100644
--- a/net/core/flow.c
+++ b/net/core/flow.c
@@ -172,29 +172,26 @@ static void flow_new_hash_rnd(struct flow_cache *fc,
static u32 flow_hash_code(struct flow_cache *fc,
struct flow_cache_percpu *fcp,
- const struct flowi *key)
+ const struct flowi *key,
+ size_t keysize)
{
const u32 *k = (const u32 *) key;
+ const u32 length = keysize * sizeof(flow_compare_t) / sizeof(u32);
- return jhash2(k, (sizeof(*key) / sizeof(u32)), fcp->hash_rnd)
+ return jhash2(k, length, fcp->hash_rnd)
& (flow_cache_hash_size(fc) - 1);
}
-typedef unsigned long flow_compare_t;
-
/* I hear what you're saying, use memcmp. But memcmp cannot make
- * important assumptions that we can here, such as alignment and
- * constant size.
+ * important assumptions that we can here, such as alignment.
*/
-static int flow_key_compare(const struct flowi *key1, const struct flowi *key2)
+static int flow_key_compare(const struct flowi *key1, const struct flowi *key2,
+ size_t keysize)
{
const flow_compare_t *k1, *k1_lim, *k2;
- const int n_elem = sizeof(struct flowi) / sizeof(flow_compare_t);
-
- BUILD_BUG_ON(sizeof(struct flowi) % sizeof(flow_compare_t));
k1 = (const flow_compare_t *) key1;
- k1_lim = k1 + n_elem;
+ k1_lim = k1 + keysize;
k2 = (const flow_compare_t *) key2;
@@ -215,6 +212,7 @@ flow_cache_lookup(struct net *net, const struct flowi *key, u16 family, u8 dir,
struct flow_cache_entry *fle, *tfle;
struct hlist_node *entry;
struct flow_cache_object *flo;
+ size_t keysize;
unsigned int hash;
local_bh_disable();
@@ -222,6 +220,11 @@ flow_cache_lookup(struct net *net, const struct flowi *key, u16 family, u8 dir,
fle = NULL;
flo = NULL;
+
+ keysize = flow_key_size(family);
+ if (!keysize)
+ goto nocache;
+
/* Packet really early in init? Making flow_cache_init a
* pre-smp initcall would solve this. --RR */
if (!fcp->hash_table)
@@ -230,11 +233,11 @@ flow_cache_lookup(struct net *net, const struct flowi *key, u16 family, u8 dir,
if (fcp->hash_rnd_recalc)
flow_new_hash_rnd(fc, fcp);
- hash = flow_hash_code(fc, fcp, key);
+ hash = flow_hash_code(fc, fcp, key, keysize);
hlist_for_each_entry(tfle, entry, &fcp->hash_table[hash], u.hlist) {
if (tfle->family == family &&
tfle->dir == dir &&
- flow_key_compare(key, &tfle->key) == 0) {
+ flow_key_compare(key, &tfle->key, keysize) == 0) {
fle = tfle;
break;
}
@@ -248,7 +251,7 @@ flow_cache_lookup(struct net *net, const struct flowi *key, u16 family, u8 dir,
if (fle) {
fle->family = family;
fle->dir = dir;
- memcpy(&fle->key, key, sizeof(*key));
+ memcpy(&fle->key, key, keysize * sizeof(flow_compare_t));
fle->object = NULL;
hlist_add_head(&fle->u.hlist, &fcp->hash_table[hash]);
fcp->hash_count++;
--
1.7.7.2
next prev parent reply other threads:[~2011-11-04 19:48 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-10-23 7:58 [PATCH] ipv4: fix ipsec forward performance regression Yan, Zheng
2011-10-23 9:03 ` Eric Dumazet
2011-10-24 7:02 ` David Miller
2011-11-01 23:50 ` Kim Phillips
2011-11-02 0:34 ` David Miller
2011-11-03 18:58 ` Kim Phillips
2011-11-04 2:43 ` David Miller
2011-11-04 19:46 ` Kim Phillips [this message]
2011-11-04 20:41 ` [stable] net: Handle different key sizes between address families in flow cache David Miller
2011-11-04 21:24 ` Greg KH
2011-11-05 2:29 ` Kim Phillips
2011-11-08 16:53 ` Greg KH
2011-11-08 19:44 ` [stable v2] " Kim Phillips
2011-10-23 14:52 ` [PATCH] ipv4: fix ipsec forward performance regression Julian Anastasov
2011-10-24 0:41 ` Yan, Zheng
2011-10-24 7:01 ` 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=20111104144659.3e0a30afffffd9dd73cf740e@freescale.com \
--to=kim.phillips@freescale.com \
--cc=davem@davemloft.net \
--cc=david.ward@ll.mit.edu \
--cc=eric.dumazet@gmail.com \
--cc=netdev@vger.kernel.org \
--cc=stable@vger.kernel.org \
--cc=zheng.z.yan@intel.com \
/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).