From cb820f8e4b7f73d1a32175e6591735b25bb5398d Mon Sep 17 00:00:00 2001 From: Richard Cochran Date: Fri, 19 Jul 2013 19:40:09 +0200 Subject: net: Provide a generic socket error queue delivery method for Tx time stamps. This patch moves the private error queue delivery function from the af_packet code to the core socket method. In this way, network layers only needing the error queue for transmit time stamping can share common code. Signed-off-by: Richard Cochran Signed-off-by: David S. Miller --- net/core/sock.c | 47 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 47 insertions(+) (limited to 'net/core/sock.c') diff --git a/net/core/sock.c b/net/core/sock.c index 548d716c5f62..85e8de1bc7fd 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -93,6 +93,7 @@ #include #include +#include #include #include #include @@ -2425,6 +2426,52 @@ void sock_enable_timestamp(struct sock *sk, int flag) } } +int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, + int level, int type) +{ + struct sock_exterr_skb *serr; + struct sk_buff *skb, *skb2; + int copied, err; + + err = -EAGAIN; + skb = skb_dequeue(&sk->sk_error_queue); + if (skb == NULL) + goto out; + + copied = skb->len; + if (copied > len) { + msg->msg_flags |= MSG_TRUNC; + copied = len; + } + err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); + if (err) + goto out_free_skb; + + sock_recv_timestamp(msg, sk, skb); + + serr = SKB_EXT_ERR(skb); + put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee); + + msg->msg_flags |= MSG_ERRQUEUE; + err = copied; + + /* Reset and regenerate socket error */ + spin_lock_bh(&sk->sk_error_queue.lock); + sk->sk_err = 0; + if ((skb2 = skb_peek(&sk->sk_error_queue)) != NULL) { + sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno; + spin_unlock_bh(&sk->sk_error_queue.lock); + sk->sk_error_report(sk); + } else + spin_unlock_bh(&sk->sk_error_queue.lock); + +out_free_skb: + kfree_skb(skb); +out: + return err; +} +EXPORT_SYMBOL(sock_recv_errqueue); + /* * Get a socket option on an socket. * -- cgit v1.2.3-59-g8ed1b From f2f872f9272a79a1048877ea14c15576f46c225e Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 30 Jul 2013 17:55:08 -0700 Subject: netem: Introduce skb_orphan_partial() helper Commit 547669d483e578 ("tcp: xps: fix reordering issues") added unexpected reorders in case netem is used in a MQ setup for high performance test bed. ETH=eth0 tc qd del dev $ETH root 2>/dev/null tc qd add dev $ETH root handle 1: mq for i in `seq 1 32` do tc qd add dev $ETH parent 1:$i netem delay 100ms done As all tcp packets are orphaned by netem, TCP stack believes it can set skb->ooo_okay on all packets. In order to allow producers to send more packets, we want to keep sk_wmem_alloc from reaching sk_sndbuf limit. We can do that by accounting one byte per skb in netem queues, so that TCP stack is not fooled too much. Tested: With above MQ/netem setup, scaling number of concurrent flows gives linear results and no reorders/retransmits lpq83:~# for n in 1 10 20 30 40 50 60 70 80 90 100 do echo -n "n:$n " ; ./super_netperf $n -H 10.7.7.84; done n:1 198.46 n:10 2002.69 n:20 4000.98 n:30 6006.35 n:40 8020.93 n:50 10032.3 n:60 12081.9 n:70 13971.3 n:80 16009.7 n:90 17117.3 n:100 17425.5 Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 1 + net/core/sock.c | 19 +++++++++++++++++++ net/sched/sch_netem.c | 5 +---- 3 files changed, 21 insertions(+), 4 deletions(-) (limited to 'net/core/sock.c') diff --git a/include/net/sock.h b/include/net/sock.h index b9f2b095b1ab..53d4714709a1 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -1520,6 +1520,7 @@ extern struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force, gfp_t priority); extern void sock_wfree(struct sk_buff *skb); +extern void skb_orphan_partial(struct sk_buff *skb); extern void sock_rfree(struct sk_buff *skb); extern void sock_edemux(struct sk_buff *skb); diff --git a/net/core/sock.c b/net/core/sock.c index 85e8de1bc7fd..a753d97434dc 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -1576,6 +1576,25 @@ void sock_wfree(struct sk_buff *skb) } EXPORT_SYMBOL(sock_wfree); +void skb_orphan_partial(struct sk_buff *skb) +{ + /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc, + * so we do not completely orphan skb, but transfert all + * accounted bytes but one, to avoid unexpected reorders. + */ + if (skb->destructor == sock_wfree +#ifdef CONFIG_INET + || skb->destructor == tcp_wfree +#endif + ) { + atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc); + skb->truesize = 1; + } else { + skb_orphan(skb); + } +} +EXPORT_SYMBOL(skb_orphan_partial); + /* * Read buffer destructor automatically called from kfree_skb. */ diff --git a/net/sched/sch_netem.c b/net/sched/sch_netem.c index 82f6016d89ab..a6d788d45216 100644 --- a/net/sched/sch_netem.c +++ b/net/sched/sch_netem.c @@ -412,12 +412,9 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch) /* If a delay is expected, orphan the skb. (orphaning usually takes * place at TX completion time, so _before_ the link transit delay) - * Ideally, this orphaning should be done after the rate limiting - * module, because this breaks TCP Small Queue, and other mechanisms - * based on socket sk_wmem_alloc. */ if (q->latency || q->jitter) - skb_orphan(skb); + skb_orphan_partial(skb); /* * If we need to duplicate packet, then re-insert at top of the -- cgit v1.2.3-59-g8ed1b From 28d6427109d13b0f447cba5761f88d3548e83605 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Thu, 8 Aug 2013 14:38:47 -0700 Subject: net: attempt high order allocations in sock_alloc_send_pskb() Adding paged frags skbs to af_unix sockets introduced a performance regression on large sends because of additional page allocations, even if each skb could carry at least 100% more payload than before. We can instruct sock_alloc_send_pskb() to attempt high order allocations. Most of the time, it does a single page allocation instead of 8. I added an additional parameter to sock_alloc_send_pskb() to let other users to opt-in for this new feature on followup patches. Tested: Before patch : $ netperf -t STREAM_STREAM STREAM STREAM TEST Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 2304 212992 212992 10.00 46861.15 After patch : $ netperf -t STREAM_STREAM STREAM STREAM TEST Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 2304 212992 212992 10.00 57981.11 Signed-off-by: Eric Dumazet Cc: David Rientjes Signed-off-by: David S. Miller --- drivers/net/macvtap.c | 2 +- drivers/net/tun.c | 2 +- include/net/sock.h | 3 +- net/core/sock.c | 100 +++++++++++++++++++++++++------------------------ net/packet/af_packet.c | 2 +- net/unix/af_unix.c | 6 ++- 6 files changed, 60 insertions(+), 55 deletions(-) (limited to 'net/core/sock.c') diff --git a/drivers/net/macvtap.c b/drivers/net/macvtap.c index 182364abfa35..8f6056d9ba97 100644 --- a/drivers/net/macvtap.c +++ b/drivers/net/macvtap.c @@ -524,7 +524,7 @@ static inline struct sk_buff *macvtap_alloc_skb(struct sock *sk, size_t prepad, linear = len; skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, - err); + err, 0); if (!skb) return NULL; diff --git a/drivers/net/tun.c b/drivers/net/tun.c index b1630479d4a1..978d8654b14a 100644 --- a/drivers/net/tun.c +++ b/drivers/net/tun.c @@ -949,7 +949,7 @@ static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, linear = len; skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, - &err); + &err, 0); if (!skb) return ERR_PTR(err); diff --git a/include/net/sock.h b/include/net/sock.h index ab6a8b75207e..e4bbcbfd07ea 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -1539,7 +1539,8 @@ extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len, unsigned long data_len, int noblock, - int *errcode); + int *errcode, + int max_page_order); extern void *sock_kmalloc(struct sock *sk, int size, gfp_t priority); extern void sock_kfree_s(struct sock *sk, void *mem, int size); diff --git a/net/core/sock.c b/net/core/sock.c index 83667de45ec9..5b6beba494a3 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -1741,24 +1741,23 @@ static long sock_wait_for_wmem(struct sock *sk, long timeo) struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len, unsigned long data_len, int noblock, - int *errcode) + int *errcode, int max_page_order) { - struct sk_buff *skb; + struct sk_buff *skb = NULL; + unsigned long chunk; gfp_t gfp_mask; long timeo; int err; int npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT; + struct page *page; + int i; err = -EMSGSIZE; if (npages > MAX_SKB_FRAGS) goto failure; - gfp_mask = sk->sk_allocation; - if (gfp_mask & __GFP_WAIT) - gfp_mask |= __GFP_REPEAT; - timeo = sock_sndtimeo(sk, noblock); - while (1) { + while (!skb) { err = sock_error(sk); if (err != 0) goto failure; @@ -1767,50 +1766,52 @@ struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len, if (sk->sk_shutdown & SEND_SHUTDOWN) goto failure; - if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) { - skb = alloc_skb(header_len, gfp_mask); - if (skb) { - int i; - - /* No pages, we're done... */ - if (!data_len) - break; - - skb->truesize += data_len; - skb_shinfo(skb)->nr_frags = npages; - for (i = 0; i < npages; i++) { - struct page *page; - - page = alloc_pages(sk->sk_allocation, 0); - if (!page) { - err = -ENOBUFS; - skb_shinfo(skb)->nr_frags = i; - kfree_skb(skb); - goto failure; - } - - __skb_fill_page_desc(skb, i, - page, 0, - (data_len >= PAGE_SIZE ? - PAGE_SIZE : - data_len)); - data_len -= PAGE_SIZE; - } + if (atomic_read(&sk->sk_wmem_alloc) >= sk->sk_sndbuf) { + set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags); + set_bit(SOCK_NOSPACE, &sk->sk_socket->flags); + err = -EAGAIN; + if (!timeo) + goto failure; + if (signal_pending(current)) + goto interrupted; + timeo = sock_wait_for_wmem(sk, timeo); + continue; + } - /* Full success... */ - break; - } - err = -ENOBUFS; + err = -ENOBUFS; + gfp_mask = sk->sk_allocation; + if (gfp_mask & __GFP_WAIT) + gfp_mask |= __GFP_REPEAT; + + skb = alloc_skb(header_len, gfp_mask); + if (!skb) goto failure; + + skb->truesize += data_len; + + for (i = 0; npages > 0; i++) { + int order = max_page_order; + + while (order) { + if (npages >= 1 << order) { + page = alloc_pages(sk->sk_allocation | + __GFP_COMP | __GFP_NOWARN, + order); + if (page) + goto fill_page; + } + order--; + } + page = alloc_page(sk->sk_allocation); + if (!page) + goto failure; +fill_page: + chunk = min_t(unsigned long, data_len, + PAGE_SIZE << order); + skb_fill_page_desc(skb, i, page, 0, chunk); + data_len -= chunk; + npages -= 1 << order; } - set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags); - set_bit(SOCK_NOSPACE, &sk->sk_socket->flags); - err = -EAGAIN; - if (!timeo) - goto failure; - if (signal_pending(current)) - goto interrupted; - timeo = sock_wait_for_wmem(sk, timeo); } skb_set_owner_w(skb, sk); @@ -1819,6 +1820,7 @@ struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len, interrupted: err = sock_intr_errno(timeo); failure: + kfree_skb(skb); *errcode = err; return NULL; } @@ -1827,7 +1829,7 @@ EXPORT_SYMBOL(sock_alloc_send_pskb); struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size, int noblock, int *errcode) { - return sock_alloc_send_pskb(sk, size, 0, noblock, errcode); + return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0); } EXPORT_SYMBOL(sock_alloc_send_skb); diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c index 4cb28a7f639b..6c53dd9f5ccc 100644 --- a/net/packet/af_packet.c +++ b/net/packet/af_packet.c @@ -2181,7 +2181,7 @@ static struct sk_buff *packet_alloc_skb(struct sock *sk, size_t prepad, linear = len; skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, - err); + err, 0); if (!skb) return NULL; diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c index 99dc760cdd95..fee9e3397cd1 100644 --- a/net/unix/af_unix.c +++ b/net/unix/af_unix.c @@ -1479,7 +1479,8 @@ static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock, MAX_SKB_FRAGS * PAGE_SIZE); skb = sock_alloc_send_pskb(sk, len - data_len, data_len, - msg->msg_flags & MSG_DONTWAIT, &err); + msg->msg_flags & MSG_DONTWAIT, &err, + PAGE_ALLOC_COSTLY_ORDER); if (skb == NULL) goto out; @@ -1651,7 +1652,8 @@ static int unix_stream_sendmsg(struct kiocb *kiocb, struct socket *sock, data_len = max_t(int, 0, size - SKB_MAX_HEAD(0)); skb = sock_alloc_send_pskb(sk, size - data_len, data_len, - msg->msg_flags & MSG_DONTWAIT, &err); + msg->msg_flags & MSG_DONTWAIT, &err, + get_order(UNIX_SKB_FRAGS_SZ)); if (!skb) goto out_err; -- cgit v1.2.3-59-g8ed1b