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/config.h>
59 #include <linux/errno.h>
60 #include <linux/types.h>
61 #include <linux/socket.h>
63 #include <linux/kernel.h>
64 #include <linux/sockios.h>
65 #include <linux/net.h>
66 #include <linux/netdevice.h>
67 #include <linux/inet.h>
68 #include <linux/route.h>
69 #include <linux/in_route.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 <asm/errno.h>
81 #include <net/neighbour.h>
85 #include <net/dn_dev.h>
86 #include <net/dn_nsp.h>
87 #include <net/dn_route.h>
88 #include <net/dn_neigh.h>
89 #include <net/dn_fib.h>
91 struct dn_rt_hash_bucket
93 struct dn_route
*chain
;
95 } __attribute__((__aligned__(8)));
97 extern struct neigh_table dn_neigh_table
;
100 static unsigned char dn_hiord_addr
[6] = {0xAA,0x00,0x04,0x00,0x00,0x00};
102 static const int dn_rt_min_delay
= 2 * HZ
;
103 static const int dn_rt_max_delay
= 10 * HZ
;
104 static const int dn_rt_mtu_expires
= 10 * 60 * HZ
;
106 static unsigned long dn_rt_deadline
;
108 static int dn_dst_gc(void);
109 static struct dst_entry
*dn_dst_check(struct dst_entry
*, __u32
);
110 static struct dst_entry
*dn_dst_negative_advice(struct dst_entry
*);
111 static void dn_dst_link_failure(struct sk_buff
*);
112 static void dn_dst_update_pmtu(struct dst_entry
*dst
, u32 mtu
);
113 static int dn_route_input(struct sk_buff
*);
114 static void dn_run_flush(unsigned long dummy
);
116 static struct dn_rt_hash_bucket
*dn_rt_hash_table
;
117 static unsigned dn_rt_hash_mask
;
119 static struct timer_list dn_route_timer
;
120 static DEFINE_TIMER(dn_rt_flush_timer
, dn_run_flush
, 0, 0);
121 int decnet_dst_gc_interval
= 2;
123 static struct dst_ops dn_dst_ops
= {
125 .protocol
= __constant_htons(ETH_P_DNA_RT
),
128 .check
= dn_dst_check
,
129 .negative_advice
= dn_dst_negative_advice
,
130 .link_failure
= dn_dst_link_failure
,
131 .update_pmtu
= dn_dst_update_pmtu
,
132 .entry_size
= sizeof(struct dn_route
),
133 .entries
= ATOMIC_INIT(0),
136 static __inline__
unsigned dn_hash(__le16 src
, __le16 dst
)
138 __u16 tmp
= (__u16 __force
)(src
^ dst
);
142 return dn_rt_hash_mask
& (unsigned)tmp
;
145 static inline void dnrt_free(struct dn_route
*rt
)
147 call_rcu_bh(&rt
->u
.dst
.rcu_head
, dst_rcu_free
);
150 static inline void dnrt_drop(struct dn_route
*rt
)
152 dst_release(&rt
->u
.dst
);
153 call_rcu_bh(&rt
->u
.dst
.rcu_head
, dst_rcu_free
);
156 static void dn_dst_check_expire(unsigned long dummy
)
159 struct dn_route
*rt
, **rtp
;
160 unsigned long now
= jiffies
;
161 unsigned long expire
= 120 * HZ
;
163 for(i
= 0; i
<= dn_rt_hash_mask
; i
++) {
164 rtp
= &dn_rt_hash_table
[i
].chain
;
166 spin_lock(&dn_rt_hash_table
[i
].lock
);
167 while((rt
=*rtp
) != NULL
) {
168 if (atomic_read(&rt
->u
.dst
.__refcnt
) ||
169 (now
- rt
->u
.dst
.lastuse
) < expire
) {
170 rtp
= &rt
->u
.rt_next
;
173 *rtp
= rt
->u
.rt_next
;
174 rt
->u
.rt_next
= NULL
;
177 spin_unlock(&dn_rt_hash_table
[i
].lock
);
179 if ((jiffies
- now
) > 0)
183 mod_timer(&dn_route_timer
, now
+ decnet_dst_gc_interval
* HZ
);
186 static int dn_dst_gc(void)
188 struct dn_route
*rt
, **rtp
;
190 unsigned long now
= jiffies
;
191 unsigned long expire
= 10 * HZ
;
193 for(i
= 0; i
<= dn_rt_hash_mask
; i
++) {
195 spin_lock_bh(&dn_rt_hash_table
[i
].lock
);
196 rtp
= &dn_rt_hash_table
[i
].chain
;
198 while((rt
=*rtp
) != NULL
) {
199 if (atomic_read(&rt
->u
.dst
.__refcnt
) ||
200 (now
- rt
->u
.dst
.lastuse
) < expire
) {
201 rtp
= &rt
->u
.rt_next
;
204 *rtp
= rt
->u
.rt_next
;
205 rt
->u
.rt_next
= NULL
;
209 spin_unlock_bh(&dn_rt_hash_table
[i
].lock
);
216 * The decnet standards don't impose a particular minimum mtu, what they
217 * do insist on is that the routing layer accepts a datagram of at least
218 * 230 bytes long. Here we have to subtract the routing header length from
219 * 230 to get the minimum acceptable mtu. If there is no neighbour, then we
220 * assume the worst and use a long header size.
222 * We update both the mtu and the advertised mss (i.e. the segment size we
223 * advertise to the other end).
225 static void dn_dst_update_pmtu(struct dst_entry
*dst
, u32 mtu
)
228 struct dn_dev
*dn
= dst
->neighbour
?
229 (struct dn_dev
*)dst
->neighbour
->dev
->dn_ptr
: NULL
;
231 if (dn
&& dn
->use_long
== 0)
236 if (dst
->metrics
[RTAX_MTU
-1] > mtu
&& mtu
>= min_mtu
) {
237 if (!(dst_metric_locked(dst
, RTAX_MTU
))) {
238 dst
->metrics
[RTAX_MTU
-1] = mtu
;
239 dst_set_expires(dst
, dn_rt_mtu_expires
);
241 if (!(dst_metric_locked(dst
, RTAX_ADVMSS
))) {
242 u32 mss
= mtu
- DN_MAX_NSP_DATA_HEADER
;
243 if (dst
->metrics
[RTAX_ADVMSS
-1] > mss
)
244 dst
->metrics
[RTAX_ADVMSS
-1] = mss
;
250 * When a route has been marked obsolete. (e.g. routing cache flush)
252 static struct dst_entry
*dn_dst_check(struct dst_entry
*dst
, __u32 cookie
)
257 static struct dst_entry
*dn_dst_negative_advice(struct dst_entry
*dst
)
263 static void dn_dst_link_failure(struct sk_buff
*skb
)
268 static inline int compare_keys(struct flowi
*fl1
, struct flowi
*fl2
)
270 return memcmp(&fl1
->nl_u
.dn_u
, &fl2
->nl_u
.dn_u
, sizeof(fl1
->nl_u
.dn_u
)) == 0 &&
271 fl1
->oif
== fl2
->oif
&&
272 fl1
->iif
== fl2
->iif
;
275 static int dn_insert_route(struct dn_route
*rt
, unsigned hash
, struct dn_route
**rp
)
277 struct dn_route
*rth
, **rthp
;
278 unsigned long now
= jiffies
;
280 rthp
= &dn_rt_hash_table
[hash
].chain
;
282 spin_lock_bh(&dn_rt_hash_table
[hash
].lock
);
283 while((rth
= *rthp
) != NULL
) {
284 if (compare_keys(&rth
->fl
, &rt
->fl
)) {
286 *rthp
= rth
->u
.rt_next
;
287 rcu_assign_pointer(rth
->u
.rt_next
,
288 dn_rt_hash_table
[hash
].chain
);
289 rcu_assign_pointer(dn_rt_hash_table
[hash
].chain
, rth
);
292 dst_hold(&rth
->u
.dst
);
293 rth
->u
.dst
.lastuse
= now
;
294 spin_unlock_bh(&dn_rt_hash_table
[hash
].lock
);
300 rthp
= &rth
->u
.rt_next
;
303 rcu_assign_pointer(rt
->u
.rt_next
, dn_rt_hash_table
[hash
].chain
);
304 rcu_assign_pointer(dn_rt_hash_table
[hash
].chain
, rt
);
306 dst_hold(&rt
->u
.dst
);
308 rt
->u
.dst
.lastuse
= now
;
309 spin_unlock_bh(&dn_rt_hash_table
[hash
].lock
);
314 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
.rt_next
;
327 rt
->u
.rt_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
->nh
.raw
);
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
->nh
.raw
);
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 dn_ntohs(cb
->src
), dn_ntohs(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
->h
.raw
= skb
->data
;
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
->h
.raw
= skb
->data
;
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
= dn_ntohs(*(__le16
*)skb
->data
);
579 struct dn_dev
*dn
= (struct dn_dev
*)dev
->dn_ptr
;
580 unsigned char padlen
= 0;
585 if ((skb
= skb_share_check(skb
, GFP_ATOMIC
)) == NULL
)
588 if (!pskb_may_pull(skb
, 3))
602 cb
->iif
= dev
->ifindex
;
605 * If we have padding, remove it.
607 if (flags
& DN_RT_F_PF
) {
608 padlen
= flags
& ~DN_RT_F_PF
;
609 if (!pskb_may_pull(skb
, padlen
+ 1))
611 skb_pull(skb
, padlen
);
615 skb
->nh
.raw
= skb
->data
;
618 * Weed out future version DECnet
620 if (flags
& DN_RT_F_VER
)
623 cb
->rt_flags
= flags
;
625 if (decnet_debug_level
& 1)
627 "dn_route_rcv: got 0x%02x from %s [%d %d %d]\n",
628 (int)flags
, (dev
) ? dev
->name
: "???", len
, skb
->len
,
631 if (flags
& DN_RT_PKT_CNTL
) {
632 if (unlikely(skb_linearize(skb
)))
635 switch(flags
& DN_RT_CNTL_MSK
) {
637 dn_dev_init_pkt(skb
);
640 dn_dev_veri_pkt(skb
);
644 if (dn
->parms
.state
!= DN_DEV_S_RU
)
647 switch(flags
& DN_RT_CNTL_MSK
) {
649 return NF_HOOK(PF_DECnet
, NF_DN_HELLO
, skb
, skb
->dev
, NULL
, dn_route_ptp_hello
);
653 return NF_HOOK(PF_DECnet
, NF_DN_ROUTE
, skb
, skb
->dev
, NULL
, dn_route_discard
);
655 return NF_HOOK(PF_DECnet
, NF_DN_HELLO
, skb
, skb
->dev
, NULL
, dn_neigh_router_hello
);
658 return NF_HOOK(PF_DECnet
, NF_DN_HELLO
, skb
, skb
->dev
, NULL
, dn_neigh_endnode_hello
);
661 if (dn
->parms
.state
!= DN_DEV_S_RU
)
664 skb_pull(skb
, 1); /* Pull flags */
666 switch(flags
& DN_RT_PKT_MSK
) {
668 return dn_route_rx_long(skb
);
669 case DN_RT_PKT_SHORT
:
670 return dn_route_rx_short(skb
);
680 static int dn_output(struct sk_buff
*skb
)
682 struct dst_entry
*dst
= skb
->dst
;
683 struct dn_route
*rt
= (struct dn_route
*)dst
;
684 struct net_device
*dev
= dst
->dev
;
685 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
686 struct neighbour
*neigh
;
690 if ((neigh
= dst
->neighbour
) == NULL
)
695 cb
->src
= rt
->rt_saddr
;
696 cb
->dst
= rt
->rt_daddr
;
699 * Always set the Intra-Ethernet bit on all outgoing packets
700 * originated on this node. Only valid flag from upper layers
701 * is return-to-sender-requested. Set hop count to 0 too.
703 cb
->rt_flags
&= ~DN_RT_F_RQR
;
704 cb
->rt_flags
|= DN_RT_F_IE
;
707 return NF_HOOK(PF_DECnet
, NF_DN_LOCAL_OUT
, skb
, NULL
, dev
, neigh
->output
);
711 printk(KERN_DEBUG
"dn_output: This should not happen\n");
718 static int dn_forward(struct sk_buff
*skb
)
720 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
721 struct dst_entry
*dst
= skb
->dst
;
722 struct dn_dev
*dn_db
= dst
->dev
->dn_ptr
;
724 struct neighbour
*neigh
= dst
->neighbour
;
726 #ifdef CONFIG_NETFILTER
727 struct net_device
*dev
= skb
->dev
;
730 if (skb
->pkt_type
!= PACKET_HOST
)
733 /* Ensure that we have enough space for headers */
734 rt
= (struct dn_route
*)skb
->dst
;
735 header_len
= dn_db
->use_long
? 21 : 6;
736 if (skb_cow(skb
, LL_RESERVED_SPACE(rt
->u
.dst
.dev
)+header_len
))
740 * Hop count exceeded.
745 skb
->dev
= rt
->u
.dst
.dev
;
748 * If packet goes out same interface it came in on, then set
749 * the Intra-Ethernet bit. This has no effect for short
750 * packets, so we don't need to test for them here.
752 cb
->rt_flags
&= ~DN_RT_F_IE
;
753 if (rt
->rt_flags
& RTCF_DOREDIRECT
)
754 cb
->rt_flags
|= DN_RT_F_IE
;
756 return NF_HOOK(PF_DECnet
, NF_DN_FORWARD
, skb
, dev
, skb
->dev
, neigh
->output
);
764 * Drop packet. This is used for endnodes and for
765 * when we should not be forwarding packets from
768 static int dn_blackhole(struct sk_buff
*skb
)
775 * Used to catch bugs. This should never normally get
778 static int dn_rt_bug(struct sk_buff
*skb
)
780 if (net_ratelimit()) {
781 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
783 printk(KERN_DEBUG
"dn_rt_bug: skb from:%04x to:%04x\n",
784 dn_ntohs(cb
->src
), dn_ntohs(cb
->dst
));
792 static int dn_rt_set_next_hop(struct dn_route
*rt
, struct dn_fib_res
*res
)
794 struct dn_fib_info
*fi
= res
->fi
;
795 struct net_device
*dev
= rt
->u
.dst
.dev
;
800 if (DN_FIB_RES_GW(*res
) &&
801 DN_FIB_RES_NH(*res
).nh_scope
== RT_SCOPE_LINK
)
802 rt
->rt_gateway
= DN_FIB_RES_GW(*res
);
803 memcpy(rt
->u
.dst
.metrics
, fi
->fib_metrics
,
804 sizeof(rt
->u
.dst
.metrics
));
806 rt
->rt_type
= res
->type
;
808 if (dev
!= NULL
&& rt
->u
.dst
.neighbour
== NULL
) {
809 n
= __neigh_lookup_errno(&dn_neigh_table
, &rt
->rt_gateway
, dev
);
812 rt
->u
.dst
.neighbour
= n
;
815 if (rt
->u
.dst
.metrics
[RTAX_MTU
-1] == 0 ||
816 rt
->u
.dst
.metrics
[RTAX_MTU
-1] > rt
->u
.dst
.dev
->mtu
)
817 rt
->u
.dst
.metrics
[RTAX_MTU
-1] = rt
->u
.dst
.dev
->mtu
;
818 mss
= dn_mss_from_pmtu(dev
, dst_mtu(&rt
->u
.dst
));
819 if (rt
->u
.dst
.metrics
[RTAX_ADVMSS
-1] == 0 ||
820 rt
->u
.dst
.metrics
[RTAX_ADVMSS
-1] > mss
)
821 rt
->u
.dst
.metrics
[RTAX_ADVMSS
-1] = mss
;
825 static inline int dn_match_addr(__le16 addr1
, __le16 addr2
)
827 __u16 tmp
= dn_ntohs(addr1
) ^ dn_ntohs(addr2
);
836 static __le16
dnet_select_source(const struct net_device
*dev
, __le16 daddr
, int scope
)
839 struct dn_dev
*dn_db
= dev
->dn_ptr
;
840 struct dn_ifaddr
*ifa
;
844 read_lock(&dev_base_lock
);
845 for(ifa
= dn_db
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
) {
846 if (ifa
->ifa_scope
> scope
)
849 saddr
= ifa
->ifa_local
;
852 ret
= dn_match_addr(daddr
, ifa
->ifa_local
);
853 if (ret
> best_match
)
854 saddr
= ifa
->ifa_local
;
856 saddr
= ifa
->ifa_local
;
858 read_unlock(&dev_base_lock
);
863 static inline __le16
__dn_fib_res_prefsrc(struct dn_fib_res
*res
)
865 return dnet_select_source(DN_FIB_RES_DEV(*res
), DN_FIB_RES_GW(*res
), res
->scope
);
868 static inline __le16
dn_fib_rules_map_destination(__le16 daddr
, struct dn_fib_res
*res
)
870 __le16 mask
= dnet_make_mask(res
->prefixlen
);
871 return (daddr
&~mask
)|res
->fi
->fib_nh
->nh_gw
;
874 static int dn_route_output_slow(struct dst_entry
**pprt
, const struct flowi
*oldflp
, int try_hard
)
876 struct flowi fl
= { .nl_u
= { .dn_u
=
877 { .daddr
= oldflp
->fld_dst
,
878 .saddr
= oldflp
->fld_src
,
879 .scope
= RT_SCOPE_UNIVERSE
,
880 #ifdef CONFIG_DECNET_ROUTE_FWMARK
881 .fwmark
= oldflp
->fld_fwmark
884 .iif
= loopback_dev
.ifindex
,
885 .oif
= oldflp
->oif
};
886 struct dn_route
*rt
= NULL
;
887 struct net_device
*dev_out
= NULL
;
888 struct neighbour
*neigh
= NULL
;
891 struct dn_fib_res res
= { .fi
= NULL
, .type
= RTN_UNICAST
};
896 if (decnet_debug_level
& 16)
898 "dn_route_output_slow: dst=%04x src=%04x mark=%d"
899 " iif=%d oif=%d\n", dn_ntohs(oldflp
->fld_dst
),
900 dn_ntohs(oldflp
->fld_src
),
901 oldflp
->fld_fwmark
, loopback_dev
.ifindex
, oldflp
->oif
);
903 /* If we have an output interface, verify its a DECnet device */
905 dev_out
= dev_get_by_index(oldflp
->oif
);
907 if (dev_out
&& dev_out
->dn_ptr
== NULL
) {
915 /* If we have a source address, verify that its a local address */
916 if (oldflp
->fld_src
) {
917 err
= -EADDRNOTAVAIL
;
920 if (dn_dev_islocal(dev_out
, oldflp
->fld_src
))
925 read_lock(&dev_base_lock
);
926 for(dev_out
= dev_base
; dev_out
; dev_out
= dev_out
->next
) {
927 if (!dev_out
->dn_ptr
)
929 if (dn_dev_islocal(dev_out
, oldflp
->fld_src
))
932 read_unlock(&dev_base_lock
);
940 /* No destination? Assume its local */
942 fl
.fld_dst
= fl
.fld_src
;
944 err
= -EADDRNOTAVAIL
;
947 dev_out
= &loopback_dev
;
951 fl
.fld_src
= dnet_select_source(dev_out
, 0,
956 fl
.oif
= loopback_dev
.ifindex
;
957 res
.type
= RTN_LOCAL
;
961 if (decnet_debug_level
& 16)
963 "dn_route_output_slow: initial checks complete."
964 " dst=%o4x src=%04x oif=%d try_hard=%d\n",
965 dn_ntohs(fl
.fld_dst
), dn_ntohs(fl
.fld_src
),
969 * N.B. If the kernel is compiled without router support then
970 * dn_fib_lookup() will evaluate to non-zero so this if () block
971 * will always be executed.
974 if (try_hard
|| (err
= dn_fib_lookup(&fl
, &res
)) != 0) {
975 struct dn_dev
*dn_db
;
979 * Here the fallback is basically the standard algorithm for
980 * routing in endnodes which is described in the DECnet routing
983 * If we are not trying hard, look in neighbour cache.
984 * The result is tested to ensure that if a specific output
985 * device/source address was requested, then we honour that
989 neigh
= neigh_lookup_nodev(&dn_neigh_table
, &fl
.fld_dst
);
992 (neigh
->dev
->ifindex
!= oldflp
->oif
)) ||
994 (!dn_dev_islocal(neigh
->dev
,
995 oldflp
->fld_src
)))) {
996 neigh_release(neigh
);
1001 if (dn_dev_islocal(neigh
->dev
, fl
.fld_dst
)) {
1002 dev_out
= &loopback_dev
;
1003 res
.type
= RTN_LOCAL
;
1005 dev_out
= neigh
->dev
;
1013 /* Not there? Perhaps its a local address */
1014 if (dev_out
== NULL
)
1015 dev_out
= dn_dev_get_default();
1017 if (dev_out
== NULL
)
1019 dn_db
= dev_out
->dn_ptr
;
1020 /* Possible improvement - check all devices for local addr */
1021 if (dn_dev_islocal(dev_out
, fl
.fld_dst
)) {
1023 dev_out
= &loopback_dev
;
1025 res
.type
= RTN_LOCAL
;
1028 /* Not local either.... try sending it to the default router */
1029 neigh
= neigh_clone(dn_db
->router
);
1030 BUG_ON(neigh
&& neigh
->dev
!= dev_out
);
1032 /* Ok then, we assume its directly connected and move on */
1035 gateway
= ((struct dn_neigh
*)neigh
)->addr
;
1037 gateway
= fl
.fld_dst
;
1038 if (fl
.fld_src
== 0) {
1039 fl
.fld_src
= dnet_select_source(dev_out
, gateway
,
1040 res
.type
== RTN_LOCAL
?
1043 if (fl
.fld_src
== 0 && res
.type
!= RTN_LOCAL
)
1046 fl
.oif
= dev_out
->ifindex
;
1051 if (res
.type
== RTN_NAT
)
1054 if (res
.type
== RTN_LOCAL
) {
1056 fl
.fld_src
= fl
.fld_dst
;
1059 dev_out
= &loopback_dev
;
1061 fl
.oif
= dev_out
->ifindex
;
1063 dn_fib_info_put(res
.fi
);
1068 if (res
.fi
->fib_nhs
> 1 && fl
.oif
== 0)
1069 dn_fib_select_multipath(&fl
, &res
);
1072 * We could add some logic to deal with default routes here and
1073 * get rid of some of the special casing above.
1077 fl
.fld_src
= DN_FIB_RES_PREFSRC(res
);
1081 dev_out
= DN_FIB_RES_DEV(res
);
1083 fl
.oif
= dev_out
->ifindex
;
1084 gateway
= DN_FIB_RES_GW(res
);
1087 if (dev_out
->flags
& IFF_LOOPBACK
)
1088 flags
|= RTCF_LOCAL
;
1090 rt
= dst_alloc(&dn_dst_ops
);
1094 atomic_set(&rt
->u
.dst
.__refcnt
, 1);
1095 rt
->u
.dst
.flags
= DST_HOST
;
1097 rt
->fl
.fld_src
= oldflp
->fld_src
;
1098 rt
->fl
.fld_dst
= oldflp
->fld_dst
;
1099 rt
->fl
.oif
= oldflp
->oif
;
1101 #ifdef CONFIG_DECNET_ROUTE_FWMARK
1102 rt
->fl
.fld_fwmark
= oldflp
->fld_fwmark
;
1105 rt
->rt_saddr
= fl
.fld_src
;
1106 rt
->rt_daddr
= fl
.fld_dst
;
1107 rt
->rt_gateway
= gateway
? gateway
: fl
.fld_dst
;
1108 rt
->rt_local_src
= fl
.fld_src
;
1110 rt
->rt_dst_map
= fl
.fld_dst
;
1111 rt
->rt_src_map
= fl
.fld_src
;
1113 rt
->u
.dst
.dev
= dev_out
;
1115 rt
->u
.dst
.neighbour
= neigh
;
1118 rt
->u
.dst
.lastuse
= jiffies
;
1119 rt
->u
.dst
.output
= dn_output
;
1120 rt
->u
.dst
.input
= dn_rt_bug
;
1121 rt
->rt_flags
= flags
;
1122 if (flags
& RTCF_LOCAL
)
1123 rt
->u
.dst
.input
= dn_nsp_rx
;
1125 err
= dn_rt_set_next_hop(rt
, &res
);
1129 hash
= dn_hash(rt
->fl
.fld_src
, rt
->fl
.fld_dst
);
1130 dn_insert_route(rt
, hash
, (struct dn_route
**)pprt
);
1134 neigh_release(neigh
);
1136 dn_fib_res_put(&res
);
1143 err
= -EADDRNOTAVAIL
;
1152 dst_free(&rt
->u
.dst
);
1158 * N.B. The flags may be moved into the flowi at some future stage.
1160 static int __dn_route_output_key(struct dst_entry
**pprt
, const struct flowi
*flp
, int flags
)
1162 unsigned hash
= dn_hash(flp
->fld_src
, flp
->fld_dst
);
1163 struct dn_route
*rt
= NULL
;
1165 if (!(flags
& MSG_TRYHARD
)) {
1167 for(rt
= rcu_dereference(dn_rt_hash_table
[hash
].chain
); rt
;
1168 rt
= rcu_dereference(rt
->u
.rt_next
)) {
1169 if ((flp
->fld_dst
== rt
->fl
.fld_dst
) &&
1170 (flp
->fld_src
== rt
->fl
.fld_src
) &&
1171 #ifdef CONFIG_DECNET_ROUTE_FWMARK
1172 (flp
->fld_fwmark
== rt
->fl
.fld_fwmark
) &&
1174 (rt
->fl
.iif
== 0) &&
1175 (rt
->fl
.oif
== flp
->oif
)) {
1176 rt
->u
.dst
.lastuse
= jiffies
;
1177 dst_hold(&rt
->u
.dst
);
1179 rcu_read_unlock_bh();
1184 rcu_read_unlock_bh();
1187 return dn_route_output_slow(pprt
, flp
, flags
);
1190 static int dn_route_output_key(struct dst_entry
**pprt
, struct flowi
*flp
, int flags
)
1194 err
= __dn_route_output_key(pprt
, flp
, flags
);
1195 if (err
== 0 && flp
->proto
) {
1196 err
= xfrm_lookup(pprt
, flp
, NULL
, 0);
1201 int dn_route_output_sock(struct dst_entry
**pprt
, struct flowi
*fl
, struct sock
*sk
, int flags
)
1205 err
= __dn_route_output_key(pprt
, fl
, flags
& MSG_TRYHARD
);
1206 if (err
== 0 && fl
->proto
) {
1207 err
= xfrm_lookup(pprt
, fl
, sk
, !(flags
& MSG_DONTWAIT
));
1212 static int dn_route_input_slow(struct sk_buff
*skb
)
1214 struct dn_route
*rt
= NULL
;
1215 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
1216 struct net_device
*in_dev
= skb
->dev
;
1217 struct net_device
*out_dev
= NULL
;
1218 struct dn_dev
*dn_db
;
1219 struct neighbour
*neigh
= NULL
;
1223 __le16 local_src
= 0;
1224 struct flowi fl
= { .nl_u
= { .dn_u
=
1227 .scope
= RT_SCOPE_UNIVERSE
,
1228 #ifdef CONFIG_DECNET_ROUTE_FWMARK
1229 .fwmark
= skb
->nfmark
1232 .iif
= skb
->dev
->ifindex
};
1233 struct dn_fib_res res
= { .fi
= NULL
, .type
= RTN_UNREACHABLE
};
1239 if ((dn_db
= in_dev
->dn_ptr
) == NULL
)
1242 /* Zero source addresses are not allowed */
1243 if (fl
.fld_src
== 0)
1247 * In this case we've just received a packet from a source
1248 * outside ourselves pretending to come from us. We don't
1249 * allow it any further to prevent routing loops, spoofing and
1250 * other nasties. Loopback packets already have the dst attached
1251 * so this only affects packets which have originated elsewhere.
1254 if (dn_dev_islocal(in_dev
, cb
->src
))
1257 err
= dn_fib_lookup(&fl
, &res
);
1262 * Is the destination us ?
1264 if (!dn_dev_islocal(in_dev
, cb
->dst
))
1267 res
.type
= RTN_LOCAL
;
1268 flags
|= RTCF_DIRECTSRC
;
1270 __le16 src_map
= fl
.fld_src
;
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");
1283 src_map
= dn_fib_rules_policy(fl
.fld_src
, &res
, &flags
);
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
);
1290 if (dn_fib_lookup(&fl
, &res
))
1293 if (res
.type
!= RTN_UNICAST
)
1296 gateway
= fl
.fld_dst
;
1298 fl
.fld_src
= src_map
;
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
1309 if (dn_db
->parms
.forwarding
== 0)
1312 if (res
.fi
->fib_nhs
> 1 && fl
.oif
== 0)
1313 dn_fib_select_multipath(&fl
, &res
);
1316 * 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.
1321 if (out_dev
== in_dev
&& !(flags
& RTCF_NAT
))
1322 flags
|= RTCF_DOREDIRECT
;
1324 local_src
= DN_FIB_RES_PREFSRC(res
);
1327 case RTN_UNREACHABLE
:
1330 flags
|= RTCF_LOCAL
;
1331 fl
.fld_src
= cb
->dst
;
1332 fl
.fld_dst
= cb
->src
;
1334 /* Routing tables gave us a gateway */
1338 /* Packet was intra-ethernet, so we know its on-link */
1339 if (cb
->rt_flags
| DN_RT_F_IE
) {
1341 flags
|= RTCF_DIRECTSRC
;
1345 /* Use the default router if there is one */
1346 neigh
= neigh_clone(dn_db
->router
);
1348 gateway
= ((struct dn_neigh
*)neigh
)->addr
;
1352 /* Close eyes and pray */
1354 flags
|= RTCF_DIRECTSRC
;
1361 rt
= dst_alloc(&dn_dst_ops
);
1365 rt
->rt_saddr
= fl
.fld_src
;
1366 rt
->rt_daddr
= fl
.fld_dst
;
1367 rt
->rt_gateway
= fl
.fld_dst
;
1369 rt
->rt_gateway
= gateway
;
1370 rt
->rt_local_src
= local_src
? local_src
: rt
->rt_saddr
;
1372 rt
->rt_dst_map
= fl
.fld_dst
;
1373 rt
->rt_src_map
= fl
.fld_src
;
1375 rt
->fl
.fld_src
= cb
->src
;
1376 rt
->fl
.fld_dst
= cb
->dst
;
1378 rt
->fl
.iif
= in_dev
->ifindex
;
1379 rt
->fl
.fld_fwmark
= fl
.fld_fwmark
;
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
;
1388 rt
->u
.dst
.input
= dn_forward
;
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
;
1397 case RTN_UNREACHABLE
:
1399 rt
->u
.dst
.input
= dn_blackhole
;
1401 rt
->rt_flags
= flags
;
1403 dev_hold(rt
->u
.dst
.dev
);
1405 err
= dn_rt_set_next_hop(rt
, &res
);
1409 hash
= dn_hash(rt
->fl
.fld_src
, rt
->fl
.fld_dst
);
1410 dn_insert_route(rt
, hash
, (struct dn_route
**)&skb
->dst
);
1414 neigh_release(neigh
);
1416 dn_fib_res_put(&res
);
1432 dst_free(&rt
->u
.dst
);
1436 int dn_route_input(struct sk_buff
*skb
)
1438 struct dn_route
*rt
;
1439 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
1440 unsigned hash
= dn_hash(cb
->src
, cb
->dst
);
1446 for(rt
= rcu_dereference(dn_rt_hash_table
[hash
].chain
); rt
!= NULL
;
1447 rt
= rcu_dereference(rt
->u
.rt_next
)) {
1448 if ((rt
->fl
.fld_src
== cb
->src
) &&
1449 (rt
->fl
.fld_dst
== cb
->dst
) &&
1450 (rt
->fl
.oif
== 0) &&
1451 #ifdef CONFIG_DECNET_ROUTE_FWMARK
1452 (rt
->fl
.fld_fwmark
== skb
->nfmark
) &&
1454 (rt
->fl
.iif
== cb
->iif
)) {
1455 rt
->u
.dst
.lastuse
= jiffies
;
1456 dst_hold(&rt
->u
.dst
);
1459 skb
->dst
= (struct dst_entry
*)rt
;
1465 return dn_route_input_slow(skb
);
1468 static int dn_rt_fill_info(struct sk_buff
*skb
, u32 pid
, u32 seq
,
1469 int event
, int nowait
, unsigned int flags
)
1471 struct dn_route
*rt
= (struct dn_route
*)skb
->dst
;
1473 struct nlmsghdr
*nlh
;
1474 unsigned char *b
= skb
->tail
;
1475 struct rta_cacheinfo ci
;
1477 nlh
= NLMSG_NEW(skb
, pid
, seq
, event
, sizeof(*r
), flags
);
1478 r
= NLMSG_DATA(nlh
);
1479 r
->rtm_family
= AF_DECnet
;
1480 r
->rtm_dst_len
= 16;
1483 r
->rtm_table
= RT_TABLE_MAIN
;
1484 r
->rtm_type
= rt
->rt_type
;
1485 r
->rtm_flags
= (rt
->rt_flags
& ~0xFFFF) | RTM_F_CLONED
;
1486 r
->rtm_scope
= RT_SCOPE_UNIVERSE
;
1487 r
->rtm_protocol
= RTPROT_UNSPEC
;
1488 if (rt
->rt_flags
& RTCF_NOTIFY
)
1489 r
->rtm_flags
|= RTM_F_NOTIFY
;
1490 RTA_PUT(skb
, RTA_DST
, 2, &rt
->rt_daddr
);
1491 if (rt
->fl
.fld_src
) {
1492 r
->rtm_src_len
= 16;
1493 RTA_PUT(skb
, RTA_SRC
, 2, &rt
->fl
.fld_src
);
1496 RTA_PUT(skb
, RTA_OIF
, sizeof(int), &rt
->u
.dst
.dev
->ifindex
);
1498 * Note to self - change this if input routes reverse direction when
1499 * they deal only with inputs and not with replies like they do
1502 RTA_PUT(skb
, RTA_PREFSRC
, 2, &rt
->rt_local_src
);
1503 if (rt
->rt_daddr
!= rt
->rt_gateway
)
1504 RTA_PUT(skb
, RTA_GATEWAY
, 2, &rt
->rt_gateway
);
1505 if (rtnetlink_put_metrics(skb
, rt
->u
.dst
.metrics
) < 0)
1506 goto rtattr_failure
;
1507 ci
.rta_lastuse
= jiffies_to_clock_t(jiffies
- rt
->u
.dst
.lastuse
);
1508 ci
.rta_used
= rt
->u
.dst
.__use
;
1509 ci
.rta_clntref
= atomic_read(&rt
->u
.dst
.__refcnt
);
1510 if (rt
->u
.dst
.expires
)
1511 ci
.rta_expires
= jiffies_to_clock_t(rt
->u
.dst
.expires
- jiffies
);
1514 ci
.rta_error
= rt
->u
.dst
.error
;
1515 ci
.rta_id
= ci
.rta_ts
= ci
.rta_tsage
= 0;
1516 RTA_PUT(skb
, RTA_CACHEINFO
, sizeof(ci
), &ci
);
1518 RTA_PUT(skb
, RTA_IIF
, sizeof(int), &rt
->fl
.iif
);
1520 nlh
->nlmsg_len
= skb
->tail
- b
;
1525 skb_trim(skb
, b
- skb
->data
);
1530 * This is called by both endnodes and routers now.
1532 int dn_cache_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
, void *arg
)
1534 struct rtattr
**rta
= arg
;
1535 struct rtmsg
*rtm
= NLMSG_DATA(nlh
);
1536 struct dn_route
*rt
= NULL
;
1537 struct dn_skb_cb
*cb
;
1539 struct sk_buff
*skb
;
1542 memset(&fl
, 0, sizeof(fl
));
1543 fl
.proto
= DNPROTO_NSP
;
1545 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
1548 skb
->mac
.raw
= skb
->data
;
1549 cb
= DN_SKB_CB(skb
);
1552 memcpy(&fl
.fld_src
, RTA_DATA(rta
[RTA_SRC
-1]), 2);
1554 memcpy(&fl
.fld_dst
, RTA_DATA(rta
[RTA_DST
-1]), 2);
1556 memcpy(&fl
.iif
, RTA_DATA(rta
[RTA_IIF
-1]), sizeof(int));
1559 struct net_device
*dev
;
1560 if ((dev
= dev_get_by_index(fl
.iif
)) == NULL
) {
1569 skb
->protocol
= __constant_htons(ETH_P_DNA_RT
);
1571 cb
->src
= fl
.fld_src
;
1572 cb
->dst
= fl
.fld_dst
;
1574 err
= dn_route_input(skb
);
1576 memset(cb
, 0, sizeof(struct dn_skb_cb
));
1577 rt
= (struct dn_route
*)skb
->dst
;
1578 if (!err
&& -rt
->u
.dst
.error
)
1579 err
= rt
->u
.dst
.error
;
1582 if (rta
[RTA_OIF
- 1])
1583 memcpy(&oif
, RTA_DATA(rta
[RTA_OIF
- 1]), sizeof(int));
1585 err
= dn_route_output_key((struct dst_entry
**)&rt
, &fl
, 0);
1593 skb
->dst
= &rt
->u
.dst
;
1594 if (rtm
->rtm_flags
& RTM_F_NOTIFY
)
1595 rt
->rt_flags
|= RTCF_NOTIFY
;
1597 NETLINK_CB(skb
).dst_pid
= NETLINK_CB(in_skb
).pid
;
1599 err
= dn_rt_fill_info(skb
, NETLINK_CB(in_skb
).pid
, nlh
->nlmsg_seq
, RTM_NEWROUTE
, 0, 0);
1608 err
= netlink_unicast(rtnl
, skb
, NETLINK_CB(in_skb
).pid
, MSG_DONTWAIT
);
1618 * For routers, this is called from dn_fib_dump, but for endnodes its
1619 * called directly from the rtnetlink dispatch table.
1621 int dn_cache_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1623 struct dn_route
*rt
;
1627 if (NLMSG_PAYLOAD(cb
->nlh
, 0) < sizeof(struct rtmsg
))
1629 if (!(((struct rtmsg
*)NLMSG_DATA(cb
->nlh
))->rtm_flags
&RTM_F_CLONED
))
1633 s_idx
= idx
= cb
->args
[1];
1634 for(h
= 0; h
<= dn_rt_hash_mask
; h
++) {
1640 for(rt
= rcu_dereference(dn_rt_hash_table
[h
].chain
), idx
= 0;
1642 rt
= rcu_dereference(rt
->u
.rt_next
), idx
++) {
1645 skb
->dst
= dst_clone(&rt
->u
.dst
);
1646 if (dn_rt_fill_info(skb
, NETLINK_CB(cb
->skb
).pid
,
1647 cb
->nlh
->nlmsg_seq
, RTM_NEWROUTE
,
1648 1, NLM_F_MULTI
) <= 0) {
1649 dst_release(xchg(&skb
->dst
, NULL
));
1650 rcu_read_unlock_bh();
1653 dst_release(xchg(&skb
->dst
, NULL
));
1655 rcu_read_unlock_bh();
1664 #ifdef CONFIG_PROC_FS
1665 struct dn_rt_cache_iter_state
{
1669 static struct dn_route
*dn_rt_cache_get_first(struct seq_file
*seq
)
1671 struct dn_route
*rt
= NULL
;
1672 struct dn_rt_cache_iter_state
*s
= seq
->private;
1674 for(s
->bucket
= dn_rt_hash_mask
; s
->bucket
>= 0; --s
->bucket
) {
1676 rt
= dn_rt_hash_table
[s
->bucket
].chain
;
1679 rcu_read_unlock_bh();
1684 static struct dn_route
*dn_rt_cache_get_next(struct seq_file
*seq
, struct dn_route
*rt
)
1686 struct dn_rt_cache_iter_state
*s
= rcu_dereference(seq
->private);
1690 rcu_read_unlock_bh();
1691 if (--s
->bucket
< 0)
1694 rt
= dn_rt_hash_table
[s
->bucket
].chain
;
1699 static void *dn_rt_cache_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1701 struct dn_route
*rt
= dn_rt_cache_get_first(seq
);
1704 while(*pos
&& (rt
= dn_rt_cache_get_next(seq
, rt
)))
1707 return *pos
? NULL
: rt
;
1710 static void *dn_rt_cache_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1712 struct dn_route
*rt
= dn_rt_cache_get_next(seq
, v
);
1717 static void dn_rt_cache_seq_stop(struct seq_file
*seq
, void *v
)
1720 rcu_read_unlock_bh();
1723 static int dn_rt_cache_seq_show(struct seq_file
*seq
, void *v
)
1725 struct dn_route
*rt
= v
;
1726 char buf1
[DN_ASCBUF_LEN
], buf2
[DN_ASCBUF_LEN
];
1728 seq_printf(seq
, "%-8s %-7s %-7s %04d %04d %04d\n",
1729 rt
->u
.dst
.dev
? rt
->u
.dst
.dev
->name
: "*",
1730 dn_addr2asc(dn_ntohs(rt
->rt_daddr
), buf1
),
1731 dn_addr2asc(dn_ntohs(rt
->rt_saddr
), buf2
),
1732 atomic_read(&rt
->u
.dst
.__refcnt
),
1734 (int) dst_metric(&rt
->u
.dst
, RTAX_RTT
));
1738 static struct seq_operations dn_rt_cache_seq_ops
= {
1739 .start
= dn_rt_cache_seq_start
,
1740 .next
= dn_rt_cache_seq_next
,
1741 .stop
= dn_rt_cache_seq_stop
,
1742 .show
= dn_rt_cache_seq_show
,
1745 static int dn_rt_cache_seq_open(struct inode
*inode
, struct file
*file
)
1747 struct seq_file
*seq
;
1749 struct dn_rt_cache_iter_state
*s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
1753 rc
= seq_open(file
, &dn_rt_cache_seq_ops
);
1756 seq
= file
->private_data
;
1758 memset(s
, 0, sizeof(*s
));
1766 static struct file_operations dn_rt_cache_seq_fops
= {
1767 .owner
= THIS_MODULE
,
1768 .open
= dn_rt_cache_seq_open
,
1770 .llseek
= seq_lseek
,
1771 .release
= seq_release_private
,
1774 #endif /* CONFIG_PROC_FS */
1776 void __init
dn_route_init(void)
1780 dn_dst_ops
.kmem_cachep
= kmem_cache_create("dn_dst_cache",
1781 sizeof(struct dn_route
),
1782 0, SLAB_HWCACHE_ALIGN
,
1785 if (!dn_dst_ops
.kmem_cachep
)
1786 panic("DECnet: Failed to allocate dn_dst_cache\n");
1788 init_timer(&dn_route_timer
);
1789 dn_route_timer
.function
= dn_dst_check_expire
;
1790 dn_route_timer
.expires
= jiffies
+ decnet_dst_gc_interval
* HZ
;
1791 add_timer(&dn_route_timer
);
1793 goal
= num_physpages
>> (26 - PAGE_SHIFT
);
1795 for(order
= 0; (1UL << order
) < goal
; order
++)
1799 * Only want 1024 entries max, since the table is very, very unlikely
1800 * to be larger than that.
1802 while(order
&& ((((1UL << order
) * PAGE_SIZE
) /
1803 sizeof(struct dn_rt_hash_bucket
)) >= 2048))
1807 dn_rt_hash_mask
= (1UL << order
) * PAGE_SIZE
/
1808 sizeof(struct dn_rt_hash_bucket
);
1809 while(dn_rt_hash_mask
& (dn_rt_hash_mask
- 1))
1811 dn_rt_hash_table
= (struct dn_rt_hash_bucket
*)
1812 __get_free_pages(GFP_ATOMIC
, order
);
1813 } while (dn_rt_hash_table
== NULL
&& --order
> 0);
1815 if (!dn_rt_hash_table
)
1816 panic("Failed to allocate DECnet route cache hash table\n");
1819 "DECnet: Routing cache hash table of %u buckets, %ldKbytes\n",
1821 (long)(dn_rt_hash_mask
*sizeof(struct dn_rt_hash_bucket
))/1024);
1824 for(i
= 0; i
<= dn_rt_hash_mask
; i
++) {
1825 spin_lock_init(&dn_rt_hash_table
[i
].lock
);
1826 dn_rt_hash_table
[i
].chain
= NULL
;
1829 dn_dst_ops
.gc_thresh
= (dn_rt_hash_mask
+ 1);
1831 proc_net_fops_create("decnet_cache", S_IRUGO
, &dn_rt_cache_seq_fops
);
1834 void __exit
dn_route_cleanup(void)
1836 del_timer(&dn_route_timer
);
1839 proc_net_remove("decnet_cache");