diff options
author | Eric Dumazet <eric.dumazet@gmail.com> | 2010-04-14 05:55:35 -0400 |
---|---|---|
committer | David S. Miller <davem@davemloft.net> | 2010-04-21 19:19:29 -0400 |
commit | 989a2979205dd34269382b357e6d4b4b6956b889 (patch) | |
tree | 2f504e9f4d8d418dd8fb2d042b076c1318232360 /net | |
parent | e5700aff144fbbba46be40049f0c55fb57283777 (diff) |
fasync: RCU and fine grained locking
kill_fasync() uses a central rwlock, candidate for RCU conversion, to
avoid cache line ping pongs on SMP.
fasync_remove_entry() and fasync_add_entry() can disable IRQS on a short
section instead during whole list scan.
Use a spinlock per fasync_struct to synchronize kill_fasync_rcu() and
fasync_{remove|add}_entry(). This spinlock is IRQ safe, so sock_fasync()
doesnt need its own implementation and can use fasync_helper(), to
reduce code size and complexity.
We can remove __kill_fasync() direct use in net/socket.c, and rename it
to kill_fasync_rcu().
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net')
-rw-r--r-- | net/socket.c | 73 |
1 files changed, 11 insertions, 62 deletions
diff --git a/net/socket.c b/net/socket.c index 35bc198bbf68..9822081eab38 100644 --- a/net/socket.c +++ b/net/socket.c | |||
@@ -1067,78 +1067,27 @@ static int sock_close(struct inode *inode, struct file *filp) | |||
1067 | * 1. fasync_list is modified only under process context socket lock | 1067 | * 1. fasync_list is modified only under process context socket lock |
1068 | * i.e. under semaphore. | 1068 | * i.e. under semaphore. |
1069 | * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) | 1069 | * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) |
1070 | * or under socket lock. | 1070 | * or under socket lock |
1071 | * 3. fasync_list can be used from softirq context, so that | ||
1072 | * modification under socket lock have to be enhanced with | ||
1073 | * write_lock_bh(&sk->sk_callback_lock). | ||
1074 | * --ANK (990710) | ||
1075 | */ | 1071 | */ |
1076 | 1072 | ||
1077 | static int sock_fasync(int fd, struct file *filp, int on) | 1073 | static int sock_fasync(int fd, struct file *filp, int on) |
1078 | { | 1074 | { |
1079 | struct fasync_struct *fa, *fna = NULL, **prev; | 1075 | struct socket *sock = filp->private_data; |
1080 | struct socket *sock; | 1076 | struct sock *sk = sock->sk; |
1081 | struct sock *sk; | ||
1082 | |||
1083 | if (on) { | ||
1084 | fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL); | ||
1085 | if (fna == NULL) | ||
1086 | return -ENOMEM; | ||
1087 | } | ||
1088 | |||
1089 | sock = filp->private_data; | ||
1090 | 1077 | ||
1091 | sk = sock->sk; | 1078 | if (sk == NULL) |
1092 | if (sk == NULL) { | ||
1093 | kfree(fna); | ||
1094 | return -EINVAL; | 1079 | return -EINVAL; |
1095 | } | ||
1096 | 1080 | ||
1097 | lock_sock(sk); | 1081 | lock_sock(sk); |
1098 | 1082 | ||
1099 | spin_lock(&filp->f_lock); | 1083 | fasync_helper(fd, filp, on, &sock->fasync_list); |
1100 | if (on) | ||
1101 | filp->f_flags |= FASYNC; | ||
1102 | else | ||
1103 | filp->f_flags &= ~FASYNC; | ||
1104 | spin_unlock(&filp->f_lock); | ||
1105 | |||
1106 | prev = &(sock->fasync_list); | ||
1107 | 1084 | ||
1108 | for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev) | 1085 | if (!sock->fasync_list) |
1109 | if (fa->fa_file == filp) | 1086 | sock_reset_flag(sk, SOCK_FASYNC); |
1110 | break; | 1087 | else |
1111 | |||
1112 | if (on) { | ||
1113 | if (fa != NULL) { | ||
1114 | write_lock_bh(&sk->sk_callback_lock); | ||
1115 | fa->fa_fd = fd; | ||
1116 | write_unlock_bh(&sk->sk_callback_lock); | ||
1117 | |||
1118 | kfree(fna); | ||
1119 | goto out; | ||
1120 | } | ||
1121 | fna->fa_file = filp; | ||
1122 | fna->fa_fd = fd; | ||
1123 | fna->magic = FASYNC_MAGIC; | ||
1124 | fna->fa_next = sock->fasync_list; | ||
1125 | write_lock_bh(&sk->sk_callback_lock); | ||
1126 | sock->fasync_list = fna; | ||
1127 | sock_set_flag(sk, SOCK_FASYNC); | 1088 | sock_set_flag(sk, SOCK_FASYNC); |
1128 | write_unlock_bh(&sk->sk_callback_lock); | ||
1129 | } else { | ||
1130 | if (fa != NULL) { | ||
1131 | write_lock_bh(&sk->sk_callback_lock); | ||
1132 | *prev = fa->fa_next; | ||
1133 | if (!sock->fasync_list) | ||
1134 | sock_reset_flag(sk, SOCK_FASYNC); | ||
1135 | write_unlock_bh(&sk->sk_callback_lock); | ||
1136 | kfree(fa); | ||
1137 | } | ||
1138 | } | ||
1139 | 1089 | ||
1140 | out: | 1090 | release_sock(sk); |
1141 | release_sock(sock->sk); | ||
1142 | return 0; | 1091 | return 0; |
1143 | } | 1092 | } |
1144 | 1093 | ||
@@ -1159,10 +1108,10 @@ int sock_wake_async(struct socket *sock, int how, int band) | |||
1159 | /* fall through */ | 1108 | /* fall through */ |
1160 | case SOCK_WAKE_IO: | 1109 | case SOCK_WAKE_IO: |
1161 | call_kill: | 1110 | call_kill: |
1162 | __kill_fasync(sock->fasync_list, SIGIO, band); | 1111 | kill_fasync(&sock->fasync_list, SIGIO, band); |
1163 | break; | 1112 | break; |
1164 | case SOCK_WAKE_URG: | 1113 | case SOCK_WAKE_URG: |
1165 | __kill_fasync(sock->fasync_list, SIGURG, band); | 1114 | kill_fasync(&sock->fasync_list, SIGURG, band); |
1166 | } | 1115 | } |
1167 | return 0; | 1116 | return 0; |
1168 | } | 1117 | } |