File: /usr/src/linux/net/802/tr.c
1 /*
2 * NET3: Token ring device handling subroutines
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
8 *
9 * Fixes: 3 Feb 97 Paul Norton <pnorton@cts.com> Minor routing fixes.
10 * Added rif table to /proc/net/tr_rif and rif timeout to
11 * /proc/sys/net/token-ring/rif_timeout.
12 * 22 Jun 98 Paul Norton <p.norton@computer.org> Rearranged
13 * tr_header and tr_type_trans to handle passing IPX SNAP and
14 * 802.2 through the correct layers. Eliminated tr_reformat.
15 *
16 */
17
18 #include <asm/uaccess.h>
19 #include <asm/system.h>
20 #include <linux/config.h>
21 #include <linux/types.h>
22 #include <linux/kernel.h>
23 #include <linux/sched.h>
24 #include <linux/string.h>
25 #include <linux/mm.h>
26 #include <linux/socket.h>
27 #include <linux/in.h>
28 #include <linux/inet.h>
29 #include <linux/netdevice.h>
30 #include <linux/trdevice.h>
31 #include <linux/skbuff.h>
32 #include <linux/errno.h>
33 #include <linux/timer.h>
34 #include <linux/net.h>
35 #include <linux/proc_fs.h>
36 #include <linux/init.h>
37 #include <net/arp.h>
38
39 static void tr_source_route(struct sk_buff *skb, struct trh_hdr *trh, struct net_device *dev);
40 static void tr_add_rif_info(struct trh_hdr *trh, struct net_device *dev);
41 static void rif_check_expire(unsigned long dummy);
42
43 #define TR_SR_DEBUG 0
44
45 typedef struct rif_cache_s *rif_cache;
46
47 /*
48 * Each RIF entry we learn is kept this way
49 */
50
51 struct rif_cache_s {
52 unsigned char addr[TR_ALEN];
53 int iface;
54 __u16 rcf;
55 __u16 rseg[8];
56 rif_cache next;
57 unsigned long last_used;
58 unsigned char local_ring;
59 };
60
61 #define RIF_TABLE_SIZE 32
62
63 /*
64 * We hash the RIF cache 32 ways. We do after all have to look it
65 * up a lot.
66 */
67
68 rif_cache rif_table[RIF_TABLE_SIZE]={ NULL, };
69
70 static spinlock_t rif_lock = SPIN_LOCK_UNLOCKED;
71
72 #define RIF_TIMEOUT 60*10*HZ
73 #define RIF_CHECK_INTERVAL 60*HZ
74
75 /*
76 * Garbage disposal timer.
77 */
78
79 static struct timer_list rif_timer;
80
81 int sysctl_tr_rif_timeout = RIF_TIMEOUT;
82
83 /*
84 * Put the headers on a token ring packet. Token ring source routing
85 * makes this a little more exciting than on ethernet.
86 */
87
88 int tr_header(struct sk_buff *skb, struct net_device *dev, unsigned short type,
89 void *daddr, void *saddr, unsigned len)
90 {
91 struct trh_hdr *trh;
92 int hdr_len;
93
94 /*
95 * Add the 802.2 SNAP header if IP as the IPv4 code calls
96 * dev->hard_header directly.
97 */
98 if (type == ETH_P_IP || type == ETH_P_ARP)
99 {
100 struct trllc *trllc=(struct trllc *)(trh+1);
101
102 hdr_len = sizeof(struct trh_hdr) + sizeof(struct trllc);
103 trh = (struct trh_hdr *)skb_push(skb, hdr_len);
104 trllc = (struct trllc *)(trh+1);
105 trllc->dsap = trllc->ssap = EXTENDED_SAP;
106 trllc->llc = UI_CMD;
107 trllc->protid[0] = trllc->protid[1] = trllc->protid[2] = 0x00;
108 trllc->ethertype = htons(type);
109 }
110 else
111 {
112 hdr_len = sizeof(struct trh_hdr);
113 trh = (struct trh_hdr *)skb_push(skb, hdr_len);
114 }
115
116 trh->ac=AC;
117 trh->fc=LLC_FRAME;
118
119 if(saddr)
120 memcpy(trh->saddr,saddr,dev->addr_len);
121 else
122 memcpy(trh->saddr,dev->dev_addr,dev->addr_len);
123
124 /*
125 * Build the destination and then source route the frame
126 */
127
128 if(daddr)
129 {
130 memcpy(trh->daddr,daddr,dev->addr_len);
131 tr_source_route(skb,trh,dev);
132 return(hdr_len);
133 }
134
135 return -hdr_len;
136 }
137
138 /*
139 * A neighbour discovery of some species (eg arp) has completed. We
140 * can now send the packet.
141 */
142
143 int tr_rebuild_header(struct sk_buff *skb)
144 {
145 struct trh_hdr *trh=(struct trh_hdr *)skb->data;
146 struct trllc *trllc=(struct trllc *)(skb->data+sizeof(struct trh_hdr));
147 struct net_device *dev = skb->dev;
148
149 /*
150 * FIXME: We don't yet support IPv6 over token rings
151 */
152
153 if(trllc->ethertype != htons(ETH_P_IP)) {
154 printk("tr_rebuild_header: Don't know how to resolve type %04X addresses ?\n",(unsigned int)htons(trllc->ethertype));
155 return 0;
156 }
157
158 #ifdef CONFIG_INET
159 if(arp_find(trh->daddr, skb)) {
160 return 1;
161 }
162 else
163 #endif
164 {
165 tr_source_route(skb,trh,dev);
166 return 0;
167 }
168 }
169
170 /*
171 * Some of this is a bit hackish. We intercept RIF information
172 * used for source routing. We also grab IP directly and don't feed
173 * it via SNAP.
174 */
175
176 unsigned short tr_type_trans(struct sk_buff *skb, struct net_device *dev)
177 {
178
179 struct trh_hdr *trh=(struct trh_hdr *)skb->data;
180 struct trllc *trllc;
181 unsigned riflen=0;
182
183 skb->mac.raw = skb->data;
184
185 if(trh->saddr[0] & TR_RII)
186 riflen = (ntohs(trh->rcf) & TR_RCF_LEN_MASK) >> 8;
187
188 trllc = (struct trllc *)(skb->data+sizeof(struct trh_hdr)-TR_MAXRIFLEN+riflen);
189
190 skb_pull(skb,sizeof(struct trh_hdr)-TR_MAXRIFLEN+riflen);
191
192 if(*trh->daddr & 0x80)
193 {
194 if(!memcmp(trh->daddr,dev->broadcast,TR_ALEN))
195 skb->pkt_type=PACKET_BROADCAST;
196 else
197 skb->pkt_type=PACKET_MULTICAST;
198 }
199 else if ( (trh->daddr[0] & 0x01) && (trh->daddr[1] & 0x00) && (trh->daddr[2] & 0x5E))
200 {
201 skb->pkt_type=PACKET_MULTICAST;
202 }
203 else if(dev->flags & IFF_PROMISC)
204 {
205 if(memcmp(trh->daddr, dev->dev_addr, TR_ALEN))
206 skb->pkt_type=PACKET_OTHERHOST;
207 }
208
209 if ((skb->pkt_type != PACKET_BROADCAST) &&
210 (skb->pkt_type != PACKET_MULTICAST))
211 tr_add_rif_info(trh,dev) ;
212
213 /*
214 * Strip the SNAP header from ARP packets since we don't
215 * pass them through to the 802.2/SNAP layers.
216 */
217
218 if (trllc->dsap == EXTENDED_SAP &&
219 (trllc->ethertype == ntohs(ETH_P_IP) ||
220 trllc->ethertype == ntohs(ETH_P_ARP)))
221 {
222 skb_pull(skb, sizeof(struct trllc));
223 return trllc->ethertype;
224 }
225
226 return ntohs(ETH_P_802_2);
227 }
228
229 /*
230 * We try to do source routing...
231 */
232
233 static void tr_source_route(struct sk_buff *skb,struct trh_hdr *trh,struct net_device *dev)
234 {
235 int i, slack;
236 unsigned int hash;
237 rif_cache entry;
238 unsigned char *olddata;
239 unsigned char mcast_func_addr[] = {0xC0,0x00,0x00,0x04,0x00,0x00};
240 unsigned long flags ;
241
242 spin_lock_irqsave(&rif_lock,flags);
243
244 /*
245 * Broadcasts are single route as stated in RFC 1042
246 */
247 if( (!memcmp(&(trh->daddr[0]),&(dev->broadcast[0]),TR_ALEN)) ||
248 (!memcmp(&(trh->daddr[0]),&(mcast_func_addr[0]), TR_ALEN)) )
249 {
250 trh->rcf=htons((((sizeof(trh->rcf)) << 8) & TR_RCF_LEN_MASK)
251 | TR_RCF_FRAME2K | TR_RCF_LIMITED_BROADCAST);
252 trh->saddr[0]|=TR_RII;
253 }
254 else
255 {
256 for(i=0,hash=0;i<TR_ALEN;hash+=trh->daddr[i++]);
257 hash&=RIF_TABLE_SIZE-1;
258 /*
259 * Walk the hash table and look for an entry
260 */
261 for(entry=rif_table[hash];entry && memcmp(&(entry->addr[0]),&(trh->daddr[0]),TR_ALEN);entry=entry->next);
262
263 /*
264 * If we found an entry we can route the frame.
265 */
266 if(entry)
267 {
268 #if TR_SR_DEBUG
269 printk("source routing for %02X:%02X:%02X:%02X:%02X:%02X\n",trh->daddr[0],
270 trh->daddr[1],trh->daddr[2],trh->daddr[3],trh->daddr[4],trh->daddr[5]);
271 #endif
272 if(!entry->local_ring && (ntohs(entry->rcf) & TR_RCF_LEN_MASK) >> 8)
273 {
274 trh->rcf=entry->rcf;
275 memcpy(&trh->rseg[0],&entry->rseg[0],8*sizeof(unsigned short));
276 trh->rcf^=htons(TR_RCF_DIR_BIT);
277 trh->rcf&=htons(0x1fff); /* Issam Chehab <ichehab@madge1.demon.co.uk> */
278
279 trh->saddr[0]|=TR_RII;
280 #if TR_SR_DEBUG
281 printk("entry found with rcf %04x\n", entry->rcf);
282 }
283 else
284 {
285 printk("entry found but without rcf length, local=%02x\n", entry->local_ring);
286 #endif
287 }
288 entry->last_used=jiffies;
289 }
290 else
291 {
292 /*
293 * Without the information we simply have to shout
294 * on the wire. The replies should rapidly clean this
295 * situation up.
296 */
297 trh->rcf=htons((((sizeof(trh->rcf)) << 8) & TR_RCF_LEN_MASK)
298 | TR_RCF_FRAME2K | TR_RCF_LIMITED_BROADCAST);
299 trh->saddr[0]|=TR_RII;
300 #if TR_SR_DEBUG
301 printk("no entry in rif table found - broadcasting frame\n");
302 #endif
303 }
304 }
305
306 /* Compress the RIF here so we don't have to do it in the driver(s) */
307 if (!(trh->saddr[0] & 0x80))
308 slack = 18;
309 else
310 slack = 18 - ((ntohs(trh->rcf) & TR_RCF_LEN_MASK)>>8);
311 olddata = skb->data;
312 spin_unlock_irqrestore(&rif_lock,flags);
313
314 skb_pull(skb, slack);
315 memmove(skb->data, olddata, sizeof(struct trh_hdr) - slack);
316 }
317
318 /*
319 * We have learned some new RIF information for our source
320 * routing.
321 */
322
323 static void tr_add_rif_info(struct trh_hdr *trh, struct net_device *dev)
324 {
325 int i;
326 unsigned int hash, rii_p = 0;
327 rif_cache entry;
328
329
330 spin_lock_bh(&rif_lock);
331
332 /*
333 * Firstly see if the entry exists
334 */
335
336 if(trh->saddr[0] & TR_RII)
337 {
338 trh->saddr[0]&=0x7f;
339 if (((ntohs(trh->rcf) & TR_RCF_LEN_MASK) >> 8) > 2)
340 {
341 rii_p = 1;
342 }
343 }
344
345 for(i=0,hash=0;i<TR_ALEN;hash+=trh->saddr[i++]);
346 hash&=RIF_TABLE_SIZE-1;
347 for(entry=rif_table[hash];entry && memcmp(&(entry->addr[0]),&(trh->saddr[0]),TR_ALEN);entry=entry->next);
348
349 if(entry==NULL)
350 {
351 #if TR_SR_DEBUG
352 printk("adding rif_entry: addr:%02X:%02X:%02X:%02X:%02X:%02X rcf:%04X\n",
353 trh->saddr[0],trh->saddr[1],trh->saddr[2],
354 trh->saddr[3],trh->saddr[4],trh->saddr[5],
355 ntohs(trh->rcf));
356 #endif
357 /*
358 * Allocate our new entry. A failure to allocate loses
359 * use the information. This is harmless.
360 *
361 * FIXME: We ought to keep some kind of cache size
362 * limiting and adjust the timers to suit.
363 */
364 entry=kmalloc(sizeof(struct rif_cache_s),GFP_ATOMIC);
365
366 if(!entry)
367 {
368 printk(KERN_DEBUG "tr.c: Couldn't malloc rif cache entry !\n");
369 spin_unlock_bh(&rif_lock);
370 return;
371 }
372
373 memcpy(&(entry->addr[0]),&(trh->saddr[0]),TR_ALEN);
374 entry->iface = dev->ifindex;
375 entry->next=rif_table[hash];
376 entry->last_used=jiffies;
377 rif_table[hash]=entry;
378
379 if (rii_p)
380 {
381 entry->rcf = trh->rcf & htons((unsigned short)~TR_RCF_BROADCAST_MASK);
382 memcpy(&(entry->rseg[0]),&(trh->rseg[0]),8*sizeof(unsigned short));
383 entry->local_ring = 0;
384 trh->saddr[0]|=TR_RII; /* put the routing indicator back for tcpdump */
385 }
386 else
387 {
388 entry->local_ring = 1;
389 }
390 }
391 else /* Y. Tahara added */
392 {
393 /*
394 * Update existing entries
395 */
396 if (!entry->local_ring)
397 if (entry->rcf != (trh->rcf & htons((unsigned short)~TR_RCF_BROADCAST_MASK)) &&
398 !(trh->rcf & htons(TR_RCF_BROADCAST_MASK)))
399 {
400 #if TR_SR_DEBUG
401 printk("updating rif_entry: addr:%02X:%02X:%02X:%02X:%02X:%02X rcf:%04X\n",
402 trh->saddr[0],trh->saddr[1],trh->saddr[2],
403 trh->saddr[3],trh->saddr[4],trh->saddr[5],
404 ntohs(trh->rcf));
405 #endif
406 entry->rcf = trh->rcf & htons((unsigned short)~TR_RCF_BROADCAST_MASK);
407 memcpy(&(entry->rseg[0]),&(trh->rseg[0]),8*sizeof(unsigned short));
408 }
409 entry->last_used=jiffies;
410 }
411 spin_unlock_bh(&rif_lock);
412 }
413
414 /*
415 * Scan the cache with a timer and see what we need to throw out.
416 */
417
418 static void rif_check_expire(unsigned long dummy)
419 {
420 int i;
421 unsigned long now=jiffies;
422 unsigned long flags ;
423
424 spin_lock_irqsave(&rif_lock,flags);
425
426 for(i=0; i < RIF_TABLE_SIZE;i++)
427 {
428 rif_cache entry, *pentry=rif_table+i;
429 while((entry=*pentry))
430 {
431 /*
432 * Out it goes
433 */
434 if((now-entry->last_used) > sysctl_tr_rif_timeout)
435 {
436 *pentry=entry->next;
437 kfree(entry);
438 }
439 else
440 pentry=&entry->next;
441 }
442 }
443
444 spin_unlock_irqrestore(&rif_lock,flags);
445
446 /*
447 * Reset the timer
448 */
449
450 mod_timer(&rif_timer, jiffies+sysctl_tr_rif_timeout);
451
452 }
453
454 /*
455 * Generate the /proc/net information for the token ring RIF
456 * routing.
457 */
458
459 #ifndef CONFIG_PROC_FS
460 static int rif_get_info(char *buffer,char **start, off_t offset, int length) { return 0;}
461 #else
462 static int rif_get_info(char *buffer,char **start, off_t offset, int length)
463 {
464 int len=0;
465 off_t begin=0;
466 off_t pos=0;
467 int size,i,j,rcf_len,segment,brdgnmb;
468 unsigned long now=jiffies;
469
470 rif_cache entry;
471
472 size=sprintf(buffer,
473 "if TR address TTL rcf routing segments\n");
474 pos+=size;
475 len+=size;
476
477 spin_lock_bh(&rif_lock);
478 for(i=0;i < RIF_TABLE_SIZE;i++)
479 {
480 for(entry=rif_table[i];entry;entry=entry->next) {
481 struct net_device *dev = __dev_get_by_index(entry->iface);
482
483 size=sprintf(buffer+len,"%s %02X:%02X:%02X:%02X:%02X:%02X %7li ",
484 dev?dev->name:"?",entry->addr[0],entry->addr[1],entry->addr[2],entry->addr[3],entry->addr[4],entry->addr[5],
485 sysctl_tr_rif_timeout-(now-entry->last_used));
486 len+=size;
487 pos=begin+len;
488 if (entry->local_ring)
489 size=sprintf(buffer+len,"local\n");
490 else {
491 size=sprintf(buffer+len,"%04X", ntohs(entry->rcf));
492 rcf_len = ((ntohs(entry->rcf) & TR_RCF_LEN_MASK)>>8)-2;
493 if (rcf_len)
494 rcf_len >>= 1;
495 for(j = 1; j < rcf_len; j++) {
496 if(j==1) {
497 segment=ntohs(entry->rseg[j-1])>>4;
498 len+=size;
499 pos=begin+len;
500 size=sprintf(buffer+len," %03X",segment);
501 };
502 segment=ntohs(entry->rseg[j])>>4;
503 brdgnmb=ntohs(entry->rseg[j-1])&0x00f;
504 len+=size;
505 pos=begin+len;
506 size=sprintf(buffer+len,"-%01X-%03X",brdgnmb,segment);
507 }
508 len+=size;
509 pos=begin+len;
510 size=sprintf(buffer+len,"\n");
511 }
512 len+=size;
513 pos=begin+len;
514
515 if(pos<offset)
516 {
517 len=0;
518 begin=pos;
519 }
520 if(pos>offset+length)
521 break;
522 }
523 if(pos>offset+length)
524 break;
525 }
526 spin_unlock_bh(&rif_lock);
527
528 *start=buffer+(offset-begin); /* Start of wanted data */
529 len-=(offset-begin); /* Start slop */
530 if(len>length)
531 len=length; /* Ending slop */
532 if (len<0)
533 len=0;
534 return len;
535 }
536 #endif
537
538 /*
539 * Called during bootup. We don't actually have to initialise
540 * too much for this.
541 */
542
543 static int __init rif_init(void)
544 {
545 rif_timer.expires = RIF_TIMEOUT;
546 rif_timer.data = 0L;
547 rif_timer.function = rif_check_expire;
548 init_timer(&rif_timer);
549 add_timer(&rif_timer);
550
551 proc_net_create("tr_rif",0,rif_get_info);
552 return 0;
553 }
554
555 module_init(rif_init);
556