From: Eric Dumazet <eric.dumazet@gmail.com>
To: David Miller <davem@davemloft.net>
Cc: netdev <netdev@vger.kernel.org>,
Netfilter Development Mailinglist
<netfilter-devel@vger.kernel.org>,
Patrick McHardy <kaber@trash.net>,
Pablo Neira Ayuso <pablo@netfilter.org>
Subject: [PATCH net-next] netfilter: use jump_label for nf_hooks
Date: Sat, 19 Nov 2011 04:32:46 +0100 [thread overview]
Message-ID: <1321673566.3157.14.camel@edumazet-laptop> (raw)
On configs where CONFIG_JUMP_LABEL=y, we can replace in fast path a
load/compare/conditional jump by a single jump with no dcache reference.
Jump target is modified as soon as nf_hooks[pf][hook] switches from
empty state to non empty states. jump_label state is kept outside of
nf_hooks array so has no cost on cpu caches.
This patch removes the test on CONFIG_NETFILTER_DEBUG : No need to call
nf_hook_slow() at all if nf_hooks[pf][hook] is empty, this didnt give
useful information, but slowed down things a lot.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Patrick McHardy <kaber@trash.net>
CC: Pablo Neira Ayuso <pablo@netfilter.org>
---
This is obviously a netfilter patch, but since David committed two other
jump_label patches lately, maybe its better he takes this patch as well
directly ? Thanks !
include/linux/netfilter.h | 26 +++++++++++++++++++++-----
net/netfilter/core.c | 13 ++++++++++++-
2 files changed, 33 insertions(+), 6 deletions(-)
diff --git a/include/linux/netfilter.h b/include/linux/netfilter.h
index 857f502..b809265 100644
--- a/include/linux/netfilter.h
+++ b/include/linux/netfilter.h
@@ -162,6 +162,24 @@ extern struct ctl_path nf_net_ipv4_netfilter_sysctl_path[];
extern struct list_head nf_hooks[NFPROTO_NUMPROTO][NF_MAX_HOOKS];
+#if defined(CONFIG_JUMP_LABEL)
+#include <linux/jump_label.h>
+extern struct jump_label_key nf_hooks_needed[NFPROTO_NUMPROTO][NF_MAX_HOOKS];
+static inline bool nf_hooks_active(u_int8_t pf, unsigned int hook)
+{
+ if (__builtin_constant_p(pf) &&
+ __builtin_constant_p(hook))
+ return static_branch(&nf_hooks_needed[pf][hook]);
+
+ return !list_empty(&nf_hooks[pf][hook]);
+}
+#else
+static inline bool nf_hooks_active(u_int8_t pf, unsigned int hook)
+{
+ return !list_empty(&nf_hooks[pf][hook]);
+}
+#endif
+
int nf_hook_slow(u_int8_t pf, unsigned int hook, struct sk_buff *skb,
struct net_device *indev, struct net_device *outdev,
int (*okfn)(struct sk_buff *), int thresh);
@@ -179,11 +197,9 @@ static inline int nf_hook_thresh(u_int8_t pf, unsigned int hook,
struct net_device *outdev,
int (*okfn)(struct sk_buff *), int thresh)
{
-#ifndef CONFIG_NETFILTER_DEBUG
- if (list_empty(&nf_hooks[pf][hook]))
- return 1;
-#endif
- return nf_hook_slow(pf, hook, skb, indev, outdev, okfn, thresh);
+ if (nf_hooks_active(pf, hook))
+ return nf_hook_slow(pf, hook, skb, indev, outdev, okfn, thresh);
+ return 1;
}
static inline int nf_hook(u_int8_t pf, unsigned int hook, struct sk_buff *skb,
diff --git a/net/netfilter/core.c b/net/netfilter/core.c
index afca6c7..4aa0f4b 100644
--- a/net/netfilter/core.c
+++ b/net/netfilter/core.c
@@ -54,6 +54,12 @@ EXPORT_SYMBOL_GPL(nf_unregister_afinfo);
struct list_head nf_hooks[NFPROTO_NUMPROTO][NF_MAX_HOOKS] __read_mostly;
EXPORT_SYMBOL(nf_hooks);
+
+#if defined(CONFIG_JUMP_LABEL)
+struct jump_label_key nf_hooks_needed[NFPROTO_NUMPROTO][NF_MAX_HOOKS];
+EXPORT_SYMBOL(nf_hooks_needed);
+#endif
+
static DEFINE_MUTEX(nf_hook_mutex);
int nf_register_hook(struct nf_hook_ops *reg)
@@ -70,6 +76,9 @@ int nf_register_hook(struct nf_hook_ops *reg)
}
list_add_rcu(®->list, elem->list.prev);
mutex_unlock(&nf_hook_mutex);
+#if defined(CONFIG_JUMP_LABEL)
+ jump_label_inc(&nf_hooks_needed[reg->pf][reg->hooknum]);
+#endif
return 0;
}
EXPORT_SYMBOL(nf_register_hook);
@@ -79,7 +88,9 @@ void nf_unregister_hook(struct nf_hook_ops *reg)
mutex_lock(&nf_hook_mutex);
list_del_rcu(®->list);
mutex_unlock(&nf_hook_mutex);
-
+#if defined(CONFIG_JUMP_LABEL)
+ jump_label_dec(&nf_hooks_needed[reg->pf][reg->hooknum]);
+#endif
synchronize_net();
}
EXPORT_SYMBOL(nf_unregister_hook);
next reply other threads:[~2011-11-19 3:32 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-11-19 3:32 Eric Dumazet [this message]
2011-11-21 21:39 ` [PATCH net-next] netfilter: use jump_label for nf_hooks 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=1321673566.3157.14.camel@edumazet-laptop \
--to=eric.dumazet@gmail.com \
--cc=davem@davemloft.net \
--cc=kaber@trash.net \
--cc=netdev@vger.kernel.org \
--cc=netfilter-devel@vger.kernel.org \
--cc=pablo@netfilter.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