1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * DECnet An implementation of the DECnet protocol suite for the LINUX
4 * operating system. DECnet is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * DECnet Routing Functions (Endnode and Router)
9 * Authors: Steve Whitehouse <SteveW@ACM.org>
10 * Eduardo Marcelo Serrat <emserrat@geocities.com>
13 * Steve Whitehouse : Fixes to allow "intra-ethernet" and
14 * "return-to-sender" bits on outgoing
16 * Steve Whitehouse : Timeouts for cached routes.
17 * Steve Whitehouse : Use dst cache for input routes too.
18 * Steve Whitehouse : Fixed error values in dn_send_skb.
19 * Steve Whitehouse : Rework routing functions to better fit
20 * DECnet routing design
21 * Alexey Kuznetsov : New SMP locking
22 * Steve Whitehouse : More SMP locking changes & dn_cache_dump()
23 * Steve Whitehouse : Prerouting NF hook, now really is prerouting.
24 * Fixed possible skb leak in rtnetlink funcs.
25 * Steve Whitehouse : Dave Miller's dynamic hash table sizing and
26 * Alexey Kuznetsov's finer grained locking
28 * Steve Whitehouse : Routing is now starting to look like a
29 * sensible set of code now, mainly due to
30 * my copying the IPv4 routing code. The
31 * hooks here are modified and will continue
32 * to evolve for a while.
33 * Steve Whitehouse : Real SMP at last :-) Also new netfilter
34 * stuff. Look out raw sockets your days
36 * Steve Whitehouse : Added return-to-sender functions. Added
37 * backlog congestion level return codes.
38 * Steve Whitehouse : Fixed bug where routes were set up with
39 * no ref count on net devices.
40 * Steve Whitehouse : RCU for the route cache
41 * Steve Whitehouse : Preparations for the flow cache
42 * Steve Whitehouse : Prepare for nonlinear skbs
45 /******************************************************************************
46 (c) 1995-1998 E.M. Serrat emserrat@geocities.com
48 *******************************************************************************/
50 #include <linux/errno.h>
51 #include <linux/types.h>
52 #include <linux/socket.h>
54 #include <linux/kernel.h>
55 #include <linux/sockios.h>
56 #include <linux/net.h>
57 #include <linux/netdevice.h>
58 #include <linux/inet.h>
59 #include <linux/route.h>
60 #include <linux/in_route.h>
61 #include <linux/slab.h>
64 #include <linux/proc_fs.h>
65 #include <linux/seq_file.h>
66 #include <linux/init.h>
67 #include <linux/rtnetlink.h>
68 #include <linux/string.h>
69 #include <linux/netfilter_decnet.h>
70 #include <linux/rcupdate.h>
71 #include <linux/times.h>
72 #include <linux/export.h>
73 #include <asm/errno.h>
74 #include <net/net_namespace.h>
75 #include <net/netlink.h>
76 #include <net/neighbour.h>
79 #include <net/fib_rules.h>
81 #include <net/dn_dev.h>
82 #include <net/dn_nsp.h>
83 #include <net/dn_route.h>
84 #include <net/dn_neigh.h>
85 #include <net/dn_fib.h>
87 struct dn_rt_hash_bucket
89 struct dn_route __rcu
*chain
;
93 extern struct neigh_table dn_neigh_table
;
96 static unsigned char dn_hiord_addr
[6] = {0xAA,0x00,0x04,0x00,0x00,0x00};
98 static const int dn_rt_min_delay
= 2 * HZ
;
99 static const int dn_rt_max_delay
= 10 * HZ
;
100 static const int dn_rt_mtu_expires
= 10 * 60 * HZ
;
102 static unsigned long dn_rt_deadline
;
104 static int dn_dst_gc(struct dst_ops
*ops
);
105 static struct dst_entry
*dn_dst_check(struct dst_entry
*, __u32
);
106 static unsigned int dn_dst_default_advmss(const struct dst_entry
*dst
);
107 static unsigned int dn_dst_mtu(const struct dst_entry
*dst
);
108 static void dn_dst_destroy(struct dst_entry
*);
109 static void dn_dst_ifdown(struct dst_entry
*, struct net_device
*dev
, int how
);
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
, struct sock
*sk
,
113 struct sk_buff
*skb
, u32 mtu
,
115 static void dn_dst_redirect(struct dst_entry
*dst
, struct sock
*sk
,
116 struct sk_buff
*skb
);
117 static struct neighbour
*dn_dst_neigh_lookup(const struct dst_entry
*dst
,
120 static int dn_route_input(struct sk_buff
*);
121 static void dn_run_flush(struct timer_list
*unused
);
123 static struct dn_rt_hash_bucket
*dn_rt_hash_table
;
124 static unsigned int dn_rt_hash_mask
;
126 static struct timer_list dn_route_timer
;
127 static DEFINE_TIMER(dn_rt_flush_timer
, dn_run_flush
);
128 int decnet_dst_gc_interval
= 2;
130 static struct dst_ops dn_dst_ops
= {
134 .check
= dn_dst_check
,
135 .default_advmss
= dn_dst_default_advmss
,
137 .cow_metrics
= dst_cow_metrics_generic
,
138 .destroy
= dn_dst_destroy
,
139 .ifdown
= dn_dst_ifdown
,
140 .negative_advice
= dn_dst_negative_advice
,
141 .link_failure
= dn_dst_link_failure
,
142 .update_pmtu
= dn_dst_update_pmtu
,
143 .redirect
= dn_dst_redirect
,
144 .neigh_lookup
= dn_dst_neigh_lookup
,
147 static void dn_dst_destroy(struct dst_entry
*dst
)
149 struct dn_route
*rt
= (struct dn_route
*) dst
;
152 neigh_release(rt
->n
);
153 dst_destroy_metrics_generic(dst
);
156 static void dn_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
, int how
)
159 struct dn_route
*rt
= (struct dn_route
*) dst
;
160 struct neighbour
*n
= rt
->n
;
162 if (n
&& n
->dev
== dev
) {
163 n
->dev
= dev_net(dev
)->loopback_dev
;
170 static __inline__
unsigned int dn_hash(__le16 src
, __le16 dst
)
172 __u16 tmp
= (__u16 __force
)(src
^ dst
);
176 return dn_rt_hash_mask
& (unsigned int)tmp
;
179 static void dn_dst_check_expire(struct timer_list
*unused
)
183 struct dn_route __rcu
**rtp
;
184 unsigned long now
= jiffies
;
185 unsigned long expire
= 120 * HZ
;
187 for (i
= 0; i
<= dn_rt_hash_mask
; i
++) {
188 rtp
= &dn_rt_hash_table
[i
].chain
;
190 spin_lock(&dn_rt_hash_table
[i
].lock
);
191 while ((rt
= rcu_dereference_protected(*rtp
,
192 lockdep_is_held(&dn_rt_hash_table
[i
].lock
))) != NULL
) {
193 if (atomic_read(&rt
->dst
.__refcnt
) > 1 ||
194 (now
- rt
->dst
.lastuse
) < expire
) {
200 dst_dev_put(&rt
->dst
);
201 dst_release(&rt
->dst
);
203 spin_unlock(&dn_rt_hash_table
[i
].lock
);
205 if ((jiffies
- now
) > 0)
209 mod_timer(&dn_route_timer
, now
+ decnet_dst_gc_interval
* HZ
);
212 static int dn_dst_gc(struct dst_ops
*ops
)
215 struct dn_route __rcu
**rtp
;
217 unsigned long now
= jiffies
;
218 unsigned long expire
= 10 * HZ
;
220 for (i
= 0; i
<= dn_rt_hash_mask
; i
++) {
222 spin_lock_bh(&dn_rt_hash_table
[i
].lock
);
223 rtp
= &dn_rt_hash_table
[i
].chain
;
225 while ((rt
= rcu_dereference_protected(*rtp
,
226 lockdep_is_held(&dn_rt_hash_table
[i
].lock
))) != NULL
) {
227 if (atomic_read(&rt
->dst
.__refcnt
) > 1 ||
228 (now
- rt
->dst
.lastuse
) < expire
) {
234 dst_dev_put(&rt
->dst
);
235 dst_release(&rt
->dst
);
238 spin_unlock_bh(&dn_rt_hash_table
[i
].lock
);
245 * The decnet standards don't impose a particular minimum mtu, what they
246 * do insist on is that the routing layer accepts a datagram of at least
247 * 230 bytes long. Here we have to subtract the routing header length from
248 * 230 to get the minimum acceptable mtu. If there is no neighbour, then we
249 * assume the worst and use a long header size.
251 * We update both the mtu and the advertised mss (i.e. the segment size we
252 * advertise to the other end).
254 static void dn_dst_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
255 struct sk_buff
*skb
, u32 mtu
,
258 struct dn_route
*rt
= (struct dn_route
*) dst
;
259 struct neighbour
*n
= rt
->n
;
263 dn
= n
? rcu_dereference_raw(n
->dev
->dn_ptr
) : NULL
;
265 if (dn
&& dn
->use_long
== 0)
270 if (dst_metric(dst
, RTAX_MTU
) > mtu
&& mtu
>= min_mtu
) {
271 if (!(dst_metric_locked(dst
, RTAX_MTU
))) {
272 dst_metric_set(dst
, RTAX_MTU
, mtu
);
273 dst_set_expires(dst
, dn_rt_mtu_expires
);
275 if (!(dst_metric_locked(dst
, RTAX_ADVMSS
))) {
276 u32 mss
= mtu
- DN_MAX_NSP_DATA_HEADER
;
277 u32 existing_mss
= dst_metric_raw(dst
, RTAX_ADVMSS
);
278 if (!existing_mss
|| existing_mss
> mss
)
279 dst_metric_set(dst
, RTAX_ADVMSS
, mss
);
284 static void dn_dst_redirect(struct dst_entry
*dst
, struct sock
*sk
,
290 * When a route has been marked obsolete. (e.g. routing cache flush)
292 static struct dst_entry
*dn_dst_check(struct dst_entry
*dst
, __u32 cookie
)
297 static struct dst_entry
*dn_dst_negative_advice(struct dst_entry
*dst
)
303 static void dn_dst_link_failure(struct sk_buff
*skb
)
307 static inline int compare_keys(struct flowidn
*fl1
, struct flowidn
*fl2
)
309 return ((fl1
->daddr
^ fl2
->daddr
) |
310 (fl1
->saddr
^ fl2
->saddr
) |
311 (fl1
->flowidn_mark
^ fl2
->flowidn_mark
) |
312 (fl1
->flowidn_scope
^ fl2
->flowidn_scope
) |
313 (fl1
->flowidn_oif
^ fl2
->flowidn_oif
) |
314 (fl1
->flowidn_iif
^ fl2
->flowidn_iif
)) == 0;
317 static int dn_insert_route(struct dn_route
*rt
, unsigned int hash
, struct dn_route
**rp
)
319 struct dn_route
*rth
;
320 struct dn_route __rcu
**rthp
;
321 unsigned long now
= jiffies
;
323 rthp
= &dn_rt_hash_table
[hash
].chain
;
325 spin_lock_bh(&dn_rt_hash_table
[hash
].lock
);
326 while ((rth
= rcu_dereference_protected(*rthp
,
327 lockdep_is_held(&dn_rt_hash_table
[hash
].lock
))) != NULL
) {
328 if (compare_keys(&rth
->fld
, &rt
->fld
)) {
330 *rthp
= rth
->dn_next
;
331 rcu_assign_pointer(rth
->dn_next
,
332 dn_rt_hash_table
[hash
].chain
);
333 rcu_assign_pointer(dn_rt_hash_table
[hash
].chain
, rth
);
335 dst_hold_and_use(&rth
->dst
, now
);
336 spin_unlock_bh(&dn_rt_hash_table
[hash
].lock
);
338 dst_release_immediate(&rt
->dst
);
342 rthp
= &rth
->dn_next
;
345 rcu_assign_pointer(rt
->dn_next
, dn_rt_hash_table
[hash
].chain
);
346 rcu_assign_pointer(dn_rt_hash_table
[hash
].chain
, rt
);
348 dst_hold_and_use(&rt
->dst
, now
);
349 spin_unlock_bh(&dn_rt_hash_table
[hash
].lock
);
354 static void dn_run_flush(struct timer_list
*unused
)
357 struct dn_route
*rt
, *next
;
359 for (i
= 0; i
< dn_rt_hash_mask
; i
++) {
360 spin_lock_bh(&dn_rt_hash_table
[i
].lock
);
362 if ((rt
= xchg((struct dn_route
**)&dn_rt_hash_table
[i
].chain
, NULL
)) == NULL
)
363 goto nothing_to_declare
;
365 for(; rt
; rt
= next
) {
366 next
= rcu_dereference_raw(rt
->dn_next
);
367 RCU_INIT_POINTER(rt
->dn_next
, NULL
);
368 dst_dev_put(&rt
->dst
);
369 dst_release(&rt
->dst
);
373 spin_unlock_bh(&dn_rt_hash_table
[i
].lock
);
377 static DEFINE_SPINLOCK(dn_rt_flush_lock
);
379 void dn_rt_cache_flush(int delay
)
381 unsigned long now
= jiffies
;
382 int user_mode
= !in_interrupt();
385 delay
= dn_rt_min_delay
;
387 spin_lock_bh(&dn_rt_flush_lock
);
389 if (del_timer(&dn_rt_flush_timer
) && delay
> 0 && dn_rt_deadline
) {
390 long tmo
= (long)(dn_rt_deadline
- now
);
392 if (user_mode
&& tmo
< dn_rt_max_delay
- dn_rt_min_delay
)
400 spin_unlock_bh(&dn_rt_flush_lock
);
405 if (dn_rt_deadline
== 0)
406 dn_rt_deadline
= now
+ dn_rt_max_delay
;
408 dn_rt_flush_timer
.expires
= now
+ delay
;
409 add_timer(&dn_rt_flush_timer
);
410 spin_unlock_bh(&dn_rt_flush_lock
);
414 * dn_return_short - Return a short packet to its sender
415 * @skb: The packet to return
418 static int dn_return_short(struct sk_buff
*skb
)
420 struct dn_skb_cb
*cb
;
425 /* Add back headers */
426 skb_push(skb
, skb
->data
- skb_network_header(skb
));
428 if ((skb
= skb_unshare(skb
, GFP_ATOMIC
)) == NULL
)
432 /* Skip packet length and point to flags */
434 *ptr
++ = (cb
->rt_flags
& ~DN_RT_F_RQR
) | DN_RT_F_RTS
;
440 *ptr
= 0; /* Zero hop count */
444 skb
->pkt_type
= PACKET_OUTGOING
;
445 dn_rt_finish_output(skb
, NULL
, NULL
);
446 return NET_RX_SUCCESS
;
450 * dn_return_long - Return a long packet to its sender
451 * @skb: The long format packet to return
454 static int dn_return_long(struct sk_buff
*skb
)
456 struct dn_skb_cb
*cb
;
458 unsigned char *src_addr
, *dst_addr
;
459 unsigned char tmp
[ETH_ALEN
];
461 /* Add back all headers */
462 skb_push(skb
, skb
->data
- skb_network_header(skb
));
464 if ((skb
= skb_unshare(skb
, GFP_ATOMIC
)) == NULL
)
468 /* Ignore packet length and point to flags */
472 if (*ptr
& DN_RT_F_PF
) {
473 char padlen
= (*ptr
& ~DN_RT_F_PF
);
477 *ptr
++ = (cb
->rt_flags
& ~DN_RT_F_RQR
) | DN_RT_F_RTS
;
483 *ptr
= 0; /* Zero hop count */
485 /* Swap source and destination */
486 memcpy(tmp
, src_addr
, ETH_ALEN
);
487 memcpy(src_addr
, dst_addr
, ETH_ALEN
);
488 memcpy(dst_addr
, tmp
, ETH_ALEN
);
490 skb
->pkt_type
= PACKET_OUTGOING
;
491 dn_rt_finish_output(skb
, dst_addr
, src_addr
);
492 return NET_RX_SUCCESS
;
496 * dn_route_rx_packet - Try and find a route for an incoming packet
497 * @skb: The packet to find a route for
499 * Returns: result of input function if route is found, error code otherwise
501 static int dn_route_rx_packet(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
503 struct dn_skb_cb
*cb
;
506 if ((err
= dn_route_input(skb
)) == 0)
507 return dst_input(skb
);
510 if (decnet_debug_level
& 4) {
511 char *devname
= skb
->dev
? skb
->dev
->name
: "???";
514 "DECnet: dn_route_rx_packet: rt_flags=0x%02x dev=%s len=%d src=0x%04hx dst=0x%04hx err=%d type=%d\n",
515 (int)cb
->rt_flags
, devname
, skb
->len
,
516 le16_to_cpu(cb
->src
), le16_to_cpu(cb
->dst
),
520 if ((skb
->pkt_type
== PACKET_HOST
) && (cb
->rt_flags
& DN_RT_F_RQR
)) {
521 switch (cb
->rt_flags
& DN_RT_PKT_MSK
) {
522 case DN_RT_PKT_SHORT
:
523 return dn_return_short(skb
);
525 return dn_return_long(skb
);
533 static int dn_route_rx_long(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
, 21)) /* 20 for long header, 1 for shortest nsp */
542 skb_reset_transport_header(skb
);
544 /* Destination info */
546 cb
->dst
= dn_eth2dn(ptr
);
547 if (memcmp(ptr
, dn_hiord_addr
, 4) != 0)
554 cb
->src
= dn_eth2dn(ptr
);
555 if (memcmp(ptr
, dn_hiord_addr
, 4) != 0)
560 cb
->hops
= *ptr
++; /* Visit Count */
562 return NF_HOOK(NFPROTO_DECNET
, NF_DN_PRE_ROUTING
,
563 &init_net
, NULL
, skb
, skb
->dev
, NULL
,
573 static int dn_route_rx_short(struct sk_buff
*skb
)
575 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
576 unsigned char *ptr
= skb
->data
;
578 if (!pskb_may_pull(skb
, 6)) /* 5 for short header + 1 for shortest nsp */
582 skb_reset_transport_header(skb
);
584 cb
->dst
= *(__le16
*)ptr
;
586 cb
->src
= *(__le16
*)ptr
;
588 cb
->hops
= *ptr
& 0x3f;
590 return NF_HOOK(NFPROTO_DECNET
, NF_DN_PRE_ROUTING
,
591 &init_net
, NULL
, skb
, skb
->dev
, NULL
,
599 static int dn_route_discard(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
602 * I know we drop the packet here, but thats considered success in
606 return NET_RX_SUCCESS
;
609 static int dn_route_ptp_hello(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
612 dn_neigh_pointopoint_hello(skb
);
613 return NET_RX_SUCCESS
;
616 int dn_route_rcv(struct sk_buff
*skb
, struct net_device
*dev
, struct packet_type
*pt
, struct net_device
*orig_dev
)
618 struct dn_skb_cb
*cb
;
619 unsigned char flags
= 0;
620 __u16 len
= le16_to_cpu(*(__le16
*)skb
->data
);
621 struct dn_dev
*dn
= rcu_dereference(dev
->dn_ptr
);
622 unsigned char padlen
= 0;
624 if (!net_eq(dev_net(dev
), &init_net
))
630 if ((skb
= skb_share_check(skb
, GFP_ATOMIC
)) == NULL
)
633 if (!pskb_may_pull(skb
, 3))
647 cb
->iif
= dev
->ifindex
;
650 * If we have padding, remove it.
652 if (flags
& DN_RT_F_PF
) {
653 padlen
= flags
& ~DN_RT_F_PF
;
654 if (!pskb_may_pull(skb
, padlen
+ 1))
656 skb_pull(skb
, padlen
);
660 skb_reset_network_header(skb
);
663 * Weed out future version DECnet
665 if (flags
& DN_RT_F_VER
)
668 cb
->rt_flags
= flags
;
670 if (decnet_debug_level
& 1)
672 "dn_route_rcv: got 0x%02x from %s [%d %d %d]\n",
673 (int)flags
, (dev
) ? dev
->name
: "???", len
, skb
->len
,
676 if (flags
& DN_RT_PKT_CNTL
) {
677 if (unlikely(skb_linearize(skb
)))
680 switch (flags
& DN_RT_CNTL_MSK
) {
682 dn_dev_init_pkt(skb
);
685 dn_dev_veri_pkt(skb
);
689 if (dn
->parms
.state
!= DN_DEV_S_RU
)
692 switch (flags
& DN_RT_CNTL_MSK
) {
694 return NF_HOOK(NFPROTO_DECNET
, NF_DN_HELLO
,
695 &init_net
, NULL
, skb
, skb
->dev
, NULL
,
700 return NF_HOOK(NFPROTO_DECNET
, NF_DN_ROUTE
,
701 &init_net
, NULL
, skb
, skb
->dev
, NULL
,
704 return NF_HOOK(NFPROTO_DECNET
, NF_DN_HELLO
,
705 &init_net
, NULL
, skb
, skb
->dev
, NULL
,
706 dn_neigh_router_hello
);
709 return NF_HOOK(NFPROTO_DECNET
, NF_DN_HELLO
,
710 &init_net
, NULL
, skb
, skb
->dev
, NULL
,
711 dn_neigh_endnode_hello
);
714 if (dn
->parms
.state
!= DN_DEV_S_RU
)
717 skb_pull(skb
, 1); /* Pull flags */
719 switch (flags
& DN_RT_PKT_MSK
) {
721 return dn_route_rx_long(skb
);
722 case DN_RT_PKT_SHORT
:
723 return dn_route_rx_short(skb
);
733 static int dn_output(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
735 struct dst_entry
*dst
= skb_dst(skb
);
736 struct dn_route
*rt
= (struct dn_route
*)dst
;
737 struct net_device
*dev
= dst
->dev
;
738 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
747 cb
->src
= rt
->rt_saddr
;
748 cb
->dst
= rt
->rt_daddr
;
751 * Always set the Intra-Ethernet bit on all outgoing packets
752 * originated on this node. Only valid flag from upper layers
753 * is return-to-sender-requested. Set hop count to 0 too.
755 cb
->rt_flags
&= ~DN_RT_F_RQR
;
756 cb
->rt_flags
|= DN_RT_F_IE
;
759 return NF_HOOK(NFPROTO_DECNET
, NF_DN_LOCAL_OUT
,
760 &init_net
, sk
, skb
, NULL
, dev
,
764 net_dbg_ratelimited("dn_output: This should not happen\n");
771 static int dn_forward(struct sk_buff
*skb
)
773 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
774 struct dst_entry
*dst
= skb_dst(skb
);
775 struct dn_dev
*dn_db
= rcu_dereference(dst
->dev
->dn_ptr
);
778 struct net_device
*dev
= skb
->dev
;
780 if (skb
->pkt_type
!= PACKET_HOST
)
783 /* Ensure that we have enough space for headers */
784 rt
= (struct dn_route
*)skb_dst(skb
);
785 header_len
= dn_db
->use_long
? 21 : 6;
786 if (skb_cow(skb
, LL_RESERVED_SPACE(rt
->dst
.dev
)+header_len
))
790 * Hop count exceeded.
795 skb
->dev
= rt
->dst
.dev
;
798 * If packet goes out same interface it came in on, then set
799 * the Intra-Ethernet bit. This has no effect for short
800 * packets, so we don't need to test for them here.
802 cb
->rt_flags
&= ~DN_RT_F_IE
;
803 if (rt
->rt_flags
& RTCF_DOREDIRECT
)
804 cb
->rt_flags
|= DN_RT_F_IE
;
806 return NF_HOOK(NFPROTO_DECNET
, NF_DN_FORWARD
,
807 &init_net
, NULL
, skb
, dev
, skb
->dev
,
816 * Used to catch bugs. This should never normally get
819 static int dn_rt_bug_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
821 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
823 net_dbg_ratelimited("dn_rt_bug: skb from:%04x to:%04x\n",
824 le16_to_cpu(cb
->src
), le16_to_cpu(cb
->dst
));
831 static int dn_rt_bug(struct sk_buff
*skb
)
833 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
835 net_dbg_ratelimited("dn_rt_bug: skb from:%04x to:%04x\n",
836 le16_to_cpu(cb
->src
), le16_to_cpu(cb
->dst
));
843 static unsigned int dn_dst_default_advmss(const struct dst_entry
*dst
)
845 return dn_mss_from_pmtu(dst
->dev
, dst_mtu(dst
));
848 static unsigned int dn_dst_mtu(const struct dst_entry
*dst
)
850 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
852 return mtu
? : dst
->dev
->mtu
;
855 static struct neighbour
*dn_dst_neigh_lookup(const struct dst_entry
*dst
,
859 return __neigh_lookup_errno(&dn_neigh_table
, daddr
, dst
->dev
);
862 static int dn_rt_set_next_hop(struct dn_route
*rt
, struct dn_fib_res
*res
)
864 struct dn_fib_info
*fi
= res
->fi
;
865 struct net_device
*dev
= rt
->dst
.dev
;
866 unsigned int mss_metric
;
870 if (DN_FIB_RES_GW(*res
) &&
871 DN_FIB_RES_NH(*res
).nh_scope
== RT_SCOPE_LINK
)
872 rt
->rt_gateway
= DN_FIB_RES_GW(*res
);
873 dst_init_metrics(&rt
->dst
, fi
->fib_metrics
, true);
875 rt
->rt_type
= res
->type
;
877 if (dev
!= NULL
&& rt
->n
== NULL
) {
878 n
= __neigh_lookup_errno(&dn_neigh_table
, &rt
->rt_gateway
, dev
);
884 if (dst_metric(&rt
->dst
, RTAX_MTU
) > rt
->dst
.dev
->mtu
)
885 dst_metric_set(&rt
->dst
, RTAX_MTU
, rt
->dst
.dev
->mtu
);
886 mss_metric
= dst_metric_raw(&rt
->dst
, RTAX_ADVMSS
);
888 unsigned int mss
= dn_mss_from_pmtu(dev
, dst_mtu(&rt
->dst
));
889 if (mss_metric
> mss
)
890 dst_metric_set(&rt
->dst
, RTAX_ADVMSS
, mss
);
895 static inline int dn_match_addr(__le16 addr1
, __le16 addr2
)
897 __u16 tmp
= le16_to_cpu(addr1
) ^ le16_to_cpu(addr2
);
906 static __le16
dnet_select_source(const struct net_device
*dev
, __le16 daddr
, int scope
)
909 struct dn_dev
*dn_db
;
910 struct dn_ifaddr
*ifa
;
915 dn_db
= rcu_dereference(dev
->dn_ptr
);
916 for (ifa
= rcu_dereference(dn_db
->ifa_list
);
918 ifa
= rcu_dereference(ifa
->ifa_next
)) {
919 if (ifa
->ifa_scope
> scope
)
922 saddr
= ifa
->ifa_local
;
925 ret
= dn_match_addr(daddr
, ifa
->ifa_local
);
926 if (ret
> best_match
)
927 saddr
= ifa
->ifa_local
;
929 saddr
= ifa
->ifa_local
;
936 static inline __le16
__dn_fib_res_prefsrc(struct dn_fib_res
*res
)
938 return dnet_select_source(DN_FIB_RES_DEV(*res
), DN_FIB_RES_GW(*res
), res
->scope
);
941 static inline __le16
dn_fib_rules_map_destination(__le16 daddr
, struct dn_fib_res
*res
)
943 __le16 mask
= dnet_make_mask(res
->prefixlen
);
944 return (daddr
&~mask
)|res
->fi
->fib_nh
->nh_gw
;
947 static int dn_route_output_slow(struct dst_entry
**pprt
, const struct flowidn
*oldflp
, int try_hard
)
949 struct flowidn fld
= {
950 .daddr
= oldflp
->daddr
,
951 .saddr
= oldflp
->saddr
,
952 .flowidn_scope
= RT_SCOPE_UNIVERSE
,
953 .flowidn_mark
= oldflp
->flowidn_mark
,
954 .flowidn_iif
= LOOPBACK_IFINDEX
,
955 .flowidn_oif
= oldflp
->flowidn_oif
,
957 struct dn_route
*rt
= NULL
;
958 struct net_device
*dev_out
= NULL
, *dev
;
959 struct neighbour
*neigh
= NULL
;
961 unsigned int flags
= 0;
962 struct dn_fib_res res
= { .fi
= NULL
, .type
= RTN_UNICAST
};
967 if (decnet_debug_level
& 16)
969 "dn_route_output_slow: dst=%04x src=%04x mark=%d"
970 " iif=%d oif=%d\n", le16_to_cpu(oldflp
->daddr
),
971 le16_to_cpu(oldflp
->saddr
),
972 oldflp
->flowidn_mark
, LOOPBACK_IFINDEX
,
973 oldflp
->flowidn_oif
);
975 /* If we have an output interface, verify its a DECnet device */
976 if (oldflp
->flowidn_oif
) {
977 dev_out
= dev_get_by_index(&init_net
, oldflp
->flowidn_oif
);
979 if (dev_out
&& dev_out
->dn_ptr
== NULL
) {
987 /* If we have a source address, verify that its a local address */
989 err
= -EADDRNOTAVAIL
;
992 if (dn_dev_islocal(dev_out
, oldflp
->saddr
))
998 for_each_netdev_rcu(&init_net
, dev
) {
1001 if (!dn_dev_islocal(dev
, oldflp
->saddr
))
1003 if ((dev
->flags
& IFF_LOOPBACK
) &&
1005 !dn_dev_islocal(dev
, oldflp
->daddr
))
1012 if (dev_out
== NULL
)
1019 /* No destination? Assume its local */
1021 fld
.daddr
= fld
.saddr
;
1026 dev_out
= init_net
.loopback_dev
;
1027 if (!dev_out
->dn_ptr
)
1029 err
= -EADDRNOTAVAIL
;
1033 fld
.saddr
= dnet_select_source(dev_out
, 0,
1038 fld
.flowidn_oif
= LOOPBACK_IFINDEX
;
1039 res
.type
= RTN_LOCAL
;
1043 if (decnet_debug_level
& 16)
1045 "dn_route_output_slow: initial checks complete."
1046 " dst=%04x src=%04x oif=%d try_hard=%d\n",
1047 le16_to_cpu(fld
.daddr
), le16_to_cpu(fld
.saddr
),
1048 fld
.flowidn_oif
, try_hard
);
1051 * N.B. If the kernel is compiled without router support then
1052 * dn_fib_lookup() will evaluate to non-zero so this if () block
1053 * will always be executed.
1056 if (try_hard
|| (err
= dn_fib_lookup(&fld
, &res
)) != 0) {
1057 struct dn_dev
*dn_db
;
1061 * Here the fallback is basically the standard algorithm for
1062 * routing in endnodes which is described in the DECnet routing
1065 * If we are not trying hard, look in neighbour cache.
1066 * The result is tested to ensure that if a specific output
1067 * device/source address was requested, then we honour that
1071 neigh
= neigh_lookup_nodev(&dn_neigh_table
, &init_net
, &fld
.daddr
);
1073 if ((oldflp
->flowidn_oif
&&
1074 (neigh
->dev
->ifindex
!= oldflp
->flowidn_oif
)) ||
1076 (!dn_dev_islocal(neigh
->dev
,
1078 neigh_release(neigh
);
1083 if (dn_dev_islocal(neigh
->dev
, fld
.daddr
)) {
1084 dev_out
= init_net
.loopback_dev
;
1085 res
.type
= RTN_LOCAL
;
1087 dev_out
= neigh
->dev
;
1095 /* Not there? Perhaps its a local address */
1096 if (dev_out
== NULL
)
1097 dev_out
= dn_dev_get_default();
1099 if (dev_out
== NULL
)
1101 dn_db
= rcu_dereference_raw(dev_out
->dn_ptr
);
1104 /* Possible improvement - check all devices for local addr */
1105 if (dn_dev_islocal(dev_out
, fld
.daddr
)) {
1107 dev_out
= init_net
.loopback_dev
;
1109 res
.type
= RTN_LOCAL
;
1112 /* Not local either.... try sending it to the default router */
1113 neigh
= neigh_clone(dn_db
->router
);
1114 BUG_ON(neigh
&& neigh
->dev
!= dev_out
);
1116 /* Ok then, we assume its directly connected and move on */
1119 gateway
= ((struct dn_neigh
*)neigh
)->addr
;
1121 gateway
= fld
.daddr
;
1122 if (fld
.saddr
== 0) {
1123 fld
.saddr
= dnet_select_source(dev_out
, gateway
,
1124 res
.type
== RTN_LOCAL
?
1127 if (fld
.saddr
== 0 && res
.type
!= RTN_LOCAL
)
1130 fld
.flowidn_oif
= dev_out
->ifindex
;
1135 if (res
.type
== RTN_NAT
)
1138 if (res
.type
== RTN_LOCAL
) {
1140 fld
.saddr
= fld
.daddr
;
1143 dev_out
= init_net
.loopback_dev
;
1145 if (!dev_out
->dn_ptr
)
1147 fld
.flowidn_oif
= dev_out
->ifindex
;
1149 dn_fib_info_put(res
.fi
);
1154 if (res
.fi
->fib_nhs
> 1 && fld
.flowidn_oif
== 0)
1155 dn_fib_select_multipath(&fld
, &res
);
1158 * We could add some logic to deal with default routes here and
1159 * get rid of some of the special casing above.
1163 fld
.saddr
= DN_FIB_RES_PREFSRC(res
);
1167 dev_out
= DN_FIB_RES_DEV(res
);
1169 fld
.flowidn_oif
= dev_out
->ifindex
;
1170 gateway
= DN_FIB_RES_GW(res
);
1173 if (dev_out
->flags
& IFF_LOOPBACK
)
1174 flags
|= RTCF_LOCAL
;
1176 rt
= dst_alloc(&dn_dst_ops
, dev_out
, 0, DST_OBSOLETE_NONE
, DST_HOST
);
1181 memset(&rt
->fld
, 0, sizeof(rt
->fld
));
1182 rt
->fld
.saddr
= oldflp
->saddr
;
1183 rt
->fld
.daddr
= oldflp
->daddr
;
1184 rt
->fld
.flowidn_oif
= oldflp
->flowidn_oif
;
1185 rt
->fld
.flowidn_iif
= 0;
1186 rt
->fld
.flowidn_mark
= oldflp
->flowidn_mark
;
1188 rt
->rt_saddr
= fld
.saddr
;
1189 rt
->rt_daddr
= fld
.daddr
;
1190 rt
->rt_gateway
= gateway
? gateway
: fld
.daddr
;
1191 rt
->rt_local_src
= fld
.saddr
;
1193 rt
->rt_dst_map
= fld
.daddr
;
1194 rt
->rt_src_map
= fld
.saddr
;
1199 rt
->dst
.lastuse
= jiffies
;
1200 rt
->dst
.output
= dn_output
;
1201 rt
->dst
.input
= dn_rt_bug
;
1202 rt
->rt_flags
= flags
;
1203 if (flags
& RTCF_LOCAL
)
1204 rt
->dst
.input
= dn_nsp_rx
;
1206 err
= dn_rt_set_next_hop(rt
, &res
);
1210 hash
= dn_hash(rt
->fld
.saddr
, rt
->fld
.daddr
);
1211 /* dn_insert_route() increments dst->__refcnt */
1212 dn_insert_route(rt
, hash
, (struct dn_route
**)pprt
);
1216 neigh_release(neigh
);
1218 dn_fib_res_put(&res
);
1225 err
= -EADDRNOTAVAIL
;
1234 dst_release_immediate(&rt
->dst
);
1240 * N.B. The flags may be moved into the flowi at some future stage.
1242 static int __dn_route_output_key(struct dst_entry
**pprt
, const struct flowidn
*flp
, int flags
)
1244 unsigned int hash
= dn_hash(flp
->saddr
, flp
->daddr
);
1245 struct dn_route
*rt
= NULL
;
1247 if (!(flags
& MSG_TRYHARD
)) {
1249 for (rt
= rcu_dereference_bh(dn_rt_hash_table
[hash
].chain
); rt
;
1250 rt
= rcu_dereference_bh(rt
->dn_next
)) {
1251 if ((flp
->daddr
== rt
->fld
.daddr
) &&
1252 (flp
->saddr
== rt
->fld
.saddr
) &&
1253 (flp
->flowidn_mark
== rt
->fld
.flowidn_mark
) &&
1254 dn_is_output_route(rt
) &&
1255 (rt
->fld
.flowidn_oif
== flp
->flowidn_oif
)) {
1256 dst_hold_and_use(&rt
->dst
, jiffies
);
1257 rcu_read_unlock_bh();
1262 rcu_read_unlock_bh();
1265 return dn_route_output_slow(pprt
, flp
, flags
);
1268 static int dn_route_output_key(struct dst_entry
**pprt
, struct flowidn
*flp
, int flags
)
1272 err
= __dn_route_output_key(pprt
, flp
, flags
);
1273 if (err
== 0 && flp
->flowidn_proto
) {
1274 *pprt
= xfrm_lookup(&init_net
, *pprt
,
1275 flowidn_to_flowi(flp
), NULL
, 0);
1276 if (IS_ERR(*pprt
)) {
1277 err
= PTR_ERR(*pprt
);
1284 int dn_route_output_sock(struct dst_entry __rcu
**pprt
, struct flowidn
*fl
, struct sock
*sk
, int flags
)
1288 err
= __dn_route_output_key(pprt
, fl
, flags
& MSG_TRYHARD
);
1289 if (err
== 0 && fl
->flowidn_proto
) {
1290 *pprt
= xfrm_lookup(&init_net
, *pprt
,
1291 flowidn_to_flowi(fl
), sk
, 0);
1292 if (IS_ERR(*pprt
)) {
1293 err
= PTR_ERR(*pprt
);
1300 static int dn_route_input_slow(struct sk_buff
*skb
)
1302 struct dn_route
*rt
= NULL
;
1303 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
1304 struct net_device
*in_dev
= skb
->dev
;
1305 struct net_device
*out_dev
= NULL
;
1306 struct dn_dev
*dn_db
;
1307 struct neighbour
*neigh
= NULL
;
1311 __le16 local_src
= 0;
1312 struct flowidn fld
= {
1315 .flowidn_scope
= RT_SCOPE_UNIVERSE
,
1316 .flowidn_mark
= skb
->mark
,
1317 .flowidn_iif
= skb
->dev
->ifindex
,
1319 struct dn_fib_res res
= { .fi
= NULL
, .type
= RTN_UNREACHABLE
};
1325 if ((dn_db
= rcu_dereference(in_dev
->dn_ptr
)) == NULL
)
1328 /* Zero source addresses are not allowed */
1333 * In this case we've just received a packet from a source
1334 * outside ourselves pretending to come from us. We don't
1335 * allow it any further to prevent routing loops, spoofing and
1336 * other nasties. Loopback packets already have the dst attached
1337 * so this only affects packets which have originated elsewhere.
1340 if (dn_dev_islocal(in_dev
, cb
->src
))
1343 err
= dn_fib_lookup(&fld
, &res
);
1348 * Is the destination us ?
1350 if (!dn_dev_islocal(in_dev
, cb
->dst
))
1353 res
.type
= RTN_LOCAL
;
1355 __le16 src_map
= fld
.saddr
;
1358 out_dev
= DN_FIB_RES_DEV(res
);
1359 if (out_dev
== NULL
) {
1360 net_crit_ratelimited("Bug in dn_route_input_slow() No output device\n");
1366 src_map
= fld
.saddr
; /* no NAT support for now */
1368 gateway
= DN_FIB_RES_GW(res
);
1369 if (res
.type
== RTN_NAT
) {
1370 fld
.daddr
= dn_fib_rules_map_destination(fld
.daddr
, &res
);
1371 dn_fib_res_put(&res
);
1373 if (dn_fib_lookup(&fld
, &res
))
1376 if (res
.type
!= RTN_UNICAST
)
1379 gateway
= fld
.daddr
;
1381 fld
.saddr
= src_map
;
1387 * Forwarding check here, we only check for forwarding
1388 * being turned off, if you want to only forward intra
1389 * area, its up to you to set the routing tables up
1392 if (dn_db
->parms
.forwarding
== 0)
1395 if (res
.fi
->fib_nhs
> 1 && fld
.flowidn_oif
== 0)
1396 dn_fib_select_multipath(&fld
, &res
);
1399 * Check for out_dev == in_dev. We use the RTCF_DOREDIRECT
1400 * flag as a hint to set the intra-ethernet bit when
1401 * forwarding. If we've got NAT in operation, we don't do
1402 * this optimisation.
1404 if (out_dev
== in_dev
&& !(flags
& RTCF_NAT
))
1405 flags
|= RTCF_DOREDIRECT
;
1407 local_src
= DN_FIB_RES_PREFSRC(res
);
1410 case RTN_UNREACHABLE
:
1413 flags
|= RTCF_LOCAL
;
1414 fld
.saddr
= cb
->dst
;
1415 fld
.daddr
= cb
->src
;
1417 /* Routing tables gave us a gateway */
1421 /* Packet was intra-ethernet, so we know its on-link */
1422 if (cb
->rt_flags
& DN_RT_F_IE
) {
1427 /* Use the default router if there is one */
1428 neigh
= neigh_clone(dn_db
->router
);
1430 gateway
= ((struct dn_neigh
*)neigh
)->addr
;
1434 /* Close eyes and pray */
1442 rt
= dst_alloc(&dn_dst_ops
, out_dev
, 1, DST_OBSOLETE_NONE
, DST_HOST
);
1447 memset(&rt
->fld
, 0, sizeof(rt
->fld
));
1448 rt
->rt_saddr
= fld
.saddr
;
1449 rt
->rt_daddr
= fld
.daddr
;
1450 rt
->rt_gateway
= fld
.daddr
;
1452 rt
->rt_gateway
= gateway
;
1453 rt
->rt_local_src
= local_src
? local_src
: rt
->rt_saddr
;
1455 rt
->rt_dst_map
= fld
.daddr
;
1456 rt
->rt_src_map
= fld
.saddr
;
1458 rt
->fld
.saddr
= cb
->src
;
1459 rt
->fld
.daddr
= cb
->dst
;
1460 rt
->fld
.flowidn_oif
= 0;
1461 rt
->fld
.flowidn_iif
= in_dev
->ifindex
;
1462 rt
->fld
.flowidn_mark
= fld
.flowidn_mark
;
1465 rt
->dst
.lastuse
= jiffies
;
1466 rt
->dst
.output
= dn_rt_bug_out
;
1469 rt
->dst
.input
= dn_forward
;
1472 rt
->dst
.output
= dn_output
;
1473 rt
->dst
.input
= dn_nsp_rx
;
1474 rt
->dst
.dev
= in_dev
;
1475 flags
|= RTCF_LOCAL
;
1478 case RTN_UNREACHABLE
:
1480 rt
->dst
.input
= dst_discard
;
1482 rt
->rt_flags
= flags
;
1484 err
= dn_rt_set_next_hop(rt
, &res
);
1488 hash
= dn_hash(rt
->fld
.saddr
, rt
->fld
.daddr
);
1489 /* dn_insert_route() increments dst->__refcnt */
1490 dn_insert_route(rt
, hash
, &rt
);
1491 skb_dst_set(skb
, &rt
->dst
);
1495 neigh_release(neigh
);
1497 dn_fib_res_put(&res
);
1513 dst_release_immediate(&rt
->dst
);
1517 static int dn_route_input(struct sk_buff
*skb
)
1519 struct dn_route
*rt
;
1520 struct dn_skb_cb
*cb
= DN_SKB_CB(skb
);
1521 unsigned int hash
= dn_hash(cb
->src
, cb
->dst
);
1527 for(rt
= rcu_dereference(dn_rt_hash_table
[hash
].chain
); rt
!= NULL
;
1528 rt
= rcu_dereference(rt
->dn_next
)) {
1529 if ((rt
->fld
.saddr
== cb
->src
) &&
1530 (rt
->fld
.daddr
== cb
->dst
) &&
1531 (rt
->fld
.flowidn_oif
== 0) &&
1532 (rt
->fld
.flowidn_mark
== skb
->mark
) &&
1533 (rt
->fld
.flowidn_iif
== cb
->iif
)) {
1534 dst_hold_and_use(&rt
->dst
, jiffies
);
1536 skb_dst_set(skb
, (struct dst_entry
*)rt
);
1542 return dn_route_input_slow(skb
);
1545 static int dn_rt_fill_info(struct sk_buff
*skb
, u32 portid
, u32 seq
,
1546 int event
, int nowait
, unsigned int flags
)
1548 struct dn_route
*rt
= (struct dn_route
*)skb_dst(skb
);
1550 struct nlmsghdr
*nlh
;
1553 nlh
= nlmsg_put(skb
, portid
, seq
, event
, sizeof(*r
), flags
);
1557 r
= nlmsg_data(nlh
);
1558 r
->rtm_family
= AF_DECnet
;
1559 r
->rtm_dst_len
= 16;
1562 r
->rtm_table
= RT_TABLE_MAIN
;
1563 r
->rtm_type
= rt
->rt_type
;
1564 r
->rtm_flags
= (rt
->rt_flags
& ~0xFFFF) | RTM_F_CLONED
;
1565 r
->rtm_scope
= RT_SCOPE_UNIVERSE
;
1566 r
->rtm_protocol
= RTPROT_UNSPEC
;
1568 if (rt
->rt_flags
& RTCF_NOTIFY
)
1569 r
->rtm_flags
|= RTM_F_NOTIFY
;
1571 if (nla_put_u32(skb
, RTA_TABLE
, RT_TABLE_MAIN
) < 0 ||
1572 nla_put_le16(skb
, RTA_DST
, rt
->rt_daddr
) < 0)
1575 if (rt
->fld
.saddr
) {
1576 r
->rtm_src_len
= 16;
1577 if (nla_put_le16(skb
, RTA_SRC
, rt
->fld
.saddr
) < 0)
1581 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
) < 0)
1585 * Note to self - change this if input routes reverse direction when
1586 * they deal only with inputs and not with replies like they do
1589 if (nla_put_le16(skb
, RTA_PREFSRC
, rt
->rt_local_src
) < 0)
1592 if (rt
->rt_daddr
!= rt
->rt_gateway
&&
1593 nla_put_le16(skb
, RTA_GATEWAY
, rt
->rt_gateway
) < 0)
1596 if (rtnetlink_put_metrics(skb
, dst_metrics_ptr(&rt
->dst
)) < 0)
1599 expires
= rt
->dst
.expires
? rt
->dst
.expires
- jiffies
: 0;
1600 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
,
1604 if (dn_is_input_route(rt
) &&
1605 nla_put_u32(skb
, RTA_IIF
, rt
->fld
.flowidn_iif
) < 0)
1608 nlmsg_end(skb
, nlh
);
1612 nlmsg_cancel(skb
, nlh
);
1616 const struct nla_policy rtm_dn_policy
[RTA_MAX
+ 1] = {
1617 [RTA_DST
] = { .type
= NLA_U16
},
1618 [RTA_SRC
] = { .type
= NLA_U16
},
1619 [RTA_IIF
] = { .type
= NLA_U32
},
1620 [RTA_OIF
] = { .type
= NLA_U32
},
1621 [RTA_GATEWAY
] = { .type
= NLA_U16
},
1622 [RTA_PRIORITY
] = { .type
= NLA_U32
},
1623 [RTA_PREFSRC
] = { .type
= NLA_U16
},
1624 [RTA_METRICS
] = { .type
= NLA_NESTED
},
1625 [RTA_MULTIPATH
] = { .type
= NLA_NESTED
},
1626 [RTA_TABLE
] = { .type
= NLA_U32
},
1627 [RTA_MARK
] = { .type
= NLA_U32
},
1631 * This is called by both endnodes and routers now.
1633 static int dn_cache_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
,
1634 struct netlink_ext_ack
*extack
)
1636 struct net
*net
= sock_net(in_skb
->sk
);
1637 struct rtmsg
*rtm
= nlmsg_data(nlh
);
1638 struct dn_route
*rt
= NULL
;
1639 struct dn_skb_cb
*cb
;
1641 struct sk_buff
*skb
;
1643 struct nlattr
*tb
[RTA_MAX
+1];
1645 if (!net_eq(net
, &init_net
))
1648 err
= nlmsg_parse_deprecated(nlh
, sizeof(*rtm
), tb
, RTA_MAX
,
1649 rtm_dn_policy
, extack
);
1653 memset(&fld
, 0, sizeof(fld
));
1654 fld
.flowidn_proto
= DNPROTO_NSP
;
1656 skb
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1659 skb_reset_mac_header(skb
);
1660 cb
= DN_SKB_CB(skb
);
1663 fld
.saddr
= nla_get_le16(tb
[RTA_SRC
]);
1666 fld
.daddr
= nla_get_le16(tb
[RTA_DST
]);
1669 fld
.flowidn_iif
= nla_get_u32(tb
[RTA_IIF
]);
1671 if (fld
.flowidn_iif
) {
1672 struct net_device
*dev
;
1673 dev
= __dev_get_by_index(&init_net
, fld
.flowidn_iif
);
1674 if (!dev
|| !dev
->dn_ptr
) {
1678 skb
->protocol
= htons(ETH_P_DNA_RT
);
1680 cb
->src
= fld
.saddr
;
1681 cb
->dst
= fld
.daddr
;
1683 err
= dn_route_input(skb
);
1685 memset(cb
, 0, sizeof(struct dn_skb_cb
));
1686 rt
= (struct dn_route
*)skb_dst(skb
);
1687 if (!err
&& -rt
->dst
.error
)
1688 err
= rt
->dst
.error
;
1691 fld
.flowidn_oif
= nla_get_u32(tb
[RTA_OIF
]);
1693 err
= dn_route_output_key((struct dst_entry
**)&rt
, &fld
, 0);
1699 skb_dst_set(skb
, &rt
->dst
);
1700 if (rtm
->rtm_flags
& RTM_F_NOTIFY
)
1701 rt
->rt_flags
|= RTCF_NOTIFY
;
1703 err
= dn_rt_fill_info(skb
, NETLINK_CB(in_skb
).portid
, nlh
->nlmsg_seq
, RTM_NEWROUTE
, 0, 0);
1709 return rtnl_unicast(skb
, &init_net
, NETLINK_CB(in_skb
).portid
);
1717 * For routers, this is called from dn_fib_dump, but for endnodes its
1718 * called directly from the rtnetlink dispatch table.
1720 int dn_cache_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1722 struct net
*net
= sock_net(skb
->sk
);
1723 struct dn_route
*rt
;
1728 if (!net_eq(net
, &init_net
))
1731 if (nlmsg_len(cb
->nlh
) < sizeof(struct rtmsg
))
1734 rtm
= nlmsg_data(cb
->nlh
);
1735 if (!(rtm
->rtm_flags
& RTM_F_CLONED
))
1739 s_idx
= idx
= cb
->args
[1];
1740 for(h
= 0; h
<= dn_rt_hash_mask
; h
++) {
1746 for(rt
= rcu_dereference_bh(dn_rt_hash_table
[h
].chain
), idx
= 0;
1748 rt
= rcu_dereference_bh(rt
->dn_next
), idx
++) {
1751 skb_dst_set(skb
, dst_clone(&rt
->dst
));
1752 if (dn_rt_fill_info(skb
, NETLINK_CB(cb
->skb
).portid
,
1753 cb
->nlh
->nlmsg_seq
, RTM_NEWROUTE
,
1754 1, NLM_F_MULTI
) < 0) {
1756 rcu_read_unlock_bh();
1761 rcu_read_unlock_bh();
1770 #ifdef CONFIG_PROC_FS
1771 struct dn_rt_cache_iter_state
{
1775 static struct dn_route
*dn_rt_cache_get_first(struct seq_file
*seq
)
1777 struct dn_route
*rt
= NULL
;
1778 struct dn_rt_cache_iter_state
*s
= seq
->private;
1780 for(s
->bucket
= dn_rt_hash_mask
; s
->bucket
>= 0; --s
->bucket
) {
1782 rt
= rcu_dereference_bh(dn_rt_hash_table
[s
->bucket
].chain
);
1785 rcu_read_unlock_bh();
1790 static struct dn_route
*dn_rt_cache_get_next(struct seq_file
*seq
, struct dn_route
*rt
)
1792 struct dn_rt_cache_iter_state
*s
= seq
->private;
1794 rt
= rcu_dereference_bh(rt
->dn_next
);
1796 rcu_read_unlock_bh();
1797 if (--s
->bucket
< 0)
1800 rt
= rcu_dereference_bh(dn_rt_hash_table
[s
->bucket
].chain
);
1805 static void *dn_rt_cache_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1807 struct dn_route
*rt
= dn_rt_cache_get_first(seq
);
1810 while(*pos
&& (rt
= dn_rt_cache_get_next(seq
, rt
)))
1813 return *pos
? NULL
: rt
;
1816 static void *dn_rt_cache_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1818 struct dn_route
*rt
= dn_rt_cache_get_next(seq
, v
);
1823 static void dn_rt_cache_seq_stop(struct seq_file
*seq
, void *v
)
1826 rcu_read_unlock_bh();
1829 static int dn_rt_cache_seq_show(struct seq_file
*seq
, void *v
)
1831 struct dn_route
*rt
= v
;
1832 char buf1
[DN_ASCBUF_LEN
], buf2
[DN_ASCBUF_LEN
];
1834 seq_printf(seq
, "%-8s %-7s %-7s %04d %04d %04d\n",
1835 rt
->dst
.dev
? rt
->dst
.dev
->name
: "*",
1836 dn_addr2asc(le16_to_cpu(rt
->rt_daddr
), buf1
),
1837 dn_addr2asc(le16_to_cpu(rt
->rt_saddr
), buf2
),
1838 atomic_read(&rt
->dst
.__refcnt
),
1843 static const struct seq_operations dn_rt_cache_seq_ops
= {
1844 .start
= dn_rt_cache_seq_start
,
1845 .next
= dn_rt_cache_seq_next
,
1846 .stop
= dn_rt_cache_seq_stop
,
1847 .show
= dn_rt_cache_seq_show
,
1849 #endif /* CONFIG_PROC_FS */
1851 void __init
dn_route_init(void)
1855 dn_dst_ops
.kmem_cachep
=
1856 kmem_cache_create("dn_dst_cache", sizeof(struct dn_route
), 0,
1857 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
1858 dst_entries_init(&dn_dst_ops
);
1859 timer_setup(&dn_route_timer
, dn_dst_check_expire
, 0);
1860 dn_route_timer
.expires
= jiffies
+ decnet_dst_gc_interval
* HZ
;
1861 add_timer(&dn_route_timer
);
1863 goal
= totalram_pages() >> (26 - PAGE_SHIFT
);
1865 for(order
= 0; (1UL << order
) < goal
; order
++)
1869 * Only want 1024 entries max, since the table is very, very unlikely
1870 * to be larger than that.
1872 while(order
&& ((((1UL << order
) * PAGE_SIZE
) /
1873 sizeof(struct dn_rt_hash_bucket
)) >= 2048))
1877 dn_rt_hash_mask
= (1UL << order
) * PAGE_SIZE
/
1878 sizeof(struct dn_rt_hash_bucket
);
1879 while(dn_rt_hash_mask
& (dn_rt_hash_mask
- 1))
1881 dn_rt_hash_table
= (struct dn_rt_hash_bucket
*)
1882 __get_free_pages(GFP_ATOMIC
, order
);
1883 } while (dn_rt_hash_table
== NULL
&& --order
> 0);
1885 if (!dn_rt_hash_table
)
1886 panic("Failed to allocate DECnet route cache hash table\n");
1889 "DECnet: Routing cache hash table of %u buckets, %ldKbytes\n",
1891 (long)(dn_rt_hash_mask
*sizeof(struct dn_rt_hash_bucket
))/1024);
1894 for(i
= 0; i
<= dn_rt_hash_mask
; i
++) {
1895 spin_lock_init(&dn_rt_hash_table
[i
].lock
);
1896 dn_rt_hash_table
[i
].chain
= NULL
;
1899 dn_dst_ops
.gc_thresh
= (dn_rt_hash_mask
+ 1);
1901 proc_create_seq_private("decnet_cache", 0444, init_net
.proc_net
,
1902 &dn_rt_cache_seq_ops
,
1903 sizeof(struct dn_rt_cache_iter_state
), NULL
);
1905 #ifdef CONFIG_DECNET_ROUTER
1906 rtnl_register_module(THIS_MODULE
, PF_DECnet
, RTM_GETROUTE
,
1907 dn_cache_getroute
, dn_fib_dump
, 0);
1909 rtnl_register_module(THIS_MODULE
, PF_DECnet
, RTM_GETROUTE
,
1910 dn_cache_getroute
, dn_cache_dump
, 0);
1914 void __exit
dn_route_cleanup(void)
1916 del_timer(&dn_route_timer
);
1919 remove_proc_entry("decnet_cache", init_net
.proc_net
);
1920 dst_entries_destroy(&dn_dst_ops
);