~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/ipv4/netfilter/ipt_CLUSTERIP.c

Version: ~ [ linux-5.4-rc3 ] ~ [ linux-5.3.6 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.79 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.149 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.196 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.196 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.75 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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/netfilter_arp.h>
 26 #include <linux/netfilter/x_tables.h>
 27 #include <linux/netfilter_ipv4/ip_tables.h>
 28 #include <linux/netfilter_ipv4/ipt_CLUSTERIP.h>
 29 #include <net/netfilter/nf_conntrack.h>
 30 #include <net/net_namespace.h>
 31 #include <net/netns/generic.h>
 32 #include <net/checksum.h>
 33 #include <net/ip.h>
 34 
 35 #define CLUSTERIP_VERSION "0.8"
 36 
 37 MODULE_LICENSE("GPL");
 38 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
 39 MODULE_DESCRIPTION("Xtables: CLUSTERIP target");
 40 
 41 struct clusterip_config {
 42         struct list_head list;                  /* list of all configs */
 43         atomic_t refcount;                      /* reference count */
 44         atomic_t entries;                       /* number of entries/rules
 45                                                  * referencing us */
 46 
 47         __be32 clusterip;                       /* the IP address */
 48         u_int8_t clustermac[ETH_ALEN];          /* the MAC address */
 49         struct net_device *dev;                 /* device */
 50         u_int16_t num_total_nodes;              /* total number of nodes */
 51         unsigned long local_nodes;              /* node number array */
 52 
 53 #ifdef CONFIG_PROC_FS
 54         struct proc_dir_entry *pde;             /* proc dir entry */
 55 #endif
 56         enum clusterip_hashmode hash_mode;      /* which hashing mode */
 57         u_int32_t hash_initval;                 /* hash initialization */
 58         struct rcu_head rcu;
 59 };
 60 
 61 #ifdef CONFIG_PROC_FS
 62 static const struct file_operations clusterip_proc_fops;
 63 #endif
 64 
 65 static int clusterip_net_id __read_mostly;
 66 
 67 struct clusterip_net {
 68         struct list_head configs;
 69         /* lock protects the configs list */
 70         spinlock_t lock;
 71 
 72 #ifdef CONFIG_PROC_FS
 73         struct proc_dir_entry *procdir;
 74 #endif
 75 };
 76 
 77 static inline void
 78 clusterip_config_get(struct clusterip_config *c)
 79 {
 80         atomic_inc(&c->refcount);
 81 }
 82 
 83 
 84 static void clusterip_config_rcu_free(struct rcu_head *head)
 85 {
 86         kfree(container_of(head, struct clusterip_config, rcu));
 87 }
 88 
 89 static inline void
 90 clusterip_config_put(struct clusterip_config *c)
 91 {
 92         if (atomic_dec_and_test(&c->refcount))
 93                 call_rcu_bh(&c->rcu, clusterip_config_rcu_free);
 94 }
 95 
 96 /* decrease the count of entries using/referencing this config.  If last
 97  * entry(rule) is removed, remove the config from lists, but don't free it
 98  * yet, since proc-files could still be holding references */
 99 static inline void
100 clusterip_config_entry_put(struct clusterip_config *c)
101 {
102         struct net *net = dev_net(c->dev);
103         struct clusterip_net *cn = net_generic(net, clusterip_net_id);
104 
105         local_bh_disable();
106         if (atomic_dec_and_lock(&c->entries, &cn->lock)) {
107                 list_del_rcu(&c->list);
108                 spin_unlock(&cn->lock);
109                 local_bh_enable();
110 
111                 dev_mc_del(c->dev, c->clustermac);
112                 dev_put(c->dev);
113 
114                 /* In case anyone still accesses the file, the open/close
115                  * functions are also incrementing the refcount on their own,
116                  * so it's safe to remove the entry even if it's in use. */
117 #ifdef CONFIG_PROC_FS
118                 proc_remove(c->pde);
119 #endif
120                 return;
121         }
122         local_bh_enable();
123 }
124 
125 static struct clusterip_config *
126 __clusterip_config_find(struct net *net, __be32 clusterip)
127 {
128         struct clusterip_config *c;
129         struct clusterip_net *cn = net_generic(net, clusterip_net_id);
130 
131         list_for_each_entry_rcu(c, &cn->configs, list) {
132                 if (c->clusterip == clusterip)
133                         return c;
134         }
135 
136         return NULL;
137 }
138 
139 static inline struct clusterip_config *
140 clusterip_config_find_get(struct net *net, __be32 clusterip, int entry)
141 {
142         struct clusterip_config *c;
143 
144         rcu_read_lock_bh();
145         c = __clusterip_config_find(net, clusterip);
146         if (c) {
147                 if (unlikely(!atomic_inc_not_zero(&c->refcount)))
148                         c = NULL;
149                 else if (entry)
150                         atomic_inc(&c->entries);
151         }
152         rcu_read_unlock_bh();
153 
154         return c;
155 }
156 
157 static void
158 clusterip_config_init_nodelist(struct clusterip_config *c,
159                                const struct ipt_clusterip_tgt_info *i)
160 {
161         int n;
162 
163         for (n = 0; n < i->num_local_nodes; n++)
164                 set_bit(i->local_nodes[n] - 1, &c->local_nodes);
165 }
166 
167 static struct clusterip_config *
168 clusterip_config_init(const struct ipt_clusterip_tgt_info *i, __be32 ip,
169                         struct net_device *dev)
170 {
171         struct clusterip_config *c;
172         struct clusterip_net *cn = net_generic(dev_net(dev), clusterip_net_id);
173 
174         c = kzalloc(sizeof(*c), GFP_ATOMIC);
175         if (!c)
176                 return NULL;
177 
178         c->dev = dev;
179         c->clusterip = ip;
180         memcpy(&c->clustermac, &i->clustermac, ETH_ALEN);
181         c->num_total_nodes = i->num_total_nodes;
182         clusterip_config_init_nodelist(c, i);
183         c->hash_mode = i->hash_mode;
184         c->hash_initval = i->hash_initval;
185         atomic_set(&c->refcount, 1);
186         atomic_set(&c->entries, 1);
187 
188 #ifdef CONFIG_PROC_FS
189         {
190                 char buffer[16];
191 
192                 /* create proc dir entry */
193                 sprintf(buffer, "%pI4", &ip);
194                 c->pde = proc_create_data(buffer, S_IWUSR|S_IRUSR,
195                                           cn->procdir,
196                                           &clusterip_proc_fops, c);
197                 if (!c->pde) {
198                         kfree(c);
199                         return NULL;
200                 }
201         }
202 #endif
203 
204         spin_lock_bh(&cn->lock);
205         list_add_rcu(&c->list, &cn->configs);
206         spin_unlock_bh(&cn->lock);
207 
208         return c;
209 }
210 
211 #ifdef CONFIG_PROC_FS
212 static int
213 clusterip_add_node(struct clusterip_config *c, u_int16_t nodenum)
214 {
215 
216         if (nodenum == 0 ||
217             nodenum > c->num_total_nodes)
218                 return 1;
219 
220         /* check if we already have this number in our bitfield */
221         if (test_and_set_bit(nodenum - 1, &c->local_nodes))
222                 return 1;
223 
224         return 0;
225 }
226 
227 static bool
228 clusterip_del_node(struct clusterip_config *c, u_int16_t nodenum)
229 {
230         if (nodenum == 0 ||
231             nodenum > c->num_total_nodes)
232                 return true;
233 
234         if (test_and_clear_bit(nodenum - 1, &c->local_nodes))
235                 return false;
236 
237         return true;
238 }
239 #endif
240 
241 static inline u_int32_t
242 clusterip_hashfn(const struct sk_buff *skb,
243                  const struct clusterip_config *config)
244 {
245         const struct iphdr *iph = ip_hdr(skb);
246         unsigned long hashval;
247         u_int16_t sport = 0, dport = 0;
248         int poff;
249 
250         poff = proto_ports_offset(iph->protocol);
251         if (poff >= 0) {
252                 const u_int16_t *ports;
253                 u16 _ports[2];
254 
255                 ports = skb_header_pointer(skb, iph->ihl * 4 + poff, 4, _ports);
256                 if (ports) {
257                         sport = ports[0];
258                         dport = ports[1];
259                 }
260         } else {
261                 net_info_ratelimited("unknown protocol %u\n", iph->protocol);
262         }
263 
264         switch (config->hash_mode) {
265         case CLUSTERIP_HASHMODE_SIP:
266                 hashval = jhash_1word(ntohl(iph->saddr),
267                                       config->hash_initval);
268                 break;
269         case CLUSTERIP_HASHMODE_SIP_SPT:
270                 hashval = jhash_2words(ntohl(iph->saddr), sport,
271                                        config->hash_initval);
272                 break;
273         case CLUSTERIP_HASHMODE_SIP_SPT_DPT:
274                 hashval = jhash_3words(ntohl(iph->saddr), sport, dport,
275                                        config->hash_initval);
276                 break;
277         default:
278                 /* to make gcc happy */
279                 hashval = 0;
280                 /* This cannot happen, unless the check function wasn't called
281                  * at rule load time */
282                 pr_info("unknown mode %u\n", config->hash_mode);
283                 BUG();
284                 break;
285         }
286 
287         /* node numbers are 1..n, not 0..n */
288         return reciprocal_scale(hashval, config->num_total_nodes) + 1;
289 }
290 
291 static inline int
292 clusterip_responsible(const struct clusterip_config *config, u_int32_t hash)
293 {
294         return test_bit(hash - 1, &config->local_nodes);
295 }
296 
297 /***********************************************************************
298  * IPTABLES TARGET
299  ***********************************************************************/
300 
301 static unsigned int
302 clusterip_tg(struct sk_buff *skb, const struct xt_action_param *par)
303 {
304         const struct ipt_clusterip_tgt_info *cipinfo = par->targinfo;
305         struct nf_conn *ct;
306         enum ip_conntrack_info ctinfo;
307         u_int32_t hash;
308 
309         /* don't need to clusterip_config_get() here, since refcount
310          * is only decremented by destroy() - and ip_tables guarantees
311          * that the ->target() function isn't called after ->destroy() */
312 
313         ct = nf_ct_get(skb, &ctinfo);
314         if (ct == NULL)
315                 return NF_DROP;
316 
317         /* special case: ICMP error handling. conntrack distinguishes between
318          * error messages (RELATED) and information requests (see below) */
319         if (ip_hdr(skb)->protocol == IPPROTO_ICMP &&
320             (ctinfo == IP_CT_RELATED ||
321              ctinfo == IP_CT_RELATED_REPLY))
322                 return XT_CONTINUE;
323 
324         /* ip_conntrack_icmp guarantees us that we only have ICMP_ECHO,
325          * TIMESTAMP, INFO_REQUEST or ADDRESS type icmp packets from here
326          * on, which all have an ID field [relevant for hashing]. */
327 
328         hash = clusterip_hashfn(skb, cipinfo->config);
329 
330         switch (ctinfo) {
331         case IP_CT_NEW:
332                 ct->mark = hash;
333                 break;
334         case IP_CT_RELATED:
335         case IP_CT_RELATED_REPLY:
336                 /* FIXME: we don't handle expectations at the moment.
337                  * They can arrive on a different node than
338                  * the master connection (e.g. FTP passive mode) */
339         case IP_CT_ESTABLISHED:
340         case IP_CT_ESTABLISHED_REPLY:
341                 break;
342         default:                        /* Prevent gcc warnings */
343                 break;
344         }
345 
346 #ifdef DEBUG
347         nf_ct_dump_tuple_ip(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple);
348 #endif
349         pr_debug("hash=%u ct_hash=%u ", hash, ct->mark);
350         if (!clusterip_responsible(cipinfo->config, hash)) {
351                 pr_debug("not responsible\n");
352                 return NF_DROP;
353         }
354         pr_debug("responsible\n");
355 
356         /* despite being received via linklayer multicast, this is
357          * actually a unicast IP packet. TCP doesn't like PACKET_MULTICAST */
358         skb->pkt_type = PACKET_HOST;
359 
360         return XT_CONTINUE;
361 }
362 
363 static int clusterip_tg_check(const struct xt_tgchk_param *par)
364 {
365         struct ipt_clusterip_tgt_info *cipinfo = par->targinfo;
366         const struct ipt_entry *e = par->entryinfo;
367         struct clusterip_config *config;
368         int ret, i;
369 
370         if (cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP &&
371             cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP_SPT &&
372             cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP_SPT_DPT) {
373                 pr_info("unknown mode %u\n", cipinfo->hash_mode);
374                 return -EINVAL;
375 
376         }
377         if (e->ip.dmsk.s_addr != htonl(0xffffffff) ||
378             e->ip.dst.s_addr == 0) {
379                 pr_info("Please specify destination IP\n");
380                 return -EINVAL;
381         }
382         if (cipinfo->num_local_nodes > ARRAY_SIZE(cipinfo->local_nodes)) {
383                 pr_info("bad num_local_nodes %u\n", cipinfo->num_local_nodes);
384                 return -EINVAL;
385         }
386         for (i = 0; i < cipinfo->num_local_nodes; i++) {
387                 if (cipinfo->local_nodes[i] - 1 >=
388                     sizeof(config->local_nodes) * 8) {
389                         pr_info("bad local_nodes[%d] %u\n",
390                                 i, cipinfo->local_nodes[i]);
391                         return -EINVAL;
392                 }
393         }
394 
395         config = clusterip_config_find_get(par->net, e->ip.dst.s_addr, 1);
396         if (!config) {
397                 if (!(cipinfo->flags & CLUSTERIP_FLAG_NEW)) {
398                         pr_info("no config found for %pI4, need 'new'\n",
399                                 &e->ip.dst.s_addr);
400                         return -EINVAL;
401                 } else {
402                         struct net_device *dev;
403 
404                         if (e->ip.iniface[0] == '\0') {
405                                 pr_info("Please specify an interface name\n");
406                                 return -EINVAL;
407                         }
408 
409                         dev = dev_get_by_name(par->net, e->ip.iniface);
410                         if (!dev) {
411                                 pr_info("no such interface %s\n",
412                                         e->ip.iniface);
413                                 return -ENOENT;
414                         }
415 
416                         config = clusterip_config_init(cipinfo,
417                                                         e->ip.dst.s_addr, dev);
418                         if (!config) {
419                                 dev_put(dev);
420                                 return -ENOMEM;
421                         }
422                         dev_mc_add(config->dev, config->clustermac);
423                 }
424         }
425         cipinfo->config = config;
426 
427         ret = nf_ct_l3proto_try_module_get(par->family);
428         if (ret < 0)
429                 pr_info("cannot load conntrack support for proto=%u\n",
430                         par->family);
431 
432         if (!par->net->xt.clusterip_deprecated_warning) {
433                 pr_info("ipt_CLUSTERIP is deprecated and it will removed soon, "
434                         "use xt_cluster instead\n");
435                 par->net->xt.clusterip_deprecated_warning = true;
436         }
437 
438         return ret;
439 }
440 
441 /* drop reference count of cluster config when rule is deleted */
442 static void clusterip_tg_destroy(const struct xt_tgdtor_param *par)
443 {
444         const struct ipt_clusterip_tgt_info *cipinfo = par->targinfo;
445 
446         /* if no more entries are referencing the config, remove it
447          * from the list and destroy the proc entry */
448         clusterip_config_entry_put(cipinfo->config);
449 
450         clusterip_config_put(cipinfo->config);
451 
452         nf_ct_l3proto_module_put(par->family);
453 }
454 
455 #ifdef CONFIG_COMPAT
456 struct compat_ipt_clusterip_tgt_info
457 {
458         u_int32_t       flags;
459         u_int8_t        clustermac[6];
460         u_int16_t       num_total_nodes;
461         u_int16_t       num_local_nodes;
462         u_int16_t       local_nodes[CLUSTERIP_MAX_NODES];
463         u_int32_t       hash_mode;
464         u_int32_t       hash_initval;
465         compat_uptr_t   config;
466 };
467 #endif /* CONFIG_COMPAT */
468 
469 static struct xt_target clusterip_tg_reg __read_mostly = {
470         .name           = "CLUSTERIP",
471         .family         = NFPROTO_IPV4,
472         .target         = clusterip_tg,
473         .checkentry     = clusterip_tg_check,
474         .destroy        = clusterip_tg_destroy,
475         .targetsize     = sizeof(struct ipt_clusterip_tgt_info),
476 #ifdef CONFIG_COMPAT
477         .compatsize     = sizeof(struct compat_ipt_clusterip_tgt_info),
478 #endif /* CONFIG_COMPAT */
479         .me             = THIS_MODULE
480 };
481 
482 
483 /***********************************************************************
484  * ARP MANGLING CODE
485  ***********************************************************************/
486 
487 /* hardcoded for 48bit ethernet and 32bit ipv4 addresses */
488 struct arp_payload {
489         u_int8_t src_hw[ETH_ALEN];
490         __be32 src_ip;
491         u_int8_t dst_hw[ETH_ALEN];
492         __be32 dst_ip;
493 } __packed;
494 
495 #ifdef DEBUG
496 static void arp_print(struct arp_payload *payload)
497 {
498 #define HBUFFERLEN 30
499         char hbuffer[HBUFFERLEN];
500         int j,k;
501 
502         for (k=0, j=0; k < HBUFFERLEN-3 && j < ETH_ALEN; j++) {
503                 hbuffer[k++] = hex_asc_hi(payload->src_hw[j]);
504                 hbuffer[k++] = hex_asc_lo(payload->src_hw[j]);
505                 hbuffer[k++]=':';
506         }
507         hbuffer[--k]='\0';
508 
509         pr_debug("src %pI4@%s, dst %pI4\n",
510                  &payload->src_ip, hbuffer, &payload->dst_ip);
511 }
512 #endif
513 
514 static unsigned int
515 arp_mangle(const struct nf_hook_ops *ops,
516            struct sk_buff *skb,
517            const struct nf_hook_state *state)
518 {
519         struct arphdr *arp = arp_hdr(skb);
520         struct arp_payload *payload;
521         struct clusterip_config *c;
522         struct net *net = dev_net(state->in ? state->in : state->out);
523 
524         /* we don't care about non-ethernet and non-ipv4 ARP */
525         if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
526             arp->ar_pro != htons(ETH_P_IP) ||
527             arp->ar_pln != 4 || arp->ar_hln != ETH_ALEN)
528                 return NF_ACCEPT;
529 
530         /* we only want to mangle arp requests and replies */
531         if (arp->ar_op != htons(ARPOP_REPLY) &&
532             arp->ar_op != htons(ARPOP_REQUEST))
533                 return NF_ACCEPT;
534 
535         payload = (void *)(arp+1);
536 
537         /* if there is no clusterip configuration for the arp reply's
538          * source ip, we don't want to mangle it */
539         c = clusterip_config_find_get(net, payload->src_ip, 0);
540         if (!c)
541                 return NF_ACCEPT;
542 
543         /* normally the linux kernel always replies to arp queries of
544          * addresses on different interfacs.  However, in the CLUSTERIP case
545          * this wouldn't work, since we didn't subscribe the mcast group on
546          * other interfaces */
547         if (c->dev != state->out) {
548                 pr_debug("not mangling arp reply on different "
549                          "interface: cip'%s'-skb'%s'\n",
550                          c->dev->name, state->out->name);
551                 clusterip_config_put(c);
552                 return NF_ACCEPT;
553         }
554 
555         /* mangle reply hardware address */
556         memcpy(payload->src_hw, c->clustermac, arp->ar_hln);
557 
558 #ifdef DEBUG
559         pr_debug("mangled arp reply: ");
560         arp_print(payload);
561 #endif
562 
563         clusterip_config_put(c);
564 
565         return NF_ACCEPT;
566 }
567 
568 static struct nf_hook_ops cip_arp_ops __read_mostly = {
569         .hook = arp_mangle,
570         .pf = NFPROTO_ARP,
571         .hooknum = NF_ARP_OUT,
572         .priority = -1
573 };
574 
575 /***********************************************************************
576  * PROC DIR HANDLING
577  ***********************************************************************/
578 
579 #ifdef CONFIG_PROC_FS
580 
581 struct clusterip_seq_position {
582         unsigned int pos;       /* position */
583         unsigned int weight;    /* number of bits set == size */
584         unsigned int bit;       /* current bit */
585         unsigned long val;      /* current value */
586 };
587 
588 static void *clusterip_seq_start(struct seq_file *s, loff_t *pos)
589 {
590         struct clusterip_config *c = s->private;
591         unsigned int weight;
592         u_int32_t local_nodes;
593         struct clusterip_seq_position *idx;
594 
595         /* FIXME: possible race */
596         local_nodes = c->local_nodes;
597         weight = hweight32(local_nodes);
598         if (*pos >= weight)
599                 return NULL;
600 
601         idx = kmalloc(sizeof(struct clusterip_seq_position), GFP_KERNEL);
602         if (!idx)
603                 return ERR_PTR(-ENOMEM);
604 
605         idx->pos = *pos;
606         idx->weight = weight;
607         idx->bit = ffs(local_nodes);
608         idx->val = local_nodes;
609         clear_bit(idx->bit - 1, &idx->val);
610 
611         return idx;
612 }
613 
614 static void *clusterip_seq_next(struct seq_file *s, void *v, loff_t *pos)
615 {
616         struct clusterip_seq_position *idx = v;
617 
618         *pos = ++idx->pos;
619         if (*pos >= idx->weight) {
620                 kfree(v);
621                 return NULL;
622         }
623         idx->bit = ffs(idx->val);
624         clear_bit(idx->bit - 1, &idx->val);
625         return idx;
626 }
627 
628 static void clusterip_seq_stop(struct seq_file *s, void *v)
629 {
630         if (!IS_ERR(v))
631                 kfree(v);
632 }
633 
634 static int clusterip_seq_show(struct seq_file *s, void *v)
635 {
636         struct clusterip_seq_position *idx = v;
637 
638         if (idx->pos != 0)
639                 seq_putc(s, ',');
640 
641         seq_printf(s, "%u", idx->bit);
642 
643         if (idx->pos == idx->weight - 1)
644                 seq_putc(s, '\n');
645 
646         return 0;
647 }
648 
649 static const struct seq_operations clusterip_seq_ops = {
650         .start  = clusterip_seq_start,
651         .next   = clusterip_seq_next,
652         .stop   = clusterip_seq_stop,
653         .show   = clusterip_seq_show,
654 };
655 
656 static int clusterip_proc_open(struct inode *inode, struct file *file)
657 {
658         int ret = seq_open(file, &clusterip_seq_ops);
659 
660         if (!ret) {
661                 struct seq_file *sf = file->private_data;
662                 struct clusterip_config *c = PDE_DATA(inode);
663 
664                 sf->private = c;
665 
666                 clusterip_config_get(c);
667         }
668 
669         return ret;
670 }
671 
672 static int clusterip_proc_release(struct inode *inode, struct file *file)
673 {
674         struct clusterip_config *c = PDE_DATA(inode);
675         int ret;
676 
677         ret = seq_release(inode, file);
678 
679         if (!ret)
680                 clusterip_config_put(c);
681 
682         return ret;
683 }
684 
685 static ssize_t clusterip_proc_write(struct file *file, const char __user *input,
686                                 size_t size, loff_t *ofs)
687 {
688         struct clusterip_config *c = PDE_DATA(file_inode(file));
689 #define PROC_WRITELEN   10
690         char buffer[PROC_WRITELEN+1];
691         unsigned long nodenum;
692         int rc;
693 
694         if (size > PROC_WRITELEN)
695                 return -EIO;
696         if (copy_from_user(buffer, input, size))
697                 return -EFAULT;
698         buffer[size] = 0;
699 
700         if (*buffer == '+') {
701                 rc = kstrtoul(buffer+1, 10, &nodenum);
702                 if (rc)
703                         return rc;
704                 if (clusterip_add_node(c, nodenum))
705                         return -ENOMEM;
706         } else if (*buffer == '-') {
707                 rc = kstrtoul(buffer+1, 10, &nodenum);
708                 if (rc)
709                         return rc;
710                 if (clusterip_del_node(c, nodenum))
711                         return -ENOENT;
712         } else
713                 return -EIO;
714 
715         return size;
716 }
717 
718 static const struct file_operations clusterip_proc_fops = {
719         .owner   = THIS_MODULE,
720         .open    = clusterip_proc_open,
721         .read    = seq_read,
722         .write   = clusterip_proc_write,
723         .llseek  = seq_lseek,
724         .release = clusterip_proc_release,
725 };
726 
727 #endif /* CONFIG_PROC_FS */
728 
729 static int clusterip_net_init(struct net *net)
730 {
731         struct clusterip_net *cn = net_generic(net, clusterip_net_id);
732 
733         INIT_LIST_HEAD(&cn->configs);
734 
735         spin_lock_init(&cn->lock);
736 
737 #ifdef CONFIG_PROC_FS
738         cn->procdir = proc_mkdir("ipt_CLUSTERIP", net->proc_net);
739         if (!cn->procdir) {
740                 pr_err("Unable to proc dir entry\n");
741                 return -ENOMEM;
742         }
743 #endif /* CONFIG_PROC_FS */
744 
745         return 0;
746 }
747 
748 static void clusterip_net_exit(struct net *net)
749 {
750 #ifdef CONFIG_PROC_FS
751         struct clusterip_net *cn = net_generic(net, clusterip_net_id);
752         proc_remove(cn->procdir);
753 #endif
754 }
755 
756 static struct pernet_operations clusterip_net_ops = {
757         .init = clusterip_net_init,
758         .exit = clusterip_net_exit,
759         .id   = &clusterip_net_id,
760         .size = sizeof(struct clusterip_net),
761 };
762 
763 static int __init clusterip_tg_init(void)
764 {
765         int ret;
766 
767         ret = register_pernet_subsys(&clusterip_net_ops);
768         if (ret < 0)
769                 return ret;
770 
771         ret = xt_register_target(&clusterip_tg_reg);
772         if (ret < 0)
773                 goto cleanup_subsys;
774 
775         ret = nf_register_hook(&cip_arp_ops);
776         if (ret < 0)
777                 goto cleanup_target;
778 
779         pr_info("ClusterIP Version %s loaded successfully\n",
780                 CLUSTERIP_VERSION);
781 
782         return 0;
783 
784 cleanup_target:
785         xt_unregister_target(&clusterip_tg_reg);
786 cleanup_subsys:
787         unregister_pernet_subsys(&clusterip_net_ops);
788         return ret;
789 }
790 
791 static void __exit clusterip_tg_exit(void)
792 {
793         pr_info("ClusterIP Version %s unloading\n", CLUSTERIP_VERSION);
794 
795         nf_unregister_hook(&cip_arp_ops);
796         xt_unregister_target(&clusterip_tg_reg);
797         unregister_pernet_subsys(&clusterip_net_ops);
798 
799         /* Wait for completion of call_rcu_bh()'s (clusterip_config_rcu_free) */
800         rcu_barrier_bh();
801 }
802 
803 module_init(clusterip_tg_init);
804 module_exit(clusterip_tg_exit);
805 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp