projects / fio.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
buffer: only set refill_buffers, it it wasn't set manually
[fio.git] / engines / rdma.c
... / ...
CommitLineData
1/*
2 * RDMA I/O engine
3 *
4 * RDMA I/O engine based on the IB verbs and RDMA/CM user space libraries.
5 * Supports both RDMA memory semantics and channel semantics
6 * for the InfiniBand, RoCE and iWARP protocols.
7 *
8 * You will need the Linux RDMA software installed, either
9 * from your Linux distributor or directly from openfabrics.org:
10 *
11 * http://www.openfabrics.org/downloads/OFED/
12 *
13 * Exchanging steps of RDMA ioengine control messages:
14 * 1. client side sends test mode (RDMA_WRITE/RDMA_READ/SEND)
15 * to server side.
16 * 2. server side parses test mode, and sends back confirmation
17 * to client side. In RDMA WRITE/READ test, this confirmation
18 * includes memory information, such as rkey, address.
19 * 3. client side initiates test loop.
20 * 4. In RDMA WRITE/READ test, client side sends a completion
21 * notification to server side. Server side updates its
22 * td->done as true.
23 *
24 */
25#include <stdio.h>
26#include <stdlib.h>
27#include <unistd.h>
28#include <errno.h>
29#include <assert.h>
30#include <netinet/in.h>
31#include <arpa/inet.h>
32#include <netdb.h>
33#include <sys/poll.h>
34#include <sys/types.h>
35#include <sys/socket.h>
36#include <sys/time.h>
37#include <sys/resource.h>
38
39#include <pthread.h>
40#include <inttypes.h>
41
42#include "../fio.h"
43#include "../hash.h"
44
45#include <rdma/rdma_cma.h>
46#include <infiniband/arch.h>
47
48#define FIO_RDMA_MAX_IO_DEPTH 512
49
50enum rdma_io_mode {
51 FIO_RDMA_UNKNOWN = 0,
52 FIO_RDMA_MEM_WRITE,
53 FIO_RDMA_MEM_READ,
54 FIO_RDMA_CHA_SEND,
55 FIO_RDMA_CHA_RECV
56};
57
58struct rdmaio_options {
59 struct thread_data *td;
60 unsigned int port;
61 enum rdma_io_mode verb;
62};
63
64static int str_hostname_cb(void *data, const char *input)
65{
66 struct rdmaio_options *o = data;
67
68 if (o->td->o.filename)
69 free(o->td->o.filename);
70 o->td->o.filename = strdup(input);
71 return 0;
72}
73
74static struct fio_option options[] = {
75 {
76 .name = "hostname",
77 .lname = "rdma engine hostname",
78 .type = FIO_OPT_STR_STORE,
79 .cb = str_hostname_cb,
80 .help = "Hostname for RDMA IO engine",
81 .category = FIO_OPT_C_ENGINE,
82 .group = FIO_OPT_G_RDMA,
83 },
84 {
85 .name = "port",
86 .lname = "rdma engine port",
87 .type = FIO_OPT_INT,
88 .off1 = offsetof(struct rdmaio_options, port),
89 .minval = 1,
90 .maxval = 65535,
91 .help = "Port to use for RDMA connections",
92 .category = FIO_OPT_C_ENGINE,
93 .group = FIO_OPT_G_RDMA,
94 },
95 {
96 .name = "verb",
97 .lname = "RDMA engine verb",
98 .alias = "proto",
99 .type = FIO_OPT_STR,
100 .off1 = offsetof(struct rdmaio_options, verb),
101 .help = "RDMA engine verb",
102 .def = "write",
103 .posval = {
104 { .ival = "write",
105 .oval = FIO_RDMA_MEM_WRITE,
106 .help = "Memory Write",
107 },
108 { .ival = "read",
109 .oval = FIO_RDMA_MEM_READ,
110 .help = "Memory Read",
111 },
112 { .ival = "send",
113 .oval = FIO_RDMA_CHA_SEND,
114 .help = "Posted Send",
115 },
116 { .ival = "recv",
117 .oval = FIO_RDMA_CHA_RECV,
118 .help = "Posted Recieve",
119 },
120 },
121 .category = FIO_OPT_C_ENGINE,
122 .group = FIO_OPT_G_RDMA,
123 },
124 {
125 .name = NULL,
126 },
127};
128
129struct remote_u {
130 uint64_t buf;
131 uint32_t rkey;
132 uint32_t size;
133};
134
135struct rdma_info_blk {
136 uint32_t mode; /* channel semantic or memory semantic */
137 uint32_t nr; /* client: io depth
138 server: number of records for memory semantic
139 */
140 struct remote_u rmt_us[FIO_RDMA_MAX_IO_DEPTH];
141};
142
143struct rdma_io_u_data {
144 uint64_t wr_id;
145 struct ibv_send_wr sq_wr;
146 struct ibv_recv_wr rq_wr;
147 struct ibv_sge rdma_sgl;
148};
149
150struct rdmaio_data {
151 int is_client;
152 enum rdma_io_mode rdma_protocol;
153 char host[64];
154 struct sockaddr_in addr;
155
156 struct ibv_recv_wr rq_wr;
157 struct ibv_sge recv_sgl;
158 struct rdma_info_blk recv_buf;
159 struct ibv_mr *recv_mr;
160
161 struct ibv_send_wr sq_wr;
162 struct ibv_sge send_sgl;
163 struct rdma_info_blk send_buf;
164 struct ibv_mr *send_mr;
165
166 struct ibv_comp_channel *channel;
167 struct ibv_cq *cq;
168 struct ibv_pd *pd;
169 struct ibv_qp *qp;
170
171 pthread_t cmthread;
172 struct rdma_event_channel *cm_channel;
173 struct rdma_cm_id *cm_id;
174 struct rdma_cm_id *child_cm_id;
175
176 int cq_event_num;
177
178 struct remote_u *rmt_us;
179 int rmt_nr;
180 struct io_u **io_us_queued;
181 int io_u_queued_nr;
182 struct io_u **io_us_flight;
183 int io_u_flight_nr;
184 struct io_u **io_us_completed;
185 int io_u_completed_nr;
186
187 struct frand_state rand_state;
188};
189
190static int client_recv(struct thread_data *td, struct ibv_wc *wc)
191{
192 struct rdmaio_data *rd = td->io_ops->data;
193
194 if (wc->byte_len != sizeof(rd->recv_buf)) {
195 log_err("Received bogus data, size %d\n", wc->byte_len);
196 return 1;
197 }
198
199 /* store mr info for MEMORY semantic */
200 if ((rd->rdma_protocol == FIO_RDMA_MEM_WRITE) ||
201 (rd->rdma_protocol == FIO_RDMA_MEM_READ)) {
202 /* struct flist_head *entry; */
203 int i = 0;
204
205 rd->rmt_nr = ntohl(rd->recv_buf.nr);
206
207 for (i = 0; i < rd->rmt_nr; i++) {
208 rd->rmt_us[i].buf = ntohll(rd->recv_buf.rmt_us[i].buf);
209 rd->rmt_us[i].rkey = ntohl(rd->recv_buf.rmt_us[i].rkey);
210 rd->rmt_us[i].size = ntohl(rd->recv_buf.rmt_us[i].size);
211
212 dprint(FD_IO,
213 "fio: Received rkey %x addr %" PRIx64
214 " len %d from peer\n", rd->rmt_us[i].rkey,
215 rd->rmt_us[i].buf, rd->rmt_us[i].size);
216 }
217 }
218
219 return 0;
220}
221
222static int server_recv(struct thread_data *td, struct ibv_wc *wc)
223{
224 struct rdmaio_data *rd = td->io_ops->data;
225
226 if (wc->wr_id == FIO_RDMA_MAX_IO_DEPTH) {
227 rd->rdma_protocol = ntohl(rd->recv_buf.mode);
228
229 /* CHANNEL semantic, do nothing */
230 if (rd->rdma_protocol == FIO_RDMA_CHA_SEND)
231 rd->rdma_protocol = FIO_RDMA_CHA_RECV;
232 }
233
234 return 0;
235}
236
237static int cq_event_handler(struct thread_data *td, enum ibv_wc_opcode opcode)
238{
239 struct rdmaio_data *rd = td->io_ops->data;
240 struct ibv_wc wc;
241 struct rdma_io_u_data *r_io_u_d;
242 int ret;
243 int compevnum = 0;
244 int i;
245
246 while ((ret = ibv_poll_cq(rd->cq, 1, &wc)) == 1) {
247 ret = 0;
248 compevnum++;
249
250 if (wc.status) {
251 log_err("fio: cq completion status %d(%s)\n",
252 wc.status, ibv_wc_status_str(wc.status));
253 return -1;
254 }
255
256 switch (wc.opcode) {
257
258 case IBV_WC_RECV:
259 if (rd->is_client == 1)
260 client_recv(td, &wc);
261 else
262 server_recv(td, &wc);
263
264 if (wc.wr_id == FIO_RDMA_MAX_IO_DEPTH)
265 break;
266
267 for (i = 0; i < rd->io_u_flight_nr; i++) {
268 r_io_u_d = rd->io_us_flight[i]->engine_data;
269
270 if (wc.wr_id == r_io_u_d->rq_wr.wr_id) {
271 rd->io_us_flight[i]->resid =
272 rd->io_us_flight[i]->buflen
273 - wc.byte_len;
274
275 rd->io_us_flight[i]->error = 0;
276
277 rd->io_us_completed[rd->
278 io_u_completed_nr]
279 = rd->io_us_flight[i];
280 rd->io_u_completed_nr++;
281 break;
282 }
283 }
284 if (i == rd->io_u_flight_nr)
285 log_err("fio: recv wr %" PRId64 " not found\n",
286 wc.wr_id);
287 else {
288 /* put the last one into middle of the list */
289 rd->io_us_flight[i] =
290 rd->io_us_flight[rd->io_u_flight_nr - 1];
291 rd->io_u_flight_nr--;
292 }
293
294 break;
295
296 case IBV_WC_SEND:
297 case IBV_WC_RDMA_WRITE:
298 case IBV_WC_RDMA_READ:
299 if (wc.wr_id == FIO_RDMA_MAX_IO_DEPTH)
300 break;
301
302 for (i = 0; i < rd->io_u_flight_nr; i++) {
303 r_io_u_d = rd->io_us_flight[i]->engine_data;
304
305 if (wc.wr_id == r_io_u_d->sq_wr.wr_id) {
306 rd->io_us_completed[rd->
307 io_u_completed_nr]
308 = rd->io_us_flight[i];
309 rd->io_u_completed_nr++;
310 break;
311 }
312 }
313 if (i == rd->io_u_flight_nr)
314 log_err("fio: send wr %" PRId64 " not found\n",
315 wc.wr_id);
316 else {
317 /* put the last one into middle of the list */
318 rd->io_us_flight[i] =
319 rd->io_us_flight[rd->io_u_flight_nr - 1];
320 rd->io_u_flight_nr--;
321 }
322
323 break;
324
325 default:
326 log_info("fio: unknown completion event %d\n",
327 wc.opcode);
328 return -1;
329 }
330 rd->cq_event_num++;
331 }
332 if (ret) {
333 log_err("fio: poll error %d\n", ret);
334 return 1;
335 }
336
337 return compevnum;
338}
339
340/*
341 * Return -1 for error and 'nr events' for a positive number
342 * of events
343 */
344static int rdma_poll_wait(struct thread_data *td, enum ibv_wc_opcode opcode)
345{
346 struct rdmaio_data *rd = td->io_ops->data;
347 struct ibv_cq *ev_cq;
348 void *ev_ctx;
349 int ret;
350
351 if (rd->cq_event_num > 0) { /* previous left */
352 rd->cq_event_num--;
353 return 0;
354 }
355
356again:
357 if (ibv_get_cq_event(rd->channel, &ev_cq, &ev_ctx) != 0) {
358 log_err("fio: Failed to get cq event!\n");
359 return -1;
360 }
361 if (ev_cq != rd->cq) {
362 log_err("fio: Unknown CQ!\n");
363 return -1;
364 }
365 if (ibv_req_notify_cq(rd->cq, 0) != 0) {
366 log_err("fio: Failed to set notify!\n");
367 return -1;
368 }
369
370 ret = cq_event_handler(td, opcode);
371 if (ret < 1)
372 goto again;
373
374 ibv_ack_cq_events(rd->cq, ret);
375
376 rd->cq_event_num--;
377
378 return ret;
379}
380
381static int fio_rdmaio_setup_qp(struct thread_data *td)
382{
383 struct rdmaio_data *rd = td->io_ops->data;
384 struct ibv_qp_init_attr init_attr;
385 int qp_depth = td->o.iodepth * 2; /* 2 times of io depth */
386
387 if (rd->is_client == 0)
388 rd->pd = ibv_alloc_pd(rd->child_cm_id->verbs);
389 else
390 rd->pd = ibv_alloc_pd(rd->cm_id->verbs);
391
392 if (rd->pd == NULL) {
393 log_err("fio: ibv_alloc_pd fail\n");
394 return 1;
395 }
396
397 if (rd->is_client == 0)
398 rd->channel = ibv_create_comp_channel(rd->child_cm_id->verbs);
399 else
400 rd->channel = ibv_create_comp_channel(rd->cm_id->verbs);
401 if (rd->channel == NULL) {
402 log_err("fio: ibv_create_comp_channel fail\n");
403 goto err1;
404 }
405
406 if (qp_depth < 16)
407 qp_depth = 16;
408
409 if (rd->is_client == 0)
410 rd->cq = ibv_create_cq(rd->child_cm_id->verbs,
411 qp_depth, rd, rd->channel, 0);
412 else
413 rd->cq = ibv_create_cq(rd->cm_id->verbs,
414 qp_depth, rd, rd->channel, 0);
415 if (rd->cq == NULL) {
416 log_err("fio: ibv_create_cq failed\n");
417 goto err2;
418 }
419
420 if (ibv_req_notify_cq(rd->cq, 0) != 0) {
421 log_err("fio: ibv_create_cq failed\n");
422 goto err3;
423 }
424
425 /* create queue pair */
426 memset(&init_attr, 0, sizeof(init_attr));
427 init_attr.cap.max_send_wr = qp_depth;
428 init_attr.cap.max_recv_wr = qp_depth;
429 init_attr.cap.max_recv_sge = 1;
430 init_attr.cap.max_send_sge = 1;
431 init_attr.qp_type = IBV_QPT_RC;
432 init_attr.send_cq = rd->cq;
433 init_attr.recv_cq = rd->cq;
434
435 if (rd->is_client == 0) {
436 if (rdma_create_qp(rd->child_cm_id, rd->pd, &init_attr) != 0) {
437 log_err("fio: rdma_create_qp failed\n");
438 goto err3;
439 }
440 rd->qp = rd->child_cm_id->qp;
441 } else {
442 if (rdma_create_qp(rd->cm_id, rd->pd, &init_attr) != 0) {
443 log_err("fio: rdma_create_qp failed\n");
444 goto err3;
445 }
446 rd->qp = rd->cm_id->qp;
447 }
448
449 return 0;
450
451err3:
452 ibv_destroy_cq(rd->cq);
453err2:
454 ibv_destroy_comp_channel(rd->channel);
455err1:
456 ibv_dealloc_pd(rd->pd);
457
458 return 1;
459}
460
461static int fio_rdmaio_setup_control_msg_buffers(struct thread_data *td)
462{
463 struct rdmaio_data *rd = td->io_ops->data;
464
465 rd->recv_mr = ibv_reg_mr(rd->pd, &rd->recv_buf, sizeof(rd->recv_buf),
466 IBV_ACCESS_LOCAL_WRITE);
467 if (rd->recv_mr == NULL) {
468 log_err("fio: recv_buf reg_mr failed\n");
469 return 1;
470 }
471
472 rd->send_mr = ibv_reg_mr(rd->pd, &rd->send_buf, sizeof(rd->send_buf),
473 0);
474 if (rd->send_mr == NULL) {
475 log_err("fio: send_buf reg_mr failed\n");
476 ibv_dereg_mr(rd->recv_mr);
477 return 1;
478 }
479
480 /* setup work request */
481 /* recv wq */
482 rd->recv_sgl.addr = (uint64_t) (unsigned long)&rd->recv_buf;
483 rd->recv_sgl.length = sizeof(rd->recv_buf);
484 rd->recv_sgl.lkey = rd->recv_mr->lkey;
485 rd->rq_wr.sg_list = &rd->recv_sgl;
486 rd->rq_wr.num_sge = 1;
487 rd->rq_wr.wr_id = FIO_RDMA_MAX_IO_DEPTH;
488
489 /* send wq */
490 rd->send_sgl.addr = (uint64_t) (unsigned long)&rd->send_buf;
491 rd->send_sgl.length = sizeof(rd->send_buf);
492 rd->send_sgl.lkey = rd->send_mr->lkey;
493
494 rd->sq_wr.opcode = IBV_WR_SEND;
495 rd->sq_wr.send_flags = IBV_SEND_SIGNALED;
496 rd->sq_wr.sg_list = &rd->send_sgl;
497 rd->sq_wr.num_sge = 1;
498 rd->sq_wr.wr_id = FIO_RDMA_MAX_IO_DEPTH;
499
500 return 0;
501}
502
503static int get_next_channel_event(struct thread_data *td,
504 struct rdma_event_channel *channel,
505 enum rdma_cm_event_type wait_event)
506{
507 struct rdmaio_data *rd = td->io_ops->data;
508 struct rdma_cm_event *event;
509 int ret;
510
511 ret = rdma_get_cm_event(channel, &event);
512 if (ret) {
513 log_err("fio: rdma_get_cm_event: %d\n", ret);
514 return 1;
515 }
516
517 if (event->event != wait_event) {
518 log_err("fio: event is %s instead of %s\n",
519 rdma_event_str(event->event),
520 rdma_event_str(wait_event));
521 return 1;
522 }
523
524 switch (event->event) {
525 case RDMA_CM_EVENT_CONNECT_REQUEST:
526 rd->child_cm_id = event->id;
527 break;
528 default:
529 break;
530 }
531
532 rdma_ack_cm_event(event);
533
534 return 0;
535}
536
537static int fio_rdmaio_prep(struct thread_data *td, struct io_u *io_u)
538{
539 struct rdmaio_data *rd = td->io_ops->data;
540 struct rdma_io_u_data *r_io_u_d;
541
542 r_io_u_d = io_u->engine_data;
543
544 switch (rd->rdma_protocol) {
545 case FIO_RDMA_MEM_WRITE:
546 case FIO_RDMA_MEM_READ:
547 r_io_u_d->rdma_sgl.addr = (uint64_t) (unsigned long)io_u->buf;
548 r_io_u_d->rdma_sgl.lkey = io_u->mr->lkey;
549 r_io_u_d->sq_wr.wr_id = r_io_u_d->wr_id;
550 r_io_u_d->sq_wr.send_flags = IBV_SEND_SIGNALED;
551 r_io_u_d->sq_wr.sg_list = &r_io_u_d->rdma_sgl;
552 r_io_u_d->sq_wr.num_sge = 1;
553 break;
554 case FIO_RDMA_CHA_SEND:
555 r_io_u_d->rdma_sgl.addr = (uint64_t) (unsigned long)io_u->buf;
556 r_io_u_d->rdma_sgl.lkey = io_u->mr->lkey;
557 r_io_u_d->rdma_sgl.length = io_u->buflen;
558 r_io_u_d->sq_wr.wr_id = r_io_u_d->wr_id;
559 r_io_u_d->sq_wr.opcode = IBV_WR_SEND;
560 r_io_u_d->sq_wr.send_flags = IBV_SEND_SIGNALED;
561 r_io_u_d->sq_wr.sg_list = &r_io_u_d->rdma_sgl;
562 r_io_u_d->sq_wr.num_sge = 1;
563 break;
564 case FIO_RDMA_CHA_RECV:
565 r_io_u_d->rdma_sgl.addr = (uint64_t) (unsigned long)io_u->buf;
566 r_io_u_d->rdma_sgl.lkey = io_u->mr->lkey;
567 r_io_u_d->rdma_sgl.length = io_u->buflen;
568 r_io_u_d->rq_wr.wr_id = r_io_u_d->wr_id;
569 r_io_u_d->rq_wr.sg_list = &r_io_u_d->rdma_sgl;
570 r_io_u_d->rq_wr.num_sge = 1;
571 break;
572 default:
573 log_err("fio: unknown rdma protocol - %d\n", rd->rdma_protocol);
574 break;
575 }
576
577 return 0;
578}
579
580static struct io_u *fio_rdmaio_event(struct thread_data *td, int event)
581{
582 struct rdmaio_data *rd = td->io_ops->data;
583 struct io_u *io_u;
584 int i;
585
586 io_u = rd->io_us_completed[0];
587 for (i = 0; i < rd->io_u_completed_nr - 1; i++)
588 rd->io_us_completed[i] = rd->io_us_completed[i + 1];
589
590 rd->io_u_completed_nr--;
591
592 dprint_io_u(io_u, "fio_rdmaio_event");
593
594 return io_u;
595}
596
597static int fio_rdmaio_getevents(struct thread_data *td, unsigned int min,
598 unsigned int max, const struct timespec *t)
599{
600 struct rdmaio_data *rd = td->io_ops->data;
601 enum ibv_wc_opcode comp_opcode;
602 struct ibv_cq *ev_cq;
603 void *ev_ctx;
604 int ret, r = 0;
605 comp_opcode = IBV_WC_RDMA_WRITE;
606
607 switch (rd->rdma_protocol) {
608 case FIO_RDMA_MEM_WRITE:
609 comp_opcode = IBV_WC_RDMA_WRITE;
610 break;
611 case FIO_RDMA_MEM_READ:
612 comp_opcode = IBV_WC_RDMA_READ;
613 break;
614 case FIO_RDMA_CHA_SEND:
615 comp_opcode = IBV_WC_SEND;
616 break;
617 case FIO_RDMA_CHA_RECV:
618 comp_opcode = IBV_WC_RECV;
619 break;
620 default:
621 log_err("fio: unknown rdma protocol - %d\n", rd->rdma_protocol);
622 break;
623 }
624
625 if (rd->cq_event_num > 0) { /* previous left */
626 rd->cq_event_num--;
627 return 0;
628 }
629
630again:
631 if (ibv_get_cq_event(rd->channel, &ev_cq, &ev_ctx) != 0) {
632 log_err("fio: Failed to get cq event!\n");
633 return -1;
634 }
635 if (ev_cq != rd->cq) {
636 log_err("fio: Unknown CQ!\n");
637 return -1;
638 }
639 if (ibv_req_notify_cq(rd->cq, 0) != 0) {
640 log_err("fio: Failed to set notify!\n");
641 return -1;
642 }
643
644 ret = cq_event_handler(td, comp_opcode);
645 if (ret < 1)
646 goto again;
647
648 ibv_ack_cq_events(rd->cq, ret);
649
650 r += ret;
651 if (r < min)
652 goto again;
653
654 rd->cq_event_num -= r;
655
656 return r;
657}
658
659static int fio_rdmaio_send(struct thread_data *td, struct io_u **io_us,
660 unsigned int nr)
661{
662 struct rdmaio_data *rd = td->io_ops->data;
663 struct ibv_send_wr *bad_wr;
664#if 0
665 enum ibv_wc_opcode comp_opcode;
666 comp_opcode = IBV_WC_RDMA_WRITE;
667#endif
668 int i;
669 long index;
670 struct rdma_io_u_data *r_io_u_d;
671
672 r_io_u_d = NULL;
673
674 for (i = 0; i < nr; i++) {
675 /* RDMA_WRITE or RDMA_READ */
676 switch (rd->rdma_protocol) {
677 case FIO_RDMA_MEM_WRITE:
678 /* compose work request */
679 r_io_u_d = io_us[i]->engine_data;
680 index = __rand(&rd->rand_state) % rd->rmt_nr;
681 r_io_u_d->sq_wr.opcode = IBV_WR_RDMA_WRITE;
682 r_io_u_d->sq_wr.wr.rdma.rkey = rd->rmt_us[index].rkey;
683 r_io_u_d->sq_wr.wr.rdma.remote_addr = \
684 rd->rmt_us[index].buf;
685 r_io_u_d->sq_wr.sg_list->length = io_us[i]->buflen;
686 break;
687 case FIO_RDMA_MEM_READ:
688 /* compose work request */
689 r_io_u_d = io_us[i]->engine_data;
690 index = __rand(&rd->rand_state) % rd->rmt_nr;
691 r_io_u_d->sq_wr.opcode = IBV_WR_RDMA_READ;
692 r_io_u_d->sq_wr.wr.rdma.rkey = rd->rmt_us[index].rkey;
693 r_io_u_d->sq_wr.wr.rdma.remote_addr = \
694 rd->rmt_us[index].buf;
695 r_io_u_d->sq_wr.sg_list->length = io_us[i]->buflen;
696 break;
697 case FIO_RDMA_CHA_SEND:
698 r_io_u_d = io_us[i]->engine_data;
699 r_io_u_d->sq_wr.opcode = IBV_WR_SEND;
700 r_io_u_d->sq_wr.send_flags = IBV_SEND_SIGNALED;
701 break;
702 default:
703 log_err("fio: unknown rdma protocol - %d\n",
704 rd->rdma_protocol);
705 break;
706 }
707
708 if (ibv_post_send(rd->qp, &r_io_u_d->sq_wr, &bad_wr) != 0) {
709 log_err("fio: ibv_post_send fail\n");
710 return -1;
711 }
712
713 dprint_io_u(io_us[i], "fio_rdmaio_send");
714 }
715
716 /* wait for completion
717 rdma_poll_wait(td, comp_opcode); */
718
719 return i;
720}
721
722static int fio_rdmaio_recv(struct thread_data *td, struct io_u **io_us,
723 unsigned int nr)
724{
725 struct rdmaio_data *rd = td->io_ops->data;
726 struct ibv_recv_wr *bad_wr;
727 struct rdma_io_u_data *r_io_u_d;
728 int i;
729
730 i = 0;
731 if (rd->rdma_protocol == FIO_RDMA_CHA_RECV) {
732 /* post io_u into recv queue */
733 for (i = 0; i < nr; i++) {
734 r_io_u_d = io_us[i]->engine_data;
735 if (ibv_post_recv(rd->qp, &r_io_u_d->rq_wr, &bad_wr) !=
736 0) {
737 log_err("fio: ibv_post_recv fail\n");
738 return 1;
739 }
740 }
741 } else if ((rd->rdma_protocol == FIO_RDMA_MEM_READ)
742 || (rd->rdma_protocol == FIO_RDMA_MEM_WRITE)) {
743 /* re-post the rq_wr */
744 if (ibv_post_recv(rd->qp, &rd->rq_wr, &bad_wr) != 0) {
745 log_err("fio: ibv_post_recv fail\n");
746 return 1;
747 }
748
749 rdma_poll_wait(td, IBV_WC_RECV);
750
751 dprint(FD_IO, "fio: recv FINISH message\n");
752 td->done = 1;
753 return 0;
754 }
755
756 return i;
757}
758
759static int fio_rdmaio_queue(struct thread_data *td, struct io_u *io_u)
760{
761 struct rdmaio_data *rd = td->io_ops->data;
762
763 fio_ro_check(td, io_u);
764
765 if (rd->io_u_queued_nr == (int)td->o.iodepth)
766 return FIO_Q_BUSY;
767
768 rd->io_us_queued[rd->io_u_queued_nr] = io_u;
769 rd->io_u_queued_nr++;
770
771 dprint_io_u(io_u, "fio_rdmaio_queue");
772
773 return FIO_Q_QUEUED;
774}
775
776static void fio_rdmaio_queued(struct thread_data *td, struct io_u **io_us,
777 unsigned int nr)
778{
779 struct rdmaio_data *rd = td->io_ops->data;
780 struct timeval now;
781 unsigned int i;
782
783 if (!fio_fill_issue_time(td))
784 return;
785
786 fio_gettime(&now, NULL);
787
788 for (i = 0; i < nr; i++) {
789 struct io_u *io_u = io_us[i];
790
791 /* queued -> flight */
792 rd->io_us_flight[rd->io_u_flight_nr] = io_u;
793 rd->io_u_flight_nr++;
794
795 memcpy(&io_u->issue_time, &now, sizeof(now));
796 io_u_queued(td, io_u);
797 }
798}
799
800static int fio_rdmaio_commit(struct thread_data *td)
801{
802 struct rdmaio_data *rd = td->io_ops->data;
803 struct io_u **io_us;
804 int ret;
805
806 if (!rd->io_us_queued)
807 return 0;
808
809 io_us = rd->io_us_queued;
810 do {
811 /* RDMA_WRITE or RDMA_READ */
812 if (rd->is_client)
813 ret = fio_rdmaio_send(td, io_us, rd->io_u_queued_nr);
814 else if (!rd->is_client)
815 ret = fio_rdmaio_recv(td, io_us, rd->io_u_queued_nr);
816 else
817 ret = 0; /* must be a SYNC */
818
819 if (ret > 0) {
820 fio_rdmaio_queued(td, io_us, ret);
821 io_u_mark_submit(td, ret);
822 rd->io_u_queued_nr -= ret;
823 io_us += ret;
824 ret = 0;
825 } else
826 break;
827 } while (rd->io_u_queued_nr);
828
829 return ret;
830}
831
832static int fio_rdmaio_connect(struct thread_data *td, struct fio_file *f)
833{
834 struct rdmaio_data *rd = td->io_ops->data;
835 struct rdma_conn_param conn_param;
836 struct ibv_send_wr *bad_wr;
837
838 memset(&conn_param, 0, sizeof(conn_param));
839 conn_param.responder_resources = 1;
840 conn_param.initiator_depth = 1;
841 conn_param.retry_count = 10;
842
843 if (rdma_connect(rd->cm_id, &conn_param) != 0) {
844 log_err("fio: rdma_connect fail\n");
845 return 1;
846 }
847
848 if (get_next_channel_event
849 (td, rd->cm_channel, RDMA_CM_EVENT_ESTABLISHED) != 0) {
850 log_err("fio: wait for RDMA_CM_EVENT_ESTABLISHED\n");
851 return 1;
852 }
853
854 /* send task request */
855 rd->send_buf.mode = htonl(rd->rdma_protocol);
856 rd->send_buf.nr = htonl(td->o.iodepth);
857
858 if (ibv_post_send(rd->qp, &rd->sq_wr, &bad_wr) != 0) {
859 log_err("fio: ibv_post_send fail");
860 return 1;
861 }
862
863 rdma_poll_wait(td, IBV_WC_SEND);
864
865 /* wait for remote MR info from server side */
866 rdma_poll_wait(td, IBV_WC_RECV);
867
868 /* In SEND/RECV test, it's a good practice to setup the iodepth of
869 * of the RECV side deeper than that of the SEND side to
870 * avoid RNR (receiver not ready) error. The
871 * SEND side may send so many unsolicited message before
872 * RECV side commits sufficient recv buffers into recv queue.
873 * This may lead to RNR error. Here, SEND side pauses for a while
874 * during which RECV side commits sufficient recv buffers.
875 */
876 usleep(500000);
877
878 return 0;
879}
880
881static int fio_rdmaio_accept(struct thread_data *td, struct fio_file *f)
882{
883 struct rdmaio_data *rd = td->io_ops->data;
884 struct rdma_conn_param conn_param;
885 struct ibv_send_wr *bad_wr;
886
887 /* rdma_accept() - then wait for accept success */
888 memset(&conn_param, 0, sizeof(conn_param));
889 conn_param.responder_resources = 1;
890 conn_param.initiator_depth = 1;
891
892 if (rdma_accept(rd->child_cm_id, &conn_param) != 0) {
893 log_err("fio: rdma_accept\n");
894 return 1;
895 }
896
897 if (get_next_channel_event
898 (td, rd->cm_channel, RDMA_CM_EVENT_ESTABLISHED) != 0) {
899 log_err("fio: wait for RDMA_CM_EVENT_ESTABLISHED\n");
900 return 1;
901 }
902
903 /* wait for request */
904 rdma_poll_wait(td, IBV_WC_RECV);
905
906 if (ibv_post_send(rd->qp, &rd->sq_wr, &bad_wr) != 0) {
907 log_err("fio: ibv_post_send fail");
908 return 1;
909 }
910
911 rdma_poll_wait(td, IBV_WC_SEND);
912
913 return 0;
914}
915
916static int fio_rdmaio_open_file(struct thread_data *td, struct fio_file *f)
917{
918 if (td_read(td))
919 return fio_rdmaio_accept(td, f);
920 else
921 return fio_rdmaio_connect(td, f);
922}
923
924static int fio_rdmaio_close_file(struct thread_data *td, struct fio_file *f)
925{
926 struct rdmaio_data *rd = td->io_ops->data;
927 struct ibv_send_wr *bad_wr;
928
929 /* unregister rdma buffer */
930
931 /*
932 * Client sends notification to the server side
933 */
934 /* refer to: http://linux.die.net/man/7/rdma_cm */
935 if ((rd->is_client == 1) && ((rd->rdma_protocol == FIO_RDMA_MEM_WRITE)
936 || (rd->rdma_protocol ==
937 FIO_RDMA_MEM_READ))) {
938 if (ibv_post_send(rd->qp, &rd->sq_wr, &bad_wr) != 0) {
939 log_err("fio: ibv_post_send fail");
940 return 1;
941 }
942
943 dprint(FD_IO, "fio: close information sent success\n");
944 rdma_poll_wait(td, IBV_WC_SEND);
945 }
946
947 if (rd->is_client == 1)
948 rdma_disconnect(rd->cm_id);
949 else {
950 rdma_disconnect(rd->child_cm_id);
951#if 0
952 rdma_disconnect(rd->cm_id);
953#endif
954 }
955
956#if 0
957 if (get_next_channel_event(td, rd->cm_channel, RDMA_CM_EVENT_DISCONNECTED) != 0) {
958 log_err("fio: wait for RDMA_CM_EVENT_DISCONNECTED\n");
959 return 1;
960 }
961#endif
962
963 ibv_destroy_cq(rd->cq);
964 ibv_destroy_qp(rd->qp);
965
966 if (rd->is_client == 1)
967 rdma_destroy_id(rd->cm_id);
968 else {
969 rdma_destroy_id(rd->child_cm_id);
970 rdma_destroy_id(rd->cm_id);
971 }
972
973 ibv_destroy_comp_channel(rd->channel);
974 ibv_dealloc_pd(rd->pd);
975
976 return 0;
977}
978
979static int fio_rdmaio_setup_connect(struct thread_data *td, const char *host,
980 unsigned short port)
981{
982 struct rdmaio_data *rd = td->io_ops->data;
983 struct ibv_recv_wr *bad_wr;
984 int err;
985
986 rd->addr.sin_family = AF_INET;
987 rd->addr.sin_port = htons(port);
988
989 if (inet_aton(host, &rd->addr.sin_addr) != 1) {
990 struct hostent *hent;
991
992 hent = gethostbyname(host);
993 if (!hent) {
994 td_verror(td, errno, "gethostbyname");
995 return 1;
996 }
997
998 memcpy(&rd->addr.sin_addr, hent->h_addr, 4);
999 }
1000
1001 /* resolve route */
1002 err = rdma_resolve_addr(rd->cm_id, NULL, (struct sockaddr *)&rd->addr, 2000);
1003 if (err != 0) {
1004 log_err("fio: rdma_resolve_addr: %d\n", err);
1005 return 1;
1006 }
1007
1008 err = get_next_channel_event(td, rd->cm_channel, RDMA_CM_EVENT_ADDR_RESOLVED);
1009 if (err != 0) {
1010 log_err("fio: get_next_channel_event: %d\n", err);
1011 return 1;
1012 }
1013
1014 /* resolve route */
1015 err = rdma_resolve_route(rd->cm_id, 2000);
1016 if (err != 0) {
1017 log_err("fio: rdma_resolve_route: %d\n", err);
1018 return 1;
1019 }
1020
1021 err = get_next_channel_event(td, rd->cm_channel, RDMA_CM_EVENT_ROUTE_RESOLVED);
1022 if (err != 0) {
1023 log_err("fio: get_next_channel_event: %d\n", err);
1024 return 1;
1025 }
1026
1027 /* create qp and buffer */
1028 if (fio_rdmaio_setup_qp(td) != 0)
1029 return 1;
1030
1031 if (fio_rdmaio_setup_control_msg_buffers(td) != 0)
1032 return 1;
1033
1034 /* post recv buf */
1035 err = ibv_post_recv(rd->qp, &rd->rq_wr, &bad_wr);
1036 if (err != 0) {
1037 log_err("fio: ibv_post_recv fail: %d\n", err);
1038 return 1;
1039 }
1040
1041 return 0;
1042}
1043
1044static int fio_rdmaio_setup_listen(struct thread_data *td, short port)
1045{
1046 struct rdmaio_data *rd = td->io_ops->data;
1047 struct ibv_recv_wr *bad_wr;
1048
1049 rd->addr.sin_family = AF_INET;
1050 rd->addr.sin_addr.s_addr = htonl(INADDR_ANY);
1051 rd->addr.sin_port = htons(port);
1052
1053 /* rdma_listen */
1054 if (rdma_bind_addr(rd->cm_id, (struct sockaddr *)&rd->addr) != 0) {
1055 log_err("fio: rdma_bind_addr fail\n");
1056 return 1;
1057 }
1058
1059 if (rdma_listen(rd->cm_id, 3) != 0) {
1060 log_err("fio: rdma_listen fail\n");
1061 return 1;
1062 }
1063
1064 /* wait for CONNECT_REQUEST */
1065 if (get_next_channel_event
1066 (td, rd->cm_channel, RDMA_CM_EVENT_CONNECT_REQUEST) != 0) {
1067 log_err("fio: wait for RDMA_CM_EVENT_CONNECT_REQUEST\n");
1068 return 1;
1069 }
1070
1071 if (fio_rdmaio_setup_qp(td) != 0)
1072 return 1;
1073
1074 if (fio_rdmaio_setup_control_msg_buffers(td) != 0)
1075 return 1;
1076
1077 /* post recv buf */
1078 if (ibv_post_recv(rd->qp, &rd->rq_wr, &bad_wr) != 0) {
1079 log_err("fio: ibv_post_recv fail\n");
1080 return 1;
1081 }
1082
1083 return 0;
1084}
1085
1086static int check_set_rlimits(struct thread_data *td)
1087{
1088#ifdef CONFIG_RLIMIT_MEMLOCK
1089 struct rlimit rl;
1090
1091 /* check RLIMIT_MEMLOCK */
1092 if (getrlimit(RLIMIT_MEMLOCK, &rl) != 0) {
1093 log_err("fio: getrlimit fail: %d(%s)\n",
1094 errno, strerror(errno));
1095 return 1;
1096 }
1097
1098 /* soft limit */
1099 if ((rl.rlim_cur != RLIM_INFINITY)
1100 && (rl.rlim_cur < td->orig_buffer_size)) {
1101 log_err("fio: soft RLIMIT_MEMLOCK is: %" PRId64 "\n",
1102 rl.rlim_cur);
1103 log_err("fio: total block size is: %zd\n",
1104 td->orig_buffer_size);
1105 /* try to set larger RLIMIT_MEMLOCK */
1106 rl.rlim_cur = rl.rlim_max;
1107 if (setrlimit(RLIMIT_MEMLOCK, &rl) != 0) {
1108 log_err("fio: setrlimit fail: %d(%s)\n",
1109 errno, strerror(errno));
1110 log_err("fio: you may try enlarge MEMLOCK by root\n");
1111 log_err("# ulimit -l unlimited\n");
1112 return 1;
1113 }
1114 }
1115#endif
1116
1117 return 0;
1118}
1119
1120static int compat_options(struct thread_data *td)
1121{
1122 // The original RDMA engine had an ugly / seperator
1123 // on the filename for it's options. This function
1124 // retains backwards compatibility with it.100
1125
1126 struct rdmaio_options *o = td->eo;
1127 char *modep, *portp;
1128 char *filename = td->o.filename;
1129
1130 if (!filename)
1131 return 0;
1132
1133 portp = strchr(filename, '/');
1134 if (portp == NULL)
1135 return 0;
1136
1137 *portp = '\0';
1138 portp++;
1139
1140 o->port = strtol(portp, NULL, 10);
1141 if (!o->port || o->port > 65535)
1142 goto bad_host;
1143
1144 modep = strchr(portp, '/');
1145 if (modep != NULL) {
1146 *modep = '\0';
1147 modep++;
1148 }
1149
1150 if (modep) {
1151 if (!strncmp("rdma_write", modep, strlen(modep)) ||
1152 !strncmp("RDMA_WRITE", modep, strlen(modep)))
1153 o->verb = FIO_RDMA_MEM_WRITE;
1154 else if (!strncmp("rdma_read", modep, strlen(modep)) ||
1155 !strncmp("RDMA_READ", modep, strlen(modep)))
1156 o->verb = FIO_RDMA_MEM_READ;
1157 else if (!strncmp("send", modep, strlen(modep)) ||
1158 !strncmp("SEND", modep, strlen(modep)))
1159 o->verb = FIO_RDMA_CHA_SEND;
1160 else
1161 goto bad_host;
1162 } else
1163 o->verb = FIO_RDMA_MEM_WRITE;
1164
1165
1166 return 0;
1167
1168bad_host:
1169 log_err("fio: bad rdma host/port/protocol: %s\n", td->o.filename);
1170 return 1;
1171}
1172
1173static int fio_rdmaio_init(struct thread_data *td)
1174{
1175 struct rdmaio_data *rd = td->io_ops->data;
1176 struct rdmaio_options *o = td->eo;
1177 unsigned int max_bs;
1178 int ret, i;
1179
1180 if (td_rw(td)) {
1181 log_err("fio: rdma connections must be read OR write\n");
1182 return 1;
1183 }
1184 if (td_random(td)) {
1185 log_err("fio: RDMA network IO can't be random\n");
1186 return 1;
1187 }
1188
1189 if (compat_options(td))
1190 return 1;
1191
1192 if (!o->port) {
1193 log_err("fio: no port has been specified which is required "
1194 "for the rdma engine\n");
1195 return 1;
1196 }
1197
1198 if (check_set_rlimits(td))
1199 return 1;
1200
1201 rd->rdma_protocol = o->verb;
1202 rd->cq_event_num = 0;
1203
1204 rd->cm_channel = rdma_create_event_channel();
1205 if (!rd->cm_channel) {
1206 log_err("fio: rdma_create_event_channel fail\n");
1207 return 1;
1208 }
1209
1210 ret = rdma_create_id(rd->cm_channel, &rd->cm_id, rd, RDMA_PS_TCP);
1211 if (ret) {
1212 log_err("fio: rdma_create_id fail\n");
1213 return 1;
1214 }
1215
1216 if ((rd->rdma_protocol == FIO_RDMA_MEM_WRITE) ||
1217 (rd->rdma_protocol == FIO_RDMA_MEM_READ)) {
1218 rd->rmt_us =
1219 malloc(FIO_RDMA_MAX_IO_DEPTH * sizeof(struct remote_u));
1220 memset(rd->rmt_us, 0,
1221 FIO_RDMA_MAX_IO_DEPTH * sizeof(struct remote_u));
1222 rd->rmt_nr = 0;
1223 }
1224
1225 rd->io_us_queued = malloc(td->o.iodepth * sizeof(struct io_u *));
1226 memset(rd->io_us_queued, 0, td->o.iodepth * sizeof(struct io_u *));
1227 rd->io_u_queued_nr = 0;
1228
1229 rd->io_us_flight = malloc(td->o.iodepth * sizeof(struct io_u *));
1230 memset(rd->io_us_flight, 0, td->o.iodepth * sizeof(struct io_u *));
1231 rd->io_u_flight_nr = 0;
1232
1233 rd->io_us_completed = malloc(td->o.iodepth * sizeof(struct io_u *));
1234 memset(rd->io_us_completed, 0, td->o.iodepth * sizeof(struct io_u *));
1235 rd->io_u_completed_nr = 0;
1236
1237 if (td_read(td)) { /* READ as the server */
1238 rd->is_client = 0;
1239 /* server rd->rdma_buf_len will be setup after got request */
1240 ret = fio_rdmaio_setup_listen(td, o->port);
1241 } else { /* WRITE as the client */
1242 rd->is_client = 1;
1243 ret = fio_rdmaio_setup_connect(td, td->o.filename, o->port);
1244 }
1245
1246 max_bs = max(td->o.max_bs[DDIR_READ], td->o.max_bs[DDIR_WRITE]);
1247 /* register each io_u in the free list */
1248 for (i = 0; i < td->io_u_freelist.nr; i++) {
1249 struct io_u *io_u = td->io_u_freelist.io_us[i];
1250
1251 io_u->engine_data = malloc(sizeof(struct rdma_io_u_data));
1252 memset(io_u->engine_data, 0, sizeof(struct rdma_io_u_data));
1253 ((struct rdma_io_u_data *)io_u->engine_data)->wr_id = i;
1254
1255 io_u->mr = ibv_reg_mr(rd->pd, io_u->buf, max_bs,
1256 IBV_ACCESS_LOCAL_WRITE |
1257 IBV_ACCESS_REMOTE_READ |
1258 IBV_ACCESS_REMOTE_WRITE);
1259 if (io_u->mr == NULL) {
1260 log_err("fio: ibv_reg_mr io_u failed\n");
1261 return 1;
1262 }
1263
1264 rd->send_buf.rmt_us[i].buf =
1265 htonll((uint64_t) (unsigned long)io_u->buf);
1266 rd->send_buf.rmt_us[i].rkey = htonl(io_u->mr->rkey);
1267 rd->send_buf.rmt_us[i].size = htonl(max_bs);
1268
1269#if 0
1270 log_info("fio: Send rkey %x addr %" PRIx64 " len %d to client\n", io_u->mr->rkey, io_u->buf, max_bs); */
1271#endif
1272 }
1273
1274 rd->send_buf.nr = htonl(i);
1275
1276 return ret;
1277}
1278
1279static void fio_rdmaio_cleanup(struct thread_data *td)
1280{
1281 struct rdmaio_data *rd = td->io_ops->data;
1282
1283 if (rd)
1284 free(rd);
1285}
1286
1287static int fio_rdmaio_setup(struct thread_data *td)
1288{
1289 struct rdmaio_data *rd;
1290
1291 if (!td->files_index) {
1292 add_file(td, td->o.filename ?: "rdma", 0, 0);
1293 td->o.nr_files = td->o.nr_files ?: 1;
1294 td->o.open_files++;
1295 }
1296
1297 if (!td->io_ops->data) {
1298 rd = malloc(sizeof(*rd));
1299
1300 memset(rd, 0, sizeof(*rd));
1301 init_rand_seed(&rd->rand_state, (unsigned int) GOLDEN_RATIO_PRIME, 0);
1302 td->io_ops->data = rd;
1303 }
1304
1305 return 0;
1306}
1307
1308static struct ioengine_ops ioengine_rw = {
1309 .name = "rdma",
1310 .version = FIO_IOOPS_VERSION,
1311 .setup = fio_rdmaio_setup,
1312 .init = fio_rdmaio_init,
1313 .prep = fio_rdmaio_prep,
1314 .queue = fio_rdmaio_queue,
1315 .commit = fio_rdmaio_commit,
1316 .getevents = fio_rdmaio_getevents,
1317 .event = fio_rdmaio_event,
1318 .cleanup = fio_rdmaio_cleanup,
1319 .open_file = fio_rdmaio_open_file,
1320 .close_file = fio_rdmaio_close_file,
1321 .flags = FIO_DISKLESSIO | FIO_UNIDIR | FIO_PIPEIO,
1322 .options = options,
1323 .option_struct_size = sizeof(struct rdmaio_options),
1324};
1325
1326static void fio_init fio_rdmaio_register(void)
1327{
1328 register_ioengine(&ioengine_rw);
1329}
1330
1331static void fio_exit fio_rdmaio_unregister(void)
1332{
1333 unregister_ioengine(&ioengine_rw);
1334}
fio - Flexible IO Tester