From def8b4faff5ca349beafbbfeb2c51f3602a6ef3a Mon Sep 17 00:00:00 2001 From: Alexey Dobriyan Date: Tue, 28 Oct 2008 13:24:06 -0700 Subject: net: reduce structures when XFRM=n ifdef out * struct sk_buff::sp (pointer) * struct dst_entry::xfrm (pointer) * struct sock::sk_policy (2 pointers) Signed-off-by: Alexey Dobriyan Signed-off-by: David S. Miller --- include/net/sock.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index ada50c04d09..d6b750a2507 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -229,7 +229,9 @@ struct sock { } sk_backlog; wait_queue_head_t *sk_sleep; struct dst_entry *sk_dst_cache; +#ifdef CONFIG_XFRM struct xfrm_policy *sk_policy[2]; +#endif rwlock_t sk_dst_lock; atomic_t sk_rmem_alloc; atomic_t sk_wmem_alloc; -- cgit v1.2.3-70-g09d2 From 645ca708f936b2fbeb79e52d7823e3eb2c0905f8 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Wed, 29 Oct 2008 01:41:45 -0700 Subject: udp: introduce struct udp_table and multiple spinlocks UDP sockets are hashed in a 128 slots hash table. This hash table is protected by *one* rwlock. This rwlock is readlocked each time an incoming UDP message is handled. This rwlock is writelocked each time a socket must be inserted in hash table (bind time), or deleted from this table (close time) This is not scalable on SMP machines : 1) Even in read mode, lock() and unlock() are atomic operations and must dirty a contended cache line, shared by all cpus. 2) A writer might be starved if many readers are 'in flight'. This can happen on a machine with some NIC receiving many UDP messages. User process can be delayed a long time at socket creation/dismantle time. This patch prepares RCU migration, by introducing 'struct udp_table and struct udp_hslot', and using one spinlock per chain, to reduce contention on central rwlock. Introducing one spinlock per chain reduces latencies, for port randomization on heavily loaded UDP servers. This also speedup bindings to specific ports. udp_lib_unhash() was uninlined, becoming to big. Some cleanups were done to ease review of following patch (RCUification of UDP Unicast lookups) Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 2 +- include/net/udp.h | 25 +++--- include/net/udplite.h | 2 +- net/ipv4/udp.c | 207 ++++++++++++++++++++++++++++++-------------------- net/ipv4/udp_impl.h | 4 +- net/ipv4/udplite.c | 13 ++-- net/ipv6/udp.c | 112 +++++++++++++++------------ net/ipv6/udp_impl.h | 4 +- net/ipv6/udplite.c | 8 +- 9 files changed, 214 insertions(+), 163 deletions(-) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index d6b750a2507..d200dfbe1ef 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -599,7 +599,7 @@ struct proto { union { struct inet_hashinfo *hashinfo; - struct hlist_head *udp_hash; + struct udp_table *udp_table; struct raw_hashinfo *raw_hash; } h; diff --git a/include/net/udp.h b/include/net/udp.h index 1e205095ea6..df2bfe54537 100644 --- a/include/net/udp.h +++ b/include/net/udp.h @@ -50,8 +50,15 @@ struct udp_skb_cb { }; #define UDP_SKB_CB(__skb) ((struct udp_skb_cb *)((__skb)->cb)) -extern struct hlist_head udp_hash[UDP_HTABLE_SIZE]; -extern rwlock_t udp_hash_lock; +struct udp_hslot { + struct hlist_head head; + spinlock_t lock; +} __attribute__((aligned(2 * sizeof(long)))); +struct udp_table { + struct udp_hslot hash[UDP_HTABLE_SIZE]; +}; +extern struct udp_table udp_table; +extern void udp_table_init(struct udp_table *); /* Note: this must match 'valbool' in sock_setsockopt */ @@ -110,15 +117,7 @@ static inline void udp_lib_hash(struct sock *sk) BUG(); } -static inline void udp_lib_unhash(struct sock *sk) -{ - write_lock_bh(&udp_hash_lock); - if (sk_del_node_init(sk)) { - inet_sk(sk)->num = 0; - sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); - } - write_unlock_bh(&udp_hash_lock); -} +extern void udp_lib_unhash(struct sock *sk); static inline void udp_lib_close(struct sock *sk, long timeout) { @@ -187,7 +186,7 @@ extern struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, struct udp_seq_afinfo { char *name; sa_family_t family; - struct hlist_head *hashtable; + struct udp_table *udp_table; struct file_operations seq_fops; struct seq_operations seq_ops; }; @@ -196,7 +195,7 @@ struct udp_iter_state { struct seq_net_private p; sa_family_t family; int bucket; - struct hlist_head *hashtable; + struct udp_table *udp_table; }; #ifdef CONFIG_PROC_FS diff --git a/include/net/udplite.h b/include/net/udplite.h index b76b2e377af..afdffe607b2 100644 --- a/include/net/udplite.h +++ b/include/net/udplite.h @@ -11,7 +11,7 @@ #define UDPLITE_RECV_CSCOV 11 /* receiver partial coverage (threshold ) */ extern struct proto udplite_prot; -extern struct hlist_head udplite_hash[UDP_HTABLE_SIZE]; +extern struct udp_table udplite_table; /* * Checksum computation is all in software, hence simpler getfrag. diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c index 2095abc3cab..2a6c491f97d 100644 --- a/net/ipv4/udp.c +++ b/net/ipv4/udp.c @@ -104,12 +104,8 @@ #include #include "udp_impl.h" -/* - * Snmp MIB for the UDP layer - */ - -struct hlist_head udp_hash[UDP_HTABLE_SIZE]; -DEFINE_RWLOCK(udp_hash_lock); +struct udp_table udp_table; +EXPORT_SYMBOL(udp_table); int sysctl_udp_mem[3] __read_mostly; int sysctl_udp_rmem_min __read_mostly; @@ -123,7 +119,7 @@ atomic_t udp_memory_allocated; EXPORT_SYMBOL(udp_memory_allocated); static int udp_lib_lport_inuse(struct net *net, __u16 num, - const struct hlist_head udptable[], + const struct udp_hslot *hslot, struct sock *sk, int (*saddr_comp)(const struct sock *sk1, const struct sock *sk2)) @@ -131,7 +127,7 @@ static int udp_lib_lport_inuse(struct net *net, __u16 num, struct sock *sk2; struct hlist_node *node; - sk_for_each(sk2, node, &udptable[udp_hashfn(net, num)]) + sk_for_each(sk2, node, &hslot->head) if (net_eq(sock_net(sk2), net) && sk2 != sk && sk2->sk_hash == num && @@ -154,12 +150,11 @@ int udp_lib_get_port(struct sock *sk, unsigned short snum, int (*saddr_comp)(const struct sock *sk1, const struct sock *sk2 ) ) { - struct hlist_head *udptable = sk->sk_prot->h.udp_hash; + struct udp_hslot *hslot; + struct udp_table *udptable = sk->sk_prot->h.udp_table; int error = 1; struct net *net = sock_net(sk); - write_lock_bh(&udp_hash_lock); - if (!snum) { int low, high, remaining; unsigned rand; @@ -171,26 +166,34 @@ int udp_lib_get_port(struct sock *sk, unsigned short snum, rand = net_random(); snum = first = rand % remaining + low; rand |= 1; - while (udp_lib_lport_inuse(net, snum, udptable, sk, - saddr_comp)) { + for (;;) { + hslot = &udptable->hash[udp_hashfn(net, snum)]; + spin_lock_bh(&hslot->lock); + if (!udp_lib_lport_inuse(net, snum, hslot, sk, saddr_comp)) + break; + spin_unlock_bh(&hslot->lock); do { snum = snum + rand; } while (snum < low || snum > high); if (snum == first) goto fail; } - } else if (udp_lib_lport_inuse(net, snum, udptable, sk, saddr_comp)) - goto fail; - + } else { + hslot = &udptable->hash[udp_hashfn(net, snum)]; + spin_lock_bh(&hslot->lock); + if (udp_lib_lport_inuse(net, snum, hslot, sk, saddr_comp)) + goto fail_unlock; + } inet_sk(sk)->num = snum; sk->sk_hash = snum; if (sk_unhashed(sk)) { - sk_add_node(sk, &udptable[udp_hashfn(net, snum)]); + sk_add_node(sk, &hslot->head); sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); } error = 0; +fail_unlock: + spin_unlock_bh(&hslot->lock); fail: - write_unlock_bh(&udp_hash_lock); return error; } @@ -208,63 +211,73 @@ int udp_v4_get_port(struct sock *sk, unsigned short snum) return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal); } +static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr, + unsigned short hnum, + __be16 sport, __be32 daddr, __be16 dport, int dif) +{ + int score = -1; + + if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum && + !ipv6_only_sock(sk)) { + struct inet_sock *inet = inet_sk(sk); + + score = (sk->sk_family == PF_INET ? 1 : 0); + if (inet->rcv_saddr) { + if (inet->rcv_saddr != daddr) + return -1; + score += 2; + } + if (inet->daddr) { + if (inet->daddr != saddr) + return -1; + score += 2; + } + if (inet->dport) { + if (inet->dport != sport) + return -1; + score += 2; + } + if (sk->sk_bound_dev_if) { + if (sk->sk_bound_dev_if != dif) + return -1; + score += 2; + } + } + return score; +} + /* UDP is nearly always wildcards out the wazoo, it makes no sense to try * harder than this. -DaveM */ static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, __be32 daddr, __be16 dport, - int dif, struct hlist_head udptable[]) + int dif, struct udp_table *udptable) { struct sock *sk, *result = NULL; struct hlist_node *node; unsigned short hnum = ntohs(dport); - int badness = -1; - - read_lock(&udp_hash_lock); - sk_for_each(sk, node, &udptable[udp_hashfn(net, hnum)]) { - struct inet_sock *inet = inet_sk(sk); - - if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum && - !ipv6_only_sock(sk)) { - int score = (sk->sk_family == PF_INET ? 1 : 0); - if (inet->rcv_saddr) { - if (inet->rcv_saddr != daddr) - continue; - score+=2; - } - if (inet->daddr) { - if (inet->daddr != saddr) - continue; - score+=2; - } - if (inet->dport) { - if (inet->dport != sport) - continue; - score+=2; - } - if (sk->sk_bound_dev_if) { - if (sk->sk_bound_dev_if != dif) - continue; - score+=2; - } - if (score == 9) { - result = sk; - break; - } else if (score > badness) { - result = sk; - badness = score; - } + unsigned int hash = udp_hashfn(net, hnum); + struct udp_hslot *hslot = &udptable->hash[hash]; + int score, badness = -1; + + spin_lock(&hslot->lock); + sk_for_each(sk, node, &hslot->head) { + score = compute_score(sk, net, saddr, hnum, sport, + daddr, dport, dif); + if (score > badness) { + result = sk; + badness = score; } } if (result) sock_hold(result); - read_unlock(&udp_hash_lock); + spin_unlock(&hslot->lock); return result; } static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, __be16 sport, __be16 dport, - struct hlist_head udptable[]) + struct udp_table *udptable) { struct sock *sk; const struct iphdr *iph = ip_hdr(skb); @@ -280,7 +293,7 @@ static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, __be32 daddr, __be16 dport, int dif) { - return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, udp_hash); + return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table); } EXPORT_SYMBOL_GPL(udp4_lib_lookup); @@ -323,7 +336,7 @@ found: * to find the appropriate port. */ -void __udp4_lib_err(struct sk_buff *skb, u32 info, struct hlist_head udptable[]) +void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) { struct inet_sock *inet; struct iphdr *iph = (struct iphdr*)skb->data; @@ -392,7 +405,7 @@ out: void udp_err(struct sk_buff *skb, u32 info) { - __udp4_lib_err(skb, info, udp_hash); + __udp4_lib_err(skb, info, &udp_table); } /* @@ -933,6 +946,21 @@ int udp_disconnect(struct sock *sk, int flags) return 0; } +void udp_lib_unhash(struct sock *sk) +{ + struct udp_table *udptable = sk->sk_prot->h.udp_table; + unsigned int hash = udp_hashfn(sock_net(sk), sk->sk_hash); + struct udp_hslot *hslot = &udptable->hash[hash]; + + spin_lock(&hslot->lock); + if (sk_del_node_init(sk)) { + inet_sk(sk)->num = 0; + sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); + } + spin_unlock(&hslot->lock); +} +EXPORT_SYMBOL(udp_lib_unhash); + static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) { int is_udplite = IS_UDPLITE(sk); @@ -1071,13 +1099,14 @@ drop: static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, struct udphdr *uh, __be32 saddr, __be32 daddr, - struct hlist_head udptable[]) + struct udp_table *udptable) { struct sock *sk; + struct udp_hslot *hslot = &udptable->hash[udp_hashfn(net, ntohs(uh->dest))]; int dif; - read_lock(&udp_hash_lock); - sk = sk_head(&udptable[udp_hashfn(net, ntohs(uh->dest))]); + spin_lock(&hslot->lock); + sk = sk_head(&hslot->head); dif = skb->dev->ifindex; sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif); if (sk) { @@ -1102,7 +1131,7 @@ static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, } while (sknext); } else kfree_skb(skb); - read_unlock(&udp_hash_lock); + spin_unlock(&hslot->lock); return 0; } @@ -1148,7 +1177,7 @@ static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, * All we need to do is get the socket, and then do a checksum. */ -int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[], +int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, int proto) { struct sock *sk; @@ -1246,7 +1275,7 @@ drop: int udp_rcv(struct sk_buff *skb) { - return __udp4_lib_rcv(skb, udp_hash, IPPROTO_UDP); + return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); } void udp_destroy_sock(struct sock *sk) @@ -1488,7 +1517,7 @@ struct proto udp_prot = { .sysctl_wmem = &sysctl_udp_wmem_min, .sysctl_rmem = &sysctl_udp_rmem_min, .obj_size = sizeof(struct udp_sock), - .h.udp_hash = udp_hash, + .h.udp_table = &udp_table, #ifdef CONFIG_COMPAT .compat_setsockopt = compat_udp_setsockopt, .compat_getsockopt = compat_udp_getsockopt, @@ -1498,20 +1527,23 @@ struct proto udp_prot = { /* ------------------------------------------------------------------------ */ #ifdef CONFIG_PROC_FS -static struct sock *udp_get_first(struct seq_file *seq) +static struct sock *udp_get_first(struct seq_file *seq, int start) { struct sock *sk; struct udp_iter_state *state = seq->private; struct net *net = seq_file_net(seq); - for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) { + for (state->bucket = start; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) { struct hlist_node *node; - sk_for_each(sk, node, state->hashtable + state->bucket) { + struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; + spin_lock_bh(&hslot->lock); + sk_for_each(sk, node, &hslot->head) { if (!net_eq(sock_net(sk), net)) continue; if (sk->sk_family == state->family) goto found; } + spin_unlock_bh(&hslot->lock); } sk = NULL; found: @@ -1525,20 +1557,18 @@ static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) do { sk = sk_next(sk); -try_again: - ; } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); - if (!sk && ++state->bucket < UDP_HTABLE_SIZE) { - sk = sk_head(state->hashtable + state->bucket); - goto try_again; + if (!sk) { + spin_unlock(&state->udp_table->hash[state->bucket].lock); + return udp_get_first(seq, state->bucket + 1); } return sk; } static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) { - struct sock *sk = udp_get_first(seq); + struct sock *sk = udp_get_first(seq, 0); if (sk) while (pos && (sk = udp_get_next(seq, sk)) != NULL) @@ -1547,9 +1577,7 @@ static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) } static void *udp_seq_start(struct seq_file *seq, loff_t *pos) - __acquires(udp_hash_lock) { - read_lock(&udp_hash_lock); return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; } @@ -1567,9 +1595,11 @@ static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) } static void udp_seq_stop(struct seq_file *seq, void *v) - __releases(udp_hash_lock) { - read_unlock(&udp_hash_lock); + struct udp_iter_state *state = seq->private; + + if (state->bucket < UDP_HTABLE_SIZE) + spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); } static int udp_seq_open(struct inode *inode, struct file *file) @@ -1585,7 +1615,7 @@ static int udp_seq_open(struct inode *inode, struct file *file) s = ((struct seq_file *)file->private_data)->private; s->family = afinfo->family; - s->hashtable = afinfo->hashtable; + s->udp_table = afinfo->udp_table; return err; } @@ -1657,7 +1687,7 @@ int udp4_seq_show(struct seq_file *seq, void *v) static struct udp_seq_afinfo udp4_seq_afinfo = { .name = "udp", .family = AF_INET, - .hashtable = udp_hash, + .udp_table = &udp_table, .seq_fops = { .owner = THIS_MODULE, }, @@ -1692,10 +1722,21 @@ void udp4_proc_exit(void) } #endif /* CONFIG_PROC_FS */ +void __init udp_table_init(struct udp_table *table) +{ + int i; + + for (i = 0; i < UDP_HTABLE_SIZE; i++) { + INIT_HLIST_HEAD(&table->hash[i].head); + spin_lock_init(&table->hash[i].lock); + } +} + void __init udp_init(void) { unsigned long limit; + udp_table_init(&udp_table); /* Set the pressure threshold up by the same strategy of TCP. It is a * fraction of global memory that is up to 1/2 at 256 MB, decreasing * toward zero with the amount of memory, with a floor of 128 pages. @@ -1712,8 +1753,6 @@ void __init udp_init(void) } EXPORT_SYMBOL(udp_disconnect); -EXPORT_SYMBOL(udp_hash); -EXPORT_SYMBOL(udp_hash_lock); EXPORT_SYMBOL(udp_ioctl); EXPORT_SYMBOL(udp_prot); EXPORT_SYMBOL(udp_sendmsg); diff --git a/net/ipv4/udp_impl.h b/net/ipv4/udp_impl.h index 2e9bad2fa1b..9f4a6165f72 100644 --- a/net/ipv4/udp_impl.h +++ b/net/ipv4/udp_impl.h @@ -5,8 +5,8 @@ #include #include -extern int __udp4_lib_rcv(struct sk_buff *, struct hlist_head [], int ); -extern void __udp4_lib_err(struct sk_buff *, u32, struct hlist_head []); +extern int __udp4_lib_rcv(struct sk_buff *, struct udp_table *, int ); +extern void __udp4_lib_err(struct sk_buff *, u32, struct udp_table *); extern int udp_v4_get_port(struct sock *sk, unsigned short snum); diff --git a/net/ipv4/udplite.c b/net/ipv4/udplite.c index 3c807964da9..d8ea8e5f5ea 100644 --- a/net/ipv4/udplite.c +++ b/net/ipv4/udplite.c @@ -12,16 +12,17 @@ */ #include "udp_impl.h" -struct hlist_head udplite_hash[UDP_HTABLE_SIZE]; +struct udp_table udplite_table; +EXPORT_SYMBOL(udplite_table); static int udplite_rcv(struct sk_buff *skb) { - return __udp4_lib_rcv(skb, udplite_hash, IPPROTO_UDPLITE); + return __udp4_lib_rcv(skb, &udplite_table, IPPROTO_UDPLITE); } static void udplite_err(struct sk_buff *skb, u32 info) { - __udp4_lib_err(skb, info, udplite_hash); + __udp4_lib_err(skb, info, &udplite_table); } static struct net_protocol udplite_protocol = { @@ -50,7 +51,7 @@ struct proto udplite_prot = { .unhash = udp_lib_unhash, .get_port = udp_v4_get_port, .obj_size = sizeof(struct udp_sock), - .h.udp_hash = udplite_hash, + .h.udp_table = &udplite_table, #ifdef CONFIG_COMPAT .compat_setsockopt = compat_udp_setsockopt, .compat_getsockopt = compat_udp_getsockopt, @@ -71,7 +72,7 @@ static struct inet_protosw udplite4_protosw = { static struct udp_seq_afinfo udplite4_seq_afinfo = { .name = "udplite", .family = AF_INET, - .hashtable = udplite_hash, + .udp_table = &udplite_table, .seq_fops = { .owner = THIS_MODULE, }, @@ -108,6 +109,7 @@ static inline int udplite4_proc_init(void) void __init udplite4_register(void) { + udp_table_init(&udplite_table); if (proto_register(&udplite_prot, 1)) goto out_register_err; @@ -126,5 +128,4 @@ out_register_err: printk(KERN_CRIT "%s: Cannot add UDP-Lite protocol.\n", __func__); } -EXPORT_SYMBOL(udplite_hash); EXPORT_SYMBOL(udplite_prot); diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c index e51da8c092f..ccee7244ca0 100644 --- a/net/ipv6/udp.c +++ b/net/ipv6/udp.c @@ -54,62 +54,73 @@ int udp_v6_get_port(struct sock *sk, unsigned short snum) return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal); } +static inline int compute_score(struct sock *sk, struct net *net, + unsigned short hnum, + struct in6_addr *saddr, __be16 sport, + struct in6_addr *daddr, __be16 dport, + int dif) +{ + int score = -1; + + if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum && + sk->sk_family == PF_INET6) { + struct ipv6_pinfo *np = inet6_sk(sk); + struct inet_sock *inet = inet_sk(sk); + + score = 0; + if (inet->dport) { + if (inet->dport != sport) + return -1; + score++; + } + if (!ipv6_addr_any(&np->rcv_saddr)) { + if (!ipv6_addr_equal(&np->rcv_saddr, daddr)) + return -1; + score++; + } + if (!ipv6_addr_any(&np->daddr)) { + if (!ipv6_addr_equal(&np->daddr, saddr)) + return -1; + score++; + } + if (sk->sk_bound_dev_if) { + if (sk->sk_bound_dev_if != dif) + return -1; + score++; + } + } + return score; +} + static struct sock *__udp6_lib_lookup(struct net *net, struct in6_addr *saddr, __be16 sport, struct in6_addr *daddr, __be16 dport, - int dif, struct hlist_head udptable[]) + int dif, struct udp_table *udptable) { struct sock *sk, *result = NULL; struct hlist_node *node; unsigned short hnum = ntohs(dport); - int badness = -1; - - read_lock(&udp_hash_lock); - sk_for_each(sk, node, &udptable[udp_hashfn(net, hnum)]) { - struct inet_sock *inet = inet_sk(sk); - - if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum && - sk->sk_family == PF_INET6) { - struct ipv6_pinfo *np = inet6_sk(sk); - int score = 0; - if (inet->dport) { - if (inet->dport != sport) - continue; - score++; - } - if (!ipv6_addr_any(&np->rcv_saddr)) { - if (!ipv6_addr_equal(&np->rcv_saddr, daddr)) - continue; - score++; - } - if (!ipv6_addr_any(&np->daddr)) { - if (!ipv6_addr_equal(&np->daddr, saddr)) - continue; - score++; - } - if (sk->sk_bound_dev_if) { - if (sk->sk_bound_dev_if != dif) - continue; - score++; - } - if (score == 4) { - result = sk; - break; - } else if (score > badness) { - result = sk; - badness = score; - } + unsigned int hash = udp_hashfn(net, hnum); + struct udp_hslot *hslot = &udptable->hash[hash]; + int score, badness = -1; + + spin_lock(&hslot->lock); + sk_for_each(sk, node, &hslot->head) { + score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif); + if (score > badness) { + result = sk; + badness = score; } } if (result) sock_hold(result); - read_unlock(&udp_hash_lock); + spin_unlock(&hslot->lock); return result; } static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb, __be16 sport, __be16 dport, - struct hlist_head udptable[]) + struct udp_table *udptable) { struct sock *sk; struct ipv6hdr *iph = ipv6_hdr(skb); @@ -239,7 +250,7 @@ csum_copy_err: void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt, int type, int code, int offset, __be32 info, - struct hlist_head udptable[] ) + struct udp_table *udptable) { struct ipv6_pinfo *np; struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data; @@ -275,7 +286,7 @@ static __inline__ void udpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, int type, int code, int offset, __be32 info ) { - __udp6_lib_err(skb, opt, type, code, offset, info, udp_hash); + __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table); } int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb) @@ -374,14 +385,15 @@ static struct sock *udp_v6_mcast_next(struct sock *sk, */ static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, struct in6_addr *saddr, struct in6_addr *daddr, - struct hlist_head udptable[]) + struct udp_table *udptable) { struct sock *sk, *sk2; const struct udphdr *uh = udp_hdr(skb); + struct udp_hslot *hslot = &udptable->hash[udp_hashfn(net, ntohs(uh->dest))]; int dif; - read_lock(&udp_hash_lock); - sk = sk_head(&udptable[udp_hashfn(net, ntohs(uh->dest))]); + spin_lock(&hslot->lock); + sk = sk_head(&hslot->head); dif = inet6_iif(skb); sk = udp_v6_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif); if (!sk) { @@ -409,7 +421,7 @@ static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, sk_add_backlog(sk, skb); bh_unlock_sock(sk); out: - read_unlock(&udp_hash_lock); + spin_unlock(&hslot->lock); return 0; } @@ -447,7 +459,7 @@ static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh, return 0; } -int __udp6_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[], +int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, int proto) { struct sock *sk; @@ -544,7 +556,7 @@ discard: static __inline__ int udpv6_rcv(struct sk_buff *skb) { - return __udp6_lib_rcv(skb, udp_hash, IPPROTO_UDP); + return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP); } /* @@ -1008,7 +1020,7 @@ int udp6_seq_show(struct seq_file *seq, void *v) static struct udp_seq_afinfo udp6_seq_afinfo = { .name = "udp6", .family = AF_INET6, - .hashtable = udp_hash, + .udp_table = &udp_table, .seq_fops = { .owner = THIS_MODULE, }, @@ -1050,7 +1062,7 @@ struct proto udpv6_prot = { .sysctl_wmem = &sysctl_udp_wmem_min, .sysctl_rmem = &sysctl_udp_rmem_min, .obj_size = sizeof(struct udp6_sock), - .h.udp_hash = udp_hash, + .h.udp_table = &udp_table, #ifdef CONFIG_COMPAT .compat_setsockopt = compat_udpv6_setsockopt, .compat_getsockopt = compat_udpv6_getsockopt, diff --git a/net/ipv6/udp_impl.h b/net/ipv6/udp_impl.h index 92dd7da766d..23779208c33 100644 --- a/net/ipv6/udp_impl.h +++ b/net/ipv6/udp_impl.h @@ -7,9 +7,9 @@ #include #include -extern int __udp6_lib_rcv(struct sk_buff *, struct hlist_head [], int ); +extern int __udp6_lib_rcv(struct sk_buff *, struct udp_table *, int ); extern void __udp6_lib_err(struct sk_buff *, struct inet6_skb_parm *, - int , int , int , __be32 , struct hlist_head []); + int , int , int , __be32 , struct udp_table *); extern int udp_v6_get_port(struct sock *sk, unsigned short snum); diff --git a/net/ipv6/udplite.c b/net/ipv6/udplite.c index 3cd1a1ac3d6..f1e892a99e0 100644 --- a/net/ipv6/udplite.c +++ b/net/ipv6/udplite.c @@ -15,14 +15,14 @@ static int udplitev6_rcv(struct sk_buff *skb) { - return __udp6_lib_rcv(skb, udplite_hash, IPPROTO_UDPLITE); + return __udp6_lib_rcv(skb, &udplite_table, IPPROTO_UDPLITE); } static void udplitev6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, int type, int code, int offset, __be32 info) { - __udp6_lib_err(skb, opt, type, code, offset, info, udplite_hash); + __udp6_lib_err(skb, opt, type, code, offset, info, &udplite_table); } static struct inet6_protocol udplitev6_protocol = { @@ -49,7 +49,7 @@ struct proto udplitev6_prot = { .unhash = udp_lib_unhash, .get_port = udp_v6_get_port, .obj_size = sizeof(struct udp6_sock), - .h.udp_hash = udplite_hash, + .h.udp_table = &udplite_table, #ifdef CONFIG_COMPAT .compat_setsockopt = compat_udpv6_setsockopt, .compat_getsockopt = compat_udpv6_getsockopt, @@ -95,7 +95,7 @@ void udplitev6_exit(void) static struct udp_seq_afinfo udplite6_seq_afinfo = { .name = "udplite6", .family = AF_INET6, - .hashtable = udplite_hash, + .udp_table = &udplite_table, .seq_fops = { .owner = THIS_MODULE, }, -- cgit v1.2.3-70-g09d2 From 271b72c7fa82c2c7a795bc16896149933110672d Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Wed, 29 Oct 2008 02:11:14 -0700 Subject: udp: RCU handling for Unicast packets. Goals are : 1) Optimizing handling of incoming Unicast UDP frames, so that no memory writes should happen in the fast path. Note: Multicasts and broadcasts still will need to take a lock, because doing a full lockless lookup in this case is difficult. 2) No expensive operations in the socket bind/unhash phases : - No expensive synchronize_rcu() calls. - No added rcu_head in socket structure, increasing memory needs, but more important, forcing us to use call_rcu() calls, that have the bad property of making sockets structure cold. (rcu grace period between socket freeing and its potential reuse make this socket being cold in CPU cache). David did a previous patch using call_rcu() and noticed a 20% impact on TCP connection rates. Quoting Cristopher Lameter : "Right. That results in cacheline cooldown. You'd want to recycle the object as they are cache hot on a per cpu basis. That is screwed up by the delayed regular rcu processing. We have seen multiple regressions due to cacheline cooldown. The only choice in cacheline hot sensitive areas is to deal with the complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU." - Because udp sockets are allocated from dedicated kmem_cache, use of SLAB_DESTROY_BY_RCU can help here. Theory of operation : --------------------- As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()), special attention must be taken by readers and writers. Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed, reused, inserted in a different chain or in worst case in the same chain while readers could do lookups in the same time. In order to avoid loops, a reader must check each socket found in a chain really belongs to the chain the reader was traversing. If it finds a mismatch, lookup must start again at the begining. This *restart* loop is the reason we had to use rdlock for the multicast case, because we dont want to send same message several times to the same socket. We use RCU only for fast path. Thus, /proc/net/udp still takes spinlocks. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 37 ++++++++++++++++++++++++++++++++++++- net/core/sock.c | 3 ++- net/ipv4/udp.c | 35 ++++++++++++++++++++++++++--------- net/ipv4/udplite.c | 1 + net/ipv6/udp.c | 31 ++++++++++++++++++++++++------- net/ipv6/udplite.c | 1 + 6 files changed, 90 insertions(+), 18 deletions(-) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index d200dfbe1ef..0bea25db547 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -363,6 +363,27 @@ static __inline__ int sk_del_node_init(struct sock *sk) return rc; } +static __inline__ int __sk_del_node_init_rcu(struct sock *sk) +{ + if (sk_hashed(sk)) { + hlist_del_init_rcu(&sk->sk_node); + return 1; + } + return 0; +} + +static __inline__ int sk_del_node_init_rcu(struct sock *sk) +{ + int rc = __sk_del_node_init_rcu(sk); + + if (rc) { + /* paranoid for a while -acme */ + WARN_ON(atomic_read(&sk->sk_refcnt) == 1); + __sock_put(sk); + } + return rc; +} + static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) { hlist_add_head(&sk->sk_node, list); @@ -374,6 +395,17 @@ static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) __sk_add_node(sk, list); } +static __inline__ void __sk_add_node_rcu(struct sock *sk, struct hlist_head *list) +{ + hlist_add_head_rcu(&sk->sk_node, list); +} + +static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) +{ + sock_hold(sk); + __sk_add_node_rcu(sk, list); +} + static __inline__ void __sk_del_bind_node(struct sock *sk) { __hlist_del(&sk->sk_bind_node); @@ -387,6 +419,8 @@ static __inline__ void sk_add_bind_node(struct sock *sk, #define sk_for_each(__sk, node, list) \ hlist_for_each_entry(__sk, node, list, sk_node) +#define sk_for_each_rcu(__sk, node, list) \ + hlist_for_each_entry_rcu(__sk, node, list, sk_node) #define sk_for_each_from(__sk, node) \ if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ hlist_for_each_entry_from(__sk, node, sk_node) @@ -589,8 +623,9 @@ struct proto { int *sysctl_rmem; int max_header; - struct kmem_cache *slab; + struct kmem_cache *slab; unsigned int obj_size; + int slab_flags; atomic_t *orphan_count; diff --git a/net/core/sock.c b/net/core/sock.c index 5e2a3132a8c..ded1eb5d2fd 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -2042,7 +2042,8 @@ int proto_register(struct proto *prot, int alloc_slab) if (alloc_slab) { prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0, - SLAB_HWCACHE_ALIGN, NULL); + SLAB_HWCACHE_ALIGN | prot->slab_flags, + NULL); if (prot->slab == NULL) { printk(KERN_CRIT "%s: Can't create sock SLAB cache!\n", diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c index 2a6c491f97d..0ea974bf796 100644 --- a/net/ipv4/udp.c +++ b/net/ipv4/udp.c @@ -187,7 +187,7 @@ int udp_lib_get_port(struct sock *sk, unsigned short snum, inet_sk(sk)->num = snum; sk->sk_hash = snum; if (sk_unhashed(sk)) { - sk_add_node(sk, &hslot->head); + sk_add_node_rcu(sk, &hslot->head); sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); } error = 0; @@ -253,15 +253,24 @@ static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, __be32 daddr, __be16 dport, int dif, struct udp_table *udptable) { - struct sock *sk, *result = NULL; + struct sock *sk, *result; struct hlist_node *node; unsigned short hnum = ntohs(dport); unsigned int hash = udp_hashfn(net, hnum); struct udp_hslot *hslot = &udptable->hash[hash]; - int score, badness = -1; + int score, badness; - spin_lock(&hslot->lock); - sk_for_each(sk, node, &hslot->head) { + rcu_read_lock(); +begin: + result = NULL; + badness = -1; + sk_for_each_rcu(sk, node, &hslot->head) { + /* + * lockless reader, and SLAB_DESTROY_BY_RCU items: + * We must check this item was not moved to another chain + */ + if (udp_hashfn(net, sk->sk_hash) != hash) + goto begin; score = compute_score(sk, net, saddr, hnum, sport, daddr, dport, dif); if (score > badness) { @@ -269,9 +278,16 @@ static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, badness = score; } } - if (result) - sock_hold(result); - spin_unlock(&hslot->lock); + if (result) { + if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt))) + result = NULL; + else if (unlikely(compute_score(result, net, saddr, hnum, sport, + daddr, dport, dif) < badness)) { + sock_put(result); + goto begin; + } + } + rcu_read_unlock(); return result; } @@ -953,7 +969,7 @@ void udp_lib_unhash(struct sock *sk) struct udp_hslot *hslot = &udptable->hash[hash]; spin_lock(&hslot->lock); - if (sk_del_node_init(sk)) { + if (sk_del_node_init_rcu(sk)) { inet_sk(sk)->num = 0; sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); } @@ -1517,6 +1533,7 @@ struct proto udp_prot = { .sysctl_wmem = &sysctl_udp_wmem_min, .sysctl_rmem = &sysctl_udp_rmem_min, .obj_size = sizeof(struct udp_sock), + .slab_flags = SLAB_DESTROY_BY_RCU, .h.udp_table = &udp_table, #ifdef CONFIG_COMPAT .compat_setsockopt = compat_udp_setsockopt, diff --git a/net/ipv4/udplite.c b/net/ipv4/udplite.c index d8ea8e5f5ea..c784891cb7e 100644 --- a/net/ipv4/udplite.c +++ b/net/ipv4/udplite.c @@ -51,6 +51,7 @@ struct proto udplite_prot = { .unhash = udp_lib_unhash, .get_port = udp_v4_get_port, .obj_size = sizeof(struct udp_sock), + .slab_flags = SLAB_DESTROY_BY_RCU, .h.udp_table = &udplite_table, #ifdef CONFIG_COMPAT .compat_setsockopt = compat_udp_setsockopt, diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c index ccee7244ca0..1d9790e43df 100644 --- a/net/ipv6/udp.c +++ b/net/ipv6/udp.c @@ -97,24 +97,40 @@ static struct sock *__udp6_lib_lookup(struct net *net, struct in6_addr *daddr, __be16 dport, int dif, struct udp_table *udptable) { - struct sock *sk, *result = NULL; + struct sock *sk, *result; struct hlist_node *node; unsigned short hnum = ntohs(dport); unsigned int hash = udp_hashfn(net, hnum); struct udp_hslot *hslot = &udptable->hash[hash]; - int score, badness = -1; + int score, badness; - spin_lock(&hslot->lock); - sk_for_each(sk, node, &hslot->head) { + rcu_read_lock(); +begin: + result = NULL; + badness = -1; + sk_for_each_rcu(sk, node, &hslot->head) { + /* + * lockless reader, and SLAB_DESTROY_BY_RCU items: + * We must check this item was not moved to another chain + */ + if (udp_hashfn(net, sk->sk_hash) != hash) + goto begin; score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif); if (score > badness) { result = sk; badness = score; } } - if (result) - sock_hold(result); - spin_unlock(&hslot->lock); + if (result) { + if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt))) + result = NULL; + else if (unlikely(compute_score(result, net, hnum, saddr, sport, + daddr, dport, dif) < badness)) { + sock_put(result); + goto begin; + } + } + rcu_read_unlock(); return result; } @@ -1062,6 +1078,7 @@ struct proto udpv6_prot = { .sysctl_wmem = &sysctl_udp_wmem_min, .sysctl_rmem = &sysctl_udp_rmem_min, .obj_size = sizeof(struct udp6_sock), + .slab_flags = SLAB_DESTROY_BY_RCU, .h.udp_table = &udp_table, #ifdef CONFIG_COMPAT .compat_setsockopt = compat_udpv6_setsockopt, diff --git a/net/ipv6/udplite.c b/net/ipv6/udplite.c index f1e892a99e0..ba162a82458 100644 --- a/net/ipv6/udplite.c +++ b/net/ipv6/udplite.c @@ -49,6 +49,7 @@ struct proto udplitev6_prot = { .unhash = udp_lib_unhash, .get_port = udp_v6_get_port, .obj_size = sizeof(struct udp6_sock), + .slab_flags = SLAB_DESTROY_BY_RCU, .h.udp_table = &udplite_table, #ifdef CONFIG_COMPAT .compat_setsockopt = compat_udpv6_setsockopt, -- cgit v1.2.3-70-g09d2 From 96631ed16c514cf8b28fab991a076985ce378c26 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Wed, 29 Oct 2008 11:19:58 -0700 Subject: udp: introduce sk_for_each_rcu_safenext() Corey Minyard found a race added in commit 271b72c7fa82c2c7a795bc16896149933110672d (udp: RCU handling for Unicast packets.) "If the socket is moved from one list to another list in-between the time the hash is calculated and the next field is accessed, and the socket has moved to the end of the new list, the traversal will not complete properly on the list it should have, since the socket will be on the end of the new list and there's not a way to tell it's on a new list and restart the list traversal. I think that this can be solved by pre-fetching the "next" field (with proper barriers) before checking the hash." This patch corrects this problem, introducing a new sk_for_each_rcu_safenext() macro. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/linux/rculist.h | 17 +++++++++++++++++ include/net/sock.h | 4 ++-- net/ipv4/udp.c | 4 ++-- net/ipv6/udp.c | 4 ++-- 4 files changed, 23 insertions(+), 6 deletions(-) (limited to 'include/net/sock.h') diff --git a/include/linux/rculist.h b/include/linux/rculist.h index e649bd3f2c9..3ba2998b22b 100644 --- a/include/linux/rculist.h +++ b/include/linux/rculist.h @@ -383,5 +383,22 @@ static inline void hlist_add_after_rcu(struct hlist_node *prev, ({ tpos = hlist_entry(pos, typeof(*tpos), member); 1; }); \ pos = rcu_dereference(pos->next)) +/** + * hlist_for_each_entry_rcu_safenext - iterate over rcu list of given type + * @tpos: the type * to use as a loop cursor. + * @pos: the &struct hlist_node to use as a loop cursor. + * @head: the head for your list. + * @member: the name of the hlist_node within the struct. + * @next: the &struct hlist_node to use as a next cursor + * + * Special version of hlist_for_each_entry_rcu that make sure + * each next pointer is fetched before each iteration. + */ +#define hlist_for_each_entry_rcu_safenext(tpos, pos, head, member, next) \ + for (pos = rcu_dereference((head)->first); \ + pos && ({ next = pos->next; smp_rmb(); prefetch(next); 1; }) && \ + ({ tpos = hlist_entry(pos, typeof(*tpos), member); 1; }); \ + pos = rcu_dereference(next)) + #endif /* __KERNEL__ */ #endif diff --git a/include/net/sock.h b/include/net/sock.h index 0bea25db547..a4f6d3fc047 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -419,8 +419,8 @@ static __inline__ void sk_add_bind_node(struct sock *sk, #define sk_for_each(__sk, node, list) \ hlist_for_each_entry(__sk, node, list, sk_node) -#define sk_for_each_rcu(__sk, node, list) \ - hlist_for_each_entry_rcu(__sk, node, list, sk_node) +#define sk_for_each_rcu_safenext(__sk, node, list, next) \ + hlist_for_each_entry_rcu_safenext(__sk, node, list, sk_node, next) #define sk_for_each_from(__sk, node) \ if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ hlist_for_each_entry_from(__sk, node, sk_node) diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c index ced820318f9..c3ecec8a9e1 100644 --- a/net/ipv4/udp.c +++ b/net/ipv4/udp.c @@ -256,7 +256,7 @@ static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, int dif, struct udp_table *udptable) { struct sock *sk, *result; - struct hlist_node *node; + struct hlist_node *node, *next; unsigned short hnum = ntohs(dport); unsigned int hash = udp_hashfn(net, hnum); struct udp_hslot *hslot = &udptable->hash[hash]; @@ -266,7 +266,7 @@ static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, begin: result = NULL; badness = -1; - sk_for_each_rcu(sk, node, &hslot->head) { + sk_for_each_rcu_safenext(sk, node, &hslot->head, next) { /* * lockless reader, and SLAB_DESTROY_BY_RCU items: * We must check this item was not moved to another chain diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c index 1d9790e43df..32d914db6c4 100644 --- a/net/ipv6/udp.c +++ b/net/ipv6/udp.c @@ -98,7 +98,7 @@ static struct sock *__udp6_lib_lookup(struct net *net, int dif, struct udp_table *udptable) { struct sock *sk, *result; - struct hlist_node *node; + struct hlist_node *node, *next; unsigned short hnum = ntohs(dport); unsigned int hash = udp_hashfn(net, hnum); struct udp_hslot *hslot = &udptable->hash[hash]; @@ -108,7 +108,7 @@ static struct sock *__udp6_lib_lookup(struct net *net, begin: result = NULL; badness = -1; - sk_for_each_rcu(sk, node, &hslot->head) { + sk_for_each_rcu_safenext(sk, node, &hslot->head, next) { /* * lockless reader, and SLAB_DESTROY_BY_RCU items: * We must check this item was not moved to another chain -- cgit v1.2.3-70-g09d2 From d5f642384e9da75393160350f75bbb9a527f7c58 Mon Sep 17 00:00:00 2001 From: Alexey Dobriyan Date: Tue, 4 Nov 2008 14:45:58 -0800 Subject: net: #ifdef ->sk_security Signed-off-by: Alexey Dobriyan Signed-off-by: David S. Miller --- include/net/sock.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index 941ad7c830a..08291c1be41 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -271,7 +271,9 @@ struct sock { struct sk_buff *sk_send_head; __u32 sk_sndmsg_off; int sk_write_pending; +#ifdef CONFIG_SECURITY void *sk_security; +#endif __u32 sk_mark; /* XXX 4 bytes hole on 64 bit */ void (*sk_state_change)(struct sock *sk); -- cgit v1.2.3-70-g09d2 From 2378982487c492541d17adc0a870e7e83b07ba43 Mon Sep 17 00:00:00 2001 From: Alexey Dobriyan Date: Wed, 12 Nov 2008 23:25:32 -0800 Subject: net: ifdef struct sock::sk_async_wait_queue Every user is under CONFIG_NET_DMA already, so ifdef field as well. Signed-off-by: Alexey Dobriyan Signed-off-by: David S. Miller --- include/net/sock.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index 08291c1be41..8b2b82131b6 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -239,7 +239,9 @@ struct sock { int sk_sndbuf; struct sk_buff_head sk_receive_queue; struct sk_buff_head sk_write_queue; +#ifdef CONFIG_NET_DMA struct sk_buff_head sk_async_wait_queue; +#endif int sk_wmem_queued; int sk_forward_alloc; gfp_t sk_allocation; -- cgit v1.2.3-70-g09d2 From 88ab1932eac721c6e7336708558fa5ed02c85c80 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Sun, 16 Nov 2008 19:39:21 -0800 Subject: udp: Use hlist_nulls in UDP RCU code This is a straightforward patch, using hlist_nulls infrastructure. RCUification already done on UDP two weeks ago. Using hlist_nulls permits us to avoid some memory barriers, both at lookup time and delete time. Patch is large because it adds new macros to include/net/sock.h. These macros will be used by TCP & DCCP in next patch. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/linux/rculist.h | 17 --------------- include/net/sock.h | 57 +++++++++++++++++++++++++++++++++++++++---------- include/net/udp.h | 2 +- net/ipv4/udp.c | 47 +++++++++++++++++++--------------------- net/ipv6/udp.c | 26 +++++++++++----------- 5 files changed, 83 insertions(+), 66 deletions(-) (limited to 'include/net/sock.h') diff --git a/include/linux/rculist.h b/include/linux/rculist.h index 3ba2998b22b..e649bd3f2c9 100644 --- a/include/linux/rculist.h +++ b/include/linux/rculist.h @@ -383,22 +383,5 @@ static inline void hlist_add_after_rcu(struct hlist_node *prev, ({ tpos = hlist_entry(pos, typeof(*tpos), member); 1; }); \ pos = rcu_dereference(pos->next)) -/** - * hlist_for_each_entry_rcu_safenext - iterate over rcu list of given type - * @tpos: the type * to use as a loop cursor. - * @pos: the &struct hlist_node to use as a loop cursor. - * @head: the head for your list. - * @member: the name of the hlist_node within the struct. - * @next: the &struct hlist_node to use as a next cursor - * - * Special version of hlist_for_each_entry_rcu that make sure - * each next pointer is fetched before each iteration. - */ -#define hlist_for_each_entry_rcu_safenext(tpos, pos, head, member, next) \ - for (pos = rcu_dereference((head)->first); \ - pos && ({ next = pos->next; smp_rmb(); prefetch(next); 1; }) && \ - ({ tpos = hlist_entry(pos, typeof(*tpos), member); 1; }); \ - pos = rcu_dereference(next)) - #endif /* __KERNEL__ */ #endif diff --git a/include/net/sock.h b/include/net/sock.h index 8b2b82131b6..0a638948868 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -42,6 +42,7 @@ #include #include +#include #include #include #include @@ -52,6 +53,7 @@ #include #include +#include #include #include @@ -106,6 +108,7 @@ struct net; * @skc_reuse: %SO_REUSEADDR setting * @skc_bound_dev_if: bound device index if != 0 * @skc_node: main hash linkage for various protocol lookup tables + * @skc_nulls_node: main hash linkage for UDP/UDP-Lite protocol * @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 @@ -120,7 +123,10 @@ struct sock_common { volatile unsigned char skc_state; unsigned char skc_reuse; int skc_bound_dev_if; - struct hlist_node skc_node; + union { + struct hlist_node skc_node; + struct hlist_nulls_node skc_nulls_node; + }; struct hlist_node skc_bind_node; atomic_t skc_refcnt; unsigned int skc_hash; @@ -206,6 +212,7 @@ struct sock { #define sk_reuse __sk_common.skc_reuse #define sk_bound_dev_if __sk_common.skc_bound_dev_if #define sk_node __sk_common.skc_node +#define sk_nulls_node __sk_common.skc_nulls_node #define sk_bind_node __sk_common.skc_bind_node #define sk_refcnt __sk_common.skc_refcnt #define sk_hash __sk_common.skc_hash @@ -300,12 +307,30 @@ static inline struct sock *sk_head(const struct hlist_head *head) return hlist_empty(head) ? NULL : __sk_head(head); } +static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) +{ + return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); +} + +static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) +{ + return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); +} + static inline struct sock *sk_next(const struct sock *sk) { return sk->sk_node.next ? hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; } +static inline struct sock *sk_nulls_next(const struct sock *sk) +{ + return (!is_a_nulls(sk->sk_nulls_node.next)) ? + hlist_nulls_entry(sk->sk_nulls_node.next, + struct sock, sk_nulls_node) : + NULL; +} + static inline int sk_unhashed(const struct sock *sk) { return hlist_unhashed(&sk->sk_node); @@ -321,6 +346,11 @@ static __inline__ void sk_node_init(struct hlist_node *node) node->pprev = NULL; } +static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node) +{ + node->pprev = NULL; +} + static __inline__ void __sk_del_node(struct sock *sk) { __hlist_del(&sk->sk_node); @@ -367,18 +397,18 @@ static __inline__ int sk_del_node_init(struct sock *sk) return rc; } -static __inline__ int __sk_del_node_init_rcu(struct sock *sk) +static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk) { if (sk_hashed(sk)) { - hlist_del_init_rcu(&sk->sk_node); + hlist_nulls_del_init_rcu(&sk->sk_nulls_node); return 1; } return 0; } -static __inline__ int sk_del_node_init_rcu(struct sock *sk) +static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk) { - int rc = __sk_del_node_init_rcu(sk); + int rc = __sk_nulls_del_node_init_rcu(sk); if (rc) { /* paranoid for a while -acme */ @@ -399,15 +429,15 @@ static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) __sk_add_node(sk, list); } -static __inline__ void __sk_add_node_rcu(struct sock *sk, struct hlist_head *list) +static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) { - hlist_add_head_rcu(&sk->sk_node, list); + hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); } -static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) +static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) { sock_hold(sk); - __sk_add_node_rcu(sk, list); + __sk_nulls_add_node_rcu(sk, list); } static __inline__ void __sk_del_bind_node(struct sock *sk) @@ -423,11 +453,16 @@ static __inline__ void sk_add_bind_node(struct sock *sk, #define sk_for_each(__sk, node, list) \ hlist_for_each_entry(__sk, node, list, sk_node) -#define sk_for_each_rcu_safenext(__sk, node, list, next) \ - hlist_for_each_entry_rcu_safenext(__sk, node, list, sk_node, next) +#define sk_nulls_for_each(__sk, node, list) \ + hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) +#define sk_nulls_for_each_rcu(__sk, node, list) \ + hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) #define sk_for_each_from(__sk, node) \ if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ hlist_for_each_entry_from(__sk, node, sk_node) +#define sk_nulls_for_each_from(__sk, node) \ + if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ + hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) #define sk_for_each_continue(__sk, node) \ if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ hlist_for_each_entry_continue(__sk, node, sk_node) diff --git a/include/net/udp.h b/include/net/udp.h index df2bfe54537..90e6ce56be6 100644 --- a/include/net/udp.h +++ b/include/net/udp.h @@ -51,7 +51,7 @@ struct udp_skb_cb { #define UDP_SKB_CB(__skb) ((struct udp_skb_cb *)((__skb)->cb)) struct udp_hslot { - struct hlist_head head; + struct hlist_nulls_head head; spinlock_t lock; } __attribute__((aligned(2 * sizeof(long)))); struct udp_table { diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c index 54badc9a019..fea2d873dd4 100644 --- a/net/ipv4/udp.c +++ b/net/ipv4/udp.c @@ -127,9 +127,9 @@ static int udp_lib_lport_inuse(struct net *net, __u16 num, const struct sock *sk2)) { struct sock *sk2; - struct hlist_node *node; + struct hlist_nulls_node *node; - sk_for_each(sk2, node, &hslot->head) + sk_nulls_for_each(sk2, node, &hslot->head) if (net_eq(sock_net(sk2), net) && sk2 != sk && sk2->sk_hash == num && @@ -189,12 +189,7 @@ int udp_lib_get_port(struct sock *sk, unsigned short snum, inet_sk(sk)->num = snum; sk->sk_hash = snum; if (sk_unhashed(sk)) { - /* - * We need that previous write to sk->sk_hash committed - * before write to sk->next done in following add_node() variant - */ - smp_wmb(); - sk_add_node_rcu(sk, &hslot->head); + sk_nulls_add_node_rcu(sk, &hslot->head); sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); } error = 0; @@ -261,7 +256,7 @@ static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, int dif, struct udp_table *udptable) { struct sock *sk, *result; - struct hlist_node *node, *next; + struct hlist_nulls_node *node; unsigned short hnum = ntohs(dport); unsigned int hash = udp_hashfn(net, hnum); struct udp_hslot *hslot = &udptable->hash[hash]; @@ -271,13 +266,7 @@ static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, begin: result = NULL; badness = -1; - sk_for_each_rcu_safenext(sk, node, &hslot->head, next) { - /* - * lockless reader, and SLAB_DESTROY_BY_RCU items: - * We must check this item was not moved to another chain - */ - if (udp_hashfn(net, sk->sk_hash) != hash) - goto begin; + sk_nulls_for_each_rcu(sk, node, &hslot->head) { score = compute_score(sk, net, saddr, hnum, sport, daddr, dport, dif); if (score > badness) { @@ -285,6 +274,14 @@ begin: badness = score; } } + /* + * if the nulls value we got at the end of this lookup is + * not the expected one, we must restart lookup. + * We probably met an item that was moved to another chain. + */ + if (get_nulls_value(node) != hash) + goto begin; + if (result) { if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt))) result = NULL; @@ -325,11 +322,11 @@ static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk, __be16 rmt_port, __be32 rmt_addr, int dif) { - struct hlist_node *node; + struct hlist_nulls_node *node; struct sock *s = sk; unsigned short hnum = ntohs(loc_port); - sk_for_each_from(s, node) { + sk_nulls_for_each_from(s, node) { struct inet_sock *inet = inet_sk(s); if (!net_eq(sock_net(s), net) || @@ -977,7 +974,7 @@ void udp_lib_unhash(struct sock *sk) struct udp_hslot *hslot = &udptable->hash[hash]; spin_lock_bh(&hslot->lock); - if (sk_del_node_init_rcu(sk)) { + if (sk_nulls_del_node_init_rcu(sk)) { inet_sk(sk)->num = 0; sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); } @@ -1130,7 +1127,7 @@ static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, int dif; spin_lock(&hslot->lock); - sk = sk_head(&hslot->head); + sk = sk_nulls_head(&hslot->head); dif = skb->dev->ifindex; sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif); if (sk) { @@ -1139,7 +1136,7 @@ static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, do { struct sk_buff *skb1 = skb; - sknext = udp_v4_mcast_next(net, sk_next(sk), uh->dest, + sknext = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest, daddr, uh->source, saddr, dif); if (sknext) @@ -1560,10 +1557,10 @@ static struct sock *udp_get_first(struct seq_file *seq, int start) struct net *net = seq_file_net(seq); for (state->bucket = start; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) { - struct hlist_node *node; + struct hlist_nulls_node *node; struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; spin_lock_bh(&hslot->lock); - sk_for_each(sk, node, &hslot->head) { + sk_nulls_for_each(sk, node, &hslot->head) { if (!net_eq(sock_net(sk), net)) continue; if (sk->sk_family == state->family) @@ -1582,7 +1579,7 @@ static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) struct net *net = seq_file_net(seq); do { - sk = sk_next(sk); + sk = sk_nulls_next(sk); } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); if (!sk) { @@ -1753,7 +1750,7 @@ void __init udp_table_init(struct udp_table *table) int i; for (i = 0; i < UDP_HTABLE_SIZE; i++) { - INIT_HLIST_HEAD(&table->hash[i].head); + INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i); spin_lock_init(&table->hash[i].lock); } } diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c index 8dafa36b1ba..fd2d9ad4a8a 100644 --- a/net/ipv6/udp.c +++ b/net/ipv6/udp.c @@ -98,7 +98,7 @@ static struct sock *__udp6_lib_lookup(struct net *net, int dif, struct udp_table *udptable) { struct sock *sk, *result; - struct hlist_node *node, *next; + struct hlist_nulls_node *node; unsigned short hnum = ntohs(dport); unsigned int hash = udp_hashfn(net, hnum); struct udp_hslot *hslot = &udptable->hash[hash]; @@ -108,19 +108,21 @@ static struct sock *__udp6_lib_lookup(struct net *net, begin: result = NULL; badness = -1; - sk_for_each_rcu_safenext(sk, node, &hslot->head, next) { - /* - * lockless reader, and SLAB_DESTROY_BY_RCU items: - * We must check this item was not moved to another chain - */ - if (udp_hashfn(net, sk->sk_hash) != hash) - goto begin; + sk_nulls_for_each_rcu(sk, node, &hslot->head) { score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif); if (score > badness) { result = sk; badness = score; } } + /* + * if the nulls value we got at the end of this lookup is + * not the expected one, we must restart lookup. + * We probably met an item that was moved to another chain. + */ + if (get_nulls_value(node) != hash) + goto begin; + if (result) { if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt))) result = NULL; @@ -374,11 +376,11 @@ static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk, __be16 rmt_port, struct in6_addr *rmt_addr, int dif) { - struct hlist_node *node; + struct hlist_nulls_node *node; struct sock *s = sk; unsigned short num = ntohs(loc_port); - sk_for_each_from(s, node) { + sk_nulls_for_each_from(s, node) { struct inet_sock *inet = inet_sk(s); if (!net_eq(sock_net(s), net)) @@ -423,7 +425,7 @@ static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, int dif; spin_lock(&hslot->lock); - sk = sk_head(&hslot->head); + sk = sk_nulls_head(&hslot->head); dif = inet6_iif(skb); sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif); if (!sk) { @@ -432,7 +434,7 @@ static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, } sk2 = sk; - while ((sk2 = udp_v6_mcast_next(net, sk_next(sk2), uh->dest, daddr, + while ((sk2 = udp_v6_mcast_next(net, sk_nulls_next(sk2), uh->dest, daddr, uh->source, saddr, dif))) { struct sk_buff *buff = skb_clone(skb, GFP_ATOMIC); if (buff) { -- cgit v1.2.3-70-g09d2 From 1748376b6626acf59c24e9592ac67b3fe2a0e026 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 25 Nov 2008 21:16:35 -0800 Subject: net: Use a percpu_counter for sockets_allocated Instead of using one atomic_t per protocol, use a percpu_counter for "sockets_allocated", to reduce cache line contention on heavy duty network servers. Note : We revert commit (248969ae31e1b3276fc4399d67ce29a5d81e6fd9 net: af_unix can make unix_nr_socks visbile in /proc), since it is not anymore used after sock_prot_inuse_add() addition Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sctp/sctp.h | 1 + include/net/sock.h | 2 +- include/net/tcp.h | 2 +- net/core/sock.c | 10 +++++++--- net/ipv4/proc.c | 3 ++- net/ipv4/tcp.c | 8 ++++++-- net/ipv4/tcp_ipv4.c | 4 ++-- net/ipv6/tcp_ipv6.c | 2 +- net/sctp/protocol.c | 6 +++++- net/sctp/socket.c | 6 +++--- net/unix/af_unix.c | 1 - 11 files changed, 29 insertions(+), 16 deletions(-) (limited to 'include/net/sock.h') diff --git a/include/net/sctp/sctp.h b/include/net/sctp/sctp.h index 23797506f59..bbb7742195b 100644 --- a/include/net/sctp/sctp.h +++ b/include/net/sctp/sctp.h @@ -138,6 +138,7 @@ void sctp_write_space(struct sock *sk); unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait); void sctp_sock_rfree(struct sk_buff *skb); +extern struct percpu_counter sctp_sockets_allocated; /* * sctp/primitive.c diff --git a/include/net/sock.h b/include/net/sock.h index 00cd486d362..a2a3890959c 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -649,7 +649,7 @@ struct proto { /* Memory pressure */ void (*enter_memory_pressure)(struct sock *sk); atomic_t *memory_allocated; /* Current allocated memory. */ - atomic_t *sockets_allocated; /* Current number of sockets. */ + struct percpu_counter *sockets_allocated; /* Current number of sockets. */ /* * Pressure flag: try to collapse. * Technical note: it is used by multiple contexts non atomically. diff --git a/include/net/tcp.h b/include/net/tcp.h index e8ae90a8c35..cbca3b8a133 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -238,7 +238,7 @@ extern int sysctl_tcp_slow_start_after_idle; extern int sysctl_tcp_max_ssthresh; extern atomic_t tcp_memory_allocated; -extern atomic_t tcp_sockets_allocated; +extern struct percpu_counter tcp_sockets_allocated; extern int tcp_memory_pressure; /* diff --git a/net/core/sock.c b/net/core/sock.c index a4e840e5a05..7a081b647bf 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -1071,7 +1071,7 @@ struct sock *sk_clone(const struct sock *sk, const gfp_t priority) newsk->sk_sleep = NULL; if (newsk->sk_prot->sockets_allocated) - atomic_inc(newsk->sk_prot->sockets_allocated); + percpu_counter_inc(newsk->sk_prot->sockets_allocated); } out: return newsk; @@ -1463,8 +1463,12 @@ int __sk_mem_schedule(struct sock *sk, int size, int kind) } if (prot->memory_pressure) { - if (!*prot->memory_pressure || - prot->sysctl_mem[2] > atomic_read(prot->sockets_allocated) * + int alloc; + + if (!*prot->memory_pressure) + return 1; + alloc = percpu_counter_read_positive(prot->sockets_allocated); + if (prot->sysctl_mem[2] > alloc * sk_mem_pages(sk->sk_wmem_queued + atomic_read(&sk->sk_rmem_alloc) + sk->sk_forward_alloc)) diff --git a/net/ipv4/proc.c b/net/ipv4/proc.c index 731789bb499..4944b47ad62 100644 --- a/net/ipv4/proc.c +++ b/net/ipv4/proc.c @@ -55,7 +55,8 @@ static int sockstat_seq_show(struct seq_file *seq, void *v) seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %d\n", sock_prot_inuse_get(net, &tcp_prot), atomic_read(&tcp_orphan_count), - tcp_death_row.tw_count, atomic_read(&tcp_sockets_allocated), + tcp_death_row.tw_count, + (int)percpu_counter_sum_positive(&tcp_sockets_allocated), atomic_read(&tcp_memory_allocated)); seq_printf(seq, "UDP: inuse %d mem %d\n", sock_prot_inuse_get(net, &udp_prot), diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 044224a341e..e6fade9ebf6 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -290,9 +290,12 @@ EXPORT_SYMBOL(sysctl_tcp_rmem); EXPORT_SYMBOL(sysctl_tcp_wmem); atomic_t tcp_memory_allocated; /* Current allocated memory. */ -atomic_t tcp_sockets_allocated; /* Current number of TCP sockets. */ - EXPORT_SYMBOL(tcp_memory_allocated); + +/* + * Current number of TCP sockets. + */ +struct percpu_counter tcp_sockets_allocated; EXPORT_SYMBOL(tcp_sockets_allocated); /* @@ -2685,6 +2688,7 @@ void __init tcp_init(void) BUILD_BUG_ON(sizeof(struct tcp_skb_cb) > sizeof(skb->cb)); + percpu_counter_init(&tcp_sockets_allocated, 0); tcp_hashinfo.bind_bucket_cachep = kmem_cache_create("tcp_bind_bucket", sizeof(struct inet_bind_bucket), 0, diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c index cab2458f86f..26b9030747c 100644 --- a/net/ipv4/tcp_ipv4.c +++ b/net/ipv4/tcp_ipv4.c @@ -1797,7 +1797,7 @@ static int tcp_v4_init_sock(struct sock *sk) sk->sk_sndbuf = sysctl_tcp_wmem[1]; sk->sk_rcvbuf = sysctl_tcp_rmem[1]; - atomic_inc(&tcp_sockets_allocated); + percpu_counter_inc(&tcp_sockets_allocated); return 0; } @@ -1845,7 +1845,7 @@ void tcp_v4_destroy_sock(struct sock *sk) sk->sk_sndmsg_page = NULL; } - atomic_dec(&tcp_sockets_allocated); + percpu_counter_dec(&tcp_sockets_allocated); } EXPORT_SYMBOL(tcp_v4_destroy_sock); diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c index f259c9671f3..8702b06cb60 100644 --- a/net/ipv6/tcp_ipv6.c +++ b/net/ipv6/tcp_ipv6.c @@ -1830,7 +1830,7 @@ static int tcp_v6_init_sock(struct sock *sk) sk->sk_sndbuf = sysctl_tcp_wmem[1]; sk->sk_rcvbuf = sysctl_tcp_rmem[1]; - atomic_inc(&tcp_sockets_allocated); + percpu_counter_inc(&tcp_sockets_allocated); return 0; } diff --git a/net/sctp/protocol.c b/net/sctp/protocol.c index a8ca743241e..d5ea232c912 100644 --- a/net/sctp/protocol.c +++ b/net/sctp/protocol.c @@ -102,6 +102,8 @@ struct sock *sctp_get_ctl_sock(void) /* Set up the proc fs entry for the SCTP protocol. */ static __init int sctp_proc_init(void) { + if (percpu_counter_init(&sctp_sockets_allocated, 0)) + goto out_nomem; #ifdef CONFIG_PROC_FS if (!proc_net_sctp) { struct proc_dir_entry *ent; @@ -110,7 +112,7 @@ static __init int sctp_proc_init(void) ent->owner = THIS_MODULE; proc_net_sctp = ent; } else - goto out_nomem; + goto out_free_percpu; } if (sctp_snmp_proc_init()) @@ -135,6 +137,8 @@ out_snmp_proc_init: proc_net_sctp = NULL; remove_proc_entry("sctp", init_net.proc_net); } +out_free_percpu: + percpu_counter_destroy(&sctp_sockets_allocated); out_nomem: return -ENOMEM; #else diff --git a/net/sctp/socket.c b/net/sctp/socket.c index ba81fe3ccab..a2de585888d 100644 --- a/net/sctp/socket.c +++ b/net/sctp/socket.c @@ -114,7 +114,7 @@ extern int sysctl_sctp_wmem[3]; static int sctp_memory_pressure; static atomic_t sctp_memory_allocated; -static atomic_t sctp_sockets_allocated; +struct percpu_counter sctp_sockets_allocated; static void sctp_enter_memory_pressure(struct sock *sk) { @@ -3613,7 +3613,7 @@ SCTP_STATIC int sctp_init_sock(struct sock *sk) sp->hmac = NULL; SCTP_DBG_OBJCNT_INC(sock); - atomic_inc(&sctp_sockets_allocated); + percpu_counter_inc(&sctp_sockets_allocated); local_bh_disable(); sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); @@ -3632,7 +3632,7 @@ SCTP_STATIC void sctp_destroy_sock(struct sock *sk) /* Release our hold on the endpoint. */ ep = sctp_sk(sk)->ep; sctp_endpoint_free(ep); - atomic_dec(&sctp_sockets_allocated); + percpu_counter_dec(&sctp_sockets_allocated); local_bh_disable(); sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); local_bh_enable(); diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c index 3a35a6e8bf9..5aaf23e43f1 100644 --- a/net/unix/af_unix.c +++ b/net/unix/af_unix.c @@ -571,7 +571,6 @@ static const struct proto_ops unix_seqpacket_ops = { static struct proto unix_proto = { .name = "UNIX", .owner = THIS_MODULE, - .sockets_allocated = &unix_nr_socks, .obj_size = sizeof(struct unix_sock), }; -- cgit v1.2.3-70-g09d2 From dd24c00191d5e4a1ae896aafe33c6b8095ab4bd1 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 25 Nov 2008 21:17:14 -0800 Subject: net: Use a percpu_counter for orphan_count Instead of using one atomic_t per protocol, use a percpu_counter for "orphan_count", to reduce cache line contention on heavy duty network servers. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 2 +- include/net/tcp.h | 2 +- net/dccp/dccp.h | 2 +- net/dccp/proto.c | 16 ++++++++++------ net/ipv4/inet_connection_sock.c | 4 ++-- net/ipv4/proc.c | 2 +- net/ipv4/tcp.c | 12 +++++++----- net/ipv4/tcp_timer.c | 2 +- 8 files changed, 24 insertions(+), 18 deletions(-) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index a2a3890959c..5a3a151bd73 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -666,7 +666,7 @@ struct proto { unsigned int obj_size; int slab_flags; - atomic_t *orphan_count; + struct percpu_counter *orphan_count; struct request_sock_ops *rsk_prot; struct timewait_sock_ops *twsk_prot; diff --git a/include/net/tcp.h b/include/net/tcp.h index cbca3b8a133..de1e91d959b 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -46,7 +46,7 @@ extern struct inet_hashinfo tcp_hashinfo; -extern atomic_t tcp_orphan_count; +extern struct percpu_counter tcp_orphan_count; extern void tcp_time_wait(struct sock *sk, int state, int timeo); #define MAX_TCP_HEADER (128 + MAX_HEADER) diff --git a/net/dccp/dccp.h b/net/dccp/dccp.h index 031ce350d3c..33a1127270c 100644 --- a/net/dccp/dccp.h +++ b/net/dccp/dccp.h @@ -49,7 +49,7 @@ extern int dccp_debug; extern struct inet_hashinfo dccp_hashinfo; -extern atomic_t dccp_orphan_count; +extern struct percpu_counter dccp_orphan_count; extern void dccp_time_wait(struct sock *sk, int state, int timeo); diff --git a/net/dccp/proto.c b/net/dccp/proto.c index ea85c423cdb..db225f93cd5 100644 --- a/net/dccp/proto.c +++ b/net/dccp/proto.c @@ -40,8 +40,7 @@ DEFINE_SNMP_STAT(struct dccp_mib, dccp_statistics) __read_mostly; EXPORT_SYMBOL_GPL(dccp_statistics); -atomic_t dccp_orphan_count = ATOMIC_INIT(0); - +struct percpu_counter dccp_orphan_count; EXPORT_SYMBOL_GPL(dccp_orphan_count); struct inet_hashinfo dccp_hashinfo; @@ -1000,7 +999,7 @@ adjudge_to_death: state = sk->sk_state; sock_hold(sk); sock_orphan(sk); - atomic_inc(sk->sk_prot->orphan_count); + percpu_counter_inc(sk->sk_prot->orphan_count); /* * It is the last release_sock in its life. It will remove backlog. @@ -1064,18 +1063,21 @@ static int __init dccp_init(void) { unsigned long goal; int ehash_order, bhash_order, i; - int rc = -ENOBUFS; + int rc; BUILD_BUG_ON(sizeof(struct dccp_skb_cb) > FIELD_SIZEOF(struct sk_buff, cb)); - + rc = percpu_counter_init(&dccp_orphan_count, 0); + if (rc) + goto out; + rc = -ENOBUFS; inet_hashinfo_init(&dccp_hashinfo); dccp_hashinfo.bind_bucket_cachep = kmem_cache_create("dccp_bind_bucket", sizeof(struct inet_bind_bucket), 0, SLAB_HWCACHE_ALIGN, NULL); if (!dccp_hashinfo.bind_bucket_cachep) - goto out; + goto out_free_percpu; /* * Size and allocate the main established and bind bucket @@ -1168,6 +1170,8 @@ out_free_dccp_ehash: out_free_bind_bucket_cachep: kmem_cache_destroy(dccp_hashinfo.bind_bucket_cachep); dccp_hashinfo.bind_bucket_cachep = NULL; +out_free_percpu: + percpu_counter_destroy(&dccp_orphan_count); goto out; } diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c index 05af807ca9b..1ccdbba528b 100644 --- a/net/ipv4/inet_connection_sock.c +++ b/net/ipv4/inet_connection_sock.c @@ -561,7 +561,7 @@ void inet_csk_destroy_sock(struct sock *sk) sk_refcnt_debug_release(sk); - atomic_dec(sk->sk_prot->orphan_count); + percpu_counter_dec(sk->sk_prot->orphan_count); sock_put(sk); } @@ -641,7 +641,7 @@ void inet_csk_listen_stop(struct sock *sk) sock_orphan(child); - atomic_inc(sk->sk_prot->orphan_count); + percpu_counter_inc(sk->sk_prot->orphan_count); inet_csk_destroy_sock(child); diff --git a/net/ipv4/proc.c b/net/ipv4/proc.c index 4944b47ad62..614958b7c27 100644 --- a/net/ipv4/proc.c +++ b/net/ipv4/proc.c @@ -54,7 +54,7 @@ static int sockstat_seq_show(struct seq_file *seq, void *v) socket_seq_show(seq); seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %d\n", sock_prot_inuse_get(net, &tcp_prot), - atomic_read(&tcp_orphan_count), + (int)percpu_counter_sum_positive(&tcp_orphan_count), tcp_death_row.tw_count, (int)percpu_counter_sum_positive(&tcp_sockets_allocated), atomic_read(&tcp_memory_allocated)); diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index e6fade9ebf6..01924340862 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -277,8 +277,7 @@ int sysctl_tcp_fin_timeout __read_mostly = TCP_FIN_TIMEOUT; -atomic_t tcp_orphan_count = ATOMIC_INIT(0); - +struct percpu_counter tcp_orphan_count; EXPORT_SYMBOL_GPL(tcp_orphan_count); int sysctl_tcp_mem[3] __read_mostly; @@ -1837,7 +1836,7 @@ adjudge_to_death: state = sk->sk_state; sock_hold(sk); sock_orphan(sk); - atomic_inc(sk->sk_prot->orphan_count); + percpu_counter_inc(sk->sk_prot->orphan_count); /* It is the last release_sock in its life. It will remove backlog. */ release_sock(sk); @@ -1888,9 +1887,11 @@ adjudge_to_death: } } if (sk->sk_state != TCP_CLOSE) { + int orphan_count = percpu_counter_read_positive( + sk->sk_prot->orphan_count); + sk_mem_reclaim(sk); - if (tcp_too_many_orphans(sk, - atomic_read(sk->sk_prot->orphan_count))) { + if (tcp_too_many_orphans(sk, orphan_count)) { if (net_ratelimit()) printk(KERN_INFO "TCP: too many of orphaned " "sockets\n"); @@ -2689,6 +2690,7 @@ void __init tcp_init(void) BUILD_BUG_ON(sizeof(struct tcp_skb_cb) > sizeof(skb->cb)); percpu_counter_init(&tcp_sockets_allocated, 0); + percpu_counter_init(&tcp_orphan_count, 0); tcp_hashinfo.bind_bucket_cachep = kmem_cache_create("tcp_bind_bucket", sizeof(struct inet_bind_bucket), 0, diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c index 3df339e3e36..cc4e6d27ded 100644 --- a/net/ipv4/tcp_timer.c +++ b/net/ipv4/tcp_timer.c @@ -65,7 +65,7 @@ static void tcp_write_err(struct sock *sk) static int tcp_out_of_resources(struct sock *sk, int do_reset) { struct tcp_sock *tp = tcp_sk(sk); - int orphans = atomic_read(&tcp_orphan_count); + int orphans = percpu_counter_read_positive(&tcp_orphan_count); /* If peer does not open window for long time, or did not transmit * anything for long time, penalize it. */ -- cgit v1.2.3-70-g09d2