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/module.h> 21#include <linux/types.h> 22#include <linux/kernel.h> 23#include <linux/jiffies.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/seq_file.h> 37#include <linux/init.h> 38#include <net/arp.h> 39 40static void tr_add_rif_info(struct trh_hdr *trh, struct net_device *dev); 41static void rif_check_expire(unsigned long dummy); 42 43#define TR_SR_DEBUG 0 44 45/* 46 * Each RIF entry we learn is kept this way 47 */ 48 49struct rif_cache { 50 unsigned char addr[TR_ALEN]; 51 int iface; 52 __be16 rcf; 53 __be16 rseg[8]; 54 struct rif_cache *next; 55 unsigned long last_used; 56 unsigned char local_ring; 57}; 58 59#define RIF_TABLE_SIZE 32 60 61/* 62 * We hash the RIF cache 32 ways. We do after all have to look it 63 * up a lot. 64 */ 65 66static struct rif_cache *rif_table[RIF_TABLE_SIZE]; 67 68static DEFINE_SPINLOCK(rif_lock); 69 70 71/* 72 * Garbage disposal timer. 73 */ 74 75static struct timer_list rif_timer; 76 77int sysctl_tr_rif_timeout = 60*10*HZ; 78 79static inline unsigned long rif_hash(const unsigned char *addr) 80{ 81 unsigned long x; 82 83 x = addr[0]; 84 x = (x << 2) ^ addr[1]; 85 x = (x << 2) ^ addr[2]; 86 x = (x << 2) ^ addr[3]; 87 x = (x << 2) ^ addr[4]; 88 x = (x << 2) ^ addr[5]; 89 90 x ^= x >> 8; 91 92 return x & (RIF_TABLE_SIZE - 1); 93} 94 95/* 96 * Put the headers on a token ring packet. Token ring source routing 97 * makes this a little more exciting than on ethernet. 98 */ 99 100static int tr_header(struct sk_buff *skb, struct net_device *dev, 101 unsigned short type, 102 void *daddr, void *saddr, unsigned len) 103{ 104 struct trh_hdr *trh; 105 int hdr_len; 106 107 /* 108 * Add the 802.2 SNAP header if IP as the IPv4/IPv6 code calls 109 * dev->hard_header directly. 110 */ 111 if (type == ETH_P_IP || type == ETH_P_IPV6 || type == ETH_P_ARP) 112 { 113 struct trllc *trllc; 114 115 hdr_len = sizeof(struct trh_hdr) + sizeof(struct trllc); 116 trh = (struct trh_hdr *)skb_push(skb, hdr_len); 117 trllc = (struct trllc *)(trh+1); 118 trllc->dsap = trllc->ssap = EXTENDED_SAP; 119 trllc->llc = UI_CMD; 120 trllc->protid[0] = trllc->protid[1] = trllc->protid[2] = 0x00; 121 trllc->ethertype = htons(type); 122 } 123 else 124 { 125 hdr_len = sizeof(struct trh_hdr); 126 trh = (struct trh_hdr *)skb_push(skb, hdr_len); 127 } 128 129 trh->ac=AC; 130 trh->fc=LLC_FRAME; 131 132 if(saddr) 133 memcpy(trh->saddr,saddr,dev->addr_len); 134 else 135 memcpy(trh->saddr,dev->dev_addr,dev->addr_len); 136 137 /* 138 * Build the destination and then source route the frame 139 */ 140 141 if(daddr) 142 { 143 memcpy(trh->daddr,daddr,dev->addr_len); 144 tr_source_route(skb,trh,dev); 145 return(hdr_len); 146 } 147 148 return -hdr_len; 149} 150 151/* 152 * A neighbour discovery of some species (eg arp) has completed. We 153 * can now send the packet. 154 */ 155 156static int tr_rebuild_header(struct sk_buff *skb) 157{ 158 struct trh_hdr *trh=(struct trh_hdr *)skb->data; 159 struct trllc *trllc=(struct trllc *)(skb->data+sizeof(struct trh_hdr)); 160 struct net_device *dev = skb->dev; 161 162 163 if(trllc->ethertype != htons(ETH_P_IP)) { 164 printk("tr_rebuild_header: Don't know how to resolve type %04X addresses ?\n", ntohs(trllc->ethertype)); 165 return 0; 166 } 167 168#ifdef CONFIG_INET 169 if(arp_find(trh->daddr, skb)) { 170 return 1; 171 } 172 else 173#endif 174 { 175 tr_source_route(skb,trh,dev); 176 return 0; 177 } 178} 179 180/* 181 * Some of this is a bit hackish. We intercept RIF information 182 * used for source routing. We also grab IP directly and don't feed 183 * it via SNAP. 184 */ 185 186__be16 tr_type_trans(struct sk_buff *skb, struct net_device *dev) 187{ 188 189 struct trh_hdr *trh; 190 struct trllc *trllc; 191 unsigned riflen=0; 192 193 skb->dev = dev; 194 skb_reset_mac_header(skb); 195 trh = tr_hdr(skb); 196 197 if(trh->saddr[0] & TR_RII) 198 riflen = (ntohs(trh->rcf) & TR_RCF_LEN_MASK) >> 8; 199 200 trllc = (struct trllc *)(skb->data+sizeof(struct trh_hdr)-TR_MAXRIFLEN+riflen); 201 202 skb_pull(skb,sizeof(struct trh_hdr)-TR_MAXRIFLEN+riflen); 203 204 if(*trh->daddr & 0x80) 205 { 206 if(!memcmp(trh->daddr,dev->broadcast,TR_ALEN)) 207 skb->pkt_type=PACKET_BROADCAST; 208 else 209 skb->pkt_type=PACKET_MULTICAST; 210 } 211 else if ( (trh->daddr[0] & 0x01) && (trh->daddr[1] & 0x00) && (trh->daddr[2] & 0x5E)) 212 { 213 skb->pkt_type=PACKET_MULTICAST; 214 } 215 else if(dev->flags & IFF_PROMISC) 216 { 217 if(memcmp(trh->daddr, dev->dev_addr, TR_ALEN)) 218 skb->pkt_type=PACKET_OTHERHOST; 219 } 220 221 if ((skb->pkt_type != PACKET_BROADCAST) && 222 (skb->pkt_type != PACKET_MULTICAST)) 223 tr_add_rif_info(trh,dev) ; 224 225 /* 226 * Strip the SNAP header from ARP packets since we don't 227 * pass them through to the 802.2/SNAP layers. 228 */ 229 230 if (trllc->dsap == EXTENDED_SAP && 231 (trllc->ethertype == htons(ETH_P_IP) || 232 trllc->ethertype == htons(ETH_P_IPV6) || 233 trllc->ethertype == htons(ETH_P_ARP))) 234 { 235 skb_pull(skb, sizeof(struct trllc)); 236 return trllc->ethertype; 237 } 238 239 return htons(ETH_P_TR_802_2); 240} 241 242/* 243 * We try to do source routing... 244 */ 245 246void tr_source_route(struct sk_buff *skb,struct trh_hdr *trh,struct net_device *dev) 247{ 248 int slack; 249 unsigned int hash; 250 struct rif_cache *entry; 251 unsigned char *olddata; 252 unsigned long flags; 253 static const unsigned char mcast_func_addr[] 254 = {0xC0,0x00,0x00,0x04,0x00,0x00}; 255 256 spin_lock_irqsave(&rif_lock, flags); 257 258 /* 259 * Broadcasts are single route as stated in RFC 1042 260 */ 261 if( (!memcmp(&(trh->daddr[0]),&(dev->broadcast[0]),TR_ALEN)) || 262 (!memcmp(&(trh->daddr[0]),&(mcast_func_addr[0]), TR_ALEN)) ) 263 { 264 trh->rcf=htons((((sizeof(trh->rcf)) << 8) & TR_RCF_LEN_MASK) 265 | TR_RCF_FRAME2K | TR_RCF_LIMITED_BROADCAST); 266 trh->saddr[0]|=TR_RII; 267 } 268 else 269 { 270 hash = rif_hash(trh->daddr); 271 /* 272 * Walk the hash table and look for an entry 273 */ 274 for(entry=rif_table[hash];entry && memcmp(&(entry->addr[0]),&(trh->daddr[0]),TR_ALEN);entry=entry->next); 275 276 /* 277 * If we found an entry we can route the frame. 278 */ 279 if(entry) 280 { 281#if TR_SR_DEBUG 282printk("source routing for %02X:%02X:%02X:%02X:%02X:%02X\n",trh->daddr[0], 283 trh->daddr[1],trh->daddr[2],trh->daddr[3],trh->daddr[4],trh->daddr[5]); 284#endif 285 if(!entry->local_ring && (ntohs(entry->rcf) & TR_RCF_LEN_MASK) >> 8) 286 { 287 trh->rcf=entry->rcf; 288 memcpy(&trh->rseg[0],&entry->rseg[0],8*sizeof(unsigned short)); 289 trh->rcf^=htons(TR_RCF_DIR_BIT); 290 trh->rcf&=htons(0x1fff); /* Issam Chehab <ichehab@madge1.demon.co.uk> */ 291 292 trh->saddr[0]|=TR_RII; 293#if TR_SR_DEBUG 294 printk("entry found with rcf %04x\n", entry->rcf); 295 } 296 else 297 { 298 printk("entry found but without rcf length, local=%02x\n", entry->local_ring); 299#endif 300 } 301 entry->last_used=jiffies; 302 } 303 else 304 { 305 /* 306 * Without the information we simply have to shout 307 * on the wire. The replies should rapidly clean this 308 * situation up. 309 */ 310 trh->rcf=htons((((sizeof(trh->rcf)) << 8) & TR_RCF_LEN_MASK) 311 | TR_RCF_FRAME2K | TR_RCF_LIMITED_BROADCAST); 312 trh->saddr[0]|=TR_RII; 313#if TR_SR_DEBUG 314 printk("no entry in rif table found - broadcasting frame\n"); 315#endif 316 } 317 } 318 319 /* Compress the RIF here so we don't have to do it in the driver(s) */ 320 if (!(trh->saddr[0] & 0x80)) 321 slack = 18; 322 else 323 slack = 18 - ((ntohs(trh->rcf) & TR_RCF_LEN_MASK)>>8); 324 olddata = skb->data; 325 spin_unlock_irqrestore(&rif_lock, flags); 326 327 skb_pull(skb, slack); 328 memmove(skb->data, olddata, sizeof(struct trh_hdr) - slack); 329} 330 331/* 332 * We have learned some new RIF information for our source 333 * routing. 334 */ 335 336static void tr_add_rif_info(struct trh_hdr *trh, struct net_device *dev) 337{ 338 unsigned int hash, rii_p = 0; 339 unsigned long flags; 340 struct rif_cache *entry; 341 unsigned char saddr0; 342 343 spin_lock_irqsave(&rif_lock, flags); 344 saddr0 = trh->saddr[0]; 345 346 /* 347 * Firstly see if the entry exists 348 */ 349 350 if(trh->saddr[0] & TR_RII) 351 { 352 trh->saddr[0]&=0x7f; 353 if (((ntohs(trh->rcf) & TR_RCF_LEN_MASK) >> 8) > 2) 354 { 355 rii_p = 1; 356 } 357 } 358 359 hash = rif_hash(trh->saddr); 360 for(entry=rif_table[hash];entry && memcmp(&(entry->addr[0]),&(trh->saddr[0]),TR_ALEN);entry=entry->next); 361 362 if(entry==NULL) 363 { 364#if TR_SR_DEBUG 365printk("adding rif_entry: addr:%02X:%02X:%02X:%02X:%02X:%02X rcf:%04X\n", 366 trh->saddr[0],trh->saddr[1],trh->saddr[2], 367 trh->saddr[3],trh->saddr[4],trh->saddr[5], 368 ntohs(trh->rcf)); 369#endif 370 entry=kmalloc(sizeof(struct rif_cache),GFP_ATOMIC); 371 372 if(!entry) 373 { 374 printk(KERN_DEBUG "tr.c: Couldn't malloc rif cache entry !\n"); 375 spin_unlock_irqrestore(&rif_lock, flags); 376 return; 377 } 378 379 memcpy(&(entry->addr[0]),&(trh->saddr[0]),TR_ALEN); 380 entry->iface = dev->ifindex; 381 entry->next=rif_table[hash]; 382 entry->last_used=jiffies; 383 rif_table[hash]=entry; 384 385 if (rii_p) 386 { 387 entry->rcf = trh->rcf & htons((unsigned short)~TR_RCF_BROADCAST_MASK); 388 memcpy(&(entry->rseg[0]),&(trh->rseg[0]),8*sizeof(unsigned short)); 389 entry->local_ring = 0; 390 } 391 else 392 { 393 entry->local_ring = 1; 394 } 395 } 396 else /* Y. Tahara added */ 397 { 398 /* 399 * Update existing entries 400 */ 401 if (!entry->local_ring) 402 if (entry->rcf != (trh->rcf & htons((unsigned short)~TR_RCF_BROADCAST_MASK)) && 403 !(trh->rcf & htons(TR_RCF_BROADCAST_MASK))) 404 { 405#if TR_SR_DEBUG 406printk("updating rif_entry: addr:%02X:%02X:%02X:%02X:%02X:%02X rcf:%04X\n", 407 trh->saddr[0],trh->saddr[1],trh->saddr[2], 408 trh->saddr[3],trh->saddr[4],trh->saddr[5], 409 ntohs(trh->rcf)); 410#endif 411 entry->rcf = trh->rcf & htons((unsigned short)~TR_RCF_BROADCAST_MASK); 412 memcpy(&(entry->rseg[0]),&(trh->rseg[0]),8*sizeof(unsigned short)); 413 } 414 entry->last_used=jiffies; 415 } 416 trh->saddr[0]=saddr0; /* put the routing indicator back for tcpdump */ 417 spin_unlock_irqrestore(&rif_lock, flags); 418} 419 420/* 421 * Scan the cache with a timer and see what we need to throw out. 422 */ 423 424static void rif_check_expire(unsigned long dummy) 425{ 426 int i; 427 unsigned long flags, next_interval = jiffies + sysctl_tr_rif_timeout/2; 428 429 spin_lock_irqsave(&rif_lock, flags); 430 431 for(i =0; i < RIF_TABLE_SIZE; i++) { 432 struct rif_cache *entry, **pentry; 433 434 pentry = rif_table+i; 435 while((entry=*pentry) != NULL) { 436 unsigned long expires 437 = entry->last_used + sysctl_tr_rif_timeout; 438 439 if (time_before_eq(expires, jiffies)) { 440 *pentry = entry->next; 441 kfree(entry); 442 } else { 443 pentry = &entry->next; 444 445 if (time_before(expires, next_interval)) 446 next_interval = expires; 447 } 448 } 449 } 450 451 spin_unlock_irqrestore(&rif_lock, flags); 452 453 mod_timer(&rif_timer, next_interval); 454 455} 456 457/* 458 * Generate the /proc/net information for the token ring RIF 459 * routing. 460 */ 461 462#ifdef CONFIG_PROC_FS 463 464static struct rif_cache *rif_get_idx(loff_t pos) 465{ 466 int i; 467 struct rif_cache *entry; 468 loff_t off = 0; 469 470 for(i = 0; i < RIF_TABLE_SIZE; i++) 471 for(entry = rif_table[i]; entry; entry = entry->next) { 472 if (off == pos) 473 return entry; 474 ++off; 475 } 476 477 return NULL; 478} 479 480static void *rif_seq_start(struct seq_file *seq, loff_t *pos) 481{ 482 spin_lock_irq(&rif_lock); 483 484 return *pos ? rif_get_idx(*pos - 1) : SEQ_START_TOKEN; 485} 486 487static void *rif_seq_next(struct seq_file *seq, void *v, loff_t *pos) 488{ 489 int i; 490 struct rif_cache *ent = v; 491 492 ++*pos; 493 494 if (v == SEQ_START_TOKEN) { 495 i = -1; 496 goto scan; 497 } 498 499 if (ent->next) 500 return ent->next; 501 502 i = rif_hash(ent->addr); 503 scan: 504 while (++i < RIF_TABLE_SIZE) { 505 if ((ent = rif_table[i]) != NULL) 506 return ent; 507 } 508 return NULL; 509} 510 511static void rif_seq_stop(struct seq_file *seq, void *v) 512{ 513 spin_unlock_irq(&rif_lock); 514} 515 516static int rif_seq_show(struct seq_file *seq, void *v) 517{ 518 int j, rcf_len, segment, brdgnmb; 519 struct rif_cache *entry = v; 520 521 if (v == SEQ_START_TOKEN) 522 seq_puts(seq, 523 "if TR address TTL rcf routing segments\n"); 524 else { 525 struct net_device *dev = dev_get_by_index(entry->iface); 526 long ttl = (long) (entry->last_used + sysctl_tr_rif_timeout) 527 - (long) jiffies; 528 529 seq_printf(seq, "%s %02X:%02X:%02X:%02X:%02X:%02X %7li ", 530 dev?dev->name:"?", 531 entry->addr[0],entry->addr[1],entry->addr[2], 532 entry->addr[3],entry->addr[4],entry->addr[5], 533 ttl/HZ); 534 535 if (entry->local_ring) 536 seq_puts(seq, "local\n"); 537 else { 538 539 seq_printf(seq, "%04X", ntohs(entry->rcf)); 540 rcf_len = ((ntohs(entry->rcf) & TR_RCF_LEN_MASK)>>8)-2; 541 if (rcf_len) 542 rcf_len >>= 1; 543 for(j = 1; j < rcf_len; j++) { 544 if(j==1) { 545 segment=ntohs(entry->rseg[j-1])>>4; 546 seq_printf(seq," %03X",segment); 547 } 548 549 segment=ntohs(entry->rseg[j])>>4; 550 brdgnmb=ntohs(entry->rseg[j-1])&0x00f; 551 seq_printf(seq,"-%01X-%03X",brdgnmb,segment); 552 } 553 seq_putc(seq, '\n'); 554 } 555 } 556 return 0; 557} 558 559 560static struct seq_operations rif_seq_ops = { 561 .start = rif_seq_start, 562 .next = rif_seq_next, 563 .stop = rif_seq_stop, 564 .show = rif_seq_show, 565}; 566 567static int rif_seq_open(struct inode *inode, struct file *file) 568{ 569 return seq_open(file, &rif_seq_ops); 570} 571 572static const struct file_operations rif_seq_fops = { 573 .owner = THIS_MODULE, 574 .open = rif_seq_open, 575 .read = seq_read, 576 .llseek = seq_lseek, 577 .release = seq_release, 578}; 579 580#endif 581 582static void tr_setup(struct net_device *dev) 583{ 584 /* 585 * Configure and register 586 */ 587 588 dev->hard_header = tr_header; 589 dev->rebuild_header = tr_rebuild_header; 590 591 dev->type = ARPHRD_IEEE802_TR; 592 dev->hard_header_len = TR_HLEN; 593 dev->mtu = 2000; 594 dev->addr_len = TR_ALEN; 595 dev->tx_queue_len = 100; /* Long queues on tr */ 596 597 memset(dev->broadcast,0xFF, TR_ALEN); 598 599 /* New-style flags. */ 600 dev->flags = IFF_BROADCAST | IFF_MULTICAST ; 601} 602 603/** 604 * alloc_trdev - Register token ring device 605 * @sizeof_priv: Size of additional driver-private structure to be allocated 606 * for this token ring device 607 * 608 * Fill in the fields of the device structure with token ring-generic values. 609 * 610 * Constructs a new net device, complete with a private data area of 611 * size @sizeof_priv. A 32-byte (not bit) alignment is enforced for 612 * this private data area. 613 */ 614struct net_device *alloc_trdev(int sizeof_priv) 615{ 616 return alloc_netdev(sizeof_priv, "tr%d", tr_setup); 617} 618 619/* 620 * Called during bootup. We don't actually have to initialise 621 * too much for this. 622 */ 623 624static int __init rif_init(void) 625{ 626 init_timer(&rif_timer); 627 rif_timer.expires = sysctl_tr_rif_timeout; 628 rif_timer.data = 0L; 629 rif_timer.function = rif_check_expire; 630 add_timer(&rif_timer); 631 632 proc_net_fops_create("tr_rif", S_IRUGO, &rif_seq_fops); 633 return 0; 634} 635 636module_init(rif_init); 637 638EXPORT_SYMBOL(tr_type_trans); 639EXPORT_SYMBOL(alloc_trdev); 640