aboutsummaryrefslogtreecommitdiffstats
path: root/net
diff options
context:
space:
mode:
authorJeff Garzik <jgarzik@pobox.com>2005-12-12 15:24:45 -0500
committerJeff Garzik <jgarzik@pobox.com>2005-12-12 15:24:45 -0500
commitb1086eef813ecee09bd6b8ae364acf0fad065cba (patch)
treebc723bbdfc2898252e3fd8e14320d7fac58dca4b /net
parent003a20c81ec278595820d3829b544e90919f6f61 (diff)
parent49d7bc64283970ee83d2c954d04ba00d04e5943d (diff)
Merge branch 'master'
Diffstat (limited to 'net')
-rw-r--r--net/core/dev.c3
-rw-r--r--net/core/skbuff.c2
-rw-r--r--net/decnet/af_decnet.c25
-rw-r--r--net/decnet/sysctl_net_decnet.c33
-rw-r--r--net/ipv4/netfilter/Kconfig8
-rw-r--r--net/ipv4/netfilter/ip_conntrack_core.c20
-rw-r--r--net/ipv4/netfilter/ip_conntrack_netlink.c12
-rw-r--r--net/ipv4/netfilter/ip_conntrack_proto_tcp.c3
-rw-r--r--net/ipv4/tcp_output.c233
-rw-r--r--net/ipv4/tcp_vegas.c16
-rw-r--r--net/ipv6/esp6.c2
-rw-r--r--net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c12
-rw-r--r--net/netfilter/Kconfig4
-rw-r--r--net/netfilter/nf_conntrack_core.c3
-rw-r--r--net/netfilter/nfnetlink.c5
-rw-r--r--net/packet/af_packet.c115
16 files changed, 283 insertions, 213 deletions
diff --git a/net/core/dev.c b/net/core/dev.c
index 94e642ee6e2b..1a59f2173f0a 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -1113,7 +1113,8 @@ out:
1113void netdev_rx_csum_fault(struct net_device *dev) 1113void netdev_rx_csum_fault(struct net_device *dev)
1114{ 1114{
1115 if (net_ratelimit()) { 1115 if (net_ratelimit()) {
1116 printk(KERN_ERR "%s: hw csum failure.\n", dev->name); 1116 printk(KERN_ERR "%s: hw csum failure.\n",
1117 dev ? dev->name : "<unknown>");
1117 dump_stack(); 1118 dump_stack();
1118 } 1119 }
1119} 1120}
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index b7d13a4fff48..83fee37de38e 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -1725,7 +1725,7 @@ unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1725 * of the skb if any page alloc fails user this procedure returns -ENOMEM 1725 * of the skb if any page alloc fails user this procedure returns -ENOMEM
1726 */ 1726 */
1727int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb, 1727int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1728 int getfrag(void *from, char *to, int offset, 1728 int (*getfrag)(void *from, char *to, int offset,
1729 int len, int odd, struct sk_buff *skb), 1729 int len, int odd, struct sk_buff *skb),
1730 void *from, int length) 1730 void *from, int length)
1731{ 1731{
diff --git a/net/decnet/af_decnet.c b/net/decnet/af_decnet.c
index f89e55f814d9..d402e9020c68 100644
--- a/net/decnet/af_decnet.c
+++ b/net/decnet/af_decnet.c
@@ -153,6 +153,7 @@ static struct proto_ops dn_proto_ops;
153static DEFINE_RWLOCK(dn_hash_lock); 153static DEFINE_RWLOCK(dn_hash_lock);
154static struct hlist_head dn_sk_hash[DN_SK_HASH_SIZE]; 154static struct hlist_head dn_sk_hash[DN_SK_HASH_SIZE];
155static struct hlist_head dn_wild_sk; 155static struct hlist_head dn_wild_sk;
156static atomic_t decnet_memory_allocated;
156 157
157static int __dn_setsockopt(struct socket *sock, int level, int optname, char __user *optval, int optlen, int flags); 158static int __dn_setsockopt(struct socket *sock, int level, int optname, char __user *optval, int optlen, int flags);
158static int __dn_getsockopt(struct socket *sock, int level, int optname, char __user *optval, int __user *optlen, int flags); 159static int __dn_getsockopt(struct socket *sock, int level, int optname, char __user *optval, int __user *optlen, int flags);
@@ -446,10 +447,26 @@ static void dn_destruct(struct sock *sk)
446 dst_release(xchg(&sk->sk_dst_cache, NULL)); 447 dst_release(xchg(&sk->sk_dst_cache, NULL));
447} 448}
448 449
450static int dn_memory_pressure;
451
452static void dn_enter_memory_pressure(void)
453{
454 if (!dn_memory_pressure) {
455 dn_memory_pressure = 1;
456 }
457}
458
449static struct proto dn_proto = { 459static struct proto dn_proto = {
450 .name = "DECNET", 460 .name = "NSP",
451 .owner = THIS_MODULE, 461 .owner = THIS_MODULE,
452 .obj_size = sizeof(struct dn_sock), 462 .enter_memory_pressure = dn_enter_memory_pressure,
463 .memory_pressure = &dn_memory_pressure,
464 .memory_allocated = &decnet_memory_allocated,
465 .sysctl_mem = sysctl_decnet_mem,
466 .sysctl_wmem = sysctl_decnet_wmem,
467 .sysctl_rmem = sysctl_decnet_rmem,
468 .max_header = DN_MAX_NSP_DATA_HEADER + 64,
469 .obj_size = sizeof(struct dn_sock),
453}; 470};
454 471
455static struct sock *dn_alloc_sock(struct socket *sock, gfp_t gfp) 472static struct sock *dn_alloc_sock(struct socket *sock, gfp_t gfp)
@@ -470,6 +487,8 @@ static struct sock *dn_alloc_sock(struct socket *sock, gfp_t gfp)
470 sk->sk_family = PF_DECnet; 487 sk->sk_family = PF_DECnet;
471 sk->sk_protocol = 0; 488 sk->sk_protocol = 0;
472 sk->sk_allocation = gfp; 489 sk->sk_allocation = gfp;
490 sk->sk_sndbuf = sysctl_decnet_wmem[1];
491 sk->sk_rcvbuf = sysctl_decnet_rmem[1];
473 492
474 /* Initialization of DECnet Session Control Port */ 493 /* Initialization of DECnet Session Control Port */
475 scp = DN_SK(sk); 494 scp = DN_SK(sk);
diff --git a/net/decnet/sysctl_net_decnet.c b/net/decnet/sysctl_net_decnet.c
index 02bca49cb508..0e9d2c571165 100644
--- a/net/decnet/sysctl_net_decnet.c
+++ b/net/decnet/sysctl_net_decnet.c
@@ -10,6 +10,7 @@
10 * 10 *
11 * Changes: 11 * Changes:
12 * Steve Whitehouse - C99 changes and default device handling 12 * Steve Whitehouse - C99 changes and default device handling
13 * Steve Whitehouse - Memory buffer settings, like the tcp ones
13 * 14 *
14 */ 15 */
15#include <linux/config.h> 16#include <linux/config.h>
@@ -37,6 +38,11 @@ int decnet_dr_count = 3;
37int decnet_log_martians = 1; 38int decnet_log_martians = 1;
38int decnet_no_fc_max_cwnd = NSP_MIN_WINDOW; 39int decnet_no_fc_max_cwnd = NSP_MIN_WINDOW;
39 40
41/* Reasonable defaults, I hope, based on tcp's defaults */
42int sysctl_decnet_mem[3] = { 768 << 3, 1024 << 3, 1536 << 3 };
43int sysctl_decnet_wmem[3] = { 4 * 1024, 16 * 1024, 128 * 1024 };
44int sysctl_decnet_rmem[3] = { 4 * 1024, 87380, 87380 * 2 };
45
40#ifdef CONFIG_SYSCTL 46#ifdef CONFIG_SYSCTL
41extern int decnet_dst_gc_interval; 47extern int decnet_dst_gc_interval;
42static int min_decnet_time_wait[] = { 5 }; 48static int min_decnet_time_wait[] = { 5 };
@@ -428,6 +434,33 @@ static ctl_table dn_table[] = {
428 .extra1 = &min_decnet_no_fc_max_cwnd, 434 .extra1 = &min_decnet_no_fc_max_cwnd,
429 .extra2 = &max_decnet_no_fc_max_cwnd 435 .extra2 = &max_decnet_no_fc_max_cwnd
430 }, 436 },
437 {
438 .ctl_name = NET_DECNET_MEM,
439 .procname = "decnet_mem",
440 .data = &sysctl_decnet_mem,
441 .maxlen = sizeof(sysctl_decnet_mem),
442 .mode = 0644,
443 .proc_handler = &proc_dointvec,
444 .strategy = &sysctl_intvec,
445 },
446 {
447 .ctl_name = NET_DECNET_RMEM,
448 .procname = "decnet_rmem",
449 .data = &sysctl_decnet_rmem,
450 .maxlen = sizeof(sysctl_decnet_rmem),
451 .mode = 0644,
452 .proc_handler = &proc_dointvec,
453 .strategy = &sysctl_intvec,
454 },
455 {
456 .ctl_name = NET_DECNET_WMEM,
457 .procname = "decnet_wmem",
458 .data = &sysctl_decnet_wmem,
459 .maxlen = sizeof(sysctl_decnet_wmem),
460 .mode = 0644,
461 .proc_handler = &proc_dointvec,
462 .strategy = &sysctl_intvec,
463 },
431 { 464 {
432 .ctl_name = NET_DECNET_DEBUG_LEVEL, 465 .ctl_name = NET_DECNET_DEBUG_LEVEL,
433 .procname = "debug", 466 .procname = "debug",
diff --git a/net/ipv4/netfilter/Kconfig b/net/ipv4/netfilter/Kconfig
index 0bc00528d888..88a60650e6b8 100644
--- a/net/ipv4/netfilter/Kconfig
+++ b/net/ipv4/netfilter/Kconfig
@@ -56,8 +56,8 @@ config IP_NF_CONNTRACK_MARK
56 instead of the individual packets. 56 instead of the individual packets.
57 57
58config IP_NF_CONNTRACK_EVENTS 58config IP_NF_CONNTRACK_EVENTS
59 bool "Connection tracking events" 59 bool "Connection tracking events (EXPERIMENTAL)"
60 depends on IP_NF_CONNTRACK 60 depends on EXPERIMENTAL && IP_NF_CONNTRACK
61 help 61 help
62 If this option is enabled, the connection tracking code will 62 If this option is enabled, the connection tracking code will
63 provide a notifier chain that can be used by other kernel code 63 provide a notifier chain that can be used by other kernel code
@@ -66,8 +66,8 @@ config IP_NF_CONNTRACK_EVENTS
66 IF unsure, say `N'. 66 IF unsure, say `N'.
67 67
68config IP_NF_CONNTRACK_NETLINK 68config IP_NF_CONNTRACK_NETLINK
69 tristate 'Connection tracking netlink interface' 69 tristate 'Connection tracking netlink interface (EXPERIMENTAL)'
70 depends on IP_NF_CONNTRACK && NETFILTER_NETLINK 70 depends on EXPERIMENTAL && IP_NF_CONNTRACK && NETFILTER_NETLINK
71 depends on IP_NF_CONNTRACK!=y || NETFILTER_NETLINK!=m 71 depends on IP_NF_CONNTRACK!=y || NETFILTER_NETLINK!=m
72 help 72 help
73 This option enables support for a netlink-based userspace interface 73 This option enables support for a netlink-based userspace interface
diff --git a/net/ipv4/netfilter/ip_conntrack_core.c b/net/ipv4/netfilter/ip_conntrack_core.c
index 7a4ecddd597b..84c66dbfedaf 100644
--- a/net/ipv4/netfilter/ip_conntrack_core.c
+++ b/net/ipv4/netfilter/ip_conntrack_core.c
@@ -1345,6 +1345,11 @@ static int kill_all(struct ip_conntrack *i, void *data)
1345 return 1; 1345 return 1;
1346} 1346}
1347 1347
1348void ip_conntrack_flush(void)
1349{
1350 ip_ct_iterate_cleanup(kill_all, NULL);
1351}
1352
1348static void free_conntrack_hash(struct list_head *hash, int vmalloced,int size) 1353static void free_conntrack_hash(struct list_head *hash, int vmalloced,int size)
1349{ 1354{
1350 if (vmalloced) 1355 if (vmalloced)
@@ -1354,8 +1359,12 @@ static void free_conntrack_hash(struct list_head *hash, int vmalloced,int size)
1354 get_order(sizeof(struct list_head) * size)); 1359 get_order(sizeof(struct list_head) * size));
1355} 1360}
1356 1361
1357void ip_conntrack_flush(void) 1362/* Mishearing the voices in his head, our hero wonders how he's
1363 supposed to kill the mall. */
1364void ip_conntrack_cleanup(void)
1358{ 1365{
1366 ip_ct_attach = NULL;
1367
1359 /* This makes sure all current packets have passed through 1368 /* This makes sure all current packets have passed through
1360 netfilter framework. Roll on, two-stage module 1369 netfilter framework. Roll on, two-stage module
1361 delete... */ 1370 delete... */
@@ -1363,7 +1372,7 @@ void ip_conntrack_flush(void)
1363 1372
1364 ip_ct_event_cache_flush(); 1373 ip_ct_event_cache_flush();
1365 i_see_dead_people: 1374 i_see_dead_people:
1366 ip_ct_iterate_cleanup(kill_all, NULL); 1375 ip_conntrack_flush();
1367 if (atomic_read(&ip_conntrack_count) != 0) { 1376 if (atomic_read(&ip_conntrack_count) != 0) {
1368 schedule(); 1377 schedule();
1369 goto i_see_dead_people; 1378 goto i_see_dead_people;
@@ -1371,14 +1380,7 @@ void ip_conntrack_flush(void)
1371 /* wait until all references to ip_conntrack_untracked are dropped */ 1380 /* wait until all references to ip_conntrack_untracked are dropped */
1372 while (atomic_read(&ip_conntrack_untracked.ct_general.use) > 1) 1381 while (atomic_read(&ip_conntrack_untracked.ct_general.use) > 1)
1373 schedule(); 1382 schedule();
1374}
1375 1383
1376/* Mishearing the voices in his head, our hero wonders how he's
1377 supposed to kill the mall. */
1378void ip_conntrack_cleanup(void)
1379{
1380 ip_ct_attach = NULL;
1381 ip_conntrack_flush();
1382 kmem_cache_destroy(ip_conntrack_cachep); 1384 kmem_cache_destroy(ip_conntrack_cachep);
1383 kmem_cache_destroy(ip_conntrack_expect_cachep); 1385 kmem_cache_destroy(ip_conntrack_expect_cachep);
1384 free_conntrack_hash(ip_conntrack_hash, ip_conntrack_vmalloc, 1386 free_conntrack_hash(ip_conntrack_hash, ip_conntrack_vmalloc,
diff --git a/net/ipv4/netfilter/ip_conntrack_netlink.c b/net/ipv4/netfilter/ip_conntrack_netlink.c
index 3fce91bcc0ba..91fe8f2e38ff 100644
--- a/net/ipv4/netfilter/ip_conntrack_netlink.c
+++ b/net/ipv4/netfilter/ip_conntrack_netlink.c
@@ -503,7 +503,7 @@ ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
503} 503}
504 504
505static const size_t cta_min_proto[CTA_PROTO_MAX] = { 505static const size_t cta_min_proto[CTA_PROTO_MAX] = {
506 [CTA_PROTO_NUM-1] = sizeof(u_int16_t), 506 [CTA_PROTO_NUM-1] = sizeof(u_int8_t),
507 [CTA_PROTO_SRC_PORT-1] = sizeof(u_int16_t), 507 [CTA_PROTO_SRC_PORT-1] = sizeof(u_int16_t),
508 [CTA_PROTO_DST_PORT-1] = sizeof(u_int16_t), 508 [CTA_PROTO_DST_PORT-1] = sizeof(u_int16_t),
509 [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t), 509 [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
@@ -528,7 +528,7 @@ ctnetlink_parse_tuple_proto(struct nfattr *attr,
528 528
529 if (!tb[CTA_PROTO_NUM-1]) 529 if (!tb[CTA_PROTO_NUM-1])
530 return -EINVAL; 530 return -EINVAL;
531 tuple->dst.protonum = *(u_int16_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]); 531 tuple->dst.protonum = *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
532 532
533 proto = ip_conntrack_proto_find_get(tuple->dst.protonum); 533 proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
534 534
@@ -728,11 +728,9 @@ ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
728 return -ENOENT; 728 return -ENOENT;
729 } 729 }
730 } 730 }
731 if (del_timer(&ct->timeout)) { 731 if (del_timer(&ct->timeout))
732 ip_conntrack_put(ct);
733 ct->timeout.function((unsigned long)ct); 732 ct->timeout.function((unsigned long)ct);
734 return 0; 733
735 }
736 ip_conntrack_put(ct); 734 ip_conntrack_put(ct);
737 DEBUGP("leaving\n"); 735 DEBUGP("leaving\n");
738 736
@@ -877,7 +875,7 @@ ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
877 DEBUGP("NAT status: %lu\n", 875 DEBUGP("NAT status: %lu\n",
878 status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK)); 876 status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
879 877
880 if (ip_nat_initialized(ct, hooknum)) 878 if (ip_nat_initialized(ct, HOOK2MANIP(hooknum)))
881 return -EEXIST; 879 return -EEXIST;
882 ip_nat_setup_info(ct, &range, hooknum); 880 ip_nat_setup_info(ct, &range, hooknum);
883 881
diff --git a/net/ipv4/netfilter/ip_conntrack_proto_tcp.c b/net/ipv4/netfilter/ip_conntrack_proto_tcp.c
index aeb7353d4777..e7fa29e576dc 100644
--- a/net/ipv4/netfilter/ip_conntrack_proto_tcp.c
+++ b/net/ipv4/netfilter/ip_conntrack_proto_tcp.c
@@ -341,9 +341,10 @@ static int tcp_print_conntrack(struct seq_file *s,
341static int tcp_to_nfattr(struct sk_buff *skb, struct nfattr *nfa, 341static int tcp_to_nfattr(struct sk_buff *skb, struct nfattr *nfa,
342 const struct ip_conntrack *ct) 342 const struct ip_conntrack *ct)
343{ 343{
344 struct nfattr *nest_parms = NFA_NEST(skb, CTA_PROTOINFO_TCP); 344 struct nfattr *nest_parms;
345 345
346 read_lock_bh(&tcp_lock); 346 read_lock_bh(&tcp_lock);
347 nest_parms = NFA_NEST(skb, CTA_PROTOINFO_TCP);
347 NFA_PUT(skb, CTA_PROTOINFO_TCP_STATE, sizeof(u_int8_t), 348 NFA_PUT(skb, CTA_PROTOINFO_TCP_STATE, sizeof(u_int8_t),
348 &ct->proto.tcp.state); 349 &ct->proto.tcp.state);
349 read_unlock_bh(&tcp_lock); 350 read_unlock_bh(&tcp_lock);
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 029c70dfb585..b7325e0b406a 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -262,122 +262,139 @@ static __inline__ u16 tcp_select_window(struct sock *sk)
262 * We are working here with either a clone of the original 262 * We are working here with either a clone of the original
263 * SKB, or a fresh unique copy made by the retransmit engine. 263 * SKB, or a fresh unique copy made by the retransmit engine.
264 */ 264 */
265static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb) 265static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it, gfp_t gfp_mask)
266{ 266{
267 if (skb != NULL) { 267 const struct inet_connection_sock *icsk = inet_csk(sk);
268 const struct inet_connection_sock *icsk = inet_csk(sk); 268 struct inet_sock *inet;
269 struct inet_sock *inet = inet_sk(sk); 269 struct tcp_sock *tp;
270 struct tcp_sock *tp = tcp_sk(sk); 270 struct tcp_skb_cb *tcb;
271 struct tcp_skb_cb *tcb = TCP_SKB_CB(skb); 271 int tcp_header_size;
272 int tcp_header_size = tp->tcp_header_len; 272 struct tcphdr *th;
273 struct tcphdr *th; 273 int sysctl_flags;
274 int sysctl_flags; 274 int err;
275 int err; 275
276 BUG_ON(!skb || !tcp_skb_pcount(skb));
277
278 /* If congestion control is doing timestamping, we must
279 * take such a timestamp before we potentially clone/copy.
280 */
281 if (icsk->icsk_ca_ops->rtt_sample)
282 __net_timestamp(skb);
283
284 if (likely(clone_it)) {
285 if (unlikely(skb_cloned(skb)))
286 skb = pskb_copy(skb, gfp_mask);
287 else
288 skb = skb_clone(skb, gfp_mask);
289 if (unlikely(!skb))
290 return -ENOBUFS;
291 }
276 292
277 BUG_ON(!tcp_skb_pcount(skb)); 293 inet = inet_sk(sk);
294 tp = tcp_sk(sk);
295 tcb = TCP_SKB_CB(skb);
296 tcp_header_size = tp->tcp_header_len;
278 297
279#define SYSCTL_FLAG_TSTAMPS 0x1 298#define SYSCTL_FLAG_TSTAMPS 0x1
280#define SYSCTL_FLAG_WSCALE 0x2 299#define SYSCTL_FLAG_WSCALE 0x2
281#define SYSCTL_FLAG_SACK 0x4 300#define SYSCTL_FLAG_SACK 0x4
282 301
283 /* If congestion control is doing timestamping */ 302 sysctl_flags = 0;
284 if (icsk->icsk_ca_ops->rtt_sample) 303 if (unlikely(tcb->flags & TCPCB_FLAG_SYN)) {
285 __net_timestamp(skb); 304 tcp_header_size = sizeof(struct tcphdr) + TCPOLEN_MSS;
286 305 if(sysctl_tcp_timestamps) {
287 sysctl_flags = 0; 306 tcp_header_size += TCPOLEN_TSTAMP_ALIGNED;
288 if (tcb->flags & TCPCB_FLAG_SYN) { 307 sysctl_flags |= SYSCTL_FLAG_TSTAMPS;
289 tcp_header_size = sizeof(struct tcphdr) + TCPOLEN_MSS;
290 if(sysctl_tcp_timestamps) {
291 tcp_header_size += TCPOLEN_TSTAMP_ALIGNED;
292 sysctl_flags |= SYSCTL_FLAG_TSTAMPS;
293 }
294 if(sysctl_tcp_window_scaling) {
295 tcp_header_size += TCPOLEN_WSCALE_ALIGNED;
296 sysctl_flags |= SYSCTL_FLAG_WSCALE;
297 }
298 if(sysctl_tcp_sack) {
299 sysctl_flags |= SYSCTL_FLAG_SACK;
300 if(!(sysctl_flags & SYSCTL_FLAG_TSTAMPS))
301 tcp_header_size += TCPOLEN_SACKPERM_ALIGNED;
302 }
303 } else if (tp->rx_opt.eff_sacks) {
304 /* A SACK is 2 pad bytes, a 2 byte header, plus
305 * 2 32-bit sequence numbers for each SACK block.
306 */
307 tcp_header_size += (TCPOLEN_SACK_BASE_ALIGNED +
308 (tp->rx_opt.eff_sacks * TCPOLEN_SACK_PERBLOCK));
309 } 308 }
310 309 if (sysctl_tcp_window_scaling) {
311 if (tcp_packets_in_flight(tp) == 0) 310 tcp_header_size += TCPOLEN_WSCALE_ALIGNED;
312 tcp_ca_event(sk, CA_EVENT_TX_START); 311 sysctl_flags |= SYSCTL_FLAG_WSCALE;
313
314 th = (struct tcphdr *) skb_push(skb, tcp_header_size);
315 skb->h.th = th;
316 skb_set_owner_w(skb, sk);
317
318 /* Build TCP header and checksum it. */
319 th->source = inet->sport;
320 th->dest = inet->dport;
321 th->seq = htonl(tcb->seq);
322 th->ack_seq = htonl(tp->rcv_nxt);
323 *(((__u16 *)th) + 6) = htons(((tcp_header_size >> 2) << 12) | tcb->flags);
324 if (tcb->flags & TCPCB_FLAG_SYN) {
325 /* RFC1323: The window in SYN & SYN/ACK segments
326 * is never scaled.
327 */
328 th->window = htons(tp->rcv_wnd);
329 } else {
330 th->window = htons(tcp_select_window(sk));
331 } 312 }
332 th->check = 0; 313 if (sysctl_tcp_sack) {
333 th->urg_ptr = 0; 314 sysctl_flags |= SYSCTL_FLAG_SACK;
334 315 if (!(sysctl_flags & SYSCTL_FLAG_TSTAMPS))
335 if (tp->urg_mode && 316 tcp_header_size += TCPOLEN_SACKPERM_ALIGNED;
336 between(tp->snd_up, tcb->seq+1, tcb->seq+0xFFFF)) {
337 th->urg_ptr = htons(tp->snd_up-tcb->seq);
338 th->urg = 1;
339 } 317 }
318 } else if (unlikely(tp->rx_opt.eff_sacks)) {
319 /* A SACK is 2 pad bytes, a 2 byte header, plus
320 * 2 32-bit sequence numbers for each SACK block.
321 */
322 tcp_header_size += (TCPOLEN_SACK_BASE_ALIGNED +
323 (tp->rx_opt.eff_sacks *
324 TCPOLEN_SACK_PERBLOCK));
325 }
326
327 if (tcp_packets_in_flight(tp) == 0)
328 tcp_ca_event(sk, CA_EVENT_TX_START);
329
330 th = (struct tcphdr *) skb_push(skb, tcp_header_size);
331 skb->h.th = th;
332 skb_set_owner_w(skb, sk);
333
334 /* Build TCP header and checksum it. */
335 th->source = inet->sport;
336 th->dest = inet->dport;
337 th->seq = htonl(tcb->seq);
338 th->ack_seq = htonl(tp->rcv_nxt);
339 *(((__u16 *)th) + 6) = htons(((tcp_header_size >> 2) << 12) |
340 tcb->flags);
341
342 if (unlikely(tcb->flags & TCPCB_FLAG_SYN)) {
343 /* RFC1323: The window in SYN & SYN/ACK segments
344 * is never scaled.
345 */
346 th->window = htons(tp->rcv_wnd);
347 } else {
348 th->window = htons(tcp_select_window(sk));
349 }
350 th->check = 0;
351 th->urg_ptr = 0;
340 352
341 if (tcb->flags & TCPCB_FLAG_SYN) { 353 if (unlikely(tp->urg_mode &&
342 tcp_syn_build_options((__u32 *)(th + 1), 354 between(tp->snd_up, tcb->seq+1, tcb->seq+0xFFFF))) {
343 tcp_advertise_mss(sk), 355 th->urg_ptr = htons(tp->snd_up-tcb->seq);
344 (sysctl_flags & SYSCTL_FLAG_TSTAMPS), 356 th->urg = 1;
345 (sysctl_flags & SYSCTL_FLAG_SACK), 357 }
346 (sysctl_flags & SYSCTL_FLAG_WSCALE),
347 tp->rx_opt.rcv_wscale,
348 tcb->when,
349 tp->rx_opt.ts_recent);
350 } else {
351 tcp_build_and_update_options((__u32 *)(th + 1),
352 tp, tcb->when);
353 358
354 TCP_ECN_send(sk, tp, skb, tcp_header_size); 359 if (unlikely(tcb->flags & TCPCB_FLAG_SYN)) {
355 } 360 tcp_syn_build_options((__u32 *)(th + 1),
356 tp->af_specific->send_check(sk, th, skb->len, skb); 361 tcp_advertise_mss(sk),
362 (sysctl_flags & SYSCTL_FLAG_TSTAMPS),
363 (sysctl_flags & SYSCTL_FLAG_SACK),
364 (sysctl_flags & SYSCTL_FLAG_WSCALE),
365 tp->rx_opt.rcv_wscale,
366 tcb->when,
367 tp->rx_opt.ts_recent);
368 } else {
369 tcp_build_and_update_options((__u32 *)(th + 1),
370 tp, tcb->when);
371 TCP_ECN_send(sk, tp, skb, tcp_header_size);
372 }
357 373
358 if (tcb->flags & TCPCB_FLAG_ACK) 374 tp->af_specific->send_check(sk, th, skb->len, skb);
359 tcp_event_ack_sent(sk, tcp_skb_pcount(skb));
360 375
361 if (skb->len != tcp_header_size) 376 if (likely(tcb->flags & TCPCB_FLAG_ACK))
362 tcp_event_data_sent(tp, skb, sk); 377 tcp_event_ack_sent(sk, tcp_skb_pcount(skb));
363 378
364 TCP_INC_STATS(TCP_MIB_OUTSEGS); 379 if (skb->len != tcp_header_size)
380 tcp_event_data_sent(tp, skb, sk);
365 381
366 err = tp->af_specific->queue_xmit(skb, 0); 382 TCP_INC_STATS(TCP_MIB_OUTSEGS);
367 if (err <= 0)
368 return err;
369 383
370 tcp_enter_cwr(sk); 384 err = tp->af_specific->queue_xmit(skb, 0);
385 if (unlikely(err <= 0))
386 return err;
387
388 tcp_enter_cwr(sk);
389
390 /* NET_XMIT_CN is special. It does not guarantee,
391 * that this packet is lost. It tells that device
392 * is about to start to drop packets or already
393 * drops some packets of the same priority and
394 * invokes us to send less aggressively.
395 */
396 return err == NET_XMIT_CN ? 0 : err;
371 397
372 /* NET_XMIT_CN is special. It does not guarantee,
373 * that this packet is lost. It tells that device
374 * is about to start to drop packets or already
375 * drops some packets of the same priority and
376 * invokes us to send less aggressively.
377 */
378 return err == NET_XMIT_CN ? 0 : err;
379 }
380 return -ENOBUFS;
381#undef SYSCTL_FLAG_TSTAMPS 398#undef SYSCTL_FLAG_TSTAMPS
382#undef SYSCTL_FLAG_WSCALE 399#undef SYSCTL_FLAG_WSCALE
383#undef SYSCTL_FLAG_SACK 400#undef SYSCTL_FLAG_SACK
@@ -1036,7 +1053,7 @@ static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle)
1036 1053
1037 TCP_SKB_CB(skb)->when = tcp_time_stamp; 1054 TCP_SKB_CB(skb)->when = tcp_time_stamp;
1038 1055
1039 if (unlikely(tcp_transmit_skb(sk, skb_clone(skb, GFP_ATOMIC)))) 1056 if (unlikely(tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC)))
1040 break; 1057 break;
1041 1058
1042 /* Advance the send_head. This one is sent out. 1059 /* Advance the send_head. This one is sent out.
@@ -1109,7 +1126,7 @@ void tcp_push_one(struct sock *sk, unsigned int mss_now)
1109 /* Send it out now. */ 1126 /* Send it out now. */
1110 TCP_SKB_CB(skb)->when = tcp_time_stamp; 1127 TCP_SKB_CB(skb)->when = tcp_time_stamp;
1111 1128
1112 if (likely(!tcp_transmit_skb(sk, skb_clone(skb, sk->sk_allocation)))) { 1129 if (likely(!tcp_transmit_skb(sk, skb, 1, sk->sk_allocation))) {
1113 update_send_head(sk, tp, skb); 1130 update_send_head(sk, tp, skb);
1114 tcp_cwnd_validate(sk, tp); 1131 tcp_cwnd_validate(sk, tp);
1115 return; 1132 return;
@@ -1429,9 +1446,7 @@ int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb)
1429 */ 1446 */
1430 TCP_SKB_CB(skb)->when = tcp_time_stamp; 1447 TCP_SKB_CB(skb)->when = tcp_time_stamp;
1431 1448
1432 err = tcp_transmit_skb(sk, (skb_cloned(skb) ? 1449 err = tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
1433 pskb_copy(skb, GFP_ATOMIC):
1434 skb_clone(skb, GFP_ATOMIC)));
1435 1450
1436 if (err == 0) { 1451 if (err == 0) {
1437 /* Update global TCP statistics. */ 1452 /* Update global TCP statistics. */
@@ -1665,7 +1680,7 @@ void tcp_send_active_reset(struct sock *sk, gfp_t priority)
1665 TCP_SKB_CB(skb)->seq = tcp_acceptable_seq(sk, tp); 1680 TCP_SKB_CB(skb)->seq = tcp_acceptable_seq(sk, tp);
1666 TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq; 1681 TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;
1667 TCP_SKB_CB(skb)->when = tcp_time_stamp; 1682 TCP_SKB_CB(skb)->when = tcp_time_stamp;
1668 if (tcp_transmit_skb(sk, skb)) 1683 if (tcp_transmit_skb(sk, skb, 0, priority))
1669 NET_INC_STATS(LINUX_MIB_TCPABORTFAILED); 1684 NET_INC_STATS(LINUX_MIB_TCPABORTFAILED);
1670} 1685}
1671 1686
@@ -1700,7 +1715,7 @@ int tcp_send_synack(struct sock *sk)
1700 TCP_ECN_send_synack(tcp_sk(sk), skb); 1715 TCP_ECN_send_synack(tcp_sk(sk), skb);
1701 } 1716 }
1702 TCP_SKB_CB(skb)->when = tcp_time_stamp; 1717 TCP_SKB_CB(skb)->when = tcp_time_stamp;
1703 return tcp_transmit_skb(sk, skb_clone(skb, GFP_ATOMIC)); 1718 return tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
1704} 1719}
1705 1720
1706/* 1721/*
@@ -1861,7 +1876,7 @@ int tcp_connect(struct sock *sk)
1861 __skb_queue_tail(&sk->sk_write_queue, buff); 1876 __skb_queue_tail(&sk->sk_write_queue, buff);
1862 sk_charge_skb(sk, buff); 1877 sk_charge_skb(sk, buff);
1863 tp->packets_out += tcp_skb_pcount(buff); 1878 tp->packets_out += tcp_skb_pcount(buff);
1864 tcp_transmit_skb(sk, skb_clone(buff, GFP_KERNEL)); 1879 tcp_transmit_skb(sk, buff, 1, GFP_KERNEL);
1865 TCP_INC_STATS(TCP_MIB_ACTIVEOPENS); 1880 TCP_INC_STATS(TCP_MIB_ACTIVEOPENS);
1866 1881
1867 /* Timer for repeating the SYN until an answer. */ 1882 /* Timer for repeating the SYN until an answer. */
@@ -1957,7 +1972,7 @@ void tcp_send_ack(struct sock *sk)
1957 /* Send it off, this clears delayed acks for us. */ 1972 /* Send it off, this clears delayed acks for us. */
1958 TCP_SKB_CB(buff)->seq = TCP_SKB_CB(buff)->end_seq = tcp_acceptable_seq(sk, tp); 1973 TCP_SKB_CB(buff)->seq = TCP_SKB_CB(buff)->end_seq = tcp_acceptable_seq(sk, tp);
1959 TCP_SKB_CB(buff)->when = tcp_time_stamp; 1974 TCP_SKB_CB(buff)->when = tcp_time_stamp;
1960 tcp_transmit_skb(sk, buff); 1975 tcp_transmit_skb(sk, buff, 0, GFP_ATOMIC);
1961 } 1976 }
1962} 1977}
1963 1978
@@ -1997,7 +2012,7 @@ static int tcp_xmit_probe_skb(struct sock *sk, int urgent)
1997 TCP_SKB_CB(skb)->seq = urgent ? tp->snd_una : tp->snd_una - 1; 2012 TCP_SKB_CB(skb)->seq = urgent ? tp->snd_una : tp->snd_una - 1;
1998 TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq; 2013 TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;
1999 TCP_SKB_CB(skb)->when = tcp_time_stamp; 2014 TCP_SKB_CB(skb)->when = tcp_time_stamp;
2000 return tcp_transmit_skb(sk, skb); 2015 return tcp_transmit_skb(sk, skb, 0, GFP_ATOMIC);
2001} 2016}
2002 2017
2003int tcp_write_wakeup(struct sock *sk) 2018int tcp_write_wakeup(struct sock *sk)
@@ -2030,7 +2045,7 @@ int tcp_write_wakeup(struct sock *sk)
2030 2045
2031 TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH; 2046 TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;
2032 TCP_SKB_CB(skb)->when = tcp_time_stamp; 2047 TCP_SKB_CB(skb)->when = tcp_time_stamp;
2033 err = tcp_transmit_skb(sk, skb_clone(skb, GFP_ATOMIC)); 2048 err = tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
2034 if (!err) { 2049 if (!err) {
2035 update_send_head(sk, tp, skb); 2050 update_send_head(sk, tp, skb);
2036 } 2051 }
diff --git a/net/ipv4/tcp_vegas.c b/net/ipv4/tcp_vegas.c
index b7d296a8ac6d..13e7e6e8df16 100644
--- a/net/ipv4/tcp_vegas.c
+++ b/net/ipv4/tcp_vegas.c
@@ -215,14 +215,6 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack,
215 vegas->beg_snd_nxt = tp->snd_nxt; 215 vegas->beg_snd_nxt = tp->snd_nxt;
216 vegas->beg_snd_cwnd = tp->snd_cwnd; 216 vegas->beg_snd_cwnd = tp->snd_cwnd;
217 217
218 /* Take into account the current RTT sample too, to
219 * decrease the impact of delayed acks. This double counts
220 * this sample since we count it for the next window as well,
221 * but that's not too awful, since we're taking the min,
222 * rather than averaging.
223 */
224 tcp_vegas_rtt_calc(sk, seq_rtt * 1000);
225
226 /* We do the Vegas calculations only if we got enough RTT 218 /* We do the Vegas calculations only if we got enough RTT
227 * samples that we can be reasonably sure that we got 219 * samples that we can be reasonably sure that we got
228 * at least one RTT sample that wasn't from a delayed ACK. 220 * at least one RTT sample that wasn't from a delayed ACK.
@@ -333,11 +325,11 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack,
333 else if (tp->snd_cwnd > tp->snd_cwnd_clamp) 325 else if (tp->snd_cwnd > tp->snd_cwnd_clamp)
334 tp->snd_cwnd = tp->snd_cwnd_clamp; 326 tp->snd_cwnd = tp->snd_cwnd_clamp;
335 } 327 }
336 }
337 328
338 /* Wipe the slate clean for the next RTT. */ 329 /* Wipe the slate clean for the next RTT. */
339 vegas->cntRTT = 0; 330 vegas->cntRTT = 0;
340 vegas->minRTT = 0x7fffffff; 331 vegas->minRTT = 0x7fffffff;
332 }
341} 333}
342 334
343/* Extract info for Tcp socket info provided via netlink. */ 335/* Extract info for Tcp socket info provided via netlink. */
diff --git a/net/ipv6/esp6.c b/net/ipv6/esp6.c
index 40d9a1935ab5..8bfbe9970793 100644
--- a/net/ipv6/esp6.c
+++ b/net/ipv6/esp6.c
@@ -248,7 +248,7 @@ static u32 esp6_get_max_size(struct xfrm_state *x, int mtu)
248 if (esp->conf.padlen) 248 if (esp->conf.padlen)
249 mtu = ALIGN(mtu, esp->conf.padlen); 249 mtu = ALIGN(mtu, esp->conf.padlen);
250 250
251 return mtu + x->props.header_len + esp->auth.icv_full_len; 251 return mtu + x->props.header_len + esp->auth.icv_trunc_len;
252} 252}
253 253
254static void esp6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 254static void esp6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
diff --git a/net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c b/net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c
index c0f1da5497a9..a7e03cfacd06 100644
--- a/net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c
+++ b/net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c
@@ -68,8 +68,8 @@ static int icmpv6_invert_tuple(struct nf_conntrack_tuple *tuple,
68 [ICMPV6_NI_REPLY - 128] = ICMPV6_NI_REPLY +1 68 [ICMPV6_NI_REPLY - 128] = ICMPV6_NI_REPLY +1
69 }; 69 };
70 70
71 __u8 type = orig->dst.u.icmp.type - 128; 71 int type = orig->dst.u.icmp.type - 128;
72 if (type >= sizeof(invmap) || !invmap[type]) 72 if (type < 0 || type >= sizeof(invmap) || !invmap[type])
73 return 0; 73 return 0;
74 74
75 tuple->src.u.icmp.id = orig->src.u.icmp.id; 75 tuple->src.u.icmp.id = orig->src.u.icmp.id;
@@ -129,12 +129,12 @@ static int icmpv6_new(struct nf_conn *conntrack,
129 [ICMPV6_ECHO_REQUEST - 128] = 1, 129 [ICMPV6_ECHO_REQUEST - 128] = 1,
130 [ICMPV6_NI_QUERY - 128] = 1 130 [ICMPV6_NI_QUERY - 128] = 1
131 }; 131 };
132 int type = conntrack->tuplehash[0].tuple.dst.u.icmp.type - 128;
132 133
133 if (conntrack->tuplehash[0].tuple.dst.u.icmp.type - 128 >= sizeof(valid_new) 134 if (type < 0 || type >= sizeof(valid_new) || !valid_new[type]) {
134 || !valid_new[conntrack->tuplehash[0].tuple.dst.u.icmp.type - 128]) {
135 /* Can't create a new ICMPv6 `conn' with this. */ 135 /* Can't create a new ICMPv6 `conn' with this. */
136 DEBUGP("icmp: can't create new conn with type %u\n", 136 DEBUGP("icmpv6: can't create new conn with type %u\n",
137 conntrack->tuplehash[0].tuple.dst.u.icmp.type); 137 type + 128);
138 NF_CT_DUMP_TUPLE(&conntrack->tuplehash[0].tuple); 138 NF_CT_DUMP_TUPLE(&conntrack->tuplehash[0].tuple);
139 return 0; 139 return 0;
140 } 140 }
diff --git a/net/netfilter/Kconfig b/net/netfilter/Kconfig
index a84f9221e5f0..794c41d19b28 100644
--- a/net/netfilter/Kconfig
+++ b/net/netfilter/Kconfig
@@ -61,8 +61,8 @@ config NF_CONNTRACK_MARK
61 instead of the individual packets. 61 instead of the individual packets.
62 62
63config NF_CONNTRACK_EVENTS 63config NF_CONNTRACK_EVENTS
64 bool "Connection tracking events" 64 bool "Connection tracking events (EXPERIMENTAL)"
65 depends on NF_CONNTRACK 65 depends on EXPERIMENTAL && NF_CONNTRACK
66 help 66 help
67 If this option is enabled, the connection tracking code will 67 If this option is enabled, the connection tracking code will
68 provide a notifier chain that can be used by other kernel code 68 provide a notifier chain that can be used by other kernel code
diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c
index 1da678303d78..a7c7b490cf22 100644
--- a/net/netfilter/nf_conntrack_core.c
+++ b/net/netfilter/nf_conntrack_core.c
@@ -1383,6 +1383,9 @@ void nf_conntrack_cleanup(void)
1383 schedule(); 1383 schedule();
1384 goto i_see_dead_people; 1384 goto i_see_dead_people;
1385 } 1385 }
1386 /* wait until all references to nf_conntrack_untracked are dropped */
1387 while (atomic_read(&nf_conntrack_untracked.ct_general.use) > 1)
1388 schedule();
1386 1389
1387 for (i = 0; i < NF_CT_F_NUM; i++) { 1390 for (i = 0; i < NF_CT_F_NUM; i++) {
1388 if (nf_ct_cache[i].use == 0) 1391 if (nf_ct_cache[i].use == 0)
diff --git a/net/netfilter/nfnetlink.c b/net/netfilter/nfnetlink.c
index a60c59b97631..95fdf04f1d88 100644
--- a/net/netfilter/nfnetlink.c
+++ b/net/netfilter/nfnetlink.c
@@ -162,7 +162,7 @@ nfnetlink_check_attributes(struct nfnetlink_subsystem *subsys,
162 return -EINVAL; 162 return -EINVAL;
163 } 163 }
164 164
165 min_len = NLMSG_ALIGN(sizeof(struct nfgenmsg)); 165 min_len = NLMSG_SPACE(sizeof(struct nfgenmsg));
166 if (unlikely(nlh->nlmsg_len < min_len)) 166 if (unlikely(nlh->nlmsg_len < min_len))
167 return -EINVAL; 167 return -EINVAL;
168 168
@@ -236,8 +236,7 @@ static inline int nfnetlink_rcv_msg(struct sk_buff *skb,
236 } 236 }
237 237
238 /* All the messages must at least contain nfgenmsg */ 238 /* All the messages must at least contain nfgenmsg */
239 if (nlh->nlmsg_len < 239 if (nlh->nlmsg_len < NLMSG_SPACE(sizeof(struct nfgenmsg))) {
240 NLMSG_LENGTH(NLMSG_ALIGN(sizeof(struct nfgenmsg)))) {
241 DEBUGP("received message was too short\n"); 240 DEBUGP("received message was too short\n");
242 return 0; 241 return 0;
243 } 242 }
diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c
index 499ae3df4a44..3e2462760413 100644
--- a/net/packet/af_packet.c
+++ b/net/packet/af_packet.c
@@ -1587,23 +1587,47 @@ static inline struct page *pg_vec_endpage(char *one_pg_vec, unsigned int order)
1587 return virt_to_page(one_pg_vec + (PAGE_SIZE << order) - 1); 1587 return virt_to_page(one_pg_vec + (PAGE_SIZE << order) - 1);
1588} 1588}
1589 1589
1590static void free_pg_vec(char **pg_vec, unsigned order, unsigned len) 1590static void free_pg_vec(char **pg_vec, unsigned int order, unsigned int len)
1591{ 1591{
1592 int i; 1592 int i;
1593 1593
1594 for (i=0; i<len; i++) { 1594 for (i = 0; i < len; i++) {
1595 if (pg_vec[i]) { 1595 if (likely(pg_vec[i]))
1596 struct page *page, *pend; 1596 free_pages((unsigned long) pg_vec[i], order);
1597
1598 pend = pg_vec_endpage(pg_vec[i], order);
1599 for (page = virt_to_page(pg_vec[i]); page <= pend; page++)
1600 ClearPageReserved(page);
1601 free_pages((unsigned long)pg_vec[i], order);
1602 }
1603 } 1597 }
1604 kfree(pg_vec); 1598 kfree(pg_vec);
1605} 1599}
1606 1600
1601static inline char *alloc_one_pg_vec_page(unsigned long order)
1602{
1603 return (char *) __get_free_pages(GFP_KERNEL | __GFP_COMP | __GFP_ZERO,
1604 order);
1605}
1606
1607static char **alloc_pg_vec(struct tpacket_req *req, int order)
1608{
1609 unsigned int block_nr = req->tp_block_nr;
1610 char **pg_vec;
1611 int i;
1612
1613 pg_vec = kzalloc(block_nr * sizeof(char *), GFP_KERNEL);
1614 if (unlikely(!pg_vec))
1615 goto out;
1616
1617 for (i = 0; i < block_nr; i++) {
1618 pg_vec[i] = alloc_one_pg_vec_page(order);
1619 if (unlikely(!pg_vec[i]))
1620 goto out_free_pgvec;
1621 }
1622
1623out:
1624 return pg_vec;
1625
1626out_free_pgvec:
1627 free_pg_vec(pg_vec, order, block_nr);
1628 pg_vec = NULL;
1629 goto out;
1630}
1607 1631
1608static int packet_set_ring(struct sock *sk, struct tpacket_req *req, int closing) 1632static int packet_set_ring(struct sock *sk, struct tpacket_req *req, int closing)
1609{ 1633{
@@ -1617,64 +1641,46 @@ static int packet_set_ring(struct sock *sk, struct tpacket_req *req, int closing
1617 1641
1618 /* Sanity tests and some calculations */ 1642 /* Sanity tests and some calculations */
1619 1643
1620 if (po->pg_vec) 1644 if (unlikely(po->pg_vec))
1621 return -EBUSY; 1645 return -EBUSY;
1622 1646
1623 if ((int)req->tp_block_size <= 0) 1647 if (unlikely((int)req->tp_block_size <= 0))
1624 return -EINVAL; 1648 return -EINVAL;
1625 if (req->tp_block_size&(PAGE_SIZE-1)) 1649 if (unlikely(req->tp_block_size & (PAGE_SIZE - 1)))
1626 return -EINVAL; 1650 return -EINVAL;
1627 if (req->tp_frame_size < TPACKET_HDRLEN) 1651 if (unlikely(req->tp_frame_size < TPACKET_HDRLEN))
1628 return -EINVAL; 1652 return -EINVAL;
1629 if (req->tp_frame_size&(TPACKET_ALIGNMENT-1)) 1653 if (unlikely(req->tp_frame_size & (TPACKET_ALIGNMENT - 1)))
1630 return -EINVAL; 1654 return -EINVAL;
1631 1655
1632 po->frames_per_block = req->tp_block_size/req->tp_frame_size; 1656 po->frames_per_block = req->tp_block_size/req->tp_frame_size;
1633 if (po->frames_per_block <= 0) 1657 if (unlikely(po->frames_per_block <= 0))
1634 return -EINVAL; 1658 return -EINVAL;
1635 if (po->frames_per_block*req->tp_block_nr != req->tp_frame_nr) 1659 if (unlikely((po->frames_per_block * req->tp_block_nr) !=
1660 req->tp_frame_nr))
1636 return -EINVAL; 1661 return -EINVAL;
1637 /* OK! */
1638
1639 /* Allocate page vector */
1640 while ((PAGE_SIZE<<order) < req->tp_block_size)
1641 order++;
1642 1662
1643 err = -ENOMEM; 1663 err = -ENOMEM;
1644 1664 order = get_order(req->tp_block_size);
1645 pg_vec = kmalloc(req->tp_block_nr*sizeof(char *), GFP_KERNEL); 1665 pg_vec = alloc_pg_vec(req, order);
1646 if (pg_vec == NULL) 1666 if (unlikely(!pg_vec))
1647 goto out; 1667 goto out;
1648 memset(pg_vec, 0, req->tp_block_nr*sizeof(char **));
1649
1650 for (i=0; i<req->tp_block_nr; i++) {
1651 struct page *page, *pend;
1652 pg_vec[i] = (char *)__get_free_pages(GFP_KERNEL, order);
1653 if (!pg_vec[i])
1654 goto out_free_pgvec;
1655
1656 pend = pg_vec_endpage(pg_vec[i], order);
1657 for (page = virt_to_page(pg_vec[i]); page <= pend; page++)
1658 SetPageReserved(page);
1659 }
1660 /* Page vector is allocated */
1661 1668
1662 l = 0; 1669 l = 0;
1663 for (i=0; i<req->tp_block_nr; i++) { 1670 for (i = 0; i < req->tp_block_nr; i++) {
1664 char *ptr = pg_vec[i]; 1671 char *ptr = pg_vec[i];
1665 struct tpacket_hdr *header; 1672 struct tpacket_hdr *header;
1666 int k; 1673 int k;
1667 1674
1668 for (k=0; k<po->frames_per_block; k++) { 1675 for (k = 0; k < po->frames_per_block; k++) {
1669 1676 header = (struct tpacket_hdr *) ptr;
1670 header = (struct tpacket_hdr*)ptr;
1671 header->tp_status = TP_STATUS_KERNEL; 1677 header->tp_status = TP_STATUS_KERNEL;
1672 ptr += req->tp_frame_size; 1678 ptr += req->tp_frame_size;
1673 } 1679 }
1674 } 1680 }
1675 /* Done */ 1681 /* Done */
1676 } else { 1682 } else {
1677 if (req->tp_frame_nr) 1683 if (unlikely(req->tp_frame_nr))
1678 return -EINVAL; 1684 return -EINVAL;
1679 } 1685 }
1680 1686
@@ -1701,7 +1707,7 @@ static int packet_set_ring(struct sock *sk, struct tpacket_req *req, int closing
1701 1707
1702 spin_lock_bh(&sk->sk_receive_queue.lock); 1708 spin_lock_bh(&sk->sk_receive_queue.lock);
1703 pg_vec = XC(po->pg_vec, pg_vec); 1709 pg_vec = XC(po->pg_vec, pg_vec);
1704 po->frame_max = req->tp_frame_nr-1; 1710 po->frame_max = (req->tp_frame_nr - 1);
1705 po->head = 0; 1711 po->head = 0;
1706 po->frame_size = req->tp_frame_size; 1712 po->frame_size = req->tp_frame_size;
1707 spin_unlock_bh(&sk->sk_receive_queue.lock); 1713 spin_unlock_bh(&sk->sk_receive_queue.lock);
@@ -1728,7 +1734,6 @@ static int packet_set_ring(struct sock *sk, struct tpacket_req *req, int closing
1728 1734
1729 release_sock(sk); 1735 release_sock(sk);
1730 1736
1731out_free_pgvec:
1732 if (pg_vec) 1737 if (pg_vec)
1733 free_pg_vec(pg_vec, order, req->tp_block_nr); 1738 free_pg_vec(pg_vec, order, req->tp_block_nr);
1734out: 1739out:
@@ -1755,17 +1760,19 @@ static int packet_mmap(struct file *file, struct socket *sock, struct vm_area_st
1755 if (size != po->pg_vec_len*po->pg_vec_pages*PAGE_SIZE) 1760 if (size != po->pg_vec_len*po->pg_vec_pages*PAGE_SIZE)
1756 goto out; 1761 goto out;
1757 1762
1758 atomic_inc(&po->mapped);
1759 start = vma->vm_start; 1763 start = vma->vm_start;
1760 err = -EAGAIN; 1764 for (i = 0; i < po->pg_vec_len; i++) {
1761 for (i=0; i<po->pg_vec_len; i++) { 1765 struct page *page = virt_to_page(po->pg_vec[i]);
1762 if (remap_pfn_range(vma, start, 1766 int pg_num;
1763 __pa(po->pg_vec[i]) >> PAGE_SHIFT, 1767
1764 po->pg_vec_pages*PAGE_SIZE, 1768 for (pg_num = 0; pg_num < po->pg_vec_pages; pg_num++, page++) {
1765 vma->vm_page_prot)) 1769 err = vm_insert_page(vma, start, page);
1766 goto out; 1770 if (unlikely(err))
1767 start += po->pg_vec_pages*PAGE_SIZE; 1771 goto out;
1772 start += PAGE_SIZE;
1773 }
1768 } 1774 }
1775 atomic_inc(&po->mapped);
1769 vma->vm_ops = &packet_mmap_ops; 1776 vma->vm_ops = &packet_mmap_ops;
1770 err = 0; 1777 err = 0;
1771 1778
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-08 13:54:26 -0500