inotify: fix race
[linux-2.6.22.y-op.git] / net / decnet / dn_neigh.c
blob4bf066c416e2d8b0ba1e6160b377cb687081b0b7
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.
6 * DECnet Neighbour Functions (Adjacency Database and
7 * On-Ethernet Cache)
9 * Author: Steve Whitehouse <SteveW@ACM.org>
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
27 #include <linux/net.h>
28 #include <linux/module.h>
29 #include <linux/socket.h>
30 #include <linux/if_arp.h>
31 #include <linux/if_ether.h>
32 #include <linux/init.h>
33 #include <linux/proc_fs.h>
34 #include <linux/string.h>
35 #include <linux/netfilter_decnet.h>
36 #include <linux/spinlock.h>
37 #include <linux/seq_file.h>
38 #include <linux/rcupdate.h>
39 #include <linux/jhash.h>
40 #include <asm/atomic.h>
41 #include <net/neighbour.h>
42 #include <net/dst.h>
43 #include <net/flow.h>
44 #include <net/dn.h>
45 #include <net/dn_dev.h>
46 #include <net/dn_neigh.h>
47 #include <net/dn_route.h>
49 static u32 dn_neigh_hash(const void *pkey, const struct net_device *dev);
50 static int dn_neigh_construct(struct neighbour *);
51 static void dn_long_error_report(struct neighbour *, struct sk_buff *);
52 static void dn_short_error_report(struct neighbour *, struct sk_buff *);
53 static int dn_long_output(struct sk_buff *);
54 static int dn_short_output(struct sk_buff *);
55 static int dn_phase3_output(struct sk_buff *);
59 * For talking to broadcast devices: Ethernet & PPP
61 static struct neigh_ops dn_long_ops = {
62 .family = AF_DECnet,
63 .error_report = dn_long_error_report,
64 .output = dn_long_output,
65 .connected_output = dn_long_output,
66 .hh_output = dev_queue_xmit,
67 .queue_xmit = dev_queue_xmit,
71 * For talking to pointopoint and multidrop devices: DDCMP and X.25
73 static struct neigh_ops dn_short_ops = {
74 .family = AF_DECnet,
75 .error_report = dn_short_error_report,
76 .output = dn_short_output,
77 .connected_output = dn_short_output,
78 .hh_output = dev_queue_xmit,
79 .queue_xmit = dev_queue_xmit,
83 * For talking to DECnet phase III nodes
85 static struct neigh_ops dn_phase3_ops = {
86 .family = AF_DECnet,
87 .error_report = dn_short_error_report, /* Can use short version here */
88 .output = dn_phase3_output,
89 .connected_output = dn_phase3_output,
90 .hh_output = dev_queue_xmit,
91 .queue_xmit = dev_queue_xmit
94 struct neigh_table dn_neigh_table = {
95 .family = PF_DECnet,
96 .entry_size = sizeof(struct dn_neigh),
97 .key_len = sizeof(__le16),
98 .hash = dn_neigh_hash,
99 .constructor = dn_neigh_construct,
100 .id = "dn_neigh_cache",
101 .parms ={
102 .tbl = &dn_neigh_table,
103 .base_reachable_time = 30 * HZ,
104 .retrans_time = 1 * HZ,
105 .gc_staletime = 60 * HZ,
106 .reachable_time = 30 * HZ,
107 .delay_probe_time = 5 * HZ,
108 .queue_len = 3,
109 .ucast_probes = 0,
110 .app_probes = 0,
111 .mcast_probes = 0,
112 .anycast_delay = 0,
113 .proxy_delay = 0,
114 .proxy_qlen = 0,
115 .locktime = 1 * HZ,
117 .gc_interval = 30 * HZ,
118 .gc_thresh1 = 128,
119 .gc_thresh2 = 512,
120 .gc_thresh3 = 1024,
123 static u32 dn_neigh_hash(const void *pkey, const struct net_device *dev)
125 return jhash_2words(*(__u16 *)pkey, 0, dn_neigh_table.hash_rnd);
128 static int dn_neigh_construct(struct neighbour *neigh)
130 struct net_device *dev = neigh->dev;
131 struct dn_neigh *dn = (struct dn_neigh *)neigh;
132 struct dn_dev *dn_db;
133 struct neigh_parms *parms;
135 rcu_read_lock();
136 dn_db = rcu_dereference(dev->dn_ptr);
137 if (dn_db == NULL) {
138 rcu_read_unlock();
139 return -EINVAL;
142 parms = dn_db->neigh_parms;
143 if (!parms) {
144 rcu_read_unlock();
145 return -EINVAL;
148 __neigh_parms_put(neigh->parms);
149 neigh->parms = neigh_parms_clone(parms);
151 if (dn_db->use_long)
152 neigh->ops = &dn_long_ops;
153 else
154 neigh->ops = &dn_short_ops;
155 rcu_read_unlock();
157 if (dn->flags & DN_NDFLAG_P3)
158 neigh->ops = &dn_phase3_ops;
160 neigh->nud_state = NUD_NOARP;
161 neigh->output = neigh->ops->connected_output;
163 if ((dev->type == ARPHRD_IPGRE) || (dev->flags & IFF_POINTOPOINT))
164 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
165 else if ((dev->type == ARPHRD_ETHER) || (dev->type == ARPHRD_LOOPBACK))
166 dn_dn2eth(neigh->ha, dn->addr);
167 else {
168 if (net_ratelimit())
169 printk(KERN_DEBUG "Trying to create neigh for hw %d\n", dev->type);
170 return -EINVAL;
174 * Make an estimate of the remote block size by assuming that its
175 * two less then the device mtu, which it true for ethernet (and
176 * other things which support long format headers) since there is
177 * an extra length field (of 16 bits) which isn't part of the
178 * ethernet headers and which the DECnet specs won't admit is part
179 * of the DECnet routing headers either.
181 * If we over estimate here its no big deal, the NSP negotiations
182 * will prevent us from sending packets which are too large for the
183 * remote node to handle. In any case this figure is normally updated
184 * by a hello message in most cases.
186 dn->blksize = dev->mtu - 2;
188 return 0;
191 static void dn_long_error_report(struct neighbour *neigh, struct sk_buff *skb)
193 printk(KERN_DEBUG "dn_long_error_report: called\n");
194 kfree_skb(skb);
198 static void dn_short_error_report(struct neighbour *neigh, struct sk_buff *skb)
200 printk(KERN_DEBUG "dn_short_error_report: called\n");
201 kfree_skb(skb);
204 static int dn_neigh_output_packet(struct sk_buff *skb)
206 struct dst_entry *dst = skb->dst;
207 struct dn_route *rt = (struct dn_route *)dst;
208 struct neighbour *neigh = dst->neighbour;
209 struct net_device *dev = neigh->dev;
210 char mac_addr[ETH_ALEN];
212 dn_dn2eth(mac_addr, rt->rt_local_src);
213 if (!dev->hard_header || dev->hard_header(skb, dev, ntohs(skb->protocol), neigh->ha, mac_addr, skb->len) >= 0)
214 return neigh->ops->queue_xmit(skb);
216 if (net_ratelimit())
217 printk(KERN_DEBUG "dn_neigh_output_packet: oops, can't send packet\n");
219 kfree_skb(skb);
220 return -EINVAL;
223 static int dn_long_output(struct sk_buff *skb)
225 struct dst_entry *dst = skb->dst;
226 struct neighbour *neigh = dst->neighbour;
227 struct net_device *dev = neigh->dev;
228 int headroom = dev->hard_header_len + sizeof(struct dn_long_packet) + 3;
229 unsigned char *data;
230 struct dn_long_packet *lp;
231 struct dn_skb_cb *cb = DN_SKB_CB(skb);
234 if (skb_headroom(skb) < headroom) {
235 struct sk_buff *skb2 = skb_realloc_headroom(skb, headroom);
236 if (skb2 == NULL) {
237 if (net_ratelimit())
238 printk(KERN_CRIT "dn_long_output: no memory\n");
239 kfree_skb(skb);
240 return -ENOBUFS;
242 kfree_skb(skb);
243 skb = skb2;
244 if (net_ratelimit())
245 printk(KERN_INFO "dn_long_output: Increasing headroom\n");
248 data = skb_push(skb, sizeof(struct dn_long_packet) + 3);
249 lp = (struct dn_long_packet *)(data+3);
251 *((__le16 *)data) = dn_htons(skb->len - 2);
252 *(data + 2) = 1 | DN_RT_F_PF; /* Padding */
254 lp->msgflg = DN_RT_PKT_LONG|(cb->rt_flags&(DN_RT_F_IE|DN_RT_F_RQR|DN_RT_F_RTS));
255 lp->d_area = lp->d_subarea = 0;
256 dn_dn2eth(lp->d_id, cb->dst);
257 lp->s_area = lp->s_subarea = 0;
258 dn_dn2eth(lp->s_id, cb->src);
259 lp->nl2 = 0;
260 lp->visit_ct = cb->hops & 0x3f;
261 lp->s_class = 0;
262 lp->pt = 0;
264 skb_reset_network_header(skb);
266 return NF_HOOK(PF_DECnet, NF_DN_POST_ROUTING, skb, NULL, neigh->dev, dn_neigh_output_packet);
269 static int dn_short_output(struct sk_buff *skb)
271 struct dst_entry *dst = skb->dst;
272 struct neighbour *neigh = dst->neighbour;
273 struct net_device *dev = neigh->dev;
274 int headroom = dev->hard_header_len + sizeof(struct dn_short_packet) + 2;
275 struct dn_short_packet *sp;
276 unsigned char *data;
277 struct dn_skb_cb *cb = DN_SKB_CB(skb);
280 if (skb_headroom(skb) < headroom) {
281 struct sk_buff *skb2 = skb_realloc_headroom(skb, headroom);
282 if (skb2 == NULL) {
283 if (net_ratelimit())
284 printk(KERN_CRIT "dn_short_output: no memory\n");
285 kfree_skb(skb);
286 return -ENOBUFS;
288 kfree_skb(skb);
289 skb = skb2;
290 if (net_ratelimit())
291 printk(KERN_INFO "dn_short_output: Increasing headroom\n");
294 data = skb_push(skb, sizeof(struct dn_short_packet) + 2);
295 *((__le16 *)data) = dn_htons(skb->len - 2);
296 sp = (struct dn_short_packet *)(data+2);
298 sp->msgflg = DN_RT_PKT_SHORT|(cb->rt_flags&(DN_RT_F_RQR|DN_RT_F_RTS));
299 sp->dstnode = cb->dst;
300 sp->srcnode = cb->src;
301 sp->forward = cb->hops & 0x3f;
303 skb_reset_network_header(skb);
305 return NF_HOOK(PF_DECnet, NF_DN_POST_ROUTING, skb, NULL, neigh->dev, dn_neigh_output_packet);
309 * Phase 3 output is the same is short output, execpt that
310 * it clears the area bits before transmission.
312 static int dn_phase3_output(struct sk_buff *skb)
314 struct dst_entry *dst = skb->dst;
315 struct neighbour *neigh = dst->neighbour;
316 struct net_device *dev = neigh->dev;
317 int headroom = dev->hard_header_len + sizeof(struct dn_short_packet) + 2;
318 struct dn_short_packet *sp;
319 unsigned char *data;
320 struct dn_skb_cb *cb = DN_SKB_CB(skb);
322 if (skb_headroom(skb) < headroom) {
323 struct sk_buff *skb2 = skb_realloc_headroom(skb, headroom);
324 if (skb2 == NULL) {
325 if (net_ratelimit())
326 printk(KERN_CRIT "dn_phase3_output: no memory\n");
327 kfree_skb(skb);
328 return -ENOBUFS;
330 kfree_skb(skb);
331 skb = skb2;
332 if (net_ratelimit())
333 printk(KERN_INFO "dn_phase3_output: Increasing headroom\n");
336 data = skb_push(skb, sizeof(struct dn_short_packet) + 2);
337 *((__le16 *)data) = dn_htons(skb->len - 2);
338 sp = (struct dn_short_packet *)(data + 2);
340 sp->msgflg = DN_RT_PKT_SHORT|(cb->rt_flags&(DN_RT_F_RQR|DN_RT_F_RTS));
341 sp->dstnode = cb->dst & dn_htons(0x03ff);
342 sp->srcnode = cb->src & dn_htons(0x03ff);
343 sp->forward = cb->hops & 0x3f;
345 skb_reset_network_header(skb);
347 return NF_HOOK(PF_DECnet, NF_DN_POST_ROUTING, skb, NULL, neigh->dev, dn_neigh_output_packet);
351 * Unfortunately, the neighbour code uses the device in its hash
352 * function, so we don't get any advantage from it. This function
353 * basically does a neigh_lookup(), but without comparing the device
354 * field. This is required for the On-Ethernet cache
358 * Pointopoint link receives a hello message
360 void dn_neigh_pointopoint_hello(struct sk_buff *skb)
362 kfree_skb(skb);
366 * Ethernet router hello message received
368 int dn_neigh_router_hello(struct sk_buff *skb)
370 struct rtnode_hello_message *msg = (struct rtnode_hello_message *)skb->data;
372 struct neighbour *neigh;
373 struct dn_neigh *dn;
374 struct dn_dev *dn_db;
375 __le16 src;
377 src = dn_eth2dn(msg->id);
379 neigh = __neigh_lookup(&dn_neigh_table, &src, skb->dev, 1);
381 dn = (struct dn_neigh *)neigh;
383 if (neigh) {
384 write_lock(&neigh->lock);
386 neigh->used = jiffies;
387 dn_db = (struct dn_dev *)neigh->dev->dn_ptr;
389 if (!(neigh->nud_state & NUD_PERMANENT)) {
390 neigh->updated = jiffies;
392 if (neigh->dev->type == ARPHRD_ETHER)
393 memcpy(neigh->ha, &eth_hdr(skb)->h_source, ETH_ALEN);
395 dn->blksize = dn_ntohs(msg->blksize);
396 dn->priority = msg->priority;
398 dn->flags &= ~DN_NDFLAG_P3;
400 switch(msg->iinfo & DN_RT_INFO_TYPE) {
401 case DN_RT_INFO_L1RT:
402 dn->flags &=~DN_NDFLAG_R2;
403 dn->flags |= DN_NDFLAG_R1;
404 break;
405 case DN_RT_INFO_L2RT:
406 dn->flags |= DN_NDFLAG_R2;
410 /* Only use routers in our area */
411 if ((dn_ntohs(src)>>10) == (dn_ntohs((decnet_address))>>10)) {
412 if (!dn_db->router) {
413 dn_db->router = neigh_clone(neigh);
414 } else {
415 if (msg->priority > ((struct dn_neigh *)dn_db->router)->priority)
416 neigh_release(xchg(&dn_db->router, neigh_clone(neigh)));
419 write_unlock(&neigh->lock);
420 neigh_release(neigh);
423 kfree_skb(skb);
424 return 0;
428 * Endnode hello message received
430 int dn_neigh_endnode_hello(struct sk_buff *skb)
432 struct endnode_hello_message *msg = (struct endnode_hello_message *)skb->data;
433 struct neighbour *neigh;
434 struct dn_neigh *dn;
435 __le16 src;
437 src = dn_eth2dn(msg->id);
439 neigh = __neigh_lookup(&dn_neigh_table, &src, skb->dev, 1);
441 dn = (struct dn_neigh *)neigh;
443 if (neigh) {
444 write_lock(&neigh->lock);
446 neigh->used = jiffies;
448 if (!(neigh->nud_state & NUD_PERMANENT)) {
449 neigh->updated = jiffies;
451 if (neigh->dev->type == ARPHRD_ETHER)
452 memcpy(neigh->ha, &eth_hdr(skb)->h_source, ETH_ALEN);
453 dn->flags &= ~(DN_NDFLAG_R1 | DN_NDFLAG_R2);
454 dn->blksize = dn_ntohs(msg->blksize);
455 dn->priority = 0;
458 write_unlock(&neigh->lock);
459 neigh_release(neigh);
462 kfree_skb(skb);
463 return 0;
466 static char *dn_find_slot(char *base, int max, int priority)
468 int i;
469 unsigned char *min = NULL;
471 base += 6; /* skip first id */
473 for(i = 0; i < max; i++) {
474 if (!min || (*base < *min))
475 min = base;
476 base += 7; /* find next priority */
479 if (!min)
480 return NULL;
482 return (*min < priority) ? (min - 6) : NULL;
485 struct elist_cb_state {
486 struct net_device *dev;
487 unsigned char *ptr;
488 unsigned char *rs;
489 int t, n;
492 static void neigh_elist_cb(struct neighbour *neigh, void *_info)
494 struct elist_cb_state *s = _info;
495 struct dn_neigh *dn;
497 if (neigh->dev != s->dev)
498 return;
500 dn = (struct dn_neigh *) neigh;
501 if (!(dn->flags & (DN_NDFLAG_R1|DN_NDFLAG_R2)))
502 return;
504 if (s->t == s->n)
505 s->rs = dn_find_slot(s->ptr, s->n, dn->priority);
506 else
507 s->t++;
508 if (s->rs == NULL)
509 return;
511 dn_dn2eth(s->rs, dn->addr);
512 s->rs += 6;
513 *(s->rs) = neigh->nud_state & NUD_CONNECTED ? 0x80 : 0x0;
514 *(s->rs) |= dn->priority;
515 s->rs++;
518 int dn_neigh_elist(struct net_device *dev, unsigned char *ptr, int n)
520 struct elist_cb_state state;
522 state.dev = dev;
523 state.t = 0;
524 state.n = n;
525 state.ptr = ptr;
526 state.rs = ptr;
528 neigh_for_each(&dn_neigh_table, neigh_elist_cb, &state);
530 return state.t;
534 #ifdef CONFIG_PROC_FS
536 static inline void dn_neigh_format_entry(struct seq_file *seq,
537 struct neighbour *n)
539 struct dn_neigh *dn = (struct dn_neigh *) n;
540 char buf[DN_ASCBUF_LEN];
542 read_lock(&n->lock);
543 seq_printf(seq, "%-7s %s%s%s %02x %02d %07ld %-8s\n",
544 dn_addr2asc(dn_ntohs(dn->addr), buf),
545 (dn->flags&DN_NDFLAG_R1) ? "1" : "-",
546 (dn->flags&DN_NDFLAG_R2) ? "2" : "-",
547 (dn->flags&DN_NDFLAG_P3) ? "3" : "-",
548 dn->n.nud_state,
549 atomic_read(&dn->n.refcnt),
550 dn->blksize,
551 (dn->n.dev) ? dn->n.dev->name : "?");
552 read_unlock(&n->lock);
555 static int dn_neigh_seq_show(struct seq_file *seq, void *v)
557 if (v == SEQ_START_TOKEN) {
558 seq_puts(seq, "Addr Flags State Use Blksize Dev\n");
559 } else {
560 dn_neigh_format_entry(seq, v);
563 return 0;
566 static void *dn_neigh_seq_start(struct seq_file *seq, loff_t *pos)
568 return neigh_seq_start(seq, pos, &dn_neigh_table,
569 NEIGH_SEQ_NEIGH_ONLY);
572 static struct seq_operations dn_neigh_seq_ops = {
573 .start = dn_neigh_seq_start,
574 .next = neigh_seq_next,
575 .stop = neigh_seq_stop,
576 .show = dn_neigh_seq_show,
579 static int dn_neigh_seq_open(struct inode *inode, struct file *file)
581 struct seq_file *seq;
582 int rc = -ENOMEM;
583 struct neigh_seq_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
585 if (!s)
586 goto out;
588 rc = seq_open(file, &dn_neigh_seq_ops);
589 if (rc)
590 goto out_kfree;
592 seq = file->private_data;
593 seq->private = s;
594 out:
595 return rc;
596 out_kfree:
597 kfree(s);
598 goto out;
601 static const struct file_operations dn_neigh_seq_fops = {
602 .owner = THIS_MODULE,
603 .open = dn_neigh_seq_open,
604 .read = seq_read,
605 .llseek = seq_lseek,
606 .release = seq_release_private,
609 #endif
611 void __init dn_neigh_init(void)
613 neigh_table_init(&dn_neigh_table);
614 proc_net_fops_create("decnet_neigh", S_IRUGO, &dn_neigh_seq_fops);
617 void __exit dn_neigh_cleanup(void)
619 proc_net_remove("decnet_neigh");
620 neigh_table_clear(&dn_neigh_table);