API tools faq
paste
Guest User

Untitled

a guest
Mar 1st, 2018
169
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 124.81 KB | None | 0 0
raw download clone embed print report
  1. diff -u linux-2.6.20.y/net/ipv4/ipvs/Kconfig linux-2.6.27.y/net/ipv4/ipvs/Kconfig
  2. --- linux-2.6.20.y/net/ipv4/ipvs/Kconfig 2009-02-16 11:57:22.000000000 -0400
  3. +++ linux-2.6.27.y/net/ipv4/ipvs/Kconfig 2009-02-16 12:56:22.000000000 -0400
  4. @@ -1,10 +1,7 @@
  5. #
  6. # IP Virtual Server configuration
  7. #
  8. -menu "IP: Virtual Server Configuration"
  9. - depends on NETFILTER
  10. -
  11. -config IP_VS
  12. +menuconfig IP_VS
  13. tristate "IP virtual server support (EXPERIMENTAL)"
  14. depends on NETFILTER
  15. ---help---
  16. @@ -25,9 +22,10 @@
  17. If you want to compile it in kernel, say Y. To compile it as a
  18. module, choose M here. If unsure, say N.
  19.  
  20. +if IP_VS
  21. +
  22. config IP_VS_DEBUG
  23. bool "IP virtual server debugging"
  24. - depends on IP_VS
  25. ---help---
  26. Say Y here if you want to get additional messages useful in
  27. debugging the IP virtual server code. You can change the debug
  28. @@ -35,7 +33,6 @@
  29.  
  30. config IP_VS_TAB_BITS
  31. int "IPVS connection table size (the Nth power of 2)"
  32. - depends on IP_VS
  33. default "12"
  34. ---help---
  35. The IPVS connection hash table uses the chaining scheme to handle
  36. @@ -61,42 +58,35 @@
  37. needed for your box.
  38.  
  39. comment "IPVS transport protocol load balancing support"
  40. - depends on IP_VS
  41.  
  42. config IP_VS_PROTO_TCP
  43. bool "TCP load balancing support"
  44. - depends on IP_VS
  45. ---help---
  46. This option enables support for load balancing TCP transport
  47. protocol. Say Y if unsure.
  48.  
  49. config IP_VS_PROTO_UDP
  50. bool "UDP load balancing support"
  51. - depends on IP_VS
  52. ---help---
  53. This option enables support for load balancing UDP transport
  54. protocol. Say Y if unsure.
  55.  
  56. config IP_VS_PROTO_ESP
  57. bool "ESP load balancing support"
  58. - depends on IP_VS
  59. ---help---
  60. This option enables support for load balancing ESP (Encapsulation
  61. Security Payload) transport protocol. Say Y if unsure.
  62.  
  63. config IP_VS_PROTO_AH
  64. bool "AH load balancing support"
  65. - depends on IP_VS
  66. ---help---
  67. This option enables support for load balancing AH (Authentication
  68. Header) transport protocol. Say Y if unsure.
  69.  
  70. comment "IPVS scheduler"
  71. - depends on IP_VS
  72.  
  73. config IP_VS_RR
  74. tristate "round-robin scheduling"
  75. - depends on IP_VS
  76. ---help---
  77. The robin-robin scheduling algorithm simply directs network
  78. connections to different real servers in a round-robin manner.
  79. @@ -106,7 +96,6 @@
  80.  
  81. config IP_VS_WRR
  82. tristate "weighted round-robin scheduling"
  83. - depends on IP_VS
  84. ---help---
  85. The weighted robin-robin scheduling algorithm directs network
  86. connections to different real servers based on server weights
  87. @@ -120,7 +109,6 @@
  88.  
  89. config IP_VS_LC
  90. tristate "least-connection scheduling"
  91. - depends on IP_VS
  92. ---help---
  93. The least-connection scheduling algorithm directs network
  94. connections to the server with the least number of active
  95. @@ -131,7 +119,6 @@
  96.  
  97. config IP_VS_WLC
  98. tristate "weighted least-connection scheduling"
  99. - depends on IP_VS
  100. ---help---
  101. The weighted least-connection scheduling algorithm directs network
  102. connections to the server with the least active connections
  103. @@ -142,7 +129,6 @@
  104.  
  105. config IP_VS_LBLC
  106. tristate "locality-based least-connection scheduling"
  107. - depends on IP_VS
  108. ---help---
  109. The locality-based least-connection scheduling algorithm is for
  110. destination IP load balancing. It is usually used in cache cluster.
  111. @@ -157,7 +143,6 @@
  112.  
  113. config IP_VS_LBLCR
  114. tristate "locality-based least-connection with replication scheduling"
  115. - depends on IP_VS
  116. ---help---
  117. The locality-based least-connection with replication scheduling
  118. algorithm is also for destination IP load balancing. It is
  119. @@ -176,7 +161,6 @@
  120.  
  121. config IP_VS_DH
  122. tristate "destination hashing scheduling"
  123. - depends on IP_VS
  124. ---help---
  125. The destination hashing scheduling algorithm assigns network
  126. connections to the servers through looking up a statically assigned
  127. @@ -187,7 +171,6 @@
  128.  
  129. config IP_VS_SH
  130. tristate "source hashing scheduling"
  131. - depends on IP_VS
  132. ---help---
  133. The source hashing scheduling algorithm assigns network
  134. connections to the servers through looking up a statically assigned
  135. @@ -198,7 +181,6 @@
  136.  
  137. config IP_VS_SED
  138. tristate "shortest expected delay scheduling"
  139. - depends on IP_VS
  140. ---help---
  141. The shortest expected delay scheduling algorithm assigns network
  142. connections to the server with the shortest expected delay. The
  143. @@ -212,7 +194,6 @@
  144.  
  145. config IP_VS_NQ
  146. tristate "never queue scheduling"
  147. - depends on IP_VS
  148. ---help---
  149. The never queue scheduling algorithm adopts a two-speed model.
  150. When there is an idle server available, the job will be sent to
  151. @@ -225,11 +206,10 @@
  152. module, choose M here. If unsure, say N.
  153.  
  154. comment 'IPVS application helper'
  155. - depends on IP_VS
  156.  
  157. config IP_VS_FTP
  158. tristate "FTP protocol helper"
  159. - depends on IP_VS && IP_VS_PROTO_TCP
  160. + depends on IP_VS_PROTO_TCP
  161. ---help---
  162. FTP is a protocol that transfers IP address and/or port number in
  163. the payload. In the virtual server via Network Address Translation,
  164. @@ -241,4 +221,4 @@
  165. If you want to compile it in kernel, say Y. To compile it as a
  166. module, choose M here. If unsure, say N.
  167.  
  168. -endmenu
  169. +endif # IP_VS
  170. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_app.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_app.c
  171. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_app.c 2009-02-16 11:57:22.000000000 -0400
  172. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_app.c 2009-02-16 12:56:22.000000000 -0400
  173. @@ -1,8 +1,6 @@
  174. /*
  175. * ip_vs_app.c: Application module support for IPVS
  176. *
  177. - * Version: $Id: ip_vs_app.c,v 1.17 2003/03/22 06:31:21 wensong Exp $
  178. - *
  179. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  180. *
  181. * This program is free software; you can redistribute it and/or
  182. @@ -25,6 +23,8 @@
  183. #include <linux/skbuff.h>
  184. #include <linux/in.h>
  185. #include <linux/ip.h>
  186. +#include <linux/netfilter.h>
  187. +#include <net/net_namespace.h>
  188. #include <net/protocol.h>
  189. #include <net/tcp.h>
  190. #include <asm/system.h>
  191. @@ -49,18 +49,13 @@
  192. */
  193. static inline int ip_vs_app_get(struct ip_vs_app *app)
  194. {
  195. - /* test and get the module atomically */
  196. - if (app->module)
  197. - return try_module_get(app->module);
  198. - else
  199. - return 1;
  200. + return try_module_get(app->module);
  201. }
  202.  
  203.  
  204. static inline void ip_vs_app_put(struct ip_vs_app *app)
  205. {
  206. - if (app->module)
  207. - module_put(app->module);
  208. + module_put(app->module);
  209. }
  210.  
  211.  
  212. @@ -327,18 +322,18 @@
  213. spin_unlock(&cp->lock);
  214. }
  215.  
  216. -static inline int app_tcp_pkt_out(struct ip_vs_conn *cp, struct sk_buff **pskb,
  217. +static inline int app_tcp_pkt_out(struct ip_vs_conn *cp, struct sk_buff *skb,
  218. struct ip_vs_app *app)
  219. {
  220. int diff;
  221. - unsigned int tcp_offset = (*pskb)->nh.iph->ihl*4;
  222. + const unsigned int tcp_offset = ip_hdrlen(skb);
  223. struct tcphdr *th;
  224. __u32 seq;
  225.  
  226. - if (!ip_vs_make_skb_writable(pskb, tcp_offset + sizeof(*th)))
  227. + if (!skb_make_writable(skb, tcp_offset + sizeof(*th)))
  228. return 0;
  229.  
  230. - th = (struct tcphdr *)((*pskb)->nh.raw + tcp_offset);
  231. + th = (struct tcphdr *)(skb_network_header(skb) + tcp_offset);
  232.  
  233. /*
  234. * Remember seq number in case this pkt gets resized
  235. @@ -359,7 +354,7 @@
  236. if (app->pkt_out == NULL)
  237. return 1;
  238.  
  239. - if (!app->pkt_out(app, cp, pskb, &diff))
  240. + if (!app->pkt_out(app, cp, skb, &diff))
  241. return 0;
  242.  
  243. /*
  244. @@ -377,7 +372,7 @@
  245. * called by ipvs packet handler, assumes previously checked cp!=NULL
  246. * returns false if it can't handle packet (oom)
  247. */
  248. -int ip_vs_app_pkt_out(struct ip_vs_conn *cp, struct sk_buff **pskb)
  249. +int ip_vs_app_pkt_out(struct ip_vs_conn *cp, struct sk_buff *skb)
  250. {
  251. struct ip_vs_app *app;
  252.  
  253. @@ -390,7 +385,7 @@
  254.  
  255. /* TCP is complicated */
  256. if (cp->protocol == IPPROTO_TCP)
  257. - return app_tcp_pkt_out(cp, pskb, app);
  258. + return app_tcp_pkt_out(cp, skb, app);
  259.  
  260. /*
  261. * Call private output hook function
  262. @@ -398,22 +393,22 @@
  263. if (app->pkt_out == NULL)
  264. return 1;
  265.  
  266. - return app->pkt_out(app, cp, pskb, NULL);
  267. + return app->pkt_out(app, cp, skb, NULL);
  268. }
  269.  
  270.  
  271. -static inline int app_tcp_pkt_in(struct ip_vs_conn *cp, struct sk_buff **pskb,
  272. +static inline int app_tcp_pkt_in(struct ip_vs_conn *cp, struct sk_buff *skb,
  273. struct ip_vs_app *app)
  274. {
  275. int diff;
  276. - unsigned int tcp_offset = (*pskb)->nh.iph->ihl*4;
  277. + const unsigned int tcp_offset = ip_hdrlen(skb);
  278. struct tcphdr *th;
  279. __u32 seq;
  280.  
  281. - if (!ip_vs_make_skb_writable(pskb, tcp_offset + sizeof(*th)))
  282. + if (!skb_make_writable(skb, tcp_offset + sizeof(*th)))
  283. return 0;
  284.  
  285. - th = (struct tcphdr *)((*pskb)->nh.raw + tcp_offset);
  286. + th = (struct tcphdr *)(skb_network_header(skb) + tcp_offset);
  287.  
  288. /*
  289. * Remember seq number in case this pkt gets resized
  290. @@ -434,7 +429,7 @@
  291. if (app->pkt_in == NULL)
  292. return 1;
  293.  
  294. - if (!app->pkt_in(app, cp, pskb, &diff))
  295. + if (!app->pkt_in(app, cp, skb, &diff))
  296. return 0;
  297.  
  298. /*
  299. @@ -452,7 +447,7 @@
  300. * called by ipvs packet handler, assumes previously checked cp!=NULL.
  301. * returns false if can't handle packet (oom).
  302. */
  303. -int ip_vs_app_pkt_in(struct ip_vs_conn *cp, struct sk_buff **pskb)
  304. +int ip_vs_app_pkt_in(struct ip_vs_conn *cp, struct sk_buff *skb)
  305. {
  306. struct ip_vs_app *app;
  307.  
  308. @@ -465,7 +460,7 @@
  309.  
  310. /* TCP is complicated */
  311. if (cp->protocol == IPPROTO_TCP)
  312. - return app_tcp_pkt_in(cp, pskb, app);
  313. + return app_tcp_pkt_in(cp, skb, app);
  314.  
  315. /*
  316. * Call private input hook function
  317. @@ -473,7 +468,7 @@
  318. if (app->pkt_in == NULL)
  319. return 1;
  320.  
  321. - return app->pkt_in(app, cp, pskb, NULL);
  322. + return app->pkt_in(app, cp, skb, NULL);
  323. }
  324.  
  325.  
  326. @@ -549,7 +544,7 @@
  327. return 0;
  328. }
  329.  
  330. -static struct seq_operations ip_vs_app_seq_ops = {
  331. +static const struct seq_operations ip_vs_app_seq_ops = {
  332. .start = ip_vs_app_seq_start,
  333. .next = ip_vs_app_seq_next,
  334. .stop = ip_vs_app_seq_stop,
  335. @@ -561,7 +556,7 @@
  336. return seq_open(file, &ip_vs_app_seq_ops);
  337. }
  338.  
  339. -static struct file_operations ip_vs_app_fops = {
  340. +static const struct file_operations ip_vs_app_fops = {
  341. .owner = THIS_MODULE,
  342. .open = ip_vs_app_open,
  343. .read = seq_read,
  344. @@ -577,7 +572,6 @@
  345. int ip_vs_skb_replace(struct sk_buff *skb, gfp_t pri,
  346. char *o_buf, int o_len, char *n_buf, int n_len)
  347. {
  348. - struct iphdr *iph;
  349. int diff;
  350. int o_offset;
  351. int o_left;
  352. @@ -603,27 +597,26 @@
  353. skb_put(skb, diff);
  354. memmove(skb->data + o_offset + n_len,
  355. skb->data + o_offset + o_len, o_left);
  356. - memcpy(skb->data + o_offset, n_buf, n_len);
  357. + skb_copy_to_linear_data_offset(skb, o_offset, n_buf, n_len);
  358. }
  359.  
  360. /* must update the iph total length here */
  361. - iph = skb->nh.iph;
  362. - iph->tot_len = htons(skb->len);
  363. + ip_hdr(skb)->tot_len = htons(skb->len);
  364.  
  365. LeaveFunction(9);
  366. return 0;
  367. }
  368.  
  369.  
  370. -int ip_vs_app_init(void)
  371. +int __init ip_vs_app_init(void)
  372. {
  373. /* we will replace it with proc_net_ipvs_create() soon */
  374. - proc_net_fops_create("ip_vs_app", 0, &ip_vs_app_fops);
  375. + proc_net_fops_create(&init_net, "ip_vs_app", 0, &ip_vs_app_fops);
  376. return 0;
  377. }
  378.  
  379.  
  380. void ip_vs_app_cleanup(void)
  381. {
  382. - proc_net_remove("ip_vs_app");
  383. + proc_net_remove(&init_net, "ip_vs_app");
  384. }
  385. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_conn.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_conn.c
  386. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_conn.c 2009-02-16 11:57:22.000000000 -0400
  387. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_conn.c 2009-02-16 12:56:22.000000000 -0400
  388. @@ -5,8 +5,6 @@
  389. * high-performance and highly available server based on a
  390. * cluster of servers.
  391. *
  392. - * Version: $Id: ip_vs_conn.c,v 1.31 2003/04/18 09:03:16 wensong Exp $
  393. - *
  394. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  395. * Peter Kese <peter.kese@ijs.si>
  396. * Julian Anastasov <ja@ssi.bg>
  397. @@ -35,6 +33,7 @@
  398. #include <linux/jhash.h>
  399. #include <linux/random.h>
  400.  
  401. +#include <net/net_namespace.h>
  402. #include <net/ip_vs.h>
  403.  
  404.  
  405. @@ -392,7 +391,15 @@
  406. atomic_inc(&dest->refcnt);
  407.  
  408. /* Bind with the destination and its corresponding transmitter */
  409. - cp->flags |= atomic_read(&dest->conn_flags);
  410. + if ((cp->flags & IP_VS_CONN_F_SYNC) &&
  411. + (!(cp->flags & IP_VS_CONN_F_TEMPLATE)))
  412. + /* if the connection is not template and is created
  413. + * by sync, preserve the activity flag.
  414. + */
  415. + cp->flags |= atomic_read(&dest->conn_flags) &
  416. + (~IP_VS_CONN_F_INACTIVE);
  417. + else
  418. + cp->flags |= atomic_read(&dest->conn_flags);
  419. cp->dest = dest;
  420.  
  421. IP_VS_DBG(7, "Bind-dest %s c:%u.%u.%u.%u:%d v:%u.%u.%u.%u:%d "
  422. @@ -411,7 +418,11 @@
  423. /* It is a normal connection, so increase the inactive
  424. connection counter because it is in TCP SYNRECV
  425. state (inactive) or other protocol inacive state */
  426. - atomic_inc(&dest->inactconns);
  427. + if ((cp->flags & IP_VS_CONN_F_SYNC) &&
  428. + (!(cp->flags & IP_VS_CONN_F_INACTIVE)))
  429. + atomic_inc(&dest->activeconns);
  430. + else
  431. + atomic_inc(&dest->inactconns);
  432. } else {
  433. /* It is a persistent connection/template, so increase
  434. the peristent connection counter */
  435. @@ -425,6 +436,24 @@
  436.  
  437.  
  438. /*
  439. + * Check if there is a destination for the connection, if so
  440. + * bind the connection to the destination.
  441. + */
  442. +struct ip_vs_dest *ip_vs_try_bind_dest(struct ip_vs_conn *cp)
  443. +{
  444. + struct ip_vs_dest *dest;
  445. +
  446. + if ((cp) && (!cp->dest)) {
  447. + dest = ip_vs_find_dest(cp->daddr, cp->dport,
  448. + cp->vaddr, cp->vport, cp->protocol);
  449. + ip_vs_bind_dest(cp, dest);
  450. + return dest;
  451. + } else
  452. + return NULL;
  453. +}
  454. +
  455. +
  456. +/*
  457. * Unbind a connection entry with its VS destination
  458. * Called by the ip_vs_conn_expire function.
  459. */
  460. @@ -494,8 +523,8 @@
  461. * Checking the dest server status.
  462. */
  463. if ((dest == NULL) ||
  464. - !(dest->flags & IP_VS_DEST_F_AVAILABLE) ||
  465. - (sysctl_ip_vs_expire_quiescent_template &&
  466. + !(dest->flags & IP_VS_DEST_F_AVAILABLE) ||
  467. + (sysctl_ip_vs_expire_quiescent_template &&
  468. (atomic_read(&dest->weight) == 0))) {
  469. IP_VS_DBG(9, "check_template: dest not available for "
  470. "protocol %s s:%u.%u.%u.%u:%d v:%u.%u.%u.%u:%d "
  471. @@ -603,17 +632,14 @@
  472. struct ip_vs_conn *cp;
  473. struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
  474.  
  475. - cp = kmem_cache_alloc(ip_vs_conn_cachep, GFP_ATOMIC);
  476. + cp = kmem_cache_zalloc(ip_vs_conn_cachep, GFP_ATOMIC);
  477. if (cp == NULL) {
  478. IP_VS_ERR_RL("ip_vs_conn_new: no memory available.\n");
  479. return NULL;
  480. }
  481.  
  482. - memset(cp, 0, sizeof(*cp));
  483. INIT_LIST_HEAD(&cp->c_list);
  484. - init_timer(&cp->timer);
  485. - cp->timer.data = (unsigned long)cp;
  486. - cp->timer.function = ip_vs_conn_expire;
  487. + setup_timer(&cp->timer, ip_vs_conn_expire, (unsigned long)cp);
  488. cp->protocol = proto;
  489. cp->caddr = caddr;
  490. cp->cport = cport;
  491. @@ -667,7 +693,7 @@
  492. {
  493. int idx;
  494. struct ip_vs_conn *cp;
  495. -
  496. +
  497. for(idx = 0; idx < IP_VS_CONN_TAB_SIZE; idx++) {
  498. ct_read_lock_bh(idx);
  499. list_for_each_entry(cp, &ip_vs_conn_tab[idx], c_list) {
  500. @@ -695,7 +721,7 @@
  501. int idx;
  502.  
  503. ++*pos;
  504. - if (v == SEQ_START_TOKEN)
  505. + if (v == SEQ_START_TOKEN)
  506. return ip_vs_conn_array(seq, 0);
  507.  
  508. /* more on same hash chain? */
  509. @@ -710,7 +736,7 @@
  510. list_for_each_entry(cp, &ip_vs_conn_tab[idx], c_list) {
  511. seq->private = &ip_vs_conn_tab[idx];
  512. return cp;
  513. - }
  514. + }
  515. ct_read_unlock_bh(idx);
  516. }
  517. seq->private = NULL;
  518. @@ -746,7 +772,7 @@
  519. return 0;
  520. }
  521.  
  522. -static struct seq_operations ip_vs_conn_seq_ops = {
  523. +static const struct seq_operations ip_vs_conn_seq_ops = {
  524. .start = ip_vs_conn_seq_start,
  525. .next = ip_vs_conn_seq_next,
  526. .stop = ip_vs_conn_seq_stop,
  527. @@ -758,13 +784,64 @@
  528. return seq_open(file, &ip_vs_conn_seq_ops);
  529. }
  530.  
  531. -static struct file_operations ip_vs_conn_fops = {
  532. +static const struct file_operations ip_vs_conn_fops = {
  533. .owner = THIS_MODULE,
  534. .open = ip_vs_conn_open,
  535. .read = seq_read,
  536. .llseek = seq_lseek,
  537. .release = seq_release,
  538. };
  539. +
  540. +static const char *ip_vs_origin_name(unsigned flags)
  541. +{
  542. + if (flags & IP_VS_CONN_F_SYNC)
  543. + return "SYNC";
  544. + else
  545. + return "LOCAL";
  546. +}
  547. +
  548. +static int ip_vs_conn_sync_seq_show(struct seq_file *seq, void *v)
  549. +{
  550. +
  551. + if (v == SEQ_START_TOKEN)
  552. + seq_puts(seq,
  553. + "Pro FromIP FPrt ToIP TPrt DestIP DPrt State Origin Expires\n");
  554. + else {
  555. + const struct ip_vs_conn *cp = v;
  556. +
  557. + seq_printf(seq,
  558. + "%-3s %08X %04X %08X %04X %08X %04X %-11s %-6s %7lu\n",
  559. + ip_vs_proto_name(cp->protocol),
  560. + ntohl(cp->caddr), ntohs(cp->cport),
  561. + ntohl(cp->vaddr), ntohs(cp->vport),
  562. + ntohl(cp->daddr), ntohs(cp->dport),
  563. + ip_vs_state_name(cp->protocol, cp->state),
  564. + ip_vs_origin_name(cp->flags),
  565. + (cp->timer.expires-jiffies)/HZ);
  566. + }
  567. + return 0;
  568. +}
  569. +
  570. +static const struct seq_operations ip_vs_conn_sync_seq_ops = {
  571. + .start = ip_vs_conn_seq_start,
  572. + .next = ip_vs_conn_seq_next,
  573. + .stop = ip_vs_conn_seq_stop,
  574. + .show = ip_vs_conn_sync_seq_show,
  575. +};
  576. +
  577. +static int ip_vs_conn_sync_open(struct inode *inode, struct file *file)
  578. +{
  579. + return seq_open(file, &ip_vs_conn_sync_seq_ops);
  580. +}
  581. +
  582. +static const struct file_operations ip_vs_conn_sync_fops = {
  583. + .owner = THIS_MODULE,
  584. + .open = ip_vs_conn_sync_open,
  585. + .read = seq_read,
  586. + .llseek = seq_lseek,
  587. + .release = seq_release,
  588. +};
  589. +
  590. #endif
  591.  
  592.  
  593. @@ -888,7 +965,7 @@
  594. }
  595.  
  596.  
  597. -int ip_vs_conn_init(void)
  598. +int __init ip_vs_conn_init(void)
  599. {
  600. int idx;
  601.  
  602. @@ -902,7 +979,7 @@
  603. /* Allocate ip_vs_conn slab cache */
  604. ip_vs_conn_cachep = kmem_cache_create("ip_vs_conn",
  605. sizeof(struct ip_vs_conn), 0,
  606. - SLAB_HWCACHE_ALIGN, NULL, NULL);
  607. + SLAB_HWCACHE_ALIGN, NULL);
  608. if (!ip_vs_conn_cachep) {
  609. vfree(ip_vs_conn_tab);
  610. return -ENOMEM;
  611. @@ -923,7 +1000,8 @@
  612. rwlock_init(&__ip_vs_conntbl_lock_array[idx].l);
  613. }
  614.  
  615. - proc_net_fops_create("ip_vs_conn", 0, &ip_vs_conn_fops);
  616. + proc_net_fops_create(&init_net, "ip_vs_conn", 0, &ip_vs_conn_fops);
  617. + proc_net_fops_create(&init_net, "ip_vs_conn_sync", 0, &ip_vs_conn_sync_fops);
  618.  
  619. /* calculate the random value for connection hash */
  620. get_random_bytes(&ip_vs_conn_rnd, sizeof(ip_vs_conn_rnd));
  621. @@ -939,6 +1017,7 @@
  622.  
  623. /* Release the empty cache */
  624. kmem_cache_destroy(ip_vs_conn_cachep);
  625. - proc_net_remove("ip_vs_conn");
  626. + proc_net_remove(&init_net, "ip_vs_conn");
  627. + proc_net_remove(&init_net, "ip_vs_conn_sync");
  628. vfree(ip_vs_conn_tab);
  629. }
  630. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_core.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_core.c
  631. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_core.c 2009-02-16 11:57:22.000000000 -0400
  632. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_core.c 2009-02-16 12:56:22.000000000 -0400
  633. @@ -5,8 +5,6 @@
  634. * high-performance and highly available server based on a
  635. * cluster of servers.
  636. *
  637. - * Version: $Id: ip_vs_core.c,v 1.34 2003/05/10 03:05:23 wensong Exp $
  638. - *
  639. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  640. * Peter Kese <peter.kese@ijs.si>
  641. * Julian Anastasov <ja@ssi.bg>
  642. @@ -58,7 +56,6 @@
  643. #ifdef CONFIG_IP_VS_DEBUG
  644. EXPORT_SYMBOL(ip_vs_get_debug_level);
  645. #endif
  646. -EXPORT_SYMBOL(ip_vs_make_skb_writable);
  647.  
  648.  
  649. /* ID used in ICMP lookups */
  650. @@ -163,42 +160,6 @@
  651. }
  652.  
  653.  
  654. -int ip_vs_make_skb_writable(struct sk_buff **pskb, int writable_len)
  655. -{
  656. - struct sk_buff *skb = *pskb;
  657. -
  658. - /* skb is already used, better copy skb and its payload */
  659. - if (unlikely(skb_shared(skb) || skb->sk))
  660. - goto copy_skb;
  661. -
  662. - /* skb data is already used, copy it */
  663. - if (unlikely(skb_cloned(skb)))
  664. - goto copy_data;
  665. -
  666. - return pskb_may_pull(skb, writable_len);
  667. -
  668. - copy_data:
  669. - if (unlikely(writable_len > skb->len))
  670. - return 0;
  671. - return !pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
  672. -
  673. - copy_skb:
  674. - if (unlikely(writable_len > skb->len))
  675. - return 0;
  676. - skb = skb_copy(skb, GFP_ATOMIC);
  677. - if (!skb)
  678. - return 0;
  679. - BUG_ON(skb_is_nonlinear(skb));
  680. -
  681. - /* Rest of kernel will get very unhappy if we pass it a
  682. - suddenly-orphaned skbuff */
  683. - if ((*pskb)->sk)
  684. - skb_set_owner_w(skb, (*pskb)->sk);
  685. - kfree_skb(*pskb);
  686. - *pskb = skb;
  687. - return 1;
  688. -}
  689. -
  690. /*
  691. * IPVS persistent scheduling function
  692. * It creates a connection entry according to its template if exists,
  693. @@ -212,7 +173,7 @@
  694. __be16 ports[2])
  695. {
  696. struct ip_vs_conn *cp = NULL;
  697. - struct iphdr *iph = skb->nh.iph;
  698. + struct iphdr *iph = ip_hdr(skb);
  699. struct ip_vs_dest *dest;
  700. struct ip_vs_conn *ct;
  701. __be16 dport; /* destination port to forward */
  702. @@ -381,7 +342,7 @@
  703. ip_vs_schedule(struct ip_vs_service *svc, const struct sk_buff *skb)
  704. {
  705. struct ip_vs_conn *cp = NULL;
  706. - struct iphdr *iph = skb->nh.iph;
  707. + struct iphdr *iph = ip_hdr(skb);
  708. struct ip_vs_dest *dest;
  709. __be16 _ports[2], *pptr;
  710.  
  711. @@ -447,7 +408,7 @@
  712. struct ip_vs_protocol *pp)
  713. {
  714. __be16 _ports[2], *pptr;
  715. - struct iphdr *iph = skb->nh.iph;
  716. + struct iphdr *iph = ip_hdr(skb);
  717.  
  718. pptr = skb_header_pointer(skb, iph->ihl*4,
  719. sizeof(_ports), _ports);
  720. @@ -460,7 +421,7 @@
  721. and the destination is RTN_UNICAST (and not local), then create
  722. a cache_bypass connection entry */
  723. if (sysctl_ip_vs_cache_bypass && svc->fwmark
  724. - && (inet_addr_type(iph->daddr) == RTN_UNICAST)) {
  725. + && (inet_addr_type(&init_net, iph->daddr) == RTN_UNICAST)) {
  726. int ret, cs;
  727. struct ip_vs_conn *cp;
  728.  
  729. @@ -518,19 +479,19 @@
  730.  
  731.  
  732. /*
  733. - * It is hooked before NF_IP_PRI_NAT_SRC at the NF_IP_POST_ROUTING
  734. + * It is hooked before NF_IP_PRI_NAT_SRC at the NF_INET_POST_ROUTING
  735. * chain, and is used for VS/NAT.
  736. * It detects packets for VS/NAT connections and sends the packets
  737. * immediately. This can avoid that iptable_nat mangles the packets
  738. * for VS/NAT.
  739. */
  740. static unsigned int ip_vs_post_routing(unsigned int hooknum,
  741. - struct sk_buff **pskb,
  742. + struct sk_buff *skb,
  743. const struct net_device *in,
  744. const struct net_device *out,
  745. int (*okfn)(struct sk_buff *))
  746. {
  747. - if (!((*pskb)->ipvs_property))
  748. + if (!skb->ipvs_property)
  749. return NF_ACCEPT;
  750. /* The packet was sent from IPVS, exit this chain */
  751. return NF_STOP;
  752. @@ -541,13 +502,14 @@
  753. return csum_fold(skb_checksum(skb, offset, skb->len - offset, 0));
  754. }
  755.  
  756. -static inline struct sk_buff *
  757. -ip_vs_gather_frags(struct sk_buff *skb, u_int32_t user)
  758. +static inline int ip_vs_gather_frags(struct sk_buff *skb, u_int32_t user)
  759. {
  760. - skb = ip_defrag(skb, user);
  761. - if (skb)
  762. - ip_send_check(skb->nh.iph);
  763. - return skb;
  764. + int err = ip_defrag(skb, user);
  765. +
  766. + if (!err)
  767. + ip_send_check(ip_hdr(skb));
  768. +
  769. + return err;
  770. }
  771.  
  772. /*
  773. @@ -557,9 +519,10 @@
  774. void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
  775. struct ip_vs_conn *cp, int inout)
  776. {
  777. - struct iphdr *iph = skb->nh.iph;
  778. + struct iphdr *iph = ip_hdr(skb);
  779. unsigned int icmp_offset = iph->ihl*4;
  780. - struct icmphdr *icmph = (struct icmphdr *)(skb->nh.raw + icmp_offset);
  781. + struct icmphdr *icmph = (struct icmphdr *)(skb_network_header(skb) +
  782. + icmp_offset);
  783. struct iphdr *ciph = (struct iphdr *)(icmph + 1);
  784.  
  785. if (inout) {
  786. @@ -604,9 +567,8 @@
  787. * Currently handles error types - unreachable, quench, ttl exceeded.
  788. * (Only used in VS/NAT)
  789. */
  790. -static int ip_vs_out_icmp(struct sk_buff **pskb, int *related)
  791. +static int ip_vs_out_icmp(struct sk_buff *skb, int *related)
  792. {
  793. - struct sk_buff *skb = *pskb;
  794. struct iphdr *iph;
  795. struct icmphdr _icmph, *ic;
  796. struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */
  797. @@ -617,14 +579,12 @@
  798. *related = 1;
  799.  
  800. /* reassemble IP fragments */
  801. - if (skb->nh.iph->frag_off & __constant_htons(IP_MF|IP_OFFSET)) {
  802. - skb = ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT);
  803. - if (!skb)
  804. + if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) {
  805. + if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT))
  806. return NF_STOLEN;
  807. - *pskb = skb;
  808. }
  809.  
  810. - iph = skb->nh.iph;
  811. + iph = ip_hdr(skb);
  812. offset = ihl = iph->ihl * 4;
  813. ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph);
  814. if (ic == NULL)
  815. @@ -659,7 +619,7 @@
  816. return NF_ACCEPT;
  817.  
  818. /* Is the embedded protocol header present? */
  819. - if (unlikely(cih->frag_off & __constant_htons(IP_OFFSET) &&
  820. + if (unlikely(cih->frag_off & htons(IP_OFFSET) &&
  821. pp->dont_defrag))
  822. return NF_ACCEPT;
  823.  
  824. @@ -675,13 +635,12 @@
  825. verdict = NF_DROP;
  826.  
  827. if (IP_VS_FWD_METHOD(cp) != 0) {
  828. - IP_VS_ERR("shouldn't reach here, because the box is on the"
  829. + IP_VS_ERR("shouldn't reach here, because the box is on the "
  830. "half connection in the tun/dr module.\n");
  831. }
  832.  
  833. /* Ensure the checksum is correct */
  834. - if (skb->ip_summed != CHECKSUM_UNNECESSARY &&
  835. - ip_vs_checksum_complete(skb, ihl)) {
  836. + if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) {
  837. /* Failed checksum! */
  838. IP_VS_DBG(1, "Forward ICMP: failed checksum from %d.%d.%d.%d!\n",
  839. NIPQUAD(iph->saddr));
  840. @@ -690,9 +649,8 @@
  841.  
  842. if (IPPROTO_TCP == cih->protocol || IPPROTO_UDP == cih->protocol)
  843. offset += 2 * sizeof(__u16);
  844. - if (!ip_vs_make_skb_writable(pskb, offset))
  845. + if (!skb_make_writable(skb, offset))
  846. goto out;
  847. - skb = *pskb;
  848.  
  849. ip_vs_nat_icmp(skb, pp, cp, 1);
  850.  
  851. @@ -712,24 +670,22 @@
  852. {
  853. struct tcphdr _tcph, *th;
  854.  
  855. - th = skb_header_pointer(skb, skb->nh.iph->ihl * 4,
  856. - sizeof(_tcph), &_tcph);
  857. + th = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_tcph), &_tcph);
  858. if (th == NULL)
  859. return 0;
  860. return th->rst;
  861. }
  862.  
  863. /*
  864. - * It is hooked at the NF_IP_FORWARD chain, used only for VS/NAT.
  865. + * It is hooked at the NF_INET_FORWARD chain, used only for VS/NAT.
  866. * Check if outgoing packet belongs to the established ip_vs_conn,
  867. * rewrite addresses of the packet and send it on its way...
  868. */
  869. static unsigned int
  870. -ip_vs_out(unsigned int hooknum, struct sk_buff **pskb,
  871. +ip_vs_out(unsigned int hooknum, struct sk_buff *skb,
  872. const struct net_device *in, const struct net_device *out,
  873. int (*okfn)(struct sk_buff *))
  874. {
  875. - struct sk_buff *skb = *pskb;
  876. struct iphdr *iph;
  877. struct ip_vs_protocol *pp;
  878. struct ip_vs_conn *cp;
  879. @@ -740,14 +696,13 @@
  880. if (skb->ipvs_property)
  881. return NF_ACCEPT;
  882.  
  883. - iph = skb->nh.iph;
  884. + iph = ip_hdr(skb);
  885. if (unlikely(iph->protocol == IPPROTO_ICMP)) {
  886. - int related, verdict = ip_vs_out_icmp(pskb, &related);
  887. + int related, verdict = ip_vs_out_icmp(skb, &related);
  888.  
  889. if (related)
  890. return verdict;
  891. - skb = *pskb;
  892. - iph = skb->nh.iph;
  893. + iph = ip_hdr(skb);
  894. }
  895.  
  896. pp = ip_vs_proto_get(iph->protocol);
  897. @@ -755,13 +710,11 @@
  898. return NF_ACCEPT;
  899.  
  900. /* reassemble IP fragments */
  901. - if (unlikely(iph->frag_off & __constant_htons(IP_MF|IP_OFFSET) &&
  902. + if (unlikely(iph->frag_off & htons(IP_MF|IP_OFFSET) &&
  903. !pp->dont_defrag)) {
  904. - skb = ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT);
  905. - if (!skb)
  906. + if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT))
  907. return NF_STOLEN;
  908. - iph = skb->nh.iph;
  909. - *pskb = skb;
  910. + iph = ip_hdr(skb);
  911. }
  912.  
  913. ihl = iph->ihl << 2;
  914. @@ -803,25 +756,23 @@
  915.  
  916. IP_VS_DBG_PKT(11, pp, skb, 0, "Outgoing packet");
  917.  
  918. - if (!ip_vs_make_skb_writable(pskb, ihl))
  919. + if (!skb_make_writable(skb, ihl))
  920. goto drop;
  921.  
  922. /* mangle the packet */
  923. - if (pp->snat_handler && !pp->snat_handler(pskb, pp, cp))
  924. + if (pp->snat_handler && !pp->snat_handler(skb, pp, cp))
  925. + goto drop;
  926. + ip_hdr(skb)->saddr = cp->vaddr;
  927. + ip_send_check(ip_hdr(skb));
  928. +
  929. + /* For policy routing, packets originating from this
  930. + * machine itself may be routed differently to packets
  931. + * passing through. We want this packet to be routed as
  932. + * if it came from this machine itself. So re-compute
  933. + * the routing information.
  934. + */
  935. + if (ip_route_me_harder(skb, RTN_LOCAL) != 0)
  936. goto drop;
  937. - skb = *pskb;
  938. - skb->nh.iph->saddr = cp->vaddr;
  939. - ip_send_check(skb->nh.iph);
  940. -
  941. - /* For policy routing, packets originating from this
  942. - * machine itself may be routed differently to packets
  943. - * passing through. We want this packet to be routed as
  944. - * if it came from this machine itself. So re-compute
  945. - * the routing information.
  946. - */
  947. - if (ip_route_me_harder(pskb, RTN_LOCAL) != 0)
  948. - goto drop;
  949. - skb = *pskb;
  950.  
  951. IP_VS_DBG_PKT(10, pp, skb, 0, "After SNAT");
  952.  
  953. @@ -836,7 +787,7 @@
  954.  
  955. drop:
  956. ip_vs_conn_put(cp);
  957. - kfree_skb(*pskb);
  958. + kfree_skb(skb);
  959. return NF_STOLEN;
  960. }
  961.  
  962. @@ -847,10 +798,9 @@
  963. * forward to the right destination host if relevant.
  964. * Currently handles error types - unreachable, quench, ttl exceeded.
  965. */
  966. -static int
  967. -ip_vs_in_icmp(struct sk_buff **pskb, int *related, unsigned int hooknum)
  968. +static int
  969. +ip_vs_in_icmp(struct sk_buff *skb, int *related, unsigned int hooknum)
  970. {
  971. - struct sk_buff *skb = *pskb;
  972. struct iphdr *iph;
  973. struct icmphdr _icmph, *ic;
  974. struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */
  975. @@ -861,16 +811,13 @@
  976. *related = 1;
  977.  
  978. /* reassemble IP fragments */
  979. - if (skb->nh.iph->frag_off & __constant_htons(IP_MF|IP_OFFSET)) {
  980. - skb = ip_vs_gather_frags(skb,
  981. - hooknum == NF_IP_LOCAL_IN ?
  982. - IP_DEFRAG_VS_IN : IP_DEFRAG_VS_FWD);
  983. - if (!skb)
  984. + if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) {
  985. + if (ip_vs_gather_frags(skb, hooknum == NF_INET_LOCAL_IN ?
  986. + IP_DEFRAG_VS_IN : IP_DEFRAG_VS_FWD))
  987. return NF_STOLEN;
  988. - *pskb = skb;
  989. }
  990.  
  991. - iph = skb->nh.iph;
  992. + iph = ip_hdr(skb);
  993. offset = ihl = iph->ihl * 4;
  994. ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph);
  995. if (ic == NULL)
  996. @@ -905,7 +852,7 @@
  997. return NF_ACCEPT;
  998.  
  999. /* Is the embedded protocol header present? */
  1000. - if (unlikely(cih->frag_off & __constant_htons(IP_OFFSET) &&
  1001. + if (unlikely(cih->frag_off & htons(IP_OFFSET) &&
  1002. pp->dont_defrag))
  1003. return NF_ACCEPT;
  1004.  
  1005. @@ -921,8 +868,7 @@
  1006. verdict = NF_DROP;
  1007.  
  1008. /* Ensure the checksum is correct */
  1009. - if (skb->ip_summed != CHECKSUM_UNNECESSARY &&
  1010. - ip_vs_checksum_complete(skb, ihl)) {
  1011. + if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) {
  1012. /* Failed checksum! */
  1013. IP_VS_DBG(1, "Incoming ICMP: failed checksum from %d.%d.%d.%d!\n",
  1014. NIPQUAD(iph->saddr));
  1015. @@ -947,11 +893,10 @@
  1016. * and send it on its way...
  1017. */
  1018. static unsigned int
  1019. -ip_vs_in(unsigned int hooknum, struct sk_buff **pskb,
  1020. +ip_vs_in(unsigned int hooknum, struct sk_buff *skb,
  1021. const struct net_device *in, const struct net_device *out,
  1022. int (*okfn)(struct sk_buff *))
  1023. {
  1024. - struct sk_buff *skb = *pskb;
  1025. struct iphdr *iph;
  1026. struct ip_vs_protocol *pp;
  1027. struct ip_vs_conn *cp;
  1028. @@ -963,22 +908,21 @@
  1029. * ... don't know why 1st test DOES NOT include 2nd (?)
  1030. */
  1031. if (unlikely(skb->pkt_type != PACKET_HOST
  1032. - || skb->dev == &loopback_dev || skb->sk)) {
  1033. + || skb->dev->flags & IFF_LOOPBACK || skb->sk)) {
  1034. IP_VS_DBG(12, "packet type=%d proto=%d daddr=%d.%d.%d.%d ignored\n",
  1035. skb->pkt_type,
  1036. - skb->nh.iph->protocol,
  1037. - NIPQUAD(skb->nh.iph->daddr));
  1038. + ip_hdr(skb)->protocol,
  1039. + NIPQUAD(ip_hdr(skb)->daddr));
  1040. return NF_ACCEPT;
  1041. }
  1042.  
  1043. - iph = skb->nh.iph;
  1044. + iph = ip_hdr(skb);
  1045. if (unlikely(iph->protocol == IPPROTO_ICMP)) {
  1046. - int related, verdict = ip_vs_in_icmp(pskb, &related, hooknum);
  1047. + int related, verdict = ip_vs_in_icmp(skb, &related, hooknum);
  1048.  
  1049. if (related)
  1050. return verdict;
  1051. - skb = *pskb;
  1052. - iph = skb->nh.iph;
  1053. + iph = ip_hdr(skb);
  1054. }
  1055.  
  1056. /* Protocol supported? */
  1057. @@ -1033,15 +977,24 @@
  1058. ret = NF_ACCEPT;
  1059. }
  1060.  
  1061. - /* increase its packet counter and check if it is needed
  1062. - to be synchronized */
  1063. + /* Increase its packet counter and check if it is needed
  1064. + * to be synchronized
  1065. + *
  1066. + * Sync connection if it is about to close to
  1067. + * encorage the standby servers to update the connections timeout
  1068. + */
  1069. atomic_inc(&cp->in_pkts);
  1070. if ((ip_vs_sync_state & IP_VS_STATE_MASTER) &&
  1071. - (cp->protocol != IPPROTO_TCP ||
  1072. - cp->state == IP_VS_TCP_S_ESTABLISHED) &&
  1073. - (atomic_read(&cp->in_pkts) % sysctl_ip_vs_sync_threshold[1]
  1074. - == sysctl_ip_vs_sync_threshold[0]))
  1075. + (((cp->protocol != IPPROTO_TCP ||
  1076. + cp->state == IP_VS_TCP_S_ESTABLISHED) &&
  1077. + (atomic_read(&cp->in_pkts) % sysctl_ip_vs_sync_threshold[1]
  1078. + == sysctl_ip_vs_sync_threshold[0])) ||
  1079. + ((cp->protocol == IPPROTO_TCP) && (cp->old_state != cp->state) &&
  1080. + ((cp->state == IP_VS_TCP_S_FIN_WAIT) ||
  1081. + (cp->state == IP_VS_TCP_S_CLOSE_WAIT) ||
  1082. + (cp->state == IP_VS_TCP_S_TIME_WAIT)))))
  1083. ip_vs_sync_conn(cp);
  1084. + cp->old_state = cp->state;
  1085.  
  1086. ip_vs_conn_put(cp);
  1087. return ret;
  1088. @@ -1049,65 +1002,64 @@
  1089.  
  1090.  
  1091. /*
  1092. - * It is hooked at the NF_IP_FORWARD chain, in order to catch ICMP
  1093. + * It is hooked at the NF_INET_FORWARD chain, in order to catch ICMP
  1094. * related packets destined for 0.0.0.0/0.
  1095. * When fwmark-based virtual service is used, such as transparent
  1096. * cache cluster, TCP packets can be marked and routed to ip_vs_in,
  1097. * but ICMP destined for 0.0.0.0/0 cannot not be easily marked and
  1098. - * sent to ip_vs_in_icmp. So, catch them at the NF_IP_FORWARD chain
  1099. + * sent to ip_vs_in_icmp. So, catch them at the NF_INET_FORWARD chain
  1100. * and send them to ip_vs_in_icmp.
  1101. */
  1102. static unsigned int
  1103. -ip_vs_forward_icmp(unsigned int hooknum, struct sk_buff **pskb,
  1104. +ip_vs_forward_icmp(unsigned int hooknum, struct sk_buff *skb,
  1105. const struct net_device *in, const struct net_device *out,
  1106. int (*okfn)(struct sk_buff *))
  1107. {
  1108. int r;
  1109.  
  1110. - if ((*pskb)->nh.iph->protocol != IPPROTO_ICMP)
  1111. + if (ip_hdr(skb)->protocol != IPPROTO_ICMP)
  1112. return NF_ACCEPT;
  1113.  
  1114. - return ip_vs_in_icmp(pskb, &r, hooknum);
  1115. + return ip_vs_in_icmp(skb, &r, hooknum);
  1116. }
  1117.  
  1118.  
  1119. -/* After packet filtering, forward packet through VS/DR, VS/TUN,
  1120. - or VS/NAT(change destination), so that filtering rules can be
  1121. - applied to IPVS. */
  1122. -static struct nf_hook_ops ip_vs_in_ops = {
  1123. - .hook = ip_vs_in,
  1124. - .owner = THIS_MODULE,
  1125. - .pf = PF_INET,
  1126. - .hooknum = NF_IP_LOCAL_IN,
  1127. - .priority = 100,
  1128. -};
  1129. -
  1130. -/* After packet filtering, change source only for VS/NAT */
  1131. -static struct nf_hook_ops ip_vs_out_ops = {
  1132. - .hook = ip_vs_out,
  1133. - .owner = THIS_MODULE,
  1134. - .pf = PF_INET,
  1135. - .hooknum = NF_IP_FORWARD,
  1136. - .priority = 100,
  1137. -};
  1138. -
  1139. -/* After packet filtering (but before ip_vs_out_icmp), catch icmp
  1140. - destined for 0.0.0.0/0, which is for incoming IPVS connections */
  1141. -static struct nf_hook_ops ip_vs_forward_icmp_ops = {
  1142. - .hook = ip_vs_forward_icmp,
  1143. - .owner = THIS_MODULE,
  1144. - .pf = PF_INET,
  1145. - .hooknum = NF_IP_FORWARD,
  1146. - .priority = 99,
  1147. -};
  1148. -
  1149. -/* Before the netfilter connection tracking, exit from POST_ROUTING */
  1150. -static struct nf_hook_ops ip_vs_post_routing_ops = {
  1151. - .hook = ip_vs_post_routing,
  1152. - .owner = THIS_MODULE,
  1153. - .pf = PF_INET,
  1154. - .hooknum = NF_IP_POST_ROUTING,
  1155. - .priority = NF_IP_PRI_NAT_SRC-1,
  1156. +static struct nf_hook_ops ip_vs_ops[] __read_mostly = {
  1157. + /* After packet filtering, forward packet through VS/DR, VS/TUN,
  1158. + * or VS/NAT(change destination), so that filtering rules can be
  1159. + * applied to IPVS. */
  1160. + {
  1161. + .hook = ip_vs_in,
  1162. + .owner = THIS_MODULE,
  1163. + .pf = PF_INET,
  1164. + .hooknum = NF_INET_LOCAL_IN,
  1165. + .priority = 100,
  1166. + },
  1167. + /* After packet filtering, change source only for VS/NAT */
  1168. + {
  1169. + .hook = ip_vs_out,
  1170. + .owner = THIS_MODULE,
  1171. + .pf = PF_INET,
  1172. + .hooknum = NF_INET_FORWARD,
  1173. + .priority = 100,
  1174. + },
  1175. + /* After packet filtering (but before ip_vs_out_icmp), catch icmp
  1176. + * destined for 0.0.0.0/0, which is for incoming IPVS connections */
  1177. + {
  1178. + .hook = ip_vs_forward_icmp,
  1179. + .owner = THIS_MODULE,
  1180. + .pf = PF_INET,
  1181. + .hooknum = NF_INET_FORWARD,
  1182. + .priority = 99,
  1183. + },
  1184. + /* Before the netfilter connection tracking, exit from POST_ROUTING */
  1185. + {
  1186. + .hook = ip_vs_post_routing,
  1187. + .owner = THIS_MODULE,
  1188. + .pf = PF_INET,
  1189. + .hooknum = NF_INET_POST_ROUTING,
  1190. + .priority = NF_IP_PRI_NAT_SRC-1,
  1191. + },
  1192. };
  1193.  
  1194.  
  1195. @@ -1138,37 +1090,15 @@
  1196. goto cleanup_app;
  1197. }
  1198.  
  1199. - ret = nf_register_hook(&ip_vs_in_ops);
  1200. + ret = nf_register_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops));
  1201. if (ret < 0) {
  1202. - IP_VS_ERR("can't register in hook.\n");
  1203. + IP_VS_ERR("can't register hooks.\n");
  1204. goto cleanup_conn;
  1205. }
  1206.  
  1207. - ret = nf_register_hook(&ip_vs_out_ops);
  1208. - if (ret < 0) {
  1209. - IP_VS_ERR("can't register out hook.\n");
  1210. - goto cleanup_inops;
  1211. - }
  1212. - ret = nf_register_hook(&ip_vs_post_routing_ops);
  1213. - if (ret < 0) {
  1214. - IP_VS_ERR("can't register post_routing hook.\n");
  1215. - goto cleanup_outops;
  1216. - }
  1217. - ret = nf_register_hook(&ip_vs_forward_icmp_ops);
  1218. - if (ret < 0) {
  1219. - IP_VS_ERR("can't register forward_icmp hook.\n");
  1220. - goto cleanup_postroutingops;
  1221. - }
  1222. -
  1223. IP_VS_INFO("ipvs loaded.\n");
  1224. return ret;
  1225.  
  1226. - cleanup_postroutingops:
  1227. - nf_unregister_hook(&ip_vs_post_routing_ops);
  1228. - cleanup_outops:
  1229. - nf_unregister_hook(&ip_vs_out_ops);
  1230. - cleanup_inops:
  1231. - nf_unregister_hook(&ip_vs_in_ops);
  1232. cleanup_conn:
  1233. ip_vs_conn_cleanup();
  1234. cleanup_app:
  1235. @@ -1182,10 +1112,7 @@
  1236.  
  1237. static void __exit ip_vs_cleanup(void)
  1238. {
  1239. - nf_unregister_hook(&ip_vs_forward_icmp_ops);
  1240. - nf_unregister_hook(&ip_vs_post_routing_ops);
  1241. - nf_unregister_hook(&ip_vs_out_ops);
  1242. - nf_unregister_hook(&ip_vs_in_ops);
  1243. + nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops));
  1244. ip_vs_conn_cleanup();
  1245. ip_vs_app_cleanup();
  1246. ip_vs_protocol_cleanup();
  1247. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_ctl.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_ctl.c
  1248. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_ctl.c 2009-02-16 11:57:22.000000000 -0400
  1249. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_ctl.c 2009-02-16 12:56:22.000000000 -0400
  1250. @@ -5,8 +5,6 @@
  1251. * high-performance and highly available server based on a
  1252. * cluster of servers.
  1253. *
  1254. - * Version: $Id: ip_vs_ctl.c,v 1.36 2003/06/08 09:31:19 wensong Exp $
  1255. - *
  1256. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  1257. * Peter Kese <peter.kese@ijs.si>
  1258. * Julian Anastasov <ja@ssi.bg>
  1259. @@ -29,13 +27,13 @@
  1260. #include <linux/proc_fs.h>
  1261. #include <linux/workqueue.h>
  1262. #include <linux/swap.h>
  1263. -#include <linux/proc_fs.h>
  1264. #include <linux/seq_file.h>
  1265.  
  1266. #include <linux/netfilter.h>
  1267. #include <linux/netfilter_ipv4.h>
  1268. #include <linux/mutex.h>
  1269.  
  1270. +#include <net/net_namespace.h>
  1271. #include <net/ip.h>
  1272. #include <net/route.h>
  1273. #include <net/sock.h>
  1274. @@ -579,6 +577,31 @@
  1275. return NULL;
  1276. }
  1277.  
  1278. +/*
  1279. + * Find destination by {daddr,dport,vaddr,protocol}
  1280. + * Cretaed to be used in ip_vs_process_message() in
  1281. + * the backup synchronization daemon. It finds the
  1282. + * destination to be bound to the received connection
  1283. + * on the backup.
  1284. + *
  1285. + * ip_vs_lookup_real_service() looked promissing, but
  1286. + * seems not working as expected.
  1287. + */
  1288. +struct ip_vs_dest *ip_vs_find_dest(__be32 daddr, __be16 dport,
  1289. + __be32 vaddr, __be16 vport, __u16 protocol)
  1290. +{
  1291. + struct ip_vs_dest *dest;
  1292. + struct ip_vs_service *svc;
  1293. +
  1294. + svc = ip_vs_service_get(0, protocol, vaddr, vport);
  1295. + if (!svc)
  1296. + return NULL;
  1297. + dest = ip_vs_lookup_dest(svc, daddr, dport);
  1298. + if (dest)
  1299. + atomic_inc(&dest->refcnt);
  1300. + ip_vs_service_put(svc);
  1301. + return dest;
  1302. +}
  1303.  
  1304. /*
  1305. * Lookup dest by {svc,addr,port} in the destination trash.
  1306. @@ -660,9 +683,22 @@
  1307. ip_vs_zero_stats(struct ip_vs_stats *stats)
  1308. {
  1309. spin_lock_bh(&stats->lock);
  1310. - memset(stats, 0, (char *)&stats->lock - (char *)stats);
  1311. - spin_unlock_bh(&stats->lock);
  1312. +
  1313. + stats->conns = 0;
  1314. + stats->inpkts = 0;
  1315. + stats->outpkts = 0;
  1316. + stats->inbytes = 0;
  1317. + stats->outbytes = 0;
  1318. +
  1319. + stats->cps = 0;
  1320. + stats->inpps = 0;
  1321. + stats->outpps = 0;
  1322. + stats->inbps = 0;
  1323. + stats->outbps = 0;
  1324. +
  1325. ip_vs_zero_estimator(stats);
  1326. +
  1327. + spin_unlock_bh(&stats->lock);
  1328. }
  1329.  
  1330. /*
  1331. @@ -679,7 +715,7 @@
  1332. conn_flags = udest->conn_flags | IP_VS_CONN_F_INACTIVE;
  1333.  
  1334. /* check if local node and update the flags */
  1335. - if (inet_addr_type(udest->addr) == RTN_LOCAL) {
  1336. + if (inet_addr_type(&init_net, udest->addr) == RTN_LOCAL) {
  1337. conn_flags = (conn_flags & ~IP_VS_CONN_F_FWD_MASK)
  1338. | IP_VS_CONN_F_LOCALNODE;
  1339. }
  1340. @@ -731,7 +767,7 @@
  1341.  
  1342. EnterFunction(2);
  1343.  
  1344. - atype = inet_addr_type(udest->addr);
  1345. + atype = inet_addr_type(&init_net, udest->addr);
  1346. if (atype != RTN_LOCAL && atype != RTN_UNICAST)
  1347. return -EINVAL;
  1348.  
  1349. @@ -909,7 +945,7 @@
  1350. write_lock_bh(&__ip_vs_svc_lock);
  1351.  
  1352. /* Wait until all other svc users go away */
  1353. - while (atomic_read(&svc->usecnt) > 1) {};
  1354. + IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
  1355.  
  1356. /* call the update_service, because server weight may be changed */
  1357. svc->scheduler->update_service(svc);
  1358. @@ -1399,7 +1435,6 @@
  1359.  
  1360. static struct ctl_table vs_vars[] = {
  1361. {
  1362. - .ctl_name = NET_IPV4_VS_AMEMTHRESH,
  1363. .procname = "amemthresh",
  1364. .data = &sysctl_ip_vs_amemthresh,
  1365. .maxlen = sizeof(int),
  1366. @@ -1408,7 +1443,6 @@
  1367. },
  1368. #ifdef CONFIG_IP_VS_DEBUG
  1369. {
  1370. - .ctl_name = NET_IPV4_VS_DEBUG_LEVEL,
  1371. .procname = "debug_level",
  1372. .data = &sysctl_ip_vs_debug_level,
  1373. .maxlen = sizeof(int),
  1374. @@ -1417,7 +1451,6 @@
  1375. },
  1376. #endif
  1377. {
  1378. - .ctl_name = NET_IPV4_VS_AMDROPRATE,
  1379. .procname = "am_droprate",
  1380. .data = &sysctl_ip_vs_am_droprate,
  1381. .maxlen = sizeof(int),
  1382. @@ -1425,7 +1458,6 @@
  1383. .proc_handler = &proc_dointvec,
  1384. },
  1385. {
  1386. - .ctl_name = NET_IPV4_VS_DROP_ENTRY,
  1387. .procname = "drop_entry",
  1388. .data = &sysctl_ip_vs_drop_entry,
  1389. .maxlen = sizeof(int),
  1390. @@ -1433,7 +1465,6 @@
  1391. .proc_handler = &proc_do_defense_mode,
  1392. },
  1393. {
  1394. - .ctl_name = NET_IPV4_VS_DROP_PACKET,
  1395. .procname = "drop_packet",
  1396. .data = &sysctl_ip_vs_drop_packet,
  1397. .maxlen = sizeof(int),
  1398. @@ -1441,7 +1472,6 @@
  1399. .proc_handler = &proc_do_defense_mode,
  1400. },
  1401. {
  1402. - .ctl_name = NET_IPV4_VS_SECURE_TCP,
  1403. .procname = "secure_tcp",
  1404. .data = &sysctl_ip_vs_secure_tcp,
  1405. .maxlen = sizeof(int),
  1406. @@ -1450,7 +1480,6 @@
  1407. },
  1408. #if 0
  1409. {
  1410. - .ctl_name = NET_IPV4_VS_TO_ES,
  1411. .procname = "timeout_established",
  1412. .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
  1413. .maxlen = sizeof(int),
  1414. @@ -1458,7 +1487,6 @@
  1415. .proc_handler = &proc_dointvec_jiffies,
  1416. },
  1417. {
  1418. - .ctl_name = NET_IPV4_VS_TO_SS,
  1419. .procname = "timeout_synsent",
  1420. .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
  1421. .maxlen = sizeof(int),
  1422. @@ -1466,7 +1494,6 @@
  1423. .proc_handler = &proc_dointvec_jiffies,
  1424. },
  1425. {
  1426. - .ctl_name = NET_IPV4_VS_TO_SR,
  1427. .procname = "timeout_synrecv",
  1428. .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
  1429. .maxlen = sizeof(int),
  1430. @@ -1474,7 +1501,6 @@
  1431. .proc_handler = &proc_dointvec_jiffies,
  1432. },
  1433. {
  1434. - .ctl_name = NET_IPV4_VS_TO_FW,
  1435. .procname = "timeout_finwait",
  1436. .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
  1437. .maxlen = sizeof(int),
  1438. @@ -1482,7 +1508,6 @@
  1439. .proc_handler = &proc_dointvec_jiffies,
  1440. },
  1441. {
  1442. - .ctl_name = NET_IPV4_VS_TO_TW,
  1443. .procname = "timeout_timewait",
  1444. .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
  1445. .maxlen = sizeof(int),
  1446. @@ -1490,7 +1515,6 @@
  1447. .proc_handler = &proc_dointvec_jiffies,
  1448. },
  1449. {
  1450. - .ctl_name = NET_IPV4_VS_TO_CL,
  1451. .procname = "timeout_close",
  1452. .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
  1453. .maxlen = sizeof(int),
  1454. @@ -1498,7 +1522,6 @@
  1455. .proc_handler = &proc_dointvec_jiffies,
  1456. },
  1457. {
  1458. - .ctl_name = NET_IPV4_VS_TO_CW,
  1459. .procname = "timeout_closewait",
  1460. .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
  1461. .maxlen = sizeof(int),
  1462. @@ -1506,7 +1529,6 @@
  1463. .proc_handler = &proc_dointvec_jiffies,
  1464. },
  1465. {
  1466. - .ctl_name = NET_IPV4_VS_TO_LA,
  1467. .procname = "timeout_lastack",
  1468. .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
  1469. .maxlen = sizeof(int),
  1470. @@ -1514,7 +1536,6 @@
  1471. .proc_handler = &proc_dointvec_jiffies,
  1472. },
  1473. {
  1474. - .ctl_name = NET_IPV4_VS_TO_LI,
  1475. .procname = "timeout_listen",
  1476. .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
  1477. .maxlen = sizeof(int),
  1478. @@ -1522,7 +1543,6 @@
  1479. .proc_handler = &proc_dointvec_jiffies,
  1480. },
  1481. {
  1482. - .ctl_name = NET_IPV4_VS_TO_SA,
  1483. .procname = "timeout_synack",
  1484. .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
  1485. .maxlen = sizeof(int),
  1486. @@ -1530,7 +1550,6 @@
  1487. .proc_handler = &proc_dointvec_jiffies,
  1488. },
  1489. {
  1490. - .ctl_name = NET_IPV4_VS_TO_UDP,
  1491. .procname = "timeout_udp",
  1492. .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
  1493. .maxlen = sizeof(int),
  1494. @@ -1538,7 +1557,6 @@
  1495. .proc_handler = &proc_dointvec_jiffies,
  1496. },
  1497. {
  1498. - .ctl_name = NET_IPV4_VS_TO_ICMP,
  1499. .procname = "timeout_icmp",
  1500. .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
  1501. .maxlen = sizeof(int),
  1502. @@ -1547,7 +1565,6 @@
  1503. },
  1504. #endif
  1505. {
  1506. - .ctl_name = NET_IPV4_VS_CACHE_BYPASS,
  1507. .procname = "cache_bypass",
  1508. .data = &sysctl_ip_vs_cache_bypass,
  1509. .maxlen = sizeof(int),
  1510. @@ -1555,7 +1572,6 @@
  1511. .proc_handler = &proc_dointvec,
  1512. },
  1513. {
  1514. - .ctl_name = NET_IPV4_VS_EXPIRE_NODEST_CONN,
  1515. .procname = "expire_nodest_conn",
  1516. .data = &sysctl_ip_vs_expire_nodest_conn,
  1517. .maxlen = sizeof(int),
  1518. @@ -1563,7 +1579,6 @@
  1519. .proc_handler = &proc_dointvec,
  1520. },
  1521. {
  1522. - .ctl_name = NET_IPV4_VS_EXPIRE_QUIESCENT_TEMPLATE,
  1523. .procname = "expire_quiescent_template",
  1524. .data = &sysctl_ip_vs_expire_quiescent_template,
  1525. .maxlen = sizeof(int),
  1526. @@ -1571,7 +1586,6 @@
  1527. .proc_handler = &proc_dointvec,
  1528. },
  1529. {
  1530. - .ctl_name = NET_IPV4_VS_SYNC_THRESHOLD,
  1531. .procname = "sync_threshold",
  1532. .data = &sysctl_ip_vs_sync_threshold,
  1533. .maxlen = sizeof(sysctl_ip_vs_sync_threshold),
  1534. @@ -1579,7 +1593,6 @@
  1535. .proc_handler = &proc_do_sync_threshold,
  1536. },
  1537. {
  1538. - .ctl_name = NET_IPV4_VS_NAT_ICMP_SEND,
  1539. .procname = "nat_icmp_send",
  1540. .data = &sysctl_ip_vs_nat_icmp_send,
  1541. .maxlen = sizeof(int),
  1542. @@ -1589,35 +1602,13 @@
  1543. { .ctl_name = 0 }
  1544. };
  1545.  
  1546. -static ctl_table vs_table[] = {
  1547. - {
  1548. - .ctl_name = NET_IPV4_VS,
  1549. - .procname = "vs",
  1550. - .mode = 0555,
  1551. - .child = vs_vars
  1552. - },
  1553. - { .ctl_name = 0 }
  1554. -};
  1555. -
  1556. -static ctl_table ipvs_ipv4_table[] = {
  1557. - {
  1558. - .ctl_name = NET_IPV4,
  1559. - .procname = "ipv4",
  1560. - .mode = 0555,
  1561. - .child = vs_table,
  1562. - },
  1563. - { .ctl_name = 0 }
  1564. -};
  1565. -
  1566. -static ctl_table vs_root_table[] = {
  1567. - {
  1568. - .ctl_name = CTL_NET,
  1569. - .procname = "net",
  1570. - .mode = 0555,
  1571. - .child = ipvs_ipv4_table,
  1572. - },
  1573. - { .ctl_name = 0 }
  1574. +const struct ctl_path net_vs_ctl_path[] = {
  1575. + { .procname = "net", .ctl_name = CTL_NET, },
  1576. + { .procname = "ipv4", .ctl_name = NET_IPV4, },
  1577. + { .procname = "vs", },
  1578. + { }
  1579. };
  1580. +EXPORT_SYMBOL_GPL(net_vs_ctl_path);
  1581.  
  1582. static struct ctl_table_header * sysctl_header;
  1583.  
  1584. @@ -1783,7 +1774,7 @@
  1585. return 0;
  1586. }
  1587.  
  1588. -static struct seq_operations ip_vs_info_seq_ops = {
  1589. +static const struct seq_operations ip_vs_info_seq_ops = {
  1590. .start = ip_vs_info_seq_start,
  1591. .next = ip_vs_info_seq_next,
  1592. .stop = ip_vs_info_seq_stop,
  1593. @@ -1792,27 +1783,11 @@
  1594.  
  1595. static int ip_vs_info_open(struct inode *inode, struct file *file)
  1596. {
  1597. - struct seq_file *seq;
  1598. - int rc = -ENOMEM;
  1599. - struct ip_vs_iter *s = kzalloc(sizeof(*s), GFP_KERNEL);
  1600. -
  1601. - if (!s)
  1602. - goto out;
  1603. -
  1604. - rc = seq_open(file, &ip_vs_info_seq_ops);
  1605. - if (rc)
  1606. - goto out_kfree;
  1607. -
  1608. - seq = file->private_data;
  1609. - seq->private = s;
  1610. -out:
  1611. - return rc;
  1612. -out_kfree:
  1613. - kfree(s);
  1614. - goto out;
  1615. + return seq_open_private(file, &ip_vs_info_seq_ops,
  1616. + sizeof(struct ip_vs_iter));
  1617. }
  1618.  
  1619. -static struct file_operations ip_vs_info_fops = {
  1620. +static const struct file_operations ip_vs_info_fops = {
  1621. .owner = THIS_MODULE,
  1622. .open = ip_vs_info_open,
  1623. .read = seq_read,
  1624. @@ -1822,7 +1797,9 @@
  1625.  
  1626. #endif
  1627.  
  1628. -struct ip_vs_stats ip_vs_stats;
  1629. +struct ip_vs_stats ip_vs_stats = {
  1630. + .lock = __SPIN_LOCK_UNLOCKED(ip_vs_stats.lock),
  1631. +};
  1632.  
  1633. #ifdef CONFIG_PROC_FS
  1634. static int ip_vs_stats_show(struct seq_file *seq, void *v)
  1635. @@ -1859,7 +1836,7 @@
  1636. return single_open(file, ip_vs_stats_show, NULL);
  1637. }
  1638.  
  1639. -static struct file_operations ip_vs_stats_fops = {
  1640. +static const struct file_operations ip_vs_stats_fops = {
  1641. .owner = THIS_MODULE,
  1642. .open = ip_vs_stats_seq_open,
  1643. .read = seq_read,
  1644. @@ -2340,10 +2317,11 @@
  1645. .get_optmin = IP_VS_BASE_CTL,
  1646. .get_optmax = IP_VS_SO_GET_MAX+1,
  1647. .get = do_ip_vs_get_ctl,
  1648. + .owner = THIS_MODULE,
  1649. };
  1650.  
  1651.  
  1652. -int ip_vs_control_init(void)
  1653. +int __init ip_vs_control_init(void)
  1654. {
  1655. int ret;
  1656. int idx;
  1657. @@ -2356,10 +2334,10 @@
  1658. return ret;
  1659. }
  1660.  
  1661. - proc_net_fops_create("ip_vs", 0, &ip_vs_info_fops);
  1662. - proc_net_fops_create("ip_vs_stats",0, &ip_vs_stats_fops);
  1663. + proc_net_fops_create(&init_net, "ip_vs", 0, &ip_vs_info_fops);
  1664. + proc_net_fops_create(&init_net, "ip_vs_stats",0, &ip_vs_stats_fops);
  1665.  
  1666. - sysctl_header = register_sysctl_table(vs_root_table, 0);
  1667. + sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars);
  1668.  
  1669. /* Initialize ip_vs_svc_table, ip_vs_svc_fwm_table, ip_vs_rtable */
  1670. for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
  1671. @@ -2370,8 +2348,6 @@
  1672. INIT_LIST_HEAD(&ip_vs_rtable[idx]);
  1673. }
  1674.  
  1675. - memset(&ip_vs_stats, 0, sizeof(ip_vs_stats));
  1676. - spin_lock_init(&ip_vs_stats.lock);
  1677. ip_vs_new_estimator(&ip_vs_stats);
  1678.  
  1679. /* Hook the defense timer */
  1680. @@ -2387,10 +2363,11 @@
  1681. EnterFunction(2);
  1682. ip_vs_trash_cleanup();
  1683. cancel_rearming_delayed_work(&defense_work);
  1684. + cancel_work_sync(&defense_work.work);
  1685. ip_vs_kill_estimator(&ip_vs_stats);
  1686. unregister_sysctl_table(sysctl_header);
  1687. - proc_net_remove("ip_vs_stats");
  1688. - proc_net_remove("ip_vs");
  1689. + proc_net_remove(&init_net, "ip_vs_stats");
  1690. + proc_net_remove(&init_net, "ip_vs");
  1691. nf_unregister_sockopt(&ip_vs_sockopts);
  1692. LeaveFunction(2);
  1693. }
  1694. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_dh.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_dh.c
  1695. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_dh.c 2009-02-16 11:57:22.000000000 -0400
  1696. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_dh.c 2009-02-16 12:56:22.000000000 -0400
  1697. @@ -1,8 +1,6 @@
  1698. /*
  1699. * IPVS: Destination Hashing scheduling module
  1700. *
  1701. - * Version: $Id: ip_vs_dh.c,v 1.5 2002/09/15 08:14:08 wensong Exp $
  1702. - *
  1703. * Authors: Wensong Zhang <wensong@gnuchina.org>
  1704. *
  1705. * Inspired by the consistent hashing scheduler patch from
  1706. @@ -204,7 +202,7 @@
  1707. {
  1708. struct ip_vs_dest *dest;
  1709. struct ip_vs_dh_bucket *tbl;
  1710. - struct iphdr *iph = skb->nh.iph;
  1711. + struct iphdr *iph = ip_hdr(skb);
  1712.  
  1713. IP_VS_DBG(6, "ip_vs_dh_schedule(): Scheduling...\n");
  1714.  
  1715. @@ -235,6 +233,7 @@
  1716. .name = "dh",
  1717. .refcnt = ATOMIC_INIT(0),
  1718. .module = THIS_MODULE,
  1719. + .n_list = LIST_HEAD_INIT(ip_vs_dh_scheduler.n_list),
  1720. .init_service = ip_vs_dh_init_svc,
  1721. .done_service = ip_vs_dh_done_svc,
  1722. .update_service = ip_vs_dh_update_svc,
  1723. @@ -244,7 +243,6 @@
  1724.  
  1725. static int __init ip_vs_dh_init(void)
  1726. {
  1727. - INIT_LIST_HEAD(&ip_vs_dh_scheduler.n_list);
  1728. return register_ip_vs_scheduler(&ip_vs_dh_scheduler);
  1729. }
  1730.  
  1731. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_est.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_est.c
  1732. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_est.c 2009-02-16 11:57:22.000000000 -0400
  1733. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_est.c 2009-02-16 12:56:22.000000000 -0400
  1734. @@ -1,8 +1,6 @@
  1735. /*
  1736. * ip_vs_est.c: simple rate estimator for IPVS
  1737. *
  1738. - * Version: $Id: ip_vs_est.c,v 1.4 2002/11/30 01:50:35 wensong Exp $
  1739. - *
  1740. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  1741. *
  1742. * This program is free software; you can redistribute it and/or
  1743. @@ -18,6 +16,8 @@
  1744. #include <linux/slab.h>
  1745. #include <linux/types.h>
  1746. #include <linux/interrupt.h>
  1747. +#include <linux/sysctl.h>
  1748. +#include <linux/list.h>
  1749.  
  1750. #include <net/ip_vs.h>
  1751.  
  1752. @@ -45,28 +45,11 @@
  1753. */
  1754.  
  1755.  
  1756. -struct ip_vs_estimator
  1757. -{
  1758. - struct ip_vs_estimator *next;
  1759. - struct ip_vs_stats *stats;
  1760. +static void estimation_timer(unsigned long arg);
  1761.  
  1762. - u32 last_conns;
  1763. - u32 last_inpkts;
  1764. - u32 last_outpkts;
  1765. - u64 last_inbytes;
  1766. - u64 last_outbytes;
  1767. -
  1768. - u32 cps;
  1769. - u32 inpps;
  1770. - u32 outpps;
  1771. - u32 inbps;
  1772. - u32 outbps;
  1773. -};
  1774. -
  1775. -
  1776. -static struct ip_vs_estimator *est_list = NULL;
  1777. -static DEFINE_RWLOCK(est_lock);
  1778. -static struct timer_list est_timer;
  1779. +static LIST_HEAD(est_list);
  1780. +static DEFINE_SPINLOCK(est_lock);
  1781. +static DEFINE_TIMER(est_timer, estimation_timer, 0, 0);
  1782.  
  1783. static void estimation_timer(unsigned long arg)
  1784. {
  1785. @@ -77,9 +60,9 @@
  1786. u64 n_inbytes, n_outbytes;
  1787. u32 rate;
  1788.  
  1789. - read_lock(&est_lock);
  1790. - for (e = est_list; e; e = e->next) {
  1791. - s = e->stats;
  1792. + spin_lock(&est_lock);
  1793. + list_for_each_entry(e, &est_list, list) {
  1794. + s = container_of(e, struct ip_vs_stats, est);
  1795.  
  1796. spin_lock(&s->lock);
  1797. n_conns = s->conns;
  1798. @@ -115,19 +98,16 @@
  1799. s->outbps = (e->outbps+0xF)>>5;
  1800. spin_unlock(&s->lock);
  1801. }
  1802. - read_unlock(&est_lock);
  1803. + spin_unlock(&est_lock);
  1804. mod_timer(&est_timer, jiffies + 2*HZ);
  1805. }
  1806.  
  1807. -int ip_vs_new_estimator(struct ip_vs_stats *stats)
  1808. +void ip_vs_new_estimator(struct ip_vs_stats *stats)
  1809. {
  1810. - struct ip_vs_estimator *est;
  1811. + struct ip_vs_estimator *est = &stats->est;
  1812.  
  1813. - est = kzalloc(sizeof(*est), GFP_KERNEL);
  1814. - if (est == NULL)
  1815. - return -ENOMEM;
  1816. + INIT_LIST_HEAD(&est->list);
  1817.  
  1818. - est->stats = stats;
  1819. est->last_conns = stats->conns;
  1820. est->cps = stats->cps<<10;
  1821.  
  1822. @@ -143,60 +123,40 @@
  1823. est->last_outbytes = stats->outbytes;
  1824. est->outbps = stats->outbps<<5;
  1825.  
  1826. - write_lock_bh(&est_lock);
  1827. - est->next = est_list;
  1828. - if (est->next == NULL) {
  1829. - init_timer(&est_timer);
  1830. - est_timer.expires = jiffies + 2*HZ;
  1831. - est_timer.function = estimation_timer;
  1832. - add_timer(&est_timer);
  1833. - }
  1834. - est_list = est;
  1835. - write_unlock_bh(&est_lock);
  1836. - return 0;
  1837. + spin_lock_bh(&est_lock);
  1838. + if (list_empty(&est_list))
  1839. + mod_timer(&est_timer, jiffies + 2 * HZ);
  1840. + list_add(&est->list, &est_list);
  1841. + spin_unlock_bh(&est_lock);
  1842. }
  1843.  
  1844. void ip_vs_kill_estimator(struct ip_vs_stats *stats)
  1845. {
  1846. - struct ip_vs_estimator *est, **pest;
  1847. - int killed = 0;
  1848. + struct ip_vs_estimator *est = &stats->est;
  1849.  
  1850. - write_lock_bh(&est_lock);
  1851. - pest = &est_list;
  1852. - while ((est=*pest) != NULL) {
  1853. - if (est->stats != stats) {
  1854. - pest = &est->next;
  1855. - continue;
  1856. - }
  1857. - *pest = est->next;
  1858. - kfree(est);
  1859. - killed++;
  1860. + spin_lock_bh(&est_lock);
  1861. + list_del(&est->list);
  1862. + while (list_empty(&est_list) && try_to_del_timer_sync(&est_timer) < 0) {
  1863. + spin_unlock_bh(&est_lock);
  1864. + cpu_relax();
  1865. + spin_lock_bh(&est_lock);
  1866. }
  1867. - if (killed && est_list == NULL)
  1868. - del_timer_sync(&est_timer);
  1869. - write_unlock_bh(&est_lock);
  1870. + spin_unlock_bh(&est_lock);
  1871. }
  1872.  
  1873. void ip_vs_zero_estimator(struct ip_vs_stats *stats)
  1874. {
  1875. - struct ip_vs_estimator *e;
  1876. + struct ip_vs_estimator *est = &stats->est;
  1877.  
  1878. - write_lock_bh(&est_lock);
  1879. - for (e = est_list; e; e = e->next) {
  1880. - if (e->stats != stats)
  1881. - continue;
  1882. -
  1883. - /* set counters zero */
  1884. - e->last_conns = 0;
  1885. - e->last_inpkts = 0;
  1886. - e->last_outpkts = 0;
  1887. - e->last_inbytes = 0;
  1888. - e->last_outbytes = 0;
  1889. - e->cps = 0;
  1890. - e->inpps = 0;
  1891. - e->outpps = 0;
  1892. - e->inbps = 0;
  1893. - e->outbps = 0;
  1894. - }
  1895. - write_unlock_bh(&est_lock);
  1896. + /* set counters zero, caller must hold the stats->lock lock */
  1897. + est->last_inbytes = 0;
  1898. + est->last_outbytes = 0;
  1899. + est->last_conns = 0;
  1900. + est->last_inpkts = 0;
  1901. + est->last_outpkts = 0;
  1902. + est->cps = 0;
  1903. + est->inpps = 0;
  1904. + est->outpps = 0;
  1905. + est->inbps = 0;
  1906. + est->outbps = 0;
  1907. }
  1908. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_ftp.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_ftp.c
  1909. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_ftp.c 2009-02-16 11:57:22.000000000 -0400
  1910. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_ftp.c 2009-02-16 12:56:22.000000000 -0400
  1911. @@ -1,8 +1,6 @@
  1912. /*
  1913. * ip_vs_ftp.c: IPVS ftp application module
  1914. *
  1915. - * Version: $Id: ip_vs_ftp.c,v 1.13 2002/09/15 08:14:08 wensong Exp $
  1916. - *
  1917. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  1918. *
  1919. * Changes:
  1920. @@ -30,6 +28,7 @@
  1921. #include <linux/skbuff.h>
  1922. #include <linux/in.h>
  1923. #include <linux/ip.h>
  1924. +#include <linux/netfilter.h>
  1925. #include <net/protocol.h>
  1926. #include <net/tcp.h>
  1927. #include <asm/unaligned.h>
  1928. @@ -135,7 +134,7 @@
  1929. * xxx,xxx,xxx,xxx is the server address, ppp,ppp is the server port number.
  1930. */
  1931. static int ip_vs_ftp_out(struct ip_vs_app *app, struct ip_vs_conn *cp,
  1932. - struct sk_buff **pskb, int *diff)
  1933. + struct sk_buff *skb, int *diff)
  1934. {
  1935. struct iphdr *iph;
  1936. struct tcphdr *th;
  1937. @@ -155,14 +154,14 @@
  1938. return 1;
  1939.  
  1940. /* Linear packets are much easier to deal with. */
  1941. - if (!ip_vs_make_skb_writable(pskb, (*pskb)->len))
  1942. + if (!skb_make_writable(skb, skb->len))
  1943. return 0;
  1944.  
  1945. if (cp->app_data == &ip_vs_ftp_pasv) {
  1946. - iph = (*pskb)->nh.iph;
  1947. + iph = ip_hdr(skb);
  1948. th = (struct tcphdr *)&(((char *)iph)[iph->ihl*4]);
  1949. data = (char *)th + (th->doff << 2);
  1950. - data_limit = (*pskb)->tail;
  1951. + data_limit = skb_tail_pointer(skb);
  1952.  
  1953. if (ip_vs_ftp_get_addrport(data, data_limit,
  1954. SERVER_STRING,
  1955. @@ -213,7 +212,7 @@
  1956. memcpy(start, buf, buf_len);
  1957. ret = 1;
  1958. } else {
  1959. - ret = !ip_vs_skb_replace(*pskb, GFP_ATOMIC, start,
  1960. + ret = !ip_vs_skb_replace(skb, GFP_ATOMIC, start,
  1961. end-start, buf, buf_len);
  1962. }
  1963.  
  1964. @@ -238,7 +237,7 @@
  1965. * the client.
  1966. */
  1967. static int ip_vs_ftp_in(struct ip_vs_app *app, struct ip_vs_conn *cp,
  1968. - struct sk_buff **pskb, int *diff)
  1969. + struct sk_buff *skb, int *diff)
  1970. {
  1971. struct iphdr *iph;
  1972. struct tcphdr *th;
  1973. @@ -256,20 +255,20 @@
  1974. return 1;
  1975.  
  1976. /* Linear packets are much easier to deal with. */
  1977. - if (!ip_vs_make_skb_writable(pskb, (*pskb)->len))
  1978. + if (!skb_make_writable(skb, skb->len))
  1979. return 0;
  1980.  
  1981. /*
  1982. * Detecting whether it is passive
  1983. */
  1984. - iph = (*pskb)->nh.iph;
  1985. + iph = ip_hdr(skb);
  1986. th = (struct tcphdr *)&(((char *)iph)[iph->ihl*4]);
  1987.  
  1988. /* Since there may be OPTIONS in the TCP packet and the HLEN is
  1989. the length of the header in 32-bit multiples, it is accurate
  1990. to calculate data address by th+HLEN*4 */
  1991. data = data_start = (char *)th + (th->doff << 2);
  1992. - data_limit = (*pskb)->tail;
  1993. + data_limit = skb_tail_pointer(skb);
  1994.  
  1995. while (data <= data_limit - 6) {
  1996. if (strnicmp(data, "PASV\r\n", 6) == 0) {
  1997. @@ -370,7 +369,7 @@
  1998. if (ret)
  1999. break;
  2000. IP_VS_INFO("%s: loaded support on port[%d] = %d\n",
  2001. - app->name, i, ports[i]);
  2002. + app->name, i, ports[i]);
  2003. }
  2004.  
  2005. if (ret)
  2006. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_lblc.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_lblc.c
  2007. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_lblc.c 2009-02-16 11:57:22.000000000 -0400
  2008. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_lblc.c 2009-02-16 12:56:22.000000000 -0400
  2009. @@ -1,8 +1,6 @@
  2010. /*
  2011. * IPVS: Locality-Based Least-Connection scheduling module
  2012. *
  2013. - * Version: $Id: ip_vs_lblc.c,v 1.10 2002/09/15 08:14:08 wensong Exp $
  2014. - *
  2015. * Authors: Wensong Zhang <wensong@gnuchina.org>
  2016. *
  2017. * This program is free software; you can redistribute it and/or
  2018. @@ -114,46 +112,15 @@
  2019.  
  2020. static ctl_table vs_vars_table[] = {
  2021. {
  2022. - .ctl_name = NET_IPV4_VS_LBLC_EXPIRE,
  2023. .procname = "lblc_expiration",
  2024. .data = &sysctl_ip_vs_lblc_expiration,
  2025. .maxlen = sizeof(int),
  2026. - .mode = 0644,
  2027. + .mode = 0644,
  2028. .proc_handler = &proc_dointvec_jiffies,
  2029. },
  2030. { .ctl_name = 0 }
  2031. };
  2032.  
  2033. -static ctl_table vs_table[] = {
  2034. - {
  2035. - .ctl_name = NET_IPV4_VS,
  2036. - .procname = "vs",
  2037. - .mode = 0555,
  2038. - .child = vs_vars_table
  2039. - },
  2040. - { .ctl_name = 0 }
  2041. -};
  2042. -
  2043. -static ctl_table ipvs_ipv4_table[] = {
  2044. - {
  2045. - .ctl_name = NET_IPV4,
  2046. - .procname = "ipv4",
  2047. - .mode = 0555,
  2048. - .child = vs_table
  2049. - },
  2050. - { .ctl_name = 0 }
  2051. -};
  2052. -
  2053. -static ctl_table lblc_root_table[] = {
  2054. - {
  2055. - .ctl_name = CTL_NET,
  2056. - .procname = "net",
  2057. - .mode = 0555,
  2058. - .child = ipvs_ipv4_table
  2059. - },
  2060. - { .ctl_name = 0 }
  2061. -};
  2062. -
  2063. static struct ctl_table_header * sysctl_header;
  2064.  
  2065. /*
  2066. @@ -288,7 +255,7 @@
  2067.  
  2068. write_lock(&tbl->lock);
  2069. list_for_each_entry_safe(en, nxt, &tbl->bucket[j], list) {
  2070. - if (time_before(now,
  2071. + if (time_before(now,
  2072. en->lastuse + sysctl_ip_vs_lblc_expiration))
  2073. continue;
  2074.  
  2075. @@ -393,9 +360,8 @@
  2076. /*
  2077. * Hook periodic timer for garbage collection
  2078. */
  2079. - init_timer(&tbl->periodic_timer);
  2080. - tbl->periodic_timer.data = (unsigned long)tbl;
  2081. - tbl->periodic_timer.function = ip_vs_lblc_check_expire;
  2082. + setup_timer(&tbl->periodic_timer, ip_vs_lblc_check_expire,
  2083. + (unsigned long)tbl);
  2084. tbl->periodic_timer.expires = jiffies+CHECK_EXPIRE_INTERVAL;
  2085. add_timer(&tbl->periodic_timer);
  2086.  
  2087. @@ -521,7 +487,7 @@
  2088. struct ip_vs_dest *dest;
  2089. struct ip_vs_lblc_table *tbl;
  2090. struct ip_vs_lblc_entry *en;
  2091. - struct iphdr *iph = skb->nh.iph;
  2092. + struct iphdr *iph = ip_hdr(skb);
  2093.  
  2094. IP_VS_DBG(6, "ip_vs_lblc_schedule(): Scheduling...\n");
  2095.  
  2096. @@ -573,6 +539,7 @@
  2097. .name = "lblc",
  2098. .refcnt = ATOMIC_INIT(0),
  2099. .module = THIS_MODULE,
  2100. + .n_list = LIST_HEAD_INIT(ip_vs_lblc_scheduler.n_list),
  2101. .init_service = ip_vs_lblc_init_svc,
  2102. .done_service = ip_vs_lblc_done_svc,
  2103. .update_service = ip_vs_lblc_update_svc,
  2104. @@ -582,9 +549,13 @@
  2105.  
  2106. static int __init ip_vs_lblc_init(void)
  2107. {
  2108. - INIT_LIST_HEAD(&ip_vs_lblc_scheduler.n_list);
  2109. - sysctl_header = register_sysctl_table(lblc_root_table, 0);
  2110. - return register_ip_vs_scheduler(&ip_vs_lblc_scheduler);
  2111. + int ret;
  2112. +
  2113. + sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars_table);
  2114. + ret = register_ip_vs_scheduler(&ip_vs_lblc_scheduler);
  2115. + if (ret)
  2116. + unregister_sysctl_table(sysctl_header);
  2117. + return ret;
  2118. }
  2119.  
  2120.  
  2121. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_lblcr.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_lblcr.c
  2122. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_lblcr.c 2009-02-16 11:57:22.000000000 -0400
  2123. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_lblcr.c 2009-02-16 12:56:22.000000000 -0400
  2124. @@ -1,8 +1,6 @@
  2125. /*
  2126. * IPVS: Locality-Based Least-Connection with Replication scheduler
  2127. *
  2128. - * Version: $Id: ip_vs_lblcr.c,v 1.11 2002/09/15 08:14:08 wensong Exp $
  2129. - *
  2130. * Authors: Wensong Zhang <wensong@gnuchina.org>
  2131. *
  2132. * This program is free software; you can redistribute it and/or
  2133. @@ -48,8 +46,7 @@
  2134. /* for sysctl */
  2135. #include <linux/fs.h>
  2136. #include <linux/sysctl.h>
  2137. -/* for proc_net_create/proc_net_remove */
  2138. -#include <linux/proc_fs.h>
  2139. +#include <net/net_namespace.h>
  2140.  
  2141. #include <net/ip_vs.h>
  2142.  
  2143. @@ -303,46 +300,15 @@
  2144.  
  2145. static ctl_table vs_vars_table[] = {
  2146. {
  2147. - .ctl_name = NET_IPV4_VS_LBLCR_EXPIRE,
  2148. .procname = "lblcr_expiration",
  2149. .data = &sysctl_ip_vs_lblcr_expiration,
  2150. .maxlen = sizeof(int),
  2151. - .mode = 0644,
  2152. + .mode = 0644,
  2153. .proc_handler = &proc_dointvec_jiffies,
  2154. },
  2155. { .ctl_name = 0 }
  2156. };
  2157.  
  2158. -static ctl_table vs_table[] = {
  2159. - {
  2160. - .ctl_name = NET_IPV4_VS,
  2161. - .procname = "vs",
  2162. - .mode = 0555,
  2163. - .child = vs_vars_table
  2164. - },
  2165. - { .ctl_name = 0 }
  2166. -};
  2167. -
  2168. -static ctl_table ipvs_ipv4_table[] = {
  2169. - {
  2170. - .ctl_name = NET_IPV4,
  2171. - .procname = "ipv4",
  2172. - .mode = 0555,
  2173. - .child = vs_table
  2174. - },
  2175. - { .ctl_name = 0 }
  2176. -};
  2177. -
  2178. -static ctl_table lblcr_root_table[] = {
  2179. - {
  2180. - .ctl_name = CTL_NET,
  2181. - .procname = "net",
  2182. - .mode = 0555,
  2183. - .child = ipvs_ipv4_table
  2184. - },
  2185. - { .ctl_name = 0 }
  2186. -};
  2187. -
  2188. static struct ctl_table_header * sysctl_header;
  2189.  
  2190. /*
  2191. @@ -546,71 +512,6 @@
  2192. mod_timer(&tbl->periodic_timer, jiffies+CHECK_EXPIRE_INTERVAL);
  2193. }
  2194.  
  2195. -
  2196. -#ifdef CONFIG_IP_VS_LBLCR_DEBUG
  2197. -static struct ip_vs_lblcr_table *lblcr_table_list;
  2198. -
  2199. -/*
  2200. - * /proc/net/ip_vs_lblcr to display the mappings of
  2201. - * destination IP address <==> its serverSet
  2202. - */
  2203. -static int
  2204. -ip_vs_lblcr_getinfo(char *buffer, char **start, off_t offset, int length)
  2205. -{
  2206. - off_t pos=0, begin;
  2207. - int len=0, size;
  2208. - struct ip_vs_lblcr_table *tbl;
  2209. - unsigned long now = jiffies;
  2210. - int i;
  2211. - struct ip_vs_lblcr_entry *en;
  2212. -
  2213. - tbl = lblcr_table_list;
  2214. -
  2215. - size = sprintf(buffer, "LastTime Dest IP address Server set\n");
  2216. - pos += size;
  2217. - len += size;
  2218. -
  2219. - for (i=0; i<IP_VS_LBLCR_TAB_SIZE; i++) {
  2220. - read_lock_bh(&tbl->lock);
  2221. - list_for_each_entry(en, &tbl->bucket[i], list) {
  2222. - char tbuf[16];
  2223. - struct ip_vs_dest_list *d;
  2224. -
  2225. - sprintf(tbuf, "%u.%u.%u.%u", NIPQUAD(en->addr));
  2226. - size = sprintf(buffer+len, "%8lu %-16s ",
  2227. - now-en->lastuse, tbuf);
  2228. -
  2229. - read_lock(&en->set.lock);
  2230. - for (d=en->set.list; d!=NULL; d=d->next) {
  2231. - size += sprintf(buffer+len+size,
  2232. - "%u.%u.%u.%u ",
  2233. - NIPQUAD(d->dest->addr));
  2234. - }
  2235. - read_unlock(&en->set.lock);
  2236. - size += sprintf(buffer+len+size, "\n");
  2237. - len += size;
  2238. - pos += size;
  2239. - if (pos <= offset)
  2240. - len=0;
  2241. - if (pos >= offset+length) {
  2242. - read_unlock_bh(&tbl->lock);
  2243. - goto done;
  2244. - }
  2245. - }
  2246. - read_unlock_bh(&tbl->lock);
  2247. - }
  2248. -
  2249. - done:
  2250. - begin = len - (pos - offset);
  2251. - *start = buffer + begin;
  2252. - len -= begin;
  2253. - if(len>length)
  2254. - len = length;
  2255. - return len;
  2256. -}
  2257. -#endif
  2258. -
  2259. -
  2260. static int ip_vs_lblcr_init_svc(struct ip_vs_service *svc)
  2261. {
  2262. int i;
  2263. @@ -643,15 +544,11 @@
  2264. /*
  2265. * Hook periodic timer for garbage collection
  2266. */
  2267. - init_timer(&tbl->periodic_timer);
  2268. - tbl->periodic_timer.data = (unsigned long)tbl;
  2269. - tbl->periodic_timer.function = ip_vs_lblcr_check_expire;
  2270. + setup_timer(&tbl->periodic_timer, ip_vs_lblcr_check_expire,
  2271. + (unsigned long)tbl);
  2272. tbl->periodic_timer.expires = jiffies+CHECK_EXPIRE_INTERVAL;
  2273. add_timer(&tbl->periodic_timer);
  2274.  
  2275. -#ifdef CONFIG_IP_VS_LBLCR_DEBUG
  2276. - lblcr_table_list = tbl;
  2277. -#endif
  2278. return 0;
  2279. }
  2280.  
  2281. @@ -775,7 +672,7 @@
  2282. struct ip_vs_dest *dest;
  2283. struct ip_vs_lblcr_table *tbl;
  2284. struct ip_vs_lblcr_entry *en;
  2285. - struct iphdr *iph = skb->nh.iph;
  2286. + struct iphdr *iph = ip_hdr(skb);
  2287.  
  2288. IP_VS_DBG(6, "ip_vs_lblcr_schedule(): Scheduling...\n");
  2289.  
  2290. @@ -831,6 +728,7 @@
  2291. .name = "lblcr",
  2292. .refcnt = ATOMIC_INIT(0),
  2293. .module = THIS_MODULE,
  2294. + .n_list = LIST_HEAD_INIT(ip_vs_lblcr_scheduler.n_list),
  2295. .init_service = ip_vs_lblcr_init_svc,
  2296. .done_service = ip_vs_lblcr_done_svc,
  2297. .update_service = ip_vs_lblcr_update_svc,
  2298. @@ -840,20 +738,18 @@
  2299.  
  2300. static int __init ip_vs_lblcr_init(void)
  2301. {
  2302. - INIT_LIST_HEAD(&ip_vs_lblcr_scheduler.n_list);
  2303. - sysctl_header = register_sysctl_table(lblcr_root_table, 0);
  2304. -#ifdef CONFIG_IP_VS_LBLCR_DEBUG
  2305. - proc_net_create("ip_vs_lblcr", 0, ip_vs_lblcr_getinfo);
  2306. -#endif
  2307. - return register_ip_vs_scheduler(&ip_vs_lblcr_scheduler);
  2308. + int ret;
  2309. +
  2310. + sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars_table);
  2311. + ret = register_ip_vs_scheduler(&ip_vs_lblcr_scheduler);
  2312. + if (ret)
  2313. + unregister_sysctl_table(sysctl_header);
  2314. + return ret;
  2315. }
  2316.  
  2317.  
  2318. static void __exit ip_vs_lblcr_cleanup(void)
  2319. {
  2320. -#ifdef CONFIG_IP_VS_LBLCR_DEBUG
  2321. - proc_net_remove("ip_vs_lblcr");
  2322. -#endif
  2323. unregister_sysctl_table(sysctl_header);
  2324. unregister_ip_vs_scheduler(&ip_vs_lblcr_scheduler);
  2325. }
  2326. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_lc.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_lc.c
  2327. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_lc.c 2009-02-16 11:57:22.000000000 -0400
  2328. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_lc.c 2009-02-16 12:56:22.000000000 -0400
  2329. @@ -1,8 +1,6 @@
  2330. /*
  2331. * IPVS: Least-Connection Scheduling module
  2332. *
  2333. - * Version: $Id: ip_vs_lc.c,v 1.10 2003/04/18 09:03:16 wensong Exp $
  2334. - *
  2335. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  2336. *
  2337. * This program is free software; you can redistribute it and/or
  2338. @@ -100,6 +98,7 @@
  2339. .name = "lc",
  2340. .refcnt = ATOMIC_INIT(0),
  2341. .module = THIS_MODULE,
  2342. + .n_list = LIST_HEAD_INIT(ip_vs_lc_scheduler.n_list),
  2343. .init_service = ip_vs_lc_init_svc,
  2344. .done_service = ip_vs_lc_done_svc,
  2345. .update_service = ip_vs_lc_update_svc,
  2346. @@ -109,7 +108,6 @@
  2347.  
  2348. static int __init ip_vs_lc_init(void)
  2349. {
  2350. - INIT_LIST_HEAD(&ip_vs_lc_scheduler.n_list);
  2351. return register_ip_vs_scheduler(&ip_vs_lc_scheduler) ;
  2352. }
  2353.  
  2354. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_nq.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_nq.c
  2355. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_nq.c 2009-02-16 11:57:22.000000000 -0400
  2356. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_nq.c 2009-02-16 12:56:22.000000000 -0400
  2357. @@ -1,8 +1,6 @@
  2358. /*
  2359. * IPVS: Never Queue scheduling module
  2360. *
  2361. - * Version: $Id: ip_vs_nq.c,v 1.2 2003/06/08 09:31:19 wensong Exp $
  2362. - *
  2363. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  2364. *
  2365. * This program is free software; you can redistribute it and/or
  2366. @@ -138,6 +136,7 @@
  2367. .name = "nq",
  2368. .refcnt = ATOMIC_INIT(0),
  2369. .module = THIS_MODULE,
  2370. + .n_list = LIST_HEAD_INIT(ip_vs_nq_scheduler.n_list),
  2371. .init_service = ip_vs_nq_init_svc,
  2372. .done_service = ip_vs_nq_done_svc,
  2373. .update_service = ip_vs_nq_update_svc,
  2374. @@ -147,7 +146,6 @@
  2375.  
  2376. static int __init ip_vs_nq_init(void)
  2377. {
  2378. - INIT_LIST_HEAD(&ip_vs_nq_scheduler.n_list);
  2379. return register_ip_vs_scheduler(&ip_vs_nq_scheduler);
  2380. }
  2381.  
  2382. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_proto.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_proto.c
  2383. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_proto.c 2009-02-16 11:57:22.000000000 -0400
  2384. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_proto.c 2009-02-16 12:56:22.000000000 -0400
  2385. @@ -1,8 +1,6 @@
  2386. /*
  2387. * ip_vs_proto.c: transport protocol load balancing support for IPVS
  2388. *
  2389. - * Version: $Id: ip_vs_proto.c,v 1.2 2003/04/18 09:03:16 wensong Exp $
  2390. - *
  2391. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  2392. * Julian Anastasov <ja@ssi.bg>
  2393. *
  2394. @@ -45,7 +43,7 @@
  2395. /*
  2396. * register an ipvs protocol
  2397. */
  2398. -static int register_ip_vs_protocol(struct ip_vs_protocol *pp)
  2399. +static int __used __init register_ip_vs_protocol(struct ip_vs_protocol *pp)
  2400. {
  2401. unsigned hash = IP_VS_PROTO_HASH(pp->protocol);
  2402.  
  2403. @@ -148,7 +146,7 @@
  2404. struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
  2405.  
  2406. if (pp == NULL || pp->state_name == NULL)
  2407. - return "ERR!";
  2408. + return (IPPROTO_IP == proto) ? "NONE" : "ERR!";
  2409. return pp->state_name(state);
  2410. }
  2411.  
  2412. @@ -165,7 +163,7 @@
  2413. ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
  2414. if (ih == NULL)
  2415. sprintf(buf, "%s TRUNCATED", pp->name);
  2416. - else if (ih->frag_off & __constant_htons(IP_OFFSET))
  2417. + else if (ih->frag_off & htons(IP_OFFSET))
  2418. sprintf(buf, "%s %u.%u.%u.%u->%u.%u.%u.%u frag",
  2419. pp->name, NIPQUAD(ih->saddr),
  2420. NIPQUAD(ih->daddr));
  2421. @@ -192,7 +190,7 @@
  2422. }
  2423.  
  2424.  
  2425. -int ip_vs_protocol_init(void)
  2426. +int __init ip_vs_protocol_init(void)
  2427. {
  2428. char protocols[64];
  2429. #define REGISTER_PROTOCOL(p) \
  2430. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_proto_ah.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_proto_ah.c
  2431. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_proto_ah.c 2009-02-16 11:57:22.000000000 -0400
  2432. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_proto_ah.c 2009-02-16 12:56:22.000000000 -0400
  2433. @@ -1,8 +1,6 @@
  2434. /*
  2435. * ip_vs_proto_ah.c: AH IPSec load balancing support for IPVS
  2436. *
  2437. - * Version: $Id: ip_vs_proto_ah.c,v 1.1 2003/07/04 15:04:37 wensong Exp $
  2438. - *
  2439. * Authors: Julian Anastasov <ja@ssi.bg>, February 2002
  2440. * Wensong Zhang <wensong@linuxvirtualserver.org>
  2441. *
  2442. @@ -52,15 +50,15 @@
  2443. if (likely(!inverse)) {
  2444. cp = ip_vs_conn_in_get(IPPROTO_UDP,
  2445. iph->saddr,
  2446. - __constant_htons(PORT_ISAKMP),
  2447. + htons(PORT_ISAKMP),
  2448. iph->daddr,
  2449. - __constant_htons(PORT_ISAKMP));
  2450. + htons(PORT_ISAKMP));
  2451. } else {
  2452. cp = ip_vs_conn_in_get(IPPROTO_UDP,
  2453. iph->daddr,
  2454. - __constant_htons(PORT_ISAKMP),
  2455. + htons(PORT_ISAKMP),
  2456. iph->saddr,
  2457. - __constant_htons(PORT_ISAKMP));
  2458. + htons(PORT_ISAKMP));
  2459. }
  2460.  
  2461. if (!cp) {
  2462. @@ -89,15 +87,15 @@
  2463. if (likely(!inverse)) {
  2464. cp = ip_vs_conn_out_get(IPPROTO_UDP,
  2465. iph->saddr,
  2466. - __constant_htons(PORT_ISAKMP),
  2467. + htons(PORT_ISAKMP),
  2468. iph->daddr,
  2469. - __constant_htons(PORT_ISAKMP));
  2470. + htons(PORT_ISAKMP));
  2471. } else {
  2472. cp = ip_vs_conn_out_get(IPPROTO_UDP,
  2473. iph->daddr,
  2474. - __constant_htons(PORT_ISAKMP),
  2475. + htons(PORT_ISAKMP),
  2476. iph->saddr,
  2477. - __constant_htons(PORT_ISAKMP));
  2478. + htons(PORT_ISAKMP));
  2479. }
  2480.  
  2481. if (!cp) {
  2482. @@ -160,6 +158,7 @@
  2483. struct ip_vs_protocol ip_vs_protocol_ah = {
  2484. .name = "AH",
  2485. .protocol = IPPROTO_AH,
  2486. + .num_states = 1,
  2487. .dont_defrag = 1,
  2488. .init = ah_init,
  2489. .exit = ah_exit,
  2490. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_proto_esp.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_proto_esp.c
  2491. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_proto_esp.c 2009-02-16 11:57:22.000000000 -0400
  2492. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_proto_esp.c 2009-02-16 12:56:22.000000000 -0400
  2493. @@ -1,8 +1,6 @@
  2494. /*
  2495. * ip_vs_proto_esp.c: ESP IPSec load balancing support for IPVS
  2496. *
  2497. - * Version: $Id: ip_vs_proto_esp.c,v 1.1 2003/07/04 15:04:37 wensong Exp $
  2498. - *
  2499. * Authors: Julian Anastasov <ja@ssi.bg>, February 2002
  2500. * Wensong Zhang <wensong@linuxvirtualserver.org>
  2501. *
  2502. @@ -52,15 +50,15 @@
  2503. if (likely(!inverse)) {
  2504. cp = ip_vs_conn_in_get(IPPROTO_UDP,
  2505. iph->saddr,
  2506. - __constant_htons(PORT_ISAKMP),
  2507. + htons(PORT_ISAKMP),
  2508. iph->daddr,
  2509. - __constant_htons(PORT_ISAKMP));
  2510. + htons(PORT_ISAKMP));
  2511. } else {
  2512. cp = ip_vs_conn_in_get(IPPROTO_UDP,
  2513. iph->daddr,
  2514. - __constant_htons(PORT_ISAKMP),
  2515. + htons(PORT_ISAKMP),
  2516. iph->saddr,
  2517. - __constant_htons(PORT_ISAKMP));
  2518. + htons(PORT_ISAKMP));
  2519. }
  2520.  
  2521. if (!cp) {
  2522. @@ -89,15 +87,15 @@
  2523. if (likely(!inverse)) {
  2524. cp = ip_vs_conn_out_get(IPPROTO_UDP,
  2525. iph->saddr,
  2526. - __constant_htons(PORT_ISAKMP),
  2527. + htons(PORT_ISAKMP),
  2528. iph->daddr,
  2529. - __constant_htons(PORT_ISAKMP));
  2530. + htons(PORT_ISAKMP));
  2531. } else {
  2532. cp = ip_vs_conn_out_get(IPPROTO_UDP,
  2533. iph->daddr,
  2534. - __constant_htons(PORT_ISAKMP),
  2535. + htons(PORT_ISAKMP),
  2536. iph->saddr,
  2537. - __constant_htons(PORT_ISAKMP));
  2538. + htons(PORT_ISAKMP));
  2539. }
  2540.  
  2541. if (!cp) {
  2542. @@ -159,6 +157,7 @@
  2543. struct ip_vs_protocol ip_vs_protocol_esp = {
  2544. .name = "ESP",
  2545. .protocol = IPPROTO_ESP,
  2546. + .num_states = 1,
  2547. .dont_defrag = 1,
  2548. .init = esp_init,
  2549. .exit = esp_exit,
  2550. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_proto_tcp.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_proto_tcp.c
  2551. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_proto_tcp.c 2009-02-16 11:57:22.000000000 -0400
  2552. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_proto_tcp.c 2009-02-16 12:56:22.000000000 -0400
  2553. @@ -1,8 +1,6 @@
  2554. /*
  2555. * ip_vs_proto_tcp.c: TCP load balancing support for IPVS
  2556. *
  2557. - * Version: $Id: ip_vs_proto_tcp.c,v 1.3 2002/11/30 01:50:35 wensong Exp $
  2558. - *
  2559. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  2560. * Julian Anastasov <ja@ssi.bg>
  2561. *
  2562. @@ -20,6 +18,7 @@
  2563. #include <linux/tcp.h> /* for tcphdr */
  2564. #include <net/ip.h>
  2565. #include <net/tcp.h> /* for csum_tcpudp_magic */
  2566. +#include <linux/netfilter.h>
  2567. #include <linux/netfilter_ipv4.h>
  2568.  
  2569. #include <net/ip_vs.h>
  2570. @@ -76,16 +75,15 @@
  2571. struct ip_vs_service *svc;
  2572. struct tcphdr _tcph, *th;
  2573.  
  2574. - th = skb_header_pointer(skb, skb->nh.iph->ihl*4,
  2575. - sizeof(_tcph), &_tcph);
  2576. + th = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_tcph), &_tcph);
  2577. if (th == NULL) {
  2578. *verdict = NF_DROP;
  2579. return 0;
  2580. }
  2581.  
  2582. if (th->syn &&
  2583. - (svc = ip_vs_service_get(skb->mark, skb->nh.iph->protocol,
  2584. - skb->nh.iph->daddr, th->dest))) {
  2585. + (svc = ip_vs_service_get(skb->mark, ip_hdr(skb)->protocol,
  2586. + ip_hdr(skb)->daddr, th->dest))) {
  2587. if (ip_vs_todrop()) {
  2588. /*
  2589. * It seems that we are very loaded.
  2590. @@ -123,27 +121,27 @@
  2591.  
  2592.  
  2593. static int
  2594. -tcp_snat_handler(struct sk_buff **pskb,
  2595. +tcp_snat_handler(struct sk_buff *skb,
  2596. struct ip_vs_protocol *pp, struct ip_vs_conn *cp)
  2597. {
  2598. struct tcphdr *tcph;
  2599. - unsigned int tcphoff = (*pskb)->nh.iph->ihl * 4;
  2600. + const unsigned int tcphoff = ip_hdrlen(skb);
  2601.  
  2602. /* csum_check requires unshared skb */
  2603. - if (!ip_vs_make_skb_writable(pskb, tcphoff+sizeof(*tcph)))
  2604. + if (!skb_make_writable(skb, tcphoff+sizeof(*tcph)))
  2605. return 0;
  2606.  
  2607. if (unlikely(cp->app != NULL)) {
  2608. /* Some checks before mangling */
  2609. - if (pp->csum_check && !pp->csum_check(*pskb, pp))
  2610. + if (pp->csum_check && !pp->csum_check(skb, pp))
  2611. return 0;
  2612.  
  2613. /* Call application helper if needed */
  2614. - if (!ip_vs_app_pkt_out(cp, pskb))
  2615. + if (!ip_vs_app_pkt_out(cp, skb))
  2616. return 0;
  2617. }
  2618.  
  2619. - tcph = (void *)(*pskb)->nh.iph + tcphoff;
  2620. + tcph = (void *)ip_hdr(skb) + tcphoff;
  2621. tcph->source = cp->vport;
  2622.  
  2623. /* Adjust TCP checksums */
  2624. @@ -151,17 +149,15 @@
  2625. /* Only port and addr are changed, do fast csum update */
  2626. tcp_fast_csum_update(tcph, cp->daddr, cp->vaddr,
  2627. cp->dport, cp->vport);
  2628. - if ((*pskb)->ip_summed == CHECKSUM_COMPLETE)
  2629. - (*pskb)->ip_summed = CHECKSUM_NONE;
  2630. + if (skb->ip_summed == CHECKSUM_COMPLETE)
  2631. + skb->ip_summed = CHECKSUM_NONE;
  2632. } else {
  2633. /* full checksum calculation */
  2634. tcph->check = 0;
  2635. - (*pskb)->csum = skb_checksum(*pskb, tcphoff,
  2636. - (*pskb)->len - tcphoff, 0);
  2637. + skb->csum = skb_checksum(skb, tcphoff, skb->len - tcphoff, 0);
  2638. tcph->check = csum_tcpudp_magic(cp->vaddr, cp->caddr,
  2639. - (*pskb)->len - tcphoff,
  2640. - cp->protocol,
  2641. - (*pskb)->csum);
  2642. + skb->len - tcphoff,
  2643. + cp->protocol, skb->csum);
  2644. IP_VS_DBG(11, "O-pkt: %s O-csum=%d (+%zd)\n",
  2645. pp->name, tcph->check,
  2646. (char*)&(tcph->check) - (char*)tcph);
  2647. @@ -171,30 +167,30 @@
  2648.  
  2649.  
  2650. static int
  2651. -tcp_dnat_handler(struct sk_buff **pskb,
  2652. +tcp_dnat_handler(struct sk_buff *skb,
  2653. struct ip_vs_protocol *pp, struct ip_vs_conn *cp)
  2654. {
  2655. struct tcphdr *tcph;
  2656. - unsigned int tcphoff = (*pskb)->nh.iph->ihl * 4;
  2657. + const unsigned int tcphoff = ip_hdrlen(skb);
  2658.  
  2659. /* csum_check requires unshared skb */
  2660. - if (!ip_vs_make_skb_writable(pskb, tcphoff+sizeof(*tcph)))
  2661. + if (!skb_make_writable(skb, tcphoff+sizeof(*tcph)))
  2662. return 0;
  2663.  
  2664. if (unlikely(cp->app != NULL)) {
  2665. /* Some checks before mangling */
  2666. - if (pp->csum_check && !pp->csum_check(*pskb, pp))
  2667. + if (pp->csum_check && !pp->csum_check(skb, pp))
  2668. return 0;
  2669.  
  2670. /*
  2671. * Attempt ip_vs_app call.
  2672. * It will fix ip_vs_conn and iph ack_seq stuff
  2673. */
  2674. - if (!ip_vs_app_pkt_in(cp, pskb))
  2675. + if (!ip_vs_app_pkt_in(cp, skb))
  2676. return 0;
  2677. }
  2678.  
  2679. - tcph = (void *)(*pskb)->nh.iph + tcphoff;
  2680. + tcph = (void *)ip_hdr(skb) + tcphoff;
  2681. tcph->dest = cp->dport;
  2682.  
  2683. /*
  2684. @@ -204,18 +200,16 @@
  2685. /* Only port and addr are changed, do fast csum update */
  2686. tcp_fast_csum_update(tcph, cp->vaddr, cp->daddr,
  2687. cp->vport, cp->dport);
  2688. - if ((*pskb)->ip_summed == CHECKSUM_COMPLETE)
  2689. - (*pskb)->ip_summed = CHECKSUM_NONE;
  2690. + if (skb->ip_summed == CHECKSUM_COMPLETE)
  2691. + skb->ip_summed = CHECKSUM_NONE;
  2692. } else {
  2693. /* full checksum calculation */
  2694. tcph->check = 0;
  2695. - (*pskb)->csum = skb_checksum(*pskb, tcphoff,
  2696. - (*pskb)->len - tcphoff, 0);
  2697. + skb->csum = skb_checksum(skb, tcphoff, skb->len - tcphoff, 0);
  2698. tcph->check = csum_tcpudp_magic(cp->caddr, cp->daddr,
  2699. - (*pskb)->len - tcphoff,
  2700. - cp->protocol,
  2701. - (*pskb)->csum);
  2702. - (*pskb)->ip_summed = CHECKSUM_UNNECESSARY;
  2703. + skb->len - tcphoff,
  2704. + cp->protocol, skb->csum);
  2705. + skb->ip_summed = CHECKSUM_UNNECESSARY;
  2706. }
  2707. return 1;
  2708. }
  2709. @@ -224,15 +218,15 @@
  2710. static int
  2711. tcp_csum_check(struct sk_buff *skb, struct ip_vs_protocol *pp)
  2712. {
  2713. - unsigned int tcphoff = skb->nh.iph->ihl*4;
  2714. + const unsigned int tcphoff = ip_hdrlen(skb);
  2715.  
  2716. switch (skb->ip_summed) {
  2717. case CHECKSUM_NONE:
  2718. skb->csum = skb_checksum(skb, tcphoff, skb->len - tcphoff, 0);
  2719. case CHECKSUM_COMPLETE:
  2720. - if (csum_tcpudp_magic(skb->nh.iph->saddr, skb->nh.iph->daddr,
  2721. + if (csum_tcpudp_magic(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
  2722. skb->len - tcphoff,
  2723. - skb->nh.iph->protocol, skb->csum)) {
  2724. + ip_hdr(skb)->protocol, skb->csum)) {
  2725. IP_VS_DBG_RL_PKT(0, pp, skb, 0,
  2726. "Failed checksum for");
  2727. return 0;
  2728. @@ -467,8 +461,7 @@
  2729. {
  2730. struct tcphdr _tcph, *th;
  2731.  
  2732. - th = skb_header_pointer(skb, skb->nh.iph->ihl*4,
  2733. - sizeof(_tcph), &_tcph);
  2734. + th = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_tcph), &_tcph);
  2735. if (th == NULL)
  2736. return 0;
  2737.  
  2738. @@ -555,7 +548,7 @@
  2739.  
  2740. IP_VS_DBG(9, "%s: Binding conn %u.%u.%u.%u:%u->"
  2741. "%u.%u.%u.%u:%u to app %s on port %u\n",
  2742. - __FUNCTION__,
  2743. + __func__,
  2744. NIPQUAD(cp->caddr), ntohs(cp->cport),
  2745. NIPQUAD(cp->vaddr), ntohs(cp->vport),
  2746. inc->name, ntohs(inc->port));
  2747. @@ -599,6 +592,7 @@
  2748. struct ip_vs_protocol ip_vs_protocol_tcp = {
  2749. .name = "TCP",
  2750. .protocol = IPPROTO_TCP,
  2751. + .num_states = IP_VS_TCP_S_LAST,
  2752. .dont_defrag = 0,
  2753. .appcnt = ATOMIC_INIT(0),
  2754. .init = ip_vs_tcp_init,
  2755. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_proto_udp.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_proto_udp.c
  2756. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_proto_udp.c 2009-02-16 11:57:22.000000000 -0400
  2757. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_proto_udp.c 2009-02-16 12:56:22.000000000 -0400
  2758. @@ -1,8 +1,6 @@
  2759. /*
  2760. * ip_vs_proto_udp.c: UDP load balancing support for IPVS
  2761. *
  2762. - * Version: $Id: ip_vs_proto_udp.c,v 1.3 2002/11/30 01:50:35 wensong Exp $
  2763. - *
  2764. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  2765. * Julian Anastasov <ja@ssi.bg>
  2766. *
  2767. @@ -18,11 +16,12 @@
  2768. #include <linux/in.h>
  2769. #include <linux/ip.h>
  2770. #include <linux/kernel.h>
  2771. +#include <linux/netfilter.h>
  2772. #include <linux/netfilter_ipv4.h>
  2773. #include <linux/udp.h>
  2774.  
  2775. #include <net/ip_vs.h>
  2776. -
  2777. +#include <net/ip.h>
  2778.  
  2779. static struct ip_vs_conn *
  2780. udp_conn_in_get(const struct sk_buff *skb, struct ip_vs_protocol *pp,
  2781. @@ -56,7 +55,7 @@
  2782. struct ip_vs_conn *cp;
  2783. __be16 _ports[2], *pptr;
  2784.  
  2785. - pptr = skb_header_pointer(skb, skb->nh.iph->ihl*4,
  2786. + pptr = skb_header_pointer(skb, ip_hdrlen(skb),
  2787. sizeof(_ports), _ports);
  2788. if (pptr == NULL)
  2789. return NULL;
  2790. @@ -82,15 +81,15 @@
  2791. struct ip_vs_service *svc;
  2792. struct udphdr _udph, *uh;
  2793.  
  2794. - uh = skb_header_pointer(skb, skb->nh.iph->ihl*4,
  2795. + uh = skb_header_pointer(skb, ip_hdrlen(skb),
  2796. sizeof(_udph), &_udph);
  2797. if (uh == NULL) {
  2798. *verdict = NF_DROP;
  2799. return 0;
  2800. }
  2801.  
  2802. - if ((svc = ip_vs_service_get(skb->mark, skb->nh.iph->protocol,
  2803. - skb->nh.iph->daddr, uh->dest))) {
  2804. + if ((svc = ip_vs_service_get(skb->mark, ip_hdr(skb)->protocol,
  2805. + ip_hdr(skb)->daddr, uh->dest))) {
  2806. if (ip_vs_todrop()) {
  2807. /*
  2808. * It seems that we are very loaded.
  2809. @@ -129,29 +128,29 @@
  2810. }
  2811.  
  2812. static int
  2813. -udp_snat_handler(struct sk_buff **pskb,
  2814. +udp_snat_handler(struct sk_buff *skb,
  2815. struct ip_vs_protocol *pp, struct ip_vs_conn *cp)
  2816. {
  2817. struct udphdr *udph;
  2818. - unsigned int udphoff = (*pskb)->nh.iph->ihl * 4;
  2819. + const unsigned int udphoff = ip_hdrlen(skb);
  2820.  
  2821. /* csum_check requires unshared skb */
  2822. - if (!ip_vs_make_skb_writable(pskb, udphoff+sizeof(*udph)))
  2823. + if (!skb_make_writable(skb, udphoff+sizeof(*udph)))
  2824. return 0;
  2825.  
  2826. if (unlikely(cp->app != NULL)) {
  2827. /* Some checks before mangling */
  2828. - if (pp->csum_check && !pp->csum_check(*pskb, pp))
  2829. + if (pp->csum_check && !pp->csum_check(skb, pp))
  2830. return 0;
  2831.  
  2832. /*
  2833. * Call application helper if needed
  2834. */
  2835. - if (!ip_vs_app_pkt_out(cp, pskb))
  2836. + if (!ip_vs_app_pkt_out(cp, skb))
  2837. return 0;
  2838. }
  2839.  
  2840. - udph = (void *)(*pskb)->nh.iph + udphoff;
  2841. + udph = (void *)ip_hdr(skb) + udphoff;
  2842. udph->source = cp->vport;
  2843.  
  2844. /*
  2845. @@ -161,17 +160,15 @@
  2846. /* Only port and addr are changed, do fast csum update */
  2847. udp_fast_csum_update(udph, cp->daddr, cp->vaddr,
  2848. cp->dport, cp->vport);
  2849. - if ((*pskb)->ip_summed == CHECKSUM_COMPLETE)
  2850. - (*pskb)->ip_summed = CHECKSUM_NONE;
  2851. + if (skb->ip_summed == CHECKSUM_COMPLETE)
  2852. + skb->ip_summed = CHECKSUM_NONE;
  2853. } else {
  2854. /* full checksum calculation */
  2855. udph->check = 0;
  2856. - (*pskb)->csum = skb_checksum(*pskb, udphoff,
  2857. - (*pskb)->len - udphoff, 0);
  2858. + skb->csum = skb_checksum(skb, udphoff, skb->len - udphoff, 0);
  2859. udph->check = csum_tcpudp_magic(cp->vaddr, cp->caddr,
  2860. - (*pskb)->len - udphoff,
  2861. - cp->protocol,
  2862. - (*pskb)->csum);
  2863. + skb->len - udphoff,
  2864. + cp->protocol, skb->csum);
  2865. if (udph->check == 0)
  2866. udph->check = CSUM_MANGLED_0;
  2867. IP_VS_DBG(11, "O-pkt: %s O-csum=%d (+%zd)\n",
  2868. @@ -183,30 +180,30 @@
  2869.  
  2870.  
  2871. static int
  2872. -udp_dnat_handler(struct sk_buff **pskb,
  2873. +udp_dnat_handler(struct sk_buff *skb,
  2874. struct ip_vs_protocol *pp, struct ip_vs_conn *cp)
  2875. {
  2876. struct udphdr *udph;
  2877. - unsigned int udphoff = (*pskb)->nh.iph->ihl * 4;
  2878. + unsigned int udphoff = ip_hdrlen(skb);
  2879.  
  2880. /* csum_check requires unshared skb */
  2881. - if (!ip_vs_make_skb_writable(pskb, udphoff+sizeof(*udph)))
  2882. + if (!skb_make_writable(skb, udphoff+sizeof(*udph)))
  2883. return 0;
  2884.  
  2885. if (unlikely(cp->app != NULL)) {
  2886. /* Some checks before mangling */
  2887. - if (pp->csum_check && !pp->csum_check(*pskb, pp))
  2888. + if (pp->csum_check && !pp->csum_check(skb, pp))
  2889. return 0;
  2890.  
  2891. /*
  2892. * Attempt ip_vs_app call.
  2893. * It will fix ip_vs_conn
  2894. */
  2895. - if (!ip_vs_app_pkt_in(cp, pskb))
  2896. + if (!ip_vs_app_pkt_in(cp, skb))
  2897. return 0;
  2898. }
  2899.  
  2900. - udph = (void *)(*pskb)->nh.iph + udphoff;
  2901. + udph = (void *)ip_hdr(skb) + udphoff;
  2902. udph->dest = cp->dport;
  2903.  
  2904. /*
  2905. @@ -216,20 +213,18 @@
  2906. /* Only port and addr are changed, do fast csum update */
  2907. udp_fast_csum_update(udph, cp->vaddr, cp->daddr,
  2908. cp->vport, cp->dport);
  2909. - if ((*pskb)->ip_summed == CHECKSUM_COMPLETE)
  2910. - (*pskb)->ip_summed = CHECKSUM_NONE;
  2911. + if (skb->ip_summed == CHECKSUM_COMPLETE)
  2912. + skb->ip_summed = CHECKSUM_NONE;
  2913. } else {
  2914. /* full checksum calculation */
  2915. udph->check = 0;
  2916. - (*pskb)->csum = skb_checksum(*pskb, udphoff,
  2917. - (*pskb)->len - udphoff, 0);
  2918. + skb->csum = skb_checksum(skb, udphoff, skb->len - udphoff, 0);
  2919. udph->check = csum_tcpudp_magic(cp->caddr, cp->daddr,
  2920. - (*pskb)->len - udphoff,
  2921. - cp->protocol,
  2922. - (*pskb)->csum);
  2923. + skb->len - udphoff,
  2924. + cp->protocol, skb->csum);
  2925. if (udph->check == 0)
  2926. udph->check = CSUM_MANGLED_0;
  2927. - (*pskb)->ip_summed = CHECKSUM_UNNECESSARY;
  2928. + skb->ip_summed = CHECKSUM_UNNECESSARY;
  2929. }
  2930. return 1;
  2931. }
  2932. @@ -239,7 +234,7 @@
  2933. udp_csum_check(struct sk_buff *skb, struct ip_vs_protocol *pp)
  2934. {
  2935. struct udphdr _udph, *uh;
  2936. - unsigned int udphoff = skb->nh.iph->ihl*4;
  2937. + const unsigned int udphoff = ip_hdrlen(skb);
  2938.  
  2939. uh = skb_header_pointer(skb, udphoff, sizeof(_udph), &_udph);
  2940. if (uh == NULL)
  2941. @@ -251,10 +246,10 @@
  2942. skb->csum = skb_checksum(skb, udphoff,
  2943. skb->len - udphoff, 0);
  2944. case CHECKSUM_COMPLETE:
  2945. - if (csum_tcpudp_magic(skb->nh.iph->saddr,
  2946. - skb->nh.iph->daddr,
  2947. + if (csum_tcpudp_magic(ip_hdr(skb)->saddr,
  2948. + ip_hdr(skb)->daddr,
  2949. skb->len - udphoff,
  2950. - skb->nh.iph->protocol,
  2951. + ip_hdr(skb)->protocol,
  2952. skb->csum)) {
  2953. IP_VS_DBG_RL_PKT(0, pp, skb, 0,
  2954. "Failed checksum for");
  2955. @@ -347,7 +342,7 @@
  2956.  
  2957. IP_VS_DBG(9, "%s: Binding conn %u.%u.%u.%u:%u->"
  2958. "%u.%u.%u.%u:%u to app %s on port %u\n",
  2959. - __FUNCTION__,
  2960. + __func__,
  2961. NIPQUAD(cp->caddr), ntohs(cp->cport),
  2962. NIPQUAD(cp->vaddr), ntohs(cp->vport),
  2963. inc->name, ntohs(inc->port));
  2964. @@ -412,6 +407,7 @@
  2965. struct ip_vs_protocol ip_vs_protocol_udp = {
  2966. .name = "UDP",
  2967. .protocol = IPPROTO_UDP,
  2968. + .num_states = IP_VS_UDP_S_LAST,
  2969. .dont_defrag = 0,
  2970. .init = udp_init,
  2971. .exit = udp_exit,
  2972. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_rr.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_rr.c
  2973. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_rr.c 2009-02-16 11:57:22.000000000 -0400
  2974. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_rr.c 2009-02-16 12:56:22.000000000 -0400
  2975. @@ -1,8 +1,6 @@
  2976. /*
  2977. * IPVS: Round-Robin Scheduling module
  2978. *
  2979. - * Version: $Id: ip_vs_rr.c,v 1.9 2002/09/15 08:14:08 wensong Exp $
  2980. - *
  2981. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  2982. * Peter Kese <peter.kese@ijs.si>
  2983. *
  2984. @@ -68,7 +66,7 @@
  2985. q = q->next;
  2986. continue;
  2987. }
  2988. -
  2989. +
  2990. dest = list_entry(q, struct ip_vs_dest, n_list);
  2991. if (!(dest->flags & IP_VS_DEST_F_OVERLOAD) &&
  2992. atomic_read(&dest->weight) > 0)
  2993. @@ -96,6 +94,7 @@
  2994. .name = "rr", /* name */
  2995. .refcnt = ATOMIC_INIT(0),
  2996. .module = THIS_MODULE,
  2997. + .n_list = LIST_HEAD_INIT(ip_vs_rr_scheduler.n_list),
  2998. .init_service = ip_vs_rr_init_svc,
  2999. .done_service = ip_vs_rr_done_svc,
  3000. .update_service = ip_vs_rr_update_svc,
  3001. @@ -104,7 +103,6 @@
  3002.  
  3003. static int __init ip_vs_rr_init(void)
  3004. {
  3005. - INIT_LIST_HEAD(&ip_vs_rr_scheduler.n_list);
  3006. return register_ip_vs_scheduler(&ip_vs_rr_scheduler);
  3007. }
  3008.  
  3009. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_sched.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_sched.c
  3010. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_sched.c 2009-02-16 11:57:22.000000000 -0400
  3011. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_sched.c 2009-02-16 12:56:22.000000000 -0400
  3012. @@ -5,8 +5,6 @@
  3013. * high-performance and highly available server based on a
  3014. * cluster of servers.
  3015. *
  3016. - * Version: $Id: ip_vs_sched.c,v 1.13 2003/05/10 03:05:23 wensong Exp $
  3017. - *
  3018. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  3019. * Peter Kese <peter.kese@ijs.si>
  3020. *
  3021. @@ -20,11 +18,11 @@
  3022. */
  3023.  
  3024. #include <linux/module.h>
  3025. -#include <linux/sched.h>
  3026. #include <linux/spinlock.h>
  3027. #include <linux/interrupt.h>
  3028. #include <asm/string.h>
  3029. #include <linux/kmod.h>
  3030. +#include <linux/sysctl.h>
  3031.  
  3032. #include <net/ip_vs.h>
  3033.  
  3034. @@ -184,22 +182,9 @@
  3035. /* increase the module use count */
  3036. ip_vs_use_count_inc();
  3037.  
  3038. - /*
  3039. - * Make sure that the scheduler with this name doesn't exist
  3040. - * in the scheduler list.
  3041. - */
  3042. - sched = ip_vs_sched_getbyname(scheduler->name);
  3043. - if (sched) {
  3044. - ip_vs_scheduler_put(sched);
  3045. - ip_vs_use_count_dec();
  3046. - IP_VS_ERR("register_ip_vs_scheduler(): [%s] scheduler "
  3047. - "already existed in the system\n", scheduler->name);
  3048. - return -EINVAL;
  3049. - }
  3050. -
  3051. write_lock_bh(&__ip_vs_sched_lock);
  3052.  
  3053. - if (scheduler->n_list.next != &scheduler->n_list) {
  3054. + if (!list_empty(&scheduler->n_list)) {
  3055. write_unlock_bh(&__ip_vs_sched_lock);
  3056. ip_vs_use_count_dec();
  3057. IP_VS_ERR("register_ip_vs_scheduler(): [%s] scheduler "
  3058. @@ -208,6 +193,20 @@
  3059. }
  3060.  
  3061. /*
  3062. + * Make sure that the scheduler with this name doesn't exist
  3063. + * in the scheduler list.
  3064. + */
  3065. + list_for_each_entry(sched, &ip_vs_schedulers, n_list) {
  3066. + if (strcmp(scheduler->name, sched->name) == 0) {
  3067. + write_unlock_bh(&__ip_vs_sched_lock);
  3068. + ip_vs_use_count_dec();
  3069. + IP_VS_ERR("register_ip_vs_scheduler(): [%s] scheduler "
  3070. + "already existed in the system\n",
  3071. + scheduler->name);
  3072. + return -EINVAL;
  3073. + }
  3074. + }
  3075. + /*
  3076. * Add it into the d-linked scheduler list
  3077. */
  3078. list_add(&scheduler->n_list, &ip_vs_schedulers);
  3079. @@ -230,7 +229,7 @@
  3080. }
  3081.  
  3082. write_lock_bh(&__ip_vs_sched_lock);
  3083. - if (scheduler->n_list.next == &scheduler->n_list) {
  3084. + if (list_empty(&scheduler->n_list)) {
  3085. write_unlock_bh(&__ip_vs_sched_lock);
  3086. IP_VS_ERR("unregister_ip_vs_scheduler(): [%s] scheduler "
  3087. "is not in the list. failed\n", scheduler->name);
  3088. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_sed.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_sed.c
  3089. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_sed.c 2009-02-16 11:57:22.000000000 -0400
  3090. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_sed.c 2009-02-16 12:56:22.000000000 -0400
  3091. @@ -1,8 +1,6 @@
  3092. /*
  3093. * IPVS: Shortest Expected Delay scheduling module
  3094. *
  3095. - * Version: $Id: ip_vs_sed.c,v 1.1 2003/05/10 03:06:08 wensong Exp $
  3096. - *
  3097. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  3098. *
  3099. * This program is free software; you can redistribute it and/or
  3100. @@ -18,7 +16,7 @@
  3101. * The SED algorithm attempts to minimize each job's expected delay until
  3102. * completion. The expected delay that the job will experience is
  3103. * (Ci + 1) / Ui if sent to the ith server, in which Ci is the number of
  3104. - * jobs on the the ith server and Ui is the fixed service rate (weight) of
  3105. + * jobs on the ith server and Ui is the fixed service rate (weight) of
  3106. * the ith server. The SED algorithm adopts a greedy policy that each does
  3107. * what is in its own best interest, i.e. to join the queue which would
  3108. * minimize its expected delay of completion.
  3109. @@ -140,6 +138,7 @@
  3110. .name = "sed",
  3111. .refcnt = ATOMIC_INIT(0),
  3112. .module = THIS_MODULE,
  3113. + .n_list = LIST_HEAD_INIT(ip_vs_sed_scheduler.n_list),
  3114. .init_service = ip_vs_sed_init_svc,
  3115. .done_service = ip_vs_sed_done_svc,
  3116. .update_service = ip_vs_sed_update_svc,
  3117. @@ -149,7 +148,6 @@
  3118.  
  3119. static int __init ip_vs_sed_init(void)
  3120. {
  3121. - INIT_LIST_HEAD(&ip_vs_sed_scheduler.n_list);
  3122. return register_ip_vs_scheduler(&ip_vs_sed_scheduler);
  3123. }
  3124.  
  3125. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_sh.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_sh.c
  3126. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_sh.c 2009-02-16 11:57:22.000000000 -0400
  3127. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_sh.c 2009-02-16 12:56:22.000000000 -0400
  3128. @@ -1,8 +1,6 @@
  3129. /*
  3130. * IPVS: Source Hashing scheduling module
  3131. *
  3132. - * Version: $Id: ip_vs_sh.c,v 1.5 2002/09/15 08:14:08 wensong Exp $
  3133. - *
  3134. * Authors: Wensong Zhang <wensong@gnuchina.org>
  3135. *
  3136. * This program is free software; you can redistribute it and/or
  3137. @@ -201,7 +199,7 @@
  3138. {
  3139. struct ip_vs_dest *dest;
  3140. struct ip_vs_sh_bucket *tbl;
  3141. - struct iphdr *iph = skb->nh.iph;
  3142. + struct iphdr *iph = ip_hdr(skb);
  3143.  
  3144. IP_VS_DBG(6, "ip_vs_sh_schedule(): Scheduling...\n");
  3145.  
  3146. @@ -232,6 +230,7 @@
  3147. .name = "sh",
  3148. .refcnt = ATOMIC_INIT(0),
  3149. .module = THIS_MODULE,
  3150. + .n_list = LIST_HEAD_INIT(ip_vs_sh_scheduler.n_list),
  3151. .init_service = ip_vs_sh_init_svc,
  3152. .done_service = ip_vs_sh_done_svc,
  3153. .update_service = ip_vs_sh_update_svc,
  3154. @@ -241,7 +240,6 @@
  3155.  
  3156. static int __init ip_vs_sh_init(void)
  3157. {
  3158. - INIT_LIST_HEAD(&ip_vs_sh_scheduler.n_list);
  3159. return register_ip_vs_scheduler(&ip_vs_sh_scheduler);
  3160. }
  3161.  
  3162. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_sync.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_sync.c
  3163. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_sync.c 2009-02-16 11:57:22.000000000 -0400
  3164. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_sync.c 2009-02-16 12:56:22.000000000 -0400
  3165. @@ -5,8 +5,6 @@
  3166. * high-performance and highly available server based on a
  3167. * cluster of servers.
  3168. *
  3169. - * Version: $Id: ip_vs_sync.c,v 1.13 2003/06/08 09:31:19 wensong Exp $
  3170. - *
  3171. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  3172. *
  3173. * ip_vs_sync: sync connection info from master load balancer to backups
  3174. @@ -29,10 +27,12 @@
  3175. #include <linux/in.h>
  3176. #include <linux/igmp.h> /* for ip_mc_join_group */
  3177. #include <linux/udp.h>
  3178. +#include <linux/err.h>
  3179. +#include <linux/kthread.h>
  3180. +#include <linux/wait.h>
  3181.  
  3182. #include <net/ip.h>
  3183. #include <net/sock.h>
  3184. -#include <asm/uaccess.h> /* for get_fs and set_fs */
  3185.  
  3186. #include <net/ip_vs.h>
  3187.  
  3188. @@ -67,7 +67,11 @@
  3189. struct ip_vs_seq out_seq; /* outgoing seq. struct */
  3190. };
  3191.  
  3192. -#define IP_VS_SYNC_CONN_TIMEOUT (3*60*HZ)
  3193. +struct ip_vs_sync_thread_data {
  3194. + struct socket *sock;
  3195. + char *buf;
  3196. +};
  3197. +
  3198. #define SIMPLE_CONN_SIZE (sizeof(struct ip_vs_sync_conn))
  3199. #define FULL_CONN_SIZE \
  3200. (sizeof(struct ip_vs_sync_conn) + sizeof(struct ip_vs_sync_conn_options))
  3201. @@ -136,18 +140,19 @@
  3202. char ip_vs_master_mcast_ifn[IP_VS_IFNAME_MAXLEN];
  3203. char ip_vs_backup_mcast_ifn[IP_VS_IFNAME_MAXLEN];
  3204.  
  3205. -/* multicast addr */
  3206. -static struct sockaddr_in mcast_addr;
  3207. +/* sync daemon tasks */
  3208. +static struct task_struct *sync_master_thread;
  3209. +static struct task_struct *sync_backup_thread;
  3210.  
  3211. +/* multicast addr */
  3212. +static struct sockaddr_in mcast_addr = {
  3213. + .sin_family = AF_INET,
  3214. + .sin_port = __constant_htons(IP_VS_SYNC_PORT),
  3215. + .sin_addr.s_addr = __constant_htonl(IP_VS_SYNC_GROUP),
  3216. +};
  3217.  
  3218. -static inline void sb_queue_tail(struct ip_vs_sync_buff *sb)
  3219. -{
  3220. - spin_lock(&ip_vs_sync_lock);
  3221. - list_add_tail(&sb->list, &ip_vs_sync_queue);
  3222. - spin_unlock(&ip_vs_sync_lock);
  3223. -}
  3224.  
  3225. -static inline struct ip_vs_sync_buff * sb_dequeue(void)
  3226. +static inline struct ip_vs_sync_buff *sb_dequeue(void)
  3227. {
  3228. struct ip_vs_sync_buff *sb;
  3229.  
  3230. @@ -191,6 +196,16 @@
  3231. kfree(sb);
  3232. }
  3233.  
  3234. +static inline void sb_queue_tail(struct ip_vs_sync_buff *sb)
  3235. +{
  3236. + spin_lock(&ip_vs_sync_lock);
  3237. + if (ip_vs_sync_state & IP_VS_STATE_MASTER)
  3238. + list_add_tail(&sb->list, &ip_vs_sync_queue);
  3239. + else
  3240. + ip_vs_sync_buff_release(sb);
  3241. + spin_unlock(&ip_vs_sync_lock);
  3242. +}
  3243. +
  3244. /*
  3245. * Get the current sync buffer if it has been created for more
  3246. * than the specified time or the specified time is zero.
  3247. @@ -279,14 +294,21 @@
  3248. struct ip_vs_sync_conn *s;
  3249. struct ip_vs_sync_conn_options *opt;
  3250. struct ip_vs_conn *cp;
  3251. + struct ip_vs_protocol *pp;
  3252. + struct ip_vs_dest *dest;
  3253. char *p;
  3254. int i;
  3255.  
  3256. + if (buflen < sizeof(struct ip_vs_sync_mesg)) {
  3257. + IP_VS_ERR_RL("sync message header too short\n");
  3258. + return;
  3259. + }
  3260. +
  3261. /* Convert size back to host byte order */
  3262. m->size = ntohs(m->size);
  3263.  
  3264. if (buflen != m->size) {
  3265. - IP_VS_ERR("bogus message\n");
  3266. + IP_VS_ERR_RL("bogus sync message size\n");
  3267. return;
  3268. }
  3269.  
  3270. @@ -299,10 +321,50 @@
  3271.  
  3272. p = (char *)buffer + sizeof(struct ip_vs_sync_mesg);
  3273. for (i=0; i<m->nr_conns; i++) {
  3274. - unsigned flags;
  3275. + unsigned flags, state;
  3276. +
  3277. + if (p + SIMPLE_CONN_SIZE > buffer+buflen) {
  3278. + IP_VS_ERR_RL("bogus conn in sync message\n");
  3279. + return;
  3280. + }
  3281. + s = (struct ip_vs_sync_conn *) p;
  3282. + flags = ntohs(s->flags) | IP_VS_CONN_F_SYNC;
  3283. + flags &= ~IP_VS_CONN_F_HASHED;
  3284. + if (flags & IP_VS_CONN_F_SEQ_MASK) {
  3285. + opt = (struct ip_vs_sync_conn_options *)&s[1];
  3286. + p += FULL_CONN_SIZE;
  3287. + if (p > buffer+buflen) {
  3288. + IP_VS_ERR_RL("bogus conn options in sync message\n");
  3289. + return;
  3290. + }
  3291. + } else {
  3292. + opt = NULL;
  3293. + p += SIMPLE_CONN_SIZE;
  3294. + }
  3295. +
  3296. + state = ntohs(s->state);
  3297. + if (!(flags & IP_VS_CONN_F_TEMPLATE)) {
  3298. + pp = ip_vs_proto_get(s->protocol);
  3299. + if (!pp) {
  3300. + IP_VS_ERR_RL("Unsupported protocol %u in sync msg\n",
  3301. + s->protocol);
  3302. + continue;
  3303. + }
  3304. + if (state >= pp->num_states) {
  3305. + IP_VS_DBG(2, "Invalid %s state %u in sync msg\n",
  3306. + pp->name, state);
  3307. + continue;
  3308. + }
  3309. + } else {
  3310. + /* protocol in templates is not used for state/timeout */
  3311. + pp = NULL;
  3312. + if (state > 0) {
  3313. + IP_VS_DBG(2, "Invalid template state %u in sync msg\n",
  3314. + state);
  3315. + state = 0;
  3316. + }
  3317. + }
  3318.  
  3319. - s = (struct ip_vs_sync_conn *)p;
  3320. - flags = ntohs(s->flags);
  3321. if (!(flags & IP_VS_CONN_F_TEMPLATE))
  3322. cp = ip_vs_conn_in_get(s->protocol,
  3323. s->caddr, s->cport,
  3324. @@ -312,38 +374,69 @@
  3325. s->caddr, s->cport,
  3326. s->vaddr, s->vport);
  3327. if (!cp) {
  3328. + /*
  3329. + * Find the appropriate destination for the connection.
  3330. + * If it is not found the connection will remain unbound
  3331. + * but still handled.
  3332. + */
  3333. + dest = ip_vs_find_dest(s->daddr, s->dport,
  3334. + s->vaddr, s->vport,
  3335. + s->protocol);
  3336. + /* Set the approprite ativity flag */
  3337. + if (s->protocol == IPPROTO_TCP) {
  3338. + if (state != IP_VS_TCP_S_ESTABLISHED)
  3339. + flags |= IP_VS_CONN_F_INACTIVE;
  3340. + else
  3341. + flags &= ~IP_VS_CONN_F_INACTIVE;
  3342. + }
  3343. cp = ip_vs_conn_new(s->protocol,
  3344. s->caddr, s->cport,
  3345. s->vaddr, s->vport,
  3346. s->daddr, s->dport,
  3347. - flags, NULL);
  3348. + flags, dest);
  3349. + if (dest)
  3350. + atomic_dec(&dest->refcnt);
  3351. if (!cp) {
  3352. IP_VS_ERR("ip_vs_conn_new failed\n");
  3353. return;
  3354. }
  3355. - cp->state = ntohs(s->state);
  3356. } else if (!cp->dest) {
  3357. - /* it is an entry created by the synchronization */
  3358. - cp->state = ntohs(s->state);
  3359. - cp->flags = flags | IP_VS_CONN_F_HASHED;
  3360. - } /* Note that we don't touch its state and flags
  3361. - if it is a normal entry. */
  3362. + dest = ip_vs_try_bind_dest(cp);
  3363. + if (dest)
  3364. + atomic_dec(&dest->refcnt);
  3365. + } else if ((cp->dest) && (cp->protocol == IPPROTO_TCP) &&
  3366. + (cp->state != state)) {
  3367. + /* update active/inactive flag for the connection */
  3368. + dest = cp->dest;
  3369. + if (!(cp->flags & IP_VS_CONN_F_INACTIVE) &&
  3370. + (state != IP_VS_TCP_S_ESTABLISHED)) {
  3371. + atomic_dec(&dest->activeconns);
  3372. + atomic_inc(&dest->inactconns);
  3373. + cp->flags |= IP_VS_CONN_F_INACTIVE;
  3374. + } else if ((cp->flags & IP_VS_CONN_F_INACTIVE) &&
  3375. + (state == IP_VS_TCP_S_ESTABLISHED)) {
  3376. + atomic_inc(&dest->activeconns);
  3377. + atomic_dec(&dest->inactconns);
  3378. + cp->flags &= ~IP_VS_CONN_F_INACTIVE;
  3379. + }
  3380. + }
  3381.  
  3382. - if (flags & IP_VS_CONN_F_SEQ_MASK) {
  3383. - opt = (struct ip_vs_sync_conn_options *)&s[1];
  3384. + if (opt)
  3385. memcpy(&cp->in_seq, opt, sizeof(*opt));
  3386. - p += FULL_CONN_SIZE;
  3387. - } else
  3388. - p += SIMPLE_CONN_SIZE;
  3389. -
  3390. atomic_set(&cp->in_pkts, sysctl_ip_vs_sync_threshold[0]);
  3391. - cp->timeout = IP_VS_SYNC_CONN_TIMEOUT;
  3392. + cp->state = state;
  3393. + cp->old_state = cp->state;
  3394. + /*
  3395. + * We can not recover the right timeout for templates
  3396. + * in all cases, we can not find the right fwmark
  3397. + * virtual service. If needed, we can do it for
  3398. + * non-fwmark persistent services.
  3399. + */
  3400. + if (!(flags & IP_VS_CONN_F_TEMPLATE) && pp->timeout_table)
  3401. + cp->timeout = pp->timeout_table[state];
  3402. + else
  3403. + cp->timeout = (3*60*HZ);
  3404. ip_vs_conn_put(cp);
  3405. -
  3406. - if (p > buffer+buflen) {
  3407. - IP_VS_ERR("bogus message\n");
  3408. - return;
  3409. - }
  3410. }
  3411. }
  3412.  
  3413. @@ -382,7 +475,7 @@
  3414. struct net_device *dev;
  3415. struct inet_sock *inet = inet_sk(sk);
  3416.  
  3417. - if ((dev = __dev_get_by_name(ifname)) == NULL)
  3418. + if ((dev = __dev_get_by_name(&init_net, ifname)) == NULL)
  3419. return -ENODEV;
  3420.  
  3421. if (sk->sk_bound_dev_if && dev->ifindex != sk->sk_bound_dev_if)
  3422. @@ -407,7 +500,7 @@
  3423. int num;
  3424.  
  3425. if (sync_state == IP_VS_STATE_MASTER) {
  3426. - if ((dev = __dev_get_by_name(ip_vs_master_mcast_ifn)) == NULL)
  3427. + if ((dev = __dev_get_by_name(&init_net, ip_vs_master_mcast_ifn)) == NULL)
  3428. return -ENODEV;
  3429.  
  3430. num = (dev->mtu - sizeof(struct iphdr) -
  3431. @@ -418,7 +511,7 @@
  3432. IP_VS_DBG(7, "setting the maximum length of sync sending "
  3433. "message %d.\n", sync_send_mesg_maxlen);
  3434. } else if (sync_state == IP_VS_STATE_BACKUP) {
  3435. - if ((dev = __dev_get_by_name(ip_vs_backup_mcast_ifn)) == NULL)
  3436. + if ((dev = __dev_get_by_name(&init_net, ip_vs_backup_mcast_ifn)) == NULL)
  3437. return -ENODEV;
  3438.  
  3439. sync_recv_mesg_maxlen = dev->mtu -
  3440. @@ -446,7 +539,7 @@
  3441. memset(&mreq, 0, sizeof(mreq));
  3442. memcpy(&mreq.imr_multiaddr, addr, sizeof(struct in_addr));
  3443.  
  3444. - if ((dev = __dev_get_by_name(ifname)) == NULL)
  3445. + if ((dev = __dev_get_by_name(&init_net, ifname)) == NULL)
  3446. return -ENODEV;
  3447. if (sk->sk_bound_dev_if && dev->ifindex != sk->sk_bound_dev_if)
  3448. return -EINVAL;
  3449. @@ -467,7 +560,7 @@
  3450. __be32 addr;
  3451. struct sockaddr_in sin;
  3452.  
  3453. - if ((dev = __dev_get_by_name(ifname)) == NULL)
  3454. + if ((dev = __dev_get_by_name(&init_net, ifname)) == NULL)
  3455. return -ENODEV;
  3456.  
  3457. addr = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
  3458. @@ -492,14 +585,17 @@
  3459. static struct socket * make_send_sock(void)
  3460. {
  3461. struct socket *sock;
  3462. + int result;
  3463.  
  3464. /* First create a socket */
  3465. - if (sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock) < 0) {
  3466. + result = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
  3467. + if (result < 0) {
  3468. IP_VS_ERR("Error during creation of socket; terminating\n");
  3469. - return NULL;
  3470. + return ERR_PTR(result);
  3471. }
  3472.  
  3473. - if (set_mcast_if(sock->sk, ip_vs_master_mcast_ifn) < 0) {
  3474. + result = set_mcast_if(sock->sk, ip_vs_master_mcast_ifn);
  3475. + if (result < 0) {
  3476. IP_VS_ERR("Error setting outbound mcast interface\n");
  3477. goto error;
  3478. }
  3479. @@ -507,14 +603,15 @@
  3480. set_mcast_loop(sock->sk, 0);
  3481. set_mcast_ttl(sock->sk, 1);
  3482.  
  3483. - if (bind_mcastif_addr(sock, ip_vs_master_mcast_ifn) < 0) {
  3484. + result = bind_mcastif_addr(sock, ip_vs_master_mcast_ifn);
  3485. + if (result < 0) {
  3486. IP_VS_ERR("Error binding address of the mcast interface\n");
  3487. goto error;
  3488. }
  3489.  
  3490. - if (sock->ops->connect(sock,
  3491. - (struct sockaddr*)&mcast_addr,
  3492. - sizeof(struct sockaddr), 0) < 0) {
  3493. + result = sock->ops->connect(sock, (struct sockaddr *) &mcast_addr,
  3494. + sizeof(struct sockaddr), 0);
  3495. + if (result < 0) {
  3496. IP_VS_ERR("Error connecting to the multicast addr\n");
  3497. goto error;
  3498. }
  3499. @@ -523,7 +620,7 @@
  3500.  
  3501. error:
  3502. sock_release(sock);
  3503. - return NULL;
  3504. + return ERR_PTR(result);
  3505. }
  3506.  
  3507.  
  3508. @@ -533,27 +630,30 @@
  3509. static struct socket * make_receive_sock(void)
  3510. {
  3511. struct socket *sock;
  3512. + int result;
  3513.  
  3514. /* First create a socket */
  3515. - if (sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock) < 0) {
  3516. + result = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
  3517. + if (result < 0) {
  3518. IP_VS_ERR("Error during creation of socket; terminating\n");
  3519. - return NULL;
  3520. + return ERR_PTR(result);
  3521. }
  3522.  
  3523. /* it is equivalent to the REUSEADDR option in user-space */
  3524. sock->sk->sk_reuse = 1;
  3525.  
  3526. - if (sock->ops->bind(sock,
  3527. - (struct sockaddr*)&mcast_addr,
  3528. - sizeof(struct sockaddr)) < 0) {
  3529. + result = sock->ops->bind(sock, (struct sockaddr *) &mcast_addr,
  3530. + sizeof(struct sockaddr));
  3531. + if (result < 0) {
  3532. IP_VS_ERR("Error binding to the multicast addr\n");
  3533. goto error;
  3534. }
  3535.  
  3536. /* join the multicast group */
  3537. - if (join_mcast_group(sock->sk,
  3538. - (struct in_addr*)&mcast_addr.sin_addr,
  3539. - ip_vs_backup_mcast_ifn) < 0) {
  3540. + result = join_mcast_group(sock->sk,
  3541. + (struct in_addr *) &mcast_addr.sin_addr,
  3542. + ip_vs_backup_mcast_ifn);
  3543. + if (result < 0) {
  3544. IP_VS_ERR("Error joining to the multicast group\n");
  3545. goto error;
  3546. }
  3547. @@ -562,7 +662,7 @@
  3548.  
  3549. error:
  3550. sock_release(sock);
  3551. - return NULL;
  3552. + return ERR_PTR(result);
  3553. }
  3554.  
  3555.  
  3556. @@ -620,44 +720,29 @@
  3557. }
  3558.  
  3559.  
  3560. -static DECLARE_WAIT_QUEUE_HEAD(sync_wait);
  3561. -static pid_t sync_master_pid = 0;
  3562. -static pid_t sync_backup_pid = 0;
  3563. -
  3564. -static DECLARE_WAIT_QUEUE_HEAD(stop_sync_wait);
  3565. -static int stop_master_sync = 0;
  3566. -static int stop_backup_sync = 0;
  3567. -
  3568. -static void sync_master_loop(void)
  3569. +static int sync_thread_master(void *data)
  3570. {
  3571. - struct socket *sock;
  3572. + struct ip_vs_sync_thread_data *tinfo = data;
  3573. struct ip_vs_sync_buff *sb;
  3574.  
  3575. - /* create the sending multicast socket */
  3576. - sock = make_send_sock();
  3577. - if (!sock)
  3578. - return;
  3579. -
  3580. IP_VS_INFO("sync thread started: state = MASTER, mcast_ifn = %s, "
  3581. "syncid = %d\n",
  3582. ip_vs_master_mcast_ifn, ip_vs_master_syncid);
  3583.  
  3584. - for (;;) {
  3585. - while ((sb=sb_dequeue())) {
  3586. - ip_vs_send_sync_msg(sock, sb->mesg);
  3587. + while (!kthread_should_stop()) {
  3588. + while ((sb = sb_dequeue())) {
  3589. + ip_vs_send_sync_msg(tinfo->sock, sb->mesg);
  3590. ip_vs_sync_buff_release(sb);
  3591. }
  3592.  
  3593. /* check if entries stay in curr_sb for 2 seconds */
  3594. - if ((sb = get_curr_sync_buff(2*HZ))) {
  3595. - ip_vs_send_sync_msg(sock, sb->mesg);
  3596. + sb = get_curr_sync_buff(2 * HZ);
  3597. + if (sb) {
  3598. + ip_vs_send_sync_msg(tinfo->sock, sb->mesg);
  3599. ip_vs_sync_buff_release(sb);
  3600. }
  3601.  
  3602. - if (stop_master_sync)
  3603. - break;
  3604. -
  3605. - msleep_interruptible(1000);
  3606. + schedule_timeout_interruptible(HZ);
  3607. }
  3608.  
  3609. /* clean up the sync_buff queue */
  3610. @@ -671,235 +756,175 @@
  3611. }
  3612.  
  3613. /* release the sending multicast socket */
  3614. - sock_release(sock);
  3615. + sock_release(tinfo->sock);
  3616. + kfree(tinfo);
  3617. +
  3618. + return 0;
  3619. }
  3620.  
  3621.  
  3622. -static void sync_backup_loop(void)
  3623. +static int sync_thread_backup(void *data)
  3624. {
  3625. - struct socket *sock;
  3626. - char *buf;
  3627. + struct ip_vs_sync_thread_data *tinfo = data;
  3628. int len;
  3629.  
  3630. - if (!(buf = kmalloc(sync_recv_mesg_maxlen, GFP_ATOMIC))) {
  3631. - IP_VS_ERR("sync_backup_loop: kmalloc error\n");
  3632. - return;
  3633. - }
  3634. -
  3635. - /* create the receiving multicast socket */
  3636. - sock = make_receive_sock();
  3637. - if (!sock)
  3638. - goto out;
  3639. -
  3640. IP_VS_INFO("sync thread started: state = BACKUP, mcast_ifn = %s, "
  3641. "syncid = %d\n",
  3642. ip_vs_backup_mcast_ifn, ip_vs_backup_syncid);
  3643.  
  3644. - for (;;) {
  3645. - /* do you have data now? */
  3646. - while (!skb_queue_empty(&(sock->sk->sk_receive_queue))) {
  3647. - if ((len =
  3648. - ip_vs_receive(sock, buf,
  3649. - sync_recv_mesg_maxlen)) <= 0) {
  3650. + while (!kthread_should_stop()) {
  3651. + wait_event_interruptible(*tinfo->sock->sk->sk_sleep,
  3652. + !skb_queue_empty(&tinfo->sock->sk->sk_receive_queue)
  3653. + || kthread_should_stop());
  3654. +
  3655. + /* do we have data now? */
  3656. + while (!skb_queue_empty(&(tinfo->sock->sk->sk_receive_queue))) {
  3657. + len = ip_vs_receive(tinfo->sock, tinfo->buf,
  3658. + sync_recv_mesg_maxlen);
  3659. + if (len <= 0) {
  3660. IP_VS_ERR("receiving message error\n");
  3661. break;
  3662. }
  3663. - /* disable bottom half, because it accessed the data
  3664. +
  3665. + /* disable bottom half, because it accesses the data
  3666. shared by softirq while getting/creating conns */
  3667. local_bh_disable();
  3668. - ip_vs_process_message(buf, len);
  3669. + ip_vs_process_message(tinfo->buf, len);
  3670. local_bh_enable();
  3671. }
  3672. -
  3673. - if (stop_backup_sync)
  3674. - break;
  3675. -
  3676. - msleep_interruptible(1000);
  3677. }
  3678.  
  3679. /* release the sending multicast socket */
  3680. - sock_release(sock);
  3681. + sock_release(tinfo->sock);
  3682. + kfree(tinfo->buf);
  3683. + kfree(tinfo);
  3684.  
  3685. - out:
  3686. - kfree(buf);
  3687. + return 0;
  3688. }
  3689.  
  3690.  
  3691. -static void set_sync_pid(int sync_state, pid_t sync_pid)
  3692. -{
  3693. - if (sync_state == IP_VS_STATE_MASTER)
  3694. - sync_master_pid = sync_pid;
  3695. - else if (sync_state == IP_VS_STATE_BACKUP)
  3696. - sync_backup_pid = sync_pid;
  3697. -}
  3698. -
  3699. -static void set_stop_sync(int sync_state, int set)
  3700. +int start_sync_thread(int state, char *mcast_ifn, __u8 syncid)
  3701. {
  3702. - if (sync_state == IP_VS_STATE_MASTER)
  3703. - stop_master_sync = set;
  3704. - else if (sync_state == IP_VS_STATE_BACKUP)
  3705. - stop_backup_sync = set;
  3706. - else {
  3707. - stop_master_sync = set;
  3708. - stop_backup_sync = set;
  3709. - }
  3710. -}
  3711. + struct ip_vs_sync_thread_data *tinfo;
  3712. + struct task_struct **realtask, *task;
  3713. + struct socket *sock;
  3714. + char *name, *buf = NULL;
  3715. + int (*threadfn)(void *data);
  3716. + int result = -ENOMEM;
  3717.  
  3718. -static int sync_thread(void *startup)
  3719. -{
  3720. - DECLARE_WAITQUEUE(wait, current);
  3721. - mm_segment_t oldmm;
  3722. - int state;
  3723. - const char *name;
  3724. + IP_VS_DBG(7, "%s: pid %d\n", __func__, task_pid_nr(current));
  3725. + IP_VS_DBG(7, "Each ip_vs_sync_conn entry needs %Zd bytes\n",
  3726. + sizeof(struct ip_vs_sync_conn));
  3727.  
  3728. - /* increase the module use count */
  3729. - ip_vs_use_count_inc();
  3730. + if (state == IP_VS_STATE_MASTER) {
  3731. + if (sync_master_thread)
  3732. + return -EEXIST;
  3733.  
  3734. - if (ip_vs_sync_state & IP_VS_STATE_MASTER && !sync_master_pid) {
  3735. - state = IP_VS_STATE_MASTER;
  3736. + strlcpy(ip_vs_master_mcast_ifn, mcast_ifn,
  3737. + sizeof(ip_vs_master_mcast_ifn));
  3738. + ip_vs_master_syncid = syncid;
  3739. + realtask = &sync_master_thread;
  3740. name = "ipvs_syncmaster";
  3741. - } else if (ip_vs_sync_state & IP_VS_STATE_BACKUP && !sync_backup_pid) {
  3742. - state = IP_VS_STATE_BACKUP;
  3743. + threadfn = sync_thread_master;
  3744. + sock = make_send_sock();
  3745. + } else if (state == IP_VS_STATE_BACKUP) {
  3746. + if (sync_backup_thread)
  3747. + return -EEXIST;
  3748. +
  3749. + strlcpy(ip_vs_backup_mcast_ifn, mcast_ifn,
  3750. + sizeof(ip_vs_backup_mcast_ifn));
  3751. + ip_vs_backup_syncid = syncid;
  3752. + realtask = &sync_backup_thread;
  3753. name = "ipvs_syncbackup";
  3754. + threadfn = sync_thread_backup;
  3755. + sock = make_receive_sock();
  3756. } else {
  3757. - IP_VS_BUG();
  3758. - ip_vs_use_count_dec();
  3759. return -EINVAL;
  3760. }
  3761.  
  3762. - daemonize(name);
  3763. -
  3764. - oldmm = get_fs();
  3765. - set_fs(KERNEL_DS);
  3766. -
  3767. - /* Block all signals */
  3768. - spin_lock_irq(&current->sighand->siglock);
  3769. - siginitsetinv(&current->blocked, 0);
  3770. - recalc_sigpending();
  3771. - spin_unlock_irq(&current->sighand->siglock);
  3772. + if (IS_ERR(sock)) {
  3773. + result = PTR_ERR(sock);
  3774. + goto out;
  3775. + }
  3776.  
  3777. - /* set the maximum length of sync message */
  3778. set_sync_mesg_maxlen(state);
  3779. + if (state == IP_VS_STATE_BACKUP) {
  3780. + buf = kmalloc(sync_recv_mesg_maxlen, GFP_KERNEL);
  3781. + if (!buf)
  3782. + goto outsocket;
  3783. + }
  3784.  
  3785. - /* set up multicast address */
  3786. - mcast_addr.sin_family = AF_INET;
  3787. - mcast_addr.sin_port = htons(IP_VS_SYNC_PORT);
  3788. - mcast_addr.sin_addr.s_addr = htonl(IP_VS_SYNC_GROUP);
  3789. -
  3790. - add_wait_queue(&sync_wait, &wait);
  3791. -
  3792. - set_sync_pid(state, current->pid);
  3793. - complete((struct completion *)startup);
  3794. -
  3795. - /* processing master/backup loop here */
  3796. - if (state == IP_VS_STATE_MASTER)
  3797. - sync_master_loop();
  3798. - else if (state == IP_VS_STATE_BACKUP)
  3799. - sync_backup_loop();
  3800. - else IP_VS_BUG();
  3801. -
  3802. - remove_wait_queue(&sync_wait, &wait);
  3803. -
  3804. - /* thread exits */
  3805. - set_sync_pid(state, 0);
  3806. - IP_VS_INFO("sync thread stopped!\n");
  3807. -
  3808. - set_fs(oldmm);
  3809. -
  3810. - /* decrease the module use count */
  3811. - ip_vs_use_count_dec();
  3812. -
  3813. - set_stop_sync(state, 0);
  3814. - wake_up(&stop_sync_wait);
  3815. + tinfo = kmalloc(sizeof(*tinfo), GFP_KERNEL);
  3816. + if (!tinfo)
  3817. + goto outbuf;
  3818.  
  3819. - return 0;
  3820. -}
  3821. + tinfo->sock = sock;
  3822. + tinfo->buf = buf;
  3823.  
  3824. + task = kthread_run(threadfn, tinfo, name);
  3825. + if (IS_ERR(task)) {
  3826. + result = PTR_ERR(task);
  3827. + goto outtinfo;
  3828. + }
  3829.  
  3830. -static int fork_sync_thread(void *startup)
  3831. -{
  3832. - pid_t pid;
  3833. + /* mark as active */
  3834. + *realtask = task;
  3835. + ip_vs_sync_state |= state;
  3836.  
  3837. - /* fork the sync thread here, then the parent process of the
  3838. - sync thread is the init process after this thread exits. */
  3839. - repeat:
  3840. - if ((pid = kernel_thread(sync_thread, startup, 0)) < 0) {
  3841. - IP_VS_ERR("could not create sync_thread due to %d... "
  3842. - "retrying.\n", pid);
  3843. - msleep_interruptible(1000);
  3844. - goto repeat;
  3845. - }
  3846. + /* increase the module use count */
  3847. + ip_vs_use_count_inc();
  3848.  
  3849. return 0;
  3850. +
  3851. +outtinfo:
  3852. + kfree(tinfo);
  3853. +outbuf:
  3854. + kfree(buf);
  3855. +outsocket:
  3856. + sock_release(sock);
  3857. +out:
  3858. + return result;
  3859. }
  3860.  
  3861.  
  3862. -int start_sync_thread(int state, char *mcast_ifn, __u8 syncid)
  3863. +int stop_sync_thread(int state)
  3864. {
  3865. - DECLARE_COMPLETION_ONSTACK(startup);
  3866. - pid_t pid;
  3867. + IP_VS_DBG(7, "%s: pid %d\n", __func__, task_pid_nr(current));
  3868.  
  3869. - if ((state == IP_VS_STATE_MASTER && sync_master_pid) ||
  3870. - (state == IP_VS_STATE_BACKUP && sync_backup_pid))
  3871. - return -EEXIST;
  3872. + if (state == IP_VS_STATE_MASTER) {
  3873. + if (!sync_master_thread)
  3874. + return -ESRCH;
  3875.  
  3876. - IP_VS_DBG(7, "%s: pid %d\n", __FUNCTION__, current->pid);
  3877. - IP_VS_DBG(7, "Each ip_vs_sync_conn entry need %Zd bytes\n",
  3878. - sizeof(struct ip_vs_sync_conn));
  3879. + IP_VS_INFO("stopping master sync thread %d ...\n",
  3880. + task_pid_nr(sync_master_thread));
  3881.  
  3882. - ip_vs_sync_state |= state;
  3883. - if (state == IP_VS_STATE_MASTER) {
  3884. - strlcpy(ip_vs_master_mcast_ifn, mcast_ifn,
  3885. - sizeof(ip_vs_master_mcast_ifn));
  3886. - ip_vs_master_syncid = syncid;
  3887. + /*
  3888. + * The lock synchronizes with sb_queue_tail(), so that we don't
  3889. + * add sync buffers to the queue, when we are already in
  3890. + * progress of stopping the master sync daemon.
  3891. + */
  3892. +
  3893. + spin_lock_bh(&ip_vs_sync_lock);
  3894. + ip_vs_sync_state &= ~IP_VS_STATE_MASTER;
  3895. + spin_unlock_bh(&ip_vs_sync_lock);
  3896. + kthread_stop(sync_master_thread);
  3897. + sync_master_thread = NULL;
  3898. + } else if (state == IP_VS_STATE_BACKUP) {
  3899. + if (!sync_backup_thread)
  3900. + return -ESRCH;
  3901. +
  3902. + IP_VS_INFO("stopping backup sync thread %d ...\n",
  3903. + task_pid_nr(sync_backup_thread));
  3904. +
  3905. + ip_vs_sync_state &= ~IP_VS_STATE_BACKUP;
  3906. + kthread_stop(sync_backup_thread);
  3907. + sync_backup_thread = NULL;
  3908. } else {
  3909. - strlcpy(ip_vs_backup_mcast_ifn, mcast_ifn,
  3910. - sizeof(ip_vs_backup_mcast_ifn));
  3911. - ip_vs_backup_syncid = syncid;
  3912. - }
  3913. -
  3914. - repeat:
  3915. - if ((pid = kernel_thread(fork_sync_thread, &startup, 0)) < 0) {
  3916. - IP_VS_ERR("could not create fork_sync_thread due to %d... "
  3917. - "retrying.\n", pid);
  3918. - msleep_interruptible(1000);
  3919. - goto repeat;
  3920. + return -EINVAL;
  3921. }
  3922.  
  3923. - wait_for_completion(&startup);
  3924. -
  3925. - return 0;
  3926. -}
  3927. -
  3928. -
  3929. -int stop_sync_thread(int state)
  3930. -{
  3931. - DECLARE_WAITQUEUE(wait, current);
  3932. -
  3933. - if ((state == IP_VS_STATE_MASTER && !sync_master_pid) ||
  3934. - (state == IP_VS_STATE_BACKUP && !sync_backup_pid))
  3935. - return -ESRCH;
  3936. -
  3937. - IP_VS_DBG(7, "%s: pid %d\n", __FUNCTION__, current->pid);
  3938. - IP_VS_INFO("stopping sync thread %d ...\n",
  3939. - (state == IP_VS_STATE_MASTER) ?
  3940. - sync_master_pid : sync_backup_pid);
  3941. -
  3942. - __set_current_state(TASK_UNINTERRUPTIBLE);
  3943. - add_wait_queue(&stop_sync_wait, &wait);
  3944. - set_stop_sync(state, 1);
  3945. - ip_vs_sync_state -= state;
  3946. - wake_up(&sync_wait);
  3947. - schedule();
  3948. - __set_current_state(TASK_RUNNING);
  3949. - remove_wait_queue(&stop_sync_wait, &wait);
  3950. -
  3951. - /* Note: no need to reap the sync thread, because its parent
  3952. - process is the init process */
  3953. -
  3954. - if ((state == IP_VS_STATE_MASTER && stop_master_sync) ||
  3955. - (state == IP_VS_STATE_BACKUP && stop_backup_sync))
  3956. - IP_VS_BUG();
  3957. + /* decrease the module use count */
  3958. + ip_vs_use_count_dec();
  3959.  
  3960. return 0;
  3961. }
  3962. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_wlc.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_wlc.c
  3963. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_wlc.c 2009-02-16 11:57:22.000000000 -0400
  3964. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_wlc.c 2009-02-16 12:56:22.000000000 -0400
  3965. @@ -1,8 +1,6 @@
  3966. /*
  3967. * IPVS: Weighted Least-Connection Scheduling module
  3968. *
  3969. - * Version: $Id: ip_vs_wlc.c,v 1.13 2003/04/18 09:03:16 wensong Exp $
  3970. - *
  3971. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  3972. * Peter Kese <peter.kese@ijs.si>
  3973. *
  3974. @@ -128,6 +126,7 @@
  3975. .name = "wlc",
  3976. .refcnt = ATOMIC_INIT(0),
  3977. .module = THIS_MODULE,
  3978. + .n_list = LIST_HEAD_INIT(ip_vs_wlc_scheduler.n_list),
  3979. .init_service = ip_vs_wlc_init_svc,
  3980. .done_service = ip_vs_wlc_done_svc,
  3981. .update_service = ip_vs_wlc_update_svc,
  3982. @@ -137,7 +136,6 @@
  3983.  
  3984. static int __init ip_vs_wlc_init(void)
  3985. {
  3986. - INIT_LIST_HEAD(&ip_vs_wlc_scheduler.n_list);
  3987. return register_ip_vs_scheduler(&ip_vs_wlc_scheduler);
  3988. }
  3989.  
  3990. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_wrr.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_wrr.c
  3991. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_wrr.c 2009-02-16 11:57:22.000000000 -0400
  3992. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_wrr.c 2009-02-16 12:56:22.000000000 -0400
  3993. @@ -1,8 +1,6 @@
  3994. /*
  3995. * IPVS: Weighted Round-Robin Scheduling module
  3996. *
  3997. - * Version: $Id: ip_vs_wrr.c,v 1.12 2002/09/15 08:14:08 wensong Exp $
  3998. - *
  3999. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  4000. *
  4001. * This program is free software; you can redistribute it and/or
  4002. @@ -22,6 +20,7 @@
  4003.  
  4004. #include <linux/module.h>
  4005. #include <linux/kernel.h>
  4006. +#include <linux/net.h>
  4007.  
  4008. #include <net/ip_vs.h>
  4009.  
  4010. @@ -169,7 +168,7 @@
  4011. */
  4012. if (mark->cw == 0) {
  4013. mark->cl = &svc->destinations;
  4014. - IP_VS_INFO("ip_vs_wrr_schedule(): "
  4015. + IP_VS_ERR_RL("ip_vs_wrr_schedule(): "
  4016. "no available servers\n");
  4017. dest = NULL;
  4018. goto out;
  4019. @@ -213,6 +212,7 @@
  4020. .name = "wrr",
  4021. .refcnt = ATOMIC_INIT(0),
  4022. .module = THIS_MODULE,
  4023. + .n_list = LIST_HEAD_INIT(ip_vs_wrr_scheduler.n_list),
  4024. .init_service = ip_vs_wrr_init_svc,
  4025. .done_service = ip_vs_wrr_done_svc,
  4026. .update_service = ip_vs_wrr_update_svc,
  4027. @@ -221,7 +221,6 @@
  4028.  
  4029. static int __init ip_vs_wrr_init(void)
  4030. {
  4031. - INIT_LIST_HEAD(&ip_vs_wrr_scheduler.n_list);
  4032. return register_ip_vs_scheduler(&ip_vs_wrr_scheduler) ;
  4033. }
  4034.  
  4035. diff -u linux-2.6.20.y/net/ipv4/ipvs/ip_vs_xmit.c linux-2.6.27.y/net/ipv4/ipvs/ip_vs_xmit.c
  4036. --- linux-2.6.20.y/net/ipv4/ipvs/ip_vs_xmit.c 2009-02-16 11:57:22.000000000 -0400
  4037. +++ linux-2.6.27.y/net/ipv4/ipvs/ip_vs_xmit.c 2009-02-16 12:56:22.000000000 -0400
  4038. @@ -1,8 +1,6 @@
  4039. /*
  4040. * ip_vs_xmit.c: various packet transmitters for IPVS
  4041. *
  4042. - * Version: $Id: ip_vs_xmit.c,v 1.2 2002/11/30 01:50:35 wensong Exp $
  4043. - *
  4044. * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
  4045. * Julian Anastasov <ja@ssi.bg>
  4046. *
  4047. @@ -16,8 +14,8 @@
  4048. */
  4049.  
  4050. #include <linux/kernel.h>
  4051. -#include <linux/ip.h>
  4052. #include <linux/tcp.h> /* for tcphdr */
  4053. +#include <net/ip.h>
  4054. #include <net/tcp.h> /* for csum_tcpudp_magic */
  4055. #include <net/udp.h>
  4056. #include <net/icmp.h> /* for icmp_send */
  4057. @@ -59,7 +57,7 @@
  4058. return dst;
  4059. }
  4060.  
  4061. -static inline struct rtable *
  4062. +static struct rtable *
  4063. __ip_vs_get_out_rt(struct ip_vs_conn *cp, u32 rtos)
  4064. {
  4065. struct rtable *rt; /* Route to the other host */
  4066. @@ -78,7 +76,7 @@
  4067. .tos = rtos, } },
  4068. };
  4069.  
  4070. - if (ip_route_output_key(&rt, &fl)) {
  4071. + if (ip_route_output_key(&init_net, &rt, &fl)) {
  4072. spin_unlock(&dest->dst_lock);
  4073. IP_VS_DBG_RL("ip_route_output error, "
  4074. "dest: %u.%u.%u.%u\n",
  4075. @@ -101,7 +99,7 @@
  4076. .tos = rtos, } },
  4077. };
  4078.  
  4079. - if (ip_route_output_key(&rt, &fl)) {
  4080. + if (ip_route_output_key(&init_net, &rt, &fl)) {
  4081. IP_VS_DBG_RL("ip_route_output error, dest: "
  4082. "%u.%u.%u.%u\n", NIPQUAD(cp->daddr));
  4083. return NULL;
  4084. @@ -128,8 +126,8 @@
  4085. #define IP_VS_XMIT(skb, rt) \
  4086. do { \
  4087. (skb)->ipvs_property = 1; \
  4088. - (skb)->ip_summed = CHECKSUM_NONE; \
  4089. - NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, (skb), NULL, \
  4090. + skb_forward_csum(skb); \
  4091. + NF_HOOK(PF_INET, NF_INET_LOCAL_OUT, (skb), NULL, \
  4092. (rt)->u.dst.dev, dst_output); \
  4093. } while (0)
  4094.  
  4095. @@ -156,7 +154,7 @@
  4096. struct ip_vs_protocol *pp)
  4097. {
  4098. struct rtable *rt; /* Route to the other host */
  4099. - struct iphdr *iph = skb->nh.iph;
  4100. + struct iphdr *iph = ip_hdr(skb);
  4101. u8 tos = iph->tos;
  4102. int mtu;
  4103. struct flowi fl = {
  4104. @@ -170,7 +168,7 @@
  4105.  
  4106. EnterFunction(10);
  4107.  
  4108. - if (ip_route_output_key(&rt, &fl)) {
  4109. + if (ip_route_output_key(&init_net, &rt, &fl)) {
  4110. IP_VS_DBG_RL("ip_vs_bypass_xmit(): ip_route_output error, "
  4111. "dest: %u.%u.%u.%u\n", NIPQUAD(iph->daddr));
  4112. goto tx_error_icmp;
  4113. @@ -178,7 +176,7 @@
  4114.  
  4115. /* MTU checking */
  4116. mtu = dst_mtu(&rt->u.dst);
  4117. - if ((skb->len > mtu) && (iph->frag_off&__constant_htons(IP_DF))) {
  4118. + if ((skb->len > mtu) && (iph->frag_off & htons(IP_DF))) {
  4119. ip_rt_put(rt);
  4120. icmp_send(skb, ICMP_DEST_UNREACH,ICMP_FRAG_NEEDED, htonl(mtu));
  4121. IP_VS_DBG_RL("ip_vs_bypass_xmit(): frag needed\n");
  4122. @@ -193,7 +191,7 @@
  4123. ip_rt_put(rt);
  4124. return NF_STOLEN;
  4125. }
  4126. - ip_send_check(skb->nh.iph);
  4127. + ip_send_check(ip_hdr(skb));
  4128.  
  4129. /* drop old route */
  4130. dst_release(skb->dst);
  4131. @@ -226,7 +224,7 @@
  4132. {
  4133. struct rtable *rt; /* Route to the other host */
  4134. int mtu;
  4135. - struct iphdr *iph = skb->nh.iph;
  4136. + struct iphdr *iph = ip_hdr(skb);
  4137.  
  4138. EnterFunction(10);
  4139.  
  4140. @@ -245,7 +243,7 @@
  4141.  
  4142. /* MTU checking */
  4143. mtu = dst_mtu(&rt->u.dst);
  4144. - if ((skb->len > mtu) && (iph->frag_off&__constant_htons(IP_DF))) {
  4145. + if ((skb->len > mtu) && (iph->frag_off & htons(IP_DF))) {
  4146. ip_rt_put(rt);
  4147. icmp_send(skb, ICMP_DEST_UNREACH,ICMP_FRAG_NEEDED, htonl(mtu));
  4148. IP_VS_DBG_RL_PKT(0, pp, skb, 0, "ip_vs_nat_xmit(): frag needed for");
  4149. @@ -253,7 +251,7 @@
  4150. }
  4151.  
  4152. /* copy-on-write the packet before mangling it */
  4153. - if (!ip_vs_make_skb_writable(&skb, sizeof(struct iphdr)))
  4154. + if (!skb_make_writable(skb, sizeof(struct iphdr)))
  4155. goto tx_error_put;
  4156.  
  4157. if (skb_cow(skb, rt->u.dst.dev->hard_header_len))
  4158. @@ -264,10 +262,10 @@
  4159. skb->dst = &rt->u.dst;
  4160.  
  4161. /* mangle the packet */
  4162. - if (pp->dnat_handler && !pp->dnat_handler(&skb, pp, cp))
  4163. + if (pp->dnat_handler && !pp->dnat_handler(skb, pp, cp))
  4164. goto tx_error;
  4165. - skb->nh.iph->daddr = cp->daddr;
  4166. - ip_send_check(skb->nh.iph);
  4167. + ip_hdr(skb)->daddr = cp->daddr;
  4168. + ip_send_check(ip_hdr(skb));
  4169.  
  4170. IP_VS_DBG_PKT(10, pp, skb, 0, "After DNAT");
  4171.  
  4172. @@ -320,19 +318,20 @@
  4173. {
  4174. struct rtable *rt; /* Route to the other host */
  4175. struct net_device *tdev; /* Device to other host */
  4176. - struct iphdr *old_iph = skb->nh.iph;
  4177. + struct iphdr *old_iph = ip_hdr(skb);
  4178. u8 tos = old_iph->tos;
  4179. __be16 df = old_iph->frag_off;
  4180. + sk_buff_data_t old_transport_header = skb->transport_header;
  4181. struct iphdr *iph; /* Our new IP header */
  4182. - int max_headroom; /* The extra header space needed */
  4183. + unsigned int max_headroom; /* The extra header space needed */
  4184. int mtu;
  4185.  
  4186. EnterFunction(10);
  4187.  
  4188. - if (skb->protocol != __constant_htons(ETH_P_IP)) {
  4189. + if (skb->protocol != htons(ETH_P_IP)) {
  4190. IP_VS_DBG_RL("ip_vs_tunnel_xmit(): protocol error, "
  4191. "ETH_P_IP: %d, skb protocol: %d\n",
  4192. - __constant_htons(ETH_P_IP), skb->protocol);
  4193. + htons(ETH_P_IP), skb->protocol);
  4194. goto tx_error;
  4195. }
  4196.  
  4197. @@ -350,9 +349,9 @@
  4198. if (skb->dst)
  4199. skb->dst->ops->update_pmtu(skb->dst, mtu);
  4200.  
  4201. - df |= (old_iph->frag_off&__constant_htons(IP_DF));
  4202. + df |= (old_iph->frag_off & htons(IP_DF));
  4203.  
  4204. - if ((old_iph->frag_off&__constant_htons(IP_DF))
  4205. + if ((old_iph->frag_off & htons(IP_DF))
  4206. && mtu < ntohs(old_iph->tot_len)) {
  4207. icmp_send(skb, ICMP_DEST_UNREACH,ICMP_FRAG_NEEDED, htonl(mtu));
  4208. ip_rt_put(rt);
  4209. @@ -377,15 +376,16 @@
  4210. }
  4211. kfree_skb(skb);
  4212. skb = new_skb;
  4213. - old_iph = skb->nh.iph;
  4214. + old_iph = ip_hdr(skb);
  4215. }
  4216.  
  4217. - skb->h.raw = (void *) old_iph;
  4218. + skb->transport_header = old_transport_header;
  4219.  
  4220. /* fix old IP header checksum */
  4221. ip_send_check(old_iph);
  4222.  
  4223. - skb->nh.raw = skb_push(skb, sizeof(struct iphdr));
  4224. + skb_push(skb, sizeof(struct iphdr));
  4225. + skb_reset_network_header(skb);
  4226. memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
  4227.  
  4228. /* drop old route */
  4229. @@ -395,7 +395,7 @@
  4230. /*
  4231. * Push down and install the IPIP header.
  4232. */
  4233. - iph = skb->nh.iph;
  4234. + iph = ip_hdr(skb);
  4235. iph->version = 4;
  4236. iph->ihl = sizeof(struct iphdr)>>2;
  4237. iph->frag_off = df;
  4238. @@ -404,14 +404,12 @@
  4239. iph->daddr = rt->rt_dst;
  4240. iph->saddr = rt->rt_src;
  4241. iph->ttl = old_iph->ttl;
  4242. - iph->tot_len = htons(skb->len);
  4243. ip_select_ident(iph, &rt->u.dst, NULL);
  4244. - ip_send_check(iph);
  4245.  
  4246. /* Another hack: avoid icmp_send in ip_fragment */
  4247. skb->local_df = 1;
  4248.  
  4249. - IP_VS_XMIT(skb, rt);
  4250. + ip_local_out(skb);
  4251.  
  4252. LeaveFunction(10);
  4253.  
  4254. @@ -435,7 +433,7 @@
  4255. struct ip_vs_protocol *pp)
  4256. {
  4257. struct rtable *rt; /* Route to the other host */
  4258. - struct iphdr *iph = skb->nh.iph;
  4259. + struct iphdr *iph = ip_hdr(skb);
  4260. int mtu;
  4261.  
  4262. EnterFunction(10);
  4263. @@ -445,7 +443,7 @@
  4264.  
  4265. /* MTU checking */
  4266. mtu = dst_mtu(&rt->u.dst);
  4267. - if ((iph->frag_off&__constant_htons(IP_DF)) && skb->len > mtu) {
  4268. + if ((iph->frag_off & htons(IP_DF)) && skb->len > mtu) {
  4269. icmp_send(skb, ICMP_DEST_UNREACH,ICMP_FRAG_NEEDED, htonl(mtu));
  4270. ip_rt_put(rt);
  4271. IP_VS_DBG_RL("ip_vs_dr_xmit(): frag needed\n");
  4272. @@ -460,7 +458,7 @@
  4273. ip_rt_put(rt);
  4274. return NF_STOLEN;
  4275. }
  4276. - ip_send_check(skb->nh.iph);
  4277. + ip_send_check(ip_hdr(skb));
  4278.  
  4279. /* drop old route */
  4280. dst_release(skb->dst);
  4281. @@ -514,12 +512,12 @@
  4282. * mangle and send the packet here (only for VS/NAT)
  4283. */
  4284.  
  4285. - if (!(rt = __ip_vs_get_out_rt(cp, RT_TOS(skb->nh.iph->tos))))
  4286. + if (!(rt = __ip_vs_get_out_rt(cp, RT_TOS(ip_hdr(skb)->tos))))
  4287. goto tx_error_icmp;
  4288.  
  4289. /* MTU checking */
  4290. mtu = dst_mtu(&rt->u.dst);
  4291. - if ((skb->len > mtu) && (skb->nh.iph->frag_off&__constant_htons(IP_DF))) {
  4292. + if ((skb->len > mtu) && (ip_hdr(skb)->frag_off & htons(IP_DF))) {
  4293. ip_rt_put(rt);
  4294. icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
  4295. IP_VS_DBG_RL("ip_vs_in_icmp(): frag needed\n");
  4296. @@ -527,7 +525,7 @@
  4297. }
  4298.  
  4299. /* copy-on-write the packet before mangling it */
  4300. - if (!ip_vs_make_skb_writable(&skb, offset))
  4301. + if (!skb_make_writable(skb, offset))
  4302. goto tx_error_put;
  4303.  
  4304. if (skb_cow(skb, rt->u.dst.dev->hard_header_len))
Add Comment
Please, Sign In to add comment
Public Pastes
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!