From eb6a24816b247c0be6b2e97e68933072874bbe54 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 3 Apr 2012 05:28:28 +0000 Subject: af_unix: reduce high order page allocations unix_dgram_sendmsg() currently builds linear skbs, and this can stress page allocator with high order page allocations. When memory gets fragmented, this can eventually fail. We can try to use order-2 allocations for skb head (SKB_MAX_ALLOC) plus up to 16 page fragments to lower pressure on buddy allocator. This patch has no effect on messages of less than 16064 bytes. (on 64bit arches with PAGE_SIZE=4096) For bigger messages (from 16065 to 81600 bytes), this patch brings reliability at the expense of performance penalty because of extra pages allocations. netperf -t DG_STREAM -T 0,2 -- -m 16064 -s 200000 ->4086040 Messages / 10s netperf -t DG_STREAM -T 0,2 -- -m 16068 -s 200000 ->3901747 Messages / 10s Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- net/unix/af_unix.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) (limited to 'net/unix') diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c index d510353ef431..eadb9020cd64 100644 --- a/net/unix/af_unix.c +++ b/net/unix/af_unix.c @@ -1442,6 +1442,7 @@ static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock, long timeo; struct scm_cookie tmp_scm; int max_level; + int data_len = 0; if (NULL == siocb->scm) siocb->scm = &tmp_scm; @@ -1475,7 +1476,13 @@ static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock, if (len > sk->sk_sndbuf - 32) goto out; - skb = sock_alloc_send_skb(sk, len, msg->msg_flags&MSG_DONTWAIT, &err); + if (len > SKB_MAX_ALLOC) + data_len = min_t(size_t, + len - SKB_MAX_ALLOC, + MAX_SKB_FRAGS * PAGE_SIZE); + + skb = sock_alloc_send_pskb(sk, len - data_len, data_len, + msg->msg_flags & MSG_DONTWAIT, &err); if (skb == NULL) goto out; @@ -1485,8 +1492,10 @@ static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock, max_level = err + 1; unix_get_secdata(siocb->scm, skb); - skb_reset_transport_header(skb); - err = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len); + skb_put(skb, len - data_len); + skb->data_len = data_len; + skb->len = len; + err = skb_copy_datagram_from_iovec(skb, 0, msg->msg_iov, 0, len); if (err) goto out_free; -- cgit v1.2.3 From 95c961747284a6b83a5e2d81240e214b0fa3464d Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Sun, 15 Apr 2012 05:58:06 +0000 Subject: net: cleanup unsigned to unsigned int Use of "unsigned int" is preferred to bare "unsigned" in net tree. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- net/unix/af_unix.c | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) (limited to 'net/unix') diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c index eadb9020cd64..641f2e47f165 100644 --- a/net/unix/af_unix.c +++ b/net/unix/af_unix.c @@ -149,9 +149,10 @@ static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) * each socket state is protected by separate spin lock. */ -static inline unsigned unix_hash_fold(__wsum n) +static inline unsigned int unix_hash_fold(__wsum n) { - unsigned hash = (__force unsigned)n; + unsigned int hash = (__force unsigned int)n; + hash ^= hash>>16; hash ^= hash>>8; return hash&(UNIX_HASH_SIZE-1); @@ -200,7 +201,7 @@ static inline void unix_release_addr(struct unix_address *addr) * - if started by zero, it is abstract name. */ -static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned *hashp) +static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned int *hashp) { if (len <= sizeof(short) || len > sizeof(*sunaddr)) return -EINVAL; @@ -250,7 +251,7 @@ static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk) static struct sock *__unix_find_socket_byname(struct net *net, struct sockaddr_un *sunname, - int len, int type, unsigned hash) + int len, int type, unsigned int hash) { struct sock *s; struct hlist_node *node; @@ -273,7 +274,7 @@ found: static inline struct sock *unix_find_socket_byname(struct net *net, struct sockaddr_un *sunname, int len, int type, - unsigned hash) + unsigned int hash) { struct sock *s; @@ -760,7 +761,7 @@ out: mutex_unlock(&u->readlock); static struct sock *unix_find_other(struct net *net, struct sockaddr_un *sunname, int len, - int type, unsigned hash, int *error) + int type, unsigned int hash, int *error) { struct sock *u; struct path path; @@ -824,7 +825,7 @@ static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) struct dentry *dentry = NULL; struct path path; int err; - unsigned hash; + unsigned int hash; struct unix_address *addr; struct hlist_head *list; @@ -964,7 +965,7 @@ static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr, struct net *net = sock_net(sk); struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr; struct sock *other; - unsigned hash; + unsigned int hash; int err; if (addr->sa_family != AF_UNSPEC) { @@ -1062,7 +1063,7 @@ static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr, struct sock *newsk = NULL; struct sock *other = NULL; struct sk_buff *skb = NULL; - unsigned hash; + unsigned int hash; int st; int err; long timeo; @@ -1437,7 +1438,7 @@ static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock, struct sock *other = NULL; int namelen = 0; /* fake GCC */ int err; - unsigned hash; + unsigned int hash; struct sk_buff *skb; long timeo; struct scm_cookie tmp_scm; -- cgit v1.2.3 From 5dd3df105b9f6cb7dd2472b59e028d0d1c878ecb Mon Sep 17 00:00:00 2001 From: "Eric W. Biederman" Date: Thu, 19 Apr 2012 13:24:33 +0000 Subject: net: Move all of the network sysctls without a namespace into init_net. This makes it clearer which sysctls are relative to your current network namespace. This makes it a little less error prone by not exposing sysctls for the initial network namespace in other namespaces. This is the same way we handle all of our other network interfaces to userspace and I can't honestly remember why we didn't do this for sysctls right from the start. Signed-off-by: Eric W. Biederman Acked-by: Pavel Emelyanov Signed-off-by: David S. Miller --- net/unix/sysctl_net_unix.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'net/unix') diff --git a/net/unix/sysctl_net_unix.c b/net/unix/sysctl_net_unix.c index 397cffebb3b6..4f6979c06f84 100644 --- a/net/unix/sysctl_net_unix.c +++ b/net/unix/sysctl_net_unix.c @@ -58,6 +58,6 @@ void unix_sysctl_unregister(struct net *net) struct ctl_table *table; table = net->unx.ctl->ctl_table_arg; - unregister_sysctl_table(net->unx.ctl); + unregister_net_sysctl_table(net->unx.ctl); kfree(table); } -- cgit v1.2.3 From ec8f23ce0f4005b74013d4d122e0d540397a93c9 Mon Sep 17 00:00:00 2001 From: "Eric W. Biederman" Date: Thu, 19 Apr 2012 13:44:49 +0000 Subject: net: Convert all sysctl registrations to register_net_sysctl This results in code with less boiler plate that is a bit easier to read. Additionally stops us from using compatibility code in the sysctl core, hastening the day when the compatibility code can be removed. Signed-off-by: Eric W. Biederman Acked-by: Pavel Emelyanov Signed-off-by: David S. Miller --- net/unix/sysctl_net_unix.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) (limited to 'net/unix') diff --git a/net/unix/sysctl_net_unix.c b/net/unix/sysctl_net_unix.c index 4f6979c06f84..b34b5b9792f0 100644 --- a/net/unix/sysctl_net_unix.c +++ b/net/unix/sysctl_net_unix.c @@ -26,12 +26,6 @@ static ctl_table unix_table[] = { { } }; -static struct ctl_path unix_path[] = { - { .procname = "net", }, - { .procname = "unix", }, - { }, -}; - int __net_init unix_sysctl_register(struct net *net) { struct ctl_table *table; @@ -41,7 +35,7 @@ int __net_init unix_sysctl_register(struct net *net) goto err_alloc; table[0].data = &net->unx.sysctl_max_dgram_qlen; - net->unx.ctl = register_net_sysctl_table(net, unix_path, table); + net->unx.ctl = register_net_sysctl(net, "net/unix", table); if (net->unx.ctl == NULL) goto err_reg; -- cgit v1.2.3 From 8dcf01fc009d12d01fd195ed95eaaee61178f21a Mon Sep 17 00:00:00 2001 From: Shan Wei Date: Tue, 24 Apr 2012 18:21:07 +0000 Subject: net: sock_diag_handler structs can be const read only, so change it to const. Signed-off-by: Shan Wei Acked-by: Pavel Emelyanov Signed-off-by: David S. Miller --- net/unix/diag.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'net/unix') diff --git a/net/unix/diag.c b/net/unix/diag.c index f0486ae9ebe6..47d3002737f5 100644 --- a/net/unix/diag.c +++ b/net/unix/diag.c @@ -310,7 +310,7 @@ static int unix_diag_handler_dump(struct sk_buff *skb, struct nlmsghdr *h) return unix_diag_get_exact(skb, h, (struct unix_diag_req *)NLMSG_DATA(h)); } -static struct sock_diag_handler unix_diag_handler = { +static const struct sock_diag_handler unix_diag_handler = { .family = AF_UNIX, .dump = unix_diag_handler_dump, }; -- cgit v1.2.3