]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - net/ipv4/netfilter/ipt_CLUSTERIP.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
netfilter: ipt_CLUSTERIP: put config instead of freeing it
[mirror_ubuntu-bionic-kernel.git] / net / ipv4 / netfilter / ipt_CLUSTERIP.c
1 /* Cluster IP hashmark target
2 * (C) 2003-2004 by Harald Welte <laforge@netfilter.org>
3 * based on ideas of Fabio Olive Leite <olive@unixforge.org>
4 *
5 * Development of this code funded by SuSE Linux AG, http://www.suse.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 *
11 */
12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13 #include <linux/module.h>
14 #include <linux/proc_fs.h>
15 #include <linux/jhash.h>
16 #include <linux/bitops.h>
17 #include <linux/skbuff.h>
18 #include <linux/slab.h>
19 #include <linux/ip.h>
20 #include <linux/tcp.h>
21 #include <linux/udp.h>
22 #include <linux/icmp.h>
23 #include <linux/if_arp.h>
24 #include <linux/seq_file.h>
25 #include <linux/refcount.h>
26 #include <linux/netfilter_arp.h>
27 #include <linux/netfilter/x_tables.h>
28 #include <linux/netfilter_ipv4/ip_tables.h>
29 #include <linux/netfilter_ipv4/ipt_CLUSTERIP.h>
30 #include <net/netfilter/nf_conntrack.h>
31 #include <net/net_namespace.h>
32 #include <net/netns/generic.h>
33 #include <net/checksum.h>
34 #include <net/ip.h>
35
36 #define CLUSTERIP_VERSION "0.8"
37
38 MODULE_LICENSE("GPL");
39 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
40 MODULE_DESCRIPTION("Xtables: CLUSTERIP target");
41
42 struct clusterip_config {
43 struct list_head list; /* list of all configs */
44 refcount_t refcount; /* reference count */
45 refcount_t entries; /* number of entries/rules
46 * referencing us */
47
48 __be32 clusterip; /* the IP address */
49 u_int8_t clustermac[ETH_ALEN]; /* the MAC address */
50 int ifindex; /* device ifindex */
51 u_int16_t num_total_nodes; /* total number of nodes */
52 unsigned long local_nodes; /* node number array */
53
54 #ifdef CONFIG_PROC_FS
55 struct proc_dir_entry *pde; /* proc dir entry */
56 #endif
57 enum clusterip_hashmode hash_mode; /* which hashing mode */
58 u_int32_t hash_initval; /* hash initialization */
59 struct rcu_head rcu;
60
61 char ifname[IFNAMSIZ]; /* device ifname */
62 struct notifier_block notifier; /* refresh c->ifindex in it */
63 };
64
65 #ifdef CONFIG_PROC_FS
66 static const struct file_operations clusterip_proc_fops;
67 #endif
68
69 static unsigned int clusterip_net_id __read_mostly;
70
71 struct clusterip_net {
72 struct list_head configs;
73 /* lock protects the configs list */
74 spinlock_t lock;
75
76 #ifdef CONFIG_PROC_FS
77 struct proc_dir_entry *procdir;
78 #endif
79 };
80
81 static inline void
82 clusterip_config_get(struct clusterip_config *c)
83 {
84 refcount_inc(&c->refcount);
85 }
86
87
88 static void clusterip_config_rcu_free(struct rcu_head *head)
89 {
90 kfree(container_of(head, struct clusterip_config, rcu));
91 }
92
93 static inline void
94 clusterip_config_put(struct clusterip_config *c)
95 {
96 if (refcount_dec_and_test(&c->refcount))
97 call_rcu_bh(&c->rcu, clusterip_config_rcu_free);
98 }
99
100 /* decrease the count of entries using/referencing this config. If last
101 * entry(rule) is removed, remove the config from lists, but don't free it
102 * yet, since proc-files could still be holding references */
103 static inline void
104 clusterip_config_entry_put(struct net *net, struct clusterip_config *c)
105 {
106 struct clusterip_net *cn = net_generic(net, clusterip_net_id);
107
108 local_bh_disable();
109 if (refcount_dec_and_lock(&c->entries, &cn->lock)) {
110 /* In case anyone still accesses the file, the open/close
111 * functions are also incrementing the refcount on their own,
112 * so it's safe to remove the entry even if it's in use. */
113 #ifdef CONFIG_PROC_FS
114 if (cn->procdir)
115 proc_remove(c->pde);
116 #endif
117 list_del_rcu(&c->list);
118 spin_unlock(&cn->lock);
119 local_bh_enable();
120
121 unregister_netdevice_notifier(&c->notifier);
122
123 return;
124 }
125 local_bh_enable();
126 }
127
128 static struct clusterip_config *
129 __clusterip_config_find(struct net *net, __be32 clusterip)
130 {
131 struct clusterip_config *c;
132 struct clusterip_net *cn = net_generic(net, clusterip_net_id);
133
134 list_for_each_entry_rcu(c, &cn->configs, list) {
135 if (c->clusterip == clusterip)
136 return c;
137 }
138
139 return NULL;
140 }
141
142 static inline struct clusterip_config *
143 clusterip_config_find_get(struct net *net, __be32 clusterip, int entry)
144 {
145 struct clusterip_config *c;
146
147 rcu_read_lock_bh();
148 c = __clusterip_config_find(net, clusterip);
149 if (c) {
150 #ifdef CONFIG_PROC_FS
151 if (!c->pde)
152 c = NULL;
153 else
154 #endif
155 if (unlikely(!refcount_inc_not_zero(&c->refcount)))
156 c = NULL;
157 else if (entry)
158 refcount_inc(&c->entries);
159 }
160 rcu_read_unlock_bh();
161
162 return c;
163 }
164
165 static void
166 clusterip_config_init_nodelist(struct clusterip_config *c,
167 const struct ipt_clusterip_tgt_info *i)
168 {
169 int n;
170
171 for (n = 0; n < i->num_local_nodes; n++)
172 set_bit(i->local_nodes[n] - 1, &c->local_nodes);
173 }
174
175 static int
176 clusterip_netdev_event(struct notifier_block *this, unsigned long event,
177 void *ptr)
178 {
179 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
180 struct clusterip_config *c;
181
182 c = container_of(this, struct clusterip_config, notifier);
183 switch (event) {
184 case NETDEV_REGISTER:
185 if (!strcmp(dev->name, c->ifname)) {
186 c->ifindex = dev->ifindex;
187 dev_mc_add(dev, c->clustermac);
188 }
189 break;
190 case NETDEV_UNREGISTER:
191 if (dev->ifindex == c->ifindex) {
192 dev_mc_del(dev, c->clustermac);
193 c->ifindex = -1;
194 }
195 break;
196 case NETDEV_CHANGENAME:
197 if (!strcmp(dev->name, c->ifname)) {
198 c->ifindex = dev->ifindex;
199 dev_mc_add(dev, c->clustermac);
200 } else if (dev->ifindex == c->ifindex) {
201 dev_mc_del(dev, c->clustermac);
202 c->ifindex = -1;
203 }
204 break;
205 }
206
207 return NOTIFY_DONE;
208 }
209
210 static struct clusterip_config *
211 clusterip_config_init(struct net *net, const struct ipt_clusterip_tgt_info *i,
212 __be32 ip, const char *iniface)
213 {
214 struct clusterip_net *cn = net_generic(net, clusterip_net_id);
215 struct clusterip_config *c;
216 int err;
217
218 c = kzalloc(sizeof(*c), GFP_ATOMIC);
219 if (!c)
220 return ERR_PTR(-ENOMEM);
221
222 strcpy(c->ifname, iniface);
223 c->ifindex = -1;
224 c->clusterip = ip;
225 memcpy(&c->clustermac, &i->clustermac, ETH_ALEN);
226 c->num_total_nodes = i->num_total_nodes;
227 clusterip_config_init_nodelist(c, i);
228 c->hash_mode = i->hash_mode;
229 c->hash_initval = i->hash_initval;
230 refcount_set(&c->refcount, 1);
231
232 spin_lock_bh(&cn->lock);
233 if (__clusterip_config_find(net, ip)) {
234 spin_unlock_bh(&cn->lock);
235 kfree(c);
236
237 return ERR_PTR(-EBUSY);
238 }
239
240 list_add_rcu(&c->list, &cn->configs);
241 spin_unlock_bh(&cn->lock);
242
243 #ifdef CONFIG_PROC_FS
244 {
245 char buffer[16];
246
247 /* create proc dir entry */
248 sprintf(buffer, "%pI4", &ip);
249 c->pde = proc_create_data(buffer, S_IWUSR|S_IRUSR,
250 cn->procdir,
251 &clusterip_proc_fops, c);
252 if (!c->pde) {
253 err = -ENOMEM;
254 goto err;
255 }
256 }
257 #endif
258
259 c->notifier.notifier_call = clusterip_netdev_event;
260 err = register_netdevice_notifier(&c->notifier);
261 if (!err) {
262 refcount_set(&c->entries, 1);
263 return c;
264 }
265
266 #ifdef CONFIG_PROC_FS
267 proc_remove(c->pde);
268 err:
269 #endif
270 spin_lock_bh(&cn->lock);
271 list_del_rcu(&c->list);
272 spin_unlock_bh(&cn->lock);
273 clusterip_config_put(c);
274
275 return ERR_PTR(err);
276 }
277
278 #ifdef CONFIG_PROC_FS
279 static int
280 clusterip_add_node(struct clusterip_config *c, u_int16_t nodenum)
281 {
282
283 if (nodenum == 0 ||
284 nodenum > c->num_total_nodes)
285 return 1;
286
287 /* check if we already have this number in our bitfield */
288 if (test_and_set_bit(nodenum - 1, &c->local_nodes))
289 return 1;
290
291 return 0;
292 }
293
294 static bool
295 clusterip_del_node(struct clusterip_config *c, u_int16_t nodenum)
296 {
297 if (nodenum == 0 ||
298 nodenum > c->num_total_nodes)
299 return true;
300
301 if (test_and_clear_bit(nodenum - 1, &c->local_nodes))
302 return false;
303
304 return true;
305 }
306 #endif
307
308 static inline u_int32_t
309 clusterip_hashfn(const struct sk_buff *skb,
310 const struct clusterip_config *config)
311 {
312 const struct iphdr *iph = ip_hdr(skb);
313 unsigned long hashval;
314 u_int16_t sport = 0, dport = 0;
315 int poff;
316
317 poff = proto_ports_offset(iph->protocol);
318 if (poff >= 0) {
319 const u_int16_t *ports;
320 u16 _ports[2];
321
322 ports = skb_header_pointer(skb, iph->ihl * 4 + poff, 4, _ports);
323 if (ports) {
324 sport = ports[0];
325 dport = ports[1];
326 }
327 } else {
328 net_info_ratelimited("unknown protocol %u\n", iph->protocol);
329 }
330
331 switch (config->hash_mode) {
332 case CLUSTERIP_HASHMODE_SIP:
333 hashval = jhash_1word(ntohl(iph->saddr),
334 config->hash_initval);
335 break;
336 case CLUSTERIP_HASHMODE_SIP_SPT:
337 hashval = jhash_2words(ntohl(iph->saddr), sport,
338 config->hash_initval);
339 break;
340 case CLUSTERIP_HASHMODE_SIP_SPT_DPT:
341 hashval = jhash_3words(ntohl(iph->saddr), sport, dport,
342 config->hash_initval);
343 break;
344 default:
345 /* to make gcc happy */
346 hashval = 0;
347 /* This cannot happen, unless the check function wasn't called
348 * at rule load time */
349 pr_info("unknown mode %u\n", config->hash_mode);
350 BUG();
351 break;
352 }
353
354 /* node numbers are 1..n, not 0..n */
355 return reciprocal_scale(hashval, config->num_total_nodes) + 1;
356 }
357
358 static inline int
359 clusterip_responsible(const struct clusterip_config *config, u_int32_t hash)
360 {
361 return test_bit(hash - 1, &config->local_nodes);
362 }
363
364 /***********************************************************************
365 * IPTABLES TARGET
366 ***********************************************************************/
367
368 static unsigned int
369 clusterip_tg(struct sk_buff *skb, const struct xt_action_param *par)
370 {
371 const struct ipt_clusterip_tgt_info *cipinfo = par->targinfo;
372 struct nf_conn *ct;
373 enum ip_conntrack_info ctinfo;
374 u_int32_t hash;
375
376 /* don't need to clusterip_config_get() here, since refcount
377 * is only decremented by destroy() - and ip_tables guarantees
378 * that the ->target() function isn't called after ->destroy() */
379
380 ct = nf_ct_get(skb, &ctinfo);
381 if (ct == NULL)
382 return NF_DROP;
383
384 /* special case: ICMP error handling. conntrack distinguishes between
385 * error messages (RELATED) and information requests (see below) */
386 if (ip_hdr(skb)->protocol == IPPROTO_ICMP &&
387 (ctinfo == IP_CT_RELATED ||
388 ctinfo == IP_CT_RELATED_REPLY))
389 return XT_CONTINUE;
390
391 /* ip_conntrack_icmp guarantees us that we only have ICMP_ECHO,
392 * TIMESTAMP, INFO_REQUEST or ADDRESS type icmp packets from here
393 * on, which all have an ID field [relevant for hashing]. */
394
395 hash = clusterip_hashfn(skb, cipinfo->config);
396
397 switch (ctinfo) {
398 case IP_CT_NEW:
399 ct->mark = hash;
400 break;
401 case IP_CT_RELATED:
402 case IP_CT_RELATED_REPLY:
403 /* FIXME: we don't handle expectations at the moment.
404 * They can arrive on a different node than
405 * the master connection (e.g. FTP passive mode) */
406 case IP_CT_ESTABLISHED:
407 case IP_CT_ESTABLISHED_REPLY:
408 break;
409 default: /* Prevent gcc warnings */
410 break;
411 }
412
413 #ifdef DEBUG
414 nf_ct_dump_tuple_ip(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple);
415 #endif
416 pr_debug("hash=%u ct_hash=%u ", hash, ct->mark);
417 if (!clusterip_responsible(cipinfo->config, hash)) {
418 pr_debug("not responsible\n");
419 return NF_DROP;
420 }
421 pr_debug("responsible\n");
422
423 /* despite being received via linklayer multicast, this is
424 * actually a unicast IP packet. TCP doesn't like PACKET_MULTICAST */
425 skb->pkt_type = PACKET_HOST;
426
427 return XT_CONTINUE;
428 }
429
430 static int clusterip_tg_check(const struct xt_tgchk_param *par)
431 {
432 struct ipt_clusterip_tgt_info *cipinfo = par->targinfo;
433 const struct ipt_entry *e = par->entryinfo;
434 struct clusterip_config *config;
435 int ret, i;
436
437 if (par->nft_compat) {
438 pr_err("cannot use CLUSTERIP target from nftables compat\n");
439 return -EOPNOTSUPP;
440 }
441
442 if (cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP &&
443 cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP_SPT &&
444 cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP_SPT_DPT) {
445 pr_info("unknown mode %u\n", cipinfo->hash_mode);
446 return -EINVAL;
447
448 }
449 if (e->ip.dmsk.s_addr != htonl(0xffffffff) ||
450 e->ip.dst.s_addr == 0) {
451 pr_info("Please specify destination IP\n");
452 return -EINVAL;
453 }
454 if (cipinfo->num_local_nodes > ARRAY_SIZE(cipinfo->local_nodes)) {
455 pr_info("bad num_local_nodes %u\n", cipinfo->num_local_nodes);
456 return -EINVAL;
457 }
458 for (i = 0; i < cipinfo->num_local_nodes; i++) {
459 if (cipinfo->local_nodes[i] - 1 >=
460 sizeof(config->local_nodes) * 8) {
461 pr_info("bad local_nodes[%d] %u\n",
462 i, cipinfo->local_nodes[i]);
463 return -EINVAL;
464 }
465 }
466
467 config = clusterip_config_find_get(par->net, e->ip.dst.s_addr, 1);
468 if (!config) {
469 if (!(cipinfo->flags & CLUSTERIP_FLAG_NEW)) {
470 pr_info("no config found for %pI4, need 'new'\n",
471 &e->ip.dst.s_addr);
472 return -EINVAL;
473 } else {
474 struct net_device *dev;
475
476 if (e->ip.iniface[0] == '\0') {
477 pr_info("Please specify an interface name\n");
478 return -EINVAL;
479 }
480
481 dev = dev_get_by_name(par->net, e->ip.iniface);
482 if (!dev) {
483 pr_info("no such interface %s\n",
484 e->ip.iniface);
485 return -ENOENT;
486 }
487 dev_put(dev);
488
489 config = clusterip_config_init(par->net, cipinfo,
490 e->ip.dst.s_addr,
491 e->ip.iniface);
492 if (IS_ERR(config))
493 return PTR_ERR(config);
494 }
495 }
496
497 ret = nf_ct_netns_get(par->net, par->family);
498 if (ret < 0) {
499 pr_info("cannot load conntrack support for proto=%u\n",
500 par->family);
501 clusterip_config_entry_put(par->net, config);
502 clusterip_config_put(config);
503 return ret;
504 }
505
506 if (!par->net->xt.clusterip_deprecated_warning) {
507 pr_info("ipt_CLUSTERIP is deprecated and it will removed soon, "
508 "use xt_cluster instead\n");
509 par->net->xt.clusterip_deprecated_warning = true;
510 }
511
512 cipinfo->config = config;
513 return ret;
514 }
515
516 /* drop reference count of cluster config when rule is deleted */
517 static void clusterip_tg_destroy(const struct xt_tgdtor_param *par)
518 {
519 const struct ipt_clusterip_tgt_info *cipinfo = par->targinfo;
520
521 /* if no more entries are referencing the config, remove it
522 * from the list and destroy the proc entry */
523 clusterip_config_entry_put(par->net, cipinfo->config);
524
525 clusterip_config_put(cipinfo->config);
526
527 nf_ct_netns_put(par->net, par->family);
528 }
529
530 #ifdef CONFIG_COMPAT
531 struct compat_ipt_clusterip_tgt_info
532 {
533 u_int32_t flags;
534 u_int8_t clustermac[6];
535 u_int16_t num_total_nodes;
536 u_int16_t num_local_nodes;
537 u_int16_t local_nodes[CLUSTERIP_MAX_NODES];
538 u_int32_t hash_mode;
539 u_int32_t hash_initval;
540 compat_uptr_t config;
541 };
542 #endif /* CONFIG_COMPAT */
543
544 static struct xt_target clusterip_tg_reg __read_mostly = {
545 .name = "CLUSTERIP",
546 .family = NFPROTO_IPV4,
547 .target = clusterip_tg,
548 .checkentry = clusterip_tg_check,
549 .destroy = clusterip_tg_destroy,
550 .targetsize = sizeof(struct ipt_clusterip_tgt_info),
551 .usersize = offsetof(struct ipt_clusterip_tgt_info, config),
552 #ifdef CONFIG_COMPAT
553 .compatsize = sizeof(struct compat_ipt_clusterip_tgt_info),
554 #endif /* CONFIG_COMPAT */
555 .me = THIS_MODULE
556 };
557
558
559 /***********************************************************************
560 * ARP MANGLING CODE
561 ***********************************************************************/
562
563 /* hardcoded for 48bit ethernet and 32bit ipv4 addresses */
564 struct arp_payload {
565 u_int8_t src_hw[ETH_ALEN];
566 __be32 src_ip;
567 u_int8_t dst_hw[ETH_ALEN];
568 __be32 dst_ip;
569 } __packed;
570
571 #ifdef DEBUG
572 static void arp_print(struct arp_payload *payload)
573 {
574 #define HBUFFERLEN 30
575 char hbuffer[HBUFFERLEN];
576 int j, k;
577
578 for (k = 0, j = 0; k < HBUFFERLEN - 3 && j < ETH_ALEN; j++) {
579 hbuffer[k++] = hex_asc_hi(payload->src_hw[j]);
580 hbuffer[k++] = hex_asc_lo(payload->src_hw[j]);
581 hbuffer[k++] = ':';
582 }
583 hbuffer[--k] = '\0';
584
585 pr_debug("src %pI4@%s, dst %pI4\n",
586 &payload->src_ip, hbuffer, &payload->dst_ip);
587 }
588 #endif
589
590 static unsigned int
591 arp_mangle(void *priv,
592 struct sk_buff *skb,
593 const struct nf_hook_state *state)
594 {
595 struct arphdr *arp = arp_hdr(skb);
596 struct arp_payload *payload;
597 struct clusterip_config *c;
598 struct net *net = state->net;
599
600 /* we don't care about non-ethernet and non-ipv4 ARP */
601 if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
602 arp->ar_pro != htons(ETH_P_IP) ||
603 arp->ar_pln != 4 || arp->ar_hln != ETH_ALEN)
604 return NF_ACCEPT;
605
606 /* we only want to mangle arp requests and replies */
607 if (arp->ar_op != htons(ARPOP_REPLY) &&
608 arp->ar_op != htons(ARPOP_REQUEST))
609 return NF_ACCEPT;
610
611 payload = (void *)(arp+1);
612
613 /* if there is no clusterip configuration for the arp reply's
614 * source ip, we don't want to mangle it */
615 c = clusterip_config_find_get(net, payload->src_ip, 0);
616 if (!c)
617 return NF_ACCEPT;
618
619 /* normally the linux kernel always replies to arp queries of
620 * addresses on different interfacs. However, in the CLUSTERIP case
621 * this wouldn't work, since we didn't subscribe the mcast group on
622 * other interfaces */
623 if (c->ifindex != state->out->ifindex) {
624 pr_debug("not mangling arp reply on different interface: cip'%d'-skb'%d'\n",
625 c->ifindex, state->out->ifindex);
626 clusterip_config_put(c);
627 return NF_ACCEPT;
628 }
629
630 /* mangle reply hardware address */
631 memcpy(payload->src_hw, c->clustermac, arp->ar_hln);
632
633 #ifdef DEBUG
634 pr_debug("mangled arp reply: ");
635 arp_print(payload);
636 #endif
637
638 clusterip_config_put(c);
639
640 return NF_ACCEPT;
641 }
642
643 static const struct nf_hook_ops cip_arp_ops = {
644 .hook = arp_mangle,
645 .pf = NFPROTO_ARP,
646 .hooknum = NF_ARP_OUT,
647 .priority = -1
648 };
649
650 /***********************************************************************
651 * PROC DIR HANDLING
652 ***********************************************************************/
653
654 #ifdef CONFIG_PROC_FS
655
656 struct clusterip_seq_position {
657 unsigned int pos; /* position */
658 unsigned int weight; /* number of bits set == size */
659 unsigned int bit; /* current bit */
660 unsigned long val; /* current value */
661 };
662
663 static void *clusterip_seq_start(struct seq_file *s, loff_t *pos)
664 {
665 struct clusterip_config *c = s->private;
666 unsigned int weight;
667 u_int32_t local_nodes;
668 struct clusterip_seq_position *idx;
669
670 /* FIXME: possible race */
671 local_nodes = c->local_nodes;
672 weight = hweight32(local_nodes);
673 if (*pos >= weight)
674 return NULL;
675
676 idx = kmalloc(sizeof(struct clusterip_seq_position), GFP_KERNEL);
677 if (!idx)
678 return ERR_PTR(-ENOMEM);
679
680 idx->pos = *pos;
681 idx->weight = weight;
682 idx->bit = ffs(local_nodes);
683 idx->val = local_nodes;
684 clear_bit(idx->bit - 1, &idx->val);
685
686 return idx;
687 }
688
689 static void *clusterip_seq_next(struct seq_file *s, void *v, loff_t *pos)
690 {
691 struct clusterip_seq_position *idx = v;
692
693 *pos = ++idx->pos;
694 if (*pos >= idx->weight) {
695 kfree(v);
696 return NULL;
697 }
698 idx->bit = ffs(idx->val);
699 clear_bit(idx->bit - 1, &idx->val);
700 return idx;
701 }
702
703 static void clusterip_seq_stop(struct seq_file *s, void *v)
704 {
705 if (!IS_ERR(v))
706 kfree(v);
707 }
708
709 static int clusterip_seq_show(struct seq_file *s, void *v)
710 {
711 struct clusterip_seq_position *idx = v;
712
713 if (idx->pos != 0)
714 seq_putc(s, ',');
715
716 seq_printf(s, "%u", idx->bit);
717
718 if (idx->pos == idx->weight - 1)
719 seq_putc(s, '\n');
720
721 return 0;
722 }
723
724 static const struct seq_operations clusterip_seq_ops = {
725 .start = clusterip_seq_start,
726 .next = clusterip_seq_next,
727 .stop = clusterip_seq_stop,
728 .show = clusterip_seq_show,
729 };
730
731 static int clusterip_proc_open(struct inode *inode, struct file *file)
732 {
733 int ret = seq_open(file, &clusterip_seq_ops);
734
735 if (!ret) {
736 struct seq_file *sf = file->private_data;
737 struct clusterip_config *c = PDE_DATA(inode);
738
739 sf->private = c;
740
741 clusterip_config_get(c);
742 }
743
744 return ret;
745 }
746
747 static int clusterip_proc_release(struct inode *inode, struct file *file)
748 {
749 struct clusterip_config *c = PDE_DATA(inode);
750 int ret;
751
752 ret = seq_release(inode, file);
753
754 if (!ret)
755 clusterip_config_put(c);
756
757 return ret;
758 }
759
760 static ssize_t clusterip_proc_write(struct file *file, const char __user *input,
761 size_t size, loff_t *ofs)
762 {
763 struct clusterip_config *c = PDE_DATA(file_inode(file));
764 #define PROC_WRITELEN 10
765 char buffer[PROC_WRITELEN+1];
766 unsigned long nodenum;
767 int rc;
768
769 if (size > PROC_WRITELEN)
770 return -EIO;
771 if (copy_from_user(buffer, input, size))
772 return -EFAULT;
773 buffer[size] = 0;
774
775 if (*buffer == '+') {
776 rc = kstrtoul(buffer+1, 10, &nodenum);
777 if (rc)
778 return rc;
779 if (clusterip_add_node(c, nodenum))
780 return -ENOMEM;
781 } else if (*buffer == '-') {
782 rc = kstrtoul(buffer+1, 10, &nodenum);
783 if (rc)
784 return rc;
785 if (clusterip_del_node(c, nodenum))
786 return -ENOENT;
787 } else
788 return -EIO;
789
790 return size;
791 }
792
793 static const struct file_operations clusterip_proc_fops = {
794 .owner = THIS_MODULE,
795 .open = clusterip_proc_open,
796 .read = seq_read,
797 .write = clusterip_proc_write,
798 .llseek = seq_lseek,
799 .release = clusterip_proc_release,
800 };
801
802 #endif /* CONFIG_PROC_FS */
803
804 static int clusterip_net_init(struct net *net)
805 {
806 struct clusterip_net *cn = net_generic(net, clusterip_net_id);
807 int ret;
808
809 INIT_LIST_HEAD(&cn->configs);
810
811 spin_lock_init(&cn->lock);
812
813 ret = nf_register_net_hook(net, &cip_arp_ops);
814 if (ret < 0)
815 return ret;
816
817 #ifdef CONFIG_PROC_FS
818 cn->procdir = proc_mkdir("ipt_CLUSTERIP", net->proc_net);
819 if (!cn->procdir) {
820 nf_unregister_net_hook(net, &cip_arp_ops);
821 pr_err("Unable to proc dir entry\n");
822 return -ENOMEM;
823 }
824 #endif /* CONFIG_PROC_FS */
825
826 return 0;
827 }
828
829 static void clusterip_net_exit(struct net *net)
830 {
831 struct clusterip_net *cn = net_generic(net, clusterip_net_id);
832 #ifdef CONFIG_PROC_FS
833 proc_remove(cn->procdir);
834 cn->procdir = NULL;
835 #endif
836 nf_unregister_net_hook(net, &cip_arp_ops);
837 WARN_ON_ONCE(!list_empty(&cn->configs));
838 }
839
840 static struct pernet_operations clusterip_net_ops = {
841 .init = clusterip_net_init,
842 .exit = clusterip_net_exit,
843 .id = &clusterip_net_id,
844 .size = sizeof(struct clusterip_net),
845 };
846
847 static int __init clusterip_tg_init(void)
848 {
849 int ret;
850
851 ret = register_pernet_subsys(&clusterip_net_ops);
852 if (ret < 0)
853 return ret;
854
855 ret = xt_register_target(&clusterip_tg_reg);
856 if (ret < 0)
857 goto cleanup_subsys;
858
859 pr_info("ClusterIP Version %s loaded successfully\n",
860 CLUSTERIP_VERSION);
861
862 return 0;
863
864 cleanup_subsys:
865 unregister_pernet_subsys(&clusterip_net_ops);
866 return ret;
867 }
868
869 static void __exit clusterip_tg_exit(void)
870 {
871 pr_info("ClusterIP Version %s unloading\n", CLUSTERIP_VERSION);
872
873 xt_unregister_target(&clusterip_tg_reg);
874 unregister_pernet_subsys(&clusterip_net_ops);
875
876 /* Wait for completion of call_rcu_bh()'s (clusterip_config_rcu_free) */
877 rcu_barrier_bh();
878 }
879
880 module_init(clusterip_tg_init);
881 module_exit(clusterip_tg_exit);
ubuntu-bionic-kernel mirror