Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / net / decnet / dn_neigh.c
1 /*
2  * DECnet       An implementation of the DECnet protocol suite for the LINUX
3  *              operating system.  DECnet is implemented using the  BSD Socket
4  *              interface as the means of communication with the user level.
5  *
6  *              DECnet Neighbour Functions (Adjacency Database and
7  *                                                        On-Ethernet Cache)
8  *
9  * Author:      Steve Whitehouse <SteveW@ACM.org>
10  *
11  *
12  * Changes:
13  *     Steve Whitehouse     : Fixed router listing routine
14  *     Steve Whitehouse     : Added error_report functions
15  *     Steve Whitehouse     : Added default router detection
16  *     Steve Whitehouse     : Hop counts in outgoing messages
17  *     Steve Whitehouse     : Fixed src/dst in outgoing messages so
18  *                            forwarding now stands a good chance of
19  *                            working.
20  *     Steve Whitehouse     : Fixed neighbour states (for now anyway).
21  *     Steve Whitehouse     : Made error_report functions dummies. This
22  *                            is not the right place to return skbs.
23  *     Steve Whitehouse     : Convert to seq_file
24  *
25  */
26
27 #include <linux/net.h>
28 #include <linux/module.h>
29 #include <linux/socket.h>
30 #include <linux/if_arp.h>
31 #include <linux/slab.h>
32 #include <linux/if_ether.h>
33 #include <linux/init.h>
34 #include <linux/proc_fs.h>
35 #include <linux/string.h>
36 #include <linux/netfilter_decnet.h>
37 #include <linux/spinlock.h>
38 #include <linux/seq_file.h>
39 #include <linux/rcupdate.h>
40 #include <linux/jhash.h>
41 #include <linux/atomic.h>
42 #include <net/net_namespace.h>
43 #include <net/neighbour.h>
44 #include <net/dst.h>
45 #include <net/flow.h>
46 #include <net/dn.h>
47 #include <net/dn_dev.h>
48 #include <net/dn_neigh.h>
49 #include <net/dn_route.h>
50
51 static int dn_neigh_construct(struct neighbour *);
52 static void dn_neigh_error_report(struct neighbour *, struct sk_buff *);
53 static int dn_neigh_output(struct neighbour *neigh, struct sk_buff *skb);
54
55 /*
56  * Operations for adding the link layer header.
57  */
58 static const struct neigh_ops dn_neigh_ops = {
59         .family =               AF_DECnet,
60         .error_report =         dn_neigh_error_report,
61         .output =               dn_neigh_output,
62         .connected_output =     dn_neigh_output,
63 };
64
65 static u32 dn_neigh_hash(const void *pkey,
66                          const struct net_device *dev,
67                          __u32 *hash_rnd)
68 {
69         return jhash_2words(*(__u16 *)pkey, 0, hash_rnd[0]);
70 }
71
72 static bool dn_key_eq(const struct neighbour *neigh, const void *pkey)
73 {
74         return neigh_key_eq16(neigh, pkey);
75 }
76
77 struct neigh_table dn_neigh_table = {
78         .family =                       PF_DECnet,
79         .entry_size =                   NEIGH_ENTRY_SIZE(sizeof(struct dn_neigh)),
80         .key_len =                      sizeof(__le16),
81         .protocol =                     cpu_to_be16(ETH_P_DNA_RT),
82         .hash =                         dn_neigh_hash,
83         .key_eq =                       dn_key_eq,
84         .constructor =                  dn_neigh_construct,
85         .id =                           "dn_neigh_cache",
86         .parms ={
87                 .tbl =                  &dn_neigh_table,
88                 .reachable_time =       30 * HZ,
89                 .data = {
90                         [NEIGH_VAR_MCAST_PROBES] = 0,
91                         [NEIGH_VAR_UCAST_PROBES] = 0,
92                         [NEIGH_VAR_APP_PROBES] = 0,
93                         [NEIGH_VAR_RETRANS_TIME] = 1 * HZ,
94                         [NEIGH_VAR_BASE_REACHABLE_TIME] = 30 * HZ,
95                         [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
96                         [NEIGH_VAR_GC_STALETIME] = 60 * HZ,
97                         [NEIGH_VAR_QUEUE_LEN_BYTES] = 64*1024,
98                         [NEIGH_VAR_PROXY_QLEN] = 0,
99                         [NEIGH_VAR_ANYCAST_DELAY] = 0,
100                         [NEIGH_VAR_PROXY_DELAY] = 0,
101                         [NEIGH_VAR_LOCKTIME] = 1 * HZ,
102                 },
103         },
104         .gc_interval =                  30 * HZ,
105         .gc_thresh1 =                   128,
106         .gc_thresh2 =                   512,
107         .gc_thresh3 =                   1024,
108 };
109
110 static int dn_neigh_construct(struct neighbour *neigh)
111 {
112         struct net_device *dev = neigh->dev;
113         struct dn_neigh *dn = (struct dn_neigh *)neigh;
114         struct dn_dev *dn_db;
115         struct neigh_parms *parms;
116
117         rcu_read_lock();
118         dn_db = rcu_dereference(dev->dn_ptr);
119         if (dn_db == NULL) {
120                 rcu_read_unlock();
121                 return -EINVAL;
122         }
123
124         parms = dn_db->neigh_parms;
125         if (!parms) {
126                 rcu_read_unlock();
127                 return -EINVAL;
128         }
129
130         __neigh_parms_put(neigh->parms);
131         neigh->parms = neigh_parms_clone(parms);
132         rcu_read_unlock();
133
134         neigh->ops = &dn_neigh_ops;
135         neigh->nud_state = NUD_NOARP;
136         neigh->output = neigh->ops->connected_output;
137
138         if ((dev->type == ARPHRD_IPGRE) || (dev->flags & IFF_POINTOPOINT))
139                 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
140         else if ((dev->type == ARPHRD_ETHER) || (dev->type == ARPHRD_LOOPBACK))
141                 dn_dn2eth(neigh->ha, dn->addr);
142         else {
143                 net_dbg_ratelimited("Trying to create neigh for hw %d\n",
144                                     dev->type);
145                 return -EINVAL;
146         }
147
148         /*
149          * Make an estimate of the remote block size by assuming that its
150          * two less then the device mtu, which it true for ethernet (and
151          * other things which support long format headers) since there is
152          * an extra length field (of 16 bits) which isn't part of the
153          * ethernet headers and which the DECnet specs won't admit is part
154          * of the DECnet routing headers either.
155          *
156          * If we over estimate here its no big deal, the NSP negotiations
157          * will prevent us from sending packets which are too large for the
158          * remote node to handle. In any case this figure is normally updated
159          * by a hello message in most cases.
160          */
161         dn->blksize = dev->mtu - 2;
162
163         return 0;
164 }
165
166 static void dn_neigh_error_report(struct neighbour *neigh, struct sk_buff *skb)
167 {
168         printk(KERN_DEBUG "dn_neigh_error_report: called\n");
169         kfree_skb(skb);
170 }
171
172 static int dn_neigh_output(struct neighbour *neigh, struct sk_buff *skb)
173 {
174         struct dst_entry *dst = skb_dst(skb);
175         struct dn_route *rt = (struct dn_route *)dst;
176         struct net_device *dev = neigh->dev;
177         char mac_addr[ETH_ALEN];
178         unsigned int seq;
179         int err;
180
181         dn_dn2eth(mac_addr, rt->rt_local_src);
182         do {
183                 seq = read_seqbegin(&neigh->ha_lock);
184                 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
185                                       neigh->ha, mac_addr, skb->len);
186         } while (read_seqretry(&neigh->ha_lock, seq));
187
188         if (err >= 0)
189                 err = dev_queue_xmit(skb);
190         else {
191                 kfree_skb(skb);
192                 err = -EINVAL;
193         }
194         return err;
195 }
196
197 static int dn_neigh_output_packet(struct sock *sk, struct sk_buff *skb)
198 {
199         struct dst_entry *dst = skb_dst(skb);
200         struct dn_route *rt = (struct dn_route *)dst;
201         struct neighbour *neigh = rt->n;
202
203         return neigh->output(neigh, skb);
204 }
205
206 /*
207  * For talking to broadcast devices: Ethernet & PPP
208  */
209 static int dn_long_output(struct neighbour *neigh, struct sock *sk,
210                           struct sk_buff *skb)
211 {
212         struct net_device *dev = neigh->dev;
213         int headroom = dev->hard_header_len + sizeof(struct dn_long_packet) + 3;
214         unsigned char *data;
215         struct dn_long_packet *lp;
216         struct dn_skb_cb *cb = DN_SKB_CB(skb);
217
218
219         if (skb_headroom(skb) < headroom) {
220                 struct sk_buff *skb2 = skb_realloc_headroom(skb, headroom);
221                 if (skb2 == NULL) {
222                         net_crit_ratelimited("dn_long_output: no memory\n");
223                         kfree_skb(skb);
224                         return -ENOBUFS;
225                 }
226                 consume_skb(skb);
227                 skb = skb2;
228                 net_info_ratelimited("dn_long_output: Increasing headroom\n");
229         }
230
231         data = skb_push(skb, sizeof(struct dn_long_packet) + 3);
232         lp = (struct dn_long_packet *)(data+3);
233
234         *((__le16 *)data) = cpu_to_le16(skb->len - 2);
235         *(data + 2) = 1 | DN_RT_F_PF; /* Padding */
236
237         lp->msgflg   = DN_RT_PKT_LONG|(cb->rt_flags&(DN_RT_F_IE|DN_RT_F_RQR|DN_RT_F_RTS));
238         lp->d_area   = lp->d_subarea = 0;
239         dn_dn2eth(lp->d_id, cb->dst);
240         lp->s_area   = lp->s_subarea = 0;
241         dn_dn2eth(lp->s_id, cb->src);
242         lp->nl2      = 0;
243         lp->visit_ct = cb->hops & 0x3f;
244         lp->s_class  = 0;
245         lp->pt       = 0;
246
247         skb_reset_network_header(skb);
248
249         return NF_HOOK(NFPROTO_DECNET, NF_DN_POST_ROUTING, sk, skb,
250                        NULL, neigh->dev, dn_neigh_output_packet);
251 }
252
253 /*
254  * For talking to pointopoint and multidrop devices: DDCMP and X.25
255  */
256 static int dn_short_output(struct neighbour *neigh, struct sock *sk,
257                            struct sk_buff *skb)
258 {
259         struct net_device *dev = neigh->dev;
260         int headroom = dev->hard_header_len + sizeof(struct dn_short_packet) + 2;
261         struct dn_short_packet *sp;
262         unsigned char *data;
263         struct dn_skb_cb *cb = DN_SKB_CB(skb);
264
265
266         if (skb_headroom(skb) < headroom) {
267                 struct sk_buff *skb2 = skb_realloc_headroom(skb, headroom);
268                 if (skb2 == NULL) {
269                         net_crit_ratelimited("dn_short_output: no memory\n");
270                         kfree_skb(skb);
271                         return -ENOBUFS;
272                 }
273                 consume_skb(skb);
274                 skb = skb2;
275                 net_info_ratelimited("dn_short_output: Increasing headroom\n");
276         }
277
278         data = skb_push(skb, sizeof(struct dn_short_packet) + 2);
279         *((__le16 *)data) = cpu_to_le16(skb->len - 2);
280         sp = (struct dn_short_packet *)(data+2);
281
282         sp->msgflg     = DN_RT_PKT_SHORT|(cb->rt_flags&(DN_RT_F_RQR|DN_RT_F_RTS));
283         sp->dstnode    = cb->dst;
284         sp->srcnode    = cb->src;
285         sp->forward    = cb->hops & 0x3f;
286
287         skb_reset_network_header(skb);
288
289         return NF_HOOK(NFPROTO_DECNET, NF_DN_POST_ROUTING, sk, skb,
290                        NULL, neigh->dev, dn_neigh_output_packet);
291 }
292
293 /*
294  * For talking to DECnet phase III nodes
295  * Phase 3 output is the same as short output, execpt that
296  * it clears the area bits before transmission.
297  */
298 static int dn_phase3_output(struct neighbour *neigh, struct sock *sk,
299                             struct sk_buff *skb)
300 {
301         struct net_device *dev = neigh->dev;
302         int headroom = dev->hard_header_len + sizeof(struct dn_short_packet) + 2;
303         struct dn_short_packet *sp;
304         unsigned char *data;
305         struct dn_skb_cb *cb = DN_SKB_CB(skb);
306
307         if (skb_headroom(skb) < headroom) {
308                 struct sk_buff *skb2 = skb_realloc_headroom(skb, headroom);
309                 if (skb2 == NULL) {
310                         net_crit_ratelimited("dn_phase3_output: no memory\n");
311                         kfree_skb(skb);
312                         return -ENOBUFS;
313                 }
314                 consume_skb(skb);
315                 skb = skb2;
316                 net_info_ratelimited("dn_phase3_output: Increasing headroom\n");
317         }
318
319         data = skb_push(skb, sizeof(struct dn_short_packet) + 2);
320         *((__le16 *)data) = cpu_to_le16(skb->len - 2);
321         sp = (struct dn_short_packet *)(data + 2);
322
323         sp->msgflg   = DN_RT_PKT_SHORT|(cb->rt_flags&(DN_RT_F_RQR|DN_RT_F_RTS));
324         sp->dstnode  = cb->dst & cpu_to_le16(0x03ff);
325         sp->srcnode  = cb->src & cpu_to_le16(0x03ff);
326         sp->forward  = cb->hops & 0x3f;
327
328         skb_reset_network_header(skb);
329
330         return NF_HOOK(NFPROTO_DECNET, NF_DN_POST_ROUTING, sk, skb,
331                        NULL, neigh->dev, dn_neigh_output_packet);
332 }
333
334 int dn_to_neigh_output(struct sock *sk, struct sk_buff *skb)
335 {
336         struct dst_entry *dst = skb_dst(skb);
337         struct dn_route *rt = (struct dn_route *) dst;
338         struct neighbour *neigh = rt->n;
339         struct dn_neigh *dn = (struct dn_neigh *)neigh;
340         struct dn_dev *dn_db;
341         bool use_long;
342
343         rcu_read_lock();
344         dn_db = rcu_dereference(neigh->dev->dn_ptr);
345         if (dn_db == NULL) {
346                 rcu_read_unlock();
347                 return -EINVAL;
348         }
349         use_long = dn_db->use_long;
350         rcu_read_unlock();
351
352         if (dn->flags & DN_NDFLAG_P3)
353                 return dn_phase3_output(neigh, sk, skb);
354         if (use_long)
355                 return dn_long_output(neigh, sk, skb);
356         else
357                 return dn_short_output(neigh, sk, skb);
358 }
359
360 /*
361  * Unfortunately, the neighbour code uses the device in its hash
362  * function, so we don't get any advantage from it. This function
363  * basically does a neigh_lookup(), but without comparing the device
364  * field. This is required for the On-Ethernet cache
365  */
366
367 /*
368  * Pointopoint link receives a hello message
369  */
370 void dn_neigh_pointopoint_hello(struct sk_buff *skb)
371 {
372         kfree_skb(skb);
373 }
374
375 /*
376  * Ethernet router hello message received
377  */
378 int dn_neigh_router_hello(struct sock *sk, struct sk_buff *skb)
379 {
380         struct rtnode_hello_message *msg = (struct rtnode_hello_message *)skb->data;
381
382         struct neighbour *neigh;
383         struct dn_neigh *dn;
384         struct dn_dev *dn_db;
385         __le16 src;
386
387         src = dn_eth2dn(msg->id);
388
389         neigh = __neigh_lookup(&dn_neigh_table, &src, skb->dev, 1);
390
391         dn = (struct dn_neigh *)neigh;
392
393         if (neigh) {
394                 write_lock(&neigh->lock);
395
396                 neigh->used = jiffies;
397                 dn_db = rcu_dereference(neigh->dev->dn_ptr);
398
399                 if (!(neigh->nud_state & NUD_PERMANENT)) {
400                         neigh->updated = jiffies;
401
402                         if (neigh->dev->type == ARPHRD_ETHER)
403                                 memcpy(neigh->ha, &eth_hdr(skb)->h_source, ETH_ALEN);
404
405                         dn->blksize  = le16_to_cpu(msg->blksize);
406                         dn->priority = msg->priority;
407
408                         dn->flags &= ~DN_NDFLAG_P3;
409
410                         switch (msg->iinfo & DN_RT_INFO_TYPE) {
411                         case DN_RT_INFO_L1RT:
412                                 dn->flags &=~DN_NDFLAG_R2;
413                                 dn->flags |= DN_NDFLAG_R1;
414                                 break;
415                         case DN_RT_INFO_L2RT:
416                                 dn->flags |= DN_NDFLAG_R2;
417                         }
418                 }
419
420                 /* Only use routers in our area */
421                 if ((le16_to_cpu(src)>>10) == (le16_to_cpu((decnet_address))>>10)) {
422                         if (!dn_db->router) {
423                                 dn_db->router = neigh_clone(neigh);
424                         } else {
425                                 if (msg->priority > ((struct dn_neigh *)dn_db->router)->priority)
426                                         neigh_release(xchg(&dn_db->router, neigh_clone(neigh)));
427                         }
428                 }
429                 write_unlock(&neigh->lock);
430                 neigh_release(neigh);
431         }
432
433         kfree_skb(skb);
434         return 0;
435 }
436
437 /*
438  * Endnode hello message received
439  */
440 int dn_neigh_endnode_hello(struct sock *sk, struct sk_buff *skb)
441 {
442         struct endnode_hello_message *msg = (struct endnode_hello_message *)skb->data;
443         struct neighbour *neigh;
444         struct dn_neigh *dn;
445         __le16 src;
446
447         src = dn_eth2dn(msg->id);
448
449         neigh = __neigh_lookup(&dn_neigh_table, &src, skb->dev, 1);
450
451         dn = (struct dn_neigh *)neigh;
452
453         if (neigh) {
454                 write_lock(&neigh->lock);
455
456                 neigh->used = jiffies;
457
458                 if (!(neigh->nud_state & NUD_PERMANENT)) {
459                         neigh->updated = jiffies;
460
461                         if (neigh->dev->type == ARPHRD_ETHER)
462                                 memcpy(neigh->ha, &eth_hdr(skb)->h_source, ETH_ALEN);
463                         dn->flags   &= ~(DN_NDFLAG_R1 | DN_NDFLAG_R2);
464                         dn->blksize  = le16_to_cpu(msg->blksize);
465                         dn->priority = 0;
466                 }
467
468                 write_unlock(&neigh->lock);
469                 neigh_release(neigh);
470         }
471
472         kfree_skb(skb);
473         return 0;
474 }
475
476 static char *dn_find_slot(char *base, int max, int priority)
477 {
478         int i;
479         unsigned char *min = NULL;
480
481         base += 6; /* skip first id */
482
483         for(i = 0; i < max; i++) {
484                 if (!min || (*base < *min))
485                         min = base;
486                 base += 7; /* find next priority */
487         }
488
489         if (!min)
490                 return NULL;
491
492         return (*min < priority) ? (min - 6) : NULL;
493 }
494
495 struct elist_cb_state {
496         struct net_device *dev;
497         unsigned char *ptr;
498         unsigned char *rs;
499         int t, n;
500 };
501
502 static void neigh_elist_cb(struct neighbour *neigh, void *_info)
503 {
504         struct elist_cb_state *s = _info;
505         struct dn_neigh *dn;
506
507         if (neigh->dev != s->dev)
508                 return;
509
510         dn = (struct dn_neigh *) neigh;
511         if (!(dn->flags & (DN_NDFLAG_R1|DN_NDFLAG_R2)))
512                 return;
513
514         if (s->t == s->n)
515                 s->rs = dn_find_slot(s->ptr, s->n, dn->priority);
516         else
517                 s->t++;
518         if (s->rs == NULL)
519                 return;
520
521         dn_dn2eth(s->rs, dn->addr);
522         s->rs += 6;
523         *(s->rs) = neigh->nud_state & NUD_CONNECTED ? 0x80 : 0x0;
524         *(s->rs) |= dn->priority;
525         s->rs++;
526 }
527
528 int dn_neigh_elist(struct net_device *dev, unsigned char *ptr, int n)
529 {
530         struct elist_cb_state state;
531
532         state.dev = dev;
533         state.t = 0;
534         state.n = n;
535         state.ptr = ptr;
536         state.rs = ptr;
537
538         neigh_for_each(&dn_neigh_table, neigh_elist_cb, &state);
539
540         return state.t;
541 }
542
543
544 #ifdef CONFIG_PROC_FS
545
546 static inline void dn_neigh_format_entry(struct seq_file *seq,
547                                          struct neighbour *n)
548 {
549         struct dn_neigh *dn = (struct dn_neigh *) n;
550         char buf[DN_ASCBUF_LEN];
551
552         read_lock(&n->lock);
553         seq_printf(seq, "%-7s %s%s%s   %02x    %02d  %07ld %-8s\n",
554                    dn_addr2asc(le16_to_cpu(dn->addr), buf),
555                    (dn->flags&DN_NDFLAG_R1) ? "1" : "-",
556                    (dn->flags&DN_NDFLAG_R2) ? "2" : "-",
557                    (dn->flags&DN_NDFLAG_P3) ? "3" : "-",
558                    dn->n.nud_state,
559                    atomic_read(&dn->n.refcnt),
560                    dn->blksize,
561                    (dn->n.dev) ? dn->n.dev->name : "?");
562         read_unlock(&n->lock);
563 }
564
565 static int dn_neigh_seq_show(struct seq_file *seq, void *v)
566 {
567         if (v == SEQ_START_TOKEN) {
568                 seq_puts(seq, "Addr    Flags State Use Blksize Dev\n");
569         } else {
570                 dn_neigh_format_entry(seq, v);
571         }
572
573         return 0;
574 }
575
576 static void *dn_neigh_seq_start(struct seq_file *seq, loff_t *pos)
577 {
578         return neigh_seq_start(seq, pos, &dn_neigh_table,
579                                NEIGH_SEQ_NEIGH_ONLY);
580 }
581
582 static const struct seq_operations dn_neigh_seq_ops = {
583         .start = dn_neigh_seq_start,
584         .next  = neigh_seq_next,
585         .stop  = neigh_seq_stop,
586         .show  = dn_neigh_seq_show,
587 };
588
589 static int dn_neigh_seq_open(struct inode *inode, struct file *file)
590 {
591         return seq_open_net(inode, file, &dn_neigh_seq_ops,
592                             sizeof(struct neigh_seq_state));
593 }
594
595 static const struct file_operations dn_neigh_seq_fops = {
596         .owner          = THIS_MODULE,
597         .open           = dn_neigh_seq_open,
598         .read           = seq_read,
599         .llseek         = seq_lseek,
600         .release        = seq_release_net,
601 };
602
603 #endif
604
605 void __init dn_neigh_init(void)
606 {
607         neigh_table_init(NEIGH_DN_TABLE, &dn_neigh_table);
608         proc_create("decnet_neigh", S_IRUGO, init_net.proc_net,
609                     &dn_neigh_seq_fops);
610 }
611
612 void __exit dn_neigh_cleanup(void)
613 {
614         remove_proc_entry("decnet_neigh", init_net.proc_net);
615         neigh_table_clear(NEIGH_DN_TABLE, &dn_neigh_table);
616 }