xref: /linux/tools/testing/selftests/net/psock_tpacket.c (revision 31881d74b6dd1a6c530cff61248def4f2da38bee)
1 /*
2  * Copyright 2013 Red Hat, Inc.
3  * Author: Daniel Borkmann <dborkman@redhat.com>
4  *
5  * A basic test of packet socket's TPACKET_V1/TPACKET_V2/TPACKET_V3 behavior.
6  *
7  * Control:
8  *   Test the setup of the TPACKET socket with different patterns that are
9  *   known to fail (TODO) resp. succeed (OK).
10  *
11  * Datapath:
12  *   Open a pair of packet sockets and send resp. receive an a priori known
13  *   packet pattern accross the sockets and check if it was received resp.
14  *   sent correctly. Fanout in combination with RX_RING is currently not
15  *   tested here.
16  *
17  *   The test currently runs for
18  *   - TPACKET_V1: RX_RING, TX_RING
19  *   - TPACKET_V2: RX_RING, TX_RING
20  *   - TPACKET_V3: RX_RING
21  *
22  * License (GPLv2):
23  *
24  * This program is free software; you can redistribute it and/or modify it
25  * under the terms and conditions of the GNU General Public License,
26  * version 2, as published by the Free Software Foundation.
27  *
28  * This program is distributed in the hope it will be useful, but WITHOUT
29  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
30  * FITNESS FOR A PARTICULAR PURPOSE. * See the GNU General Public License for
31  * more details.
32  *
33  * You should have received a copy of the GNU General Public License along with
34  * this program; if not, write to the Free Software Foundation, Inc.,
35  * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
36  */
37 
38 #include <stdio.h>
39 #include <stdlib.h>
40 #include <sys/types.h>
41 #include <sys/stat.h>
42 #include <sys/socket.h>
43 #include <sys/mman.h>
44 #include <linux/if_packet.h>
45 #include <linux/filter.h>
46 #include <ctype.h>
47 #include <fcntl.h>
48 #include <unistd.h>
49 #include <bits/wordsize.h>
50 #include <net/ethernet.h>
51 #include <netinet/ip.h>
52 #include <arpa/inet.h>
53 #include <stdint.h>
54 #include <string.h>
55 #include <assert.h>
56 #include <net/if.h>
57 #include <inttypes.h>
58 #include <poll.h>
59 
60 #include "psock_lib.h"
61 
62 #ifndef bug_on
63 # define bug_on(cond)		assert(!(cond))
64 #endif
65 
66 #ifndef __aligned_tpacket
67 # define __aligned_tpacket	__attribute__((aligned(TPACKET_ALIGNMENT)))
68 #endif
69 
70 #ifndef __align_tpacket
71 # define __align_tpacket(x)	__attribute__((aligned(TPACKET_ALIGN(x))))
72 #endif
73 
74 #define BLOCK_STATUS(x)		((x)->h1.block_status)
75 #define BLOCK_NUM_PKTS(x)	((x)->h1.num_pkts)
76 #define BLOCK_O2FP(x)		((x)->h1.offset_to_first_pkt)
77 #define BLOCK_LEN(x)		((x)->h1.blk_len)
78 #define BLOCK_SNUM(x)		((x)->h1.seq_num)
79 #define BLOCK_O2PRIV(x)		((x)->offset_to_priv)
80 #define BLOCK_PRIV(x)		((void *) ((uint8_t *) (x) + BLOCK_O2PRIV(x)))
81 #define BLOCK_HDR_LEN		(ALIGN_8(sizeof(struct block_desc)))
82 #define ALIGN_8(x)		(((x) + 8 - 1) & ~(8 - 1))
83 #define BLOCK_PLUS_PRIV(sz_pri)	(BLOCK_HDR_LEN + ALIGN_8((sz_pri)))
84 
85 #define NUM_PACKETS		100
86 
87 struct ring {
88 	struct iovec *rd;
89 	uint8_t *mm_space;
90 	size_t mm_len, rd_len;
91 	struct sockaddr_ll ll;
92 	void (*walk)(int sock, struct ring *ring);
93 	int type, rd_num, flen, version;
94 	union {
95 		struct tpacket_req  req;
96 		struct tpacket_req3 req3;
97 	};
98 };
99 
100 struct block_desc {
101 	uint32_t version;
102 	uint32_t offset_to_priv;
103 	struct tpacket_hdr_v1 h1;
104 };
105 
106 union frame_map {
107 	struct {
108 		struct tpacket_hdr tp_h __aligned_tpacket;
109 		struct sockaddr_ll s_ll __align_tpacket(sizeof(struct tpacket_hdr));
110 	} *v1;
111 	struct {
112 		struct tpacket2_hdr tp_h __aligned_tpacket;
113 		struct sockaddr_ll s_ll __align_tpacket(sizeof(struct tpacket2_hdr));
114 	} *v2;
115 	void *raw;
116 };
117 
118 static unsigned int total_packets, total_bytes;
119 
120 static int pfsocket(int ver)
121 {
122 	int ret, sock = socket(PF_PACKET, SOCK_RAW, htons(ETH_P_ALL));
123 	if (sock == -1) {
124 		perror("socket");
125 		exit(1);
126 	}
127 
128 	ret = setsockopt(sock, SOL_PACKET, PACKET_VERSION, &ver, sizeof(ver));
129 	if (ret == -1) {
130 		perror("setsockopt");
131 		exit(1);
132 	}
133 
134 	return sock;
135 }
136 
137 static void status_bar_update(void)
138 {
139 	if (total_packets % 10 == 0) {
140 		fprintf(stderr, ".");
141 		fflush(stderr);
142 	}
143 }
144 
145 static void test_payload(void *pay, size_t len)
146 {
147 	struct ethhdr *eth = pay;
148 
149 	if (len < sizeof(struct ethhdr)) {
150 		fprintf(stderr, "test_payload: packet too "
151 			"small: %zu bytes!\n", len);
152 		exit(1);
153 	}
154 
155 	if (eth->h_proto != htons(ETH_P_IP)) {
156 		fprintf(stderr, "test_payload: wrong ethernet "
157 			"type: 0x%x!\n", ntohs(eth->h_proto));
158 		exit(1);
159 	}
160 }
161 
162 static void create_payload(void *pay, size_t *len)
163 {
164 	int i;
165 	struct ethhdr *eth = pay;
166 	struct iphdr *ip = pay + sizeof(*eth);
167 
168 	/* Lets create some broken crap, that still passes
169 	 * our BPF filter.
170 	 */
171 
172 	*len = DATA_LEN + 42;
173 
174 	memset(pay, 0xff, ETH_ALEN * 2);
175 	eth->h_proto = htons(ETH_P_IP);
176 
177 	for (i = 0; i < sizeof(*ip); ++i)
178 		((uint8_t *) pay)[i + sizeof(*eth)] = (uint8_t) rand();
179 
180 	ip->ihl = 5;
181 	ip->version = 4;
182 	ip->protocol = 0x11;
183 	ip->frag_off = 0;
184 	ip->ttl = 64;
185 	ip->tot_len = htons((uint16_t) *len - sizeof(*eth));
186 
187 	ip->saddr = htonl(INADDR_LOOPBACK);
188 	ip->daddr = htonl(INADDR_LOOPBACK);
189 
190 	memset(pay + sizeof(*eth) + sizeof(*ip),
191 	       DATA_CHAR, DATA_LEN);
192 }
193 
194 static inline int __v1_rx_kernel_ready(struct tpacket_hdr *hdr)
195 {
196 	return ((hdr->tp_status & TP_STATUS_USER) == TP_STATUS_USER);
197 }
198 
199 static inline void __v1_rx_user_ready(struct tpacket_hdr *hdr)
200 {
201 	hdr->tp_status = TP_STATUS_KERNEL;
202 	__sync_synchronize();
203 }
204 
205 static inline int __v2_rx_kernel_ready(struct tpacket2_hdr *hdr)
206 {
207 	return ((hdr->tp_status & TP_STATUS_USER) == TP_STATUS_USER);
208 }
209 
210 static inline void __v2_rx_user_ready(struct tpacket2_hdr *hdr)
211 {
212 	hdr->tp_status = TP_STATUS_KERNEL;
213 	__sync_synchronize();
214 }
215 
216 static inline int __v1_v2_rx_kernel_ready(void *base, int version)
217 {
218 	switch (version) {
219 	case TPACKET_V1:
220 		return __v1_rx_kernel_ready(base);
221 	case TPACKET_V2:
222 		return __v2_rx_kernel_ready(base);
223 	default:
224 		bug_on(1);
225 		return 0;
226 	}
227 }
228 
229 static inline void __v1_v2_rx_user_ready(void *base, int version)
230 {
231 	switch (version) {
232 	case TPACKET_V1:
233 		__v1_rx_user_ready(base);
234 		break;
235 	case TPACKET_V2:
236 		__v2_rx_user_ready(base);
237 		break;
238 	}
239 }
240 
241 static void walk_v1_v2_rx(int sock, struct ring *ring)
242 {
243 	struct pollfd pfd;
244 	int udp_sock[2];
245 	union frame_map ppd;
246 	unsigned int frame_num = 0;
247 
248 	bug_on(ring->type != PACKET_RX_RING);
249 
250 	pair_udp_open(udp_sock, PORT_BASE);
251 	pair_udp_setfilter(sock);
252 
253 	memset(&pfd, 0, sizeof(pfd));
254 	pfd.fd = sock;
255 	pfd.events = POLLIN | POLLERR;
256 	pfd.revents = 0;
257 
258 	pair_udp_send(udp_sock, NUM_PACKETS);
259 
260 	while (total_packets < NUM_PACKETS * 2) {
261 		while (__v1_v2_rx_kernel_ready(ring->rd[frame_num].iov_base,
262 					       ring->version)) {
263 			ppd.raw = ring->rd[frame_num].iov_base;
264 
265 			switch (ring->version) {
266 			case TPACKET_V1:
267 				test_payload((uint8_t *) ppd.raw + ppd.v1->tp_h.tp_mac,
268 					     ppd.v1->tp_h.tp_snaplen);
269 				total_bytes += ppd.v1->tp_h.tp_snaplen;
270 				break;
271 
272 			case TPACKET_V2:
273 				test_payload((uint8_t *) ppd.raw + ppd.v2->tp_h.tp_mac,
274 					     ppd.v2->tp_h.tp_snaplen);
275 				total_bytes += ppd.v2->tp_h.tp_snaplen;
276 				break;
277 			}
278 
279 			status_bar_update();
280 			total_packets++;
281 
282 			__v1_v2_rx_user_ready(ppd.raw, ring->version);
283 
284 			frame_num = (frame_num + 1) % ring->rd_num;
285 		}
286 
287 		poll(&pfd, 1, 1);
288 	}
289 
290 	pair_udp_close(udp_sock);
291 
292 	if (total_packets != 2 * NUM_PACKETS) {
293 		fprintf(stderr, "walk_v%d_rx: received %u out of %u pkts\n",
294 			ring->version, total_packets, NUM_PACKETS);
295 		exit(1);
296 	}
297 
298 	fprintf(stderr, " %u pkts (%u bytes)", NUM_PACKETS, total_bytes >> 1);
299 }
300 
301 static inline int __v1_tx_kernel_ready(struct tpacket_hdr *hdr)
302 {
303 	return !(hdr->tp_status & (TP_STATUS_SEND_REQUEST | TP_STATUS_SENDING));
304 }
305 
306 static inline void __v1_tx_user_ready(struct tpacket_hdr *hdr)
307 {
308 	hdr->tp_status = TP_STATUS_SEND_REQUEST;
309 	__sync_synchronize();
310 }
311 
312 static inline int __v2_tx_kernel_ready(struct tpacket2_hdr *hdr)
313 {
314 	return !(hdr->tp_status & (TP_STATUS_SEND_REQUEST | TP_STATUS_SENDING));
315 }
316 
317 static inline void __v2_tx_user_ready(struct tpacket2_hdr *hdr)
318 {
319 	hdr->tp_status = TP_STATUS_SEND_REQUEST;
320 	__sync_synchronize();
321 }
322 
323 static inline int __v1_v2_tx_kernel_ready(void *base, int version)
324 {
325 	switch (version) {
326 	case TPACKET_V1:
327 		return __v1_tx_kernel_ready(base);
328 	case TPACKET_V2:
329 		return __v2_tx_kernel_ready(base);
330 	default:
331 		bug_on(1);
332 		return 0;
333 	}
334 }
335 
336 static inline void __v1_v2_tx_user_ready(void *base, int version)
337 {
338 	switch (version) {
339 	case TPACKET_V1:
340 		__v1_tx_user_ready(base);
341 		break;
342 	case TPACKET_V2:
343 		__v2_tx_user_ready(base);
344 		break;
345 	}
346 }
347 
348 static void __v1_v2_set_packet_loss_discard(int sock)
349 {
350 	int ret, discard = 1;
351 
352 	ret = setsockopt(sock, SOL_PACKET, PACKET_LOSS, (void *) &discard,
353 			 sizeof(discard));
354 	if (ret == -1) {
355 		perror("setsockopt");
356 		exit(1);
357 	}
358 }
359 
360 static void walk_v1_v2_tx(int sock, struct ring *ring)
361 {
362 	struct pollfd pfd;
363 	int rcv_sock, ret;
364 	size_t packet_len;
365 	union frame_map ppd;
366 	char packet[1024];
367 	unsigned int frame_num = 0, got = 0;
368 	struct sockaddr_ll ll = {
369 		.sll_family = PF_PACKET,
370 		.sll_halen = ETH_ALEN,
371 	};
372 
373 	bug_on(ring->type != PACKET_TX_RING);
374 	bug_on(ring->rd_num < NUM_PACKETS);
375 
376 	rcv_sock = socket(PF_PACKET, SOCK_RAW, htons(ETH_P_ALL));
377 	if (rcv_sock == -1) {
378 		perror("socket");
379 		exit(1);
380 	}
381 
382 	pair_udp_setfilter(rcv_sock);
383 
384 	ll.sll_ifindex = if_nametoindex("lo");
385 	ret = bind(rcv_sock, (struct sockaddr *) &ll, sizeof(ll));
386 	if (ret == -1) {
387 		perror("bind");
388 		exit(1);
389 	}
390 
391 	memset(&pfd, 0, sizeof(pfd));
392 	pfd.fd = sock;
393 	pfd.events = POLLOUT | POLLERR;
394 	pfd.revents = 0;
395 
396 	total_packets = NUM_PACKETS;
397 	create_payload(packet, &packet_len);
398 
399 	while (total_packets > 0) {
400 		while (__v1_v2_tx_kernel_ready(ring->rd[frame_num].iov_base,
401 					       ring->version) &&
402 		       total_packets > 0) {
403 			ppd.raw = ring->rd[frame_num].iov_base;
404 
405 			switch (ring->version) {
406 			case TPACKET_V1:
407 				ppd.v1->tp_h.tp_snaplen = packet_len;
408 				ppd.v1->tp_h.tp_len = packet_len;
409 
410 				memcpy((uint8_t *) ppd.raw + TPACKET_HDRLEN -
411 				       sizeof(struct sockaddr_ll), packet,
412 				       packet_len);
413 				total_bytes += ppd.v1->tp_h.tp_snaplen;
414 				break;
415 
416 			case TPACKET_V2:
417 				ppd.v2->tp_h.tp_snaplen = packet_len;
418 				ppd.v2->tp_h.tp_len = packet_len;
419 
420 				memcpy((uint8_t *) ppd.raw + TPACKET2_HDRLEN -
421 				       sizeof(struct sockaddr_ll), packet,
422 				       packet_len);
423 				total_bytes += ppd.v2->tp_h.tp_snaplen;
424 				break;
425 			}
426 
427 			status_bar_update();
428 			total_packets--;
429 
430 			__v1_v2_tx_user_ready(ppd.raw, ring->version);
431 
432 			frame_num = (frame_num + 1) % ring->rd_num;
433 		}
434 
435 		poll(&pfd, 1, 1);
436 	}
437 
438 	bug_on(total_packets != 0);
439 
440 	ret = sendto(sock, NULL, 0, 0, NULL, 0);
441 	if (ret == -1) {
442 		perror("sendto");
443 		exit(1);
444 	}
445 
446 	while ((ret = recvfrom(rcv_sock, packet, sizeof(packet),
447 			       0, NULL, NULL)) > 0 &&
448 	       total_packets < NUM_PACKETS) {
449 		got += ret;
450 		test_payload(packet, ret);
451 
452 		status_bar_update();
453 		total_packets++;
454 	}
455 
456 	close(rcv_sock);
457 
458 	if (total_packets != NUM_PACKETS) {
459 		fprintf(stderr, "walk_v%d_rx: received %u out of %u pkts\n",
460 			ring->version, total_packets, NUM_PACKETS);
461 		exit(1);
462 	}
463 
464 	fprintf(stderr, " %u pkts (%u bytes)", NUM_PACKETS, got);
465 }
466 
467 static void walk_v1_v2(int sock, struct ring *ring)
468 {
469 	if (ring->type == PACKET_RX_RING)
470 		walk_v1_v2_rx(sock, ring);
471 	else
472 		walk_v1_v2_tx(sock, ring);
473 }
474 
475 static uint64_t __v3_prev_block_seq_num = 0;
476 
477 void __v3_test_block_seq_num(struct block_desc *pbd)
478 {
479 	if (__v3_prev_block_seq_num + 1 != BLOCK_SNUM(pbd)) {
480 		fprintf(stderr, "\nprev_block_seq_num:%"PRIu64", expected "
481 			"seq:%"PRIu64" != actual seq:%"PRIu64"\n",
482 			__v3_prev_block_seq_num, __v3_prev_block_seq_num + 1,
483 			(uint64_t) BLOCK_SNUM(pbd));
484 		exit(1);
485 	}
486 
487 	__v3_prev_block_seq_num = BLOCK_SNUM(pbd);
488 }
489 
490 static void __v3_test_block_len(struct block_desc *pbd, uint32_t bytes, int block_num)
491 {
492 	if (BLOCK_NUM_PKTS(pbd)) {
493 		if (bytes != BLOCK_LEN(pbd)) {
494 			fprintf(stderr, "\nblock:%u with %upackets, expected "
495 				"len:%u != actual len:%u\n", block_num,
496 				BLOCK_NUM_PKTS(pbd), bytes, BLOCK_LEN(pbd));
497 			exit(1);
498 		}
499 	} else {
500 		if (BLOCK_LEN(pbd) != BLOCK_PLUS_PRIV(13)) {
501 			fprintf(stderr, "\nblock:%u, expected len:%lu != "
502 				"actual len:%u\n", block_num, BLOCK_HDR_LEN,
503 				BLOCK_LEN(pbd));
504 			exit(1);
505 		}
506 	}
507 }
508 
509 static void __v3_test_block_header(struct block_desc *pbd, const int block_num)
510 {
511 	uint32_t block_status = BLOCK_STATUS(pbd);
512 
513 	if ((block_status & TP_STATUS_USER) == 0) {
514 		fprintf(stderr, "\nblock %u: not in TP_STATUS_USER\n", block_num);
515 		exit(1);
516 	}
517 
518 	__v3_test_block_seq_num(pbd);
519 }
520 
521 static void __v3_walk_block(struct block_desc *pbd, const int block_num)
522 {
523 	int num_pkts = BLOCK_NUM_PKTS(pbd), i;
524 	unsigned long bytes = 0;
525 	unsigned long bytes_with_padding = BLOCK_PLUS_PRIV(13);
526 	struct tpacket3_hdr *ppd;
527 
528 	__v3_test_block_header(pbd, block_num);
529 
530 	ppd = (struct tpacket3_hdr *) ((uint8_t *) pbd + BLOCK_O2FP(pbd));
531 	for (i = 0; i < num_pkts; ++i) {
532 		bytes += ppd->tp_snaplen;
533 
534 		if (ppd->tp_next_offset)
535 			bytes_with_padding += ppd->tp_next_offset;
536 		else
537 			bytes_with_padding += ALIGN_8(ppd->tp_snaplen + ppd->tp_mac);
538 
539 		test_payload((uint8_t *) ppd + ppd->tp_mac, ppd->tp_snaplen);
540 
541 		status_bar_update();
542 		total_packets++;
543 
544 		ppd = (struct tpacket3_hdr *) ((uint8_t *) ppd + ppd->tp_next_offset);
545 		__sync_synchronize();
546 	}
547 
548 	__v3_test_block_len(pbd, bytes_with_padding, block_num);
549 	total_bytes += bytes;
550 }
551 
552 void __v3_flush_block(struct block_desc *pbd)
553 {
554 	BLOCK_STATUS(pbd) = TP_STATUS_KERNEL;
555 	__sync_synchronize();
556 }
557 
558 static void walk_v3_rx(int sock, struct ring *ring)
559 {
560 	unsigned int block_num = 0;
561 	struct pollfd pfd;
562 	struct block_desc *pbd;
563 	int udp_sock[2];
564 
565 	bug_on(ring->type != PACKET_RX_RING);
566 
567 	pair_udp_open(udp_sock, PORT_BASE);
568 	pair_udp_setfilter(sock);
569 
570 	memset(&pfd, 0, sizeof(pfd));
571 	pfd.fd = sock;
572 	pfd.events = POLLIN | POLLERR;
573 	pfd.revents = 0;
574 
575 	pair_udp_send(udp_sock, NUM_PACKETS);
576 
577 	while (total_packets < NUM_PACKETS * 2) {
578 		pbd = (struct block_desc *) ring->rd[block_num].iov_base;
579 
580 		while ((BLOCK_STATUS(pbd) & TP_STATUS_USER) == 0)
581 			poll(&pfd, 1, 1);
582 
583 		__v3_walk_block(pbd, block_num);
584 		__v3_flush_block(pbd);
585 
586 		block_num = (block_num + 1) % ring->rd_num;
587 	}
588 
589 	pair_udp_close(udp_sock);
590 
591 	if (total_packets != 2 * NUM_PACKETS) {
592 		fprintf(stderr, "walk_v3_rx: received %u out of %u pkts\n",
593 			total_packets, NUM_PACKETS);
594 		exit(1);
595 	}
596 
597 	fprintf(stderr, " %u pkts (%u bytes)", NUM_PACKETS, total_bytes >> 1);
598 }
599 
600 static void walk_v3(int sock, struct ring *ring)
601 {
602 	if (ring->type == PACKET_RX_RING)
603 		walk_v3_rx(sock, ring);
604 	else
605 		bug_on(1);
606 }
607 
608 static void __v1_v2_fill(struct ring *ring, unsigned int blocks)
609 {
610 	ring->req.tp_block_size = getpagesize() << 2;
611 	ring->req.tp_frame_size = TPACKET_ALIGNMENT << 7;
612 	ring->req.tp_block_nr = blocks;
613 
614 	ring->req.tp_frame_nr = ring->req.tp_block_size /
615 				ring->req.tp_frame_size *
616 				ring->req.tp_block_nr;
617 
618 	ring->mm_len = ring->req.tp_block_size * ring->req.tp_block_nr;
619 	ring->walk = walk_v1_v2;
620 	ring->rd_num = ring->req.tp_frame_nr;
621 	ring->flen = ring->req.tp_frame_size;
622 }
623 
624 static void __v3_fill(struct ring *ring, unsigned int blocks)
625 {
626 	ring->req3.tp_retire_blk_tov = 64;
627 	ring->req3.tp_sizeof_priv = 13;
628 	ring->req3.tp_feature_req_word |= TP_FT_REQ_FILL_RXHASH;
629 
630 	ring->req3.tp_block_size = getpagesize() << 2;
631 	ring->req3.tp_frame_size = TPACKET_ALIGNMENT << 7;
632 	ring->req3.tp_block_nr = blocks;
633 
634 	ring->req3.tp_frame_nr = ring->req3.tp_block_size /
635 				 ring->req3.tp_frame_size *
636 				 ring->req3.tp_block_nr;
637 
638 	ring->mm_len = ring->req3.tp_block_size * ring->req3.tp_block_nr;
639 	ring->walk = walk_v3;
640 	ring->rd_num = ring->req3.tp_block_nr;
641 	ring->flen = ring->req3.tp_block_size;
642 }
643 
644 static void setup_ring(int sock, struct ring *ring, int version, int type)
645 {
646 	int ret = 0;
647 	unsigned int blocks = 256;
648 
649 	ring->type = type;
650 	ring->version = version;
651 
652 	switch (version) {
653 	case TPACKET_V1:
654 	case TPACKET_V2:
655 		if (type == PACKET_TX_RING)
656 			__v1_v2_set_packet_loss_discard(sock);
657 		__v1_v2_fill(ring, blocks);
658 		ret = setsockopt(sock, SOL_PACKET, type, &ring->req,
659 				 sizeof(ring->req));
660 		break;
661 
662 	case TPACKET_V3:
663 		__v3_fill(ring, blocks);
664 		ret = setsockopt(sock, SOL_PACKET, type, &ring->req3,
665 				 sizeof(ring->req3));
666 		break;
667 	}
668 
669 	if (ret == -1) {
670 		perror("setsockopt");
671 		exit(1);
672 	}
673 
674 	ring->rd_len = ring->rd_num * sizeof(*ring->rd);
675 	ring->rd = malloc(ring->rd_len);
676 	if (ring->rd == NULL) {
677 		perror("malloc");
678 		exit(1);
679 	}
680 
681 	total_packets = 0;
682 	total_bytes = 0;
683 }
684 
685 static void mmap_ring(int sock, struct ring *ring)
686 {
687 	int i;
688 
689 	ring->mm_space = mmap(0, ring->mm_len, PROT_READ | PROT_WRITE,
690 			      MAP_SHARED | MAP_LOCKED | MAP_POPULATE, sock, 0);
691 	if (ring->mm_space == MAP_FAILED) {
692 		perror("mmap");
693 		exit(1);
694 	}
695 
696 	memset(ring->rd, 0, ring->rd_len);
697 	for (i = 0; i < ring->rd_num; ++i) {
698 		ring->rd[i].iov_base = ring->mm_space + (i * ring->flen);
699 		ring->rd[i].iov_len = ring->flen;
700 	}
701 }
702 
703 static void bind_ring(int sock, struct ring *ring)
704 {
705 	int ret;
706 
707 	ring->ll.sll_family = PF_PACKET;
708 	ring->ll.sll_protocol = htons(ETH_P_ALL);
709 	ring->ll.sll_ifindex = if_nametoindex("lo");
710 	ring->ll.sll_hatype = 0;
711 	ring->ll.sll_pkttype = 0;
712 	ring->ll.sll_halen = 0;
713 
714 	ret = bind(sock, (struct sockaddr *) &ring->ll, sizeof(ring->ll));
715 	if (ret == -1) {
716 		perror("bind");
717 		exit(1);
718 	}
719 }
720 
721 static void walk_ring(int sock, struct ring *ring)
722 {
723 	ring->walk(sock, ring);
724 }
725 
726 static void unmap_ring(int sock, struct ring *ring)
727 {
728 	munmap(ring->mm_space, ring->mm_len);
729 	free(ring->rd);
730 }
731 
732 static int test_kernel_bit_width(void)
733 {
734 	char in[512], *ptr;
735 	int num = 0, fd;
736 	ssize_t ret;
737 
738 	fd = open("/proc/kallsyms", O_RDONLY);
739 	if (fd == -1) {
740 		perror("open");
741 		exit(1);
742 	}
743 
744 	ret = read(fd, in, sizeof(in));
745 	if (ret <= 0) {
746 		perror("read");
747 		exit(1);
748 	}
749 
750 	close(fd);
751 
752 	ptr = in;
753 	while(!isspace(*ptr)) {
754 		num++;
755 		ptr++;
756 	}
757 
758 	return num * 4;
759 }
760 
761 static int test_user_bit_width(void)
762 {
763 	return __WORDSIZE;
764 }
765 
766 static const char *tpacket_str[] = {
767 	[TPACKET_V1] = "TPACKET_V1",
768 	[TPACKET_V2] = "TPACKET_V2",
769 	[TPACKET_V3] = "TPACKET_V3",
770 };
771 
772 static const char *type_str[] = {
773 	[PACKET_RX_RING] = "PACKET_RX_RING",
774 	[PACKET_TX_RING] = "PACKET_TX_RING",
775 };
776 
777 static int test_tpacket(int version, int type)
778 {
779 	int sock;
780 	struct ring ring;
781 
782 	fprintf(stderr, "test: %s with %s ", tpacket_str[version],
783 		type_str[type]);
784 	fflush(stderr);
785 
786 	if (version == TPACKET_V1 &&
787 	    test_kernel_bit_width() != test_user_bit_width()) {
788 		fprintf(stderr, "test: skip %s %s since user and kernel "
789 			"space have different bit width\n",
790 			tpacket_str[version], type_str[type]);
791 		return 0;
792 	}
793 
794 	sock = pfsocket(version);
795 	memset(&ring, 0, sizeof(ring));
796 	setup_ring(sock, &ring, version, type);
797 	mmap_ring(sock, &ring);
798 	bind_ring(sock, &ring);
799 	walk_ring(sock, &ring);
800 	unmap_ring(sock, &ring);
801 	close(sock);
802 
803 	fprintf(stderr, "\n");
804 	return 0;
805 }
806 
807 int main(void)
808 {
809 	int ret = 0;
810 
811 	ret |= test_tpacket(TPACKET_V1, PACKET_RX_RING);
812 	ret |= test_tpacket(TPACKET_V1, PACKET_TX_RING);
813 
814 	ret |= test_tpacket(TPACKET_V2, PACKET_RX_RING);
815 	ret |= test_tpacket(TPACKET_V2, PACKET_TX_RING);
816 
817 	ret |= test_tpacket(TPACKET_V3, PACKET_RX_RING);
818 
819 	if (ret)
820 		return 1;
821 
822 	printf("OK. All tests passed\n");
823 	return 0;
824 }
825