diff options
Diffstat (limited to 'net/ipv4')
| -rw-r--r-- | net/ipv4/fib_semantics.c | 9 | ||||
| -rw-r--r-- | net/ipv4/icmp.c | 3 | ||||
| -rw-r--r-- | net/ipv4/ip_fragment.c | 8 | ||||
| -rw-r--r-- | net/ipv4/ip_gre.c | 21 | ||||
| -rw-r--r-- | net/ipv4/ip_sockglue.c | 3 | ||||
| -rw-r--r-- | net/ipv4/ipip.c | 20 | ||||
| -rw-r--r-- | net/ipv4/ipmr.c | 6 | ||||
| -rw-r--r-- | net/ipv4/netfilter/ip_conntrack_core.c | 5 | ||||
| -rw-r--r-- | net/ipv4/netfilter/ip_nat_standalone.c | 4 | ||||
| -rw-r--r-- | net/ipv4/tcp_ipv4.c | 14 | ||||
| -rw-r--r-- | net/ipv4/tcp_output.c | 100 | ||||
| -rw-r--r-- | net/ipv4/udp.c | 34 |
12 files changed, 130 insertions, 97 deletions
diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c index c886b28ba9f5..e278cb9d0075 100644 --- a/net/ipv4/fib_semantics.c +++ b/net/ipv4/fib_semantics.c | |||
| @@ -593,10 +593,13 @@ static void fib_hash_move(struct hlist_head *new_info_hash, | |||
| 593 | struct hlist_head *new_laddrhash, | 593 | struct hlist_head *new_laddrhash, |
| 594 | unsigned int new_size) | 594 | unsigned int new_size) |
| 595 | { | 595 | { |
| 596 | struct hlist_head *old_info_hash, *old_laddrhash; | ||
| 596 | unsigned int old_size = fib_hash_size; | 597 | unsigned int old_size = fib_hash_size; |
| 597 | unsigned int i; | 598 | unsigned int i, bytes; |
| 598 | 599 | ||
| 599 | write_lock(&fib_info_lock); | 600 | write_lock(&fib_info_lock); |
| 601 | old_info_hash = fib_info_hash; | ||
| 602 | old_laddrhash = fib_info_laddrhash; | ||
| 600 | fib_hash_size = new_size; | 603 | fib_hash_size = new_size; |
| 601 | 604 | ||
| 602 | for (i = 0; i < old_size; i++) { | 605 | for (i = 0; i < old_size; i++) { |
| @@ -636,6 +639,10 @@ static void fib_hash_move(struct hlist_head *new_info_hash, | |||
| 636 | fib_info_laddrhash = new_laddrhash; | 639 | fib_info_laddrhash = new_laddrhash; |
| 637 | 640 | ||
| 638 | write_unlock(&fib_info_lock); | 641 | write_unlock(&fib_info_lock); |
| 642 | |||
| 643 | bytes = old_size * sizeof(struct hlist_head *); | ||
| 644 | fib_hash_free(old_info_hash, bytes); | ||
| 645 | fib_hash_free(old_laddrhash, bytes); | ||
| 639 | } | 646 | } |
| 640 | 647 | ||
| 641 | struct fib_info * | 648 | struct fib_info * |
diff --git a/net/ipv4/icmp.c b/net/ipv4/icmp.c index 279f57abfecb..3d78464f64ea 100644 --- a/net/ipv4/icmp.c +++ b/net/ipv4/icmp.c | |||
| @@ -936,8 +936,7 @@ int icmp_rcv(struct sk_buff *skb) | |||
| 936 | case CHECKSUM_HW: | 936 | case CHECKSUM_HW: |
| 937 | if (!(u16)csum_fold(skb->csum)) | 937 | if (!(u16)csum_fold(skb->csum)) |
| 938 | break; | 938 | break; |
| 939 | NETDEBUG(if (net_ratelimit()) | 939 | LIMIT_NETDEBUG(printk(KERN_DEBUG "icmp v4 hw csum failure\n")); |
| 940 | printk(KERN_DEBUG "icmp v4 hw csum failure\n")); | ||
| 941 | case CHECKSUM_NONE: | 940 | case CHECKSUM_NONE: |
| 942 | if ((u16)csum_fold(skb_checksum(skb, 0, skb->len, 0))) | 941 | if ((u16)csum_fold(skb_checksum(skb, 0, skb->len, 0))) |
| 943 | goto error; | 942 | goto error; |
diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c index 7f68e27eb4ea..eb377ae15305 100644 --- a/net/ipv4/ip_fragment.c +++ b/net/ipv4/ip_fragment.c | |||
| @@ -377,7 +377,7 @@ static struct ipq *ip_frag_create(unsigned hash, struct iphdr *iph, u32 user) | |||
| 377 | return ip_frag_intern(hash, qp); | 377 | return ip_frag_intern(hash, qp); |
| 378 | 378 | ||
| 379 | out_nomem: | 379 | out_nomem: |
| 380 | NETDEBUG(if (net_ratelimit()) printk(KERN_ERR "ip_frag_create: no memory left !\n")); | 380 | LIMIT_NETDEBUG(printk(KERN_ERR "ip_frag_create: no memory left !\n")); |
| 381 | return NULL; | 381 | return NULL; |
| 382 | } | 382 | } |
| 383 | 383 | ||
| @@ -625,10 +625,8 @@ static struct sk_buff *ip_frag_reasm(struct ipq *qp, struct net_device *dev) | |||
| 625 | return head; | 625 | return head; |
| 626 | 626 | ||
| 627 | out_nomem: | 627 | out_nomem: |
| 628 | NETDEBUG(if (net_ratelimit()) | 628 | LIMIT_NETDEBUG(printk(KERN_ERR "IP: queue_glue: no memory for gluing " |
| 629 | printk(KERN_ERR | 629 | "queue %p\n", qp)); |
| 630 | "IP: queue_glue: no memory for gluing queue %p\n", | ||
| 631 | qp)); | ||
| 632 | goto out_fail; | 630 | goto out_fail; |
| 633 | out_oversize: | 631 | out_oversize: |
| 634 | if (net_ratelimit()) | 632 | if (net_ratelimit()) |
diff --git a/net/ipv4/ip_gre.c b/net/ipv4/ip_gre.c index 884835522224..f0d5740d7e22 100644 --- a/net/ipv4/ip_gre.c +++ b/net/ipv4/ip_gre.c | |||
| @@ -290,7 +290,6 @@ static struct ip_tunnel * ipgre_tunnel_locate(struct ip_tunnel_parm *parms, int | |||
| 290 | 290 | ||
| 291 | dev_hold(dev); | 291 | dev_hold(dev); |
| 292 | ipgre_tunnel_link(nt); | 292 | ipgre_tunnel_link(nt); |
| 293 | /* Do not decrement MOD_USE_COUNT here. */ | ||
| 294 | return nt; | 293 | return nt; |
| 295 | 294 | ||
| 296 | failed: | 295 | failed: |
| @@ -1277,12 +1276,28 @@ err1: | |||
| 1277 | goto out; | 1276 | goto out; |
| 1278 | } | 1277 | } |
| 1279 | 1278 | ||
| 1280 | static void ipgre_fini(void) | 1279 | static void __exit ipgre_destroy_tunnels(void) |
| 1280 | { | ||
| 1281 | int prio; | ||
| 1282 | |||
| 1283 | for (prio = 0; prio < 4; prio++) { | ||
| 1284 | int h; | ||
| 1285 | for (h = 0; h < HASH_SIZE; h++) { | ||
| 1286 | struct ip_tunnel *t; | ||
| 1287 | while ((t = tunnels[prio][h]) != NULL) | ||
| 1288 | unregister_netdevice(t->dev); | ||
| 1289 | } | ||
| 1290 | } | ||
| 1291 | } | ||
| 1292 | |||
| 1293 | static void __exit ipgre_fini(void) | ||
| 1281 | { | 1294 | { |
| 1282 | if (inet_del_protocol(&ipgre_protocol, IPPROTO_GRE) < 0) | 1295 | if (inet_del_protocol(&ipgre_protocol, IPPROTO_GRE) < 0) |
| 1283 | printk(KERN_INFO "ipgre close: can't remove protocol\n"); | 1296 | printk(KERN_INFO "ipgre close: can't remove protocol\n"); |
| 1284 | 1297 | ||
| 1285 | unregister_netdev(ipgre_fb_tunnel_dev); | 1298 | rtnl_lock(); |
| 1299 | ipgre_destroy_tunnels(); | ||
| 1300 | rtnl_unlock(); | ||
| 1286 | } | 1301 | } |
| 1287 | 1302 | ||
| 1288 | module_init(ipgre_init); | 1303 | module_init(ipgre_init); |
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c index fc7c481d0d79..ff4bd067b397 100644 --- a/net/ipv4/ip_sockglue.c +++ b/net/ipv4/ip_sockglue.c | |||
| @@ -848,6 +848,9 @@ mc_msf_out: | |||
| 848 | 848 | ||
| 849 | case IP_IPSEC_POLICY: | 849 | case IP_IPSEC_POLICY: |
| 850 | case IP_XFRM_POLICY: | 850 | case IP_XFRM_POLICY: |
| 851 | err = -EPERM; | ||
| 852 | if (!capable(CAP_NET_ADMIN)) | ||
| 853 | break; | ||
| 851 | err = xfrm_user_policy(sk, optname, optval, optlen); | 854 | err = xfrm_user_policy(sk, optname, optval, optlen); |
| 852 | break; | 855 | break; |
| 853 | 856 | ||
diff --git a/net/ipv4/ipip.c b/net/ipv4/ipip.c index c3947cd566b7..c05c1df0bb04 100644 --- a/net/ipv4/ipip.c +++ b/net/ipv4/ipip.c | |||
| @@ -255,7 +255,6 @@ static struct ip_tunnel * ipip_tunnel_locate(struct ip_tunnel_parm *parms, int c | |||
| 255 | 255 | ||
| 256 | dev_hold(dev); | 256 | dev_hold(dev); |
| 257 | ipip_tunnel_link(nt); | 257 | ipip_tunnel_link(nt); |
| 258 | /* Do not decrement MOD_USE_COUNT here. */ | ||
| 259 | return nt; | 258 | return nt; |
| 260 | 259 | ||
| 261 | failed: | 260 | failed: |
| @@ -920,12 +919,29 @@ static int __init ipip_init(void) | |||
| 920 | goto out; | 919 | goto out; |
| 921 | } | 920 | } |
| 922 | 921 | ||
| 922 | static void __exit ipip_destroy_tunnels(void) | ||
| 923 | { | ||
| 924 | int prio; | ||
| 925 | |||
| 926 | for (prio = 1; prio < 4; prio++) { | ||
| 927 | int h; | ||
| 928 | for (h = 0; h < HASH_SIZE; h++) { | ||
| 929 | struct ip_tunnel *t; | ||
| 930 | while ((t = tunnels[prio][h]) != NULL) | ||
| 931 | unregister_netdevice(t->dev); | ||
| 932 | } | ||
| 933 | } | ||
| 934 | } | ||
| 935 | |||
| 923 | static void __exit ipip_fini(void) | 936 | static void __exit ipip_fini(void) |
| 924 | { | 937 | { |
| 925 | if (ipip_unregister() < 0) | 938 | if (ipip_unregister() < 0) |
| 926 | printk(KERN_INFO "ipip close: can't deregister tunnel\n"); | 939 | printk(KERN_INFO "ipip close: can't deregister tunnel\n"); |
| 927 | 940 | ||
| 928 | unregister_netdev(ipip_fb_tunnel_dev); | 941 | rtnl_lock(); |
| 942 | ipip_destroy_tunnels(); | ||
| 943 | unregister_netdevice(ipip_fb_tunnel_dev); | ||
| 944 | rtnl_unlock(); | ||
| 929 | } | 945 | } |
| 930 | 946 | ||
| 931 | module_init(ipip_init); | 947 | module_init(ipip_init); |
diff --git a/net/ipv4/ipmr.c b/net/ipv4/ipmr.c index 7833d920bdba..dc806b578427 100644 --- a/net/ipv4/ipmr.c +++ b/net/ipv4/ipmr.c | |||
| @@ -362,7 +362,7 @@ out: | |||
| 362 | 362 | ||
| 363 | /* Fill oifs list. It is called under write locked mrt_lock. */ | 363 | /* Fill oifs list. It is called under write locked mrt_lock. */ |
| 364 | 364 | ||
| 365 | static void ipmr_update_threshoulds(struct mfc_cache *cache, unsigned char *ttls) | 365 | static void ipmr_update_thresholds(struct mfc_cache *cache, unsigned char *ttls) |
| 366 | { | 366 | { |
| 367 | int vifi; | 367 | int vifi; |
| 368 | 368 | ||
| @@ -727,7 +727,7 @@ static int ipmr_mfc_add(struct mfcctl *mfc, int mrtsock) | |||
| 727 | if (c != NULL) { | 727 | if (c != NULL) { |
| 728 | write_lock_bh(&mrt_lock); | 728 | write_lock_bh(&mrt_lock); |
| 729 | c->mfc_parent = mfc->mfcc_parent; | 729 | c->mfc_parent = mfc->mfcc_parent; |
| 730 | ipmr_update_threshoulds(c, mfc->mfcc_ttls); | 730 | ipmr_update_thresholds(c, mfc->mfcc_ttls); |
| 731 | if (!mrtsock) | 731 | if (!mrtsock) |
| 732 | c->mfc_flags |= MFC_STATIC; | 732 | c->mfc_flags |= MFC_STATIC; |
| 733 | write_unlock_bh(&mrt_lock); | 733 | write_unlock_bh(&mrt_lock); |
| @@ -744,7 +744,7 @@ static int ipmr_mfc_add(struct mfcctl *mfc, int mrtsock) | |||
| 744 | c->mfc_origin=mfc->mfcc_origin.s_addr; | 744 | c->mfc_origin=mfc->mfcc_origin.s_addr; |
| 745 | c->mfc_mcastgrp=mfc->mfcc_mcastgrp.s_addr; | 745 | c->mfc_mcastgrp=mfc->mfcc_mcastgrp.s_addr; |
| 746 | c->mfc_parent=mfc->mfcc_parent; | 746 | c->mfc_parent=mfc->mfcc_parent; |
| 747 | ipmr_update_threshoulds(c, mfc->mfcc_ttls); | 747 | ipmr_update_thresholds(c, mfc->mfcc_ttls); |
| 748 | if (!mrtsock) | 748 | if (!mrtsock) |
| 749 | c->mfc_flags |= MFC_STATIC; | 749 | c->mfc_flags |= MFC_STATIC; |
| 750 | 750 | ||
diff --git a/net/ipv4/netfilter/ip_conntrack_core.c b/net/ipv4/netfilter/ip_conntrack_core.c index 86f04e41dd8e..a7f0c821a9b2 100644 --- a/net/ipv4/netfilter/ip_conntrack_core.c +++ b/net/ipv4/netfilter/ip_conntrack_core.c | |||
| @@ -513,6 +513,11 @@ init_conntrack(const struct ip_conntrack_tuple *tuple, | |||
| 513 | #ifdef CONFIG_IP_NF_CONNTRACK_MARK | 513 | #ifdef CONFIG_IP_NF_CONNTRACK_MARK |
| 514 | conntrack->mark = exp->master->mark; | 514 | conntrack->mark = exp->master->mark; |
| 515 | #endif | 515 | #endif |
| 516 | #if defined(CONFIG_IP_NF_TARGET_MASQUERADE) || \ | ||
| 517 | defined(CONFIG_IP_NF_TARGET_MASQUERADE_MODULE) | ||
| 518 | /* this is ugly, but there is no other place where to put it */ | ||
| 519 | conntrack->nat.masq_index = exp->master->nat.masq_index; | ||
| 520 | #endif | ||
| 516 | nf_conntrack_get(&conntrack->master->ct_general); | 521 | nf_conntrack_get(&conntrack->master->ct_general); |
| 517 | CONNTRACK_STAT_INC(expect_new); | 522 | CONNTRACK_STAT_INC(expect_new); |
| 518 | } else { | 523 | } else { |
diff --git a/net/ipv4/netfilter/ip_nat_standalone.c b/net/ipv4/netfilter/ip_nat_standalone.c index bc59d0d6e89e..91d5ea1dbbc9 100644 --- a/net/ipv4/netfilter/ip_nat_standalone.c +++ b/net/ipv4/netfilter/ip_nat_standalone.c | |||
| @@ -102,6 +102,10 @@ ip_nat_fn(unsigned int hooknum, | |||
| 102 | return NF_ACCEPT; | 102 | return NF_ACCEPT; |
| 103 | } | 103 | } |
| 104 | 104 | ||
| 105 | /* Don't try to NAT if this packet is not conntracked */ | ||
| 106 | if (ct == &ip_conntrack_untracked) | ||
| 107 | return NF_ACCEPT; | ||
| 108 | |||
| 105 | switch (ctinfo) { | 109 | switch (ctinfo) { |
| 106 | case IP_CT_RELATED: | 110 | case IP_CT_RELATED: |
| 107 | case IP_CT_RELATED+IP_CT_IS_REPLY: | 111 | case IP_CT_RELATED+IP_CT_IS_REPLY: |
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c index 62f62bb05c2a..5d91213d34c0 100644 --- a/net/ipv4/tcp_ipv4.c +++ b/net/ipv4/tcp_ipv4.c | |||
| @@ -1494,12 +1494,11 @@ int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb) | |||
| 1494 | * to destinations, already remembered | 1494 | * to destinations, already remembered |
| 1495 | * to the moment of synflood. | 1495 | * to the moment of synflood. |
| 1496 | */ | 1496 | */ |
| 1497 | NETDEBUG(if (net_ratelimit()) \ | 1497 | LIMIT_NETDEBUG(printk(KERN_DEBUG "TCP: drop open " |
| 1498 | printk(KERN_DEBUG "TCP: drop open " | 1498 | "request from %u.%u." |
| 1499 | "request from %u.%u." | 1499 | "%u.%u/%u\n", |
| 1500 | "%u.%u/%u\n", \ | 1500 | NIPQUAD(saddr), |
| 1501 | NIPQUAD(saddr), | 1501 | ntohs(skb->h.th->source))); |
| 1502 | ntohs(skb->h.th->source))); | ||
| 1503 | dst_release(dst); | 1502 | dst_release(dst); |
| 1504 | goto drop_and_free; | 1503 | goto drop_and_free; |
| 1505 | } | 1504 | } |
| @@ -1627,8 +1626,7 @@ static int tcp_v4_checksum_init(struct sk_buff *skb) | |||
| 1627 | skb->nh.iph->daddr, skb->csum)) | 1626 | skb->nh.iph->daddr, skb->csum)) |
| 1628 | return 0; | 1627 | return 0; |
| 1629 | 1628 | ||
| 1630 | NETDEBUG(if (net_ratelimit()) | 1629 | LIMIT_NETDEBUG(printk(KERN_DEBUG "hw tcp v4 csum failed\n")); |
| 1631 | printk(KERN_DEBUG "hw tcp v4 csum failed\n")); | ||
| 1632 | skb->ip_summed = CHECKSUM_NONE; | 1630 | skb->ip_summed = CHECKSUM_NONE; |
| 1633 | } | 1631 | } |
| 1634 | if (skb->len <= 76) { | 1632 | if (skb->len <= 76) { |
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index e3f8ea1bfa9c..3ed6fc15815b 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c | |||
| @@ -403,11 +403,9 @@ static void tcp_queue_skb(struct sock *sk, struct sk_buff *skb) | |||
| 403 | sk->sk_send_head = skb; | 403 | sk->sk_send_head = skb; |
| 404 | } | 404 | } |
| 405 | 405 | ||
| 406 | static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb) | 406 | static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb, unsigned int mss_now) |
| 407 | { | 407 | { |
| 408 | struct tcp_sock *tp = tcp_sk(sk); | 408 | if (skb->len <= mss_now || |
| 409 | |||
| 410 | if (skb->len <= tp->mss_cache || | ||
| 411 | !(sk->sk_route_caps & NETIF_F_TSO)) { | 409 | !(sk->sk_route_caps & NETIF_F_TSO)) { |
| 412 | /* Avoid the costly divide in the normal | 410 | /* Avoid the costly divide in the normal |
| 413 | * non-TSO case. | 411 | * non-TSO case. |
| @@ -417,10 +415,10 @@ static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb) | |||
| 417 | } else { | 415 | } else { |
| 418 | unsigned int factor; | 416 | unsigned int factor; |
| 419 | 417 | ||
| 420 | factor = skb->len + (tp->mss_cache - 1); | 418 | factor = skb->len + (mss_now - 1); |
| 421 | factor /= tp->mss_cache; | 419 | factor /= mss_now; |
| 422 | skb_shinfo(skb)->tso_segs = factor; | 420 | skb_shinfo(skb)->tso_segs = factor; |
| 423 | skb_shinfo(skb)->tso_size = tp->mss_cache; | 421 | skb_shinfo(skb)->tso_size = mss_now; |
| 424 | } | 422 | } |
| 425 | } | 423 | } |
| 426 | 424 | ||
| @@ -429,7 +427,7 @@ static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb) | |||
| 429 | * packet to the list. This won't be called frequently, I hope. | 427 | * packet to the list. This won't be called frequently, I hope. |
| 430 | * Remember, these are still headerless SKBs at this point. | 428 | * Remember, these are still headerless SKBs at this point. |
| 431 | */ | 429 | */ |
| 432 | static int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len) | 430 | static int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len, unsigned int mss_now) |
| 433 | { | 431 | { |
| 434 | struct tcp_sock *tp = tcp_sk(sk); | 432 | struct tcp_sock *tp = tcp_sk(sk); |
| 435 | struct sk_buff *buff; | 433 | struct sk_buff *buff; |
| @@ -492,8 +490,8 @@ static int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len) | |||
| 492 | } | 490 | } |
| 493 | 491 | ||
| 494 | /* Fix up tso_factor for both original and new SKB. */ | 492 | /* Fix up tso_factor for both original and new SKB. */ |
| 495 | tcp_set_skb_tso_segs(sk, skb); | 493 | tcp_set_skb_tso_segs(sk, skb, mss_now); |
| 496 | tcp_set_skb_tso_segs(sk, buff); | 494 | tcp_set_skb_tso_segs(sk, buff, mss_now); |
| 497 | 495 | ||
| 498 | if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST) { | 496 | if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST) { |
| 499 | tp->lost_out += tcp_skb_pcount(skb); | 497 | tp->lost_out += tcp_skb_pcount(skb); |
| @@ -569,7 +567,7 @@ int tcp_trim_head(struct sock *sk, struct sk_buff *skb, u32 len) | |||
| 569 | * factor and mss. | 567 | * factor and mss. |
| 570 | */ | 568 | */ |
| 571 | if (tcp_skb_pcount(skb) > 1) | 569 | if (tcp_skb_pcount(skb) > 1) |
| 572 | tcp_set_skb_tso_segs(sk, skb); | 570 | tcp_set_skb_tso_segs(sk, skb, tcp_current_mss(sk, 1)); |
| 573 | 571 | ||
| 574 | return 0; | 572 | return 0; |
| 575 | } | 573 | } |
| @@ -734,12 +732,14 @@ static inline unsigned int tcp_cwnd_test(struct tcp_sock *tp, struct sk_buff *sk | |||
| 734 | /* This must be invoked the first time we consider transmitting | 732 | /* This must be invoked the first time we consider transmitting |
| 735 | * SKB onto the wire. | 733 | * SKB onto the wire. |
| 736 | */ | 734 | */ |
| 737 | static inline int tcp_init_tso_segs(struct sock *sk, struct sk_buff *skb) | 735 | static inline int tcp_init_tso_segs(struct sock *sk, struct sk_buff *skb, unsigned int mss_now) |
| 738 | { | 736 | { |
| 739 | int tso_segs = tcp_skb_pcount(skb); | 737 | int tso_segs = tcp_skb_pcount(skb); |
| 740 | 738 | ||
| 741 | if (!tso_segs) { | 739 | if (!tso_segs || |
| 742 | tcp_set_skb_tso_segs(sk, skb); | 740 | (tso_segs > 1 && |
| 741 | skb_shinfo(skb)->tso_size != mss_now)) { | ||
| 742 | tcp_set_skb_tso_segs(sk, skb, mss_now); | ||
| 743 | tso_segs = tcp_skb_pcount(skb); | 743 | tso_segs = tcp_skb_pcount(skb); |
| 744 | } | 744 | } |
| 745 | return tso_segs; | 745 | return tso_segs; |
| @@ -817,7 +817,7 @@ static unsigned int tcp_snd_test(struct sock *sk, struct sk_buff *skb, | |||
| 817 | struct tcp_sock *tp = tcp_sk(sk); | 817 | struct tcp_sock *tp = tcp_sk(sk); |
| 818 | unsigned int cwnd_quota; | 818 | unsigned int cwnd_quota; |
| 819 | 819 | ||
| 820 | tcp_init_tso_segs(sk, skb); | 820 | tcp_init_tso_segs(sk, skb, cur_mss); |
| 821 | 821 | ||
| 822 | if (!tcp_nagle_test(tp, skb, cur_mss, nonagle)) | 822 | if (!tcp_nagle_test(tp, skb, cur_mss, nonagle)) |
| 823 | return 0; | 823 | return 0; |
| @@ -854,7 +854,7 @@ int tcp_may_send_now(struct sock *sk, struct tcp_sock *tp) | |||
| 854 | * know that all the data is in scatter-gather pages, and that the | 854 | * know that all the data is in scatter-gather pages, and that the |
| 855 | * packet has never been sent out before (and thus is not cloned). | 855 | * packet has never been sent out before (and thus is not cloned). |
| 856 | */ | 856 | */ |
| 857 | static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len) | 857 | static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len, unsigned int mss_now) |
| 858 | { | 858 | { |
| 859 | struct sk_buff *buff; | 859 | struct sk_buff *buff; |
| 860 | int nlen = skb->len - len; | 860 | int nlen = skb->len - len; |
| @@ -887,8 +887,8 @@ static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len) | |||
| 887 | skb_split(skb, buff, len); | 887 | skb_split(skb, buff, len); |
| 888 | 888 | ||
| 889 | /* Fix up tso_factor for both original and new SKB. */ | 889 | /* Fix up tso_factor for both original and new SKB. */ |
| 890 | tcp_set_skb_tso_segs(sk, skb); | 890 | tcp_set_skb_tso_segs(sk, skb, mss_now); |
| 891 | tcp_set_skb_tso_segs(sk, buff); | 891 | tcp_set_skb_tso_segs(sk, buff, mss_now); |
| 892 | 892 | ||
| 893 | /* Link BUFF into the send queue. */ | 893 | /* Link BUFF into the send queue. */ |
| 894 | skb_header_release(buff); | 894 | skb_header_release(buff); |
| @@ -972,19 +972,18 @@ static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle) | |||
| 972 | if (unlikely(sk->sk_state == TCP_CLOSE)) | 972 | if (unlikely(sk->sk_state == TCP_CLOSE)) |
| 973 | return 0; | 973 | return 0; |
| 974 | 974 | ||
| 975 | skb = sk->sk_send_head; | ||
| 976 | if (unlikely(!skb)) | ||
| 977 | return 0; | ||
| 978 | |||
| 979 | tso_segs = tcp_init_tso_segs(sk, skb); | ||
| 980 | cwnd_quota = tcp_cwnd_test(tp, skb); | ||
| 981 | if (unlikely(!cwnd_quota)) | ||
| 982 | goto out; | ||
| 983 | |||
| 984 | sent_pkts = 0; | 975 | sent_pkts = 0; |
| 985 | while (likely(tcp_snd_wnd_test(tp, skb, mss_now))) { | 976 | while ((skb = sk->sk_send_head)) { |
| 977 | tso_segs = tcp_init_tso_segs(sk, skb, mss_now); | ||
| 986 | BUG_ON(!tso_segs); | 978 | BUG_ON(!tso_segs); |
| 987 | 979 | ||
| 980 | cwnd_quota = tcp_cwnd_test(tp, skb); | ||
| 981 | if (!cwnd_quota) | ||
| 982 | break; | ||
| 983 | |||
| 984 | if (unlikely(!tcp_snd_wnd_test(tp, skb, mss_now))) | ||
| 985 | break; | ||
| 986 | |||
| 988 | if (tso_segs == 1) { | 987 | if (tso_segs == 1) { |
| 989 | if (unlikely(!tcp_nagle_test(tp, skb, mss_now, | 988 | if (unlikely(!tcp_nagle_test(tp, skb, mss_now, |
| 990 | (tcp_skb_is_last(sk, skb) ? | 989 | (tcp_skb_is_last(sk, skb) ? |
| @@ -1006,11 +1005,11 @@ static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle) | |||
| 1006 | limit = skb->len - trim; | 1005 | limit = skb->len - trim; |
| 1007 | } | 1006 | } |
| 1008 | if (skb->len > limit) { | 1007 | if (skb->len > limit) { |
| 1009 | if (tso_fragment(sk, skb, limit)) | 1008 | if (tso_fragment(sk, skb, limit, mss_now)) |
| 1010 | break; | 1009 | break; |
| 1011 | } | 1010 | } |
| 1012 | } else if (unlikely(skb->len > mss_now)) { | 1011 | } else if (unlikely(skb->len > mss_now)) { |
| 1013 | if (unlikely(tcp_fragment(sk, skb, mss_now))) | 1012 | if (unlikely(tcp_fragment(sk, skb, mss_now, mss_now))) |
| 1014 | break; | 1013 | break; |
| 1015 | } | 1014 | } |
| 1016 | 1015 | ||
| @@ -1026,27 +1025,12 @@ static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle) | |||
| 1026 | 1025 | ||
| 1027 | tcp_minshall_update(tp, mss_now, skb); | 1026 | tcp_minshall_update(tp, mss_now, skb); |
| 1028 | sent_pkts++; | 1027 | sent_pkts++; |
| 1029 | |||
| 1030 | /* Do not optimize this to use tso_segs. If we chopped up | ||
| 1031 | * the packet above, tso_segs will no longer be valid. | ||
| 1032 | */ | ||
| 1033 | cwnd_quota -= tcp_skb_pcount(skb); | ||
| 1034 | |||
| 1035 | BUG_ON(cwnd_quota < 0); | ||
| 1036 | if (!cwnd_quota) | ||
| 1037 | break; | ||
| 1038 | |||
| 1039 | skb = sk->sk_send_head; | ||
| 1040 | if (!skb) | ||
| 1041 | break; | ||
| 1042 | tso_segs = tcp_init_tso_segs(sk, skb); | ||
| 1043 | } | 1028 | } |
| 1044 | 1029 | ||
| 1045 | if (likely(sent_pkts)) { | 1030 | if (likely(sent_pkts)) { |
| 1046 | tcp_cwnd_validate(sk, tp); | 1031 | tcp_cwnd_validate(sk, tp); |
| 1047 | return 0; | 1032 | return 0; |
| 1048 | } | 1033 | } |
| 1049 | out: | ||
| 1050 | return !tp->packets_out && sk->sk_send_head; | 1034 | return !tp->packets_out && sk->sk_send_head; |
| 1051 | } | 1035 | } |
| 1052 | 1036 | ||
| @@ -1076,7 +1060,7 @@ void tcp_push_one(struct sock *sk, unsigned int mss_now) | |||
| 1076 | 1060 | ||
| 1077 | BUG_ON(!skb || skb->len < mss_now); | 1061 | BUG_ON(!skb || skb->len < mss_now); |
| 1078 | 1062 | ||
| 1079 | tso_segs = tcp_init_tso_segs(sk, skb); | 1063 | tso_segs = tcp_init_tso_segs(sk, skb, mss_now); |
| 1080 | cwnd_quota = tcp_snd_test(sk, skb, mss_now, TCP_NAGLE_PUSH); | 1064 | cwnd_quota = tcp_snd_test(sk, skb, mss_now, TCP_NAGLE_PUSH); |
| 1081 | 1065 | ||
| 1082 | if (likely(cwnd_quota)) { | 1066 | if (likely(cwnd_quota)) { |
| @@ -1093,11 +1077,11 @@ void tcp_push_one(struct sock *sk, unsigned int mss_now) | |||
| 1093 | limit = skb->len - trim; | 1077 | limit = skb->len - trim; |
| 1094 | } | 1078 | } |
| 1095 | if (skb->len > limit) { | 1079 | if (skb->len > limit) { |
| 1096 | if (unlikely(tso_fragment(sk, skb, limit))) | 1080 | if (unlikely(tso_fragment(sk, skb, limit, mss_now))) |
| 1097 | return; | 1081 | return; |
| 1098 | } | 1082 | } |
| 1099 | } else if (unlikely(skb->len > mss_now)) { | 1083 | } else if (unlikely(skb->len > mss_now)) { |
| 1100 | if (unlikely(tcp_fragment(sk, skb, mss_now))) | 1084 | if (unlikely(tcp_fragment(sk, skb, mss_now, mss_now))) |
| 1101 | return; | 1085 | return; |
| 1102 | } | 1086 | } |
| 1103 | 1087 | ||
| @@ -1386,15 +1370,21 @@ int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb) | |||
| 1386 | 1370 | ||
| 1387 | if (skb->len > cur_mss) { | 1371 | if (skb->len > cur_mss) { |
| 1388 | int old_factor = tcp_skb_pcount(skb); | 1372 | int old_factor = tcp_skb_pcount(skb); |
| 1389 | int new_factor; | 1373 | int diff; |
| 1390 | 1374 | ||
| 1391 | if (tcp_fragment(sk, skb, cur_mss)) | 1375 | if (tcp_fragment(sk, skb, cur_mss, cur_mss)) |
| 1392 | return -ENOMEM; /* We'll try again later. */ | 1376 | return -ENOMEM; /* We'll try again later. */ |
| 1393 | 1377 | ||
| 1394 | /* New SKB created, account for it. */ | 1378 | /* New SKB created, account for it. */ |
| 1395 | new_factor = tcp_skb_pcount(skb); | 1379 | diff = old_factor - tcp_skb_pcount(skb) - |
| 1396 | tp->packets_out -= old_factor - new_factor; | 1380 | tcp_skb_pcount(skb->next); |
| 1397 | tp->packets_out += tcp_skb_pcount(skb->next); | 1381 | tp->packets_out -= diff; |
| 1382 | |||
| 1383 | if (diff > 0) { | ||
| 1384 | tp->fackets_out -= diff; | ||
| 1385 | if ((int)tp->fackets_out < 0) | ||
| 1386 | tp->fackets_out = 0; | ||
| 1387 | } | ||
| 1398 | } | 1388 | } |
| 1399 | 1389 | ||
| 1400 | /* Collapse two adjacent packets if worthwhile and we can. */ | 1390 | /* Collapse two adjacent packets if worthwhile and we can. */ |
| @@ -1991,7 +1981,7 @@ int tcp_write_wakeup(struct sock *sk) | |||
| 1991 | skb->len > mss) { | 1981 | skb->len > mss) { |
| 1992 | seg_size = min(seg_size, mss); | 1982 | seg_size = min(seg_size, mss); |
| 1993 | TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH; | 1983 | TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH; |
| 1994 | if (tcp_fragment(sk, skb, seg_size)) | 1984 | if (tcp_fragment(sk, skb, seg_size, mss)) |
| 1995 | return -1; | 1985 | return -1; |
| 1996 | /* SWS override triggered forced fragmentation. | 1986 | /* SWS override triggered forced fragmentation. |
| 1997 | * Disable TSO, the connection is too sick. */ | 1987 | * Disable TSO, the connection is too sick. */ |
| @@ -2000,7 +1990,7 @@ int tcp_write_wakeup(struct sock *sk) | |||
| 2000 | sk->sk_route_caps &= ~NETIF_F_TSO; | 1990 | sk->sk_route_caps &= ~NETIF_F_TSO; |
| 2001 | } | 1991 | } |
| 2002 | } else if (!tcp_skb_pcount(skb)) | 1992 | } else if (!tcp_skb_pcount(skb)) |
| 2003 | tcp_set_skb_tso_segs(sk, skb); | 1993 | tcp_set_skb_tso_segs(sk, skb, mss); |
| 2004 | 1994 | ||
| 2005 | TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH; | 1995 | TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH; |
| 2006 | TCP_SKB_CB(skb)->when = tcp_time_stamp; | 1996 | TCP_SKB_CB(skb)->when = tcp_time_stamp; |
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c index 7c24e64b443f..dc4d07357e3a 100644 --- a/net/ipv4/udp.c +++ b/net/ipv4/udp.c | |||
| @@ -628,7 +628,7 @@ back_from_confirm: | |||
| 628 | /* ... which is an evident application bug. --ANK */ | 628 | /* ... which is an evident application bug. --ANK */ |
| 629 | release_sock(sk); | 629 | release_sock(sk); |
| 630 | 630 | ||
| 631 | NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG "udp cork app bug 2\n")); | 631 | LIMIT_NETDEBUG(printk(KERN_DEBUG "udp cork app bug 2\n")); |
| 632 | err = -EINVAL; | 632 | err = -EINVAL; |
| 633 | goto out; | 633 | goto out; |
| 634 | } | 634 | } |
| @@ -693,7 +693,7 @@ static int udp_sendpage(struct sock *sk, struct page *page, int offset, | |||
| 693 | if (unlikely(!up->pending)) { | 693 | if (unlikely(!up->pending)) { |
| 694 | release_sock(sk); | 694 | release_sock(sk); |
| 695 | 695 | ||
| 696 | NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG "udp cork app bug 3\n")); | 696 | LIMIT_NETDEBUG(printk(KERN_DEBUG "udp cork app bug 3\n")); |
| 697 | return -EINVAL; | 697 | return -EINVAL; |
| 698 | } | 698 | } |
| 699 | 699 | ||
| @@ -1102,7 +1102,7 @@ static int udp_checksum_init(struct sk_buff *skb, struct udphdr *uh, | |||
| 1102 | skb->ip_summed = CHECKSUM_UNNECESSARY; | 1102 | skb->ip_summed = CHECKSUM_UNNECESSARY; |
| 1103 | if (!udp_check(uh, ulen, saddr, daddr, skb->csum)) | 1103 | if (!udp_check(uh, ulen, saddr, daddr, skb->csum)) |
| 1104 | return 0; | 1104 | return 0; |
| 1105 | NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG "udp v4 hw csum failure.\n")); | 1105 | LIMIT_NETDEBUG(printk(KERN_DEBUG "udp v4 hw csum failure.\n")); |
| 1106 | skb->ip_summed = CHECKSUM_NONE; | 1106 | skb->ip_summed = CHECKSUM_NONE; |
| 1107 | } | 1107 | } |
| 1108 | if (skb->ip_summed != CHECKSUM_UNNECESSARY) | 1108 | if (skb->ip_summed != CHECKSUM_UNNECESSARY) |
| @@ -1181,14 +1181,13 @@ int udp_rcv(struct sk_buff *skb) | |||
| 1181 | return(0); | 1181 | return(0); |
| 1182 | 1182 | ||
| 1183 | short_packet: | 1183 | short_packet: |
| 1184 | NETDEBUG(if (net_ratelimit()) | 1184 | LIMIT_NETDEBUG(printk(KERN_DEBUG "UDP: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n", |
| 1185 | printk(KERN_DEBUG "UDP: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n", | 1185 | NIPQUAD(saddr), |
| 1186 | NIPQUAD(saddr), | 1186 | ntohs(uh->source), |
| 1187 | ntohs(uh->source), | 1187 | ulen, |
| 1188 | ulen, | 1188 | len, |
| 1189 | len, | 1189 | NIPQUAD(daddr), |
| 1190 | NIPQUAD(daddr), | 1190 | ntohs(uh->dest))); |
| 1191 | ntohs(uh->dest))); | ||
| 1192 | no_header: | 1191 | no_header: |
| 1193 | UDP_INC_STATS_BH(UDP_MIB_INERRORS); | 1192 | UDP_INC_STATS_BH(UDP_MIB_INERRORS); |
| 1194 | kfree_skb(skb); | 1193 | kfree_skb(skb); |
| @@ -1199,13 +1198,12 @@ csum_error: | |||
| 1199 | * RFC1122: OK. Discards the bad packet silently (as far as | 1198 | * RFC1122: OK. Discards the bad packet silently (as far as |
| 1200 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | 1199 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). |
| 1201 | */ | 1200 | */ |
| 1202 | NETDEBUG(if (net_ratelimit()) | 1201 | LIMIT_NETDEBUG(printk(KERN_DEBUG "UDP: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n", |
| 1203 | printk(KERN_DEBUG "UDP: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n", | 1202 | NIPQUAD(saddr), |
| 1204 | NIPQUAD(saddr), | 1203 | ntohs(uh->source), |
| 1205 | ntohs(uh->source), | 1204 | NIPQUAD(daddr), |
| 1206 | NIPQUAD(daddr), | 1205 | ntohs(uh->dest), |
| 1207 | ntohs(uh->dest), | 1206 | ulen)); |
| 1208 | ulen)); | ||
| 1209 | drop: | 1207 | drop: |
| 1210 | UDP_INC_STATS_BH(UDP_MIB_INERRORS); | 1208 | UDP_INC_STATS_BH(UDP_MIB_INERRORS); |
| 1211 | kfree_skb(skb); | 1209 | kfree_skb(skb); |
