xref: /openbmc/linux/net/tipc/msg.c (revision 50fc8d92)
1 /*
2  * net/tipc/msg.c: TIPC message header routines
3  *
4  * Copyright (c) 2000-2006, 2014-2015, Ericsson AB
5  * Copyright (c) 2005, 2010-2011, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 #include <net/sock.h>
38 #include "core.h"
39 #include "msg.h"
40 #include "addr.h"
41 #include "name_table.h"
42 #include "crypto.h"
43 
44 #define MAX_FORWARD_SIZE 1024
45 #ifdef CONFIG_TIPC_CRYPTO
46 #define BUF_HEADROOM ALIGN(((LL_MAX_HEADER + 48) + EHDR_MAX_SIZE), 16)
47 #define BUF_TAILROOM (TIPC_AES_GCM_TAG_SIZE)
48 #else
49 #define BUF_HEADROOM (LL_MAX_HEADER + 48)
50 #define BUF_TAILROOM 16
51 #endif
52 
53 static unsigned int align(unsigned int i)
54 {
55 	return (i + 3) & ~3u;
56 }
57 
58 /**
59  * tipc_buf_acquire - creates a TIPC message buffer
60  * @size: message size (including TIPC header)
61  * @gfp: memory allocation flags
62  *
63  * Return: a new buffer with data pointers set to the specified size.
64  *
65  * NOTE:
66  * Headroom is reserved to allow prepending of a data link header.
67  * There may also be unrequested tailroom present at the buffer's end.
68  */
69 struct sk_buff *tipc_buf_acquire(u32 size, gfp_t gfp)
70 {
71 	struct sk_buff *skb;
72 #ifdef CONFIG_TIPC_CRYPTO
73 	unsigned int buf_size = (BUF_HEADROOM + size + BUF_TAILROOM + 3) & ~3u;
74 #else
75 	unsigned int buf_size = (BUF_HEADROOM + size + 3) & ~3u;
76 #endif
77 
78 	skb = alloc_skb_fclone(buf_size, gfp);
79 	if (skb) {
80 		skb_reserve(skb, BUF_HEADROOM);
81 		skb_put(skb, size);
82 		skb->next = NULL;
83 	}
84 	return skb;
85 }
86 
87 void tipc_msg_init(u32 own_node, struct tipc_msg *m, u32 user, u32 type,
88 		   u32 hsize, u32 dnode)
89 {
90 	memset(m, 0, hsize);
91 	msg_set_version(m);
92 	msg_set_user(m, user);
93 	msg_set_hdr_sz(m, hsize);
94 	msg_set_size(m, hsize);
95 	msg_set_prevnode(m, own_node);
96 	msg_set_type(m, type);
97 	if (hsize > SHORT_H_SIZE) {
98 		msg_set_orignode(m, own_node);
99 		msg_set_destnode(m, dnode);
100 	}
101 }
102 
103 struct sk_buff *tipc_msg_create(uint user, uint type,
104 				uint hdr_sz, uint data_sz, u32 dnode,
105 				u32 onode, u32 dport, u32 oport, int errcode)
106 {
107 	struct tipc_msg *msg;
108 	struct sk_buff *buf;
109 
110 	buf = tipc_buf_acquire(hdr_sz + data_sz, GFP_ATOMIC);
111 	if (unlikely(!buf))
112 		return NULL;
113 
114 	msg = buf_msg(buf);
115 	tipc_msg_init(onode, msg, user, type, hdr_sz, dnode);
116 	msg_set_size(msg, hdr_sz + data_sz);
117 	msg_set_origport(msg, oport);
118 	msg_set_destport(msg, dport);
119 	msg_set_errcode(msg, errcode);
120 	if (hdr_sz > SHORT_H_SIZE) {
121 		msg_set_orignode(msg, onode);
122 		msg_set_destnode(msg, dnode);
123 	}
124 	return buf;
125 }
126 
127 /* tipc_buf_append(): Append a buffer to the fragment list of another buffer
128  * @*headbuf: in:  NULL for first frag, otherwise value returned from prev call
129  *            out: set when successful non-complete reassembly, otherwise NULL
130  * @*buf:     in:  the buffer to append. Always defined
131  *            out: head buf after successful complete reassembly, otherwise NULL
132  * Returns 1 when reassembly complete, otherwise 0
133  */
134 int tipc_buf_append(struct sk_buff **headbuf, struct sk_buff **buf)
135 {
136 	struct sk_buff *head = *headbuf;
137 	struct sk_buff *frag = *buf;
138 	struct sk_buff *tail = NULL;
139 	struct tipc_msg *msg;
140 	u32 fragid;
141 	int delta;
142 	bool headstolen;
143 
144 	if (!frag)
145 		goto err;
146 
147 	msg = buf_msg(frag);
148 	fragid = msg_type(msg);
149 	frag->next = NULL;
150 	skb_pull(frag, msg_hdr_sz(msg));
151 
152 	if (fragid == FIRST_FRAGMENT) {
153 		if (unlikely(head))
154 			goto err;
155 		*buf = NULL;
156 		frag = skb_unshare(frag, GFP_ATOMIC);
157 		if (unlikely(!frag))
158 			goto err;
159 		head = *headbuf = frag;
160 		TIPC_SKB_CB(head)->tail = NULL;
161 		if (skb_is_nonlinear(head)) {
162 			skb_walk_frags(head, tail) {
163 				TIPC_SKB_CB(head)->tail = tail;
164 			}
165 		} else {
166 			skb_frag_list_init(head);
167 		}
168 		return 0;
169 	}
170 
171 	if (!head)
172 		goto err;
173 
174 	if (skb_try_coalesce(head, frag, &headstolen, &delta)) {
175 		kfree_skb_partial(frag, headstolen);
176 	} else {
177 		tail = TIPC_SKB_CB(head)->tail;
178 		if (!skb_has_frag_list(head))
179 			skb_shinfo(head)->frag_list = frag;
180 		else
181 			tail->next = frag;
182 		head->truesize += frag->truesize;
183 		head->data_len += frag->len;
184 		head->len += frag->len;
185 		TIPC_SKB_CB(head)->tail = frag;
186 	}
187 
188 	if (fragid == LAST_FRAGMENT) {
189 		TIPC_SKB_CB(head)->validated = 0;
190 		if (unlikely(!tipc_msg_validate(&head)))
191 			goto err;
192 		*buf = head;
193 		TIPC_SKB_CB(head)->tail = NULL;
194 		*headbuf = NULL;
195 		return 1;
196 	}
197 	*buf = NULL;
198 	return 0;
199 err:
200 	kfree_skb(*buf);
201 	kfree_skb(*headbuf);
202 	*buf = *headbuf = NULL;
203 	return 0;
204 }
205 
206 /**
207  * tipc_msg_append(): Append data to tail of an existing buffer queue
208  * @_hdr: header to be used
209  * @m: the data to be appended
210  * @mss: max allowable size of buffer
211  * @dlen: size of data to be appended
212  * @txq: queue to append to
213  *
214  * Return: the number of 1k blocks appended or errno value
215  */
216 int tipc_msg_append(struct tipc_msg *_hdr, struct msghdr *m, int dlen,
217 		    int mss, struct sk_buff_head *txq)
218 {
219 	struct sk_buff *skb;
220 	int accounted, total, curr;
221 	int mlen, cpy, rem = dlen;
222 	struct tipc_msg *hdr;
223 
224 	skb = skb_peek_tail(txq);
225 	accounted = skb ? msg_blocks(buf_msg(skb)) : 0;
226 	total = accounted;
227 
228 	do {
229 		if (!skb || skb->len >= mss) {
230 			skb = tipc_buf_acquire(mss, GFP_KERNEL);
231 			if (unlikely(!skb))
232 				return -ENOMEM;
233 			skb_orphan(skb);
234 			skb_trim(skb, MIN_H_SIZE);
235 			hdr = buf_msg(skb);
236 			skb_copy_to_linear_data(skb, _hdr, MIN_H_SIZE);
237 			msg_set_hdr_sz(hdr, MIN_H_SIZE);
238 			msg_set_size(hdr, MIN_H_SIZE);
239 			__skb_queue_tail(txq, skb);
240 			total += 1;
241 		}
242 		hdr = buf_msg(skb);
243 		curr = msg_blocks(hdr);
244 		mlen = msg_size(hdr);
245 		cpy = min_t(size_t, rem, mss - mlen);
246 		if (cpy != copy_from_iter(skb->data + mlen, cpy, &m->msg_iter))
247 			return -EFAULT;
248 		msg_set_size(hdr, mlen + cpy);
249 		skb_put(skb, cpy);
250 		rem -= cpy;
251 		total += msg_blocks(hdr) - curr;
252 	} while (rem > 0);
253 	return total - accounted;
254 }
255 
256 /* tipc_msg_validate - validate basic format of received message
257  *
258  * This routine ensures a TIPC message has an acceptable header, and at least
259  * as much data as the header indicates it should.  The routine also ensures
260  * that the entire message header is stored in the main fragment of the message
261  * buffer, to simplify future access to message header fields.
262  *
263  * Note: Having extra info present in the message header or data areas is OK.
264  * TIPC will ignore the excess, under the assumption that it is optional info
265  * introduced by a later release of the protocol.
266  */
267 bool tipc_msg_validate(struct sk_buff **_skb)
268 {
269 	struct sk_buff *skb = *_skb;
270 	struct tipc_msg *hdr;
271 	int msz, hsz;
272 
273 	/* Ensure that flow control ratio condition is satisfied */
274 	if (unlikely(skb->truesize / buf_roundup_len(skb) >= 4)) {
275 		skb = skb_copy_expand(skb, BUF_HEADROOM, 0, GFP_ATOMIC);
276 		if (!skb)
277 			return false;
278 		kfree_skb(*_skb);
279 		*_skb = skb;
280 	}
281 
282 	if (unlikely(TIPC_SKB_CB(skb)->validated))
283 		return true;
284 
285 	if (unlikely(!pskb_may_pull(skb, MIN_H_SIZE)))
286 		return false;
287 
288 	hsz = msg_hdr_sz(buf_msg(skb));
289 	if (unlikely(hsz < MIN_H_SIZE) || (hsz > MAX_H_SIZE))
290 		return false;
291 	if (unlikely(!pskb_may_pull(skb, hsz)))
292 		return false;
293 
294 	hdr = buf_msg(skb);
295 	if (unlikely(msg_version(hdr) != TIPC_VERSION))
296 		return false;
297 
298 	msz = msg_size(hdr);
299 	if (unlikely(msz < hsz))
300 		return false;
301 	if (unlikely((msz - hsz) > TIPC_MAX_USER_MSG_SIZE))
302 		return false;
303 	if (unlikely(skb->len < msz))
304 		return false;
305 
306 	TIPC_SKB_CB(skb)->validated = 1;
307 	return true;
308 }
309 
310 /**
311  * tipc_msg_fragment - build a fragment skb list for TIPC message
312  *
313  * @skb: TIPC message skb
314  * @hdr: internal msg header to be put on the top of the fragments
315  * @pktmax: max size of a fragment incl. the header
316  * @frags: returned fragment skb list
317  *
318  * Return: 0 if the fragmentation is successful, otherwise: -EINVAL
319  * or -ENOMEM
320  */
321 int tipc_msg_fragment(struct sk_buff *skb, const struct tipc_msg *hdr,
322 		      int pktmax, struct sk_buff_head *frags)
323 {
324 	int pktno, nof_fragms, dsz, dmax, eat;
325 	struct tipc_msg *_hdr;
326 	struct sk_buff *_skb;
327 	u8 *data;
328 
329 	/* Non-linear buffer? */
330 	if (skb_linearize(skb))
331 		return -ENOMEM;
332 
333 	data = (u8 *)skb->data;
334 	dsz = msg_size(buf_msg(skb));
335 	dmax = pktmax - INT_H_SIZE;
336 	if (dsz <= dmax || !dmax)
337 		return -EINVAL;
338 
339 	nof_fragms = dsz / dmax + 1;
340 	for (pktno = 1; pktno <= nof_fragms; pktno++) {
341 		if (pktno < nof_fragms)
342 			eat = dmax;
343 		else
344 			eat = dsz % dmax;
345 		/* Allocate a new fragment */
346 		_skb = tipc_buf_acquire(INT_H_SIZE + eat, GFP_ATOMIC);
347 		if (!_skb)
348 			goto error;
349 		skb_orphan(_skb);
350 		__skb_queue_tail(frags, _skb);
351 		/* Copy header & data to the fragment */
352 		skb_copy_to_linear_data(_skb, hdr, INT_H_SIZE);
353 		skb_copy_to_linear_data_offset(_skb, INT_H_SIZE, data, eat);
354 		data += eat;
355 		/* Update the fragment's header */
356 		_hdr = buf_msg(_skb);
357 		msg_set_fragm_no(_hdr, pktno);
358 		msg_set_nof_fragms(_hdr, nof_fragms);
359 		msg_set_size(_hdr, INT_H_SIZE + eat);
360 	}
361 	return 0;
362 
363 error:
364 	__skb_queue_purge(frags);
365 	__skb_queue_head_init(frags);
366 	return -ENOMEM;
367 }
368 
369 /**
370  * tipc_msg_build - create buffer chain containing specified header and data
371  * @mhdr: Message header, to be prepended to data
372  * @m: User message
373  * @offset: buffer offset for fragmented messages (FIXME)
374  * @dsz: Total length of user data
375  * @pktmax: Max packet size that can be used
376  * @list: Buffer or chain of buffers to be returned to caller
377  *
378  * Note that the recursive call we are making here is safe, since it can
379  * logically go only one further level down.
380  *
381  * Return: message data size or errno: -ENOMEM, -EFAULT
382  */
383 int tipc_msg_build(struct tipc_msg *mhdr, struct msghdr *m, int offset,
384 		   int dsz, int pktmax, struct sk_buff_head *list)
385 {
386 	int mhsz = msg_hdr_sz(mhdr);
387 	struct tipc_msg pkthdr;
388 	int msz = mhsz + dsz;
389 	int pktrem = pktmax;
390 	struct sk_buff *skb;
391 	int drem = dsz;
392 	int pktno = 1;
393 	char *pktpos;
394 	int pktsz;
395 	int rc;
396 
397 	msg_set_size(mhdr, msz);
398 
399 	/* No fragmentation needed? */
400 	if (likely(msz <= pktmax)) {
401 		skb = tipc_buf_acquire(msz, GFP_KERNEL);
402 
403 		/* Fall back to smaller MTU if node local message */
404 		if (unlikely(!skb)) {
405 			if (pktmax != MAX_MSG_SIZE)
406 				return -ENOMEM;
407 			rc = tipc_msg_build(mhdr, m, offset, dsz, FB_MTU, list);
408 			if (rc != dsz)
409 				return rc;
410 			if (tipc_msg_assemble(list))
411 				return dsz;
412 			return -ENOMEM;
413 		}
414 		skb_orphan(skb);
415 		__skb_queue_tail(list, skb);
416 		skb_copy_to_linear_data(skb, mhdr, mhsz);
417 		pktpos = skb->data + mhsz;
418 		if (copy_from_iter_full(pktpos, dsz, &m->msg_iter))
419 			return dsz;
420 		rc = -EFAULT;
421 		goto error;
422 	}
423 
424 	/* Prepare reusable fragment header */
425 	tipc_msg_init(msg_prevnode(mhdr), &pkthdr, MSG_FRAGMENTER,
426 		      FIRST_FRAGMENT, INT_H_SIZE, msg_destnode(mhdr));
427 	msg_set_size(&pkthdr, pktmax);
428 	msg_set_fragm_no(&pkthdr, pktno);
429 	msg_set_importance(&pkthdr, msg_importance(mhdr));
430 
431 	/* Prepare first fragment */
432 	skb = tipc_buf_acquire(pktmax, GFP_KERNEL);
433 	if (!skb)
434 		return -ENOMEM;
435 	skb_orphan(skb);
436 	__skb_queue_tail(list, skb);
437 	pktpos = skb->data;
438 	skb_copy_to_linear_data(skb, &pkthdr, INT_H_SIZE);
439 	pktpos += INT_H_SIZE;
440 	pktrem -= INT_H_SIZE;
441 	skb_copy_to_linear_data_offset(skb, INT_H_SIZE, mhdr, mhsz);
442 	pktpos += mhsz;
443 	pktrem -= mhsz;
444 
445 	do {
446 		if (drem < pktrem)
447 			pktrem = drem;
448 
449 		if (!copy_from_iter_full(pktpos, pktrem, &m->msg_iter)) {
450 			rc = -EFAULT;
451 			goto error;
452 		}
453 		drem -= pktrem;
454 
455 		if (!drem)
456 			break;
457 
458 		/* Prepare new fragment: */
459 		if (drem < (pktmax - INT_H_SIZE))
460 			pktsz = drem + INT_H_SIZE;
461 		else
462 			pktsz = pktmax;
463 		skb = tipc_buf_acquire(pktsz, GFP_KERNEL);
464 		if (!skb) {
465 			rc = -ENOMEM;
466 			goto error;
467 		}
468 		skb_orphan(skb);
469 		__skb_queue_tail(list, skb);
470 		msg_set_type(&pkthdr, FRAGMENT);
471 		msg_set_size(&pkthdr, pktsz);
472 		msg_set_fragm_no(&pkthdr, ++pktno);
473 		skb_copy_to_linear_data(skb, &pkthdr, INT_H_SIZE);
474 		pktpos = skb->data + INT_H_SIZE;
475 		pktrem = pktsz - INT_H_SIZE;
476 
477 	} while (1);
478 	msg_set_type(buf_msg(skb), LAST_FRAGMENT);
479 	return dsz;
480 error:
481 	__skb_queue_purge(list);
482 	__skb_queue_head_init(list);
483 	return rc;
484 }
485 
486 /**
487  * tipc_msg_bundle - Append contents of a buffer to tail of an existing one
488  * @bskb: the bundle buffer to append to
489  * @msg: message to be appended
490  * @max: max allowable size for the bundle buffer
491  *
492  * Return: "true" if bundling has been performed, otherwise "false"
493  */
494 static bool tipc_msg_bundle(struct sk_buff *bskb, struct tipc_msg *msg,
495 			    u32 max)
496 {
497 	struct tipc_msg *bmsg = buf_msg(bskb);
498 	u32 msz, bsz, offset, pad;
499 
500 	msz = msg_size(msg);
501 	bsz = msg_size(bmsg);
502 	offset = align(bsz);
503 	pad = offset - bsz;
504 
505 	if (unlikely(skb_tailroom(bskb) < (pad + msz)))
506 		return false;
507 	if (unlikely(max < (offset + msz)))
508 		return false;
509 
510 	skb_put(bskb, pad + msz);
511 	skb_copy_to_linear_data_offset(bskb, offset, msg, msz);
512 	msg_set_size(bmsg, offset + msz);
513 	msg_set_msgcnt(bmsg, msg_msgcnt(bmsg) + 1);
514 	return true;
515 }
516 
517 /**
518  * tipc_msg_try_bundle - Try to bundle a new message to the last one
519  * @tskb: the last/target message to which the new one will be appended
520  * @skb: the new message skb pointer
521  * @mss: max message size (header inclusive)
522  * @dnode: destination node for the message
523  * @new_bundle: if this call made a new bundle or not
524  *
525  * Return: "true" if the new message skb is potential for bundling this time or
526  * later, in the case a bundling has been done this time, the skb is consumed
527  * (the skb pointer = NULL).
528  * Otherwise, "false" if the skb cannot be bundled at all.
529  */
530 bool tipc_msg_try_bundle(struct sk_buff *tskb, struct sk_buff **skb, u32 mss,
531 			 u32 dnode, bool *new_bundle)
532 {
533 	struct tipc_msg *msg, *inner, *outer;
534 	u32 tsz;
535 
536 	/* First, check if the new buffer is suitable for bundling */
537 	msg = buf_msg(*skb);
538 	if (msg_user(msg) == MSG_FRAGMENTER)
539 		return false;
540 	if (msg_user(msg) == TUNNEL_PROTOCOL)
541 		return false;
542 	if (msg_user(msg) == BCAST_PROTOCOL)
543 		return false;
544 	if (mss <= INT_H_SIZE + msg_size(msg))
545 		return false;
546 
547 	/* Ok, but the last/target buffer can be empty? */
548 	if (unlikely(!tskb))
549 		return true;
550 
551 	/* Is it a bundle already? Try to bundle the new message to it */
552 	if (msg_user(buf_msg(tskb)) == MSG_BUNDLER) {
553 		*new_bundle = false;
554 		goto bundle;
555 	}
556 
557 	/* Make a new bundle of the two messages if possible */
558 	tsz = msg_size(buf_msg(tskb));
559 	if (unlikely(mss < align(INT_H_SIZE + tsz) + msg_size(msg)))
560 		return true;
561 	if (unlikely(pskb_expand_head(tskb, INT_H_SIZE, mss - tsz - INT_H_SIZE,
562 				      GFP_ATOMIC)))
563 		return true;
564 	inner = buf_msg(tskb);
565 	skb_push(tskb, INT_H_SIZE);
566 	outer = buf_msg(tskb);
567 	tipc_msg_init(msg_prevnode(inner), outer, MSG_BUNDLER, 0, INT_H_SIZE,
568 		      dnode);
569 	msg_set_importance(outer, msg_importance(inner));
570 	msg_set_size(outer, INT_H_SIZE + tsz);
571 	msg_set_msgcnt(outer, 1);
572 	*new_bundle = true;
573 
574 bundle:
575 	if (likely(tipc_msg_bundle(tskb, msg, mss))) {
576 		consume_skb(*skb);
577 		*skb = NULL;
578 	}
579 	return true;
580 }
581 
582 /**
583  *  tipc_msg_extract(): extract bundled inner packet from buffer
584  *  @skb: buffer to be extracted from.
585  *  @iskb: extracted inner buffer, to be returned
586  *  @pos: position in outer message of msg to be extracted.
587  *  Returns position of next msg.
588  *  Consumes outer buffer when last packet extracted
589  *  Return: true when there is an extracted buffer, otherwise false
590  */
591 bool tipc_msg_extract(struct sk_buff *skb, struct sk_buff **iskb, int *pos)
592 {
593 	struct tipc_msg *hdr, *ihdr;
594 	int imsz;
595 
596 	*iskb = NULL;
597 	if (unlikely(skb_linearize(skb)))
598 		goto none;
599 
600 	hdr = buf_msg(skb);
601 	if (unlikely(*pos > (msg_data_sz(hdr) - MIN_H_SIZE)))
602 		goto none;
603 
604 	ihdr = (struct tipc_msg *)(msg_data(hdr) + *pos);
605 	imsz = msg_size(ihdr);
606 
607 	if ((*pos + imsz) > msg_data_sz(hdr))
608 		goto none;
609 
610 	*iskb = tipc_buf_acquire(imsz, GFP_ATOMIC);
611 	if (!*iskb)
612 		goto none;
613 
614 	skb_copy_to_linear_data(*iskb, ihdr, imsz);
615 	if (unlikely(!tipc_msg_validate(iskb)))
616 		goto none;
617 
618 	*pos += align(imsz);
619 	return true;
620 none:
621 	kfree_skb(skb);
622 	kfree_skb(*iskb);
623 	*iskb = NULL;
624 	return false;
625 }
626 
627 /**
628  * tipc_msg_reverse(): swap source and destination addresses and add error code
629  * @own_node: originating node id for reversed message
630  * @skb:  buffer containing message to be reversed; will be consumed
631  * @err:  error code to be set in message, if any
632  * Replaces consumed buffer with new one when successful
633  * Return: true if success, otherwise false
634  */
635 bool tipc_msg_reverse(u32 own_node,  struct sk_buff **skb, int err)
636 {
637 	struct sk_buff *_skb = *skb;
638 	struct tipc_msg *_hdr, *hdr;
639 	int hlen, dlen;
640 
641 	if (skb_linearize(_skb))
642 		goto exit;
643 	_hdr = buf_msg(_skb);
644 	dlen = min_t(uint, msg_data_sz(_hdr), MAX_FORWARD_SIZE);
645 	hlen = msg_hdr_sz(_hdr);
646 
647 	if (msg_dest_droppable(_hdr))
648 		goto exit;
649 	if (msg_errcode(_hdr))
650 		goto exit;
651 
652 	/* Never return SHORT header */
653 	if (hlen == SHORT_H_SIZE)
654 		hlen = BASIC_H_SIZE;
655 
656 	/* Don't return data along with SYN+, - sender has a clone */
657 	if (msg_is_syn(_hdr) && err == TIPC_ERR_OVERLOAD)
658 		dlen = 0;
659 
660 	/* Allocate new buffer to return */
661 	*skb = tipc_buf_acquire(hlen + dlen, GFP_ATOMIC);
662 	if (!*skb)
663 		goto exit;
664 	memcpy((*skb)->data, _skb->data, msg_hdr_sz(_hdr));
665 	memcpy((*skb)->data + hlen, msg_data(_hdr), dlen);
666 
667 	/* Build reverse header in new buffer */
668 	hdr = buf_msg(*skb);
669 	msg_set_hdr_sz(hdr, hlen);
670 	msg_set_errcode(hdr, err);
671 	msg_set_non_seq(hdr, 0);
672 	msg_set_origport(hdr, msg_destport(_hdr));
673 	msg_set_destport(hdr, msg_origport(_hdr));
674 	msg_set_destnode(hdr, msg_prevnode(_hdr));
675 	msg_set_prevnode(hdr, own_node);
676 	msg_set_orignode(hdr, own_node);
677 	msg_set_size(hdr, hlen + dlen);
678 	skb_orphan(_skb);
679 	kfree_skb(_skb);
680 	return true;
681 exit:
682 	kfree_skb(_skb);
683 	*skb = NULL;
684 	return false;
685 }
686 
687 bool tipc_msg_skb_clone(struct sk_buff_head *msg, struct sk_buff_head *cpy)
688 {
689 	struct sk_buff *skb, *_skb;
690 
691 	skb_queue_walk(msg, skb) {
692 		_skb = skb_clone(skb, GFP_ATOMIC);
693 		if (!_skb) {
694 			__skb_queue_purge(cpy);
695 			pr_err_ratelimited("Failed to clone buffer chain\n");
696 			return false;
697 		}
698 		__skb_queue_tail(cpy, _skb);
699 	}
700 	return true;
701 }
702 
703 /**
704  * tipc_msg_lookup_dest(): try to find new destination for named message
705  * @net: pointer to associated network namespace
706  * @skb: the buffer containing the message.
707  * @err: error code to be used by caller if lookup fails
708  * Does not consume buffer
709  * Return: true if a destination is found, false otherwise
710  */
711 bool tipc_msg_lookup_dest(struct net *net, struct sk_buff *skb, int *err)
712 {
713 	struct tipc_msg *msg = buf_msg(skb);
714 	u32 dport, dnode;
715 	u32 onode = tipc_own_addr(net);
716 
717 	if (!msg_isdata(msg))
718 		return false;
719 	if (!msg_named(msg))
720 		return false;
721 	if (msg_errcode(msg))
722 		return false;
723 	*err = TIPC_ERR_NO_NAME;
724 	if (skb_linearize(skb))
725 		return false;
726 	msg = buf_msg(skb);
727 	if (msg_reroute_cnt(msg))
728 		return false;
729 	dnode = tipc_scope2node(net, msg_lookup_scope(msg));
730 	dport = tipc_nametbl_translate(net, msg_nametype(msg),
731 				       msg_nameinst(msg), &dnode);
732 	if (!dport)
733 		return false;
734 	msg_incr_reroute_cnt(msg);
735 	if (dnode != onode)
736 		msg_set_prevnode(msg, onode);
737 	msg_set_destnode(msg, dnode);
738 	msg_set_destport(msg, dport);
739 	*err = TIPC_OK;
740 
741 	return true;
742 }
743 
744 /* tipc_msg_assemble() - assemble chain of fragments into one message
745  */
746 bool tipc_msg_assemble(struct sk_buff_head *list)
747 {
748 	struct sk_buff *skb, *tmp = NULL;
749 
750 	if (skb_queue_len(list) == 1)
751 		return true;
752 
753 	while ((skb = __skb_dequeue(list))) {
754 		skb->next = NULL;
755 		if (tipc_buf_append(&tmp, &skb)) {
756 			__skb_queue_tail(list, skb);
757 			return true;
758 		}
759 		if (!tmp)
760 			break;
761 	}
762 	__skb_queue_purge(list);
763 	__skb_queue_head_init(list);
764 	pr_warn("Failed do assemble buffer\n");
765 	return false;
766 }
767 
768 /* tipc_msg_reassemble() - clone a buffer chain of fragments and
769  *                         reassemble the clones into one message
770  */
771 bool tipc_msg_reassemble(struct sk_buff_head *list, struct sk_buff_head *rcvq)
772 {
773 	struct sk_buff *skb, *_skb;
774 	struct sk_buff *frag = NULL;
775 	struct sk_buff *head = NULL;
776 	int hdr_len;
777 
778 	/* Copy header if single buffer */
779 	if (skb_queue_len(list) == 1) {
780 		skb = skb_peek(list);
781 		hdr_len = skb_headroom(skb) + msg_hdr_sz(buf_msg(skb));
782 		_skb = __pskb_copy(skb, hdr_len, GFP_ATOMIC);
783 		if (!_skb)
784 			return false;
785 		__skb_queue_tail(rcvq, _skb);
786 		return true;
787 	}
788 
789 	/* Clone all fragments and reassemble */
790 	skb_queue_walk(list, skb) {
791 		frag = skb_clone(skb, GFP_ATOMIC);
792 		if (!frag)
793 			goto error;
794 		frag->next = NULL;
795 		if (tipc_buf_append(&head, &frag))
796 			break;
797 		if (!head)
798 			goto error;
799 	}
800 	__skb_queue_tail(rcvq, frag);
801 	return true;
802 error:
803 	pr_warn("Failed do clone local mcast rcv buffer\n");
804 	kfree_skb(head);
805 	return false;
806 }
807 
808 bool tipc_msg_pskb_copy(u32 dst, struct sk_buff_head *msg,
809 			struct sk_buff_head *cpy)
810 {
811 	struct sk_buff *skb, *_skb;
812 
813 	skb_queue_walk(msg, skb) {
814 		_skb = pskb_copy(skb, GFP_ATOMIC);
815 		if (!_skb) {
816 			__skb_queue_purge(cpy);
817 			return false;
818 		}
819 		msg_set_destnode(buf_msg(_skb), dst);
820 		__skb_queue_tail(cpy, _skb);
821 	}
822 	return true;
823 }
824 
825 /* tipc_skb_queue_sorted(); sort pkt into list according to sequence number
826  * @list: list to be appended to
827  * @seqno: sequence number of buffer to add
828  * @skb: buffer to add
829  */
830 bool __tipc_skb_queue_sorted(struct sk_buff_head *list, u16 seqno,
831 			     struct sk_buff *skb)
832 {
833 	struct sk_buff *_skb, *tmp;
834 
835 	if (skb_queue_empty(list) || less(seqno, buf_seqno(skb_peek(list)))) {
836 		__skb_queue_head(list, skb);
837 		return true;
838 	}
839 
840 	if (more(seqno, buf_seqno(skb_peek_tail(list)))) {
841 		__skb_queue_tail(list, skb);
842 		return true;
843 	}
844 
845 	skb_queue_walk_safe(list, _skb, tmp) {
846 		if (more(seqno, buf_seqno(_skb)))
847 			continue;
848 		if (seqno == buf_seqno(_skb))
849 			break;
850 		__skb_queue_before(list, _skb, skb);
851 		return true;
852 	}
853 	kfree_skb(skb);
854 	return false;
855 }
856 
857 void tipc_skb_reject(struct net *net, int err, struct sk_buff *skb,
858 		     struct sk_buff_head *xmitq)
859 {
860 	if (tipc_msg_reverse(tipc_own_addr(net), &skb, err))
861 		__skb_queue_tail(xmitq, skb);
862 }
863