• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright 2013 Red Hat, Inc.
3  * Author: Daniel Borkmann <dborkman@redhat.com>
4  *         Chetan Loke <loke.chetan@gmail.com> (TPACKET_V3 usage example)
5  *
6  * A basic test of packet socket's TPACKET_V1/TPACKET_V2/TPACKET_V3 behavior.
7  *
8  * Control:
9  *   Test the setup of the TPACKET socket with different patterns that are
10  *   known to fail (TODO) resp. succeed (OK).
11  *
12  * Datapath:
13  *   Open a pair of packet sockets and send resp. receive an a priori known
14  *   packet pattern accross the sockets and check if it was received resp.
15  *   sent correctly. Fanout in combination with RX_RING is currently not
16  *   tested here.
17  *
18  *   The test currently runs for
19  *   - TPACKET_V1: RX_RING, TX_RING
20  *   - TPACKET_V2: RX_RING, TX_RING
21  *   - TPACKET_V3: RX_RING
22  *
23  * License (GPLv2):
24  *
25  * This program is free software; you can redistribute it and/or modify it
26  * under the terms and conditions of the GNU General Public License,
27  * version 2, as published by the Free Software Foundation.
28  *
29  * This program is distributed in the hope it will be useful, but WITHOUT
30  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
31  * FITNESS FOR A PARTICULAR PURPOSE. * See the GNU General Public License for
32  * more details.
33  *
34  * You should have received a copy of the GNU General Public License along with
35  * this program; if not, write to the Free Software Foundation, Inc.,
36  * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
37  */
38 
39 #include <stdio.h>
40 #include <stdlib.h>
41 #include <sys/types.h>
42 #include <sys/stat.h>
43 #include <sys/socket.h>
44 #include <sys/mman.h>
45 #include <sys/utsname.h>
46 #include <linux/if_packet.h>
47 #include <linux/filter.h>
48 #include <ctype.h>
49 #include <fcntl.h>
50 #include <unistd.h>
51 #ifndef __ANDROID__
52 #include <bits/wordsize.h>
53 #endif
54 #include <net/ethernet.h>
55 #include <netinet/ip.h>
56 #include <arpa/inet.h>
57 #include <stdint.h>
58 #include <string.h>
59 #include <assert.h>
60 #include <net/if.h>
61 #include <inttypes.h>
62 #include <poll.h>
63 
64 #include "psock_lib.h"
65 
66 #ifndef bug_on
67 # define bug_on(cond)		assert(!(cond))
68 #endif
69 
70 #ifndef __aligned_tpacket
71 # define __aligned_tpacket	__attribute__((aligned(TPACKET_ALIGNMENT)))
72 #endif
73 
74 #ifndef __align_tpacket
75 # define __align_tpacket(x)	__attribute__((aligned(TPACKET_ALIGN(x))))
76 #endif
77 
78 #define NUM_PACKETS		100
79 #define ALIGN_8(x)		(((x) + 8 - 1) & ~(8 - 1))
80 
81 struct ring {
82 	struct iovec *rd;
83 	uint8_t *mm_space;
84 	size_t mm_len, rd_len;
85 	struct sockaddr_ll ll;
86 	void (*walk)(int sock, struct ring *ring);
87 	int type, rd_num, flen, version;
88 	union {
89 		struct tpacket_req  req;
90 		struct tpacket_req3 req3;
91 	};
92 };
93 
94 struct block_desc {
95 	uint32_t version;
96 	uint32_t offset_to_priv;
97 	struct tpacket_hdr_v1 h1;
98 };
99 
100 union frame_map {
101 	struct {
102 		struct tpacket_hdr tp_h __aligned_tpacket;
103 		struct sockaddr_ll s_ll __align_tpacket(sizeof(struct tpacket_hdr));
104 	} *v1;
105 	struct {
106 		struct tpacket2_hdr tp_h __aligned_tpacket;
107 		struct sockaddr_ll s_ll __align_tpacket(sizeof(struct tpacket2_hdr));
108 	} *v2;
109 	void *raw;
110 };
111 
112 static unsigned int total_packets, total_bytes;
113 
pfsocket(int ver)114 static int pfsocket(int ver)
115 {
116 	int ret, sock = socket(PF_PACKET, SOCK_RAW, 0);
117 	if (sock == -1) {
118 		perror("socket");
119 		exit(1);
120 	}
121 
122 	ret = setsockopt(sock, SOL_PACKET, PACKET_VERSION, &ver, sizeof(ver));
123 	if (ret == -1) {
124 		perror("setsockopt");
125 		exit(1);
126 	}
127 
128 	return sock;
129 }
130 
status_bar_update(void)131 static void status_bar_update(void)
132 {
133 	if (total_packets % 10 == 0) {
134 		fprintf(stderr, ".");
135 		fflush(stderr);
136 	}
137 }
138 
test_payload(void * pay,size_t len)139 static void test_payload(void *pay, size_t len)
140 {
141 	struct ethhdr *eth = pay;
142 
143 	if (len < sizeof(struct ethhdr)) {
144 		fprintf(stderr, "test_payload: packet too "
145 			"small: %zu bytes!\n", len);
146 		exit(1);
147 	}
148 
149 	if (eth->h_proto != htons(ETH_P_IP)) {
150 		fprintf(stderr, "test_payload: wrong ethernet "
151 			"type: 0x%x!\n", ntohs(eth->h_proto));
152 		exit(1);
153 	}
154 }
155 
create_payload(void * pay,size_t * len)156 static void create_payload(void *pay, size_t *len)
157 {
158 	int i;
159 	struct ethhdr *eth = pay;
160 	struct iphdr *ip = pay + sizeof(*eth);
161 
162 	/* Lets create some broken crap, that still passes
163 	 * our BPF filter.
164 	 */
165 
166 	*len = DATA_LEN + 42;
167 
168 	memset(pay, 0xff, ETH_ALEN * 2);
169 	eth->h_proto = htons(ETH_P_IP);
170 
171 	for (i = 0; i < sizeof(*ip); ++i)
172 		((uint8_t *) pay)[i + sizeof(*eth)] = (uint8_t) rand();
173 
174 	ip->ihl = 5;
175 	ip->version = 4;
176 	ip->protocol = 0x11;
177 	ip->frag_off = 0;
178 	ip->ttl = 64;
179 	ip->tot_len = htons((uint16_t) *len - sizeof(*eth));
180 
181 	ip->saddr = htonl(INADDR_LOOPBACK);
182 	ip->daddr = htonl(INADDR_LOOPBACK);
183 
184 	memset(pay + sizeof(*eth) + sizeof(*ip),
185 	       DATA_CHAR, DATA_LEN);
186 }
187 
__v1_rx_kernel_ready(struct tpacket_hdr * hdr)188 static inline int __v1_rx_kernel_ready(struct tpacket_hdr *hdr)
189 {
190 	return ((hdr->tp_status & TP_STATUS_USER) == TP_STATUS_USER);
191 }
192 
__v1_rx_user_ready(struct tpacket_hdr * hdr)193 static inline void __v1_rx_user_ready(struct tpacket_hdr *hdr)
194 {
195 	hdr->tp_status = TP_STATUS_KERNEL;
196 	__sync_synchronize();
197 }
198 
__v2_rx_kernel_ready(struct tpacket2_hdr * hdr)199 static inline int __v2_rx_kernel_ready(struct tpacket2_hdr *hdr)
200 {
201 	return ((hdr->tp_status & TP_STATUS_USER) == TP_STATUS_USER);
202 }
203 
__v2_rx_user_ready(struct tpacket2_hdr * hdr)204 static inline void __v2_rx_user_ready(struct tpacket2_hdr *hdr)
205 {
206 	hdr->tp_status = TP_STATUS_KERNEL;
207 	__sync_synchronize();
208 }
209 
__v1_v2_rx_kernel_ready(void * base,int version)210 static inline int __v1_v2_rx_kernel_ready(void *base, int version)
211 {
212 	switch (version) {
213 	case TPACKET_V1:
214 		return __v1_rx_kernel_ready(base);
215 	case TPACKET_V2:
216 		return __v2_rx_kernel_ready(base);
217 	default:
218 		bug_on(1);
219 		return 0;
220 	}
221 }
222 
__v1_v2_rx_user_ready(void * base,int version)223 static inline void __v1_v2_rx_user_ready(void *base, int version)
224 {
225 	switch (version) {
226 	case TPACKET_V1:
227 		__v1_rx_user_ready(base);
228 		break;
229 	case TPACKET_V2:
230 		__v2_rx_user_ready(base);
231 		break;
232 	}
233 }
234 
walk_v1_v2_rx(int sock,struct ring * ring)235 static void walk_v1_v2_rx(int sock, struct ring *ring)
236 {
237 	struct pollfd pfd;
238 	int udp_sock[2];
239 	union frame_map ppd;
240 	unsigned int frame_num = 0;
241 
242 	bug_on(ring->type != PACKET_RX_RING);
243 
244 	pair_udp_open(udp_sock, PORT_BASE);
245 
246 	memset(&pfd, 0, sizeof(pfd));
247 	pfd.fd = sock;
248 	pfd.events = POLLIN | POLLERR;
249 	pfd.revents = 0;
250 
251 	pair_udp_send(udp_sock, NUM_PACKETS);
252 
253 	while (total_packets < NUM_PACKETS * 2) {
254 		while (__v1_v2_rx_kernel_ready(ring->rd[frame_num].iov_base,
255 					       ring->version)) {
256 			ppd.raw = ring->rd[frame_num].iov_base;
257 
258 			switch (ring->version) {
259 			case TPACKET_V1:
260 				test_payload((uint8_t *) ppd.raw + ppd.v1->tp_h.tp_mac,
261 					     ppd.v1->tp_h.tp_snaplen);
262 				total_bytes += ppd.v1->tp_h.tp_snaplen;
263 				break;
264 
265 			case TPACKET_V2:
266 				test_payload((uint8_t *) ppd.raw + ppd.v2->tp_h.tp_mac,
267 					     ppd.v2->tp_h.tp_snaplen);
268 				total_bytes += ppd.v2->tp_h.tp_snaplen;
269 				break;
270 			}
271 
272 			status_bar_update();
273 			total_packets++;
274 
275 			__v1_v2_rx_user_ready(ppd.raw, ring->version);
276 
277 			frame_num = (frame_num + 1) % ring->rd_num;
278 		}
279 
280 		poll(&pfd, 1, 1);
281 	}
282 
283 	pair_udp_close(udp_sock);
284 
285 	if (total_packets != 2 * NUM_PACKETS) {
286 		fprintf(stderr, "walk_v%d_rx: received %u out of %u pkts\n",
287 			ring->version, total_packets, NUM_PACKETS);
288 		exit(1);
289 	}
290 
291 	fprintf(stderr, " %u pkts (%u bytes)", NUM_PACKETS, total_bytes >> 1);
292 }
293 
__v1_tx_kernel_ready(struct tpacket_hdr * hdr)294 static inline int __v1_tx_kernel_ready(struct tpacket_hdr *hdr)
295 {
296 	return !(hdr->tp_status & (TP_STATUS_SEND_REQUEST | TP_STATUS_SENDING));
297 }
298 
__v1_tx_user_ready(struct tpacket_hdr * hdr)299 static inline void __v1_tx_user_ready(struct tpacket_hdr *hdr)
300 {
301 	hdr->tp_status = TP_STATUS_SEND_REQUEST;
302 	__sync_synchronize();
303 }
304 
__v2_tx_kernel_ready(struct tpacket2_hdr * hdr)305 static inline int __v2_tx_kernel_ready(struct tpacket2_hdr *hdr)
306 {
307 	return !(hdr->tp_status & (TP_STATUS_SEND_REQUEST | TP_STATUS_SENDING));
308 }
309 
__v2_tx_user_ready(struct tpacket2_hdr * hdr)310 static inline void __v2_tx_user_ready(struct tpacket2_hdr *hdr)
311 {
312 	hdr->tp_status = TP_STATUS_SEND_REQUEST;
313 	__sync_synchronize();
314 }
315 
__v3_tx_kernel_ready(struct tpacket3_hdr * hdr)316 static inline int __v3_tx_kernel_ready(struct tpacket3_hdr *hdr)
317 {
318 	return !(hdr->tp_status & (TP_STATUS_SEND_REQUEST | TP_STATUS_SENDING));
319 }
320 
__v3_tx_user_ready(struct tpacket3_hdr * hdr)321 static inline void __v3_tx_user_ready(struct tpacket3_hdr *hdr)
322 {
323 	hdr->tp_status = TP_STATUS_SEND_REQUEST;
324 	__sync_synchronize();
325 }
326 
__tx_kernel_ready(void * base,int version)327 static inline int __tx_kernel_ready(void *base, int version)
328 {
329 	switch (version) {
330 	case TPACKET_V1:
331 		return __v1_tx_kernel_ready(base);
332 	case TPACKET_V2:
333 		return __v2_tx_kernel_ready(base);
334 	case TPACKET_V3:
335 		return __v3_tx_kernel_ready(base);
336 	default:
337 		bug_on(1);
338 		return 0;
339 	}
340 }
341 
__tx_user_ready(void * base,int version)342 static inline void __tx_user_ready(void *base, int version)
343 {
344 	switch (version) {
345 	case TPACKET_V1:
346 		__v1_tx_user_ready(base);
347 		break;
348 	case TPACKET_V2:
349 		__v2_tx_user_ready(base);
350 		break;
351 	case TPACKET_V3:
352 		__v3_tx_user_ready(base);
353 		break;
354 	}
355 }
356 
__v1_v2_set_packet_loss_discard(int sock)357 static void __v1_v2_set_packet_loss_discard(int sock)
358 {
359 	int ret, discard = 1;
360 
361 	ret = setsockopt(sock, SOL_PACKET, PACKET_LOSS, (void *) &discard,
362 			 sizeof(discard));
363 	if (ret == -1) {
364 		perror("setsockopt");
365 		exit(1);
366 	}
367 }
368 
get_next_frame(struct ring * ring,int n)369 static inline void *get_next_frame(struct ring *ring, int n)
370 {
371 	uint8_t *f0 = ring->rd[0].iov_base;
372 
373 	switch (ring->version) {
374 	case TPACKET_V1:
375 	case TPACKET_V2:
376 		return ring->rd[n].iov_base;
377 	case TPACKET_V3:
378 		return f0 + (n * ring->req3.tp_frame_size);
379 	default:
380 		bug_on(1);
381 		return NULL;
382 	}
383 }
384 
walk_tx(int sock,struct ring * ring)385 static void walk_tx(int sock, struct ring *ring)
386 {
387 	struct pollfd pfd;
388 	int rcv_sock, ret;
389 	size_t packet_len;
390 	union frame_map ppd;
391 	char packet[1024];
392 	unsigned int frame_num = 0, got = 0;
393 	struct sockaddr_ll ll = {
394 		.sll_family = PF_PACKET,
395 		.sll_halen = ETH_ALEN,
396 	};
397 	int nframes;
398 
399 	/* TPACKET_V{1,2} sets up the ring->rd* related variables based
400 	 * on frames (e.g., rd_num is tp_frame_nr) whereas V3 sets these
401 	 * up based on blocks (e.g, rd_num is  tp_block_nr)
402 	 */
403 	if (ring->version <= TPACKET_V2)
404 		nframes = ring->rd_num;
405 	else
406 		nframes = ring->req3.tp_frame_nr;
407 
408 	bug_on(ring->type != PACKET_TX_RING);
409 	bug_on(nframes < NUM_PACKETS);
410 
411 	rcv_sock = socket(PF_PACKET, SOCK_RAW, htons(ETH_P_ALL));
412 	if (rcv_sock == -1) {
413 		perror("socket");
414 		exit(1);
415 	}
416 
417 	pair_udp_setfilter(rcv_sock);
418 
419 	ll.sll_ifindex = if_nametoindex("lo");
420 	ret = bind(rcv_sock, (struct sockaddr *) &ll, sizeof(ll));
421 	if (ret == -1) {
422 		perror("bind");
423 		exit(1);
424 	}
425 
426 	memset(&pfd, 0, sizeof(pfd));
427 	pfd.fd = sock;
428 	pfd.events = POLLOUT | POLLERR;
429 	pfd.revents = 0;
430 
431 	total_packets = NUM_PACKETS;
432 	create_payload(packet, &packet_len);
433 
434 	while (total_packets > 0) {
435 		void *next = get_next_frame(ring, frame_num);
436 
437 		while (__tx_kernel_ready(next, ring->version) &&
438 		       total_packets > 0) {
439 			ppd.raw = next;
440 
441 			switch (ring->version) {
442 			case TPACKET_V1:
443 				ppd.v1->tp_h.tp_snaplen = packet_len;
444 				ppd.v1->tp_h.tp_len = packet_len;
445 
446 				memcpy((uint8_t *) ppd.raw + TPACKET_HDRLEN -
447 				       sizeof(struct sockaddr_ll), packet,
448 				       packet_len);
449 				total_bytes += ppd.v1->tp_h.tp_snaplen;
450 				break;
451 
452 			case TPACKET_V2:
453 				ppd.v2->tp_h.tp_snaplen = packet_len;
454 				ppd.v2->tp_h.tp_len = packet_len;
455 
456 				memcpy((uint8_t *) ppd.raw + TPACKET2_HDRLEN -
457 				       sizeof(struct sockaddr_ll), packet,
458 				       packet_len);
459 				total_bytes += ppd.v2->tp_h.tp_snaplen;
460 				break;
461 			case TPACKET_V3: {
462 				struct tpacket3_hdr *tx = next;
463 
464 				tx->tp_snaplen = packet_len;
465 				tx->tp_len = packet_len;
466 				tx->tp_next_offset = 0;
467 
468 				memcpy((uint8_t *)tx + TPACKET3_HDRLEN -
469 				       sizeof(struct sockaddr_ll), packet,
470 				       packet_len);
471 				total_bytes += tx->tp_snaplen;
472 				break;
473 			}
474 			}
475 
476 			status_bar_update();
477 			total_packets--;
478 
479 			__tx_user_ready(next, ring->version);
480 
481 			frame_num = (frame_num + 1) % nframes;
482 		}
483 
484 		poll(&pfd, 1, 1);
485 	}
486 
487 	bug_on(total_packets != 0);
488 
489 	ret = sendto(sock, NULL, 0, 0, NULL, 0);
490 	if (ret == -1) {
491 		perror("sendto");
492 		exit(1);
493 	}
494 
495 	while ((ret = recvfrom(rcv_sock, packet, sizeof(packet),
496 			       0, NULL, NULL)) > 0 &&
497 	       total_packets < NUM_PACKETS) {
498 		got += ret;
499 		test_payload(packet, ret);
500 
501 		status_bar_update();
502 		total_packets++;
503 	}
504 
505 	close(rcv_sock);
506 
507 	if (total_packets != NUM_PACKETS) {
508 		fprintf(stderr, "walk_v%d_rx: received %u out of %u pkts\n",
509 			ring->version, total_packets, NUM_PACKETS);
510 		exit(1);
511 	}
512 
513 	fprintf(stderr, " %u pkts (%u bytes)", NUM_PACKETS, got);
514 }
515 
walk_v1_v2(int sock,struct ring * ring)516 static void walk_v1_v2(int sock, struct ring *ring)
517 {
518 	if (ring->type == PACKET_RX_RING)
519 		walk_v1_v2_rx(sock, ring);
520 	else
521 		walk_tx(sock, ring);
522 }
523 
524 static uint64_t __v3_prev_block_seq_num = 0;
525 
__v3_test_block_seq_num(struct block_desc * pbd)526 void __v3_test_block_seq_num(struct block_desc *pbd)
527 {
528 	if (__v3_prev_block_seq_num + 1 != pbd->h1.seq_num) {
529 		fprintf(stderr, "\nprev_block_seq_num:%"PRIu64", expected "
530 			"seq:%"PRIu64" != actual seq:%"PRIu64"\n",
531 			__v3_prev_block_seq_num, __v3_prev_block_seq_num + 1,
532 			(uint64_t) pbd->h1.seq_num);
533 		exit(1);
534 	}
535 
536 	__v3_prev_block_seq_num = pbd->h1.seq_num;
537 }
538 
__v3_test_block_len(struct block_desc * pbd,uint32_t bytes,int block_num)539 static void __v3_test_block_len(struct block_desc *pbd, uint32_t bytes, int block_num)
540 {
541 	if (pbd->h1.num_pkts && bytes != pbd->h1.blk_len) {
542 		fprintf(stderr, "\nblock:%u with %upackets, expected "
543 			"len:%u != actual len:%u\n", block_num,
544 			pbd->h1.num_pkts, bytes, pbd->h1.blk_len);
545 		exit(1);
546 	}
547 }
548 
__v3_test_block_header(struct block_desc * pbd,const int block_num)549 static void __v3_test_block_header(struct block_desc *pbd, const int block_num)
550 {
551 	if ((pbd->h1.block_status & TP_STATUS_USER) == 0) {
552 		fprintf(stderr, "\nblock %u: not in TP_STATUS_USER\n", block_num);
553 		exit(1);
554 	}
555 
556 	__v3_test_block_seq_num(pbd);
557 }
558 
__v3_walk_block(struct block_desc * pbd,const int block_num)559 static void __v3_walk_block(struct block_desc *pbd, const int block_num)
560 {
561 	int num_pkts = pbd->h1.num_pkts, i;
562 	unsigned long bytes = 0, bytes_with_padding = ALIGN_8(sizeof(*pbd));
563 	struct tpacket3_hdr *ppd;
564 
565 	__v3_test_block_header(pbd, block_num);
566 
567 	ppd = (struct tpacket3_hdr *) ((uint8_t *) pbd +
568 				       pbd->h1.offset_to_first_pkt);
569 
570 	for (i = 0; i < num_pkts; ++i) {
571 		bytes += ppd->tp_snaplen;
572 
573 		if (ppd->tp_next_offset)
574 			bytes_with_padding += ppd->tp_next_offset;
575 		else
576 			bytes_with_padding += ALIGN_8(ppd->tp_snaplen + ppd->tp_mac);
577 
578 		test_payload((uint8_t *) ppd + ppd->tp_mac, ppd->tp_snaplen);
579 
580 		status_bar_update();
581 		total_packets++;
582 
583 		ppd = (struct tpacket3_hdr *) ((uint8_t *) ppd + ppd->tp_next_offset);
584 		__sync_synchronize();
585 	}
586 
587 	__v3_test_block_len(pbd, bytes_with_padding, block_num);
588 	total_bytes += bytes;
589 }
590 
__v3_flush_block(struct block_desc * pbd)591 void __v3_flush_block(struct block_desc *pbd)
592 {
593 	pbd->h1.block_status = TP_STATUS_KERNEL;
594 	__sync_synchronize();
595 }
596 
walk_v3_rx(int sock,struct ring * ring)597 static void walk_v3_rx(int sock, struct ring *ring)
598 {
599 	unsigned int block_num = 0;
600 	struct pollfd pfd;
601 	struct block_desc *pbd;
602 	int udp_sock[2];
603 
604 	bug_on(ring->type != PACKET_RX_RING);
605 
606 	pair_udp_open(udp_sock, PORT_BASE);
607 
608 	memset(&pfd, 0, sizeof(pfd));
609 	pfd.fd = sock;
610 	pfd.events = POLLIN | POLLERR;
611 	pfd.revents = 0;
612 
613 	pair_udp_send(udp_sock, NUM_PACKETS);
614 
615 	while (total_packets < NUM_PACKETS * 2) {
616 		pbd = (struct block_desc *) ring->rd[block_num].iov_base;
617 
618 		while ((pbd->h1.block_status & TP_STATUS_USER) == 0)
619 			poll(&pfd, 1, 1);
620 
621 		__v3_walk_block(pbd, block_num);
622 		__v3_flush_block(pbd);
623 
624 		block_num = (block_num + 1) % ring->rd_num;
625 	}
626 
627 	pair_udp_close(udp_sock);
628 
629 	if (total_packets != 2 * NUM_PACKETS) {
630 		fprintf(stderr, "walk_v3_rx: received %u out of %u pkts\n",
631 			total_packets, NUM_PACKETS);
632 		exit(1);
633 	}
634 
635 	fprintf(stderr, " %u pkts (%u bytes)", NUM_PACKETS, total_bytes >> 1);
636 }
637 
walk_v3(int sock,struct ring * ring)638 static void walk_v3(int sock, struct ring *ring)
639 {
640 	if (ring->type == PACKET_RX_RING)
641 		walk_v3_rx(sock, ring);
642 	else
643 		walk_tx(sock, ring);
644 }
645 
__v1_v2_fill(struct ring * ring,unsigned int blocks)646 static void __v1_v2_fill(struct ring *ring, unsigned int blocks)
647 {
648 	ring->req.tp_block_size = getpagesize() << 2;
649 	ring->req.tp_frame_size = TPACKET_ALIGNMENT << 7;
650 	ring->req.tp_block_nr = blocks;
651 
652 	ring->req.tp_frame_nr = ring->req.tp_block_size /
653 				ring->req.tp_frame_size *
654 				ring->req.tp_block_nr;
655 
656 	ring->mm_len = ring->req.tp_block_size * ring->req.tp_block_nr;
657 	ring->walk = walk_v1_v2;
658 	ring->rd_num = ring->req.tp_frame_nr;
659 	ring->flen = ring->req.tp_frame_size;
660 }
661 
__v3_fill(struct ring * ring,unsigned int blocks,int type)662 static void __v3_fill(struct ring *ring, unsigned int blocks, int type)
663 {
664 	if (type == PACKET_RX_RING) {
665 		ring->req3.tp_retire_blk_tov = 64;
666 		ring->req3.tp_sizeof_priv = 0;
667 		ring->req3.tp_feature_req_word = TP_FT_REQ_FILL_RXHASH;
668 	}
669 	ring->req3.tp_block_size = getpagesize() << 2;
670 	ring->req3.tp_frame_size = TPACKET_ALIGNMENT << 7;
671 	ring->req3.tp_block_nr = blocks;
672 
673 	ring->req3.tp_frame_nr = ring->req3.tp_block_size /
674 				 ring->req3.tp_frame_size *
675 				 ring->req3.tp_block_nr;
676 
677 	ring->mm_len = ring->req3.tp_block_size * ring->req3.tp_block_nr;
678 	ring->walk = walk_v3;
679 	ring->rd_num = ring->req3.tp_block_nr;
680 	ring->flen = ring->req3.tp_block_size;
681 }
682 
setup_ring(int sock,struct ring * ring,int version,int type)683 static void setup_ring(int sock, struct ring *ring, int version, int type)
684 {
685 	int ret = 0;
686 	unsigned int blocks = 256;
687 
688 	ring->type = type;
689 	ring->version = version;
690 
691 	switch (version) {
692 	case TPACKET_V1:
693 	case TPACKET_V2:
694 		if (type == PACKET_TX_RING)
695 			__v1_v2_set_packet_loss_discard(sock);
696 		__v1_v2_fill(ring, blocks);
697 		ret = setsockopt(sock, SOL_PACKET, type, &ring->req,
698 				 sizeof(ring->req));
699 		break;
700 
701 	case TPACKET_V3:
702 		__v3_fill(ring, blocks, type);
703 		ret = setsockopt(sock, SOL_PACKET, type, &ring->req3,
704 				 sizeof(ring->req3));
705 		break;
706 	}
707 
708 	if (ret == -1) {
709 		perror("setsockopt");
710 		exit(1);
711 	}
712 
713 	ring->rd_len = ring->rd_num * sizeof(*ring->rd);
714 	ring->rd = malloc(ring->rd_len);
715 	if (ring->rd == NULL) {
716 		perror("malloc");
717 		exit(1);
718 	}
719 
720 	total_packets = 0;
721 	total_bytes = 0;
722 }
723 
mmap_ring(int sock,struct ring * ring)724 static void mmap_ring(int sock, struct ring *ring)
725 {
726 	int i;
727 
728 	ring->mm_space = mmap(0, ring->mm_len, PROT_READ | PROT_WRITE,
729 			      MAP_SHARED | MAP_LOCKED | MAP_POPULATE, sock, 0);
730 	if (ring->mm_space == MAP_FAILED) {
731 		perror("mmap");
732 		exit(1);
733 	}
734 
735 	memset(ring->rd, 0, ring->rd_len);
736 	for (i = 0; i < ring->rd_num; ++i) {
737 		ring->rd[i].iov_base = ring->mm_space + (i * ring->flen);
738 		ring->rd[i].iov_len = ring->flen;
739 	}
740 }
741 
bind_ring(int sock,struct ring * ring)742 static void bind_ring(int sock, struct ring *ring)
743 {
744 	int ret;
745 
746 	pair_udp_setfilter(sock);
747 
748 	ring->ll.sll_family = PF_PACKET;
749 	ring->ll.sll_protocol = htons(ETH_P_ALL);
750 	ring->ll.sll_ifindex = if_nametoindex("lo");
751 	ring->ll.sll_hatype = 0;
752 	ring->ll.sll_pkttype = 0;
753 	ring->ll.sll_halen = 0;
754 
755 	ret = bind(sock, (struct sockaddr *) &ring->ll, sizeof(ring->ll));
756 	if (ret == -1) {
757 		perror("bind");
758 		exit(1);
759 	}
760 }
761 
walk_ring(int sock,struct ring * ring)762 static void walk_ring(int sock, struct ring *ring)
763 {
764 	ring->walk(sock, ring);
765 }
766 
unmap_ring(int sock,struct ring * ring)767 static void unmap_ring(int sock, struct ring *ring)
768 {
769 	munmap(ring->mm_space, ring->mm_len);
770 	free(ring->rd);
771 }
772 
test_kernel_bit_width(void)773 static int test_kernel_bit_width(void)
774 {
775 	char in[512], *ptr;
776 	int num = 0, fd;
777 	ssize_t ret;
778 
779 	fd = open("/proc/kallsyms", O_RDONLY);
780 	if (fd == -1) {
781 		perror("open");
782 		exit(1);
783 	}
784 
785 	ret = read(fd, in, sizeof(in));
786 	if (ret <= 0) {
787 		perror("read");
788 		exit(1);
789 	}
790 
791 	close(fd);
792 
793 	ptr = in;
794 	while(!isspace(*ptr)) {
795 		num++;
796 		ptr++;
797 	}
798 
799 	return num * 4;
800 }
801 
test_user_bit_width(void)802 static int test_user_bit_width(void)
803 {
804 	return __WORDSIZE;
805 }
806 
807 static const char *tpacket_str[] = {
808 	[TPACKET_V1] = "TPACKET_V1",
809 	[TPACKET_V2] = "TPACKET_V2",
810 	[TPACKET_V3] = "TPACKET_V3",
811 };
812 
813 static const char *type_str[] = {
814 	[PACKET_RX_RING] = "PACKET_RX_RING",
815 	[PACKET_TX_RING] = "PACKET_TX_RING",
816 };
817 
test_tpacket(int version,int type)818 static int test_tpacket(int version, int type)
819 {
820 	int sock;
821 	struct ring ring;
822 
823 	fprintf(stderr, "test: %s with %s ", tpacket_str[version],
824 		type_str[type]);
825 	fflush(stderr);
826 
827 	if (version == TPACKET_V1 &&
828 	    test_kernel_bit_width() != test_user_bit_width()) {
829 		fprintf(stderr, "test: skip %s %s since user and kernel "
830 			"space have different bit width\n",
831 			tpacket_str[version], type_str[type]);
832 		return 0;
833 	}
834 
835 	sock = pfsocket(version);
836 	memset(&ring, 0, sizeof(ring));
837 	setup_ring(sock, &ring, version, type);
838 	mmap_ring(sock, &ring);
839 	bind_ring(sock, &ring);
840 	walk_ring(sock, &ring);
841 	unmap_ring(sock, &ring);
842 	close(sock);
843 
844 	fprintf(stderr, "\n");
845 	return 0;
846 }
847 
get_kernel_version(int * version,int * patchlevel)848 void get_kernel_version(int *version, int *patchlevel)
849 {
850 	int ret, sublevel;
851 	struct utsname utsname;
852 
853 	ret = uname(&utsname);
854 	if (ret) {
855 		perror("uname");
856 		exit(1);
857 	}
858 
859 	ret = sscanf(utsname.release, "%d.%d.%d", version, patchlevel,
860 		     &sublevel);
861 	if (ret < 0) {
862 		perror("sscanf");
863 		exit(1);
864 	} else if (ret != 3) {
865 		printf("Malformed kernel version %s\n", &utsname.release);
866 		exit(1);
867 	}
868 }
869 
main(void)870 int main(void)
871 {
872 	int ret = 0;
873 	int version, patchlevel;
874 
875 	get_kernel_version(&version, &patchlevel);
876 
877 	ret |= test_tpacket(TPACKET_V1, PACKET_RX_RING);
878 	ret |= test_tpacket(TPACKET_V1, PACKET_TX_RING);
879 
880 	ret |= test_tpacket(TPACKET_V2, PACKET_RX_RING);
881 	ret |= test_tpacket(TPACKET_V2, PACKET_TX_RING);
882 
883 	ret |= test_tpacket(TPACKET_V3, PACKET_RX_RING);
884 	if (version > 4 || (version == 4 && patchlevel >= 11))
885 		ret |= test_tpacket(TPACKET_V3, PACKET_TX_RING);
886 
887 	if (ret)
888 		return 1;
889 
890 	printf("OK. All tests passed\n");
891 	return 0;
892 }
893