[INET]: speedup inet (tcp/dccp) lookups

Arnaldo and I agreed it could be applied now, because I have other
pending patches depending on this one (Thank you Arnaldo)

(The other important patch moves skc_refcnt in a separate cache line,
so that the SMP/NUMA performance doesnt suffer from cache line ping pongs)

1) First some performance data :
--------------------------------

tcp_v4_rcv() wastes a *lot* of time in __inet_lookup_established()

The most time critical code is :

sk_for_each(sk, node, &head->chain) {
     if (INET_MATCH(sk, acookie, saddr, daddr, ports, dif))
         goto hit; /* You sunk my battleship! */
}

The sk_for_each() does use prefetch() hints but only the begining of
"struct sock" is prefetched.

As INET_MATCH first comparison uses inet_sk(__sk)->daddr, wich is far
away from the begining of "struct sock", it has to bring into CPU
cache cold cache line. Each iteration has to use at least 2 cache
lines.

This can be problematic if some chains are very long.

2) The goal
-----------

The idea I had is to change things so that INET_MATCH() may return
FALSE in 99% of cases only using the data already in the CPU cache,
using one cache line per iteration.

3) Description of the patch
---------------------------

Adds a new 'unsigned int skc_hash' field in 'struct sock_common',
filling a 32 bits hole on 64 bits platform.

struct sock_common {
	unsigned short		skc_family;
	volatile unsigned char	skc_state;
	unsigned char		skc_reuse;
	int			skc_bound_dev_if;
	struct hlist_node	skc_node;
	struct hlist_node	skc_bind_node;
	atomic_t		skc_refcnt;
+	unsigned int		skc_hash;
	struct proto		*skc_prot;
};

Store in this 32 bits field the full hash, not masked by (ehash_size -
1) Using this full hash as the first comparison done in INET_MATCH
permits us immediatly skip the element without touching a second cache
line in case of a miss.

Suppress the sk_hashent/tw_hashent fields since skc_hash (aliased to
sk_hash and tw_hash) already contains the slot number if we mask with
(ehash_size - 1)

File include/net/inet_hashtables.h

64 bits platforms :
#define INET_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
     (((__sk)->sk_hash == (__hash))
     ((*((__u64 *)&(inet_sk(__sk)->daddr)))== (__cookie))   &&  \
     ((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports))   &&  \
     (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))

32bits platforms:
#define TCP_IPV4_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
     (((__sk)->sk_hash == (__hash))                 &&  \
     (inet_sk(__sk)->daddr          == (__saddr))   &&  \
     (inet_sk(__sk)->rcv_saddr      == (__daddr))   &&  \
     (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))


- Adds a prefetch(head->chain.first) in 
__inet_lookup_established()/__tcp_v4_check_established() and 
__inet6_lookup_established()/__tcp_v6_check_established() and 
__dccp_v4_check_established() to bring into cache the first element of the 
list, before the {read|write}_lock(&head->lock);

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

6 files changed, 56 insertions, 43 deletions

diff --git a/include/linux/ipv6.h b/include/linux/ipv6.h
index bb6f88e14061..e0b922785d98 100644
--- a/include/linux/ipv6.h
+++ b/include/linux/ipv6.h
@@ -372,8 +372,9 @@ static inline struct raw6_sock *raw6_sk(const struct sock *sk)
 #define inet_v6_ipv6only(__sk)          0
 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */
 
-#define INET6_MATCH(__sk, __saddr, __daddr, __ports, __dif)        \
-        (((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports))   && \
+#define INET6_MATCH(__sk, __hash, __saddr, __daddr, __ports, __dif)\
+        (((__sk)->sk_hash == (__hash))                          && \
+         ((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports))   && \
          ((__sk)->sk_family             == AF_INET6)            && \
          ipv6_addr_equal(&inet6_sk(__sk)->daddr, (__saddr))     && \
          ipv6_addr_equal(&inet6_sk(__sk)->rcv_saddr, (__daddr)) && \
diff --git a/include/linux/tc_ematch/tc_em_meta.h b/include/linux/tc_ematch/tc_em_meta.h
index 081b1ee8516e..e21937cf91d0 100644
--- a/include/linux/tc_ematch/tc_em_meta.h
+++ b/include/linux/tc_ematch/tc_em_meta.h
@@ -71,7 +71,7 @@ enum
         TCF_META_ID_SK_SNDBUF,
         TCF_META_ID_SK_ALLOCS,
         TCF_META_ID_SK_ROUTE_CAPS,
-        TCF_META_ID_SK_HASHENT,
+        TCF_META_ID_SK_HASH,
         TCF_META_ID_SK_LINGERTIME,
         TCF_META_ID_SK_ACK_BACKLOG,
         TCF_META_ID_SK_MAX_ACK_BACKLOG,
diff --git a/include/net/inet6_hashtables.h b/include/net/inet6_hashtables.h
index 03df3b157960..5a2beed5a770 100644
--- a/include/net/inet6_hashtables.h
+++ b/include/net/inet6_hashtables.h
@@ -26,19 +26,18 @@
 struct inet_hashinfo;
 
 /* I have no idea if this is a good hash for v6 or not. -DaveM */
-static inline int inet6_ehashfn(const struct in6_addr *laddr, const u16 lport,
-                                const struct in6_addr *faddr, const u16 fport,
-                                const int ehash_size)
+static inline unsigned int inet6_ehashfn(const struct in6_addr *laddr, const u16 lport,
+                                const struct in6_addr *faddr, const u16 fport)
 {
-        int hashent = (lport ^ fport);
+        unsigned int hashent = (lport ^ fport);
 
         hashent ^= (laddr->s6_addr32[3] ^ faddr->s6_addr32[3]);
         hashent ^= hashent >> 16;
         hashent ^= hashent >> 8;
-        return (hashent & (ehash_size - 1));
+        return hashent;
 }
 
-static inline int inet6_sk_ehashfn(const struct sock *sk, const int ehash_size)
+static inline int inet6_sk_ehashfn(const struct sock *sk)
 {
         const struct inet_sock *inet = inet_sk(sk);
         const struct ipv6_pinfo *np = inet6_sk(sk);
@@ -46,7 +45,7 @@ static inline int inet6_sk_ehashfn(const struct sock *sk, const int ehash_size)
         const struct in6_addr *faddr = &np->daddr;
         const __u16 lport = inet->num;
         const __u16 fport = inet->dport;
-        return inet6_ehashfn(laddr, lport, faddr, fport, ehash_size);
+        return inet6_ehashfn(laddr, lport, faddr, fport);
 }
 
 /*
@@ -69,14 +68,14 @@ static inline struct sock *
         /* Optimize here for direct hit, only listening connections can
          * have wildcards anyways.
          */
-        const int hash = inet6_ehashfn(daddr, hnum, saddr, sport,
-                                       hashinfo->ehash_size);
-        struct inet_ehash_bucket *head = &hashinfo->ehash[hash];
+        unsigned int hash = inet6_ehashfn(daddr, hnum, saddr, sport);
+        struct inet_ehash_bucket *head = inet_ehash_bucket(hashinfo, hash);
 
+        prefetch(head->chain.first);
         read_lock(&head->lock);
         sk_for_each(sk, node, &head->chain) {
                 /* For IPV6 do the cheaper port and family tests first. */
-                if (INET6_MATCH(sk, saddr, daddr, ports, dif))
+                if (INET6_MATCH(sk, hash, saddr, daddr, ports, dif))
                         goto hit; /* You sunk my battleship! */
         }
         /* Must check for a TIME_WAIT'er before going to listener hash. */
diff --git a/include/net/inet_hashtables.h b/include/net/inet_hashtables.h
index 646b6ea7fe26..35f49e65e295 100644
--- a/include/net/inet_hashtables.h
+++ b/include/net/inet_hashtables.h
@@ -108,7 +108,7 @@ struct inet_hashinfo {
         struct inet_bind_hashbucket     *bhash;
 
         int                             bhash_size;
-        int                             ehash_size;
+        unsigned int                    ehash_size;
 
         /* All sockets in TCP_LISTEN state will be in here.  This is the only
          * table where wildcard'd TCP sockets can exist.  Hash function here
@@ -130,17 +130,16 @@ struct inet_hashinfo {
         int                             port_rover;
 };
 
-static inline int inet_ehashfn(const __u32 laddr, const __u16 lport,
-                               const __u32 faddr, const __u16 fport,
-                               const int ehash_size)
+static inline unsigned int inet_ehashfn(const __u32 laddr, const __u16 lport,
+                               const __u32 faddr, const __u16 fport)
 {
-        int h = (laddr ^ lport) ^ (faddr ^ fport);
+        unsigned int h = (laddr ^ lport) ^ (faddr ^ fport);
         h ^= h >> 16;
         h ^= h >> 8;
-        return h & (ehash_size - 1);
+        return h;
 }
 
-static inline int inet_sk_ehashfn(const struct sock *sk, const int ehash_size)
+static inline int inet_sk_ehashfn(const struct sock *sk)
 {
         const struct inet_sock *inet = inet_sk(sk);
         const __u32 laddr = inet->rcv_saddr;
@@ -148,7 +147,14 @@ static inline int inet_sk_ehashfn(const struct sock *sk, const int ehash_size)
         const __u32 faddr = inet->daddr;
         const __u16 fport = inet->dport;
 
-        return inet_ehashfn(laddr, lport, faddr, fport, ehash_size);
+        return inet_ehashfn(laddr, lport, faddr, fport);
+}
+
+static inline struct inet_ehash_bucket *inet_ehash_bucket(
+        struct inet_hashinfo *hashinfo,
+        unsigned int hash)
+{
+        return &hashinfo->ehash[hash & (hashinfo->ehash_size - 1)];
 }
 
 extern struct inet_bind_bucket *
@@ -235,9 +241,11 @@ static inline void __inet_hash(struct inet_hashinfo *hashinfo,
                 lock = &hashinfo->lhash_lock;
                 inet_listen_wlock(hashinfo);
         } else {
-                sk->sk_hashent = inet_sk_ehashfn(sk, hashinfo->ehash_size);
-                list = &hashinfo->ehash[sk->sk_hashent].chain;
-                lock = &hashinfo->ehash[sk->sk_hashent].lock;
+                struct inet_ehash_bucket *head;
+                sk->sk_hash = inet_sk_ehashfn(sk);
+                head = inet_ehash_bucket(hashinfo, sk->sk_hash);
+                list = &head->chain;
+                lock = &head->lock;
                 write_lock(lock);
         }
         __sk_add_node(sk, list);
@@ -268,9 +276,8 @@ static inline void inet_unhash(struct inet_hashinfo *hashinfo, struct sock *sk)
                 inet_listen_wlock(hashinfo);
                 lock = &hashinfo->lhash_lock;
         } else {
-                struct inet_ehash_bucket *head = &hashinfo->ehash[sk->sk_hashent];
-                lock = &head->lock;
-                write_lock_bh(&head->lock);
+                lock = &inet_ehash_bucket(hashinfo, sk->sk_hash)->lock;
+                write_lock_bh(lock);
         }
 
         if (__sk_del_node_init(sk))
@@ -337,23 +344,27 @@ sherry_cache:
 #define INET_ADDR_COOKIE(__name, __saddr, __daddr) \
         const __u64 __name = (((__u64)(__daddr)) << 32) | ((__u64)(__saddr));
 #endif /* __BIG_ENDIAN */
-#define INET_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif)\
-        (((*((__u64 *)&(inet_sk(__sk)->daddr))) == (__cookie))  &&      \
+#define INET_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
+        (((__sk)->sk_hash == (__hash))                          &&      \
+         ((*((__u64 *)&(inet_sk(__sk)->daddr))) == (__cookie))  &&      \
          ((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports))   &&      \
          (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
-#define INET_TW_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif)\
-        (((*((__u64 *)&(inet_twsk(__sk)->tw_daddr))) == (__cookie)) &&  \
+#define INET_TW_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
+        (((__sk)->sk_hash == (__hash))                          &&      \
+         ((*((__u64 *)&(inet_twsk(__sk)->tw_daddr))) == (__cookie)) &&  \
          ((*((__u32 *)&(inet_twsk(__sk)->tw_dport))) == (__ports)) &&   \
          (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
 #else /* 32-bit arch */
 #define INET_ADDR_COOKIE(__name, __saddr, __daddr)
-#define INET_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif)    \
-        ((inet_sk(__sk)->daddr          == (__saddr))           &&      \
+#define INET_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)    \
+        (((__sk)->sk_hash == (__hash))                          &&      \
+         (inet_sk(__sk)->daddr          == (__saddr))           &&      \
          (inet_sk(__sk)->rcv_saddr      == (__daddr))           &&      \
          ((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports))   &&      \
          (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
-#define INET_TW_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif) \
-        ((inet_twsk(__sk)->tw_daddr     == (__saddr))           &&      \
+#define INET_TW_MATCH(__sk, __hash,__cookie, __saddr, __daddr, __ports, __dif)  \
+        (((__sk)->sk_hash == (__hash))                          &&      \
+         (inet_twsk(__sk)->tw_daddr     == (__saddr))           &&      \
          (inet_twsk(__sk)->tw_rcv_saddr == (__daddr))           &&      \
          ((*((__u32 *)&(inet_twsk(__sk)->tw_dport))) == (__ports)) &&   \
          (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
@@ -378,18 +389,19 @@ static inline struct sock *
         /* Optimize here for direct hit, only listening connections can
          * have wildcards anyways.
          */
-        const int hash = inet_ehashfn(daddr, hnum, saddr, sport, hashinfo->ehash_size);
-        struct inet_ehash_bucket *head = &hashinfo->ehash[hash];
+        unsigned int hash = inet_ehashfn(daddr, hnum, saddr, sport);
+        struct inet_ehash_bucket *head = inet_ehash_bucket(hashinfo, hash);
 
+        prefetch(head->chain.first);
         read_lock(&head->lock);
         sk_for_each(sk, node, &head->chain) {
-                if (INET_MATCH(sk, acookie, saddr, daddr, ports, dif))
+                if (INET_MATCH(sk, hash, acookie, saddr, daddr, ports, dif))
                         goto hit; /* You sunk my battleship! */
         }
 
         /* Must check for a TIME_WAIT'er before going to listener hash. */
         sk_for_each(sk, node, &(head + hashinfo->ehash_size)->chain) {
-                if (INET_TW_MATCH(sk, acookie, saddr, daddr, ports, dif))
+                if (INET_TW_MATCH(sk, hash, acookie, saddr, daddr, ports, dif))
                         goto hit;
         }
         sk = NULL;
diff --git a/include/net/inet_timewait_sock.h b/include/net/inet_timewait_sock.h
index 3b070352e869..4ade56ef3a4d 100644
--- a/include/net/inet_timewait_sock.h
+++ b/include/net/inet_timewait_sock.h
@@ -112,6 +112,7 @@ struct inet_timewait_sock {
 #define tw_node                 __tw_common.skc_node
 #define tw_bind_node            __tw_common.skc_bind_node
 #define tw_refcnt               __tw_common.skc_refcnt
+#define tw_hash                 __tw_common.skc_hash
 #define tw_prot                 __tw_common.skc_prot
         volatile unsigned char  tw_substate;
         /* 3 bits hole, try to pack */
@@ -126,7 +127,6 @@ struct inet_timewait_sock {
         /* And these are ours. */
         __u8                    tw_ipv6only:1;
         /* 31 bits hole, try to pack */
-        int                     tw_hashent;
         int                     tw_timeout;
         unsigned long           tw_ttd;
         struct inet_bind_bucket *tw_tb;
diff --git a/include/net/sock.h b/include/net/sock.h
index 8c48fbecb7cf..b6440805c420 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -99,6 +99,7 @@ struct proto;
  *      @skc_node: main hash linkage for various protocol lookup tables
  *      @skc_bind_node: bind hash linkage for various protocol lookup tables
  *      @skc_refcnt: reference count
+ *      @skc_hash: hash value used with various protocol lookup tables
  *      @skc_prot: protocol handlers inside a network family
  *
  *      This is the minimal network layer representation of sockets, the header
@@ -112,6 +113,7 @@ struct sock_common {
         struct hlist_node       skc_node;
         struct hlist_node       skc_bind_node;
         atomic_t                skc_refcnt;
+        unsigned int            skc_hash;
         struct proto            *skc_prot;
 };
 
@@ -139,7 +141,6 @@ struct sock_common {
   *     @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
   *     @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
   *     @sk_lingertime: %SO_LINGER l_linger setting
-  *     @sk_hashent: hash entry in several tables (e.g. inet_hashinfo.ehash)
   *     @sk_backlog: always used with the per-socket spinlock held
   *     @sk_callback_lock: used with the callbacks in the end of this struct
   *     @sk_error_queue: rarely used
@@ -186,6 +187,7 @@ struct sock {
 #define sk_node                 __sk_common.skc_node
 #define sk_bind_node            __sk_common.skc_bind_node
 #define sk_refcnt               __sk_common.skc_refcnt
+#define sk_hash                 __sk_common.skc_hash
 #define sk_prot                 __sk_common.skc_prot
         unsigned char           sk_shutdown : 2,
                                 sk_no_check : 2,
@@ -208,7 +210,6 @@ struct sock {
         unsigned int            sk_allocation;
         int                     sk_sndbuf;
         int                     sk_route_caps;
-        int                     sk_hashent;
         unsigned long           sk_flags;
         unsigned long           sk_lingertime;
         /*