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 Routing Functions (Endnode and Router)
8 * Authors: Steve Whitehouse <SteveW@ACM.org>
9 * Eduardo Marcelo Serrat <emserrat@geocities.com>
12 * Steve Whitehouse : Fixes to allow "intra-ethernet" and
13 * "return-to-sender" bits on outgoing
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
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
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
44 /******************************************************************************
45 (c) 1995-1998 E.M. Serrat emserrat@geocities.com
47 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
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 *******************************************************************************/
58 #include <linux/errno.h>
59 #include <linux/types.h>
60 #include <linux/socket.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 <linux/slab.h>
72 #include <linux/proc_fs.h>
73 #include <linux/seq_file.h>
74 #include <linux/init.h>
75 #include <linux/rtnetlink.h>
76 #include <linux/string.h>
77 #include <linux/netfilter_decnet.h>
78 #include <linux/rcupdate.h>
79 #include <linux/times.h>
80 #include <linux/export.h>
81 #include <asm/errno.h>
82 #include <net/net_namespace.h>
83 #include <net/netlink.h>
84 #include <net/neighbour.h>
87 #include <net/fib_rules.h>
89 #include <net/dn_dev.h>
90 #include <net/dn_nsp.h>
91 #include <net/dn_route.h>
92 #include <net/dn_neigh.h>
93 #include <net/dn_fib.h>
95 struct dn_rt_hash_bucket
97 struct dn_route __rcu *chain;
101 extern struct neigh_table dn_neigh_table;
104 static unsigned char dn_hiord_addr[6] = {0xAA,0x00,0x04,0x00,0x00,0x00};
106 static const int dn_rt_min_delay = 2 * HZ;
107 static const int dn_rt_max_delay = 10 * HZ;
108 static const int dn_rt_mtu_expires = 10 * 60 * HZ;
110 static unsigned long dn_rt_deadline;
112 static int dn_dst_gc(struct dst_ops *ops);
113 static struct dst_entry *dn_dst_check(struct dst_entry *, __u32);
114 static unsigned int dn_dst_default_advmss(const struct dst_entry *dst);
115 static unsigned int dn_dst_mtu(const struct dst_entry *dst);
116 static void dn_dst_destroy(struct dst_entry *);
117 static struct dst_entry *dn_dst_negative_advice(struct dst_entry *);
118 static void dn_dst_link_failure(struct sk_buff *);
119 static void dn_dst_update_pmtu(struct dst_entry *dst, u32 mtu);
120 static struct neighbour *dn_dst_neigh_lookup(const struct dst_entry *dst, const void *daddr);
121 static int dn_route_input(struct sk_buff *);
122 static void dn_run_flush(unsigned long dummy);
124 static struct dn_rt_hash_bucket *dn_rt_hash_table;
125 static unsigned dn_rt_hash_mask;
127 static struct timer_list dn_route_timer;
128 static DEFINE_TIMER(dn_rt_flush_timer, dn_run_flush, 0, 0);
129 int decnet_dst_gc_interval = 2;
131 static struct dst_ops dn_dst_ops = {
133 .protocol = cpu_to_be16(ETH_P_DNA_RT),
136 .check = dn_dst_check,
137 .default_advmss = dn_dst_default_advmss,
139 .cow_metrics = dst_cow_metrics_generic,
140 .destroy = dn_dst_destroy,
141 .negative_advice = dn_dst_negative_advice,
142 .link_failure = dn_dst_link_failure,
143 .update_pmtu = dn_dst_update_pmtu,
144 .neigh_lookup = dn_dst_neigh_lookup,
147 static void dn_dst_destroy(struct dst_entry *dst)
149 dst_destroy_metrics_generic(dst);
152 static __inline__ unsigned dn_hash(__le16 src, __le16 dst)
154 __u16 tmp = (__u16 __force)(src ^ dst);
158 return dn_rt_hash_mask & (unsigned)tmp;
161 static inline void dnrt_free(struct dn_route *rt)
163 call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free);
166 static inline void dnrt_drop(struct dn_route *rt)
168 dst_release(&rt->dst);
169 call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free);
172 static void dn_dst_check_expire(unsigned long dummy)
176 struct dn_route __rcu **rtp;
177 unsigned long now = jiffies;
178 unsigned long expire = 120 * HZ;
180 for (i = 0; i <= dn_rt_hash_mask; i++) {
181 rtp = &dn_rt_hash_table[i].chain;
183 spin_lock(&dn_rt_hash_table[i].lock);
184 while ((rt = rcu_dereference_protected(*rtp,
185 lockdep_is_held(&dn_rt_hash_table[i].lock))) != NULL) {
186 if (atomic_read(&rt->dst.__refcnt) ||
187 (now - rt->dst.lastuse) < expire) {
188 rtp = &rt->dst.dn_next;
191 *rtp = rt->dst.dn_next;
192 rt->dst.dn_next = NULL;
195 spin_unlock(&dn_rt_hash_table[i].lock);
197 if ((jiffies - now) > 0)
201 mod_timer(&dn_route_timer, now + decnet_dst_gc_interval * HZ);
204 static int dn_dst_gc(struct dst_ops *ops)
207 struct dn_route __rcu **rtp;
209 unsigned long now = jiffies;
210 unsigned long expire = 10 * HZ;
212 for (i = 0; i <= dn_rt_hash_mask; i++) {
214 spin_lock_bh(&dn_rt_hash_table[i].lock);
215 rtp = &dn_rt_hash_table[i].chain;
217 while ((rt = rcu_dereference_protected(*rtp,
218 lockdep_is_held(&dn_rt_hash_table[i].lock))) != NULL) {
219 if (atomic_read(&rt->dst.__refcnt) ||
220 (now - rt->dst.lastuse) < expire) {
221 rtp = &rt->dst.dn_next;
224 *rtp = rt->dst.dn_next;
225 rt->dst.dn_next = NULL;
229 spin_unlock_bh(&dn_rt_hash_table[i].lock);
236 * The decnet standards don't impose a particular minimum mtu, what they
237 * do insist on is that the routing layer accepts a datagram of at least
238 * 230 bytes long. Here we have to subtract the routing header length from
239 * 230 to get the minimum acceptable mtu. If there is no neighbour, then we
240 * assume the worst and use a long header size.
242 * We update both the mtu and the advertised mss (i.e. the segment size we
243 * advertise to the other end).
245 static void dn_dst_update_pmtu(struct dst_entry *dst, u32 mtu)
247 struct neighbour *n = dst_get_neighbour(dst);
251 dn = n ? rcu_dereference_raw(n->dev->dn_ptr) : NULL;
253 if (dn && dn->use_long == 0)
258 if (dst_metric(dst, RTAX_MTU) > mtu && mtu >= min_mtu) {
259 if (!(dst_metric_locked(dst, RTAX_MTU))) {
260 dst_metric_set(dst, RTAX_MTU, mtu);
261 dst_set_expires(dst, dn_rt_mtu_expires);
263 if (!(dst_metric_locked(dst, RTAX_ADVMSS))) {
264 u32 mss = mtu - DN_MAX_NSP_DATA_HEADER;
265 u32 existing_mss = dst_metric_raw(dst, RTAX_ADVMSS);
266 if (!existing_mss || existing_mss > mss)
267 dst_metric_set(dst, RTAX_ADVMSS, mss);
273 * When a route has been marked obsolete. (e.g. routing cache flush)
275 static struct dst_entry *dn_dst_check(struct dst_entry *dst, __u32 cookie)
280 static struct dst_entry *dn_dst_negative_advice(struct dst_entry *dst)
286 static void dn_dst_link_failure(struct sk_buff *skb)
290 static inline int compare_keys(struct flowidn *fl1, struct flowidn *fl2)
292 return ((fl1->daddr ^ fl2->daddr) |
293 (fl1->saddr ^ fl2->saddr) |
294 (fl1->flowidn_mark ^ fl2->flowidn_mark) |
295 (fl1->flowidn_scope ^ fl2->flowidn_scope) |
296 (fl1->flowidn_oif ^ fl2->flowidn_oif) |
297 (fl1->flowidn_iif ^ fl2->flowidn_iif)) == 0;
300 static int dn_insert_route(struct dn_route *rt, unsigned hash, struct dn_route **rp)
302 struct dn_route *rth;
303 struct dn_route __rcu **rthp;
304 unsigned long now = jiffies;
306 rthp = &dn_rt_hash_table[hash].chain;
308 spin_lock_bh(&dn_rt_hash_table[hash].lock);
309 while ((rth = rcu_dereference_protected(*rthp,
310 lockdep_is_held(&dn_rt_hash_table[hash].lock))) != NULL) {
311 if (compare_keys(&rth->fld, &rt->fld)) {
313 *rthp = rth->dst.dn_next;
314 rcu_assign_pointer(rth->dst.dn_next,
315 dn_rt_hash_table[hash].chain);
316 rcu_assign_pointer(dn_rt_hash_table[hash].chain, rth);
318 dst_use(&rth->dst, now);
319 spin_unlock_bh(&dn_rt_hash_table[hash].lock);
325 rthp = &rth->dst.dn_next;
328 rcu_assign_pointer(rt->dst.dn_next, dn_rt_hash_table[hash].chain);
329 rcu_assign_pointer(dn_rt_hash_table[hash].chain, rt);
331 dst_use(&rt->dst, now);
332 spin_unlock_bh(&dn_rt_hash_table[hash].lock);
337 static void dn_run_flush(unsigned long dummy)
340 struct dn_route *rt, *next;
342 for (i = 0; i < dn_rt_hash_mask; i++) {
343 spin_lock_bh(&dn_rt_hash_table[i].lock);
345 if ((rt = xchg((struct dn_route **)&dn_rt_hash_table[i].chain, NULL)) == NULL)
346 goto nothing_to_declare;
348 for(; rt; rt = next) {
349 next = rcu_dereference_raw(rt->dst.dn_next);
350 RCU_INIT_POINTER(rt->dst.dn_next, NULL);
351 dst_free((struct dst_entry *)rt);
355 spin_unlock_bh(&dn_rt_hash_table[i].lock);
359 static DEFINE_SPINLOCK(dn_rt_flush_lock);
361 void dn_rt_cache_flush(int delay)
363 unsigned long now = jiffies;
364 int user_mode = !in_interrupt();
367 delay = dn_rt_min_delay;
369 spin_lock_bh(&dn_rt_flush_lock);
371 if (del_timer(&dn_rt_flush_timer) && delay > 0 && dn_rt_deadline) {
372 long tmo = (long)(dn_rt_deadline - now);
374 if (user_mode && tmo < dn_rt_max_delay - dn_rt_min_delay)
382 spin_unlock_bh(&dn_rt_flush_lock);
387 if (dn_rt_deadline == 0)
388 dn_rt_deadline = now + dn_rt_max_delay;
390 dn_rt_flush_timer.expires = now + delay;
391 add_timer(&dn_rt_flush_timer);
392 spin_unlock_bh(&dn_rt_flush_lock);
396 * dn_return_short - Return a short packet to its sender
397 * @skb: The packet to return
400 static int dn_return_short(struct sk_buff *skb)
402 struct dn_skb_cb *cb;
407 /* Add back headers */
408 skb_push(skb, skb->data - skb_network_header(skb));
410 if ((skb = skb_unshare(skb, GFP_ATOMIC)) == NULL)
414 /* Skip packet length and point to flags */
416 *ptr++ = (cb->rt_flags & ~DN_RT_F_RQR) | DN_RT_F_RTS;
422 *ptr = 0; /* Zero hop count */
426 skb->pkt_type = PACKET_OUTGOING;
427 dn_rt_finish_output(skb, NULL, NULL);
428 return NET_RX_SUCCESS;
432 * dn_return_long - Return a long packet to its sender
433 * @skb: The long format packet to return
436 static int dn_return_long(struct sk_buff *skb)
438 struct dn_skb_cb *cb;
440 unsigned char *src_addr, *dst_addr;
441 unsigned char tmp[ETH_ALEN];
443 /* Add back all headers */
444 skb_push(skb, skb->data - skb_network_header(skb));
446 if ((skb = skb_unshare(skb, GFP_ATOMIC)) == NULL)
450 /* Ignore packet length and point to flags */
454 if (*ptr & DN_RT_F_PF) {
455 char padlen = (*ptr & ~DN_RT_F_PF);
459 *ptr++ = (cb->rt_flags & ~DN_RT_F_RQR) | DN_RT_F_RTS;
465 *ptr = 0; /* Zero hop count */
467 /* Swap source and destination */
468 memcpy(tmp, src_addr, ETH_ALEN);
469 memcpy(src_addr, dst_addr, ETH_ALEN);
470 memcpy(dst_addr, tmp, ETH_ALEN);
472 skb->pkt_type = PACKET_OUTGOING;
473 dn_rt_finish_output(skb, dst_addr, src_addr);
474 return NET_RX_SUCCESS;
478 * dn_route_rx_packet - Try and find a route for an incoming packet
479 * @skb: The packet to find a route for
481 * Returns: result of input function if route is found, error code otherwise
483 static int dn_route_rx_packet(struct sk_buff *skb)
485 struct dn_skb_cb *cb;
488 if ((err = dn_route_input(skb)) == 0)
489 return dst_input(skb);
492 if (decnet_debug_level & 4) {
493 char *devname = skb->dev ? skb->dev->name : "???";
496 "DECnet: dn_route_rx_packet: rt_flags=0x%02x dev=%s len=%d src=0x%04hx dst=0x%04hx err=%d type=%d\n",
497 (int)cb->rt_flags, devname, skb->len,
498 le16_to_cpu(cb->src), le16_to_cpu(cb->dst),
502 if ((skb->pkt_type == PACKET_HOST) && (cb->rt_flags & DN_RT_F_RQR)) {
503 switch (cb->rt_flags & DN_RT_PKT_MSK) {
504 case DN_RT_PKT_SHORT:
505 return dn_return_short(skb);
507 return dn_return_long(skb);
515 static int dn_route_rx_long(struct sk_buff *skb)
517 struct dn_skb_cb *cb = DN_SKB_CB(skb);
518 unsigned char *ptr = skb->data;
520 if (!pskb_may_pull(skb, 21)) /* 20 for long header, 1 for shortest nsp */
524 skb_reset_transport_header(skb);
526 /* Destination info */
528 cb->dst = dn_eth2dn(ptr);
529 if (memcmp(ptr, dn_hiord_addr, 4) != 0)
536 cb->src = dn_eth2dn(ptr);
537 if (memcmp(ptr, dn_hiord_addr, 4) != 0)
542 cb->hops = *ptr++; /* Visit Count */
544 return NF_HOOK(NFPROTO_DECNET, NF_DN_PRE_ROUTING, skb, skb->dev, NULL,
554 static int dn_route_rx_short(struct sk_buff *skb)
556 struct dn_skb_cb *cb = DN_SKB_CB(skb);
557 unsigned char *ptr = skb->data;
559 if (!pskb_may_pull(skb, 6)) /* 5 for short header + 1 for shortest nsp */
563 skb_reset_transport_header(skb);
565 cb->dst = *(__le16 *)ptr;
567 cb->src = *(__le16 *)ptr;
569 cb->hops = *ptr & 0x3f;
571 return NF_HOOK(NFPROTO_DECNET, NF_DN_PRE_ROUTING, skb, skb->dev, NULL,
579 static int dn_route_discard(struct sk_buff *skb)
582 * I know we drop the packet here, but thats considered success in
586 return NET_RX_SUCCESS;
589 static int dn_route_ptp_hello(struct sk_buff *skb)
592 dn_neigh_pointopoint_hello(skb);
593 return NET_RX_SUCCESS;
596 int dn_route_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
598 struct dn_skb_cb *cb;
599 unsigned char flags = 0;
600 __u16 len = le16_to_cpu(*(__le16 *)skb->data);
601 struct dn_dev *dn = rcu_dereference(dev->dn_ptr);
602 unsigned char padlen = 0;
604 if (!net_eq(dev_net(dev), &init_net))
610 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL)
613 if (!pskb_may_pull(skb, 3))
627 cb->iif = dev->ifindex;
630 * If we have padding, remove it.
632 if (flags & DN_RT_F_PF) {
633 padlen = flags & ~DN_RT_F_PF;
634 if (!pskb_may_pull(skb, padlen + 1))
636 skb_pull(skb, padlen);
640 skb_reset_network_header(skb);
643 * Weed out future version DECnet
645 if (flags & DN_RT_F_VER)
648 cb->rt_flags = flags;
650 if (decnet_debug_level & 1)
652 "dn_route_rcv: got 0x%02x from %s [%d %d %d]\n",
653 (int)flags, (dev) ? dev->name : "???", len, skb->len,
656 if (flags & DN_RT_PKT_CNTL) {
657 if (unlikely(skb_linearize(skb)))
660 switch (flags & DN_RT_CNTL_MSK) {
662 dn_dev_init_pkt(skb);
665 dn_dev_veri_pkt(skb);
669 if (dn->parms.state != DN_DEV_S_RU)
672 switch (flags & DN_RT_CNTL_MSK) {
674 return NF_HOOK(NFPROTO_DECNET, NF_DN_HELLO,
680 return NF_HOOK(NFPROTO_DECNET, NF_DN_ROUTE,
684 return NF_HOOK(NFPROTO_DECNET, NF_DN_HELLO,
686 dn_neigh_router_hello);
689 return NF_HOOK(NFPROTO_DECNET, NF_DN_HELLO,
691 dn_neigh_endnode_hello);
694 if (dn->parms.state != DN_DEV_S_RU)
697 skb_pull(skb, 1); /* Pull flags */
699 switch (flags & DN_RT_PKT_MSK) {
701 return dn_route_rx_long(skb);
702 case DN_RT_PKT_SHORT:
703 return dn_route_rx_short(skb);
713 static int dn_to_neigh_output(struct sk_buff *skb)
715 struct dst_entry *dst = skb_dst(skb);
716 struct neighbour *n = dst_get_neighbour(dst);
718 return n->output(n, skb);
721 static int dn_output(struct sk_buff *skb)
723 struct dst_entry *dst = skb_dst(skb);
724 struct dn_route *rt = (struct dn_route *)dst;
725 struct net_device *dev = dst->dev;
726 struct dn_skb_cb *cb = DN_SKB_CB(skb);
727 struct neighbour *neigh;
731 if ((neigh = dst_get_neighbour(dst)) == NULL)
736 cb->src = rt->rt_saddr;
737 cb->dst = rt->rt_daddr;
740 * Always set the Intra-Ethernet bit on all outgoing packets
741 * originated on this node. Only valid flag from upper layers
742 * is return-to-sender-requested. Set hop count to 0 too.
744 cb->rt_flags &= ~DN_RT_F_RQR;
745 cb->rt_flags |= DN_RT_F_IE;
748 return NF_HOOK(NFPROTO_DECNET, NF_DN_LOCAL_OUT, skb, NULL, dev,
753 printk(KERN_DEBUG "dn_output: This should not happen\n");
760 static int dn_forward(struct sk_buff *skb)
762 struct dn_skb_cb *cb = DN_SKB_CB(skb);
763 struct dst_entry *dst = skb_dst(skb);
764 struct dn_dev *dn_db = rcu_dereference(dst->dev->dn_ptr);
767 #ifdef CONFIG_NETFILTER
768 struct net_device *dev = skb->dev;
771 if (skb->pkt_type != PACKET_HOST)
774 /* Ensure that we have enough space for headers */
775 rt = (struct dn_route *)skb_dst(skb);
776 header_len = dn_db->use_long ? 21 : 6;
777 if (skb_cow(skb, LL_RESERVED_SPACE(rt->dst.dev)+header_len))
781 * Hop count exceeded.
786 skb->dev = rt->dst.dev;
789 * If packet goes out same interface it came in on, then set
790 * the Intra-Ethernet bit. This has no effect for short
791 * packets, so we don't need to test for them here.
793 cb->rt_flags &= ~DN_RT_F_IE;
794 if (rt->rt_flags & RTCF_DOREDIRECT)
795 cb->rt_flags |= DN_RT_F_IE;
797 return NF_HOOK(NFPROTO_DECNET, NF_DN_FORWARD, skb, dev, skb->dev,
806 * Used to catch bugs. This should never normally get
809 static int dn_rt_bug(struct sk_buff *skb)
811 if (net_ratelimit()) {
812 struct dn_skb_cb *cb = DN_SKB_CB(skb);
814 printk(KERN_DEBUG "dn_rt_bug: skb from:%04x to:%04x\n",
815 le16_to_cpu(cb->src), le16_to_cpu(cb->dst));
823 static unsigned int dn_dst_default_advmss(const struct dst_entry *dst)
825 return dn_mss_from_pmtu(dst->dev, dst_mtu(dst));
828 static unsigned int dn_dst_mtu(const struct dst_entry *dst)
830 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
832 return mtu ? : dst->dev->mtu;
835 static struct neighbour *dn_dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
837 return __neigh_lookup_errno(&dn_neigh_table, daddr, dst->dev);
840 static int dn_rt_set_next_hop(struct dn_route *rt, struct dn_fib_res *res)
842 struct dn_fib_info *fi = res->fi;
843 struct net_device *dev = rt->dst.dev;
844 unsigned int mss_metric;
848 if (DN_FIB_RES_GW(*res) &&
849 DN_FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK)
850 rt->rt_gateway = DN_FIB_RES_GW(*res);
851 dst_init_metrics(&rt->dst, fi->fib_metrics, true);
853 rt->rt_type = res->type;
855 if (dev != NULL && dst_get_neighbour(&rt->dst) == NULL) {
856 n = __neigh_lookup_errno(&dn_neigh_table, &rt->rt_gateway, dev);
859 dst_set_neighbour(&rt->dst, n);
862 if (dst_metric(&rt->dst, RTAX_MTU) > rt->dst.dev->mtu)
863 dst_metric_set(&rt->dst, RTAX_MTU, rt->dst.dev->mtu);
864 mss_metric = dst_metric_raw(&rt->dst, RTAX_ADVMSS);
866 unsigned int mss = dn_mss_from_pmtu(dev, dst_mtu(&rt->dst));
867 if (mss_metric > mss)
868 dst_metric_set(&rt->dst, RTAX_ADVMSS, mss);
873 static inline int dn_match_addr(__le16 addr1, __le16 addr2)
875 __u16 tmp = le16_to_cpu(addr1) ^ le16_to_cpu(addr2);
884 static __le16 dnet_select_source(const struct net_device *dev, __le16 daddr, int scope)
887 struct dn_dev *dn_db;
888 struct dn_ifaddr *ifa;
893 dn_db = rcu_dereference(dev->dn_ptr);
894 for (ifa = rcu_dereference(dn_db->ifa_list);
896 ifa = rcu_dereference(ifa->ifa_next)) {
897 if (ifa->ifa_scope > scope)
900 saddr = ifa->ifa_local;
903 ret = dn_match_addr(daddr, ifa->ifa_local);
904 if (ret > best_match)
905 saddr = ifa->ifa_local;
907 saddr = ifa->ifa_local;
914 static inline __le16 __dn_fib_res_prefsrc(struct dn_fib_res *res)
916 return dnet_select_source(DN_FIB_RES_DEV(*res), DN_FIB_RES_GW(*res), res->scope);
919 static inline __le16 dn_fib_rules_map_destination(__le16 daddr, struct dn_fib_res *res)
921 __le16 mask = dnet_make_mask(res->prefixlen);
922 return (daddr&~mask)|res->fi->fib_nh->nh_gw;
925 static int dn_route_output_slow(struct dst_entry **pprt, const struct flowidn *oldflp, int try_hard)
927 struct flowidn fld = {
928 .daddr = oldflp->daddr,
929 .saddr = oldflp->saddr,
930 .flowidn_scope = RT_SCOPE_UNIVERSE,
931 .flowidn_mark = oldflp->flowidn_mark,
932 .flowidn_iif = init_net.loopback_dev->ifindex,
933 .flowidn_oif = oldflp->flowidn_oif,
935 struct dn_route *rt = NULL;
936 struct net_device *dev_out = NULL, *dev;
937 struct neighbour *neigh = NULL;
940 struct dn_fib_res res = { .fi = NULL, .type = RTN_UNICAST };
945 if (decnet_debug_level & 16)
947 "dn_route_output_slow: dst=%04x src=%04x mark=%d"
948 " iif=%d oif=%d\n", le16_to_cpu(oldflp->daddr),
949 le16_to_cpu(oldflp->saddr),
950 oldflp->flowidn_mark, init_net.loopback_dev->ifindex,
951 oldflp->flowidn_oif);
953 /* If we have an output interface, verify its a DECnet device */
954 if (oldflp->flowidn_oif) {
955 dev_out = dev_get_by_index(&init_net, oldflp->flowidn_oif);
957 if (dev_out && dev_out->dn_ptr == NULL) {
965 /* If we have a source address, verify that its a local address */
967 err = -EADDRNOTAVAIL;
970 if (dn_dev_islocal(dev_out, oldflp->saddr))
976 for_each_netdev_rcu(&init_net, dev) {
979 if (!dn_dev_islocal(dev, oldflp->saddr))
981 if ((dev->flags & IFF_LOOPBACK) &&
983 !dn_dev_islocal(dev, oldflp->daddr))
997 /* No destination? Assume its local */
999 fld.daddr = fld.saddr;
1004 dev_out = init_net.loopback_dev;
1005 if (!dev_out->dn_ptr)
1007 err = -EADDRNOTAVAIL;
1011 fld.saddr = dnet_select_source(dev_out, 0,
1016 fld.flowidn_oif = init_net.loopback_dev->ifindex;
1017 res.type = RTN_LOCAL;
1021 if (decnet_debug_level & 16)
1023 "dn_route_output_slow: initial checks complete."
1024 " dst=%o4x src=%04x oif=%d try_hard=%d\n",
1025 le16_to_cpu(fld.daddr), le16_to_cpu(fld.saddr),
1026 fld.flowidn_oif, try_hard);
1029 * N.B. If the kernel is compiled without router support then
1030 * dn_fib_lookup() will evaluate to non-zero so this if () block
1031 * will always be executed.
1034 if (try_hard || (err = dn_fib_lookup(&fld, &res)) != 0) {
1035 struct dn_dev *dn_db;
1039 * Here the fallback is basically the standard algorithm for
1040 * routing in endnodes which is described in the DECnet routing
1043 * If we are not trying hard, look in neighbour cache.
1044 * The result is tested to ensure that if a specific output
1045 * device/source address was requested, then we honour that
1049 neigh = neigh_lookup_nodev(&dn_neigh_table, &init_net, &fld.daddr);
1051 if ((oldflp->flowidn_oif &&
1052 (neigh->dev->ifindex != oldflp->flowidn_oif)) ||
1054 (!dn_dev_islocal(neigh->dev,
1056 neigh_release(neigh);
1061 if (dn_dev_islocal(neigh->dev, fld.daddr)) {
1062 dev_out = init_net.loopback_dev;
1063 res.type = RTN_LOCAL;
1065 dev_out = neigh->dev;
1073 /* Not there? Perhaps its a local address */
1074 if (dev_out == NULL)
1075 dev_out = dn_dev_get_default();
1077 if (dev_out == NULL)
1079 dn_db = rcu_dereference_raw(dev_out->dn_ptr);
1082 /* Possible improvement - check all devices for local addr */
1083 if (dn_dev_islocal(dev_out, fld.daddr)) {
1085 dev_out = init_net.loopback_dev;
1087 res.type = RTN_LOCAL;
1090 /* Not local either.... try sending it to the default router */
1091 neigh = neigh_clone(dn_db->router);
1092 BUG_ON(neigh && neigh->dev != dev_out);
1094 /* Ok then, we assume its directly connected and move on */
1097 gateway = ((struct dn_neigh *)neigh)->addr;
1099 gateway = fld.daddr;
1100 if (fld.saddr == 0) {
1101 fld.saddr = dnet_select_source(dev_out, gateway,
1102 res.type == RTN_LOCAL ?
1105 if (fld.saddr == 0 && res.type != RTN_LOCAL)
1108 fld.flowidn_oif = dev_out->ifindex;
1113 if (res.type == RTN_NAT)
1116 if (res.type == RTN_LOCAL) {
1118 fld.saddr = fld.daddr;
1121 dev_out = init_net.loopback_dev;
1123 if (!dev_out->dn_ptr)
1125 fld.flowidn_oif = dev_out->ifindex;
1127 dn_fib_info_put(res.fi);
1132 if (res.fi->fib_nhs > 1 && fld.flowidn_oif == 0)
1133 dn_fib_select_multipath(&fld, &res);
1136 * We could add some logic to deal with default routes here and
1137 * get rid of some of the special casing above.
1141 fld.saddr = DN_FIB_RES_PREFSRC(res);
1145 dev_out = DN_FIB_RES_DEV(res);
1147 fld.flowidn_oif = dev_out->ifindex;
1148 gateway = DN_FIB_RES_GW(res);
1151 if (dev_out->flags & IFF_LOOPBACK)
1152 flags |= RTCF_LOCAL;
1154 rt = dst_alloc(&dn_dst_ops, dev_out, 1, 0, DST_HOST);
1158 memset(&rt->fld, 0, sizeof(rt->fld));
1159 rt->fld.saddr = oldflp->saddr;
1160 rt->fld.daddr = oldflp->daddr;
1161 rt->fld.flowidn_oif = oldflp->flowidn_oif;
1162 rt->fld.flowidn_iif = 0;
1163 rt->fld.flowidn_mark = oldflp->flowidn_mark;
1165 rt->rt_saddr = fld.saddr;
1166 rt->rt_daddr = fld.daddr;
1167 rt->rt_gateway = gateway ? gateway : fld.daddr;
1168 rt->rt_local_src = fld.saddr;
1170 rt->rt_dst_map = fld.daddr;
1171 rt->rt_src_map = fld.saddr;
1173 dst_set_neighbour(&rt->dst, neigh);
1176 rt->dst.lastuse = jiffies;
1177 rt->dst.output = dn_output;
1178 rt->dst.input = dn_rt_bug;
1179 rt->rt_flags = flags;
1180 if (flags & RTCF_LOCAL)
1181 rt->dst.input = dn_nsp_rx;
1183 err = dn_rt_set_next_hop(rt, &res);
1187 hash = dn_hash(rt->fld.saddr, rt->fld.daddr);
1188 dn_insert_route(rt, hash, (struct dn_route **)pprt);
1192 neigh_release(neigh);
1194 dn_fib_res_put(&res);
1201 err = -EADDRNOTAVAIL;
1216 * N.B. The flags may be moved into the flowi at some future stage.
1218 static int __dn_route_output_key(struct dst_entry **pprt, const struct flowidn *flp, int flags)
1220 unsigned hash = dn_hash(flp->saddr, flp->daddr);
1221 struct dn_route *rt = NULL;
1223 if (!(flags & MSG_TRYHARD)) {
1225 for (rt = rcu_dereference_bh(dn_rt_hash_table[hash].chain); rt;
1226 rt = rcu_dereference_bh(rt->dst.dn_next)) {
1227 if ((flp->daddr == rt->fld.daddr) &&
1228 (flp->saddr == rt->fld.saddr) &&
1229 (flp->flowidn_mark == rt->fld.flowidn_mark) &&
1230 dn_is_output_route(rt) &&
1231 (rt->fld.flowidn_oif == flp->flowidn_oif)) {
1232 dst_use(&rt->dst, jiffies);
1233 rcu_read_unlock_bh();
1238 rcu_read_unlock_bh();
1241 return dn_route_output_slow(pprt, flp, flags);
1244 static int dn_route_output_key(struct dst_entry **pprt, struct flowidn *flp, int flags)
1248 err = __dn_route_output_key(pprt, flp, flags);
1249 if (err == 0 && flp->flowidn_proto) {
1250 *pprt = xfrm_lookup(&init_net, *pprt,
1251 flowidn_to_flowi(flp), NULL, 0);
1252 if (IS_ERR(*pprt)) {
1253 err = PTR_ERR(*pprt);
1260 int dn_route_output_sock(struct dst_entry **pprt, struct flowidn *fl, struct sock *sk, int flags)
1264 err = __dn_route_output_key(pprt, fl, flags & MSG_TRYHARD);
1265 if (err == 0 && fl->flowidn_proto) {
1266 if (!(flags & MSG_DONTWAIT))
1267 fl->flowidn_flags |= FLOWI_FLAG_CAN_SLEEP;
1268 *pprt = xfrm_lookup(&init_net, *pprt,
1269 flowidn_to_flowi(fl), sk, 0);
1270 if (IS_ERR(*pprt)) {
1271 err = PTR_ERR(*pprt);
1278 static int dn_route_input_slow(struct sk_buff *skb)
1280 struct dn_route *rt = NULL;
1281 struct dn_skb_cb *cb = DN_SKB_CB(skb);
1282 struct net_device *in_dev = skb->dev;
1283 struct net_device *out_dev = NULL;
1284 struct dn_dev *dn_db;
1285 struct neighbour *neigh = NULL;
1289 __le16 local_src = 0;
1290 struct flowidn fld = {
1293 .flowidn_scope = RT_SCOPE_UNIVERSE,
1294 .flowidn_mark = skb->mark,
1295 .flowidn_iif = skb->dev->ifindex,
1297 struct dn_fib_res res = { .fi = NULL, .type = RTN_UNREACHABLE };
1303 if ((dn_db = rcu_dereference(in_dev->dn_ptr)) == NULL)
1306 /* Zero source addresses are not allowed */
1311 * In this case we've just received a packet from a source
1312 * outside ourselves pretending to come from us. We don't
1313 * allow it any further to prevent routing loops, spoofing and
1314 * other nasties. Loopback packets already have the dst attached
1315 * so this only affects packets which have originated elsewhere.
1318 if (dn_dev_islocal(in_dev, cb->src))
1321 err = dn_fib_lookup(&fld, &res);
1326 * Is the destination us ?
1328 if (!dn_dev_islocal(in_dev, cb->dst))
1331 res.type = RTN_LOCAL;
1333 __le16 src_map = fld.saddr;
1336 out_dev = DN_FIB_RES_DEV(res);
1337 if (out_dev == NULL) {
1338 if (net_ratelimit())
1339 printk(KERN_CRIT "Bug in dn_route_input_slow() "
1340 "No output device\n");
1346 src_map = fld.saddr; /* no NAT support for now */
1348 gateway = DN_FIB_RES_GW(res);
1349 if (res.type == RTN_NAT) {
1350 fld.daddr = dn_fib_rules_map_destination(fld.daddr, &res);
1351 dn_fib_res_put(&res);
1353 if (dn_fib_lookup(&fld, &res))
1356 if (res.type != RTN_UNICAST)
1359 gateway = fld.daddr;
1361 fld.saddr = src_map;
1367 * Forwarding check here, we only check for forwarding
1368 * being turned off, if you want to only forward intra
1369 * area, its up to you to set the routing tables up
1372 if (dn_db->parms.forwarding == 0)
1375 if (res.fi->fib_nhs > 1 && fld.flowidn_oif == 0)
1376 dn_fib_select_multipath(&fld, &res);
1379 * Check for out_dev == in_dev. We use the RTCF_DOREDIRECT
1380 * flag as a hint to set the intra-ethernet bit when
1381 * forwarding. If we've got NAT in operation, we don't do
1382 * this optimisation.
1384 if (out_dev == in_dev && !(flags & RTCF_NAT))
1385 flags |= RTCF_DOREDIRECT;
1387 local_src = DN_FIB_RES_PREFSRC(res);
1390 case RTN_UNREACHABLE:
1393 flags |= RTCF_LOCAL;
1394 fld.saddr = cb->dst;
1395 fld.daddr = cb->src;
1397 /* Routing tables gave us a gateway */
1401 /* Packet was intra-ethernet, so we know its on-link */
1402 if (cb->rt_flags & DN_RT_F_IE) {
1404 flags |= RTCF_DIRECTSRC;
1408 /* Use the default router if there is one */
1409 neigh = neigh_clone(dn_db->router);
1411 gateway = ((struct dn_neigh *)neigh)->addr;
1415 /* Close eyes and pray */
1417 flags |= RTCF_DIRECTSRC;
1424 rt = dst_alloc(&dn_dst_ops, out_dev, 0, 0, DST_HOST);
1428 memset(&rt->fld, 0, sizeof(rt->fld));
1429 rt->rt_saddr = fld.saddr;
1430 rt->rt_daddr = fld.daddr;
1431 rt->rt_gateway = fld.daddr;
1433 rt->rt_gateway = gateway;
1434 rt->rt_local_src = local_src ? local_src : rt->rt_saddr;
1436 rt->rt_dst_map = fld.daddr;
1437 rt->rt_src_map = fld.saddr;
1439 rt->fld.saddr = cb->src;
1440 rt->fld.daddr = cb->dst;
1441 rt->fld.flowidn_oif = 0;
1442 rt->fld.flowidn_iif = in_dev->ifindex;
1443 rt->fld.flowidn_mark = fld.flowidn_mark;
1445 dst_set_neighbour(&rt->dst, neigh);
1446 rt->dst.lastuse = jiffies;
1447 rt->dst.output = dn_rt_bug;
1450 rt->dst.input = dn_forward;
1453 rt->dst.output = dn_output;
1454 rt->dst.input = dn_nsp_rx;
1455 rt->dst.dev = in_dev;
1456 flags |= RTCF_LOCAL;
1459 case RTN_UNREACHABLE:
1461 rt->dst.input = dst_discard;
1463 rt->rt_flags = flags;
1465 err = dn_rt_set_next_hop(rt, &res);
1469 hash = dn_hash(rt->fld.saddr, rt->fld.daddr);
1470 dn_insert_route(rt, hash, &rt);
1471 skb_dst_set(skb, &rt->dst);
1475 neigh_release(neigh);
1477 dn_fib_res_put(&res);
1497 static int dn_route_input(struct sk_buff *skb)
1499 struct dn_route *rt;
1500 struct dn_skb_cb *cb = DN_SKB_CB(skb);
1501 unsigned hash = dn_hash(cb->src, cb->dst);
1507 for(rt = rcu_dereference(dn_rt_hash_table[hash].chain); rt != NULL;
1508 rt = rcu_dereference(rt->dst.dn_next)) {
1509 if ((rt->fld.saddr == cb->src) &&
1510 (rt->fld.daddr == cb->dst) &&
1511 (rt->fld.flowidn_oif == 0) &&
1512 (rt->fld.flowidn_mark == skb->mark) &&
1513 (rt->fld.flowidn_iif == cb->iif)) {
1514 dst_use(&rt->dst, jiffies);
1516 skb_dst_set(skb, (struct dst_entry *)rt);
1522 return dn_route_input_slow(skb);
1525 static int dn_rt_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1526 int event, int nowait, unsigned int flags)
1528 struct dn_route *rt = (struct dn_route *)skb_dst(skb);
1530 struct nlmsghdr *nlh;
1531 unsigned char *b = skb_tail_pointer(skb);
1534 nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*r), flags);
1535 r = NLMSG_DATA(nlh);
1536 r->rtm_family = AF_DECnet;
1537 r->rtm_dst_len = 16;
1540 r->rtm_table = RT_TABLE_MAIN;
1541 RTA_PUT_U32(skb, RTA_TABLE, RT_TABLE_MAIN);
1542 r->rtm_type = rt->rt_type;
1543 r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED;
1544 r->rtm_scope = RT_SCOPE_UNIVERSE;
1545 r->rtm_protocol = RTPROT_UNSPEC;
1546 if (rt->rt_flags & RTCF_NOTIFY)
1547 r->rtm_flags |= RTM_F_NOTIFY;
1548 RTA_PUT(skb, RTA_DST, 2, &rt->rt_daddr);
1549 if (rt->fld.saddr) {
1550 r->rtm_src_len = 16;
1551 RTA_PUT(skb, RTA_SRC, 2, &rt->fld.saddr);
1554 RTA_PUT(skb, RTA_OIF, sizeof(int), &rt->dst.dev->ifindex);
1556 * Note to self - change this if input routes reverse direction when
1557 * they deal only with inputs and not with replies like they do
1560 RTA_PUT(skb, RTA_PREFSRC, 2, &rt->rt_local_src);
1561 if (rt->rt_daddr != rt->rt_gateway)
1562 RTA_PUT(skb, RTA_GATEWAY, 2, &rt->rt_gateway);
1563 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
1564 goto rtattr_failure;
1565 expires = rt->dst.expires ? rt->dst.expires - jiffies : 0;
1566 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, 0, 0, expires,
1568 goto rtattr_failure;
1569 if (dn_is_input_route(rt))
1570 RTA_PUT(skb, RTA_IIF, sizeof(int), &rt->fld.flowidn_iif);
1572 nlh->nlmsg_len = skb_tail_pointer(skb) - b;
1582 * This is called by both endnodes and routers now.
1584 static int dn_cache_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh, void *arg)
1586 struct net *net = sock_net(in_skb->sk);
1587 struct rtattr **rta = arg;
1588 struct rtmsg *rtm = NLMSG_DATA(nlh);
1589 struct dn_route *rt = NULL;
1590 struct dn_skb_cb *cb;
1592 struct sk_buff *skb;
1595 if (!net_eq(net, &init_net))
1598 memset(&fld, 0, sizeof(fld));
1599 fld.flowidn_proto = DNPROTO_NSP;
1601 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1604 skb_reset_mac_header(skb);
1605 cb = DN_SKB_CB(skb);
1608 memcpy(&fld.saddr, RTA_DATA(rta[RTA_SRC-1]), 2);
1610 memcpy(&fld.daddr, RTA_DATA(rta[RTA_DST-1]), 2);
1612 memcpy(&fld.flowidn_iif, RTA_DATA(rta[RTA_IIF-1]), sizeof(int));
1614 if (fld.flowidn_iif) {
1615 struct net_device *dev;
1616 if ((dev = dev_get_by_index(&init_net, fld.flowidn_iif)) == NULL) {
1625 skb->protocol = htons(ETH_P_DNA_RT);
1627 cb->src = fld.saddr;
1628 cb->dst = fld.daddr;
1630 err = dn_route_input(skb);
1632 memset(cb, 0, sizeof(struct dn_skb_cb));
1633 rt = (struct dn_route *)skb_dst(skb);
1634 if (!err && -rt->dst.error)
1635 err = rt->dst.error;
1638 if (rta[RTA_OIF - 1])
1639 memcpy(&oif, RTA_DATA(rta[RTA_OIF - 1]), sizeof(int));
1640 fld.flowidn_oif = oif;
1641 err = dn_route_output_key((struct dst_entry **)&rt, &fld, 0);
1649 skb_dst_set(skb, &rt->dst);
1650 if (rtm->rtm_flags & RTM_F_NOTIFY)
1651 rt->rt_flags |= RTCF_NOTIFY;
1653 err = dn_rt_fill_info(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, RTM_NEWROUTE, 0, 0);
1662 return rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
1670 * For routers, this is called from dn_fib_dump, but for endnodes its
1671 * called directly from the rtnetlink dispatch table.
1673 int dn_cache_dump(struct sk_buff *skb, struct netlink_callback *cb)
1675 struct net *net = sock_net(skb->sk);
1676 struct dn_route *rt;
1680 if (!net_eq(net, &init_net))
1683 if (NLMSG_PAYLOAD(cb->nlh, 0) < sizeof(struct rtmsg))
1685 if (!(((struct rtmsg *)NLMSG_DATA(cb->nlh))->rtm_flags&RTM_F_CLONED))
1689 s_idx = idx = cb->args[1];
1690 for(h = 0; h <= dn_rt_hash_mask; h++) {
1696 for(rt = rcu_dereference_bh(dn_rt_hash_table[h].chain), idx = 0;
1698 rt = rcu_dereference_bh(rt->dst.dn_next), idx++) {
1701 skb_dst_set(skb, dst_clone(&rt->dst));
1702 if (dn_rt_fill_info(skb, NETLINK_CB(cb->skb).pid,
1703 cb->nlh->nlmsg_seq, RTM_NEWROUTE,
1704 1, NLM_F_MULTI) <= 0) {
1706 rcu_read_unlock_bh();
1711 rcu_read_unlock_bh();
1720 #ifdef CONFIG_PROC_FS
1721 struct dn_rt_cache_iter_state {
1725 static struct dn_route *dn_rt_cache_get_first(struct seq_file *seq)
1727 struct dn_route *rt = NULL;
1728 struct dn_rt_cache_iter_state *s = seq->private;
1730 for(s->bucket = dn_rt_hash_mask; s->bucket >= 0; --s->bucket) {
1732 rt = rcu_dereference_bh(dn_rt_hash_table[s->bucket].chain);
1735 rcu_read_unlock_bh();
1740 static struct dn_route *dn_rt_cache_get_next(struct seq_file *seq, struct dn_route *rt)
1742 struct dn_rt_cache_iter_state *s = seq->private;
1744 rt = rcu_dereference_bh(rt->dst.dn_next);
1746 rcu_read_unlock_bh();
1747 if (--s->bucket < 0)
1750 rt = rcu_dereference_bh(dn_rt_hash_table[s->bucket].chain);
1755 static void *dn_rt_cache_seq_start(struct seq_file *seq, loff_t *pos)
1757 struct dn_route *rt = dn_rt_cache_get_first(seq);
1760 while(*pos && (rt = dn_rt_cache_get_next(seq, rt)))
1763 return *pos ? NULL : rt;
1766 static void *dn_rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1768 struct dn_route *rt = dn_rt_cache_get_next(seq, v);
1773 static void dn_rt_cache_seq_stop(struct seq_file *seq, void *v)
1776 rcu_read_unlock_bh();
1779 static int dn_rt_cache_seq_show(struct seq_file *seq, void *v)
1781 struct dn_route *rt = v;
1782 char buf1[DN_ASCBUF_LEN], buf2[DN_ASCBUF_LEN];
1784 seq_printf(seq, "%-8s %-7s %-7s %04d %04d %04d\n",
1785 rt->dst.dev ? rt->dst.dev->name : "*",
1786 dn_addr2asc(le16_to_cpu(rt->rt_daddr), buf1),
1787 dn_addr2asc(le16_to_cpu(rt->rt_saddr), buf2),
1788 atomic_read(&rt->dst.__refcnt),
1790 (int) dst_metric(&rt->dst, RTAX_RTT));
1794 static const struct seq_operations dn_rt_cache_seq_ops = {
1795 .start = dn_rt_cache_seq_start,
1796 .next = dn_rt_cache_seq_next,
1797 .stop = dn_rt_cache_seq_stop,
1798 .show = dn_rt_cache_seq_show,
1801 static int dn_rt_cache_seq_open(struct inode *inode, struct file *file)
1803 return seq_open_private(file, &dn_rt_cache_seq_ops,
1804 sizeof(struct dn_rt_cache_iter_state));
1807 static const struct file_operations dn_rt_cache_seq_fops = {
1808 .owner = THIS_MODULE,
1809 .open = dn_rt_cache_seq_open,
1811 .llseek = seq_lseek,
1812 .release = seq_release_private,
1815 #endif /* CONFIG_PROC_FS */
1817 void __init dn_route_init(void)
1821 dn_dst_ops.kmem_cachep =
1822 kmem_cache_create("dn_dst_cache", sizeof(struct dn_route), 0,
1823 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
1824 dst_entries_init(&dn_dst_ops);
1825 setup_timer(&dn_route_timer, dn_dst_check_expire, 0);
1826 dn_route_timer.expires = jiffies + decnet_dst_gc_interval * HZ;
1827 add_timer(&dn_route_timer);
1829 goal = totalram_pages >> (26 - PAGE_SHIFT);
1831 for(order = 0; (1UL << order) < goal; order++)
1835 * Only want 1024 entries max, since the table is very, very unlikely
1836 * to be larger than that.
1838 while(order && ((((1UL << order) * PAGE_SIZE) /
1839 sizeof(struct dn_rt_hash_bucket)) >= 2048))
1843 dn_rt_hash_mask = (1UL << order) * PAGE_SIZE /
1844 sizeof(struct dn_rt_hash_bucket);
1845 while(dn_rt_hash_mask & (dn_rt_hash_mask - 1))
1847 dn_rt_hash_table = (struct dn_rt_hash_bucket *)
1848 __get_free_pages(GFP_ATOMIC, order);
1849 } while (dn_rt_hash_table == NULL && --order > 0);
1851 if (!dn_rt_hash_table)
1852 panic("Failed to allocate DECnet route cache hash table\n");
1855 "DECnet: Routing cache hash table of %u buckets, %ldKbytes\n",
1857 (long)(dn_rt_hash_mask*sizeof(struct dn_rt_hash_bucket))/1024);
1860 for(i = 0; i <= dn_rt_hash_mask; i++) {
1861 spin_lock_init(&dn_rt_hash_table[i].lock);
1862 dn_rt_hash_table[i].chain = NULL;
1865 dn_dst_ops.gc_thresh = (dn_rt_hash_mask + 1);
1867 proc_net_fops_create(&init_net, "decnet_cache", S_IRUGO, &dn_rt_cache_seq_fops);
1869 #ifdef CONFIG_DECNET_ROUTER
1870 rtnl_register(PF_DECnet, RTM_GETROUTE, dn_cache_getroute,
1873 rtnl_register(PF_DECnet, RTM_GETROUTE, dn_cache_getroute,
1874 dn_cache_dump, NULL);
1878 void __exit dn_route_cleanup(void)
1880 del_timer(&dn_route_timer);
1883 proc_net_remove(&init_net, "decnet_cache");
1884 dst_entries_destroy(&dn_dst_ops);