blob: 23aa3556e56f18f605824dddfac5feba7169c161 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 Routing Functions (Endnode and Router)
7 *
8 * Authors: Steve Whitehouse <SteveW@ACM.org>
9 * Eduardo Marcelo Serrat <emserrat@geocities.com>
10 *
11 * Changes:
12 * Steve Whitehouse : Fixes to allow "intra-ethernet" and
13 * "return-to-sender" bits on outgoing
14 * packets.
15 * Steve Whitehouse : Timeouts for cached routes.
16 * Steve Whitehouse : Use dst cache for input routes too.
17 * Steve Whitehouse : Fixed error values in dn_send_skb.
18 * Steve Whitehouse : Rework routing functions to better fit
19 * DECnet routing design
20 * Alexey Kuznetsov : New SMP locking
21 * Steve Whitehouse : More SMP locking changes & dn_cache_dump()
22 * Steve Whitehouse : Prerouting NF hook, now really is prerouting.
23 * Fixed possible skb leak in rtnetlink funcs.
24 * Steve Whitehouse : Dave Miller's dynamic hash table sizing and
25 * Alexey Kuznetsov's finer grained locking
26 * from ipv4/route.c.
27 * Steve Whitehouse : Routing is now starting to look like a
28 * sensible set of code now, mainly due to
29 * my copying the IPv4 routing code. The
30 * hooks here are modified and will continue
31 * to evolve for a while.
32 * Steve Whitehouse : Real SMP at last :-) Also new netfilter
33 * stuff. Look out raw sockets your days
34 * are numbered!
35 * Steve Whitehouse : Added return-to-sender functions. Added
36 * backlog congestion level return codes.
37 * Steve Whitehouse : Fixed bug where routes were set up with
38 * no ref count on net devices.
39 * Steve Whitehouse : RCU for the route cache
40 * Steve Whitehouse : Preparations for the flow cache
41 * Steve Whitehouse : Prepare for nonlinear skbs
42 */
43
44/******************************************************************************
45 (c) 1995-1998 E.M. Serrat emserrat@geocities.com
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +090046
Linus Torvalds1da177e2005-04-16 15:20:36 -070047 This program is free software; you can redistribute it and/or modify
48 it under the terms of the GNU General Public License as published by
49 the Free Software Foundation; either version 2 of the License, or
50 any later version.
51
52 This program is distributed in the hope that it will be useful,
53 but WITHOUT ANY WARRANTY; without even the implied warranty of
54 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
55 GNU General Public License for more details.
56*******************************************************************************/
57
Linus Torvalds1da177e2005-04-16 15:20:36 -070058#include <linux/errno.h>
59#include <linux/types.h>
60#include <linux/socket.h>
61#include <linux/in.h>
62#include <linux/kernel.h>
63#include <linux/sockios.h>
64#include <linux/net.h>
65#include <linux/netdevice.h>
66#include <linux/inet.h>
67#include <linux/route.h>
68#include <linux/in_route.h>
69#include <net/sock.h>
70#include <linux/mm.h>
71#include <linux/proc_fs.h>
72#include <linux/seq_file.h>
73#include <linux/init.h>
74#include <linux/rtnetlink.h>
75#include <linux/string.h>
76#include <linux/netfilter_decnet.h>
77#include <linux/rcupdate.h>
78#include <linux/times.h>
79#include <asm/errno.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +020080#include <net/net_namespace.h>
Arnaldo Carvalho de Melodc5fc572007-03-25 23:06:12 -070081#include <net/netlink.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070082#include <net/neighbour.h>
83#include <net/dst.h>
84#include <net/flow.h>
Steven Whitehousea8731cb2006-08-09 15:56:46 -070085#include <net/fib_rules.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070086#include <net/dn.h>
87#include <net/dn_dev.h>
88#include <net/dn_nsp.h>
89#include <net/dn_route.h>
90#include <net/dn_neigh.h>
91#include <net/dn_fib.h>
92
93struct dn_rt_hash_bucket
94{
95 struct dn_route *chain;
96 spinlock_t lock;
97} __attribute__((__aligned__(8)));
98
99extern struct neigh_table dn_neigh_table;
100
101
102static unsigned char dn_hiord_addr[6] = {0xAA,0x00,0x04,0x00,0x00,0x00};
103
104static const int dn_rt_min_delay = 2 * HZ;
105static const int dn_rt_max_delay = 10 * HZ;
106static const int dn_rt_mtu_expires = 10 * 60 * HZ;
107
108static unsigned long dn_rt_deadline;
109
110static int dn_dst_gc(void);
111static struct dst_entry *dn_dst_check(struct dst_entry *, __u32);
112static struct dst_entry *dn_dst_negative_advice(struct dst_entry *);
113static void dn_dst_link_failure(struct sk_buff *);
114static void dn_dst_update_pmtu(struct dst_entry *dst, u32 mtu);
115static int dn_route_input(struct sk_buff *);
116static void dn_run_flush(unsigned long dummy);
117
118static struct dn_rt_hash_bucket *dn_rt_hash_table;
119static unsigned dn_rt_hash_mask;
120
121static struct timer_list dn_route_timer;
Ingo Molnar8d06afa2005-09-09 13:10:40 -0700122static DEFINE_TIMER(dn_rt_flush_timer, dn_run_flush, 0, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123int decnet_dst_gc_interval = 2;
124
125static struct dst_ops dn_dst_ops = {
126 .family = PF_DECnet,
127 .protocol = __constant_htons(ETH_P_DNA_RT),
128 .gc_thresh = 128,
129 .gc = dn_dst_gc,
130 .check = dn_dst_check,
131 .negative_advice = dn_dst_negative_advice,
132 .link_failure = dn_dst_link_failure,
133 .update_pmtu = dn_dst_update_pmtu,
134 .entry_size = sizeof(struct dn_route),
135 .entries = ATOMIC_INIT(0),
136};
137
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800138static __inline__ unsigned dn_hash(__le16 src, __le16 dst)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139{
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800140 __u16 tmp = (__u16 __force)(src ^ dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 tmp ^= (tmp >> 3);
142 tmp ^= (tmp >> 5);
143 tmp ^= (tmp >> 10);
144 return dn_rt_hash_mask & (unsigned)tmp;
145}
146
147static inline void dnrt_free(struct dn_route *rt)
148{
149 call_rcu_bh(&rt->u.dst.rcu_head, dst_rcu_free);
150}
151
152static inline void dnrt_drop(struct dn_route *rt)
153{
Adrian Bunk34001122006-03-20 23:00:29 -0800154 dst_release(&rt->u.dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 call_rcu_bh(&rt->u.dst.rcu_head, dst_rcu_free);
156}
157
158static void dn_dst_check_expire(unsigned long dummy)
159{
160 int i;
161 struct dn_route *rt, **rtp;
162 unsigned long now = jiffies;
163 unsigned long expire = 120 * HZ;
164
165 for(i = 0; i <= dn_rt_hash_mask; i++) {
166 rtp = &dn_rt_hash_table[i].chain;
167
168 spin_lock(&dn_rt_hash_table[i].lock);
169 while((rt=*rtp) != NULL) {
170 if (atomic_read(&rt->u.dst.__refcnt) ||
171 (now - rt->u.dst.lastuse) < expire) {
Eric Dumazet0c195c32007-02-09 16:25:52 -0800172 rtp = &rt->u.dst.dn_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 continue;
174 }
Eric Dumazet0c195c32007-02-09 16:25:52 -0800175 *rtp = rt->u.dst.dn_next;
176 rt->u.dst.dn_next = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 dnrt_free(rt);
178 }
179 spin_unlock(&dn_rt_hash_table[i].lock);
180
181 if ((jiffies - now) > 0)
182 break;
183 }
184
185 mod_timer(&dn_route_timer, now + decnet_dst_gc_interval * HZ);
186}
187
188static int dn_dst_gc(void)
189{
190 struct dn_route *rt, **rtp;
191 int i;
192 unsigned long now = jiffies;
193 unsigned long expire = 10 * HZ;
194
195 for(i = 0; i <= dn_rt_hash_mask; i++) {
196
197 spin_lock_bh(&dn_rt_hash_table[i].lock);
198 rtp = &dn_rt_hash_table[i].chain;
199
200 while((rt=*rtp) != NULL) {
201 if (atomic_read(&rt->u.dst.__refcnt) ||
202 (now - rt->u.dst.lastuse) < expire) {
Eric Dumazet0c195c32007-02-09 16:25:52 -0800203 rtp = &rt->u.dst.dn_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 continue;
205 }
Eric Dumazet0c195c32007-02-09 16:25:52 -0800206 *rtp = rt->u.dst.dn_next;
207 rt->u.dst.dn_next = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 dnrt_drop(rt);
209 break;
210 }
211 spin_unlock_bh(&dn_rt_hash_table[i].lock);
212 }
213
214 return 0;
215}
216
217/*
218 * The decnet standards don't impose a particular minimum mtu, what they
219 * do insist on is that the routing layer accepts a datagram of at least
220 * 230 bytes long. Here we have to subtract the routing header length from
221 * 230 to get the minimum acceptable mtu. If there is no neighbour, then we
222 * assume the worst and use a long header size.
223 *
224 * We update both the mtu and the advertised mss (i.e. the segment size we
225 * advertise to the other end).
226 */
227static void dn_dst_update_pmtu(struct dst_entry *dst, u32 mtu)
228{
229 u32 min_mtu = 230;
230 struct dn_dev *dn = dst->neighbour ?
231 (struct dn_dev *)dst->neighbour->dev->dn_ptr : NULL;
232
233 if (dn && dn->use_long == 0)
234 min_mtu -= 6;
235 else
236 min_mtu -= 21;
237
238 if (dst->metrics[RTAX_MTU-1] > mtu && mtu >= min_mtu) {
239 if (!(dst_metric_locked(dst, RTAX_MTU))) {
240 dst->metrics[RTAX_MTU-1] = mtu;
241 dst_set_expires(dst, dn_rt_mtu_expires);
242 }
243 if (!(dst_metric_locked(dst, RTAX_ADVMSS))) {
244 u32 mss = mtu - DN_MAX_NSP_DATA_HEADER;
245 if (dst->metrics[RTAX_ADVMSS-1] > mss)
246 dst->metrics[RTAX_ADVMSS-1] = mss;
247 }
248 }
249}
250
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900251/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 * When a route has been marked obsolete. (e.g. routing cache flush)
253 */
254static struct dst_entry *dn_dst_check(struct dst_entry *dst, __u32 cookie)
255{
256 return NULL;
257}
258
259static struct dst_entry *dn_dst_negative_advice(struct dst_entry *dst)
260{
261 dst_release(dst);
262 return NULL;
263}
264
265static void dn_dst_link_failure(struct sk_buff *skb)
266{
267 return;
268}
269
270static inline int compare_keys(struct flowi *fl1, struct flowi *fl2)
271{
David S. Miller8238b212006-10-12 00:49:15 -0700272 return ((fl1->nl_u.dn_u.daddr ^ fl2->nl_u.dn_u.daddr) |
273 (fl1->nl_u.dn_u.saddr ^ fl2->nl_u.dn_u.saddr) |
Thomas Graf47dcf0c2006-11-09 15:20:38 -0800274 (fl1->mark ^ fl2->mark) |
David S. Miller8238b212006-10-12 00:49:15 -0700275 (fl1->nl_u.dn_u.scope ^ fl2->nl_u.dn_u.scope) |
276 (fl1->oif ^ fl2->oif) |
277 (fl1->iif ^ fl2->iif)) == 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278}
279
280static int dn_insert_route(struct dn_route *rt, unsigned hash, struct dn_route **rp)
281{
282 struct dn_route *rth, **rthp;
283 unsigned long now = jiffies;
284
285 rthp = &dn_rt_hash_table[hash].chain;
286
287 spin_lock_bh(&dn_rt_hash_table[hash].lock);
288 while((rth = *rthp) != NULL) {
289 if (compare_keys(&rth->fl, &rt->fl)) {
290 /* Put it first */
Eric Dumazet0c195c32007-02-09 16:25:52 -0800291 *rthp = rth->u.dst.dn_next;
292 rcu_assign_pointer(rth->u.dst.dn_next,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 dn_rt_hash_table[hash].chain);
294 rcu_assign_pointer(dn_rt_hash_table[hash].chain, rth);
295
Pavel Emelyanov03f49f32007-11-10 21:28:34 -0800296 dst_use(&rth->u.dst, now);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 spin_unlock_bh(&dn_rt_hash_table[hash].lock);
298
299 dnrt_drop(rt);
300 *rp = rth;
301 return 0;
302 }
Eric Dumazet0c195c32007-02-09 16:25:52 -0800303 rthp = &rth->u.dst.dn_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 }
305
Eric Dumazet0c195c32007-02-09 16:25:52 -0800306 rcu_assign_pointer(rt->u.dst.dn_next, dn_rt_hash_table[hash].chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 rcu_assign_pointer(dn_rt_hash_table[hash].chain, rt);
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900308
Pavel Emelyanov03f49f32007-11-10 21:28:34 -0800309 dst_use(&rt->u.dst, now);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 spin_unlock_bh(&dn_rt_hash_table[hash].lock);
311 *rp = rt;
312 return 0;
313}
314
315void dn_run_flush(unsigned long dummy)
316{
317 int i;
318 struct dn_route *rt, *next;
319
320 for(i = 0; i < dn_rt_hash_mask; i++) {
321 spin_lock_bh(&dn_rt_hash_table[i].lock);
322
323 if ((rt = xchg(&dn_rt_hash_table[i].chain, NULL)) == NULL)
324 goto nothing_to_declare;
325
326 for(; rt; rt=next) {
Eric Dumazet0c195c32007-02-09 16:25:52 -0800327 next = rt->u.dst.dn_next;
328 rt->u.dst.dn_next = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 dst_free((struct dst_entry *)rt);
330 }
331
332nothing_to_declare:
333 spin_unlock_bh(&dn_rt_hash_table[i].lock);
334 }
335}
336
337static DEFINE_SPINLOCK(dn_rt_flush_lock);
338
339void dn_rt_cache_flush(int delay)
340{
341 unsigned long now = jiffies;
342 int user_mode = !in_interrupt();
343
344 if (delay < 0)
345 delay = dn_rt_min_delay;
346
347 spin_lock_bh(&dn_rt_flush_lock);
348
349 if (del_timer(&dn_rt_flush_timer) && delay > 0 && dn_rt_deadline) {
350 long tmo = (long)(dn_rt_deadline - now);
351
352 if (user_mode && tmo < dn_rt_max_delay - dn_rt_min_delay)
353 tmo = 0;
354
355 if (delay > tmo)
356 delay = tmo;
357 }
358
359 if (delay <= 0) {
360 spin_unlock_bh(&dn_rt_flush_lock);
361 dn_run_flush(0);
362 return;
363 }
364
365 if (dn_rt_deadline == 0)
366 dn_rt_deadline = now + dn_rt_max_delay;
367
368 dn_rt_flush_timer.expires = now + delay;
369 add_timer(&dn_rt_flush_timer);
370 spin_unlock_bh(&dn_rt_flush_lock);
371}
372
373/**
374 * dn_return_short - Return a short packet to its sender
375 * @skb: The packet to return
376 *
377 */
378static int dn_return_short(struct sk_buff *skb)
379{
380 struct dn_skb_cb *cb;
381 unsigned char *ptr;
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800382 __le16 *src;
383 __le16 *dst;
384 __le16 tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385
386 /* Add back headers */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700387 skb_push(skb, skb->data - skb_network_header(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388
389 if ((skb = skb_unshare(skb, GFP_ATOMIC)) == NULL)
390 return NET_RX_DROP;
391
392 cb = DN_SKB_CB(skb);
393 /* Skip packet length and point to flags */
394 ptr = skb->data + 2;
395 *ptr++ = (cb->rt_flags & ~DN_RT_F_RQR) | DN_RT_F_RTS;
396
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800397 dst = (__le16 *)ptr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 ptr += 2;
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800399 src = (__le16 *)ptr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 ptr += 2;
401 *ptr = 0; /* Zero hop count */
402
403 /* Swap source and destination */
404 tmp = *src;
405 *src = *dst;
406 *dst = tmp;
407
408 skb->pkt_type = PACKET_OUTGOING;
409 dn_rt_finish_output(skb, NULL, NULL);
410 return NET_RX_SUCCESS;
411}
412
413/**
414 * dn_return_long - Return a long packet to its sender
415 * @skb: The long format packet to return
416 *
417 */
418static int dn_return_long(struct sk_buff *skb)
419{
420 struct dn_skb_cb *cb;
421 unsigned char *ptr;
422 unsigned char *src_addr, *dst_addr;
423 unsigned char tmp[ETH_ALEN];
424
425 /* Add back all headers */
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700426 skb_push(skb, skb->data - skb_network_header(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427
428 if ((skb = skb_unshare(skb, GFP_ATOMIC)) == NULL)
429 return NET_RX_DROP;
430
431 cb = DN_SKB_CB(skb);
432 /* Ignore packet length and point to flags */
433 ptr = skb->data + 2;
434
435 /* Skip padding */
436 if (*ptr & DN_RT_F_PF) {
437 char padlen = (*ptr & ~DN_RT_F_PF);
438 ptr += padlen;
439 }
440
441 *ptr++ = (cb->rt_flags & ~DN_RT_F_RQR) | DN_RT_F_RTS;
442 ptr += 2;
443 dst_addr = ptr;
444 ptr += 8;
445 src_addr = ptr;
446 ptr += 6;
447 *ptr = 0; /* Zero hop count */
448
449 /* Swap source and destination */
450 memcpy(tmp, src_addr, ETH_ALEN);
451 memcpy(src_addr, dst_addr, ETH_ALEN);
452 memcpy(dst_addr, tmp, ETH_ALEN);
453
454 skb->pkt_type = PACKET_OUTGOING;
455 dn_rt_finish_output(skb, dst_addr, src_addr);
456 return NET_RX_SUCCESS;
457}
458
459/**
460 * dn_route_rx_packet - Try and find a route for an incoming packet
461 * @skb: The packet to find a route for
462 *
463 * Returns: result of input function if route is found, error code otherwise
464 */
465static int dn_route_rx_packet(struct sk_buff *skb)
466{
467 struct dn_skb_cb *cb = DN_SKB_CB(skb);
468 int err;
469
470 if ((err = dn_route_input(skb)) == 0)
471 return dst_input(skb);
472
473 if (decnet_debug_level & 4) {
474 char *devname = skb->dev ? skb->dev->name : "???";
475 struct dn_skb_cb *cb = DN_SKB_CB(skb);
476 printk(KERN_DEBUG
477 "DECnet: dn_route_rx_packet: rt_flags=0x%02x dev=%s len=%d src=0x%04hx dst=0x%04hx err=%d type=%d\n",
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800478 (int)cb->rt_flags, devname, skb->len,
479 dn_ntohs(cb->src), dn_ntohs(cb->dst),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 err, skb->pkt_type);
481 }
482
483 if ((skb->pkt_type == PACKET_HOST) && (cb->rt_flags & DN_RT_F_RQR)) {
484 switch(cb->rt_flags & DN_RT_PKT_MSK) {
485 case DN_RT_PKT_SHORT:
486 return dn_return_short(skb);
487 case DN_RT_PKT_LONG:
488 return dn_return_long(skb);
489 }
490 }
491
492 kfree_skb(skb);
493 return NET_RX_DROP;
494}
495
496static int dn_route_rx_long(struct sk_buff *skb)
497{
498 struct dn_skb_cb *cb = DN_SKB_CB(skb);
499 unsigned char *ptr = skb->data;
500
501 if (!pskb_may_pull(skb, 21)) /* 20 for long header, 1 for shortest nsp */
502 goto drop_it;
503
504 skb_pull(skb, 20);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300505 skb_reset_transport_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900507 /* Destination info */
508 ptr += 2;
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800509 cb->dst = dn_eth2dn(ptr);
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900510 if (memcmp(ptr, dn_hiord_addr, 4) != 0)
511 goto drop_it;
512 ptr += 6;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513
514
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900515 /* Source info */
516 ptr += 2;
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800517 cb->src = dn_eth2dn(ptr);
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900518 if (memcmp(ptr, dn_hiord_addr, 4) != 0)
519 goto drop_it;
520 ptr += 6;
521 /* Other junk */
522 ptr++;
523 cb->hops = *ptr++; /* Visit Count */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524
525 return NF_HOOK(PF_DECnet, NF_DN_PRE_ROUTING, skb, skb->dev, NULL, dn_route_rx_packet);
526
527drop_it:
528 kfree_skb(skb);
529 return NET_RX_DROP;
530}
531
532
533
534static int dn_route_rx_short(struct sk_buff *skb)
535{
536 struct dn_skb_cb *cb = DN_SKB_CB(skb);
537 unsigned char *ptr = skb->data;
538
539 if (!pskb_may_pull(skb, 6)) /* 5 for short header + 1 for shortest nsp */
540 goto drop_it;
541
542 skb_pull(skb, 5);
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300543 skb_reset_transport_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800545 cb->dst = *(__le16 *)ptr;
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900546 ptr += 2;
547 cb->src = *(__le16 *)ptr;
548 ptr += 2;
549 cb->hops = *ptr & 0x3f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550
551 return NF_HOOK(PF_DECnet, NF_DN_PRE_ROUTING, skb, skb->dev, NULL, dn_route_rx_packet);
552
553drop_it:
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900554 kfree_skb(skb);
555 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556}
557
558static int dn_route_discard(struct sk_buff *skb)
559{
560 /*
561 * I know we drop the packet here, but thats considered success in
562 * this case
563 */
564 kfree_skb(skb);
565 return NET_RX_SUCCESS;
566}
567
568static int dn_route_ptp_hello(struct sk_buff *skb)
569{
570 dn_dev_hello(skb);
571 dn_neigh_pointopoint_hello(skb);
572 return NET_RX_SUCCESS;
573}
574
David S. Millerf2ccd8f2005-08-09 19:34:12 -0700575int dn_route_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576{
577 struct dn_skb_cb *cb;
578 unsigned char flags = 0;
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800579 __u16 len = dn_ntohs(*(__le16 *)skb->data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 struct dn_dev *dn = (struct dn_dev *)dev->dn_ptr;
581 unsigned char padlen = 0;
582
Eric W. Biedermane730c152007-09-17 11:53:39 -0700583 if (dev->nd_net != &init_net)
584 goto dump_it;
585
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 if (dn == NULL)
587 goto dump_it;
588
589 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL)
590 goto out;
591
592 if (!pskb_may_pull(skb, 3))
593 goto dump_it;
594
595 skb_pull(skb, 2);
596
597 if (len > skb->len)
598 goto dump_it;
599
600 skb_trim(skb, len);
601
602 flags = *skb->data;
603
604 cb = DN_SKB_CB(skb);
605 cb->stamp = jiffies;
606 cb->iif = dev->ifindex;
607
608 /*
609 * If we have padding, remove it.
610 */
611 if (flags & DN_RT_F_PF) {
612 padlen = flags & ~DN_RT_F_PF;
613 if (!pskb_may_pull(skb, padlen + 1))
614 goto dump_it;
615 skb_pull(skb, padlen);
616 flags = *skb->data;
617 }
618
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700619 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620
621 /*
622 * Weed out future version DECnet
623 */
624 if (flags & DN_RT_F_VER)
625 goto dump_it;
626
627 cb->rt_flags = flags;
628
629 if (decnet_debug_level & 1)
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900630 printk(KERN_DEBUG
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 "dn_route_rcv: got 0x%02x from %s [%d %d %d]\n",
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900632 (int)flags, (dev) ? dev->name : "???", len, skb->len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 padlen);
634
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900635 if (flags & DN_RT_PKT_CNTL) {
Herbert Xu364c6ba2006-06-09 16:10:40 -0700636 if (unlikely(skb_linearize(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 goto dump_it;
638
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900639 switch(flags & DN_RT_CNTL_MSK) {
640 case DN_RT_PKT_INIT:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 dn_dev_init_pkt(skb);
642 break;
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900643 case DN_RT_PKT_VERI:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 dn_dev_veri_pkt(skb);
645 break;
646 }
647
648 if (dn->parms.state != DN_DEV_S_RU)
649 goto dump_it;
650
651 switch(flags & DN_RT_CNTL_MSK) {
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900652 case DN_RT_PKT_HELO:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 return NF_HOOK(PF_DECnet, NF_DN_HELLO, skb, skb->dev, NULL, dn_route_ptp_hello);
654
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900655 case DN_RT_PKT_L1RT:
656 case DN_RT_PKT_L2RT:
657 return NF_HOOK(PF_DECnet, NF_DN_ROUTE, skb, skb->dev, NULL, dn_route_discard);
658 case DN_RT_PKT_ERTH:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 return NF_HOOK(PF_DECnet, NF_DN_HELLO, skb, skb->dev, NULL, dn_neigh_router_hello);
660
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900661 case DN_RT_PKT_EEDH:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 return NF_HOOK(PF_DECnet, NF_DN_HELLO, skb, skb->dev, NULL, dn_neigh_endnode_hello);
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900663 }
664 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 if (dn->parms.state != DN_DEV_S_RU)
666 goto dump_it;
667
668 skb_pull(skb, 1); /* Pull flags */
669
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900670 switch(flags & DN_RT_PKT_MSK) {
671 case DN_RT_PKT_LONG:
672 return dn_route_rx_long(skb);
673 case DN_RT_PKT_SHORT:
674 return dn_route_rx_short(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 }
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900676 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
678dump_it:
679 kfree_skb(skb);
680out:
681 return NET_RX_DROP;
682}
683
684static int dn_output(struct sk_buff *skb)
685{
686 struct dst_entry *dst = skb->dst;
687 struct dn_route *rt = (struct dn_route *)dst;
688 struct net_device *dev = dst->dev;
689 struct dn_skb_cb *cb = DN_SKB_CB(skb);
690 struct neighbour *neigh;
691
692 int err = -EINVAL;
693
694 if ((neigh = dst->neighbour) == NULL)
695 goto error;
696
697 skb->dev = dev;
698
699 cb->src = rt->rt_saddr;
700 cb->dst = rt->rt_daddr;
701
702 /*
703 * Always set the Intra-Ethernet bit on all outgoing packets
704 * originated on this node. Only valid flag from upper layers
705 * is return-to-sender-requested. Set hop count to 0 too.
706 */
707 cb->rt_flags &= ~DN_RT_F_RQR;
708 cb->rt_flags |= DN_RT_F_IE;
709 cb->hops = 0;
710
711 return NF_HOOK(PF_DECnet, NF_DN_LOCAL_OUT, skb, NULL, dev, neigh->output);
712
713error:
714 if (net_ratelimit())
715 printk(KERN_DEBUG "dn_output: This should not happen\n");
716
717 kfree_skb(skb);
718
719 return err;
720}
721
722static int dn_forward(struct sk_buff *skb)
723{
724 struct dn_skb_cb *cb = DN_SKB_CB(skb);
725 struct dst_entry *dst = skb->dst;
726 struct dn_dev *dn_db = dst->dev->dn_ptr;
727 struct dn_route *rt;
728 struct neighbour *neigh = dst->neighbour;
729 int header_len;
730#ifdef CONFIG_NETFILTER
731 struct net_device *dev = skb->dev;
732#endif
733
734 if (skb->pkt_type != PACKET_HOST)
735 goto drop;
736
737 /* Ensure that we have enough space for headers */
738 rt = (struct dn_route *)skb->dst;
739 header_len = dn_db->use_long ? 21 : 6;
740 if (skb_cow(skb, LL_RESERVED_SPACE(rt->u.dst.dev)+header_len))
741 goto drop;
742
743 /*
744 * Hop count exceeded.
745 */
746 if (++cb->hops > 30)
747 goto drop;
748
749 skb->dev = rt->u.dst.dev;
750
751 /*
752 * If packet goes out same interface it came in on, then set
753 * the Intra-Ethernet bit. This has no effect for short
754 * packets, so we don't need to test for them here.
755 */
756 cb->rt_flags &= ~DN_RT_F_IE;
757 if (rt->rt_flags & RTCF_DOREDIRECT)
758 cb->rt_flags |= DN_RT_F_IE;
759
760 return NF_HOOK(PF_DECnet, NF_DN_FORWARD, skb, dev, skb->dev, neigh->output);
761
762drop:
763 kfree_skb(skb);
764 return NET_RX_DROP;
765}
766
767/*
768 * Drop packet. This is used for endnodes and for
769 * when we should not be forwarding packets from
770 * this dest.
771 */
772static int dn_blackhole(struct sk_buff *skb)
773{
774 kfree_skb(skb);
775 return NET_RX_DROP;
776}
777
778/*
779 * Used to catch bugs. This should never normally get
780 * called.
781 */
782static int dn_rt_bug(struct sk_buff *skb)
783{
784 if (net_ratelimit()) {
785 struct dn_skb_cb *cb = DN_SKB_CB(skb);
786
787 printk(KERN_DEBUG "dn_rt_bug: skb from:%04x to:%04x\n",
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800788 dn_ntohs(cb->src), dn_ntohs(cb->dst));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 }
790
791 kfree_skb(skb);
792
793 return NET_RX_BAD;
794}
795
796static int dn_rt_set_next_hop(struct dn_route *rt, struct dn_fib_res *res)
797{
798 struct dn_fib_info *fi = res->fi;
799 struct net_device *dev = rt->u.dst.dev;
800 struct neighbour *n;
801 unsigned mss;
802
803 if (fi) {
804 if (DN_FIB_RES_GW(*res) &&
805 DN_FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK)
806 rt->rt_gateway = DN_FIB_RES_GW(*res);
807 memcpy(rt->u.dst.metrics, fi->fib_metrics,
808 sizeof(rt->u.dst.metrics));
809 }
810 rt->rt_type = res->type;
811
812 if (dev != NULL && rt->u.dst.neighbour == NULL) {
813 n = __neigh_lookup_errno(&dn_neigh_table, &rt->rt_gateway, dev);
814 if (IS_ERR(n))
815 return PTR_ERR(n);
816 rt->u.dst.neighbour = n;
817 }
818
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900819 if (rt->u.dst.metrics[RTAX_MTU-1] == 0 ||
820 rt->u.dst.metrics[RTAX_MTU-1] > rt->u.dst.dev->mtu)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 rt->u.dst.metrics[RTAX_MTU-1] = rt->u.dst.dev->mtu;
822 mss = dn_mss_from_pmtu(dev, dst_mtu(&rt->u.dst));
823 if (rt->u.dst.metrics[RTAX_ADVMSS-1] == 0 ||
824 rt->u.dst.metrics[RTAX_ADVMSS-1] > mss)
825 rt->u.dst.metrics[RTAX_ADVMSS-1] = mss;
826 return 0;
827}
828
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800829static inline int dn_match_addr(__le16 addr1, __le16 addr2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830{
831 __u16 tmp = dn_ntohs(addr1) ^ dn_ntohs(addr2);
832 int match = 16;
833 while(tmp) {
834 tmp >>= 1;
835 match--;
836 }
837 return match;
838}
839
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800840static __le16 dnet_select_source(const struct net_device *dev, __le16 daddr, int scope)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841{
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800842 __le16 saddr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 struct dn_dev *dn_db = dev->dn_ptr;
844 struct dn_ifaddr *ifa;
845 int best_match = 0;
846 int ret;
847
848 read_lock(&dev_base_lock);
849 for(ifa = dn_db->ifa_list; ifa; ifa = ifa->ifa_next) {
850 if (ifa->ifa_scope > scope)
851 continue;
852 if (!daddr) {
853 saddr = ifa->ifa_local;
854 break;
855 }
856 ret = dn_match_addr(daddr, ifa->ifa_local);
857 if (ret > best_match)
858 saddr = ifa->ifa_local;
859 if (best_match == 0)
860 saddr = ifa->ifa_local;
861 }
862 read_unlock(&dev_base_lock);
863
864 return saddr;
865}
866
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800867static inline __le16 __dn_fib_res_prefsrc(struct dn_fib_res *res)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868{
869 return dnet_select_source(DN_FIB_RES_DEV(*res), DN_FIB_RES_GW(*res), res->scope);
870}
871
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800872static inline __le16 dn_fib_rules_map_destination(__le16 daddr, struct dn_fib_res *res)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873{
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800874 __le16 mask = dnet_make_mask(res->prefixlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 return (daddr&~mask)|res->fi->fib_nh->nh_gw;
876}
877
878static int dn_route_output_slow(struct dst_entry **pprt, const struct flowi *oldflp, int try_hard)
879{
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900880 struct flowi fl = { .nl_u = { .dn_u =
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 { .daddr = oldflp->fld_dst,
882 .saddr = oldflp->fld_src,
883 .scope = RT_SCOPE_UNIVERSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 } },
Thomas Graf47dcf0c2006-11-09 15:20:38 -0800885 .mark = oldflp->mark,
Eric W. Biederman2774c7a2007-09-26 22:10:56 -0700886 .iif = init_net.loopback_dev->ifindex,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 .oif = oldflp->oif };
888 struct dn_route *rt = NULL;
Pavel Emelianov7562f872007-05-03 15:13:45 -0700889 struct net_device *dev_out = NULL, *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 struct neighbour *neigh = NULL;
891 unsigned hash;
892 unsigned flags = 0;
893 struct dn_fib_res res = { .fi = NULL, .type = RTN_UNICAST };
894 int err;
895 int free_res = 0;
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800896 __le16 gateway = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897
898 if (decnet_debug_level & 16)
899 printk(KERN_DEBUG
900 "dn_route_output_slow: dst=%04x src=%04x mark=%d"
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800901 " iif=%d oif=%d\n", dn_ntohs(oldflp->fld_dst),
902 dn_ntohs(oldflp->fld_src),
Eric W. Biederman2774c7a2007-09-26 22:10:56 -0700903 oldflp->mark, init_net.loopback_dev->ifindex, oldflp->oif);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904
905 /* If we have an output interface, verify its a DECnet device */
906 if (oldflp->oif) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700907 dev_out = dev_get_by_index(&init_net, oldflp->oif);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 err = -ENODEV;
909 if (dev_out && dev_out->dn_ptr == NULL) {
910 dev_put(dev_out);
911 dev_out = NULL;
912 }
913 if (dev_out == NULL)
914 goto out;
915 }
916
917 /* If we have a source address, verify that its a local address */
918 if (oldflp->fld_src) {
919 err = -EADDRNOTAVAIL;
920
921 if (dev_out) {
922 if (dn_dev_islocal(dev_out, oldflp->fld_src))
923 goto source_ok;
924 dev_put(dev_out);
925 goto out;
926 }
927 read_lock(&dev_base_lock);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700928 for_each_netdev(&init_net, dev) {
Pavel Emelianov7562f872007-05-03 15:13:45 -0700929 if (!dev->dn_ptr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 continue;
Pavel Emelianov7562f872007-05-03 15:13:45 -0700931 if (!dn_dev_islocal(dev, oldflp->fld_src))
Patrick Caulfield9bbf28a12006-08-02 14:14:44 -0700932 continue;
Pavel Emelianov7562f872007-05-03 15:13:45 -0700933 if ((dev->flags & IFF_LOOPBACK) &&
Patrick Caulfield9bbf28a12006-08-02 14:14:44 -0700934 oldflp->fld_dst &&
Pavel Emelianov7562f872007-05-03 15:13:45 -0700935 !dn_dev_islocal(dev, oldflp->fld_dst))
Patrick Caulfield9bbf28a12006-08-02 14:14:44 -0700936 continue;
Pavel Emelianov7562f872007-05-03 15:13:45 -0700937
938 dev_out = dev;
Patrick Caulfield9bbf28a12006-08-02 14:14:44 -0700939 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 }
941 read_unlock(&dev_base_lock);
942 if (dev_out == NULL)
943 goto out;
944 dev_hold(dev_out);
945source_ok:
946 ;
947 }
948
949 /* No destination? Assume its local */
950 if (!fl.fld_dst) {
951 fl.fld_dst = fl.fld_src;
952
953 err = -EADDRNOTAVAIL;
954 if (dev_out)
955 dev_put(dev_out);
Eric W. Biederman2774c7a2007-09-26 22:10:56 -0700956 dev_out = init_net.loopback_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 dev_hold(dev_out);
958 if (!fl.fld_dst) {
959 fl.fld_dst =
960 fl.fld_src = dnet_select_source(dev_out, 0,
961 RT_SCOPE_HOST);
962 if (!fl.fld_dst)
963 goto out;
964 }
Eric W. Biederman2774c7a2007-09-26 22:10:56 -0700965 fl.oif = init_net.loopback_dev->ifindex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 res.type = RTN_LOCAL;
967 goto make_route;
968 }
969
970 if (decnet_debug_level & 16)
971 printk(KERN_DEBUG
972 "dn_route_output_slow: initial checks complete."
Steven Whitehousec4ea94a2006-03-20 22:42:39 -0800973 " dst=%o4x src=%04x oif=%d try_hard=%d\n",
974 dn_ntohs(fl.fld_dst), dn_ntohs(fl.fld_src),
975 fl.oif, try_hard);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976
977 /*
978 * N.B. If the kernel is compiled without router support then
979 * dn_fib_lookup() will evaluate to non-zero so this if () block
980 * will always be executed.
981 */
982 err = -ESRCH;
983 if (try_hard || (err = dn_fib_lookup(&fl, &res)) != 0) {
984 struct dn_dev *dn_db;
985 if (err != -ESRCH)
986 goto out;
987 /*
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900988 * Here the fallback is basically the standard algorithm for
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 * routing in endnodes which is described in the DECnet routing
990 * docs
991 *
992 * If we are not trying hard, look in neighbour cache.
993 * The result is tested to ensure that if a specific output
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +0900994 * device/source address was requested, then we honour that
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 * here
996 */
997 if (!try_hard) {
998 neigh = neigh_lookup_nodev(&dn_neigh_table, &fl.fld_dst);
999 if (neigh) {
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001000 if ((oldflp->oif &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 (neigh->dev->ifindex != oldflp->oif)) ||
1002 (oldflp->fld_src &&
1003 (!dn_dev_islocal(neigh->dev,
1004 oldflp->fld_src)))) {
1005 neigh_release(neigh);
1006 neigh = NULL;
1007 } else {
1008 if (dev_out)
1009 dev_put(dev_out);
1010 if (dn_dev_islocal(neigh->dev, fl.fld_dst)) {
Eric W. Biederman2774c7a2007-09-26 22:10:56 -07001011 dev_out = init_net.loopback_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 res.type = RTN_LOCAL;
1013 } else {
1014 dev_out = neigh->dev;
1015 }
1016 dev_hold(dev_out);
1017 goto select_source;
1018 }
1019 }
1020 }
1021
1022 /* Not there? Perhaps its a local address */
1023 if (dev_out == NULL)
1024 dev_out = dn_dev_get_default();
1025 err = -ENODEV;
1026 if (dev_out == NULL)
1027 goto out;
1028 dn_db = dev_out->dn_ptr;
1029 /* Possible improvement - check all devices for local addr */
1030 if (dn_dev_islocal(dev_out, fl.fld_dst)) {
1031 dev_put(dev_out);
Eric W. Biederman2774c7a2007-09-26 22:10:56 -07001032 dev_out = init_net.loopback_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 dev_hold(dev_out);
1034 res.type = RTN_LOCAL;
1035 goto select_source;
1036 }
1037 /* Not local either.... try sending it to the default router */
1038 neigh = neigh_clone(dn_db->router);
1039 BUG_ON(neigh && neigh->dev != dev_out);
1040
1041 /* Ok then, we assume its directly connected and move on */
1042select_source:
1043 if (neigh)
1044 gateway = ((struct dn_neigh *)neigh)->addr;
1045 if (gateway == 0)
1046 gateway = fl.fld_dst;
1047 if (fl.fld_src == 0) {
1048 fl.fld_src = dnet_select_source(dev_out, gateway,
1049 res.type == RTN_LOCAL ?
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001050 RT_SCOPE_HOST :
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 RT_SCOPE_LINK);
1052 if (fl.fld_src == 0 && res.type != RTN_LOCAL)
1053 goto e_addr;
1054 }
1055 fl.oif = dev_out->ifindex;
1056 goto make_route;
1057 }
1058 free_res = 1;
1059
1060 if (res.type == RTN_NAT)
1061 goto e_inval;
1062
1063 if (res.type == RTN_LOCAL) {
1064 if (!fl.fld_src)
1065 fl.fld_src = fl.fld_dst;
1066 if (dev_out)
1067 dev_put(dev_out);
Eric W. Biederman2774c7a2007-09-26 22:10:56 -07001068 dev_out = init_net.loopback_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 dev_hold(dev_out);
1070 fl.oif = dev_out->ifindex;
1071 if (res.fi)
1072 dn_fib_info_put(res.fi);
1073 res.fi = NULL;
1074 goto make_route;
1075 }
1076
1077 if (res.fi->fib_nhs > 1 && fl.oif == 0)
1078 dn_fib_select_multipath(&fl, &res);
1079
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001080 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 * We could add some logic to deal with default routes here and
1082 * get rid of some of the special casing above.
1083 */
1084
1085 if (!fl.fld_src)
1086 fl.fld_src = DN_FIB_RES_PREFSRC(res);
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001087
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 if (dev_out)
1089 dev_put(dev_out);
1090 dev_out = DN_FIB_RES_DEV(res);
1091 dev_hold(dev_out);
1092 fl.oif = dev_out->ifindex;
1093 gateway = DN_FIB_RES_GW(res);
1094
1095make_route:
1096 if (dev_out->flags & IFF_LOOPBACK)
1097 flags |= RTCF_LOCAL;
1098
1099 rt = dst_alloc(&dn_dst_ops);
1100 if (rt == NULL)
1101 goto e_nobufs;
1102
1103 atomic_set(&rt->u.dst.__refcnt, 1);
1104 rt->u.dst.flags = DST_HOST;
1105
1106 rt->fl.fld_src = oldflp->fld_src;
1107 rt->fl.fld_dst = oldflp->fld_dst;
1108 rt->fl.oif = oldflp->oif;
1109 rt->fl.iif = 0;
Thomas Graf47dcf0c2006-11-09 15:20:38 -08001110 rt->fl.mark = oldflp->mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111
1112 rt->rt_saddr = fl.fld_src;
1113 rt->rt_daddr = fl.fld_dst;
1114 rt->rt_gateway = gateway ? gateway : fl.fld_dst;
1115 rt->rt_local_src = fl.fld_src;
1116
1117 rt->rt_dst_map = fl.fld_dst;
1118 rt->rt_src_map = fl.fld_src;
1119
1120 rt->u.dst.dev = dev_out;
1121 dev_hold(dev_out);
1122 rt->u.dst.neighbour = neigh;
1123 neigh = NULL;
1124
1125 rt->u.dst.lastuse = jiffies;
1126 rt->u.dst.output = dn_output;
1127 rt->u.dst.input = dn_rt_bug;
1128 rt->rt_flags = flags;
1129 if (flags & RTCF_LOCAL)
1130 rt->u.dst.input = dn_nsp_rx;
1131
1132 err = dn_rt_set_next_hop(rt, &res);
1133 if (err)
1134 goto e_neighbour;
1135
1136 hash = dn_hash(rt->fl.fld_src, rt->fl.fld_dst);
1137 dn_insert_route(rt, hash, (struct dn_route **)pprt);
1138
1139done:
1140 if (neigh)
1141 neigh_release(neigh);
1142 if (free_res)
1143 dn_fib_res_put(&res);
1144 if (dev_out)
1145 dev_put(dev_out);
1146out:
1147 return err;
1148
1149e_addr:
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001150 err = -EADDRNOTAVAIL;
1151 goto done;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152e_inval:
1153 err = -EINVAL;
1154 goto done;
1155e_nobufs:
1156 err = -ENOBUFS;
1157 goto done;
1158e_neighbour:
1159 dst_free(&rt->u.dst);
1160 goto e_nobufs;
1161}
1162
1163
1164/*
1165 * N.B. The flags may be moved into the flowi at some future stage.
1166 */
1167static int __dn_route_output_key(struct dst_entry **pprt, const struct flowi *flp, int flags)
1168{
1169 unsigned hash = dn_hash(flp->fld_src, flp->fld_dst);
1170 struct dn_route *rt = NULL;
1171
1172 if (!(flags & MSG_TRYHARD)) {
1173 rcu_read_lock_bh();
1174 for(rt = rcu_dereference(dn_rt_hash_table[hash].chain); rt;
Eric Dumazet0c195c32007-02-09 16:25:52 -08001175 rt = rcu_dereference(rt->u.dst.dn_next)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 if ((flp->fld_dst == rt->fl.fld_dst) &&
1177 (flp->fld_src == rt->fl.fld_src) &&
Thomas Graf47dcf0c2006-11-09 15:20:38 -08001178 (flp->mark == rt->fl.mark) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 (rt->fl.iif == 0) &&
1180 (rt->fl.oif == flp->oif)) {
Pavel Emelyanov03f49f32007-11-10 21:28:34 -08001181 dst_use(&rt->u.dst, jiffies);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 rcu_read_unlock_bh();
1183 *pprt = &rt->u.dst;
1184 return 0;
1185 }
1186 }
1187 rcu_read_unlock_bh();
1188 }
1189
1190 return dn_route_output_slow(pprt, flp, flags);
1191}
1192
1193static int dn_route_output_key(struct dst_entry **pprt, struct flowi *flp, int flags)
1194{
1195 int err;
1196
1197 err = __dn_route_output_key(pprt, flp, flags);
1198 if (err == 0 && flp->proto) {
1199 err = xfrm_lookup(pprt, flp, NULL, 0);
1200 }
1201 return err;
1202}
1203
1204int dn_route_output_sock(struct dst_entry **pprt, struct flowi *fl, struct sock *sk, int flags)
1205{
1206 int err;
1207
1208 err = __dn_route_output_key(pprt, fl, flags & MSG_TRYHARD);
1209 if (err == 0 && fl->proto) {
1210 err = xfrm_lookup(pprt, fl, sk, !(flags & MSG_DONTWAIT));
1211 }
1212 return err;
1213}
1214
1215static int dn_route_input_slow(struct sk_buff *skb)
1216{
1217 struct dn_route *rt = NULL;
1218 struct dn_skb_cb *cb = DN_SKB_CB(skb);
1219 struct net_device *in_dev = skb->dev;
1220 struct net_device *out_dev = NULL;
1221 struct dn_dev *dn_db;
1222 struct neighbour *neigh = NULL;
1223 unsigned hash;
1224 int flags = 0;
Steven Whitehousec4ea94a2006-03-20 22:42:39 -08001225 __le16 gateway = 0;
1226 __le16 local_src = 0;
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001227 struct flowi fl = { .nl_u = { .dn_u =
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228 { .daddr = cb->dst,
1229 .saddr = cb->src,
1230 .scope = RT_SCOPE_UNIVERSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 } },
Thomas Graf47dcf0c2006-11-09 15:20:38 -08001232 .mark = skb->mark,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 .iif = skb->dev->ifindex };
1234 struct dn_fib_res res = { .fi = NULL, .type = RTN_UNREACHABLE };
1235 int err = -EINVAL;
1236 int free_res = 0;
1237
1238 dev_hold(in_dev);
1239
1240 if ((dn_db = in_dev->dn_ptr) == NULL)
1241 goto out;
1242
1243 /* Zero source addresses are not allowed */
1244 if (fl.fld_src == 0)
1245 goto out;
1246
1247 /*
1248 * In this case we've just received a packet from a source
1249 * outside ourselves pretending to come from us. We don't
1250 * allow it any further to prevent routing loops, spoofing and
1251 * other nasties. Loopback packets already have the dst attached
1252 * so this only affects packets which have originated elsewhere.
1253 */
1254 err = -ENOTUNIQ;
1255 if (dn_dev_islocal(in_dev, cb->src))
1256 goto out;
1257
1258 err = dn_fib_lookup(&fl, &res);
1259 if (err) {
1260 if (err != -ESRCH)
1261 goto out;
1262 /*
1263 * Is the destination us ?
1264 */
1265 if (!dn_dev_islocal(in_dev, cb->dst))
1266 goto e_inval;
1267
1268 res.type = RTN_LOCAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 } else {
Steven Whitehousec4ea94a2006-03-20 22:42:39 -08001270 __le16 src_map = fl.fld_src;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 free_res = 1;
1272
1273 out_dev = DN_FIB_RES_DEV(res);
1274 if (out_dev == NULL) {
1275 if (net_ratelimit())
1276 printk(KERN_CRIT "Bug in dn_route_input_slow() "
1277 "No output device\n");
1278 goto e_inval;
1279 }
1280 dev_hold(out_dev);
1281
1282 if (res.r)
Steven Whitehousea8731cb2006-08-09 15:56:46 -07001283 src_map = fl.fld_src; /* no NAT support for now */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284
1285 gateway = DN_FIB_RES_GW(res);
1286 if (res.type == RTN_NAT) {
1287 fl.fld_dst = dn_fib_rules_map_destination(fl.fld_dst, &res);
1288 dn_fib_res_put(&res);
1289 free_res = 0;
1290 if (dn_fib_lookup(&fl, &res))
1291 goto e_inval;
1292 free_res = 1;
1293 if (res.type != RTN_UNICAST)
1294 goto e_inval;
1295 flags |= RTCF_DNAT;
1296 gateway = fl.fld_dst;
1297 }
1298 fl.fld_src = src_map;
1299 }
1300
1301 switch(res.type) {
1302 case RTN_UNICAST:
1303 /*
1304 * Forwarding check here, we only check for forwarding
1305 * being turned off, if you want to only forward intra
1306 * area, its up to you to set the routing tables up
1307 * correctly.
1308 */
1309 if (dn_db->parms.forwarding == 0)
1310 goto e_inval;
1311
1312 if (res.fi->fib_nhs > 1 && fl.oif == 0)
1313 dn_fib_select_multipath(&fl, &res);
1314
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001315 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 * Check for out_dev == in_dev. We use the RTCF_DOREDIRECT
1317 * flag as a hint to set the intra-ethernet bit when
1318 * forwarding. If we've got NAT in operation, we don't do
1319 * this optimisation.
1320 */
1321 if (out_dev == in_dev && !(flags & RTCF_NAT))
1322 flags |= RTCF_DOREDIRECT;
1323
1324 local_src = DN_FIB_RES_PREFSRC(res);
1325
1326 case RTN_BLACKHOLE:
1327 case RTN_UNREACHABLE:
1328 break;
1329 case RTN_LOCAL:
1330 flags |= RTCF_LOCAL;
1331 fl.fld_src = cb->dst;
1332 fl.fld_dst = cb->src;
1333
1334 /* Routing tables gave us a gateway */
1335 if (gateway)
1336 goto make_route;
1337
1338 /* Packet was intra-ethernet, so we know its on-link */
Steven Whitehouse3a31b9d2006-10-18 20:45:22 -07001339 if (cb->rt_flags & DN_RT_F_IE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340 gateway = cb->src;
1341 flags |= RTCF_DIRECTSRC;
1342 goto make_route;
1343 }
1344
1345 /* Use the default router if there is one */
1346 neigh = neigh_clone(dn_db->router);
1347 if (neigh) {
1348 gateway = ((struct dn_neigh *)neigh)->addr;
1349 goto make_route;
1350 }
1351
1352 /* Close eyes and pray */
1353 gateway = cb->src;
1354 flags |= RTCF_DIRECTSRC;
1355 goto make_route;
1356 default:
1357 goto e_inval;
1358 }
1359
1360make_route:
1361 rt = dst_alloc(&dn_dst_ops);
1362 if (rt == NULL)
1363 goto e_nobufs;
1364
1365 rt->rt_saddr = fl.fld_src;
1366 rt->rt_daddr = fl.fld_dst;
1367 rt->rt_gateway = fl.fld_dst;
1368 if (gateway)
1369 rt->rt_gateway = gateway;
1370 rt->rt_local_src = local_src ? local_src : rt->rt_saddr;
1371
1372 rt->rt_dst_map = fl.fld_dst;
1373 rt->rt_src_map = fl.fld_src;
1374
1375 rt->fl.fld_src = cb->src;
1376 rt->fl.fld_dst = cb->dst;
1377 rt->fl.oif = 0;
1378 rt->fl.iif = in_dev->ifindex;
Thomas Graf47dcf0c2006-11-09 15:20:38 -08001379 rt->fl.mark = fl.mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380
1381 rt->u.dst.flags = DST_HOST;
1382 rt->u.dst.neighbour = neigh;
1383 rt->u.dst.dev = out_dev;
1384 rt->u.dst.lastuse = jiffies;
1385 rt->u.dst.output = dn_rt_bug;
1386 switch(res.type) {
1387 case RTN_UNICAST:
1388 rt->u.dst.input = dn_forward;
1389 break;
1390 case RTN_LOCAL:
1391 rt->u.dst.output = dn_output;
1392 rt->u.dst.input = dn_nsp_rx;
1393 rt->u.dst.dev = in_dev;
1394 flags |= RTCF_LOCAL;
1395 break;
1396 default:
1397 case RTN_UNREACHABLE:
1398 case RTN_BLACKHOLE:
1399 rt->u.dst.input = dn_blackhole;
1400 }
1401 rt->rt_flags = flags;
1402 if (rt->u.dst.dev)
1403 dev_hold(rt->u.dst.dev);
1404
1405 err = dn_rt_set_next_hop(rt, &res);
1406 if (err)
1407 goto e_neighbour;
1408
1409 hash = dn_hash(rt->fl.fld_src, rt->fl.fld_dst);
1410 dn_insert_route(rt, hash, (struct dn_route **)&skb->dst);
1411
1412done:
1413 if (neigh)
1414 neigh_release(neigh);
1415 if (free_res)
1416 dn_fib_res_put(&res);
1417 dev_put(in_dev);
1418 if (out_dev)
1419 dev_put(out_dev);
1420out:
1421 return err;
1422
1423e_inval:
1424 err = -EINVAL;
1425 goto done;
1426
1427e_nobufs:
1428 err = -ENOBUFS;
1429 goto done;
1430
1431e_neighbour:
1432 dst_free(&rt->u.dst);
1433 goto done;
1434}
1435
1436int dn_route_input(struct sk_buff *skb)
1437{
1438 struct dn_route *rt;
1439 struct dn_skb_cb *cb = DN_SKB_CB(skb);
1440 unsigned hash = dn_hash(cb->src, cb->dst);
1441
1442 if (skb->dst)
1443 return 0;
1444
1445 rcu_read_lock();
1446 for(rt = rcu_dereference(dn_rt_hash_table[hash].chain); rt != NULL;
Eric Dumazet0c195c32007-02-09 16:25:52 -08001447 rt = rcu_dereference(rt->u.dst.dn_next)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 if ((rt->fl.fld_src == cb->src) &&
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001449 (rt->fl.fld_dst == cb->dst) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450 (rt->fl.oif == 0) &&
Thomas Graf47dcf0c2006-11-09 15:20:38 -08001451 (rt->fl.mark == skb->mark) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 (rt->fl.iif == cb->iif)) {
Pavel Emelyanov03f49f32007-11-10 21:28:34 -08001453 dst_use(&rt->u.dst, jiffies);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 rcu_read_unlock();
1455 skb->dst = (struct dst_entry *)rt;
1456 return 0;
1457 }
1458 }
1459 rcu_read_unlock();
1460
1461 return dn_route_input_slow(skb);
1462}
1463
Jamal Hadi Salimb6544c02005-06-18 22:54:12 -07001464static int dn_rt_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1465 int event, int nowait, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466{
1467 struct dn_route *rt = (struct dn_route *)skb->dst;
1468 struct rtmsg *r;
1469 struct nlmsghdr *nlh;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001470 unsigned char *b = skb_tail_pointer(skb);
Thomas Grafe3703b32006-11-27 09:27:07 -08001471 long expires;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472
Jamal Hadi Salimb6544c02005-06-18 22:54:12 -07001473 nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*r), flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 r = NLMSG_DATA(nlh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 r->rtm_family = AF_DECnet;
1476 r->rtm_dst_len = 16;
1477 r->rtm_src_len = 0;
1478 r->rtm_tos = 0;
1479 r->rtm_table = RT_TABLE_MAIN;
Patrick McHardy9e762a42006-08-10 23:09:48 -07001480 RTA_PUT_U32(skb, RTA_TABLE, RT_TABLE_MAIN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 r->rtm_type = rt->rt_type;
1482 r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED;
1483 r->rtm_scope = RT_SCOPE_UNIVERSE;
1484 r->rtm_protocol = RTPROT_UNSPEC;
1485 if (rt->rt_flags & RTCF_NOTIFY)
1486 r->rtm_flags |= RTM_F_NOTIFY;
1487 RTA_PUT(skb, RTA_DST, 2, &rt->rt_daddr);
1488 if (rt->fl.fld_src) {
1489 r->rtm_src_len = 16;
1490 RTA_PUT(skb, RTA_SRC, 2, &rt->fl.fld_src);
1491 }
1492 if (rt->u.dst.dev)
1493 RTA_PUT(skb, RTA_OIF, sizeof(int), &rt->u.dst.dev->ifindex);
1494 /*
1495 * Note to self - change this if input routes reverse direction when
1496 * they deal only with inputs and not with replies like they do
1497 * currently.
1498 */
1499 RTA_PUT(skb, RTA_PREFSRC, 2, &rt->rt_local_src);
1500 if (rt->rt_daddr != rt->rt_gateway)
1501 RTA_PUT(skb, RTA_GATEWAY, 2, &rt->rt_gateway);
1502 if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
1503 goto rtattr_failure;
Thomas Grafe3703b32006-11-27 09:27:07 -08001504 expires = rt->u.dst.expires ? rt->u.dst.expires - jiffies : 0;
1505 if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0, expires,
1506 rt->u.dst.error) < 0)
1507 goto rtattr_failure;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508 if (rt->fl.iif)
1509 RTA_PUT(skb, RTA_IIF, sizeof(int), &rt->fl.iif);
1510
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001511 nlh->nlmsg_len = skb_tail_pointer(skb) - b;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 return skb->len;
1513
1514nlmsg_failure:
1515rtattr_failure:
Arnaldo Carvalho de Melodc5fc572007-03-25 23:06:12 -07001516 nlmsg_trim(skb, b);
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001517 return -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518}
1519
1520/*
1521 * This is called by both endnodes and routers now.
1522 */
Thomas Graffa34ddd2007-03-22 11:57:46 -07001523static int dn_cache_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh, void *arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524{
1525 struct rtattr **rta = arg;
1526 struct rtmsg *rtm = NLMSG_DATA(nlh);
1527 struct dn_route *rt = NULL;
1528 struct dn_skb_cb *cb;
1529 int err;
1530 struct sk_buff *skb;
1531 struct flowi fl;
1532
1533 memset(&fl, 0, sizeof(fl));
1534 fl.proto = DNPROTO_NSP;
1535
1536 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1537 if (skb == NULL)
1538 return -ENOBUFS;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001539 skb_reset_mac_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 cb = DN_SKB_CB(skb);
1541
1542 if (rta[RTA_SRC-1])
1543 memcpy(&fl.fld_src, RTA_DATA(rta[RTA_SRC-1]), 2);
1544 if (rta[RTA_DST-1])
1545 memcpy(&fl.fld_dst, RTA_DATA(rta[RTA_DST-1]), 2);
1546 if (rta[RTA_IIF-1])
1547 memcpy(&fl.iif, RTA_DATA(rta[RTA_IIF-1]), sizeof(int));
1548
1549 if (fl.iif) {
1550 struct net_device *dev;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001551 if ((dev = dev_get_by_index(&init_net, fl.iif)) == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 kfree_skb(skb);
1553 return -ENODEV;
1554 }
1555 if (!dev->dn_ptr) {
1556 dev_put(dev);
1557 kfree_skb(skb);
1558 return -ENODEV;
1559 }
1560 skb->protocol = __constant_htons(ETH_P_DNA_RT);
1561 skb->dev = dev;
1562 cb->src = fl.fld_src;
1563 cb->dst = fl.fld_dst;
1564 local_bh_disable();
1565 err = dn_route_input(skb);
1566 local_bh_enable();
1567 memset(cb, 0, sizeof(struct dn_skb_cb));
1568 rt = (struct dn_route *)skb->dst;
1569 if (!err && -rt->u.dst.error)
1570 err = rt->u.dst.error;
1571 } else {
1572 int oif = 0;
1573 if (rta[RTA_OIF - 1])
1574 memcpy(&oif, RTA_DATA(rta[RTA_OIF - 1]), sizeof(int));
1575 fl.oif = oif;
1576 err = dn_route_output_key((struct dst_entry **)&rt, &fl, 0);
1577 }
1578
1579 if (skb->dev)
1580 dev_put(skb->dev);
1581 skb->dev = NULL;
1582 if (err)
1583 goto out_free;
1584 skb->dst = &rt->u.dst;
1585 if (rtm->rtm_flags & RTM_F_NOTIFY)
1586 rt->rt_flags |= RTCF_NOTIFY;
1587
Jamal Hadi Salimb6544c02005-06-18 22:54:12 -07001588 err = dn_rt_fill_info(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, RTM_NEWROUTE, 0, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589
1590 if (err == 0)
1591 goto out_free;
1592 if (err < 0) {
1593 err = -EMSGSIZE;
1594 goto out_free;
1595 }
1596
Thomas Graf2942e902006-08-15 00:30:25 -07001597 return rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598
1599out_free:
1600 kfree_skb(skb);
1601 return err;
1602}
1603
1604/*
1605 * For routers, this is called from dn_fib_dump, but for endnodes its
1606 * called directly from the rtnetlink dispatch table.
1607 */
1608int dn_cache_dump(struct sk_buff *skb, struct netlink_callback *cb)
1609{
1610 struct dn_route *rt;
1611 int h, s_h;
1612 int idx, s_idx;
1613
1614 if (NLMSG_PAYLOAD(cb->nlh, 0) < sizeof(struct rtmsg))
1615 return -EINVAL;
1616 if (!(((struct rtmsg *)NLMSG_DATA(cb->nlh))->rtm_flags&RTM_F_CLONED))
1617 return 0;
1618
1619 s_h = cb->args[0];
1620 s_idx = idx = cb->args[1];
1621 for(h = 0; h <= dn_rt_hash_mask; h++) {
1622 if (h < s_h)
1623 continue;
1624 if (h > s_h)
1625 s_idx = 0;
1626 rcu_read_lock_bh();
1627 for(rt = rcu_dereference(dn_rt_hash_table[h].chain), idx = 0;
1628 rt;
Eric Dumazet0c195c32007-02-09 16:25:52 -08001629 rt = rcu_dereference(rt->u.dst.dn_next), idx++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 if (idx < s_idx)
1631 continue;
1632 skb->dst = dst_clone(&rt->u.dst);
1633 if (dn_rt_fill_info(skb, NETLINK_CB(cb->skb).pid,
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001634 cb->nlh->nlmsg_seq, RTM_NEWROUTE,
Jamal Hadi Salimb6544c02005-06-18 22:54:12 -07001635 1, NLM_F_MULTI) <= 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 dst_release(xchg(&skb->dst, NULL));
1637 rcu_read_unlock_bh();
1638 goto done;
1639 }
1640 dst_release(xchg(&skb->dst, NULL));
1641 }
1642 rcu_read_unlock_bh();
1643 }
1644
1645done:
1646 cb->args[0] = h;
1647 cb->args[1] = idx;
1648 return skb->len;
1649}
1650
1651#ifdef CONFIG_PROC_FS
1652struct dn_rt_cache_iter_state {
1653 int bucket;
1654};
1655
1656static struct dn_route *dn_rt_cache_get_first(struct seq_file *seq)
1657{
1658 struct dn_route *rt = NULL;
1659 struct dn_rt_cache_iter_state *s = seq->private;
1660
1661 for(s->bucket = dn_rt_hash_mask; s->bucket >= 0; --s->bucket) {
1662 rcu_read_lock_bh();
1663 rt = dn_rt_hash_table[s->bucket].chain;
1664 if (rt)
1665 break;
1666 rcu_read_unlock_bh();
1667 }
Eric Dumazet0d89d792008-01-10 22:35:21 -08001668 return rcu_dereference(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669}
1670
1671static struct dn_route *dn_rt_cache_get_next(struct seq_file *seq, struct dn_route *rt)
1672{
Eric Dumazet0d89d792008-01-10 22:35:21 -08001673 struct dn_rt_cache_iter_state *s = seq->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674
Eric Dumazet0c195c32007-02-09 16:25:52 -08001675 rt = rt->u.dst.dn_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676 while(!rt) {
1677 rcu_read_unlock_bh();
1678 if (--s->bucket < 0)
1679 break;
1680 rcu_read_lock_bh();
1681 rt = dn_rt_hash_table[s->bucket].chain;
1682 }
Eric Dumazet0d89d792008-01-10 22:35:21 -08001683 return rcu_dereference(rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684}
1685
1686static void *dn_rt_cache_seq_start(struct seq_file *seq, loff_t *pos)
1687{
1688 struct dn_route *rt = dn_rt_cache_get_first(seq);
1689
1690 if (rt) {
1691 while(*pos && (rt = dn_rt_cache_get_next(seq, rt)))
1692 --*pos;
1693 }
1694 return *pos ? NULL : rt;
1695}
1696
1697static void *dn_rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1698{
1699 struct dn_route *rt = dn_rt_cache_get_next(seq, v);
1700 ++*pos;
1701 return rt;
1702}
1703
1704static void dn_rt_cache_seq_stop(struct seq_file *seq, void *v)
1705{
1706 if (v)
1707 rcu_read_unlock_bh();
1708}
1709
1710static int dn_rt_cache_seq_show(struct seq_file *seq, void *v)
1711{
1712 struct dn_route *rt = v;
1713 char buf1[DN_ASCBUF_LEN], buf2[DN_ASCBUF_LEN];
1714
1715 seq_printf(seq, "%-8s %-7s %-7s %04d %04d %04d\n",
1716 rt->u.dst.dev ? rt->u.dst.dev->name : "*",
1717 dn_addr2asc(dn_ntohs(rt->rt_daddr), buf1),
1718 dn_addr2asc(dn_ntohs(rt->rt_saddr), buf2),
1719 atomic_read(&rt->u.dst.__refcnt),
1720 rt->u.dst.__use,
1721 (int) dst_metric(&rt->u.dst, RTAX_RTT));
1722 return 0;
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001723}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724
Philippe De Muyter56b3d972007-07-10 23:07:31 -07001725static const struct seq_operations dn_rt_cache_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726 .start = dn_rt_cache_seq_start,
1727 .next = dn_rt_cache_seq_next,
1728 .stop = dn_rt_cache_seq_stop,
1729 .show = dn_rt_cache_seq_show,
1730};
1731
1732static int dn_rt_cache_seq_open(struct inode *inode, struct file *file)
1733{
Pavel Emelyanov31164082007-10-10 02:30:23 -07001734 return seq_open_private(file, &dn_rt_cache_seq_ops,
1735 sizeof(struct dn_rt_cache_iter_state));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736}
1737
Arjan van de Ven9a321442007-02-12 00:55:35 -08001738static const struct file_operations dn_rt_cache_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 .owner = THIS_MODULE,
1740 .open = dn_rt_cache_seq_open,
1741 .read = seq_read,
1742 .llseek = seq_lseek,
1743 .release = seq_release_private,
1744};
1745
1746#endif /* CONFIG_PROC_FS */
1747
1748void __init dn_route_init(void)
1749{
1750 int i, goal, order;
1751
Alexey Dobriyane5d679f332006-08-26 19:25:52 -07001752 dn_dst_ops.kmem_cachep =
1753 kmem_cache_create("dn_dst_cache", sizeof(struct dn_route), 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001754 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
Pavel Emelyanovb24b8a22008-01-23 21:20:07 -08001755 setup_timer(&dn_route_timer, dn_dst_check_expire, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756 dn_route_timer.expires = jiffies + decnet_dst_gc_interval * HZ;
1757 add_timer(&dn_route_timer);
1758
1759 goal = num_physpages >> (26 - PAGE_SHIFT);
1760
1761 for(order = 0; (1UL << order) < goal; order++)
1762 /* NOTHING */;
1763
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001764 /*
1765 * Only want 1024 entries max, since the table is very, very unlikely
1766 * to be larger than that.
1767 */
1768 while(order && ((((1UL << order) * PAGE_SIZE) /
1769 sizeof(struct dn_rt_hash_bucket)) >= 2048))
1770 order--;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001772 do {
1773 dn_rt_hash_mask = (1UL << order) * PAGE_SIZE /
1774 sizeof(struct dn_rt_hash_bucket);
1775 while(dn_rt_hash_mask & (dn_rt_hash_mask - 1))
1776 dn_rt_hash_mask--;
1777 dn_rt_hash_table = (struct dn_rt_hash_bucket *)
1778 __get_free_pages(GFP_ATOMIC, order);
1779 } while (dn_rt_hash_table == NULL && --order > 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780
1781 if (!dn_rt_hash_table)
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001782 panic("Failed to allocate DECnet route cache hash table\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001784 printk(KERN_INFO
1785 "DECnet: Routing cache hash table of %u buckets, %ldKbytes\n",
1786 dn_rt_hash_mask,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787 (long)(dn_rt_hash_mask*sizeof(struct dn_rt_hash_bucket))/1024);
1788
1789 dn_rt_hash_mask--;
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001790 for(i = 0; i <= dn_rt_hash_mask; i++) {
1791 spin_lock_init(&dn_rt_hash_table[i].lock);
1792 dn_rt_hash_table[i].chain = NULL;
1793 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794
YOSHIFUJI Hideaki429eb0f2007-02-09 23:24:40 +09001795 dn_dst_ops.gc_thresh = (dn_rt_hash_mask + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796
Eric W. Biederman457c4cb2007-09-12 12:01:34 +02001797 proc_net_fops_create(&init_net, "decnet_cache", S_IRUGO, &dn_rt_cache_seq_fops);
Thomas Graffa34ddd2007-03-22 11:57:46 -07001798
1799#ifdef CONFIG_DECNET_ROUTER
1800 rtnl_register(PF_DECnet, RTM_GETROUTE, dn_cache_getroute, dn_fib_dump);
1801#else
1802 rtnl_register(PF_DECnet, RTM_GETROUTE, dn_cache_getroute,
1803 dn_cache_dump);
1804#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805}
1806
1807void __exit dn_route_cleanup(void)
1808{
1809 del_timer(&dn_route_timer);
1810 dn_run_flush(0);
1811
Eric W. Biederman457c4cb2007-09-12 12:01:34 +02001812 proc_net_remove(&init_net, "decnet_cache");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813}
1814