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>
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>
80 #include <net/net_namespace.h>
81 #include <net/netlink.h>
82 #include <net/neighbour.h>
85 #include <net/fib_rules.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>
93 struct dn_rt_hash_bucket
95 struct dn_route
*chain
;
99 extern struct neigh_table dn_neigh_table
;
102 static unsigned char dn_hiord_addr
[6] = {0xAA,0x00,0x04,0x00,0x00,0x00};
104 static const int dn_rt_min_delay
= 2 * HZ
;
105 static const int dn_rt_max_delay
= 10 * HZ
;
106 static const int dn_rt_mtu_expires
= 10 * 60 * HZ
;
108 static unsigned long dn_rt_deadline
;
110 static int dn_dst_gc(struct dst_ops
*ops
);
111 static struct dst_entry
*dn_dst_check(struct dst_entry
*, __u32
);
112 static struct dst_entry
*dn_dst_negative_advice(struct dst_entry
*);
113 static void dn_dst_link_failure(struct sk_buff
*);
114 static void dn_dst_update_pmtu(struct dst_entry
*dst
, u32 mtu
);
115 static int dn_route_input(struct sk_buff
*);
116 static void dn_run_flush(unsigned long dummy
);
118 static struct dn_rt_hash_bucket
*dn_rt_hash_table
;
119 static unsigned dn_rt_hash_mask
;
121 static struct timer_list dn_route_timer
;
122 static DEFINE_TIMER(dn_rt_flush_timer
, dn_run_flush
, 0, 0);
123 int decnet_dst_gc_interval
= 2;
125 static struct dst_ops dn_dst_ops
= {
127 .protocol
= __constant_htons(ETH_P_DNA_RT
),
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 .entries
= ATOMIC_INIT(0),
137 static __inline__
unsigned dn_hash(__le16 src
, __le16 dst
)
139 __u16 tmp
= (__u16 __force
)(src
^ dst
);
143 return dn_rt_hash_mask
& (unsigned)tmp
;
146 static inline void dnrt_free(struct dn_route
*rt
)
148 call_rcu_bh(&rt
->u
.dst
.rcu_head
, dst_rcu_free
);
151 static inline void dnrt_drop(struct dn_route
*rt
)
153 dst_release(&rt
->u
.dst
);
154 call_rcu_bh(&rt
->u
.dst
.rcu_head
, dst_rcu_free
);
157 static void dn_dst_check_expire(unsigned long dummy
)
160 struct dn_route
*rt
, **rtp
;
161 unsigned long now
= jiffies
;
162 unsigned long expire
= 120 * HZ
;
164 for(i
= 0; i
<= dn_rt_hash_mask
; i
++) {
165 rtp
= &dn_rt_hash_table
[i
].chain
;
167 spin_lock(&dn_rt_hash_table
[i
].lock
);
168 while((rt
=*rtp
) != NULL
) {
169 if (atomic_read(&rt
->u
.dst
.__refcnt
) ||
170 (now
- rt
->u
.dst
.lastuse
) < expire
) {
171 rtp
= &rt
->u
.dst
.dn_next
;
174 *rtp
= rt
->u
.dst
.dn_next
;
175 rt
->u
.dst
.dn_next
= NULL
;
178 spin_unlock(&dn_rt_hash_table
[i
].lock
);
180 if ((jiffies
- now
) > 0)
184 mod_timer(&dn_route_timer
, now
+ decnet_dst_gc_interval
* HZ
);
187 static int dn_dst_gc(struct dst_ops
*ops
)
189 struct dn_route
*rt
, **rtp
;
191 unsigned long now
= jiffies
;
192 unsigned long expire
= 10 * HZ
;
194 for(i
= 0; i
<= dn_rt_hash_mask
; i
++) {
196 spin_lock_bh(&dn_rt_hash_table
[i
].lock
);
197 rtp
= &dn_rt_hash_table
[i
].chain
;
199 while((rt
=*rtp
) != NULL
) {
200 if (atomic_read(&rt
->u
.dst
.__refcnt
) ||
201 (now
- rt
->u
.dst
.lastuse
) < expire
) {
202 rtp
= &rt
->u
.dst
.dn_next
;
205 *rtp
= rt
->u
.dst
.dn_next
;
206 rt
->u
.dst
.dn_next
= NULL
;
210 spin_unlock_bh(&dn_rt_hash_table
[i
].lock
);
217 * The decnet standards don't impose a particular minimum mtu, what they
218 * do insist on is that the routing layer accepts a datagram of at least
219 * 230 bytes long. Here we have to subtract the routing header length from
220 * 230 to get the minimum acceptable mtu. If there is no neighbour, then we
221 * assume the worst and use a long header size.
223 * We update both the mtu and the advertised mss (i.e. the segment size we
224 * advertise to the other end).
226 static void dn_dst_update_pmtu(struct dst_entry
*dst
, u32 mtu
)
229 struct dn_dev
*dn
= dst
->neighbour
?
230 (struct dn_dev
*)dst
->neighbour
->dev
->dn_ptr
: NULL
;
232 if (dn
&& dn
->use_long
== 0)
237 if (dst_metric(dst
, RTAX_MTU
) > mtu
&& mtu
>= min_mtu
) {
238 if (!(dst_metric_locked(dst
, RTAX_MTU
))) {
239 dst
->metrics
[RTAX_MTU
-1] = mtu
;
240 dst_set_expires(dst
, dn_rt_mtu_expires
);
242 if (!(dst_metric_locked(dst
, RTAX_ADVMSS
))) {
243 u32 mss
= mtu
- DN_MAX_NSP_DATA_HEADER
;
244 if (dst_metric(dst
, RTAX_ADVMSS
) > mss
)
245 dst
->metrics
[RTAX_ADVMSS
-1] = mss
;
251 * When a route has been marked obsolete. (e.g. routing cache flush)
253 static struct dst_entry
*dn_dst_check(struct dst_entry
*dst
, __u32 cookie
)
258 static struct dst_entry
*dn_dst_negative_advice(struct dst_entry
*dst
)
264 static void dn_dst_link_failure(struct sk_buff
*skb
)
269 static inline int compare_keys(struct flowi
*fl1
, struct flowi
*fl2
)
271 return ((fl1
->nl_u
.dn_u
.daddr
^ fl2
->nl_u
.dn_u
.daddr
) |
272 (fl1
->nl_u
.dn_u
.saddr
^ fl2
->nl_u
.dn_u
.saddr
) |
273 (fl1
->mark
^ fl2
->mark
) |
274 (fl1
->nl_u
.dn_u
.scope
^ fl2
->nl_u
.dn_u
.scope
) |
275 (fl1
->oif
^ fl2
->oif
) |
276 (fl1
->iif
^ fl2
->iif
)) == 0;
279 static int dn_insert_route(struct dn_route
*rt
, unsigned hash
, struct dn_route
**rp
)
281 struct dn_route
*rth
, **rthp
;
282 unsigned long now
= jiffies
;
284 rthp
= &dn_rt_hash_table
[hash
].chain
;
286 spin_lock_bh(&dn_rt_hash_table
[hash
].lock
);
287 while((rth
= *rthp
) != NULL
) {
288 if (compare_keys(&rth
->fl
, &rt
->fl
)) {
290 *rthp
= rth
->u
.dst
.dn_next
;
291 rcu_assign_pointer(rth
->u
.dst
.dn_next
,
292 dn_rt_hash_table
[hash
].chain
);
293 rcu_assign_pointer(dn_rt_hash_table
[hash
].chain
, rth
);
295 dst_use(&rth
->u
.dst
, now
);
296 spin_unlock_bh(&dn_rt_hash_table
[hash
].lock
);
302 rthp
= &rth
->u
.dst
.dn_next
;
305 rcu_assign_pointer(rt
->u
.dst
.dn_next
, dn_rt_hash_table
[hash
].chain
);
306 rcu_assign_pointer(dn_rt_hash_table
[hash
].chain
, rt
);
308 dst_use(&rt
->u
.dst
, now
);
309 spin_unlock_bh(&dn_rt_hash_table
[hash
].lock
);
314 static void dn_run_flush(unsigned long dummy
)
317 struct dn_route
*rt
, *next
;
319 for(i
= 0; i
< dn_rt_hash_mask
; i
++) {
320 spin_lock_bh(&dn_rt_hash_table
[i
].lock
);
322 if ((rt
= xchg(&dn_rt_hash_table
[i
].chain
, NULL
)) == NULL
)
323 goto nothing_to_declare
;
326 next
= rt
->u
.dst
.dn_next
;
327 rt
->u
.dst
.dn_next
= NULL
;
328 dst_free((struct dst_entry
*)rt
);
332 spin_unlock_bh(&dn_rt_hash_table
[i
].lock
);
336 static DEFINE_SPINLOCK(dn_rt_flush_lock
);
338 void dn_rt_cache_flush(int delay
)
340 unsigned long now
= jiffies
;
341 int user_mode
= !in_interrupt();
344 delay
= dn_rt_min_delay
;
346 spin_lock_bh(&dn_rt_flush_lock
);
348 if (del_timer(&dn_rt_flush_timer
) && delay
> 0 && dn_rt_deadline
) {
349 long tmo
= (long)(dn_rt_deadline
- now
);
351 if (user_mode
&& tmo
< dn_rt_max_delay
- dn_rt_min_delay
)
359 spin_unlock_bh(&dn_rt_flush_lock
);
364 if (dn_rt_deadline
== 0)
365 dn_rt_deadline
= now
+ dn_rt_max_delay
;
367 dn_rt_flush_timer
.expires
= now
+ delay
;
368 add_timer(&dn_rt_flush_timer
);
369 spin_unlock_bh(&dn_rt_flush_lock
);
373 * dn_return_short - Return a short packet to its sender
374 * @skb: The packet to return
377 static int dn_return_short(struct sk_buff
*skb
)
379 struct dn_skb_cb
*cb
;
385 /* Add back headers */
386 skb_push(skb
, skb
->data
- skb_network_header(skb
));
388 if ((skb
= skb_unshare(skb
, GFP_ATOMIC
)) == NULL
)
392 /* Skip packet length and point to flags */
394 *ptr
++ = (cb
->rt_flags
& ~DN_RT_F_RQR
) | DN_RT_F_RTS
;
400 *ptr
= 0; /* Zero hop count */
402 /* Swap source and destination */
407 skb
->pkt_type
= PACKET_OUTGOING
;
408 dn_rt_finish_output(skb
, NULL
, NULL
);
409 return NET_RX_SUCCESS
;
413 * dn_return_long - Return a long packet to its sender
414 * @skb: The long format packet to return
417 static int dn_return_long(struct sk_buff
*skb
)
419 struct dn_skb_cb
*cb
;
421 unsigned char *src_addr
, *dst_addr
;
422 unsigned char tmp
[ETH_ALEN
];
424 /* Add back all headers */
425 skb_push(skb
, skb
->data
- skb_network_header(skb
));
427 if ((skb
= skb_unshare(skb
, GFP_ATOMIC
)) == NULL
)
431 /* Ignore packet length and point to flags */
435 if (*ptr
& DN_RT_F_PF
) {
436 char padlen
= (*ptr
& ~DN_RT_F_PF
);
440 *ptr
++ = (cb
->rt_flags
& ~DN_RT_F_RQR
) | DN_RT_F_RTS
;
446 *ptr
= 0; /* Zero hop count */
448 /* Swap source and destination */
449 memcpy(tmp
, src_addr
, ETH_ALEN
);
450 memcpy(src_addr
, dst_addr
, ETH_ALEN
);
451 memcpy(dst_addr
, tmp
, ETH_ALEN
);
453 skb
->pkt_type
= PACKET_OUTGOING
;
454 dn_rt_finish_output(skb
, dst_addr
, src_addr
);
455 return NET_RX_SUCCESS
;
459 * dn_route_rx_packet - Try and find a route for an incoming packet
460 * @skb: The packet to find a route for
462 * Returns: result of input function if route is found, error code otherwise
464 static int dn_route_rx_packet(struct sk_buff
*skb
)
466 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
469 if ((err
= dn_route_input(skb
)) == 0)
470 return dst_input(skb
);
472 if (decnet_debug_level
& 4) {
473 char *devname
= skb
->dev
? skb
->dev
->name
: "???";
474 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
476 "DECnet: dn_route_rx_packet: rt_flags=0x%02x dev=%s len=%d src=0x%04hx dst=0x%04hx err=%d type=%d\n",
477 (int)cb
->rt_flags
, devname
, skb
->len
,
478 le16_to_cpu(cb
->src
), le16_to_cpu(cb
->dst
),
482 if ((skb
->pkt_type
== PACKET_HOST
) && (cb
->rt_flags
& DN_RT_F_RQR
)) {
483 switch(cb
->rt_flags
& DN_RT_PKT_MSK
) {
484 case DN_RT_PKT_SHORT
:
485 return dn_return_short(skb
);
487 return dn_return_long(skb
);
495 static int dn_route_rx_long(struct sk_buff
*skb
)
497 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
498 unsigned char *ptr
= skb
->data
;
500 if (!pskb_may_pull(skb
, 21)) /* 20 for long header, 1 for shortest nsp */
504 skb_reset_transport_header(skb
);
506 /* Destination info */
508 cb
->dst
= dn_eth2dn(ptr
);
509 if (memcmp(ptr
, dn_hiord_addr
, 4) != 0)
516 cb
->src
= dn_eth2dn(ptr
);
517 if (memcmp(ptr
, dn_hiord_addr
, 4) != 0)
522 cb
->hops
= *ptr
++; /* Visit Count */
524 return NF_HOOK(PF_DECnet
, NF_DN_PRE_ROUTING
, skb
, skb
->dev
, NULL
, dn_route_rx_packet
);
533 static int dn_route_rx_short(struct sk_buff
*skb
)
535 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
536 unsigned char *ptr
= skb
->data
;
538 if (!pskb_may_pull(skb
, 6)) /* 5 for short header + 1 for shortest nsp */
542 skb_reset_transport_header(skb
);
544 cb
->dst
= *(__le16
*)ptr
;
546 cb
->src
= *(__le16
*)ptr
;
548 cb
->hops
= *ptr
& 0x3f;
550 return NF_HOOK(PF_DECnet
, NF_DN_PRE_ROUTING
, skb
, skb
->dev
, NULL
, dn_route_rx_packet
);
557 static int dn_route_discard(struct sk_buff
*skb
)
560 * I know we drop the packet here, but thats considered success in
564 return NET_RX_SUCCESS
;
567 static int dn_route_ptp_hello(struct sk_buff
*skb
)
570 dn_neigh_pointopoint_hello(skb
);
571 return NET_RX_SUCCESS
;
574 int dn_route_rcv(struct sk_buff
*skb
, struct net_device
*dev
, struct packet_type
*pt
, struct net_device
*orig_dev
)
576 struct dn_skb_cb
*cb
;
577 unsigned char flags
= 0;
578 __u16 len
= le16_to_cpu(*(__le16
*)skb
->data
);
579 struct dn_dev
*dn
= (struct dn_dev
*)dev
->dn_ptr
;
580 unsigned char padlen
= 0;
582 if (!net_eq(dev_net(dev
), &init_net
))
588 if ((skb
= skb_share_check(skb
, GFP_ATOMIC
)) == NULL
)
591 if (!pskb_may_pull(skb
, 3))
605 cb
->iif
= dev
->ifindex
;
608 * If we have padding, remove it.
610 if (flags
& DN_RT_F_PF
) {
611 padlen
= flags
& ~DN_RT_F_PF
;
612 if (!pskb_may_pull(skb
, padlen
+ 1))
614 skb_pull(skb
, padlen
);
618 skb_reset_network_header(skb
);
621 * Weed out future version DECnet
623 if (flags
& DN_RT_F_VER
)
626 cb
->rt_flags
= flags
;
628 if (decnet_debug_level
& 1)
630 "dn_route_rcv: got 0x%02x from %s [%d %d %d]\n",
631 (int)flags
, (dev
) ? dev
->name
: "???", len
, skb
->len
,
634 if (flags
& DN_RT_PKT_CNTL
) {
635 if (unlikely(skb_linearize(skb
)))
638 switch(flags
& DN_RT_CNTL_MSK
) {
640 dn_dev_init_pkt(skb
);
643 dn_dev_veri_pkt(skb
);
647 if (dn
->parms
.state
!= DN_DEV_S_RU
)
650 switch(flags
& DN_RT_CNTL_MSK
) {
652 return NF_HOOK(PF_DECnet
, NF_DN_HELLO
, skb
, skb
->dev
, NULL
, dn_route_ptp_hello
);
656 return NF_HOOK(PF_DECnet
, NF_DN_ROUTE
, skb
, skb
->dev
, NULL
, dn_route_discard
);
658 return NF_HOOK(PF_DECnet
, NF_DN_HELLO
, skb
, skb
->dev
, NULL
, dn_neigh_router_hello
);
661 return NF_HOOK(PF_DECnet
, NF_DN_HELLO
, skb
, skb
->dev
, NULL
, dn_neigh_endnode_hello
);
664 if (dn
->parms
.state
!= DN_DEV_S_RU
)
667 skb_pull(skb
, 1); /* Pull flags */
669 switch(flags
& DN_RT_PKT_MSK
) {
671 return dn_route_rx_long(skb
);
672 case DN_RT_PKT_SHORT
:
673 return dn_route_rx_short(skb
);
683 static int dn_output(struct sk_buff
*skb
)
685 struct dst_entry
*dst
= skb
->dst
;
686 struct dn_route
*rt
= (struct dn_route
*)dst
;
687 struct net_device
*dev
= dst
->dev
;
688 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
689 struct neighbour
*neigh
;
693 if ((neigh
= dst
->neighbour
) == NULL
)
698 cb
->src
= rt
->rt_saddr
;
699 cb
->dst
= rt
->rt_daddr
;
702 * Always set the Intra-Ethernet bit on all outgoing packets
703 * originated on this node. Only valid flag from upper layers
704 * is return-to-sender-requested. Set hop count to 0 too.
706 cb
->rt_flags
&= ~DN_RT_F_RQR
;
707 cb
->rt_flags
|= DN_RT_F_IE
;
710 return NF_HOOK(PF_DECnet
, NF_DN_LOCAL_OUT
, skb
, NULL
, dev
, neigh
->output
);
714 printk(KERN_DEBUG
"dn_output: This should not happen\n");
721 static int dn_forward(struct sk_buff
*skb
)
723 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
724 struct dst_entry
*dst
= skb
->dst
;
725 struct dn_dev
*dn_db
= dst
->dev
->dn_ptr
;
727 struct neighbour
*neigh
= dst
->neighbour
;
729 #ifdef CONFIG_NETFILTER
730 struct net_device
*dev
= skb
->dev
;
733 if (skb
->pkt_type
!= PACKET_HOST
)
736 /* Ensure that we have enough space for headers */
737 rt
= (struct dn_route
*)skb
->dst
;
738 header_len
= dn_db
->use_long
? 21 : 6;
739 if (skb_cow(skb
, LL_RESERVED_SPACE(rt
->u
.dst
.dev
)+header_len
))
743 * Hop count exceeded.
748 skb
->dev
= rt
->u
.dst
.dev
;
751 * If packet goes out same interface it came in on, then set
752 * the Intra-Ethernet bit. This has no effect for short
753 * packets, so we don't need to test for them here.
755 cb
->rt_flags
&= ~DN_RT_F_IE
;
756 if (rt
->rt_flags
& RTCF_DOREDIRECT
)
757 cb
->rt_flags
|= DN_RT_F_IE
;
759 return NF_HOOK(PF_DECnet
, NF_DN_FORWARD
, skb
, dev
, skb
->dev
, neigh
->output
);
767 * Used to catch bugs. This should never normally get
770 static int dn_rt_bug(struct sk_buff
*skb
)
772 if (net_ratelimit()) {
773 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
775 printk(KERN_DEBUG
"dn_rt_bug: skb from:%04x to:%04x\n",
776 le16_to_cpu(cb
->src
), le16_to_cpu(cb
->dst
));
784 static int dn_rt_set_next_hop(struct dn_route
*rt
, struct dn_fib_res
*res
)
786 struct dn_fib_info
*fi
= res
->fi
;
787 struct net_device
*dev
= rt
->u
.dst
.dev
;
792 if (DN_FIB_RES_GW(*res
) &&
793 DN_FIB_RES_NH(*res
).nh_scope
== RT_SCOPE_LINK
)
794 rt
->rt_gateway
= DN_FIB_RES_GW(*res
);
795 memcpy(rt
->u
.dst
.metrics
, fi
->fib_metrics
,
796 sizeof(rt
->u
.dst
.metrics
));
798 rt
->rt_type
= res
->type
;
800 if (dev
!= NULL
&& rt
->u
.dst
.neighbour
== NULL
) {
801 n
= __neigh_lookup_errno(&dn_neigh_table
, &rt
->rt_gateway
, dev
);
804 rt
->u
.dst
.neighbour
= n
;
807 if (dst_metric(&rt
->u
.dst
, RTAX_MTU
) == 0 ||
808 dst_metric(&rt
->u
.dst
, RTAX_MTU
) > rt
->u
.dst
.dev
->mtu
)
809 rt
->u
.dst
.metrics
[RTAX_MTU
-1] = rt
->u
.dst
.dev
->mtu
;
810 mss
= dn_mss_from_pmtu(dev
, dst_mtu(&rt
->u
.dst
));
811 if (dst_metric(&rt
->u
.dst
, RTAX_ADVMSS
) == 0 ||
812 dst_metric(&rt
->u
.dst
, RTAX_ADVMSS
) > mss
)
813 rt
->u
.dst
.metrics
[RTAX_ADVMSS
-1] = mss
;
817 static inline int dn_match_addr(__le16 addr1
, __le16 addr2
)
819 __u16 tmp
= le16_to_cpu(addr1
) ^ le16_to_cpu(addr2
);
828 static __le16
dnet_select_source(const struct net_device
*dev
, __le16 daddr
, int scope
)
831 struct dn_dev
*dn_db
= dev
->dn_ptr
;
832 struct dn_ifaddr
*ifa
;
836 read_lock(&dev_base_lock
);
837 for(ifa
= dn_db
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
) {
838 if (ifa
->ifa_scope
> scope
)
841 saddr
= ifa
->ifa_local
;
844 ret
= dn_match_addr(daddr
, ifa
->ifa_local
);
845 if (ret
> best_match
)
846 saddr
= ifa
->ifa_local
;
848 saddr
= ifa
->ifa_local
;
850 read_unlock(&dev_base_lock
);
855 static inline __le16
__dn_fib_res_prefsrc(struct dn_fib_res
*res
)
857 return dnet_select_source(DN_FIB_RES_DEV(*res
), DN_FIB_RES_GW(*res
), res
->scope
);
860 static inline __le16
dn_fib_rules_map_destination(__le16 daddr
, struct dn_fib_res
*res
)
862 __le16 mask
= dnet_make_mask(res
->prefixlen
);
863 return (daddr
&~mask
)|res
->fi
->fib_nh
->nh_gw
;
866 static int dn_route_output_slow(struct dst_entry
**pprt
, const struct flowi
*oldflp
, int try_hard
)
868 struct flowi fl
= { .nl_u
= { .dn_u
=
869 { .daddr
= oldflp
->fld_dst
,
870 .saddr
= oldflp
->fld_src
,
871 .scope
= RT_SCOPE_UNIVERSE
,
873 .mark
= oldflp
->mark
,
874 .iif
= init_net
.loopback_dev
->ifindex
,
875 .oif
= oldflp
->oif
};
876 struct dn_route
*rt
= NULL
;
877 struct net_device
*dev_out
= NULL
, *dev
;
878 struct neighbour
*neigh
= NULL
;
881 struct dn_fib_res res
= { .fi
= NULL
, .type
= RTN_UNICAST
};
886 if (decnet_debug_level
& 16)
888 "dn_route_output_slow: dst=%04x src=%04x mark=%d"
889 " iif=%d oif=%d\n", le16_to_cpu(oldflp
->fld_dst
),
890 le16_to_cpu(oldflp
->fld_src
),
891 oldflp
->mark
, init_net
.loopback_dev
->ifindex
, oldflp
->oif
);
893 /* If we have an output interface, verify its a DECnet device */
895 dev_out
= dev_get_by_index(&init_net
, oldflp
->oif
);
897 if (dev_out
&& dev_out
->dn_ptr
== NULL
) {
905 /* If we have a source address, verify that its a local address */
906 if (oldflp
->fld_src
) {
907 err
= -EADDRNOTAVAIL
;
910 if (dn_dev_islocal(dev_out
, oldflp
->fld_src
))
915 read_lock(&dev_base_lock
);
916 for_each_netdev(&init_net
, dev
) {
919 if (!dn_dev_islocal(dev
, oldflp
->fld_src
))
921 if ((dev
->flags
& IFF_LOOPBACK
) &&
923 !dn_dev_islocal(dev
, oldflp
->fld_dst
))
929 read_unlock(&dev_base_lock
);
937 /* No destination? Assume its local */
939 fl
.fld_dst
= fl
.fld_src
;
941 err
= -EADDRNOTAVAIL
;
944 dev_out
= init_net
.loopback_dev
;
948 fl
.fld_src
= dnet_select_source(dev_out
, 0,
953 fl
.oif
= init_net
.loopback_dev
->ifindex
;
954 res
.type
= RTN_LOCAL
;
958 if (decnet_debug_level
& 16)
960 "dn_route_output_slow: initial checks complete."
961 " dst=%o4x src=%04x oif=%d try_hard=%d\n",
962 le16_to_cpu(fl
.fld_dst
), le16_to_cpu(fl
.fld_src
),
966 * N.B. If the kernel is compiled without router support then
967 * dn_fib_lookup() will evaluate to non-zero so this if () block
968 * will always be executed.
971 if (try_hard
|| (err
= dn_fib_lookup(&fl
, &res
)) != 0) {
972 struct dn_dev
*dn_db
;
976 * Here the fallback is basically the standard algorithm for
977 * routing in endnodes which is described in the DECnet routing
980 * If we are not trying hard, look in neighbour cache.
981 * The result is tested to ensure that if a specific output
982 * device/source address was requested, then we honour that
986 neigh
= neigh_lookup_nodev(&dn_neigh_table
, &init_net
, &fl
.fld_dst
);
989 (neigh
->dev
->ifindex
!= oldflp
->oif
)) ||
991 (!dn_dev_islocal(neigh
->dev
,
992 oldflp
->fld_src
)))) {
993 neigh_release(neigh
);
998 if (dn_dev_islocal(neigh
->dev
, fl
.fld_dst
)) {
999 dev_out
= init_net
.loopback_dev
;
1000 res
.type
= RTN_LOCAL
;
1002 dev_out
= neigh
->dev
;
1010 /* Not there? Perhaps its a local address */
1011 if (dev_out
== NULL
)
1012 dev_out
= dn_dev_get_default();
1014 if (dev_out
== NULL
)
1016 dn_db
= dev_out
->dn_ptr
;
1017 /* Possible improvement - check all devices for local addr */
1018 if (dn_dev_islocal(dev_out
, fl
.fld_dst
)) {
1020 dev_out
= init_net
.loopback_dev
;
1022 res
.type
= RTN_LOCAL
;
1025 /* Not local either.... try sending it to the default router */
1026 neigh
= neigh_clone(dn_db
->router
);
1027 BUG_ON(neigh
&& neigh
->dev
!= dev_out
);
1029 /* Ok then, we assume its directly connected and move on */
1032 gateway
= ((struct dn_neigh
*)neigh
)->addr
;
1034 gateway
= fl
.fld_dst
;
1035 if (fl
.fld_src
== 0) {
1036 fl
.fld_src
= dnet_select_source(dev_out
, gateway
,
1037 res
.type
== RTN_LOCAL
?
1040 if (fl
.fld_src
== 0 && res
.type
!= RTN_LOCAL
)
1043 fl
.oif
= dev_out
->ifindex
;
1048 if (res
.type
== RTN_NAT
)
1051 if (res
.type
== RTN_LOCAL
) {
1053 fl
.fld_src
= fl
.fld_dst
;
1056 dev_out
= init_net
.loopback_dev
;
1058 fl
.oif
= dev_out
->ifindex
;
1060 dn_fib_info_put(res
.fi
);
1065 if (res
.fi
->fib_nhs
> 1 && fl
.oif
== 0)
1066 dn_fib_select_multipath(&fl
, &res
);
1069 * We could add some logic to deal with default routes here and
1070 * get rid of some of the special casing above.
1074 fl
.fld_src
= DN_FIB_RES_PREFSRC(res
);
1078 dev_out
= DN_FIB_RES_DEV(res
);
1080 fl
.oif
= dev_out
->ifindex
;
1081 gateway
= DN_FIB_RES_GW(res
);
1084 if (dev_out
->flags
& IFF_LOOPBACK
)
1085 flags
|= RTCF_LOCAL
;
1087 rt
= dst_alloc(&dn_dst_ops
);
1091 atomic_set(&rt
->u
.dst
.__refcnt
, 1);
1092 rt
->u
.dst
.flags
= DST_HOST
;
1094 rt
->fl
.fld_src
= oldflp
->fld_src
;
1095 rt
->fl
.fld_dst
= oldflp
->fld_dst
;
1096 rt
->fl
.oif
= oldflp
->oif
;
1098 rt
->fl
.mark
= oldflp
->mark
;
1100 rt
->rt_saddr
= fl
.fld_src
;
1101 rt
->rt_daddr
= fl
.fld_dst
;
1102 rt
->rt_gateway
= gateway
? gateway
: fl
.fld_dst
;
1103 rt
->rt_local_src
= fl
.fld_src
;
1105 rt
->rt_dst_map
= fl
.fld_dst
;
1106 rt
->rt_src_map
= fl
.fld_src
;
1108 rt
->u
.dst
.dev
= dev_out
;
1110 rt
->u
.dst
.neighbour
= neigh
;
1113 rt
->u
.dst
.lastuse
= jiffies
;
1114 rt
->u
.dst
.output
= dn_output
;
1115 rt
->u
.dst
.input
= dn_rt_bug
;
1116 rt
->rt_flags
= flags
;
1117 if (flags
& RTCF_LOCAL
)
1118 rt
->u
.dst
.input
= dn_nsp_rx
;
1120 err
= dn_rt_set_next_hop(rt
, &res
);
1124 hash
= dn_hash(rt
->fl
.fld_src
, rt
->fl
.fld_dst
);
1125 dn_insert_route(rt
, hash
, (struct dn_route
**)pprt
);
1129 neigh_release(neigh
);
1131 dn_fib_res_put(&res
);
1138 err
= -EADDRNOTAVAIL
;
1147 dst_free(&rt
->u
.dst
);
1153 * N.B. The flags may be moved into the flowi at some future stage.
1155 static int __dn_route_output_key(struct dst_entry
**pprt
, const struct flowi
*flp
, int flags
)
1157 unsigned hash
= dn_hash(flp
->fld_src
, flp
->fld_dst
);
1158 struct dn_route
*rt
= NULL
;
1160 if (!(flags
& MSG_TRYHARD
)) {
1162 for(rt
= rcu_dereference(dn_rt_hash_table
[hash
].chain
); rt
;
1163 rt
= rcu_dereference(rt
->u
.dst
.dn_next
)) {
1164 if ((flp
->fld_dst
== rt
->fl
.fld_dst
) &&
1165 (flp
->fld_src
== rt
->fl
.fld_src
) &&
1166 (flp
->mark
== rt
->fl
.mark
) &&
1167 (rt
->fl
.iif
== 0) &&
1168 (rt
->fl
.oif
== flp
->oif
)) {
1169 dst_use(&rt
->u
.dst
, jiffies
);
1170 rcu_read_unlock_bh();
1175 rcu_read_unlock_bh();
1178 return dn_route_output_slow(pprt
, flp
, flags
);
1181 static int dn_route_output_key(struct dst_entry
**pprt
, struct flowi
*flp
, int flags
)
1185 err
= __dn_route_output_key(pprt
, flp
, flags
);
1186 if (err
== 0 && flp
->proto
) {
1187 err
= xfrm_lookup(&init_net
, pprt
, flp
, NULL
, 0);
1192 int dn_route_output_sock(struct dst_entry
**pprt
, struct flowi
*fl
, struct sock
*sk
, int flags
)
1196 err
= __dn_route_output_key(pprt
, fl
, flags
& MSG_TRYHARD
);
1197 if (err
== 0 && fl
->proto
) {
1198 err
= xfrm_lookup(&init_net
, pprt
, fl
, sk
,
1199 (flags
& MSG_DONTWAIT
) ? 0 : XFRM_LOOKUP_WAIT
);
1204 static int dn_route_input_slow(struct sk_buff
*skb
)
1206 struct dn_route
*rt
= NULL
;
1207 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
1208 struct net_device
*in_dev
= skb
->dev
;
1209 struct net_device
*out_dev
= NULL
;
1210 struct dn_dev
*dn_db
;
1211 struct neighbour
*neigh
= NULL
;
1215 __le16 local_src
= 0;
1216 struct flowi fl
= { .nl_u
= { .dn_u
=
1219 .scope
= RT_SCOPE_UNIVERSE
,
1222 .iif
= skb
->dev
->ifindex
};
1223 struct dn_fib_res res
= { .fi
= NULL
, .type
= RTN_UNREACHABLE
};
1229 if ((dn_db
= in_dev
->dn_ptr
) == NULL
)
1232 /* Zero source addresses are not allowed */
1233 if (fl
.fld_src
== 0)
1237 * In this case we've just received a packet from a source
1238 * outside ourselves pretending to come from us. We don't
1239 * allow it any further to prevent routing loops, spoofing and
1240 * other nasties. Loopback packets already have the dst attached
1241 * so this only affects packets which have originated elsewhere.
1244 if (dn_dev_islocal(in_dev
, cb
->src
))
1247 err
= dn_fib_lookup(&fl
, &res
);
1252 * Is the destination us ?
1254 if (!dn_dev_islocal(in_dev
, cb
->dst
))
1257 res
.type
= RTN_LOCAL
;
1259 __le16 src_map
= fl
.fld_src
;
1262 out_dev
= DN_FIB_RES_DEV(res
);
1263 if (out_dev
== NULL
) {
1264 if (net_ratelimit())
1265 printk(KERN_CRIT
"Bug in dn_route_input_slow() "
1266 "No output device\n");
1272 src_map
= fl
.fld_src
; /* no NAT support for now */
1274 gateway
= DN_FIB_RES_GW(res
);
1275 if (res
.type
== RTN_NAT
) {
1276 fl
.fld_dst
= dn_fib_rules_map_destination(fl
.fld_dst
, &res
);
1277 dn_fib_res_put(&res
);
1279 if (dn_fib_lookup(&fl
, &res
))
1282 if (res
.type
!= RTN_UNICAST
)
1285 gateway
= fl
.fld_dst
;
1287 fl
.fld_src
= src_map
;
1293 * Forwarding check here, we only check for forwarding
1294 * being turned off, if you want to only forward intra
1295 * area, its up to you to set the routing tables up
1298 if (dn_db
->parms
.forwarding
== 0)
1301 if (res
.fi
->fib_nhs
> 1 && fl
.oif
== 0)
1302 dn_fib_select_multipath(&fl
, &res
);
1305 * Check for out_dev == in_dev. We use the RTCF_DOREDIRECT
1306 * flag as a hint to set the intra-ethernet bit when
1307 * forwarding. If we've got NAT in operation, we don't do
1308 * this optimisation.
1310 if (out_dev
== in_dev
&& !(flags
& RTCF_NAT
))
1311 flags
|= RTCF_DOREDIRECT
;
1313 local_src
= DN_FIB_RES_PREFSRC(res
);
1316 case RTN_UNREACHABLE
:
1319 flags
|= RTCF_LOCAL
;
1320 fl
.fld_src
= cb
->dst
;
1321 fl
.fld_dst
= cb
->src
;
1323 /* Routing tables gave us a gateway */
1327 /* Packet was intra-ethernet, so we know its on-link */
1328 if (cb
->rt_flags
& DN_RT_F_IE
) {
1330 flags
|= RTCF_DIRECTSRC
;
1334 /* Use the default router if there is one */
1335 neigh
= neigh_clone(dn_db
->router
);
1337 gateway
= ((struct dn_neigh
*)neigh
)->addr
;
1341 /* Close eyes and pray */
1343 flags
|= RTCF_DIRECTSRC
;
1350 rt
= dst_alloc(&dn_dst_ops
);
1354 rt
->rt_saddr
= fl
.fld_src
;
1355 rt
->rt_daddr
= fl
.fld_dst
;
1356 rt
->rt_gateway
= fl
.fld_dst
;
1358 rt
->rt_gateway
= gateway
;
1359 rt
->rt_local_src
= local_src
? local_src
: rt
->rt_saddr
;
1361 rt
->rt_dst_map
= fl
.fld_dst
;
1362 rt
->rt_src_map
= fl
.fld_src
;
1364 rt
->fl
.fld_src
= cb
->src
;
1365 rt
->fl
.fld_dst
= cb
->dst
;
1367 rt
->fl
.iif
= in_dev
->ifindex
;
1368 rt
->fl
.mark
= fl
.mark
;
1370 rt
->u
.dst
.flags
= DST_HOST
;
1371 rt
->u
.dst
.neighbour
= neigh
;
1372 rt
->u
.dst
.dev
= out_dev
;
1373 rt
->u
.dst
.lastuse
= jiffies
;
1374 rt
->u
.dst
.output
= dn_rt_bug
;
1377 rt
->u
.dst
.input
= dn_forward
;
1380 rt
->u
.dst
.output
= dn_output
;
1381 rt
->u
.dst
.input
= dn_nsp_rx
;
1382 rt
->u
.dst
.dev
= in_dev
;
1383 flags
|= RTCF_LOCAL
;
1386 case RTN_UNREACHABLE
:
1388 rt
->u
.dst
.input
= dst_discard
;
1390 rt
->rt_flags
= flags
;
1392 dev_hold(rt
->u
.dst
.dev
);
1394 err
= dn_rt_set_next_hop(rt
, &res
);
1398 hash
= dn_hash(rt
->fl
.fld_src
, rt
->fl
.fld_dst
);
1399 dn_insert_route(rt
, hash
, (struct dn_route
**)&skb
->dst
);
1403 neigh_release(neigh
);
1405 dn_fib_res_put(&res
);
1421 dst_free(&rt
->u
.dst
);
1425 static int dn_route_input(struct sk_buff
*skb
)
1427 struct dn_route
*rt
;
1428 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
1429 unsigned hash
= dn_hash(cb
->src
, cb
->dst
);
1435 for(rt
= rcu_dereference(dn_rt_hash_table
[hash
].chain
); rt
!= NULL
;
1436 rt
= rcu_dereference(rt
->u
.dst
.dn_next
)) {
1437 if ((rt
->fl
.fld_src
== cb
->src
) &&
1438 (rt
->fl
.fld_dst
== cb
->dst
) &&
1439 (rt
->fl
.oif
== 0) &&
1440 (rt
->fl
.mark
== skb
->mark
) &&
1441 (rt
->fl
.iif
== cb
->iif
)) {
1442 dst_use(&rt
->u
.dst
, jiffies
);
1444 skb
->dst
= (struct dst_entry
*)rt
;
1450 return dn_route_input_slow(skb
);
1453 static int dn_rt_fill_info(struct sk_buff
*skb
, u32 pid
, u32 seq
,
1454 int event
, int nowait
, unsigned int flags
)
1456 struct dn_route
*rt
= (struct dn_route
*)skb
->dst
;
1458 struct nlmsghdr
*nlh
;
1459 unsigned char *b
= skb_tail_pointer(skb
);
1462 nlh
= NLMSG_NEW(skb
, pid
, seq
, event
, sizeof(*r
), flags
);
1463 r
= NLMSG_DATA(nlh
);
1464 r
->rtm_family
= AF_DECnet
;
1465 r
->rtm_dst_len
= 16;
1468 r
->rtm_table
= RT_TABLE_MAIN
;
1469 RTA_PUT_U32(skb
, RTA_TABLE
, RT_TABLE_MAIN
);
1470 r
->rtm_type
= rt
->rt_type
;
1471 r
->rtm_flags
= (rt
->rt_flags
& ~0xFFFF) | RTM_F_CLONED
;
1472 r
->rtm_scope
= RT_SCOPE_UNIVERSE
;
1473 r
->rtm_protocol
= RTPROT_UNSPEC
;
1474 if (rt
->rt_flags
& RTCF_NOTIFY
)
1475 r
->rtm_flags
|= RTM_F_NOTIFY
;
1476 RTA_PUT(skb
, RTA_DST
, 2, &rt
->rt_daddr
);
1477 if (rt
->fl
.fld_src
) {
1478 r
->rtm_src_len
= 16;
1479 RTA_PUT(skb
, RTA_SRC
, 2, &rt
->fl
.fld_src
);
1482 RTA_PUT(skb
, RTA_OIF
, sizeof(int), &rt
->u
.dst
.dev
->ifindex
);
1484 * Note to self - change this if input routes reverse direction when
1485 * they deal only with inputs and not with replies like they do
1488 RTA_PUT(skb
, RTA_PREFSRC
, 2, &rt
->rt_local_src
);
1489 if (rt
->rt_daddr
!= rt
->rt_gateway
)
1490 RTA_PUT(skb
, RTA_GATEWAY
, 2, &rt
->rt_gateway
);
1491 if (rtnetlink_put_metrics(skb
, rt
->u
.dst
.metrics
) < 0)
1492 goto rtattr_failure
;
1493 expires
= rt
->u
.dst
.expires
? rt
->u
.dst
.expires
- jiffies
: 0;
1494 if (rtnl_put_cacheinfo(skb
, &rt
->u
.dst
, 0, 0, 0, expires
,
1495 rt
->u
.dst
.error
) < 0)
1496 goto rtattr_failure
;
1498 RTA_PUT(skb
, RTA_IIF
, sizeof(int), &rt
->fl
.iif
);
1500 nlh
->nlmsg_len
= skb_tail_pointer(skb
) - b
;
1510 * This is called by both endnodes and routers now.
1512 static int dn_cache_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
, void *arg
)
1514 struct net
*net
= sock_net(in_skb
->sk
);
1515 struct rtattr
**rta
= arg
;
1516 struct rtmsg
*rtm
= NLMSG_DATA(nlh
);
1517 struct dn_route
*rt
= NULL
;
1518 struct dn_skb_cb
*cb
;
1520 struct sk_buff
*skb
;
1523 if (net
!= &init_net
)
1526 memset(&fl
, 0, sizeof(fl
));
1527 fl
.proto
= DNPROTO_NSP
;
1529 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
1532 skb_reset_mac_header(skb
);
1533 cb
= DN_SKB_CB(skb
);
1536 memcpy(&fl
.fld_src
, RTA_DATA(rta
[RTA_SRC
-1]), 2);
1538 memcpy(&fl
.fld_dst
, RTA_DATA(rta
[RTA_DST
-1]), 2);
1540 memcpy(&fl
.iif
, RTA_DATA(rta
[RTA_IIF
-1]), sizeof(int));
1543 struct net_device
*dev
;
1544 if ((dev
= dev_get_by_index(&init_net
, fl
.iif
)) == NULL
) {
1553 skb
->protocol
= htons(ETH_P_DNA_RT
);
1555 cb
->src
= fl
.fld_src
;
1556 cb
->dst
= fl
.fld_dst
;
1558 err
= dn_route_input(skb
);
1560 memset(cb
, 0, sizeof(struct dn_skb_cb
));
1561 rt
= (struct dn_route
*)skb
->dst
;
1562 if (!err
&& -rt
->u
.dst
.error
)
1563 err
= rt
->u
.dst
.error
;
1566 if (rta
[RTA_OIF
- 1])
1567 memcpy(&oif
, RTA_DATA(rta
[RTA_OIF
- 1]), sizeof(int));
1569 err
= dn_route_output_key((struct dst_entry
**)&rt
, &fl
, 0);
1577 skb
->dst
= &rt
->u
.dst
;
1578 if (rtm
->rtm_flags
& RTM_F_NOTIFY
)
1579 rt
->rt_flags
|= RTCF_NOTIFY
;
1581 err
= dn_rt_fill_info(skb
, NETLINK_CB(in_skb
).pid
, nlh
->nlmsg_seq
, RTM_NEWROUTE
, 0, 0);
1590 return rtnl_unicast(skb
, &init_net
, NETLINK_CB(in_skb
).pid
);
1598 * For routers, this is called from dn_fib_dump, but for endnodes its
1599 * called directly from the rtnetlink dispatch table.
1601 int dn_cache_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1603 struct net
*net
= sock_net(skb
->sk
);
1604 struct dn_route
*rt
;
1608 if (net
!= &init_net
)
1611 if (NLMSG_PAYLOAD(cb
->nlh
, 0) < sizeof(struct rtmsg
))
1613 if (!(((struct rtmsg
*)NLMSG_DATA(cb
->nlh
))->rtm_flags
&RTM_F_CLONED
))
1617 s_idx
= idx
= cb
->args
[1];
1618 for(h
= 0; h
<= dn_rt_hash_mask
; h
++) {
1624 for(rt
= rcu_dereference(dn_rt_hash_table
[h
].chain
), idx
= 0;
1626 rt
= rcu_dereference(rt
->u
.dst
.dn_next
), idx
++) {
1629 skb
->dst
= dst_clone(&rt
->u
.dst
);
1630 if (dn_rt_fill_info(skb
, NETLINK_CB(cb
->skb
).pid
,
1631 cb
->nlh
->nlmsg_seq
, RTM_NEWROUTE
,
1632 1, NLM_F_MULTI
) <= 0) {
1633 dst_release(xchg(&skb
->dst
, NULL
));
1634 rcu_read_unlock_bh();
1637 dst_release(xchg(&skb
->dst
, NULL
));
1639 rcu_read_unlock_bh();
1648 #ifdef CONFIG_PROC_FS
1649 struct dn_rt_cache_iter_state
{
1653 static struct dn_route
*dn_rt_cache_get_first(struct seq_file
*seq
)
1655 struct dn_route
*rt
= NULL
;
1656 struct dn_rt_cache_iter_state
*s
= seq
->private;
1658 for(s
->bucket
= dn_rt_hash_mask
; s
->bucket
>= 0; --s
->bucket
) {
1660 rt
= dn_rt_hash_table
[s
->bucket
].chain
;
1663 rcu_read_unlock_bh();
1665 return rcu_dereference(rt
);
1668 static struct dn_route
*dn_rt_cache_get_next(struct seq_file
*seq
, struct dn_route
*rt
)
1670 struct dn_rt_cache_iter_state
*s
= seq
->private;
1672 rt
= rt
->u
.dst
.dn_next
;
1674 rcu_read_unlock_bh();
1675 if (--s
->bucket
< 0)
1678 rt
= dn_rt_hash_table
[s
->bucket
].chain
;
1680 return rcu_dereference(rt
);
1683 static void *dn_rt_cache_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1685 struct dn_route
*rt
= dn_rt_cache_get_first(seq
);
1688 while(*pos
&& (rt
= dn_rt_cache_get_next(seq
, rt
)))
1691 return *pos
? NULL
: rt
;
1694 static void *dn_rt_cache_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1696 struct dn_route
*rt
= dn_rt_cache_get_next(seq
, v
);
1701 static void dn_rt_cache_seq_stop(struct seq_file
*seq
, void *v
)
1704 rcu_read_unlock_bh();
1707 static int dn_rt_cache_seq_show(struct seq_file
*seq
, void *v
)
1709 struct dn_route
*rt
= v
;
1710 char buf1
[DN_ASCBUF_LEN
], buf2
[DN_ASCBUF_LEN
];
1712 seq_printf(seq
, "%-8s %-7s %-7s %04d %04d %04d\n",
1713 rt
->u
.dst
.dev
? rt
->u
.dst
.dev
->name
: "*",
1714 dn_addr2asc(le16_to_cpu(rt
->rt_daddr
), buf1
),
1715 dn_addr2asc(le16_to_cpu(rt
->rt_saddr
), buf2
),
1716 atomic_read(&rt
->u
.dst
.__refcnt
),
1718 (int) dst_metric(&rt
->u
.dst
, RTAX_RTT
));
1722 static const struct seq_operations dn_rt_cache_seq_ops
= {
1723 .start
= dn_rt_cache_seq_start
,
1724 .next
= dn_rt_cache_seq_next
,
1725 .stop
= dn_rt_cache_seq_stop
,
1726 .show
= dn_rt_cache_seq_show
,
1729 static int dn_rt_cache_seq_open(struct inode
*inode
, struct file
*file
)
1731 return seq_open_private(file
, &dn_rt_cache_seq_ops
,
1732 sizeof(struct dn_rt_cache_iter_state
));
1735 static const struct file_operations dn_rt_cache_seq_fops
= {
1736 .owner
= THIS_MODULE
,
1737 .open
= dn_rt_cache_seq_open
,
1739 .llseek
= seq_lseek
,
1740 .release
= seq_release_private
,
1743 #endif /* CONFIG_PROC_FS */
1745 void __init
dn_route_init(void)
1749 dn_dst_ops
.kmem_cachep
=
1750 kmem_cache_create("dn_dst_cache", sizeof(struct dn_route
), 0,
1751 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
1752 setup_timer(&dn_route_timer
, dn_dst_check_expire
, 0);
1753 dn_route_timer
.expires
= jiffies
+ decnet_dst_gc_interval
* HZ
;
1754 add_timer(&dn_route_timer
);
1756 goal
= num_physpages
>> (26 - PAGE_SHIFT
);
1758 for(order
= 0; (1UL << order
) < goal
; order
++)
1762 * Only want 1024 entries max, since the table is very, very unlikely
1763 * to be larger than that.
1765 while(order
&& ((((1UL << order
) * PAGE_SIZE
) /
1766 sizeof(struct dn_rt_hash_bucket
)) >= 2048))
1770 dn_rt_hash_mask
= (1UL << order
) * PAGE_SIZE
/
1771 sizeof(struct dn_rt_hash_bucket
);
1772 while(dn_rt_hash_mask
& (dn_rt_hash_mask
- 1))
1774 dn_rt_hash_table
= (struct dn_rt_hash_bucket
*)
1775 __get_free_pages(GFP_ATOMIC
, order
);
1776 } while (dn_rt_hash_table
== NULL
&& --order
> 0);
1778 if (!dn_rt_hash_table
)
1779 panic("Failed to allocate DECnet route cache hash table\n");
1782 "DECnet: Routing cache hash table of %u buckets, %ldKbytes\n",
1784 (long)(dn_rt_hash_mask
*sizeof(struct dn_rt_hash_bucket
))/1024);
1787 for(i
= 0; i
<= dn_rt_hash_mask
; i
++) {
1788 spin_lock_init(&dn_rt_hash_table
[i
].lock
);
1789 dn_rt_hash_table
[i
].chain
= NULL
;
1792 dn_dst_ops
.gc_thresh
= (dn_rt_hash_mask
+ 1);
1794 proc_net_fops_create(&init_net
, "decnet_cache", S_IRUGO
, &dn_rt_cache_seq_fops
);
1796 #ifdef CONFIG_DECNET_ROUTER
1797 rtnl_register(PF_DECnet
, RTM_GETROUTE
, dn_cache_getroute
, dn_fib_dump
);
1799 rtnl_register(PF_DECnet
, RTM_GETROUTE
, dn_cache_getroute
,
1804 void __exit
dn_route_cleanup(void)
1806 del_timer(&dn_route_timer
);
1809 proc_net_remove(&init_net
, "decnet_cache");