projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge tag 'v3.10.107' into update
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv4 / fib_frontend.c
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * IPv4 Forwarding Information Base: FIB frontend.
7 *
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
14 */
15
16 #include <linux/module.h>
17 #include <asm/uaccess.h>
18 #include <linux/bitops.h>
19 #include <linux/capability.h>
20 #include <linux/types.h>
21 #include <linux/kernel.h>
22 #include <linux/mm.h>
23 #include <linux/string.h>
24 #include <linux/socket.h>
25 #include <linux/sockios.h>
26 #include <linux/errno.h>
27 #include <linux/in.h>
28 #include <linux/inet.h>
29 #include <linux/inetdevice.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_addr.h>
32 #include <linux/if_arp.h>
33 #include <linux/skbuff.h>
34 #include <linux/cache.h>
35 #include <linux/init.h>
36 #include <linux/list.h>
37 #include <linux/slab.h>
38
39 #include <net/ip.h>
40 #include <net/protocol.h>
41 #include <net/route.h>
42 #include <net/tcp.h>
43 #include <net/sock.h>
44 #include <net/arp.h>
45 #include <net/ip_fib.h>
46 #include <net/rtnetlink.h>
47 #include <net/xfrm.h>
48
49 #ifndef CONFIG_IP_MULTIPLE_TABLES
50
51 static int __net_init fib4_rules_init(struct net *net)
52 {
53 struct fib_table *local_table, *main_table;
54
55 local_table = fib_trie_table(RT_TABLE_LOCAL);
56 if (local_table == NULL)
57 return -ENOMEM;
58
59 main_table = fib_trie_table(RT_TABLE_MAIN);
60 if (main_table == NULL)
61 goto fail;
62
63 hlist_add_head_rcu(&local_table->tb_hlist,
64 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
65 hlist_add_head_rcu(&main_table->tb_hlist,
66 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
67 return 0;
68
69 fail:
70 kfree(local_table);
71 return -ENOMEM;
72 }
73 #else
74
75 struct fib_table *fib_new_table(struct net *net, u32 id)
76 {
77 struct fib_table *tb;
78 unsigned int h;
79
80 if (id == 0)
81 id = RT_TABLE_MAIN;
82 tb = fib_get_table(net, id);
83 if (tb)
84 return tb;
85
86 tb = fib_trie_table(id);
87 if (!tb)
88 return NULL;
89
90 switch (id) {
91 case RT_TABLE_LOCAL:
92 net->ipv4.fib_local = tb;
93 break;
94
95 case RT_TABLE_MAIN:
96 net->ipv4.fib_main = tb;
97 break;
98
99 case RT_TABLE_DEFAULT:
100 net->ipv4.fib_default = tb;
101 break;
102
103 default:
104 break;
105 }
106
107 h = id & (FIB_TABLE_HASHSZ - 1);
108 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
109 return tb;
110 }
111
112 struct fib_table *fib_get_table(struct net *net, u32 id)
113 {
114 struct fib_table *tb;
115 struct hlist_head *head;
116 unsigned int h;
117
118 if (id == 0)
119 id = RT_TABLE_MAIN;
120 h = id & (FIB_TABLE_HASHSZ - 1);
121
122 rcu_read_lock();
123 head = &net->ipv4.fib_table_hash[h];
124 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
125 if (tb->tb_id == id) {
126 rcu_read_unlock();
127 return tb;
128 }
129 }
130 rcu_read_unlock();
131 return NULL;
132 }
133 #endif /* CONFIG_IP_MULTIPLE_TABLES */
134
135 static void fib_flush(struct net *net)
136 {
137 int flushed = 0;
138 struct fib_table *tb;
139 struct hlist_head *head;
140 unsigned int h;
141
142 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
143 head = &net->ipv4.fib_table_hash[h];
144 hlist_for_each_entry(tb, head, tb_hlist)
145 flushed += fib_table_flush(tb);
146 }
147
148 if (flushed)
149 rt_cache_flush(net);
150 }
151
152 /*
153 * Find address type as if only "dev" was present in the system. If
154 * on_dev is NULL then all interfaces are taken into consideration.
155 */
156 static inline unsigned int __inet_dev_addr_type(struct net *net,
157 const struct net_device *dev,
158 __be32 addr)
159 {
160 struct flowi4 fl4 = { .daddr = addr };
161 struct fib_result res;
162 unsigned int ret = RTN_BROADCAST;
163 struct fib_table *local_table;
164
165 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
166 return RTN_BROADCAST;
167 if (ipv4_is_multicast(addr))
168 return RTN_MULTICAST;
169
170 local_table = fib_get_table(net, RT_TABLE_LOCAL);
171 if (local_table) {
172 ret = RTN_UNICAST;
173 rcu_read_lock();
174 if (!fib_table_lookup(local_table, &fl4, &res, FIB_LOOKUP_NOREF)) {
175 if (!dev || dev == res.fi->fib_dev)
176 ret = res.type;
177 }
178 rcu_read_unlock();
179 }
180 return ret;
181 }
182
183 unsigned int inet_addr_type(struct net *net, __be32 addr)
184 {
185 return __inet_dev_addr_type(net, NULL, addr);
186 }
187 EXPORT_SYMBOL(inet_addr_type);
188
189 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
190 __be32 addr)
191 {
192 return __inet_dev_addr_type(net, dev, addr);
193 }
194 EXPORT_SYMBOL(inet_dev_addr_type);
195
196 __be32 fib_compute_spec_dst(struct sk_buff *skb)
197 {
198 struct net_device *dev = skb->dev;
199 struct in_device *in_dev;
200 struct fib_result res;
201 struct rtable *rt;
202 struct flowi4 fl4;
203 struct net *net;
204 int scope;
205
206 rt = skb_rtable(skb);
207 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
208 RTCF_LOCAL)
209 return ip_hdr(skb)->daddr;
210
211 in_dev = __in_dev_get_rcu(dev);
212 BUG_ON(!in_dev);
213
214 net = dev_net(dev);
215
216 scope = RT_SCOPE_UNIVERSE;
217 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
218 fl4.flowi4_oif = 0;
219 fl4.flowi4_iif = LOOPBACK_IFINDEX;
220 fl4.daddr = ip_hdr(skb)->saddr;
221 fl4.saddr = 0;
222 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
223 fl4.flowi4_scope = scope;
224 fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
225 if (!fib_lookup(net, &fl4, &res))
226 return FIB_RES_PREFSRC(net, res);
227 } else {
228 scope = RT_SCOPE_LINK;
229 }
230
231 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
232 }
233
234 /* Given (packet source, input interface) and optional (dst, oif, tos):
235 * - (main) check, that source is valid i.e. not broadcast or our local
236 * address.
237 * - figure out what "logical" interface this packet arrived
238 * and calculate "specific destination" address.
239 * - check, that packet arrived from expected physical interface.
240 * called with rcu_read_lock()
241 */
242 static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
243 u8 tos, int oif, struct net_device *dev,
244 int rpf, struct in_device *idev, u32 *itag)
245 {
246 int ret, no_addr, accept_local;
247 struct fib_result res;
248 struct flowi4 fl4;
249 struct net *net;
250 bool dev_match;
251
252 fl4.flowi4_oif = 0;
253 fl4.flowi4_iif = oif;
254 fl4.daddr = src;
255 fl4.saddr = dst;
256 fl4.flowi4_tos = tos;
257 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
258
259 no_addr = idev->ifa_list == NULL;
260
261 accept_local = IN_DEV_ACCEPT_LOCAL(idev);
262 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
263
264 net = dev_net(dev);
265 if (fib_lookup(net, &fl4, &res))
266 goto last_resort;
267 if (res.type != RTN_UNICAST) {
268 if (res.type != RTN_LOCAL || !accept_local)
269 goto e_inval;
270 }
271 fib_combine_itag(itag, &res);
272 dev_match = false;
273
274 #ifdef CONFIG_IP_ROUTE_MULTIPATH
275 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
276 struct fib_nh *nh = &res.fi->fib_nh[ret];
277
278 if (nh->nh_dev == dev) {
279 dev_match = true;
280 break;
281 }
282 }
283 #else
284 if (FIB_RES_DEV(res) == dev)
285 dev_match = true;
286 #endif
287 if (dev_match) {
288 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
289 return ret;
290 }
291 if (no_addr)
292 goto last_resort;
293 if (rpf == 1)
294 goto e_rpf;
295 fl4.flowi4_oif = dev->ifindex;
296
297 ret = 0;
298 if (fib_lookup(net, &fl4, &res) == 0) {
299 if (res.type == RTN_UNICAST)
300 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
301 }
302 return ret;
303
304 last_resort:
305 if (rpf)
306 goto e_rpf;
307 *itag = 0;
308 return 0;
309
310 e_inval:
311 return -EINVAL;
312 e_rpf:
313 return -EXDEV;
314 }
315
316 /* Ignore rp_filter for packets protected by IPsec. */
317 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
318 u8 tos, int oif, struct net_device *dev,
319 struct in_device *idev, u32 *itag)
320 {
321 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
322
323 if (!r && !fib_num_tclassid_users(dev_net(dev)) &&
324 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
325 *itag = 0;
326 return 0;
327 }
328 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
329 }
330
331 static inline __be32 sk_extract_addr(struct sockaddr *addr)
332 {
333 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
334 }
335
336 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
337 {
338 struct nlattr *nla;
339
340 nla = (struct nlattr *) ((char *) mx + len);
341 nla->nla_type = type;
342 nla->nla_len = nla_attr_size(4);
343 *(u32 *) nla_data(nla) = value;
344
345 return len + nla_total_size(4);
346 }
347
348 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
349 struct fib_config *cfg)
350 {
351 __be32 addr;
352 int plen;
353
354 memset(cfg, 0, sizeof(*cfg));
355 cfg->fc_nlinfo.nl_net = net;
356
357 if (rt->rt_dst.sa_family != AF_INET)
358 return -EAFNOSUPPORT;
359
360 /*
361 * Check mask for validity:
362 * a) it must be contiguous.
363 * b) destination must have all host bits clear.
364 * c) if application forgot to set correct family (AF_INET),
365 * reject request unless it is absolutely clear i.e.
366 * both family and mask are zero.
367 */
368 plen = 32;
369 addr = sk_extract_addr(&rt->rt_dst);
370 if (!(rt->rt_flags & RTF_HOST)) {
371 __be32 mask = sk_extract_addr(&rt->rt_genmask);
372
373 if (rt->rt_genmask.sa_family != AF_INET) {
374 if (mask || rt->rt_genmask.sa_family)
375 return -EAFNOSUPPORT;
376 }
377
378 if (bad_mask(mask, addr))
379 return -EINVAL;
380
381 plen = inet_mask_len(mask);
382 }
383
384 cfg->fc_dst_len = plen;
385 cfg->fc_dst = addr;
386
387 if (cmd != SIOCDELRT) {
388 cfg->fc_nlflags = NLM_F_CREATE;
389 cfg->fc_protocol = RTPROT_BOOT;
390 }
391
392 if (rt->rt_metric)
393 cfg->fc_priority = rt->rt_metric - 1;
394
395 if (rt->rt_flags & RTF_REJECT) {
396 cfg->fc_scope = RT_SCOPE_HOST;
397 cfg->fc_type = RTN_UNREACHABLE;
398 return 0;
399 }
400
401 cfg->fc_scope = RT_SCOPE_NOWHERE;
402 cfg->fc_type = RTN_UNICAST;
403
404 if (rt->rt_dev) {
405 char *colon;
406 struct net_device *dev;
407 char devname[IFNAMSIZ];
408
409 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
410 return -EFAULT;
411
412 devname[IFNAMSIZ-1] = 0;
413 colon = strchr(devname, ':');
414 if (colon)
415 *colon = 0;
416 #ifdef CONFIG_MTK_NET_LOGGING
417 printk(KERN_INFO "[mtk_net][RTlog info] ip_rt_ioctl rt.dev =%s \n",devname);
418 #endif
419 dev = __dev_get_by_name(net, devname);
420 if (!dev)
421 return -ENODEV;
422 cfg->fc_oif = dev->ifindex;
423 if (colon) {
424 struct in_ifaddr *ifa;
425 struct in_device *in_dev = __in_dev_get_rtnl(dev);
426 if (!in_dev)
427 return -ENODEV;
428 *colon = ':';
429 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
430 if (strcmp(ifa->ifa_label, devname) == 0)
431 break;
432 if (ifa == NULL)
433 return -ENODEV;
434 cfg->fc_prefsrc = ifa->ifa_local;
435 }
436 }
437
438 addr = sk_extract_addr(&rt->rt_gateway);
439 if (rt->rt_gateway.sa_family == AF_INET && addr) {
440 cfg->fc_gw = addr;
441 if (rt->rt_flags & RTF_GATEWAY &&
442 inet_addr_type(net, addr) == RTN_UNICAST)
443 cfg->fc_scope = RT_SCOPE_UNIVERSE;
444 }
445
446 if (cmd == SIOCDELRT)
447 return 0;
448
449 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
450 return -EINVAL;
451
452 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
453 cfg->fc_scope = RT_SCOPE_LINK;
454
455 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
456 struct nlattr *mx;
457 int len = 0;
458
459 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
460 if (mx == NULL)
461 return -ENOMEM;
462
463 if (rt->rt_flags & RTF_MTU)
464 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
465
466 if (rt->rt_flags & RTF_WINDOW)
467 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
468
469 if (rt->rt_flags & RTF_IRTT)
470 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
471
472 cfg->fc_mx = mx;
473 cfg->fc_mx_len = len;
474 }
475
476 return 0;
477 }
478
479 /*
480 * Handle IP routing ioctl calls.
481 * These are used to manipulate the routing tables
482 */
483 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
484 {
485 struct fib_config cfg;
486 struct rtentry rt;
487 int err;
488
489 switch (cmd) {
490 case SIOCADDRT: /* Add a route */
491 case SIOCDELRT: /* Delete a route */
492 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
493 return -EPERM;
494
495 if (copy_from_user(&rt, arg, sizeof(rt)))
496 return -EFAULT;
497 #ifdef CONFIG_MTK_NET_LOGGING
498 printk(KERN_INFO "[mtk_net][RTlog info]ip_rt_ioctl rt.flags =%08x, rt.rt_dst =%08x rt.rt_gateway =%08x \n",rt.rt_flags,sk_extract_addr(&(rt.rt_dst)),sk_extract_addr(&(rt.rt_gateway)));
499 #endif
500
501 rtnl_lock();
502 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
503 if (err == 0) {
504 struct fib_table *tb;
505
506 if (cmd == SIOCDELRT) {
507 #ifdef CONFIG_MTK_NET_LOGGING
508 printk(KERN_INFO "[mtk_net][RTlog delete] p_rt_ioctl: cmd == SIOCDELRT! >>\n");
509 #endif
510 tb = fib_get_table(net, cfg.fc_table);
511 if (tb)
512 err = fib_table_delete(tb, &cfg);
513 else
514 err = -ESRCH;
515 } else {
516 #ifdef CONFIG_MTK_NET_LOGGING
517 printk(KERN_INFO "[mtk_net][RTlog insert] p_rt_ioctl: cmd == SIOCADDRT! >>\n");
518 #endif
519 tb = fib_new_table(net, cfg.fc_table);
520 if (tb)
521 err = fib_table_insert(tb, &cfg);
522 else
523 err = -ENOBUFS;
524 }
525
526 /* allocated by rtentry_to_fib_config() */
527 kfree(cfg.fc_mx);
528 }
529 rtnl_unlock();
530 return err;
531 }
532 return -EINVAL;
533 }
534
535 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
536 [RTA_DST] = { .type = NLA_U32 },
537 [RTA_SRC] = { .type = NLA_U32 },
538 [RTA_IIF] = { .type = NLA_U32 },
539 [RTA_OIF] = { .type = NLA_U32 },
540 [RTA_GATEWAY] = { .type = NLA_U32 },
541 [RTA_PRIORITY] = { .type = NLA_U32 },
542 [RTA_PREFSRC] = { .type = NLA_U32 },
543 [RTA_METRICS] = { .type = NLA_NESTED },
544 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
545 [RTA_FLOW] = { .type = NLA_U32 },
546 [RTA_UID] = { .type = NLA_U32 },
547 };
548
549 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
550 struct nlmsghdr *nlh, struct fib_config *cfg)
551 {
552 struct nlattr *attr;
553 int err, remaining;
554 struct rtmsg *rtm;
555
556 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
557 if (err < 0)
558 goto errout;
559
560 memset(cfg, 0, sizeof(*cfg));
561
562 rtm = nlmsg_data(nlh);
563 cfg->fc_dst_len = rtm->rtm_dst_len;
564 cfg->fc_tos = rtm->rtm_tos;
565 cfg->fc_table = rtm->rtm_table;
566 cfg->fc_protocol = rtm->rtm_protocol;
567 cfg->fc_scope = rtm->rtm_scope;
568 cfg->fc_type = rtm->rtm_type;
569 cfg->fc_flags = rtm->rtm_flags;
570 cfg->fc_nlflags = nlh->nlmsg_flags;
571
572 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
573 cfg->fc_nlinfo.nlh = nlh;
574 cfg->fc_nlinfo.nl_net = net;
575
576 if (cfg->fc_type > RTN_MAX) {
577 err = -EINVAL;
578 goto errout;
579 }
580
581 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
582 switch (nla_type(attr)) {
583 case RTA_DST:
584 cfg->fc_dst = nla_get_be32(attr);
585 break;
586 case RTA_OIF:
587 cfg->fc_oif = nla_get_u32(attr);
588 break;
589 case RTA_GATEWAY:
590 cfg->fc_gw = nla_get_be32(attr);
591 break;
592 case RTA_PRIORITY:
593 cfg->fc_priority = nla_get_u32(attr);
594 break;
595 case RTA_PREFSRC:
596 cfg->fc_prefsrc = nla_get_be32(attr);
597 break;
598 case RTA_METRICS:
599 cfg->fc_mx = nla_data(attr);
600 cfg->fc_mx_len = nla_len(attr);
601 break;
602 case RTA_MULTIPATH:
603 cfg->fc_mp = nla_data(attr);
604 cfg->fc_mp_len = nla_len(attr);
605 break;
606 case RTA_FLOW:
607 cfg->fc_flow = nla_get_u32(attr);
608 break;
609 case RTA_TABLE:
610 cfg->fc_table = nla_get_u32(attr);
611 break;
612 }
613 }
614
615 return 0;
616 errout:
617 return err;
618 }
619
620 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
621 {
622 struct net *net = sock_net(skb->sk);
623 struct fib_config cfg;
624 struct fib_table *tb;
625 int err;
626 #ifdef CONFIG_MTK_NET_LOGGING
627 printk(KERN_INFO "[mtk_net][RTlog delete] inet_rtm_delroute !\n");
628 #endif
629 err = rtm_to_fib_config(net, skb, nlh, &cfg);
630 if (err < 0)
631 goto errout;
632 #ifdef CONFIG_MTK_NET_LOGGING
633 printk(KERN_INFO "[mtk_net][RTlog info] inet_rtm_delroute cfg.fc_dst =%08x, cfg.fc_gw =%08x\n",cfg.fc_dst,cfg.fc_gw);
634 #endif
635 tb = fib_get_table(net, cfg.fc_table);
636 if (tb == NULL) {
637 err = -ESRCH;
638 goto errout;
639 }
640
641 err = fib_table_delete(tb, &cfg);
642 errout:
643 return err;
644 }
645
646 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
647 {
648 struct net *net = sock_net(skb->sk);
649 struct fib_config cfg;
650 struct fib_table *tb;
651 int err;
652 #ifdef CONFIG_MTK_NET_LOGGING
653 printk(KERN_INFO "[mtk_net][RTlog insert] inet_rtm_newroute !\n");
654 #endif
655
656 err = rtm_to_fib_config(net, skb, nlh, &cfg);
657 if (err < 0)
658 goto errout;
659 #ifdef CONFIG_MTK_NET_LOGGING
660 printk(KERN_INFO "[mtk_net][RTlog info] inet_rtm_newroute cfg.fc_dst =%08x,cfg.fc_gw =%08x \n",cfg.fc_dst,cfg.fc_gw);
661 #endif
662
663 tb = fib_new_table(net, cfg.fc_table);
664 if (tb == NULL) {
665 err = -ENOBUFS;
666 goto errout;
667 }
668
669 err = fib_table_insert(tb, &cfg);
670 errout:
671 return err;
672 }
673
674 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
675 {
676 struct net *net = sock_net(skb->sk);
677 unsigned int h, s_h;
678 unsigned int e = 0, s_e;
679 struct fib_table *tb;
680 struct hlist_head *head;
681 int dumped = 0;
682
683 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
684 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
685 return ip_rt_dump(skb, cb);
686
687 s_h = cb->args[0];
688 s_e = cb->args[1];
689
690 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
691 e = 0;
692 head = &net->ipv4.fib_table_hash[h];
693 hlist_for_each_entry(tb, head, tb_hlist) {
694 if (e < s_e)
695 goto next;
696 if (dumped)
697 memset(&cb->args[2], 0, sizeof(cb->args) -
698 2 * sizeof(cb->args[0]));
699 if (fib_table_dump(tb, skb, cb) < 0)
700 goto out;
701 dumped = 1;
702 next:
703 e++;
704 }
705 }
706 out:
707 cb->args[1] = e;
708 cb->args[0] = h;
709
710 return skb->len;
711 }
712
713 /* Prepare and feed intra-kernel routing request.
714 * Really, it should be netlink message, but :-( netlink
715 * can be not configured, so that we feed it directly
716 * to fib engine. It is legal, because all events occur
717 * only when netlink is already locked.
718 */
719 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
720 {
721 struct net *net = dev_net(ifa->ifa_dev->dev);
722 struct fib_table *tb;
723 struct fib_config cfg = {
724 .fc_protocol = RTPROT_KERNEL,
725 .fc_type = type,
726 .fc_dst = dst,
727 .fc_dst_len = dst_len,
728 .fc_prefsrc = ifa->ifa_local,
729 .fc_oif = ifa->ifa_dev->dev->ifindex,
730 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
731 .fc_nlinfo = {
732 .nl_net = net,
733 },
734 };
735
736 if (type == RTN_UNICAST)
737 tb = fib_new_table(net, RT_TABLE_MAIN);
738 else
739 tb = fib_new_table(net, RT_TABLE_LOCAL);
740
741 if (tb == NULL)
742 return;
743
744 cfg.fc_table = tb->tb_id;
745
746 if (type != RTN_LOCAL)
747 cfg.fc_scope = RT_SCOPE_LINK;
748 else
749 cfg.fc_scope = RT_SCOPE_HOST;
750
751 if (cmd == RTM_NEWROUTE)
752 fib_table_insert(tb, &cfg);
753 else
754 fib_table_delete(tb, &cfg);
755 }
756
757 void fib_add_ifaddr(struct in_ifaddr *ifa)
758 {
759 struct in_device *in_dev = ifa->ifa_dev;
760 struct net_device *dev = in_dev->dev;
761 struct in_ifaddr *prim = ifa;
762 __be32 mask = ifa->ifa_mask;
763 __be32 addr = ifa->ifa_local;
764 __be32 prefix = ifa->ifa_address & mask;
765
766 if (ifa->ifa_flags & IFA_F_SECONDARY) {
767 prim = inet_ifa_byprefix(in_dev, prefix, mask);
768 if (prim == NULL) {
769 pr_warn("%s: bug: prim == NULL\n", __func__);
770 return;
771 }
772 }
773
774 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
775
776 if (!(dev->flags & IFF_UP))
777 return;
778
779 /* Add broadcast address, if it is explicitly assigned. */
780 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
781 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
782
783 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
784 (prefix != addr || ifa->ifa_prefixlen < 32)) {
785 /* MTK_NET_CHANGES */
786 if(0 == strncmp(dev->name, "ccmni", 2)){
787 #ifdef CONFIG_MTK_NET_LOGGING
788 printk(KERN_INFO "[mtk_net][RTlog] ignore ccmni subnet route\n");
789 #endif
790 } else {
791 fib_magic(RTM_NEWROUTE,
792 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
793 prefix, ifa->ifa_prefixlen, prim);
794 }
795 /* Add network specific broadcasts, when it takes a sense */
796 if (ifa->ifa_prefixlen < 31) {
797 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
798 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
799 32, prim);
800 }
801 }
802 }
803
804 /* Delete primary or secondary address.
805 * Optionally, on secondary address promotion consider the addresses
806 * from subnet iprim as deleted, even if they are in device list.
807 * In this case the secondary ifa can be in device list.
808 */
809 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
810 {
811 struct in_device *in_dev = ifa->ifa_dev;
812 struct net_device *dev = in_dev->dev;
813 struct in_ifaddr *ifa1;
814 struct in_ifaddr *prim = ifa, *prim1 = NULL;
815 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
816 __be32 any = ifa->ifa_address & ifa->ifa_mask;
817 #define LOCAL_OK 1
818 #define BRD_OK 2
819 #define BRD0_OK 4
820 #define BRD1_OK 8
821 unsigned int ok = 0;
822 int subnet = 0; /* Primary network */
823 int gone = 1; /* Address is missing */
824 int same_prefsrc = 0; /* Another primary with same IP */
825
826 if (ifa->ifa_flags & IFA_F_SECONDARY) {
827 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
828 if (prim == NULL) {
829 pr_warn("%s: bug: prim == NULL\n", __func__);
830 return;
831 }
832 if (iprim && iprim != prim) {
833 pr_warn("%s: bug: iprim != prim\n", __func__);
834 return;
835 }
836 } else if (!ipv4_is_zeronet(any) &&
837 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
838 fib_magic(RTM_DELROUTE,
839 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
840 any, ifa->ifa_prefixlen, prim);
841 subnet = 1;
842 }
843
844 /* Deletion is more complicated than add.
845 * We should take care of not to delete too much :-)
846 *
847 * Scan address list to be sure that addresses are really gone.
848 */
849
850 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
851 if (ifa1 == ifa) {
852 /* promotion, keep the IP */
853 gone = 0;
854 continue;
855 }
856 /* Ignore IFAs from our subnet */
857 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
858 inet_ifa_match(ifa1->ifa_address, iprim))
859 continue;
860
861 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
862 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
863 /* Another address from our subnet? */
864 if (ifa1->ifa_mask == prim->ifa_mask &&
865 inet_ifa_match(ifa1->ifa_address, prim))
866 prim1 = prim;
867 else {
868 /* We reached the secondaries, so
869 * same_prefsrc should be determined.
870 */
871 if (!same_prefsrc)
872 continue;
873 /* Search new prim1 if ifa1 is not
874 * using the current prim1
875 */
876 if (!prim1 ||
877 ifa1->ifa_mask != prim1->ifa_mask ||
878 !inet_ifa_match(ifa1->ifa_address, prim1))
879 prim1 = inet_ifa_byprefix(in_dev,
880 ifa1->ifa_address,
881 ifa1->ifa_mask);
882 if (!prim1)
883 continue;
884 if (prim1->ifa_local != prim->ifa_local)
885 continue;
886 }
887 } else {
888 if (prim->ifa_local != ifa1->ifa_local)
889 continue;
890 prim1 = ifa1;
891 if (prim != prim1)
892 same_prefsrc = 1;
893 }
894 if (ifa->ifa_local == ifa1->ifa_local)
895 ok |= LOCAL_OK;
896 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
897 ok |= BRD_OK;
898 if (brd == ifa1->ifa_broadcast)
899 ok |= BRD1_OK;
900 if (any == ifa1->ifa_broadcast)
901 ok |= BRD0_OK;
902 /* primary has network specific broadcasts */
903 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
904 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
905 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
906
907 if (!ipv4_is_zeronet(any1)) {
908 if (ifa->ifa_broadcast == brd1 ||
909 ifa->ifa_broadcast == any1)
910 ok |= BRD_OK;
911 if (brd == brd1 || brd == any1)
912 ok |= BRD1_OK;
913 if (any == brd1 || any == any1)
914 ok |= BRD0_OK;
915 }
916 }
917 }
918
919 if (!(ok & BRD_OK))
920 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
921 if (subnet && ifa->ifa_prefixlen < 31) {
922 if (!(ok & BRD1_OK))
923 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
924 if (!(ok & BRD0_OK))
925 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
926 }
927 if (!(ok & LOCAL_OK)) {
928 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
929
930 /* Check, that this local address finally disappeared. */
931 if (gone &&
932 inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
933 /* And the last, but not the least thing.
934 * We must flush stray FIB entries.
935 *
936 * First of all, we scan fib_info list searching
937 * for stray nexthop entries, then ignite fib_flush.
938 */
939 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
940 fib_flush(dev_net(dev));
941 }
942 }
943 #undef LOCAL_OK
944 #undef BRD_OK
945 #undef BRD0_OK
946 #undef BRD1_OK
947 }
948
949 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb)
950 {
951
952 struct fib_result res;
953 struct flowi4 fl4 = {
954 .flowi4_mark = frn->fl_mark,
955 .daddr = frn->fl_addr,
956 .flowi4_tos = frn->fl_tos,
957 .flowi4_scope = frn->fl_scope,
958 };
959
960 frn->err = -ENOENT;
961 if (tb) {
962 local_bh_disable();
963
964 frn->tb_id = tb->tb_id;
965 rcu_read_lock();
966 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
967
968 if (!frn->err) {
969 frn->prefixlen = res.prefixlen;
970 frn->nh_sel = res.nh_sel;
971 frn->type = res.type;
972 frn->scope = res.scope;
973 }
974 rcu_read_unlock();
975 local_bh_enable();
976 }
977 }
978
979 static void nl_fib_input(struct sk_buff *skb)
980 {
981 struct net *net;
982 struct fib_result_nl *frn;
983 struct nlmsghdr *nlh;
984 struct fib_table *tb;
985 u32 portid;
986
987 net = sock_net(skb->sk);
988 nlh = nlmsg_hdr(skb);
989 if (skb->len < nlmsg_total_size(sizeof(*frn)) ||
990 skb->len < nlh->nlmsg_len ||
991 nlmsg_len(nlh) < sizeof(*frn))
992 return;
993
994 skb = skb_clone(skb, GFP_KERNEL);
995 if (skb == NULL)
996 return;
997 nlh = nlmsg_hdr(skb);
998
999 frn = (struct fib_result_nl *) nlmsg_data(nlh);
1000 tb = fib_get_table(net, frn->tb_id_in);
1001
1002 nl_fib_lookup(frn, tb);
1003
1004 portid = NETLINK_CB(skb).portid; /* netlink portid */
1005 NETLINK_CB(skb).portid = 0; /* from kernel */
1006 NETLINK_CB(skb).dst_group = 0; /* unicast */
1007 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
1008 }
1009
1010 static int __net_init nl_fib_lookup_init(struct net *net)
1011 {
1012 struct sock *sk;
1013 struct netlink_kernel_cfg cfg = {
1014 .input = nl_fib_input,
1015 };
1016
1017 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
1018 if (sk == NULL)
1019 return -EAFNOSUPPORT;
1020 net->ipv4.fibnl = sk;
1021 return 0;
1022 }
1023
1024 static void nl_fib_lookup_exit(struct net *net)
1025 {
1026 netlink_kernel_release(net->ipv4.fibnl);
1027 net->ipv4.fibnl = NULL;
1028 }
1029
1030 static void fib_disable_ip(struct net_device *dev, int force)
1031 {
1032 if (fib_sync_down_dev(dev, force))
1033 fib_flush(dev_net(dev));
1034 rt_cache_flush(dev_net(dev));
1035 arp_ifdown(dev);
1036 }
1037
1038 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1039 {
1040 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1041 struct net_device *dev = ifa->ifa_dev->dev;
1042 struct net *net = dev_net(dev);
1043
1044 switch (event) {
1045 case NETDEV_UP:
1046 #ifdef CONFIG_MTK_NET_LOGGING
1047 printk(KERN_INFO "[mtk_net][RTlog insert] fib_inetaddr_event() %s NETDEV_UP!\n", ifa->ifa_dev->dev->name);
1048 #endif
1049 fib_add_ifaddr(ifa);
1050 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1051 fib_sync_up(dev);
1052 #endif
1053 atomic_inc(&net->ipv4.dev_addr_genid);
1054 rt_cache_flush(dev_net(dev));
1055 break;
1056 case NETDEV_DOWN:
1057 #ifdef CONFIG_MTK_NET_LOGGING
1058 printk(KERN_INFO "[mtk_net][RTlog delete] fib_inetaddr_event() %s NETDEV_DOWN!\n", ifa->ifa_dev->dev->name);
1059 #endif
1060 fib_del_ifaddr(ifa, NULL);
1061 atomic_inc(&net->ipv4.dev_addr_genid);
1062 if (ifa->ifa_dev->ifa_list == NULL) {
1063 /* Last address was deleted from this interface.
1064 * Disable IP.
1065 */
1066 fib_disable_ip(dev, 1);
1067 } else {
1068 rt_cache_flush(dev_net(dev));
1069 }
1070 break;
1071 }
1072 return NOTIFY_DONE;
1073 }
1074
1075 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1076 {
1077 struct net_device *dev = ptr;
1078 struct in_device *in_dev;
1079 struct net *net = dev_net(dev);
1080
1081 if (event == NETDEV_UNREGISTER) {
1082 fib_disable_ip(dev, 2);
1083 rt_flush_dev(dev);
1084 return NOTIFY_DONE;
1085 }
1086
1087 in_dev = __in_dev_get_rtnl(dev);
1088 if (!in_dev)
1089 return NOTIFY_DONE;
1090
1091 switch (event) {
1092 case NETDEV_UP:
1093 #ifdef CONFIG_MTK_NET_LOGGING
1094 printk(KERN_INFO "[mtk_net][RTlog insert] fib_netdev_event() %s NETDEV_UP!\n", dev->name);
1095 #endif
1096 for_ifa(in_dev) {
1097 fib_add_ifaddr(ifa);
1098 } endfor_ifa(in_dev);
1099 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1100 fib_sync_up(dev);
1101 #endif
1102 atomic_inc(&net->ipv4.dev_addr_genid);
1103 rt_cache_flush(net);
1104 break;
1105 case NETDEV_DOWN:
1106 #ifdef CONFIG_MTK_NET_LOGGING
1107 printk(KERN_INFO "[mtk_net][RTlog delete] fib_netdev_event() %s NETDEV_DOWN!\n", dev->name);
1108 #endif
1109 fib_disable_ip(dev, 0);
1110 break;
1111 case NETDEV_CHANGEMTU:
1112 case NETDEV_CHANGE:
1113 rt_cache_flush(net);
1114 break;
1115 }
1116 return NOTIFY_DONE;
1117 }
1118
1119 static struct notifier_block fib_inetaddr_notifier = {
1120 .notifier_call = fib_inetaddr_event,
1121 };
1122
1123 static struct notifier_block fib_netdev_notifier = {
1124 .notifier_call = fib_netdev_event,
1125 };
1126
1127 static int __net_init ip_fib_net_init(struct net *net)
1128 {
1129 int err;
1130 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1131
1132 /* Avoid false sharing : Use at least a full cache line */
1133 size = max_t(size_t, size, L1_CACHE_BYTES);
1134
1135 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1136 if (net->ipv4.fib_table_hash == NULL)
1137 return -ENOMEM;
1138
1139 err = fib4_rules_init(net);
1140 if (err < 0)
1141 goto fail;
1142 return 0;
1143
1144 fail:
1145 kfree(net->ipv4.fib_table_hash);
1146 return err;
1147 }
1148
1149 static void ip_fib_net_exit(struct net *net)
1150 {
1151 unsigned int i;
1152
1153 #ifdef CONFIG_IP_MULTIPLE_TABLES
1154 fib4_rules_exit(net);
1155 #endif
1156
1157 rtnl_lock();
1158 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1159 struct fib_table *tb;
1160 struct hlist_head *head;
1161 struct hlist_node *tmp;
1162
1163 head = &net->ipv4.fib_table_hash[i];
1164 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1165 hlist_del(&tb->tb_hlist);
1166 fib_table_flush(tb);
1167 fib_free_table(tb);
1168 }
1169 }
1170 rtnl_unlock();
1171 kfree(net->ipv4.fib_table_hash);
1172 }
1173
1174 static int __net_init fib_net_init(struct net *net)
1175 {
1176 int error;
1177
1178 #ifdef CONFIG_IP_ROUTE_CLASSID
1179 net->ipv4.fib_num_tclassid_users = 0;
1180 #endif
1181 error = ip_fib_net_init(net);
1182 if (error < 0)
1183 goto out;
1184 error = nl_fib_lookup_init(net);
1185 if (error < 0)
1186 goto out_nlfl;
1187 error = fib_proc_init(net);
1188 if (error < 0)
1189 goto out_proc;
1190 out:
1191 return error;
1192
1193 out_proc:
1194 nl_fib_lookup_exit(net);
1195 out_nlfl:
1196 ip_fib_net_exit(net);
1197 goto out;
1198 }
1199
1200 static void __net_exit fib_net_exit(struct net *net)
1201 {
1202 fib_proc_exit(net);
1203 nl_fib_lookup_exit(net);
1204 ip_fib_net_exit(net);
1205 }
1206
1207 static struct pernet_operations fib_net_ops = {
1208 .init = fib_net_init,
1209 .exit = fib_net_exit,
1210 };
1211
1212 void __init ip_fib_init(void)
1213 {
1214 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1215 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1216 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
1217
1218 register_pernet_subsys(&fib_net_ops);
1219 register_netdevice_notifier(&fib_netdev_notifier);
1220 register_inetaddr_notifier(&fib_inetaddr_notifier);
1221
1222 fib_trie_init();
1223 }
kernel source for telekom puls (alcatel/mediatek)