1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
|
#include "../config-host.h"
/* SPDX-License-Identifier: MIT */
/*
* Simple test case showing using send and recv bundles
*/
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <arpa/inet.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <pthread.h>
#define MSG_SIZE 128
#define NR_MIN_MSGS 4
#define NR_MAX_MSGS 32
#define SEQ_SIZE (MSG_SIZE / sizeof(unsigned long))
static int nr_msgs;
static int use_tcp;
#define RECV_BIDS 8192
#define RECV_BID_MASK (RECV_BIDS - 1)
#include "liburing.h"
#include "helpers.h"
#define PORT 10202
#define HOST "127.0.0.1"
static int use_port = PORT;
#define SEND_BGID 7
#define RECV_BGID 8
static int no_send_mshot;
struct recv_data {
pthread_barrier_t connect;
pthread_barrier_t startup;
pthread_barrier_t barrier;
pthread_barrier_t finish;
unsigned long seq;
int recv_bytes;
int accept_fd;
int abort;
unsigned int max_sends;
int to_eagain;
void *recv_buf;
int send_bundle;
int recv_bundle;
};
static int arm_recv(struct io_uring *ring, struct recv_data *rd)
{
struct io_uring_sqe *sqe;
int ret;
sqe = io_uring_get_sqe(ring);
io_uring_prep_recv_multishot(sqe, rd->accept_fd, NULL, 0, 0);
if (rd->recv_bundle && use_tcp)
sqe->ioprio |= IORING_RECVSEND_BUNDLE;
sqe->buf_group = RECV_BGID;
sqe->flags |= IOSQE_BUFFER_SELECT;
sqe->user_data = 2;
ret = io_uring_submit(ring);
if (ret != 1) {
fprintf(stderr, "submit failed: %d\n", ret);
return 1;
}
return 0;
}
static int recv_prep(struct io_uring *ring, struct recv_data *rd, int *sock)
{
struct sockaddr_in saddr;
int sockfd, ret, val, use_fd;
socklen_t socklen;
memset(&saddr, 0, sizeof(saddr));
saddr.sin_family = AF_INET;
saddr.sin_addr.s_addr = htonl(INADDR_ANY);
saddr.sin_port = htons(use_port);
if (use_tcp)
sockfd = socket(AF_INET, SOCK_STREAM, 0);
else
sockfd = socket(AF_INET, SOCK_DGRAM, 0);
if (sockfd < 0) {
perror("socket");
return 1;
}
val = 1;
setsockopt(sockfd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val));
ret = bind(sockfd, (struct sockaddr *)&saddr, sizeof(saddr));
if (ret < 0) {
perror("bind");
goto err;
}
if (use_tcp) {
ret = listen(sockfd, 1);
if (ret < 0) {
perror("listen");
goto err;
}
pthread_barrier_wait(&rd->connect);
socklen = sizeof(saddr);
use_fd = accept(sockfd, (struct sockaddr *)&saddr, &socklen);
if (use_fd < 0) {
perror("accept");
goto err;
}
} else {
use_fd = sockfd;
pthread_barrier_wait(&rd->connect);
}
rd->accept_fd = use_fd;
pthread_barrier_wait(&rd->startup);
pthread_barrier_wait(&rd->barrier);
if (arm_recv(ring, rd))
goto err;
*sock = sockfd;
return 0;
err:
close(sockfd);
return 1;
}
static int verify_seq(struct recv_data *rd, void *verify_ptr, int verify_sz,
int start_bid)
{
unsigned long *seqp;
int seq_size = verify_sz / sizeof(unsigned long);
int i;
seqp = verify_ptr;
for (i = 0; i < seq_size; i++) {
if (rd->seq != *seqp) {
fprintf(stderr, "bid=%d, got seq %lu, wanted %lu, offset %d\n", start_bid, *seqp, rd->seq, i);
return 0;
}
seqp++;
rd->seq++;
}
return 1;
}
static int recv_get_cqe(struct io_uring *ring, struct recv_data *rd,
struct io_uring_cqe **cqe)
{
struct __kernel_timespec ts = { .tv_sec = 0, .tv_nsec = 100000000LL };
int ret;
do {
ret = io_uring_wait_cqe_timeout(ring, cqe, &ts);
if (!ret)
return 0;
if (ret == -ETIME) {
if (rd->abort)
break;
continue;
}
fprintf(stderr, "wait recv: %d\n", ret);
break;
} while (1);
return 1;
}
static int do_recv(struct io_uring *ring, struct recv_data *rd)
{
struct io_uring_cqe *cqe;
int bid, next_bid = 0;
void *verify_ptr;
int verify_sz = 0;
int verify_bid = 0;
verify_ptr = malloc(rd->recv_bytes);
do {
if (recv_get_cqe(ring, rd, &cqe))
break;
if (cqe->res == -EINVAL) {
fprintf(stdout, "recv not supported, skipping\n");
return 0;
}
if (cqe->res < 0) {
fprintf(stderr, "failed recv cqe: %d\n", cqe->res);
goto err;
}
if (!(cqe->flags & IORING_CQE_F_BUFFER)) {
fprintf(stderr, "no buffer set in recv\n");
goto err;
}
bid = cqe->flags >> IORING_CQE_BUFFER_SHIFT;
if (bid != next_bid) {
fprintf(stderr, "got bid %d, wanted %d\n", bid, next_bid);
goto err;
}
if (!rd->recv_bundle && cqe->res != MSG_SIZE) {
fprintf(stderr, "recv got wrong length: %d\n", cqe->res);
goto err;
}
if (!(verify_sz % MSG_SIZE)) {
if (!verify_seq(rd, verify_ptr, verify_sz, verify_bid))
goto err;
verify_bid += verify_sz / MSG_SIZE;
verify_bid &= RECV_BID_MASK;
verify_sz = 0;
} else {
memcpy(verify_ptr + verify_sz, rd->recv_buf + (bid * MSG_SIZE), cqe->res);
verify_sz += cqe->res;
}
next_bid = bid + ((cqe->res + MSG_SIZE - 1) / MSG_SIZE);
next_bid &= RECV_BID_MASK;
rd->recv_bytes -= cqe->res;
io_uring_cqe_seen(ring, cqe);
if (!(cqe->flags & IORING_CQE_F_MORE) && rd->recv_bytes) {
if (arm_recv(ring, rd))
goto err;
}
} while (rd->recv_bytes);
if (verify_sz && !(verify_sz % MSG_SIZE) &&
!verify_seq(rd, verify_ptr, verify_sz, verify_bid))
goto err;
pthread_barrier_wait(&rd->finish);
return 0;
err:
pthread_barrier_wait(&rd->finish);
return 1;
}
static void *recv_fn(void *data)
{
struct recv_data *rd = data;
struct io_uring_params p = { };
struct io_uring ring;
struct io_uring_buf_ring *br;
void *buf, *ptr;
int ret, sock, i;
p.cq_entries = 4096;
p.flags = IORING_SETUP_CQSIZE;
ret = t_create_ring_params(16, &ring, &p);
if (ret == T_SETUP_SKIP) {
ret = 0;
goto err;
} else if (ret < 0) {
goto err;
}
if (posix_memalign(&buf, 4096, MSG_SIZE * RECV_BIDS))
goto err;
br = io_uring_setup_buf_ring(&ring, RECV_BIDS, RECV_BGID, 0, &ret);
if (!br) {
fprintf(stderr, "failed setting up recv ring %d\n", ret);
goto err;
}
ptr = buf;
for (i = 0; i < RECV_BIDS; i++) {
io_uring_buf_ring_add(br, ptr, MSG_SIZE, i, RECV_BID_MASK, i);
ptr += MSG_SIZE;
}
io_uring_buf_ring_advance(br, RECV_BIDS);
rd->recv_buf = buf;
ret = recv_prep(&ring, rd, &sock);
if (ret) {
fprintf(stderr, "recv_prep failed: %d\n", ret);
goto err;
}
ret = do_recv(&ring, rd);
close(sock);
close(rd->accept_fd);
io_uring_queue_exit(&ring);
err:
return (void *)(intptr_t)ret;
}
static int __do_send_bundle(struct recv_data *rd, struct io_uring *ring, int sockfd)
{
struct io_uring_cqe *cqe;
struct io_uring_sqe *sqe;
int bytes_needed = MSG_SIZE * nr_msgs;
int i, ret;
sqe = io_uring_get_sqe(ring);
io_uring_prep_send_bundle(sqe, sockfd, 0, 0);
sqe->flags |= IOSQE_BUFFER_SELECT;
sqe->buf_group = SEND_BGID;
sqe->user_data = 1;
ret = io_uring_submit(ring);
if (ret != 1)
return 1;
pthread_barrier_wait(&rd->barrier);
for (i = 0; i < nr_msgs; i++) {
ret = io_uring_wait_cqe(ring, &cqe);
if (ret) {
fprintf(stderr, "wait send: %d\n", ret);
return 1;
}
if (!i && cqe->res == -EINVAL) {
rd->abort = 1;
no_send_mshot = 1;
break;
}
if (cqe->res < 0) {
fprintf(stderr, "bad send cqe res: %d\n", cqe->res);
return 1;
}
bytes_needed -= cqe->res;
if (!bytes_needed) {
io_uring_cqe_seen(ring, cqe);
break;
}
if (!(cqe->flags & IORING_CQE_F_MORE)) {
fprintf(stderr, "expected more, but MORE not set\n");
return 1;
}
io_uring_cqe_seen(ring, cqe);
}
return 0;
}
static int __do_send(struct recv_data *rd, struct io_uring *ring, int sockfd)
{
struct io_uring_cqe *cqe;
struct io_uring_sqe *sqe;
int bytes_needed = MSG_SIZE * nr_msgs;
int i, ret;
for (i = 0; i < nr_msgs; i++) {
sqe = io_uring_get_sqe(ring);
io_uring_prep_send(sqe, sockfd, NULL, 0, 0);
sqe->user_data = 10 + i;
sqe->flags |= IOSQE_BUFFER_SELECT;
sqe->buf_group = SEND_BGID;
ret = io_uring_submit(ring);
if (ret != 1)
return 1;
if (!i)
pthread_barrier_wait(&rd->barrier);
ret = io_uring_wait_cqe(ring, &cqe);
if (ret) {
fprintf(stderr, "send wait cqe %d\n", ret);
return 1;
}
if (!i && cqe->res == -EINVAL) {
rd->abort = 1;
no_send_mshot = 1;
break;
}
if (cqe->res != MSG_SIZE) {
fprintf(stderr, "send failed cqe: %d\n", cqe->res);
return 1;
}
if (cqe->res < 0) {
fprintf(stderr, "bad send cqe res: %d\n", cqe->res);
return 1;
}
bytes_needed -= cqe->res;
io_uring_cqe_seen(ring, cqe);
if (!bytes_needed)
break;
}
return 0;
}
static int do_send(struct recv_data *rd)
{
struct sockaddr_in saddr;
struct io_uring ring;
unsigned long seq_buf[SEQ_SIZE], send_seq;
struct io_uring_params p = { };
struct io_uring_buf_ring *br;
int sockfd, ret, len, i;
socklen_t optlen;
void *buf, *ptr;
ret = io_uring_queue_init_params(16, &ring, &p);
if (ret) {
fprintf(stderr, "queue init failed: %d\n", ret);
return 1;
}
if (!(p.features & IORING_FEAT_RECVSEND_BUNDLE)) {
no_send_mshot = 1;
return 0;
}
if (posix_memalign(&buf, 4096, MSG_SIZE * nr_msgs))
return 1;
br = io_uring_setup_buf_ring(&ring, nr_msgs, SEND_BGID, 0, &ret);
if (!br) {
if (ret == -EINVAL) {
fprintf(stderr, "einval on br setup\n");
return 0;
}
fprintf(stderr, "failed setting up send ring %d\n", ret);
return 1;
}
ptr = buf;
for (i = 0; i < nr_msgs; i++) {
io_uring_buf_ring_add(br, ptr, MSG_SIZE, i, nr_msgs - 1, i);
ptr += MSG_SIZE;
}
io_uring_buf_ring_advance(br, nr_msgs);
memset(&saddr, 0, sizeof(saddr));
saddr.sin_family = AF_INET;
saddr.sin_port = htons(use_port);
inet_pton(AF_INET, HOST, &saddr.sin_addr);
if (use_tcp)
sockfd = socket(AF_INET, SOCK_STREAM, 0);
else
sockfd = socket(AF_INET, SOCK_DGRAM, 0);
if (sockfd < 0) {
perror("socket");
goto err2;
}
pthread_barrier_wait(&rd->connect);
ret = connect(sockfd, (struct sockaddr *)&saddr, sizeof(saddr));
if (ret < 0) {
perror("connect");
goto err;
}
pthread_barrier_wait(&rd->startup);
optlen = sizeof(len);
len = 1024 * MSG_SIZE;
setsockopt(sockfd, SOL_SOCKET, SO_SNDBUF, &len, optlen);
/* almost fill queue, leave room for one message */
send_seq = 0;
rd->to_eagain = 0;
while (rd->max_sends && rd->max_sends--) {
for (i = 0; i < SEQ_SIZE; i++)
seq_buf[i] = send_seq++;
ret = send(sockfd, seq_buf, sizeof(seq_buf), MSG_DONTWAIT);
if (ret < 0) {
if (errno == EAGAIN) {
send_seq -= SEQ_SIZE;
break;
}
perror("send");
return 1;
} else if (ret != sizeof(seq_buf)) {
fprintf(stderr, "short %d send\n", ret);
return 1;
}
rd->to_eagain++;
rd->recv_bytes += sizeof(seq_buf);
}
ptr = buf;
for (i = 0; i < nr_msgs; i++) {
unsigned long *pseq = ptr;
int j;
for (j = 0; j < SEQ_SIZE; j++)
pseq[j] = send_seq++;
ptr += MSG_SIZE;
}
/* prepare more messages, sending with bundle */
rd->recv_bytes += (nr_msgs * MSG_SIZE);
if (rd->send_bundle && use_tcp)
ret = __do_send_bundle(rd, &ring, sockfd);
else
ret = __do_send(rd, &ring, sockfd);
if (ret)
goto err;
pthread_barrier_wait(&rd->finish);
close(sockfd);
io_uring_queue_exit(&ring);
return 0;
err:
close(sockfd);
err2:
io_uring_queue_exit(&ring);
pthread_barrier_wait(&rd->finish);
return 1;
}
static int test(int backlog, unsigned int max_sends, int *to_eagain,
int send_bundle, int recv_bundle)
{
pthread_t recv_thread;
struct recv_data rd;
int ret;
void *retval;
/* backlog not reliable on UDP, skip it */
if ((backlog || max_sends) && !use_tcp)
return T_EXIT_PASS;
memset(&rd, 0, sizeof(rd));
pthread_barrier_init(&rd.connect, NULL, 2);
pthread_barrier_init(&rd.startup, NULL, 2);
pthread_barrier_init(&rd.barrier, NULL, 2);
pthread_barrier_init(&rd.finish, NULL, 2);
rd.max_sends = max_sends;
if (to_eagain)
*to_eagain = 0;
rd.send_bundle = send_bundle;
rd.recv_bundle = recv_bundle;
ret = pthread_create(&recv_thread, NULL, recv_fn, &rd);
if (ret) {
fprintf(stderr, "Thread create failed: %d\n", ret);
return 1;
}
ret = do_send(&rd);
if (no_send_mshot)
return 0;
if (ret)
return ret;
pthread_join(recv_thread, &retval);
if (to_eagain)
*to_eagain = rd.to_eagain;
return (intptr_t)retval;
}
static int run_tests(int is_udp)
{
int ret, eagain_hit;
nr_msgs = NR_MIN_MSGS;
/* test basic send bundle first */
ret = test(0, 0, NULL, 0, 0);
if (ret) {
fprintf(stderr, "test a failed\n");
return T_EXIT_FAIL;
}
if (no_send_mshot)
return T_EXIT_SKIP;
/* test recv bundle */
ret = test(0, 0, NULL, 0, 1);
if (ret) {
fprintf(stderr, "test b failed\n");
return T_EXIT_FAIL;
}
/* test bundling recv and send */
ret = test(0, 0, NULL, 1, 1);
if (ret) {
fprintf(stderr, "test c failed\n");
return T_EXIT_FAIL;
}
/* test bundling with full socket */
ret = test(1, 1000000, &eagain_hit, 1, 1);
if (ret) {
fprintf(stderr, "test d failed\n");
return T_EXIT_FAIL;
}
/* test bundling with almost full socket */
ret = test(1, eagain_hit - (nr_msgs / 2), NULL, 1, 1);
if (ret) {
fprintf(stderr, "test e failed\n");
return T_EXIT_FAIL;
}
/* test recv bundle with almost full socket */
ret = test(1, eagain_hit - (nr_msgs / 2), NULL, 0, 1);
if (ret) {
fprintf(stderr, "test f failed\n");
return T_EXIT_FAIL;
}
if (is_udp)
return T_EXIT_PASS;
/* test send bundle with almost full socket */
ret = test(1, eagain_hit - (nr_msgs / 2), &eagain_hit, 1, 0);
if (ret) {
fprintf(stderr, "test g failed\n");
return T_EXIT_FAIL;
}
/* now repeat the last three tests, but with > FAST_UIOV segments */
nr_msgs = NR_MAX_MSGS;
/* test bundling with almost full socket */
ret = test(1, eagain_hit - (nr_msgs / 2), NULL, 1, 1);
if (ret) {
fprintf(stderr, "test h failed\n");
return T_EXIT_FAIL;
}
/* test recv bundle with almost full socket */
ret = test(1, eagain_hit - (nr_msgs / 2), NULL, 0, 1);
if (ret) {
fprintf(stderr, "test i failed\n");
return T_EXIT_FAIL;
}
/* test send bundle with almost full socket */
ret = test(1, eagain_hit - (nr_msgs / 2), &eagain_hit, 1, 0);
if (ret) {
fprintf(stderr, "test j failed\n");
return T_EXIT_FAIL;
}
return T_EXIT_PASS;
}
static int test_tcp(void)
{
int ret;
use_tcp = 1;
ret = run_tests(false);
if (ret == T_EXIT_FAIL)
fprintf(stderr, "TCP test case failed\n");
return ret;
}
static int test_udp(void)
{
int ret;
use_tcp = 0;
use_port++;
ret = run_tests(true);
if (ret == T_EXIT_FAIL)
fprintf(stderr, "UDP test case failed\n");
return ret;
}
int main(int argc, char *argv[])
{
int ret;
if (argc > 1)
return T_EXIT_SKIP;
ret = test_tcp();
if (ret != T_EXIT_PASS)
return ret;
ret = test_udp();
if (ret != T_EXIT_PASS)
return ret;
return T_EXIT_PASS;
}
|