ipv6: check raw payload size correctly in ioctl
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ieee802154 / 6lowpan.c
1 /*
2 * Copyright 2011, Siemens AG
3 * written by Alexander Smirnov <alex.bluesman.smirnov@gmail.com>
4 */
5
6 /*
7 * Based on patches from Jon Smirl <jonsmirl@gmail.com>
8 * Copyright (c) 2011 Jon Smirl <jonsmirl@gmail.com>
9 *
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License version 2
12 * as published by the Free Software Foundation.
13 *
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
18 *
19 * You should have received a copy of the GNU General Public License along
20 * with this program; if not, write to the Free Software Foundation, Inc.,
21 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
22 */
23
24 /* Jon's code is based on 6lowpan implementation for Contiki which is:
25 * Copyright (c) 2008, Swedish Institute of Computer Science.
26 * All rights reserved.
27 *
28 * Redistribution and use in source and binary forms, with or without
29 * modification, are permitted provided that the following conditions
30 * are met:
31 * 1. Redistributions of source code must retain the above copyright
32 * notice, this list of conditions and the following disclaimer.
33 * 2. Redistributions in binary form must reproduce the above copyright
34 * notice, this list of conditions and the following disclaimer in the
35 * documentation and/or other materials provided with the distribution.
36 * 3. Neither the name of the Institute nor the names of its contributors
37 * may be used to endorse or promote products derived from this software
38 * without specific prior written permission.
39 *
40 * THIS SOFTWARE IS PROVIDED BY THE INSTITUTE AND CONTRIBUTORS ``AS IS'' AND
41 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
42 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
43 * ARE DISCLAIMED. IN NO EVENT SHALL THE INSTITUTE OR CONTRIBUTORS BE LIABLE
44 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
45 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
46 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
47 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
48 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
49 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
50 * SUCH DAMAGE.
51 */
52
53 #include <linux/bitops.h>
54 #include <linux/if_arp.h>
55 #include <linux/module.h>
56 #include <linux/moduleparam.h>
57 #include <linux/netdevice.h>
58 #include <net/af_ieee802154.h>
59 #include <net/ieee802154.h>
60 #include <net/ieee802154_netdev.h>
61 #include <net/ipv6.h>
62
63 #include "6lowpan.h"
64
65 /* TTL uncompression values */
66 static const u8 lowpan_ttl_values[] = {0, 1, 64, 255};
67
68 static LIST_HEAD(lowpan_devices);
69
70 /*
71 * Uncompression of linklocal:
72 * 0 -> 16 bytes from packet
73 * 1 -> 2 bytes from prefix - bunch of zeroes and 8 from packet
74 * 2 -> 2 bytes from prefix - zeroes + 2 from packet
75 * 3 -> 2 bytes from prefix - infer 8 bytes from lladdr
76 *
77 * NOTE: => the uncompress function does change 0xf to 0x10
78 * NOTE: 0x00 => no-autoconfig => unspecified
79 */
80 static const u8 lowpan_unc_llconf[] = {0x0f, 0x28, 0x22, 0x20};
81
82 /*
83 * Uncompression of ctx-based:
84 * 0 -> 0 bits from packet [unspecified / reserved]
85 * 1 -> 8 bytes from prefix - bunch of zeroes and 8 from packet
86 * 2 -> 8 bytes from prefix - zeroes + 2 from packet
87 * 3 -> 8 bytes from prefix - infer 8 bytes from lladdr
88 */
89 static const u8 lowpan_unc_ctxconf[] = {0x00, 0x88, 0x82, 0x80};
90
91 /*
92 * Uncompression of ctx-base
93 * 0 -> 0 bits from packet
94 * 1 -> 2 bytes from prefix - bunch of zeroes 5 from packet
95 * 2 -> 2 bytes from prefix - zeroes + 3 from packet
96 * 3 -> 2 bytes from prefix - infer 1 bytes from lladdr
97 */
98 static const u8 lowpan_unc_mxconf[] = {0x0f, 0x25, 0x23, 0x21};
99
100 /* Link local prefix */
101 static const u8 lowpan_llprefix[] = {0xfe, 0x80};
102
103 /* private device info */
104 struct lowpan_dev_info {
105 struct net_device *real_dev; /* real WPAN device ptr */
106 struct mutex dev_list_mtx; /* mutex for list ops */
107 unsigned short fragment_tag;
108 };
109
110 struct lowpan_dev_record {
111 struct net_device *ldev;
112 struct list_head list;
113 };
114
115 struct lowpan_fragment {
116 struct sk_buff *skb; /* skb to be assembled */
117 u16 length; /* length to be assemled */
118 u32 bytes_rcv; /* bytes received */
119 u16 tag; /* current fragment tag */
120 struct timer_list timer; /* assembling timer */
121 struct list_head list; /* fragments list */
122 };
123
124 static LIST_HEAD(lowpan_fragments);
125 static DEFINE_SPINLOCK(flist_lock);
126
127 static inline struct
128 lowpan_dev_info *lowpan_dev_info(const struct net_device *dev)
129 {
130 return netdev_priv(dev);
131 }
132
133 static inline void lowpan_address_flip(u8 *src, u8 *dest)
134 {
135 int i;
136 for (i = 0; i < IEEE802154_ADDR_LEN; i++)
137 (dest)[IEEE802154_ADDR_LEN - i - 1] = (src)[i];
138 }
139
140 /* list of all 6lowpan devices, uses for package delivering */
141 /* print data in line */
142 static inline void lowpan_raw_dump_inline(const char *caller, char *msg,
143 unsigned char *buf, int len)
144 {
145 #ifdef DEBUG
146 if (msg)
147 pr_debug("(%s) %s: ", caller, msg);
148 print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_NONE,
149 16, 1, buf, len, false);
150 #endif /* DEBUG */
151 }
152
153 /*
154 * print data in a table format:
155 *
156 * addr: xx xx xx xx xx xx
157 * addr: xx xx xx xx xx xx
158 * ...
159 */
160 static inline void lowpan_raw_dump_table(const char *caller, char *msg,
161 unsigned char *buf, int len)
162 {
163 #ifdef DEBUG
164 if (msg)
165 pr_debug("(%s) %s:\n", caller, msg);
166 print_hex_dump(KERN_DEBUG, "\t", DUMP_PREFIX_OFFSET,
167 16, 1, buf, len, false);
168 #endif /* DEBUG */
169 }
170
171 static u8
172 lowpan_compress_addr_64(u8 **hc06_ptr, u8 shift, const struct in6_addr *ipaddr,
173 const unsigned char *lladdr)
174 {
175 u8 val = 0;
176
177 if (is_addr_mac_addr_based(ipaddr, lladdr))
178 val = 3; /* 0-bits */
179 else if (lowpan_is_iid_16_bit_compressable(ipaddr)) {
180 /* compress IID to 16 bits xxxx::XXXX */
181 memcpy(*hc06_ptr, &ipaddr->s6_addr16[7], 2);
182 *hc06_ptr += 2;
183 val = 2; /* 16-bits */
184 } else {
185 /* do not compress IID => xxxx::IID */
186 memcpy(*hc06_ptr, &ipaddr->s6_addr16[4], 8);
187 *hc06_ptr += 8;
188 val = 1; /* 64-bits */
189 }
190
191 return rol8(val, shift);
192 }
193
194 static void
195 lowpan_uip_ds6_set_addr_iid(struct in6_addr *ipaddr, unsigned char *lladdr)
196 {
197 memcpy(&ipaddr->s6_addr[8], lladdr, IEEE802154_ADDR_LEN);
198 /* second bit-flip (Universe/Local) is done according RFC2464 */
199 ipaddr->s6_addr[8] ^= 0x02;
200 }
201
202 /*
203 * Uncompress addresses based on a prefix and a postfix with zeroes in
204 * between. If the postfix is zero in length it will use the link address
205 * to configure the IP address (autoconf style).
206 * pref_post_count takes a byte where the first nibble specify prefix count
207 * and the second postfix count (NOTE: 15/0xf => 16 bytes copy).
208 */
209 static int
210 lowpan_uncompress_addr(struct sk_buff *skb, struct in6_addr *ipaddr,
211 u8 const *prefix, u8 pref_post_count, unsigned char *lladdr)
212 {
213 u8 prefcount = pref_post_count >> 4;
214 u8 postcount = pref_post_count & 0x0f;
215
216 /* full nibble 15 => 16 */
217 prefcount = (prefcount == 15 ? 16 : prefcount);
218 postcount = (postcount == 15 ? 16 : postcount);
219
220 if (lladdr)
221 lowpan_raw_dump_inline(__func__, "linklocal address",
222 lladdr, IEEE802154_ADDR_LEN);
223 if (prefcount > 0)
224 memcpy(ipaddr, prefix, prefcount);
225
226 if (prefcount + postcount < 16)
227 memset(&ipaddr->s6_addr[prefcount], 0,
228 16 - (prefcount + postcount));
229
230 if (postcount > 0) {
231 memcpy(&ipaddr->s6_addr[16 - postcount], skb->data, postcount);
232 skb_pull(skb, postcount);
233 } else if (prefcount > 0) {
234 if (lladdr == NULL)
235 return -EINVAL;
236
237 /* no IID based configuration if no prefix and no data */
238 lowpan_uip_ds6_set_addr_iid(ipaddr, lladdr);
239 }
240
241 pr_debug("uncompressing %d + %d => ", prefcount, postcount);
242 lowpan_raw_dump_inline(NULL, NULL, ipaddr->s6_addr, 16);
243
244 return 0;
245 }
246
247 static void
248 lowpan_compress_udp_header(u8 **hc06_ptr, struct sk_buff *skb)
249 {
250 struct udphdr *uh = udp_hdr(skb);
251
252 if (((uh->source & LOWPAN_NHC_UDP_4BIT_MASK) ==
253 LOWPAN_NHC_UDP_4BIT_PORT) &&
254 ((uh->dest & LOWPAN_NHC_UDP_4BIT_MASK) ==
255 LOWPAN_NHC_UDP_4BIT_PORT)) {
256 pr_debug("UDP header: both ports compression to 4 bits\n");
257 **hc06_ptr = LOWPAN_NHC_UDP_CS_P_11;
258 **(hc06_ptr + 1) = /* subtraction is faster */
259 (u8)((uh->dest - LOWPAN_NHC_UDP_4BIT_PORT) +
260 ((uh->source & LOWPAN_NHC_UDP_4BIT_PORT) << 4));
261 *hc06_ptr += 2;
262 } else if ((uh->dest & LOWPAN_NHC_UDP_8BIT_MASK) ==
263 LOWPAN_NHC_UDP_8BIT_PORT) {
264 pr_debug("UDP header: remove 8 bits of dest\n");
265 **hc06_ptr = LOWPAN_NHC_UDP_CS_P_01;
266 memcpy(*hc06_ptr + 1, &uh->source, 2);
267 **(hc06_ptr + 3) = (u8)(uh->dest - LOWPAN_NHC_UDP_8BIT_PORT);
268 *hc06_ptr += 4;
269 } else if ((uh->source & LOWPAN_NHC_UDP_8BIT_MASK) ==
270 LOWPAN_NHC_UDP_8BIT_PORT) {
271 pr_debug("UDP header: remove 8 bits of source\n");
272 **hc06_ptr = LOWPAN_NHC_UDP_CS_P_10;
273 memcpy(*hc06_ptr + 1, &uh->dest, 2);
274 **(hc06_ptr + 3) = (u8)(uh->source - LOWPAN_NHC_UDP_8BIT_PORT);
275 *hc06_ptr += 4;
276 } else {
277 pr_debug("UDP header: can't compress\n");
278 **hc06_ptr = LOWPAN_NHC_UDP_CS_P_00;
279 memcpy(*hc06_ptr + 1, &uh->source, 2);
280 memcpy(*hc06_ptr + 3, &uh->dest, 2);
281 *hc06_ptr += 5;
282 }
283
284 /* checksum is always inline */
285 memcpy(*hc06_ptr, &uh->check, 2);
286 *hc06_ptr += 2;
287
288 /* skip the UDP header */
289 skb_pull(skb, sizeof(struct udphdr));
290 }
291
292 static inline int lowpan_fetch_skb_u8(struct sk_buff *skb, u8 *val)
293 {
294 if (unlikely(!pskb_may_pull(skb, 1)))
295 return -EINVAL;
296
297 *val = skb->data[0];
298 skb_pull(skb, 1);
299
300 return 0;
301 }
302
303 static inline int lowpan_fetch_skb_u16(struct sk_buff *skb, u16 *val)
304 {
305 if (unlikely(!pskb_may_pull(skb, 2)))
306 return -EINVAL;
307
308 *val = (skb->data[0] << 8) | skb->data[1];
309 skb_pull(skb, 2);
310
311 return 0;
312 }
313
314 static int
315 lowpan_uncompress_udp_header(struct sk_buff *skb, struct udphdr *uh)
316 {
317 u8 tmp;
318
319 if (!uh)
320 goto err;
321
322 if (lowpan_fetch_skb_u8(skb, &tmp))
323 goto err;
324
325 if ((tmp & LOWPAN_NHC_UDP_MASK) == LOWPAN_NHC_UDP_ID) {
326 pr_debug("UDP header uncompression\n");
327 switch (tmp & LOWPAN_NHC_UDP_CS_P_11) {
328 case LOWPAN_NHC_UDP_CS_P_00:
329 memcpy(&uh->source, &skb->data[0], 2);
330 memcpy(&uh->dest, &skb->data[2], 2);
331 skb_pull(skb, 4);
332 break;
333 case LOWPAN_NHC_UDP_CS_P_01:
334 memcpy(&uh->source, &skb->data[0], 2);
335 uh->dest =
336 skb->data[2] + LOWPAN_NHC_UDP_8BIT_PORT;
337 skb_pull(skb, 3);
338 break;
339 case LOWPAN_NHC_UDP_CS_P_10:
340 uh->source = skb->data[0] + LOWPAN_NHC_UDP_8BIT_PORT;
341 memcpy(&uh->dest, &skb->data[1], 2);
342 skb_pull(skb, 3);
343 break;
344 case LOWPAN_NHC_UDP_CS_P_11:
345 uh->source =
346 LOWPAN_NHC_UDP_4BIT_PORT + (skb->data[0] >> 4);
347 uh->dest =
348 LOWPAN_NHC_UDP_4BIT_PORT + (skb->data[0] & 0x0f);
349 skb_pull(skb, 1);
350 break;
351 default:
352 pr_debug("ERROR: unknown UDP format\n");
353 goto err;
354 break;
355 }
356
357 pr_debug("uncompressed UDP ports: src = %d, dst = %d\n",
358 uh->source, uh->dest);
359
360 /* copy checksum */
361 memcpy(&uh->check, &skb->data[0], 2);
362 skb_pull(skb, 2);
363
364 /*
365 * UDP lenght needs to be infered from the lower layers
366 * here, we obtain the hint from the remaining size of the
367 * frame
368 */
369 uh->len = htons(skb->len + sizeof(struct udphdr));
370 pr_debug("uncompressed UDP length: src = %d", uh->len);
371 } else {
372 pr_debug("ERROR: unsupported NH format\n");
373 goto err;
374 }
375
376 return 0;
377 err:
378 return -EINVAL;
379 }
380
381 static int lowpan_header_create(struct sk_buff *skb,
382 struct net_device *dev,
383 unsigned short type, const void *_daddr,
384 const void *_saddr, unsigned int len)
385 {
386 u8 tmp, iphc0, iphc1, *hc06_ptr;
387 struct ipv6hdr *hdr;
388 const u8 *saddr = _saddr;
389 const u8 *daddr = _daddr;
390 u8 head[100];
391 struct ieee802154_addr sa, da;
392
393 /* TODO:
394 * if this package isn't ipv6 one, where should it be routed?
395 */
396 if (type != ETH_P_IPV6)
397 return 0;
398
399 hdr = ipv6_hdr(skb);
400 hc06_ptr = head + 2;
401
402 pr_debug("IPv6 header dump:\n\tversion = %d\n\tlength = %d\n"
403 "\tnexthdr = 0x%02x\n\thop_lim = %d\n", hdr->version,
404 ntohs(hdr->payload_len), hdr->nexthdr, hdr->hop_limit);
405
406 lowpan_raw_dump_table(__func__, "raw skb network header dump",
407 skb_network_header(skb), sizeof(struct ipv6hdr));
408
409 if (!saddr)
410 saddr = dev->dev_addr;
411
412 lowpan_raw_dump_inline(__func__, "saddr", (unsigned char *)saddr, 8);
413
414 /*
415 * As we copy some bit-length fields, in the IPHC encoding bytes,
416 * we sometimes use |=
417 * If the field is 0, and the current bit value in memory is 1,
418 * this does not work. We therefore reset the IPHC encoding here
419 */
420 iphc0 = LOWPAN_DISPATCH_IPHC;
421 iphc1 = 0;
422
423 /* TODO: context lookup */
424
425 lowpan_raw_dump_inline(__func__, "daddr", (unsigned char *)daddr, 8);
426
427 /*
428 * Traffic class, flow label
429 * If flow label is 0, compress it. If traffic class is 0, compress it
430 * We have to process both in the same time as the offset of traffic
431 * class depends on the presence of version and flow label
432 */
433
434 /* hc06 format of TC is ECN | DSCP , original one is DSCP | ECN */
435 tmp = (hdr->priority << 4) | (hdr->flow_lbl[0] >> 4);
436 tmp = ((tmp & 0x03) << 6) | (tmp >> 2);
437
438 if (((hdr->flow_lbl[0] & 0x0F) == 0) &&
439 (hdr->flow_lbl[1] == 0) && (hdr->flow_lbl[2] == 0)) {
440 /* flow label can be compressed */
441 iphc0 |= LOWPAN_IPHC_FL_C;
442 if ((hdr->priority == 0) &&
443 ((hdr->flow_lbl[0] & 0xF0) == 0)) {
444 /* compress (elide) all */
445 iphc0 |= LOWPAN_IPHC_TC_C;
446 } else {
447 /* compress only the flow label */
448 *hc06_ptr = tmp;
449 hc06_ptr += 1;
450 }
451 } else {
452 /* Flow label cannot be compressed */
453 if ((hdr->priority == 0) &&
454 ((hdr->flow_lbl[0] & 0xF0) == 0)) {
455 /* compress only traffic class */
456 iphc0 |= LOWPAN_IPHC_TC_C;
457 *hc06_ptr = (tmp & 0xc0) | (hdr->flow_lbl[0] & 0x0F);
458 memcpy(hc06_ptr + 1, &hdr->flow_lbl[1], 2);
459 hc06_ptr += 3;
460 } else {
461 /* compress nothing */
462 memcpy(hc06_ptr, hdr, 4);
463 /* replace the top byte with new ECN | DSCP format */
464 *hc06_ptr = tmp;
465 hc06_ptr += 4;
466 }
467 }
468
469 /* NOTE: payload length is always compressed */
470
471 /* Next Header is compress if UDP */
472 if (hdr->nexthdr == UIP_PROTO_UDP)
473 iphc0 |= LOWPAN_IPHC_NH_C;
474
475 if ((iphc0 & LOWPAN_IPHC_NH_C) == 0) {
476 *hc06_ptr = hdr->nexthdr;
477 hc06_ptr += 1;
478 }
479
480 /*
481 * Hop limit
482 * if 1: compress, encoding is 01
483 * if 64: compress, encoding is 10
484 * if 255: compress, encoding is 11
485 * else do not compress
486 */
487 switch (hdr->hop_limit) {
488 case 1:
489 iphc0 |= LOWPAN_IPHC_TTL_1;
490 break;
491 case 64:
492 iphc0 |= LOWPAN_IPHC_TTL_64;
493 break;
494 case 255:
495 iphc0 |= LOWPAN_IPHC_TTL_255;
496 break;
497 default:
498 *hc06_ptr = hdr->hop_limit;
499 hc06_ptr += 1;
500 break;
501 }
502
503 /* source address compression */
504 if (is_addr_unspecified(&hdr->saddr)) {
505 pr_debug("source address is unspecified, setting SAC\n");
506 iphc1 |= LOWPAN_IPHC_SAC;
507 /* TODO: context lookup */
508 } else if (is_addr_link_local(&hdr->saddr)) {
509 pr_debug("source address is link-local\n");
510 iphc1 |= lowpan_compress_addr_64(&hc06_ptr,
511 LOWPAN_IPHC_SAM_BIT, &hdr->saddr, saddr);
512 } else {
513 pr_debug("send the full source address\n");
514 memcpy(hc06_ptr, &hdr->saddr.s6_addr16[0], 16);
515 hc06_ptr += 16;
516 }
517
518 /* destination address compression */
519 if (is_addr_mcast(&hdr->daddr)) {
520 pr_debug("destination address is multicast: ");
521 iphc1 |= LOWPAN_IPHC_M;
522 if (lowpan_is_mcast_addr_compressable8(&hdr->daddr)) {
523 pr_debug("compressed to 1 octet\n");
524 iphc1 |= LOWPAN_IPHC_DAM_11;
525 /* use last byte */
526 *hc06_ptr = hdr->daddr.s6_addr[15];
527 hc06_ptr += 1;
528 } else if (lowpan_is_mcast_addr_compressable32(&hdr->daddr)) {
529 pr_debug("compressed to 4 octets\n");
530 iphc1 |= LOWPAN_IPHC_DAM_10;
531 /* second byte + the last three */
532 *hc06_ptr = hdr->daddr.s6_addr[1];
533 memcpy(hc06_ptr + 1, &hdr->daddr.s6_addr[13], 3);
534 hc06_ptr += 4;
535 } else if (lowpan_is_mcast_addr_compressable48(&hdr->daddr)) {
536 pr_debug("compressed to 6 octets\n");
537 iphc1 |= LOWPAN_IPHC_DAM_01;
538 /* second byte + the last five */
539 *hc06_ptr = hdr->daddr.s6_addr[1];
540 memcpy(hc06_ptr + 1, &hdr->daddr.s6_addr[11], 5);
541 hc06_ptr += 6;
542 } else {
543 pr_debug("using full address\n");
544 iphc1 |= LOWPAN_IPHC_DAM_00;
545 memcpy(hc06_ptr, &hdr->daddr.s6_addr[0], 16);
546 hc06_ptr += 16;
547 }
548 } else {
549 /* TODO: context lookup */
550 if (is_addr_link_local(&hdr->daddr)) {
551 pr_debug("dest address is unicast and link-local\n");
552 iphc1 |= lowpan_compress_addr_64(&hc06_ptr,
553 LOWPAN_IPHC_DAM_BIT, &hdr->daddr, daddr);
554 } else {
555 pr_debug("dest address is unicast: using full one\n");
556 memcpy(hc06_ptr, &hdr->daddr.s6_addr16[0], 16);
557 hc06_ptr += 16;
558 }
559 }
560
561 /* UDP header compression */
562 if (hdr->nexthdr == UIP_PROTO_UDP)
563 lowpan_compress_udp_header(&hc06_ptr, skb);
564
565 head[0] = iphc0;
566 head[1] = iphc1;
567
568 skb_pull(skb, sizeof(struct ipv6hdr));
569 memcpy(skb_push(skb, hc06_ptr - head), head, hc06_ptr - head);
570
571 lowpan_raw_dump_table(__func__, "raw skb data dump", skb->data,
572 skb->len);
573
574 /*
575 * NOTE1: I'm still unsure about the fact that compression and WPAN
576 * header are created here and not later in the xmit. So wait for
577 * an opinion of net maintainers.
578 */
579 /*
580 * NOTE2: to be absolutely correct, we must derive PANid information
581 * from MAC subif of the 'dev' and 'real_dev' network devices, but
582 * this isn't implemented in mainline yet, so currently we assign 0xff
583 */
584 {
585 mac_cb(skb)->flags = IEEE802154_FC_TYPE_DATA;
586 mac_cb(skb)->seq = ieee802154_mlme_ops(dev)->get_dsn(dev);
587
588 /* prepare wpan address data */
589 sa.addr_type = IEEE802154_ADDR_LONG;
590 sa.pan_id = ieee802154_mlme_ops(dev)->get_pan_id(dev);
591
592 memcpy(&(sa.hwaddr), saddr, 8);
593 /* intra-PAN communications */
594 da.pan_id = ieee802154_mlme_ops(dev)->get_pan_id(dev);
595
596 /*
597 * if the destination address is the broadcast address, use the
598 * corresponding short address
599 */
600 if (lowpan_is_addr_broadcast(daddr)) {
601 da.addr_type = IEEE802154_ADDR_SHORT;
602 da.short_addr = IEEE802154_ADDR_BROADCAST;
603 } else {
604 da.addr_type = IEEE802154_ADDR_LONG;
605 memcpy(&(da.hwaddr), daddr, IEEE802154_ADDR_LEN);
606
607 /* request acknowledgment */
608 mac_cb(skb)->flags |= MAC_CB_FLAG_ACKREQ;
609 }
610
611 return dev_hard_header(skb, lowpan_dev_info(dev)->real_dev,
612 type, (void *)&da, (void *)&sa, skb->len);
613 }
614 }
615
616 static int lowpan_give_skb_to_devices(struct sk_buff *skb)
617 {
618 struct lowpan_dev_record *entry;
619 struct sk_buff *skb_cp;
620 int stat = NET_RX_SUCCESS;
621
622 rcu_read_lock();
623 list_for_each_entry_rcu(entry, &lowpan_devices, list)
624 if (lowpan_dev_info(entry->ldev)->real_dev == skb->dev) {
625 skb_cp = skb_copy(skb, GFP_ATOMIC);
626 if (!skb_cp) {
627 stat = -ENOMEM;
628 break;
629 }
630
631 skb_cp->dev = entry->ldev;
632 stat = netif_rx(skb_cp);
633 }
634 rcu_read_unlock();
635
636 return stat;
637 }
638
639 static int lowpan_skb_deliver(struct sk_buff *skb, struct ipv6hdr *hdr)
640 {
641 struct sk_buff *new;
642 int stat = NET_RX_SUCCESS;
643
644 new = skb_copy_expand(skb, sizeof(struct ipv6hdr), skb_tailroom(skb),
645 GFP_ATOMIC);
646 kfree_skb(skb);
647
648 if (!new)
649 return -ENOMEM;
650
651 skb_push(new, sizeof(struct ipv6hdr));
652 skb_reset_network_header(new);
653 skb_copy_to_linear_data(new, hdr, sizeof(struct ipv6hdr));
654
655 new->protocol = htons(ETH_P_IPV6);
656 new->pkt_type = PACKET_HOST;
657
658 stat = lowpan_give_skb_to_devices(new);
659
660 kfree_skb(new);
661
662 return stat;
663 }
664
665 static void lowpan_fragment_timer_expired(unsigned long entry_addr)
666 {
667 struct lowpan_fragment *entry = (struct lowpan_fragment *)entry_addr;
668
669 pr_debug("timer expired for frame with tag %d\n", entry->tag);
670
671 list_del(&entry->list);
672 dev_kfree_skb(entry->skb);
673 kfree(entry);
674 }
675
676 static struct lowpan_fragment *
677 lowpan_alloc_new_frame(struct sk_buff *skb, u16 len, u16 tag)
678 {
679 struct lowpan_fragment *frame;
680
681 frame = kzalloc(sizeof(struct lowpan_fragment),
682 GFP_ATOMIC);
683 if (!frame)
684 goto frame_err;
685
686 INIT_LIST_HEAD(&frame->list);
687
688 frame->length = len;
689 frame->tag = tag;
690
691 /* allocate buffer for frame assembling */
692 frame->skb = netdev_alloc_skb_ip_align(skb->dev, frame->length +
693 sizeof(struct ipv6hdr));
694
695 if (!frame->skb)
696 goto skb_err;
697
698 frame->skb->priority = skb->priority;
699 frame->skb->dev = skb->dev;
700
701 /* reserve headroom for uncompressed ipv6 header */
702 skb_reserve(frame->skb, sizeof(struct ipv6hdr));
703 skb_put(frame->skb, frame->length);
704
705 init_timer(&frame->timer);
706 /* time out is the same as for ipv6 - 60 sec */
707 frame->timer.expires = jiffies + LOWPAN_FRAG_TIMEOUT;
708 frame->timer.data = (unsigned long)frame;
709 frame->timer.function = lowpan_fragment_timer_expired;
710
711 add_timer(&frame->timer);
712
713 list_add_tail(&frame->list, &lowpan_fragments);
714
715 return frame;
716
717 skb_err:
718 kfree(frame);
719 frame_err:
720 return NULL;
721 }
722
723 static int
724 lowpan_process_data(struct sk_buff *skb)
725 {
726 struct ipv6hdr hdr;
727 u8 tmp, iphc0, iphc1, num_context = 0;
728 u8 *_saddr, *_daddr;
729 int err;
730
731 lowpan_raw_dump_table(__func__, "raw skb data dump", skb->data,
732 skb->len);
733 /* at least two bytes will be used for the encoding */
734 if (skb->len < 2)
735 goto drop;
736
737 if (lowpan_fetch_skb_u8(skb, &iphc0))
738 goto drop;
739
740 /* fragments assembling */
741 switch (iphc0 & LOWPAN_DISPATCH_MASK) {
742 case LOWPAN_DISPATCH_FRAG1:
743 case LOWPAN_DISPATCH_FRAGN:
744 {
745 struct lowpan_fragment *frame;
746 /* slen stores the rightmost 8 bits of the 11 bits length */
747 u8 slen, offset = 0;
748 u16 len, tag;
749 bool found = false;
750
751 if (lowpan_fetch_skb_u8(skb, &slen) || /* frame length */
752 lowpan_fetch_skb_u16(skb, &tag)) /* fragment tag */
753 goto drop;
754
755 /* adds the 3 MSB to the 8 LSB to retrieve the 11 bits length */
756 len = ((iphc0 & 7) << 8) | slen;
757
758 if ((iphc0 & LOWPAN_DISPATCH_MASK) == LOWPAN_DISPATCH_FRAG1) {
759 pr_debug("%s received a FRAG1 packet (tag: %d, "
760 "size of the entire IP packet: %d)",
761 __func__, tag, len);
762 } else { /* FRAGN */
763 if (lowpan_fetch_skb_u8(skb, &offset))
764 goto unlock_and_drop;
765 pr_debug("%s received a FRAGN packet (tag: %d, "
766 "size of the entire IP packet: %d, "
767 "offset: %d)", __func__, tag, len, offset * 8);
768 }
769
770 /*
771 * check if frame assembling with the same tag is
772 * already in progress
773 */
774 spin_lock_bh(&flist_lock);
775
776 list_for_each_entry(frame, &lowpan_fragments, list)
777 if (frame->tag == tag) {
778 found = true;
779 break;
780 }
781
782 /* alloc new frame structure */
783 if (!found) {
784 pr_debug("%s first fragment received for tag %d, "
785 "begin packet reassembly", __func__, tag);
786 frame = lowpan_alloc_new_frame(skb, len, tag);
787 if (!frame)
788 goto unlock_and_drop;
789 }
790
791 /* if payload fits buffer, copy it */
792 if (likely((offset * 8 + skb->len) <= frame->length))
793 skb_copy_to_linear_data_offset(frame->skb, offset * 8,
794 skb->data, skb->len);
795 else
796 goto unlock_and_drop;
797
798 frame->bytes_rcv += skb->len;
799
800 /* frame assembling complete */
801 if ((frame->bytes_rcv == frame->length) &&
802 frame->timer.expires > jiffies) {
803 /* if timer haven't expired - first of all delete it */
804 del_timer_sync(&frame->timer);
805 list_del(&frame->list);
806 spin_unlock_bh(&flist_lock);
807
808 pr_debug("%s successfully reassembled fragment "
809 "(tag %d)", __func__, tag);
810
811 dev_kfree_skb(skb);
812 skb = frame->skb;
813 kfree(frame);
814
815 if (lowpan_fetch_skb_u8(skb, &iphc0))
816 goto drop;
817
818 break;
819 }
820 spin_unlock_bh(&flist_lock);
821
822 return kfree_skb(skb), 0;
823 }
824 default:
825 break;
826 }
827
828 if (lowpan_fetch_skb_u8(skb, &iphc1))
829 goto drop;
830
831 _saddr = mac_cb(skb)->sa.hwaddr;
832 _daddr = mac_cb(skb)->da.hwaddr;
833
834 pr_debug("iphc0 = %02x, iphc1 = %02x\n", iphc0, iphc1);
835
836 /* another if the CID flag is set */
837 if (iphc1 & LOWPAN_IPHC_CID) {
838 pr_debug("CID flag is set, increase header with one\n");
839 if (lowpan_fetch_skb_u8(skb, &num_context))
840 goto drop;
841 }
842
843 hdr.version = 6;
844
845 /* Traffic Class and Flow Label */
846 switch ((iphc0 & LOWPAN_IPHC_TF) >> 3) {
847 /*
848 * Traffic Class and FLow Label carried in-line
849 * ECN + DSCP + 4-bit Pad + Flow Label (4 bytes)
850 */
851 case 0: /* 00b */
852 if (lowpan_fetch_skb_u8(skb, &tmp))
853 goto drop;
854
855 memcpy(&hdr.flow_lbl, &skb->data[0], 3);
856 skb_pull(skb, 3);
857 hdr.priority = ((tmp >> 2) & 0x0f);
858 hdr.flow_lbl[0] = ((tmp >> 2) & 0x30) | (tmp << 6) |
859 (hdr.flow_lbl[0] & 0x0f);
860 break;
861 /*
862 * Traffic class carried in-line
863 * ECN + DSCP (1 byte), Flow Label is elided
864 */
865 case 2: /* 10b */
866 if (lowpan_fetch_skb_u8(skb, &tmp))
867 goto drop;
868
869 hdr.priority = ((tmp >> 2) & 0x0f);
870 hdr.flow_lbl[0] = ((tmp << 6) & 0xC0) | ((tmp >> 2) & 0x30);
871 hdr.flow_lbl[1] = 0;
872 hdr.flow_lbl[2] = 0;
873 break;
874 /*
875 * Flow Label carried in-line
876 * ECN + 2-bit Pad + Flow Label (3 bytes), DSCP is elided
877 */
878 case 1: /* 01b */
879 if (lowpan_fetch_skb_u8(skb, &tmp))
880 goto drop;
881
882 hdr.flow_lbl[0] = (skb->data[0] & 0x0F) | ((tmp >> 2) & 0x30);
883 memcpy(&hdr.flow_lbl[1], &skb->data[0], 2);
884 skb_pull(skb, 2);
885 break;
886 /* Traffic Class and Flow Label are elided */
887 case 3: /* 11b */
888 hdr.priority = 0;
889 hdr.flow_lbl[0] = 0;
890 hdr.flow_lbl[1] = 0;
891 hdr.flow_lbl[2] = 0;
892 break;
893 default:
894 break;
895 }
896
897 /* Next Header */
898 if ((iphc0 & LOWPAN_IPHC_NH_C) == 0) {
899 /* Next header is carried inline */
900 if (lowpan_fetch_skb_u8(skb, &(hdr.nexthdr)))
901 goto drop;
902
903 pr_debug("NH flag is set, next header carried inline: %02x\n",
904 hdr.nexthdr);
905 }
906
907 /* Hop Limit */
908 if ((iphc0 & 0x03) != LOWPAN_IPHC_TTL_I)
909 hdr.hop_limit = lowpan_ttl_values[iphc0 & 0x03];
910 else {
911 if (lowpan_fetch_skb_u8(skb, &(hdr.hop_limit)))
912 goto drop;
913 }
914
915 /* Extract SAM to the tmp variable */
916 tmp = ((iphc1 & LOWPAN_IPHC_SAM) >> LOWPAN_IPHC_SAM_BIT) & 0x03;
917
918 /* Source address uncompression */
919 pr_debug("source address stateless compression\n");
920 err = lowpan_uncompress_addr(skb, &hdr.saddr, lowpan_llprefix,
921 lowpan_unc_llconf[tmp], skb->data);
922 if (err)
923 goto drop;
924
925 /* Extract DAM to the tmp variable */
926 tmp = ((iphc1 & LOWPAN_IPHC_DAM_11) >> LOWPAN_IPHC_DAM_BIT) & 0x03;
927
928 /* check for Multicast Compression */
929 if (iphc1 & LOWPAN_IPHC_M) {
930 if (iphc1 & LOWPAN_IPHC_DAC) {
931 pr_debug("dest: context-based mcast compression\n");
932 /* TODO: implement this */
933 } else {
934 u8 prefix[] = {0xff, 0x02};
935
936 pr_debug("dest: non context-based mcast compression\n");
937 if (0 < tmp && tmp < 3) {
938 if (lowpan_fetch_skb_u8(skb, &prefix[1]))
939 goto drop;
940 }
941
942 err = lowpan_uncompress_addr(skb, &hdr.daddr, prefix,
943 lowpan_unc_mxconf[tmp], NULL);
944 if (err)
945 goto drop;
946 }
947 } else {
948 pr_debug("dest: stateless compression\n");
949 err = lowpan_uncompress_addr(skb, &hdr.daddr, lowpan_llprefix,
950 lowpan_unc_llconf[tmp], skb->data);
951 if (err)
952 goto drop;
953 }
954
955 /* UDP data uncompression */
956 if (iphc0 & LOWPAN_IPHC_NH_C) {
957 struct udphdr uh;
958 struct sk_buff *new;
959 if (lowpan_uncompress_udp_header(skb, &uh))
960 goto drop;
961
962 /*
963 * replace the compressed UDP head by the uncompressed UDP
964 * header
965 */
966 new = skb_copy_expand(skb, sizeof(struct udphdr),
967 skb_tailroom(skb), GFP_ATOMIC);
968 kfree_skb(skb);
969
970 if (!new)
971 return -ENOMEM;
972
973 skb = new;
974
975 skb_push(skb, sizeof(struct udphdr));
976 skb_reset_transport_header(skb);
977 skb_copy_to_linear_data(skb, &uh, sizeof(struct udphdr));
978
979 lowpan_raw_dump_table(__func__, "raw UDP header dump",
980 (u8 *)&uh, sizeof(uh));
981
982 hdr.nexthdr = UIP_PROTO_UDP;
983 }
984
985 /* Not fragmented package */
986 hdr.payload_len = htons(skb->len);
987
988 pr_debug("skb headroom size = %d, data length = %d\n",
989 skb_headroom(skb), skb->len);
990
991 pr_debug("IPv6 header dump:\n\tversion = %d\n\tlength = %d\n\t"
992 "nexthdr = 0x%02x\n\thop_lim = %d\n", hdr.version,
993 ntohs(hdr.payload_len), hdr.nexthdr, hdr.hop_limit);
994
995 lowpan_raw_dump_table(__func__, "raw header dump", (u8 *)&hdr,
996 sizeof(hdr));
997 return lowpan_skb_deliver(skb, &hdr);
998
999 unlock_and_drop:
1000 spin_unlock_bh(&flist_lock);
1001 drop:
1002 kfree_skb(skb);
1003 return -EINVAL;
1004 }
1005
1006 static int lowpan_set_address(struct net_device *dev, void *p)
1007 {
1008 struct sockaddr *sa = p;
1009
1010 if (netif_running(dev))
1011 return -EBUSY;
1012
1013 /* TODO: validate addr */
1014 memcpy(dev->dev_addr, sa->sa_data, dev->addr_len);
1015
1016 return 0;
1017 }
1018
1019 static int lowpan_get_mac_header_length(struct sk_buff *skb)
1020 {
1021 /*
1022 * Currently long addressing mode is supported only, so the overall
1023 * header size is 21:
1024 * FC SeqNum DPAN DA SA Sec
1025 * 2 + 1 + 2 + 8 + 8 + 0 = 21
1026 */
1027 return 21;
1028 }
1029
1030 static int
1031 lowpan_fragment_xmit(struct sk_buff *skb, u8 *head,
1032 int mlen, int plen, int offset, int type)
1033 {
1034 struct sk_buff *frag;
1035 int hlen, ret;
1036
1037 hlen = (type == LOWPAN_DISPATCH_FRAG1) ?
1038 LOWPAN_FRAG1_HEAD_SIZE : LOWPAN_FRAGN_HEAD_SIZE;
1039
1040 lowpan_raw_dump_inline(__func__, "6lowpan fragment header", head, hlen);
1041
1042 frag = dev_alloc_skb(hlen + mlen + plen + IEEE802154_MFR_SIZE);
1043 if (!frag)
1044 return -ENOMEM;
1045
1046 frag->priority = skb->priority;
1047 frag->dev = skb->dev;
1048
1049 /* copy header, MFR and payload */
1050 memcpy(skb_put(frag, mlen), skb->data, mlen);
1051 memcpy(skb_put(frag, hlen), head, hlen);
1052
1053 if (plen)
1054 skb_copy_from_linear_data_offset(skb, offset + mlen,
1055 skb_put(frag, plen), plen);
1056
1057 lowpan_raw_dump_table(__func__, " raw fragment dump", frag->data,
1058 frag->len);
1059
1060 ret = dev_queue_xmit(frag);
1061
1062 return ret;
1063 }
1064
1065 static int
1066 lowpan_skb_fragmentation(struct sk_buff *skb, struct net_device *dev)
1067 {
1068 int err, header_length, payload_length, tag, offset = 0;
1069 u8 head[5];
1070
1071 header_length = lowpan_get_mac_header_length(skb);
1072 payload_length = skb->len - header_length;
1073 tag = lowpan_dev_info(dev)->fragment_tag++;
1074
1075 /* first fragment header */
1076 head[0] = LOWPAN_DISPATCH_FRAG1 | ((payload_length >> 8) & 0x7);
1077 head[1] = payload_length & 0xff;
1078 head[2] = tag >> 8;
1079 head[3] = tag & 0xff;
1080
1081 err = lowpan_fragment_xmit(skb, head, header_length, LOWPAN_FRAG_SIZE,
1082 0, LOWPAN_DISPATCH_FRAG1);
1083
1084 if (err) {
1085 pr_debug("%s unable to send FRAG1 packet (tag: %d)",
1086 __func__, tag);
1087 goto exit;
1088 }
1089
1090 offset = LOWPAN_FRAG_SIZE;
1091
1092 /* next fragment header */
1093 head[0] &= ~LOWPAN_DISPATCH_FRAG1;
1094 head[0] |= LOWPAN_DISPATCH_FRAGN;
1095
1096 while ((payload_length - offset > 0) && (err >= 0)) {
1097 int len = LOWPAN_FRAG_SIZE;
1098
1099 head[4] = offset / 8;
1100
1101 if (payload_length - offset < len)
1102 len = payload_length - offset;
1103
1104 err = lowpan_fragment_xmit(skb, head, header_length,
1105 len, offset, LOWPAN_DISPATCH_FRAGN);
1106 if (err) {
1107 pr_debug("%s unable to send a subsequent FRAGN packet "
1108 "(tag: %d, offset: %d", __func__, tag, offset);
1109 goto exit;
1110 }
1111
1112 offset += len;
1113 }
1114
1115 exit:
1116 return err;
1117 }
1118
1119 static netdev_tx_t lowpan_xmit(struct sk_buff *skb, struct net_device *dev)
1120 {
1121 int err = -1;
1122
1123 pr_debug("package xmit\n");
1124
1125 skb->dev = lowpan_dev_info(dev)->real_dev;
1126 if (skb->dev == NULL) {
1127 pr_debug("ERROR: no real wpan device found\n");
1128 goto error;
1129 }
1130
1131 /* Send directly if less than the MTU minus the 2 checksum bytes. */
1132 if (skb->len <= IEEE802154_MTU - IEEE802154_MFR_SIZE) {
1133 err = dev_queue_xmit(skb);
1134 goto out;
1135 }
1136
1137 pr_debug("frame is too big, fragmentation is needed\n");
1138 err = lowpan_skb_fragmentation(skb, dev);
1139 error:
1140 dev_kfree_skb(skb);
1141 out:
1142 if (err)
1143 pr_debug("ERROR: xmit failed\n");
1144
1145 return (err < 0) ? NET_XMIT_DROP : err;
1146 }
1147
1148 static struct wpan_phy *lowpan_get_phy(const struct net_device *dev)
1149 {
1150 struct net_device *real_dev = lowpan_dev_info(dev)->real_dev;
1151 return ieee802154_mlme_ops(real_dev)->get_phy(real_dev);
1152 }
1153
1154 static u16 lowpan_get_pan_id(const struct net_device *dev)
1155 {
1156 struct net_device *real_dev = lowpan_dev_info(dev)->real_dev;
1157 return ieee802154_mlme_ops(real_dev)->get_pan_id(real_dev);
1158 }
1159
1160 static u16 lowpan_get_short_addr(const struct net_device *dev)
1161 {
1162 struct net_device *real_dev = lowpan_dev_info(dev)->real_dev;
1163 return ieee802154_mlme_ops(real_dev)->get_short_addr(real_dev);
1164 }
1165
1166 static u8 lowpan_get_dsn(const struct net_device *dev)
1167 {
1168 struct net_device *real_dev = lowpan_dev_info(dev)->real_dev;
1169 return ieee802154_mlme_ops(real_dev)->get_dsn(real_dev);
1170 }
1171
1172 static struct header_ops lowpan_header_ops = {
1173 .create = lowpan_header_create,
1174 };
1175
1176 static struct lock_class_key lowpan_tx_busylock;
1177 static struct lock_class_key lowpan_netdev_xmit_lock_key;
1178
1179 static void lowpan_set_lockdep_class_one(struct net_device *dev,
1180 struct netdev_queue *txq,
1181 void *_unused)
1182 {
1183 lockdep_set_class(&txq->_xmit_lock,
1184 &lowpan_netdev_xmit_lock_key);
1185 }
1186
1187
1188 static int lowpan_dev_init(struct net_device *dev)
1189 {
1190 netdev_for_each_tx_queue(dev, lowpan_set_lockdep_class_one, NULL);
1191 dev->qdisc_tx_busylock = &lowpan_tx_busylock;
1192 return 0;
1193 }
1194
1195 static const struct net_device_ops lowpan_netdev_ops = {
1196 .ndo_init = lowpan_dev_init,
1197 .ndo_start_xmit = lowpan_xmit,
1198 .ndo_set_mac_address = lowpan_set_address,
1199 };
1200
1201 static struct ieee802154_mlme_ops lowpan_mlme = {
1202 .get_pan_id = lowpan_get_pan_id,
1203 .get_phy = lowpan_get_phy,
1204 .get_short_addr = lowpan_get_short_addr,
1205 .get_dsn = lowpan_get_dsn,
1206 };
1207
1208 static void lowpan_setup(struct net_device *dev)
1209 {
1210 dev->addr_len = IEEE802154_ADDR_LEN;
1211 memset(dev->broadcast, 0xff, IEEE802154_ADDR_LEN);
1212 dev->type = ARPHRD_IEEE802154;
1213 /* Frame Control + Sequence Number + Address fields + Security Header */
1214 dev->hard_header_len = 2 + 1 + 20 + 14;
1215 dev->needed_tailroom = 2; /* FCS */
1216 dev->mtu = 1281;
1217 dev->tx_queue_len = 0;
1218 dev->flags = IFF_BROADCAST | IFF_MULTICAST;
1219 dev->watchdog_timeo = 0;
1220
1221 dev->netdev_ops = &lowpan_netdev_ops;
1222 dev->header_ops = &lowpan_header_ops;
1223 dev->ml_priv = &lowpan_mlme;
1224 dev->destructor = free_netdev;
1225 }
1226
1227 static int lowpan_validate(struct nlattr *tb[], struct nlattr *data[])
1228 {
1229 if (tb[IFLA_ADDRESS]) {
1230 if (nla_len(tb[IFLA_ADDRESS]) != IEEE802154_ADDR_LEN)
1231 return -EINVAL;
1232 }
1233 return 0;
1234 }
1235
1236 static int lowpan_rcv(struct sk_buff *skb, struct net_device *dev,
1237 struct packet_type *pt, struct net_device *orig_dev)
1238 {
1239 struct sk_buff *local_skb;
1240
1241 if (!netif_running(dev))
1242 goto drop;
1243
1244 if (dev->type != ARPHRD_IEEE802154)
1245 goto drop;
1246
1247 /* check that it's our buffer */
1248 if (skb->data[0] == LOWPAN_DISPATCH_IPV6) {
1249 /* Copy the packet so that the IPv6 header is
1250 * properly aligned.
1251 */
1252 local_skb = skb_copy_expand(skb, NET_SKB_PAD - 1,
1253 skb_tailroom(skb), GFP_ATOMIC);
1254 if (!local_skb)
1255 goto drop;
1256
1257 local_skb->protocol = htons(ETH_P_IPV6);
1258 local_skb->pkt_type = PACKET_HOST;
1259
1260 /* Pull off the 1-byte of 6lowpan header. */
1261 skb_pull(local_skb, 1);
1262 skb_reset_network_header(local_skb);
1263 skb_set_transport_header(local_skb, sizeof(struct ipv6hdr));
1264
1265 lowpan_give_skb_to_devices(local_skb);
1266
1267 kfree_skb(local_skb);
1268 kfree_skb(skb);
1269 } else {
1270 switch (skb->data[0] & 0xe0) {
1271 case LOWPAN_DISPATCH_IPHC: /* ipv6 datagram */
1272 case LOWPAN_DISPATCH_FRAG1: /* first fragment header */
1273 case LOWPAN_DISPATCH_FRAGN: /* next fragments headers */
1274 local_skb = skb_clone(skb, GFP_ATOMIC);
1275 if (!local_skb)
1276 goto drop;
1277 lowpan_process_data(local_skb);
1278
1279 kfree_skb(skb);
1280 break;
1281 default:
1282 break;
1283 }
1284 }
1285
1286 return NET_RX_SUCCESS;
1287
1288 drop:
1289 kfree_skb(skb);
1290 return NET_RX_DROP;
1291 }
1292
1293 static int lowpan_newlink(struct net *src_net, struct net_device *dev,
1294 struct nlattr *tb[], struct nlattr *data[])
1295 {
1296 struct net_device *real_dev;
1297 struct lowpan_dev_record *entry;
1298
1299 pr_debug("adding new link\n");
1300
1301 if (!tb[IFLA_LINK])
1302 return -EINVAL;
1303 /* find and hold real wpan device */
1304 real_dev = dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1305 if (!real_dev)
1306 return -ENODEV;
1307
1308 lowpan_dev_info(dev)->real_dev = real_dev;
1309 lowpan_dev_info(dev)->fragment_tag = 0;
1310 mutex_init(&lowpan_dev_info(dev)->dev_list_mtx);
1311
1312 entry = kzalloc(sizeof(struct lowpan_dev_record), GFP_KERNEL);
1313 if (!entry) {
1314 dev_put(real_dev);
1315 lowpan_dev_info(dev)->real_dev = NULL;
1316 return -ENOMEM;
1317 }
1318
1319 entry->ldev = dev;
1320
1321 mutex_lock(&lowpan_dev_info(dev)->dev_list_mtx);
1322 INIT_LIST_HEAD(&entry->list);
1323 list_add_tail(&entry->list, &lowpan_devices);
1324 mutex_unlock(&lowpan_dev_info(dev)->dev_list_mtx);
1325
1326 register_netdevice(dev);
1327
1328 return 0;
1329 }
1330
1331 static void lowpan_dellink(struct net_device *dev, struct list_head *head)
1332 {
1333 struct lowpan_dev_info *lowpan_dev = lowpan_dev_info(dev);
1334 struct net_device *real_dev = lowpan_dev->real_dev;
1335 struct lowpan_dev_record *entry, *tmp;
1336
1337 ASSERT_RTNL();
1338
1339 mutex_lock(&lowpan_dev_info(dev)->dev_list_mtx);
1340 list_for_each_entry_safe(entry, tmp, &lowpan_devices, list) {
1341 if (entry->ldev == dev) {
1342 list_del(&entry->list);
1343 kfree(entry);
1344 }
1345 }
1346 mutex_unlock(&lowpan_dev_info(dev)->dev_list_mtx);
1347
1348 mutex_destroy(&lowpan_dev_info(dev)->dev_list_mtx);
1349
1350 unregister_netdevice_queue(dev, head);
1351
1352 dev_put(real_dev);
1353 }
1354
1355 static struct rtnl_link_ops lowpan_link_ops __read_mostly = {
1356 .kind = "lowpan",
1357 .priv_size = sizeof(struct lowpan_dev_info),
1358 .setup = lowpan_setup,
1359 .newlink = lowpan_newlink,
1360 .dellink = lowpan_dellink,
1361 .validate = lowpan_validate,
1362 };
1363
1364 static inline int __init lowpan_netlink_init(void)
1365 {
1366 return rtnl_link_register(&lowpan_link_ops);
1367 }
1368
1369 static inline void lowpan_netlink_fini(void)
1370 {
1371 rtnl_link_unregister(&lowpan_link_ops);
1372 }
1373
1374 static int lowpan_device_event(struct notifier_block *unused,
1375 unsigned long event,
1376 void *ptr)
1377 {
1378 struct net_device *dev = ptr;
1379 LIST_HEAD(del_list);
1380 struct lowpan_dev_record *entry, *tmp;
1381
1382 if (dev->type != ARPHRD_IEEE802154)
1383 goto out;
1384
1385 if (event == NETDEV_UNREGISTER) {
1386 list_for_each_entry_safe(entry, tmp, &lowpan_devices, list) {
1387 if (lowpan_dev_info(entry->ldev)->real_dev == dev)
1388 lowpan_dellink(entry->ldev, &del_list);
1389 }
1390
1391 unregister_netdevice_many(&del_list);
1392 }
1393
1394 out:
1395 return NOTIFY_DONE;
1396 }
1397
1398 static struct notifier_block lowpan_dev_notifier = {
1399 .notifier_call = lowpan_device_event,
1400 };
1401
1402 static struct packet_type lowpan_packet_type = {
1403 .type = __constant_htons(ETH_P_IEEE802154),
1404 .func = lowpan_rcv,
1405 };
1406
1407 static int __init lowpan_init_module(void)
1408 {
1409 int err = 0;
1410
1411 err = lowpan_netlink_init();
1412 if (err < 0)
1413 goto out;
1414
1415 dev_add_pack(&lowpan_packet_type);
1416
1417 err = register_netdevice_notifier(&lowpan_dev_notifier);
1418 if (err < 0) {
1419 dev_remove_pack(&lowpan_packet_type);
1420 lowpan_netlink_fini();
1421 }
1422 out:
1423 return err;
1424 }
1425
1426 static void __exit lowpan_cleanup_module(void)
1427 {
1428 struct lowpan_fragment *frame, *tframe;
1429
1430 lowpan_netlink_fini();
1431
1432 dev_remove_pack(&lowpan_packet_type);
1433
1434 unregister_netdevice_notifier(&lowpan_dev_notifier);
1435
1436 /* Now 6lowpan packet_type is removed, so no new fragments are
1437 * expected on RX, therefore that's the time to clean incomplete
1438 * fragments.
1439 */
1440 spin_lock_bh(&flist_lock);
1441 list_for_each_entry_safe(frame, tframe, &lowpan_fragments, list) {
1442 del_timer_sync(&frame->timer);
1443 list_del(&frame->list);
1444 dev_kfree_skb(frame->skb);
1445 kfree(frame);
1446 }
1447 spin_unlock_bh(&flist_lock);
1448 }
1449
1450 module_init(lowpan_init_module);
1451 module_exit(lowpan_cleanup_module);
1452 MODULE_LICENSE("GPL");
1453 MODULE_ALIAS_RTNL_LINK("lowpan");