1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * RDMA Transport Layer
4 *
5 * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved.
6 * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved.
7 * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved.
8 */
9
10 #undef pr_fmt
11 #define pr_fmt(fmt) KBUILD_MODNAME " L" __stringify(__LINE__) ": " fmt
12
13 #include <linux/module.h>
14 #include <linux/rculist.h>
15 #include <linux/random.h>
16
17 #include "rtrs-clt.h"
18 #include "rtrs-log.h"
19
20 #define RTRS_CONNECT_TIMEOUT_MS 30000
21 /*
22 * Wait a bit before trying to reconnect after a failure
23 * in order to give server time to finish clean up which
24 * leads to "false positives" failed reconnect attempts
25 */
26 #define RTRS_RECONNECT_BACKOFF 1000
27 /*
28 * Wait for additional random time between 0 and 8 seconds
29 * before starting to reconnect to avoid clients reconnecting
30 * all at once in case of a major network outage
31 */
32 #define RTRS_RECONNECT_SEED 8
33
34 #define FIRST_CONN 0x01
35 /* limit to 128 * 4k = 512k max IO */
36 #define RTRS_MAX_SEGMENTS 128
37
38 MODULE_DESCRIPTION("RDMA Transport Client");
39 MODULE_LICENSE("GPL");
40
41 static const struct rtrs_rdma_dev_pd_ops dev_pd_ops;
42 static struct rtrs_rdma_dev_pd dev_pd = {
43 .ops = &dev_pd_ops
44 };
45
46 static struct workqueue_struct *rtrs_wq;
47 static struct class *rtrs_clt_dev_class;
48
rtrs_clt_is_connected(const struct rtrs_clt * clt)49 static inline bool rtrs_clt_is_connected(const struct rtrs_clt *clt)
50 {
51 struct rtrs_clt_sess *sess;
52 bool connected = false;
53
54 rcu_read_lock();
55 list_for_each_entry_rcu(sess, &clt->paths_list, s.entry)
56 connected |= READ_ONCE(sess->state) == RTRS_CLT_CONNECTED;
57 rcu_read_unlock();
58
59 return connected;
60 }
61
62 static struct rtrs_permit *
__rtrs_get_permit(struct rtrs_clt * clt,enum rtrs_clt_con_type con_type)63 __rtrs_get_permit(struct rtrs_clt *clt, enum rtrs_clt_con_type con_type)
64 {
65 size_t max_depth = clt->queue_depth;
66 struct rtrs_permit *permit;
67 int bit;
68
69 /*
70 * Adapted from null_blk get_tag(). Callers from different cpus may
71 * grab the same bit, since find_first_zero_bit is not atomic.
72 * But then the test_and_set_bit_lock will fail for all the
73 * callers but one, so that they will loop again.
74 * This way an explicit spinlock is not required.
75 */
76 do {
77 bit = find_first_zero_bit(clt->permits_map, max_depth);
78 if (bit >= max_depth)
79 return NULL;
80 } while (test_and_set_bit_lock(bit, clt->permits_map));
81
82 permit = get_permit(clt, bit);
83 WARN_ON(permit->mem_id != bit);
84 permit->cpu_id = raw_smp_processor_id();
85 permit->con_type = con_type;
86
87 return permit;
88 }
89
__rtrs_put_permit(struct rtrs_clt * clt,struct rtrs_permit * permit)90 static inline void __rtrs_put_permit(struct rtrs_clt *clt,
91 struct rtrs_permit *permit)
92 {
93 clear_bit_unlock(permit->mem_id, clt->permits_map);
94 }
95
96 /**
97 * rtrs_clt_get_permit() - allocates permit for future RDMA operation
98 * @clt: Current session
99 * @con_type: Type of connection to use with the permit
100 * @can_wait: Wait type
101 *
102 * Description:
103 * Allocates permit for the following RDMA operation. Permit is used
104 * to preallocate all resources and to propagate memory pressure
105 * up earlier.
106 *
107 * Context:
108 * Can sleep if @wait == RTRS_PERMIT_WAIT
109 */
rtrs_clt_get_permit(struct rtrs_clt * clt,enum rtrs_clt_con_type con_type,enum wait_type can_wait)110 struct rtrs_permit *rtrs_clt_get_permit(struct rtrs_clt *clt,
111 enum rtrs_clt_con_type con_type,
112 enum wait_type can_wait)
113 {
114 struct rtrs_permit *permit;
115 DEFINE_WAIT(wait);
116
117 permit = __rtrs_get_permit(clt, con_type);
118 if (permit || !can_wait)
119 return permit;
120
121 do {
122 prepare_to_wait(&clt->permits_wait, &wait,
123 TASK_UNINTERRUPTIBLE);
124 permit = __rtrs_get_permit(clt, con_type);
125 if (permit)
126 break;
127
128 io_schedule();
129 } while (1);
130
131 finish_wait(&clt->permits_wait, &wait);
132
133 return permit;
134 }
135 EXPORT_SYMBOL(rtrs_clt_get_permit);
136
137 /**
138 * rtrs_clt_put_permit() - puts allocated permit
139 * @clt: Current session
140 * @permit: Permit to be freed
141 *
142 * Context:
143 * Does not matter
144 */
rtrs_clt_put_permit(struct rtrs_clt * clt,struct rtrs_permit * permit)145 void rtrs_clt_put_permit(struct rtrs_clt *clt, struct rtrs_permit *permit)
146 {
147 if (WARN_ON(!test_bit(permit->mem_id, clt->permits_map)))
148 return;
149
150 __rtrs_put_permit(clt, permit);
151
152 /*
153 * rtrs_clt_get_permit() adds itself to the &clt->permits_wait list
154 * before calling schedule(). So if rtrs_clt_get_permit() is sleeping
155 * it must have added itself to &clt->permits_wait before
156 * __rtrs_put_permit() finished.
157 * Hence it is safe to guard wake_up() with a waitqueue_active() test.
158 */
159 if (waitqueue_active(&clt->permits_wait))
160 wake_up(&clt->permits_wait);
161 }
162 EXPORT_SYMBOL(rtrs_clt_put_permit);
163
164 /**
165 * rtrs_permit_to_clt_con() - returns RDMA connection pointer by the permit
166 * @sess: client session pointer
167 * @permit: permit for the allocation of the RDMA buffer
168 * Note:
169 * IO connection starts from 1.
170 * 0 connection is for user messages.
171 */
172 static
rtrs_permit_to_clt_con(struct rtrs_clt_sess * sess,struct rtrs_permit * permit)173 struct rtrs_clt_con *rtrs_permit_to_clt_con(struct rtrs_clt_sess *sess,
174 struct rtrs_permit *permit)
175 {
176 int id = 0;
177
178 if (permit->con_type == RTRS_IO_CON)
179 id = (permit->cpu_id % (sess->s.irq_con_num - 1)) + 1;
180
181 return to_clt_con(sess->s.con[id]);
182 }
183
184 /**
185 * rtrs_clt_change_state() - change the session state through session state
186 * machine.
187 *
188 * @sess: client session to change the state of.
189 * @new_state: state to change to.
190 *
191 * returns true if sess's state is changed to new state, otherwise return false.
192 *
193 * Locks:
194 * state_wq lock must be hold.
195 */
rtrs_clt_change_state(struct rtrs_clt_sess * sess,enum rtrs_clt_state new_state)196 static bool rtrs_clt_change_state(struct rtrs_clt_sess *sess,
197 enum rtrs_clt_state new_state)
198 {
199 enum rtrs_clt_state old_state;
200 bool changed = false;
201
202 lockdep_assert_held(&sess->state_wq.lock);
203
204 old_state = sess->state;
205 switch (new_state) {
206 case RTRS_CLT_CONNECTING:
207 switch (old_state) {
208 case RTRS_CLT_RECONNECTING:
209 changed = true;
210 fallthrough;
211 default:
212 break;
213 }
214 break;
215 case RTRS_CLT_RECONNECTING:
216 switch (old_state) {
217 case RTRS_CLT_CONNECTED:
218 case RTRS_CLT_CONNECTING_ERR:
219 case RTRS_CLT_CLOSED:
220 changed = true;
221 fallthrough;
222 default:
223 break;
224 }
225 break;
226 case RTRS_CLT_CONNECTED:
227 switch (old_state) {
228 case RTRS_CLT_CONNECTING:
229 changed = true;
230 fallthrough;
231 default:
232 break;
233 }
234 break;
235 case RTRS_CLT_CONNECTING_ERR:
236 switch (old_state) {
237 case RTRS_CLT_CONNECTING:
238 changed = true;
239 fallthrough;
240 default:
241 break;
242 }
243 break;
244 case RTRS_CLT_CLOSING:
245 switch (old_state) {
246 case RTRS_CLT_CONNECTING:
247 case RTRS_CLT_CONNECTING_ERR:
248 case RTRS_CLT_RECONNECTING:
249 case RTRS_CLT_CONNECTED:
250 changed = true;
251 fallthrough;
252 default:
253 break;
254 }
255 break;
256 case RTRS_CLT_CLOSED:
257 switch (old_state) {
258 case RTRS_CLT_CLOSING:
259 changed = true;
260 fallthrough;
261 default:
262 break;
263 }
264 break;
265 case RTRS_CLT_DEAD:
266 switch (old_state) {
267 case RTRS_CLT_CLOSED:
268 changed = true;
269 fallthrough;
270 default:
271 break;
272 }
273 break;
274 default:
275 break;
276 }
277 if (changed) {
278 sess->state = new_state;
279 wake_up_locked(&sess->state_wq);
280 }
281
282 return changed;
283 }
284
rtrs_clt_change_state_from_to(struct rtrs_clt_sess * sess,enum rtrs_clt_state old_state,enum rtrs_clt_state new_state)285 static bool rtrs_clt_change_state_from_to(struct rtrs_clt_sess *sess,
286 enum rtrs_clt_state old_state,
287 enum rtrs_clt_state new_state)
288 {
289 bool changed = false;
290
291 spin_lock_irq(&sess->state_wq.lock);
292 if (sess->state == old_state)
293 changed = rtrs_clt_change_state(sess, new_state);
294 spin_unlock_irq(&sess->state_wq.lock);
295
296 return changed;
297 }
298
rtrs_rdma_error_recovery(struct rtrs_clt_con * con)299 static void rtrs_rdma_error_recovery(struct rtrs_clt_con *con)
300 {
301 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
302
303 if (rtrs_clt_change_state_from_to(sess,
304 RTRS_CLT_CONNECTED,
305 RTRS_CLT_RECONNECTING)) {
306 struct rtrs_clt *clt = sess->clt;
307 unsigned int delay_ms;
308
309 /*
310 * Normal scenario, reconnect if we were successfully connected
311 */
312 delay_ms = clt->reconnect_delay_sec * 1000;
313 queue_delayed_work(rtrs_wq, &sess->reconnect_dwork,
314 msecs_to_jiffies(delay_ms +
315 prandom_u32() % RTRS_RECONNECT_SEED));
316 } else {
317 /*
318 * Error can happen just on establishing new connection,
319 * so notify waiter with error state, waiter is responsible
320 * for cleaning the rest and reconnect if needed.
321 */
322 rtrs_clt_change_state_from_to(sess,
323 RTRS_CLT_CONNECTING,
324 RTRS_CLT_CONNECTING_ERR);
325 }
326 }
327
rtrs_clt_fast_reg_done(struct ib_cq * cq,struct ib_wc * wc)328 static void rtrs_clt_fast_reg_done(struct ib_cq *cq, struct ib_wc *wc)
329 {
330 struct rtrs_clt_con *con = to_clt_con(wc->qp->qp_context);
331
332 if (wc->status != IB_WC_SUCCESS) {
333 rtrs_err(con->c.sess, "Failed IB_WR_REG_MR: %s\n",
334 ib_wc_status_msg(wc->status));
335 rtrs_rdma_error_recovery(con);
336 }
337 }
338
339 static struct ib_cqe fast_reg_cqe = {
340 .done = rtrs_clt_fast_reg_done
341 };
342
343 static void complete_rdma_req(struct rtrs_clt_io_req *req, int errno,
344 bool notify, bool can_wait);
345
rtrs_clt_inv_rkey_done(struct ib_cq * cq,struct ib_wc * wc)346 static void rtrs_clt_inv_rkey_done(struct ib_cq *cq, struct ib_wc *wc)
347 {
348 struct rtrs_clt_io_req *req =
349 container_of(wc->wr_cqe, typeof(*req), inv_cqe);
350 struct rtrs_clt_con *con = to_clt_con(wc->qp->qp_context);
351
352 if (wc->status != IB_WC_SUCCESS) {
353 rtrs_err(con->c.sess, "Failed IB_WR_LOCAL_INV: %s\n",
354 ib_wc_status_msg(wc->status));
355 rtrs_rdma_error_recovery(con);
356 }
357 req->need_inv = false;
358 if (req->need_inv_comp)
359 complete(&req->inv_comp);
360 else
361 /* Complete request from INV callback */
362 complete_rdma_req(req, req->inv_errno, true, false);
363 }
364
rtrs_inv_rkey(struct rtrs_clt_io_req * req)365 static int rtrs_inv_rkey(struct rtrs_clt_io_req *req)
366 {
367 struct rtrs_clt_con *con = req->con;
368 struct ib_send_wr wr = {
369 .opcode = IB_WR_LOCAL_INV,
370 .wr_cqe = &req->inv_cqe,
371 .send_flags = IB_SEND_SIGNALED,
372 .ex.invalidate_rkey = req->mr->rkey,
373 };
374 req->inv_cqe.done = rtrs_clt_inv_rkey_done;
375
376 return ib_post_send(con->c.qp, &wr, NULL);
377 }
378
complete_rdma_req(struct rtrs_clt_io_req * req,int errno,bool notify,bool can_wait)379 static void complete_rdma_req(struct rtrs_clt_io_req *req, int errno,
380 bool notify, bool can_wait)
381 {
382 struct rtrs_clt_con *con = req->con;
383 struct rtrs_clt_sess *sess;
384 int err;
385
386 if (WARN_ON(!req->in_use))
387 return;
388 if (WARN_ON(!req->con))
389 return;
390 sess = to_clt_sess(con->c.sess);
391
392 if (req->sg_cnt) {
393 if (req->dir == DMA_FROM_DEVICE && req->need_inv) {
394 /*
395 * We are here to invalidate read requests
396 * ourselves. In normal scenario server should
397 * send INV for all read requests, but
398 * we are here, thus two things could happen:
399 *
400 * 1. this is failover, when errno != 0
401 * and can_wait == 1,
402 *
403 * 2. something totally bad happened and
404 * server forgot to send INV, so we
405 * should do that ourselves.
406 */
407
408 if (can_wait) {
409 req->need_inv_comp = true;
410 } else {
411 /* This should be IO path, so always notify */
412 WARN_ON(!notify);
413 /* Save errno for INV callback */
414 req->inv_errno = errno;
415 }
416
417 refcount_inc(&req->ref);
418 err = rtrs_inv_rkey(req);
419 if (err) {
420 rtrs_err(con->c.sess, "Send INV WR key=%#x: %d\n",
421 req->mr->rkey, err);
422 } else if (can_wait) {
423 wait_for_completion(&req->inv_comp);
424 } else {
425 /*
426 * Something went wrong, so request will be
427 * completed from INV callback.
428 */
429 WARN_ON_ONCE(1);
430
431 return;
432 }
433 if (!refcount_dec_and_test(&req->ref))
434 return;
435 }
436 ib_dma_unmap_sg(sess->s.dev->ib_dev, req->sglist,
437 req->sg_cnt, req->dir);
438 }
439 if (!refcount_dec_and_test(&req->ref))
440 return;
441 if (req->mp_policy == MP_POLICY_MIN_INFLIGHT)
442 atomic_dec(&sess->stats->inflight);
443
444 req->in_use = false;
445 req->con = NULL;
446
447 if (errno) {
448 rtrs_err_rl(con->c.sess, "IO request failed: error=%d path=%s [%s:%u] notify=%d\n",
449 errno, kobject_name(&sess->kobj), sess->hca_name,
450 sess->hca_port, notify);
451 }
452
453 if (notify)
454 req->conf(req->priv, errno);
455 }
456
rtrs_post_send_rdma(struct rtrs_clt_con * con,struct rtrs_clt_io_req * req,struct rtrs_rbuf * rbuf,u32 off,u32 imm,struct ib_send_wr * wr)457 static int rtrs_post_send_rdma(struct rtrs_clt_con *con,
458 struct rtrs_clt_io_req *req,
459 struct rtrs_rbuf *rbuf, u32 off,
460 u32 imm, struct ib_send_wr *wr)
461 {
462 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
463 enum ib_send_flags flags;
464 struct ib_sge sge;
465
466 if (!req->sg_size) {
467 rtrs_wrn(con->c.sess,
468 "Doing RDMA Write failed, no data supplied\n");
469 return -EINVAL;
470 }
471
472 /* user data and user message in the first list element */
473 sge.addr = req->iu->dma_addr;
474 sge.length = req->sg_size;
475 sge.lkey = sess->s.dev->ib_pd->local_dma_lkey;
476
477 /*
478 * From time to time we have to post signalled sends,
479 * or send queue will fill up and only QP reset can help.
480 */
481 flags = atomic_inc_return(&con->c.wr_cnt) % sess->s.signal_interval ?
482 0 : IB_SEND_SIGNALED;
483
484 ib_dma_sync_single_for_device(sess->s.dev->ib_dev, req->iu->dma_addr,
485 req->sg_size, DMA_TO_DEVICE);
486
487 return rtrs_iu_post_rdma_write_imm(&con->c, req->iu, &sge, 1,
488 rbuf->rkey, rbuf->addr + off,
489 imm, flags, wr, NULL);
490 }
491
process_io_rsp(struct rtrs_clt_sess * sess,u32 msg_id,s16 errno,bool w_inval)492 static void process_io_rsp(struct rtrs_clt_sess *sess, u32 msg_id,
493 s16 errno, bool w_inval)
494 {
495 struct rtrs_clt_io_req *req;
496
497 if (WARN_ON(msg_id >= sess->queue_depth))
498 return;
499
500 req = &sess->reqs[msg_id];
501 /* Drop need_inv if server responded with send with invalidation */
502 req->need_inv &= !w_inval;
503 complete_rdma_req(req, errno, true, false);
504 }
505
rtrs_clt_recv_done(struct rtrs_clt_con * con,struct ib_wc * wc)506 static void rtrs_clt_recv_done(struct rtrs_clt_con *con, struct ib_wc *wc)
507 {
508 struct rtrs_iu *iu;
509 int err;
510 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
511
512 WARN_ON((sess->flags & RTRS_MSG_NEW_RKEY_F) == 0);
513 iu = container_of(wc->wr_cqe, struct rtrs_iu,
514 cqe);
515 err = rtrs_iu_post_recv(&con->c, iu);
516 if (err) {
517 rtrs_err(con->c.sess, "post iu failed %d\n", err);
518 rtrs_rdma_error_recovery(con);
519 }
520 }
521
rtrs_clt_rkey_rsp_done(struct rtrs_clt_con * con,struct ib_wc * wc)522 static void rtrs_clt_rkey_rsp_done(struct rtrs_clt_con *con, struct ib_wc *wc)
523 {
524 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
525 struct rtrs_msg_rkey_rsp *msg;
526 u32 imm_type, imm_payload;
527 bool w_inval = false;
528 struct rtrs_iu *iu;
529 u32 buf_id;
530 int err;
531
532 WARN_ON((sess->flags & RTRS_MSG_NEW_RKEY_F) == 0);
533
534 iu = container_of(wc->wr_cqe, struct rtrs_iu, cqe);
535
536 if (wc->byte_len < sizeof(*msg)) {
537 rtrs_err(con->c.sess, "rkey response is malformed: size %d\n",
538 wc->byte_len);
539 goto out;
540 }
541 ib_dma_sync_single_for_cpu(sess->s.dev->ib_dev, iu->dma_addr,
542 iu->size, DMA_FROM_DEVICE);
543 msg = iu->buf;
544 if (le16_to_cpu(msg->type) != RTRS_MSG_RKEY_RSP) {
545 rtrs_err(sess->clt, "rkey response is malformed: type %d\n",
546 le16_to_cpu(msg->type));
547 goto out;
548 }
549 buf_id = le16_to_cpu(msg->buf_id);
550 if (WARN_ON(buf_id >= sess->queue_depth))
551 goto out;
552
553 rtrs_from_imm(be32_to_cpu(wc->ex.imm_data), &imm_type, &imm_payload);
554 if (imm_type == RTRS_IO_RSP_IMM ||
555 imm_type == RTRS_IO_RSP_W_INV_IMM) {
556 u32 msg_id;
557
558 w_inval = (imm_type == RTRS_IO_RSP_W_INV_IMM);
559 rtrs_from_io_rsp_imm(imm_payload, &msg_id, &err);
560
561 if (WARN_ON(buf_id != msg_id))
562 goto out;
563 sess->rbufs[buf_id].rkey = le32_to_cpu(msg->rkey);
564 process_io_rsp(sess, msg_id, err, w_inval);
565 }
566 ib_dma_sync_single_for_device(sess->s.dev->ib_dev, iu->dma_addr,
567 iu->size, DMA_FROM_DEVICE);
568 return rtrs_clt_recv_done(con, wc);
569 out:
570 rtrs_rdma_error_recovery(con);
571 }
572
573 static void rtrs_clt_rdma_done(struct ib_cq *cq, struct ib_wc *wc);
574
575 static struct ib_cqe io_comp_cqe = {
576 .done = rtrs_clt_rdma_done
577 };
578
579 /*
580 * Post x2 empty WRs: first is for this RDMA with IMM,
581 * second is for RECV with INV, which happened earlier.
582 */
rtrs_post_recv_empty_x2(struct rtrs_con * con,struct ib_cqe * cqe)583 static int rtrs_post_recv_empty_x2(struct rtrs_con *con, struct ib_cqe *cqe)
584 {
585 struct ib_recv_wr wr_arr[2], *wr;
586 int i;
587
588 memset(wr_arr, 0, sizeof(wr_arr));
589 for (i = 0; i < ARRAY_SIZE(wr_arr); i++) {
590 wr = &wr_arr[i];
591 wr->wr_cqe = cqe;
592 if (i)
593 /* Chain backwards */
594 wr->next = &wr_arr[i - 1];
595 }
596
597 return ib_post_recv(con->qp, wr, NULL);
598 }
599
rtrs_clt_rdma_done(struct ib_cq * cq,struct ib_wc * wc)600 static void rtrs_clt_rdma_done(struct ib_cq *cq, struct ib_wc *wc)
601 {
602 struct rtrs_clt_con *con = to_clt_con(wc->qp->qp_context);
603 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
604 u32 imm_type, imm_payload;
605 bool w_inval = false;
606 int err;
607
608 if (wc->status != IB_WC_SUCCESS) {
609 if (wc->status != IB_WC_WR_FLUSH_ERR) {
610 rtrs_err(sess->clt, "RDMA failed: %s\n",
611 ib_wc_status_msg(wc->status));
612 rtrs_rdma_error_recovery(con);
613 }
614 return;
615 }
616 rtrs_clt_update_wc_stats(con);
617
618 switch (wc->opcode) {
619 case IB_WC_RECV_RDMA_WITH_IMM:
620 /*
621 * post_recv() RDMA write completions of IO reqs (read/write)
622 * and hb
623 */
624 if (WARN_ON(wc->wr_cqe->done != rtrs_clt_rdma_done))
625 return;
626 rtrs_from_imm(be32_to_cpu(wc->ex.imm_data),
627 &imm_type, &imm_payload);
628 if (imm_type == RTRS_IO_RSP_IMM ||
629 imm_type == RTRS_IO_RSP_W_INV_IMM) {
630 u32 msg_id;
631
632 w_inval = (imm_type == RTRS_IO_RSP_W_INV_IMM);
633 rtrs_from_io_rsp_imm(imm_payload, &msg_id, &err);
634
635 process_io_rsp(sess, msg_id, err, w_inval);
636 } else if (imm_type == RTRS_HB_MSG_IMM) {
637 WARN_ON(con->c.cid);
638 rtrs_send_hb_ack(&sess->s);
639 if (sess->flags & RTRS_MSG_NEW_RKEY_F)
640 return rtrs_clt_recv_done(con, wc);
641 } else if (imm_type == RTRS_HB_ACK_IMM) {
642 WARN_ON(con->c.cid);
643 sess->s.hb_missed_cnt = 0;
644 sess->s.hb_cur_latency =
645 ktime_sub(ktime_get(), sess->s.hb_last_sent);
646 if (sess->flags & RTRS_MSG_NEW_RKEY_F)
647 return rtrs_clt_recv_done(con, wc);
648 } else {
649 rtrs_wrn(con->c.sess, "Unknown IMM type %u\n",
650 imm_type);
651 }
652 if (w_inval)
653 /*
654 * Post x2 empty WRs: first is for this RDMA with IMM,
655 * second is for RECV with INV, which happened earlier.
656 */
657 err = rtrs_post_recv_empty_x2(&con->c, &io_comp_cqe);
658 else
659 err = rtrs_post_recv_empty(&con->c, &io_comp_cqe);
660 if (err) {
661 rtrs_err(con->c.sess, "rtrs_post_recv_empty(): %d\n",
662 err);
663 rtrs_rdma_error_recovery(con);
664 }
665 break;
666 case IB_WC_RECV:
667 /*
668 * Key invalidations from server side
669 */
670 WARN_ON(!(wc->wc_flags & IB_WC_WITH_INVALIDATE ||
671 wc->wc_flags & IB_WC_WITH_IMM));
672 WARN_ON(wc->wr_cqe->done != rtrs_clt_rdma_done);
673 if (sess->flags & RTRS_MSG_NEW_RKEY_F) {
674 if (wc->wc_flags & IB_WC_WITH_INVALIDATE)
675 return rtrs_clt_recv_done(con, wc);
676
677 return rtrs_clt_rkey_rsp_done(con, wc);
678 }
679 break;
680 case IB_WC_RDMA_WRITE:
681 /*
682 * post_send() RDMA write completions of IO reqs (read/write)
683 * and hb.
684 */
685 break;
686
687 default:
688 rtrs_wrn(sess->clt, "Unexpected WC type: %d\n", wc->opcode);
689 return;
690 }
691 }
692
post_recv_io(struct rtrs_clt_con * con,size_t q_size)693 static int post_recv_io(struct rtrs_clt_con *con, size_t q_size)
694 {
695 int err, i;
696 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
697
698 for (i = 0; i < q_size; i++) {
699 if (sess->flags & RTRS_MSG_NEW_RKEY_F) {
700 struct rtrs_iu *iu = &con->rsp_ius[i];
701
702 err = rtrs_iu_post_recv(&con->c, iu);
703 } else {
704 err = rtrs_post_recv_empty(&con->c, &io_comp_cqe);
705 }
706 if (err)
707 return err;
708 }
709
710 return 0;
711 }
712
post_recv_sess(struct rtrs_clt_sess * sess)713 static int post_recv_sess(struct rtrs_clt_sess *sess)
714 {
715 size_t q_size = 0;
716 int err, cid;
717
718 for (cid = 0; cid < sess->s.con_num; cid++) {
719 if (cid == 0)
720 q_size = SERVICE_CON_QUEUE_DEPTH;
721 else
722 q_size = sess->queue_depth;
723
724 /*
725 * x2 for RDMA read responses + FR key invalidations,
726 * RDMA writes do not require any FR registrations.
727 */
728 q_size *= 2;
729
730 err = post_recv_io(to_clt_con(sess->s.con[cid]), q_size);
731 if (err) {
732 rtrs_err(sess->clt, "post_recv_io(), err: %d\n", err);
733 return err;
734 }
735 }
736
737 return 0;
738 }
739
740 struct path_it {
741 int i;
742 struct list_head skip_list;
743 struct rtrs_clt *clt;
744 struct rtrs_clt_sess *(*next_path)(struct path_it *it);
745 };
746
747 /**
748 * list_next_or_null_rr_rcu - get next list element in round-robin fashion.
749 * @head: the head for the list.
750 * @ptr: the list head to take the next element from.
751 * @type: the type of the struct this is embedded in.
752 * @memb: the name of the list_head within the struct.
753 *
754 * Next element returned in round-robin fashion, i.e. head will be skipped,
755 * but if list is observed as empty, NULL will be returned.
756 *
757 * This primitive may safely run concurrently with the _rcu list-mutation
758 * primitives such as list_add_rcu() as long as it's guarded by rcu_read_lock().
759 */
760 #define list_next_or_null_rr_rcu(head, ptr, type, memb) \
761 ({ \
762 list_next_or_null_rcu(head, ptr, type, memb) ?: \
763 list_next_or_null_rcu(head, READ_ONCE((ptr)->next), \
764 type, memb); \
765 })
766
767 /**
768 * get_next_path_rr() - Returns path in round-robin fashion.
769 * @it: the path pointer
770 *
771 * Related to @MP_POLICY_RR
772 *
773 * Locks:
774 * rcu_read_lock() must be hold.
775 */
get_next_path_rr(struct path_it * it)776 static struct rtrs_clt_sess *get_next_path_rr(struct path_it *it)
777 {
778 struct rtrs_clt_sess __rcu **ppcpu_path;
779 struct rtrs_clt_sess *path;
780 struct rtrs_clt *clt;
781
782 clt = it->clt;
783
784 /*
785 * Here we use two RCU objects: @paths_list and @pcpu_path
786 * pointer. See rtrs_clt_remove_path_from_arr() for details
787 * how that is handled.
788 */
789
790 ppcpu_path = this_cpu_ptr(clt->pcpu_path);
791 path = rcu_dereference(*ppcpu_path);
792 if (!path)
793 path = list_first_or_null_rcu(&clt->paths_list,
794 typeof(*path), s.entry);
795 else
796 path = list_next_or_null_rr_rcu(&clt->paths_list,
797 &path->s.entry,
798 typeof(*path),
799 s.entry);
800 rcu_assign_pointer(*ppcpu_path, path);
801
802 return path;
803 }
804
805 /**
806 * get_next_path_min_inflight() - Returns path with minimal inflight count.
807 * @it: the path pointer
808 *
809 * Related to @MP_POLICY_MIN_INFLIGHT
810 *
811 * Locks:
812 * rcu_read_lock() must be hold.
813 */
get_next_path_min_inflight(struct path_it * it)814 static struct rtrs_clt_sess *get_next_path_min_inflight(struct path_it *it)
815 {
816 struct rtrs_clt_sess *min_path = NULL;
817 struct rtrs_clt *clt = it->clt;
818 struct rtrs_clt_sess *sess;
819 int min_inflight = INT_MAX;
820 int inflight;
821
822 list_for_each_entry_rcu(sess, &clt->paths_list, s.entry) {
823 if (READ_ONCE(sess->state) != RTRS_CLT_CONNECTED)
824 continue;
825
826 if (!list_empty(raw_cpu_ptr(sess->mp_skip_entry)))
827 continue;
828
829 inflight = atomic_read(&sess->stats->inflight);
830
831 if (inflight < min_inflight) {
832 min_inflight = inflight;
833 min_path = sess;
834 }
835 }
836
837 /*
838 * add the path to the skip list, so that next time we can get
839 * a different one
840 */
841 if (min_path)
842 list_add(raw_cpu_ptr(min_path->mp_skip_entry), &it->skip_list);
843
844 return min_path;
845 }
846
847 /**
848 * get_next_path_min_latency() - Returns path with minimal latency.
849 * @it: the path pointer
850 *
851 * Return: a path with the lowest latency or NULL if all paths are tried
852 *
853 * Locks:
854 * rcu_read_lock() must be hold.
855 *
856 * Related to @MP_POLICY_MIN_LATENCY
857 *
858 * This DOES skip an already-tried path.
859 * There is a skip-list to skip a path if the path has tried but failed.
860 * It will try the minimum latency path and then the second minimum latency
861 * path and so on. Finally it will return NULL if all paths are tried.
862 * Therefore the caller MUST check the returned
863 * path is NULL and trigger the IO error.
864 */
get_next_path_min_latency(struct path_it * it)865 static struct rtrs_clt_sess *get_next_path_min_latency(struct path_it *it)
866 {
867 struct rtrs_clt_sess *min_path = NULL;
868 struct rtrs_clt *clt = it->clt;
869 struct rtrs_clt_sess *sess;
870 ktime_t min_latency = INT_MAX;
871 ktime_t latency;
872
873 list_for_each_entry_rcu(sess, &clt->paths_list, s.entry) {
874 if (READ_ONCE(sess->state) != RTRS_CLT_CONNECTED)
875 continue;
876
877 if (!list_empty(raw_cpu_ptr(sess->mp_skip_entry)))
878 continue;
879
880 latency = sess->s.hb_cur_latency;
881
882 if (latency < min_latency) {
883 min_latency = latency;
884 min_path = sess;
885 }
886 }
887
888 /*
889 * add the path to the skip list, so that next time we can get
890 * a different one
891 */
892 if (min_path)
893 list_add(raw_cpu_ptr(min_path->mp_skip_entry), &it->skip_list);
894
895 return min_path;
896 }
897
path_it_init(struct path_it * it,struct rtrs_clt * clt)898 static inline void path_it_init(struct path_it *it, struct rtrs_clt *clt)
899 {
900 INIT_LIST_HEAD(&it->skip_list);
901 it->clt = clt;
902 it->i = 0;
903
904 if (clt->mp_policy == MP_POLICY_RR)
905 it->next_path = get_next_path_rr;
906 else if (clt->mp_policy == MP_POLICY_MIN_INFLIGHT)
907 it->next_path = get_next_path_min_inflight;
908 else
909 it->next_path = get_next_path_min_latency;
910 }
911
path_it_deinit(struct path_it * it)912 static inline void path_it_deinit(struct path_it *it)
913 {
914 struct list_head *skip, *tmp;
915 /*
916 * The skip_list is used only for the MIN_INFLIGHT policy.
917 * We need to remove paths from it, so that next IO can insert
918 * paths (->mp_skip_entry) into a skip_list again.
919 */
920 list_for_each_safe(skip, tmp, &it->skip_list)
921 list_del_init(skip);
922 }
923
924 /**
925 * rtrs_clt_init_req() - Initialize an rtrs_clt_io_req holding information
926 * about an inflight IO.
927 * The user buffer holding user control message (not data) is copied into
928 * the corresponding buffer of rtrs_iu (req->iu->buf), which later on will
929 * also hold the control message of rtrs.
930 * @req: an io request holding information about IO.
931 * @sess: client session
932 * @conf: conformation callback function to notify upper layer.
933 * @permit: permit for allocation of RDMA remote buffer
934 * @priv: private pointer
935 * @vec: kernel vector containing control message
936 * @usr_len: length of the user message
937 * @sg: scater list for IO data
938 * @sg_cnt: number of scater list entries
939 * @data_len: length of the IO data
940 * @dir: direction of the IO.
941 */
rtrs_clt_init_req(struct rtrs_clt_io_req * req,struct rtrs_clt_sess * sess,void (* conf)(void * priv,int errno),struct rtrs_permit * permit,void * priv,const struct kvec * vec,size_t usr_len,struct scatterlist * sg,size_t sg_cnt,size_t data_len,int dir)942 static void rtrs_clt_init_req(struct rtrs_clt_io_req *req,
943 struct rtrs_clt_sess *sess,
944 void (*conf)(void *priv, int errno),
945 struct rtrs_permit *permit, void *priv,
946 const struct kvec *vec, size_t usr_len,
947 struct scatterlist *sg, size_t sg_cnt,
948 size_t data_len, int dir)
949 {
950 struct iov_iter iter;
951 size_t len;
952
953 req->permit = permit;
954 req->in_use = true;
955 req->usr_len = usr_len;
956 req->data_len = data_len;
957 req->sglist = sg;
958 req->sg_cnt = sg_cnt;
959 req->priv = priv;
960 req->dir = dir;
961 req->con = rtrs_permit_to_clt_con(sess, permit);
962 req->conf = conf;
963 req->need_inv = false;
964 req->need_inv_comp = false;
965 req->inv_errno = 0;
966 refcount_set(&req->ref, 1);
967 req->mp_policy = sess->clt->mp_policy;
968
969 iov_iter_kvec(&iter, READ, vec, 1, usr_len);
970 len = _copy_from_iter(req->iu->buf, usr_len, &iter);
971 WARN_ON(len != usr_len);
972
973 reinit_completion(&req->inv_comp);
974 }
975
976 static struct rtrs_clt_io_req *
rtrs_clt_get_req(struct rtrs_clt_sess * sess,void (* conf)(void * priv,int errno),struct rtrs_permit * permit,void * priv,const struct kvec * vec,size_t usr_len,struct scatterlist * sg,size_t sg_cnt,size_t data_len,int dir)977 rtrs_clt_get_req(struct rtrs_clt_sess *sess,
978 void (*conf)(void *priv, int errno),
979 struct rtrs_permit *permit, void *priv,
980 const struct kvec *vec, size_t usr_len,
981 struct scatterlist *sg, size_t sg_cnt,
982 size_t data_len, int dir)
983 {
984 struct rtrs_clt_io_req *req;
985
986 req = &sess->reqs[permit->mem_id];
987 rtrs_clt_init_req(req, sess, conf, permit, priv, vec, usr_len,
988 sg, sg_cnt, data_len, dir);
989 return req;
990 }
991
992 static struct rtrs_clt_io_req *
rtrs_clt_get_copy_req(struct rtrs_clt_sess * alive_sess,struct rtrs_clt_io_req * fail_req)993 rtrs_clt_get_copy_req(struct rtrs_clt_sess *alive_sess,
994 struct rtrs_clt_io_req *fail_req)
995 {
996 struct rtrs_clt_io_req *req;
997 struct kvec vec = {
998 .iov_base = fail_req->iu->buf,
999 .iov_len = fail_req->usr_len
1000 };
1001
1002 req = &alive_sess->reqs[fail_req->permit->mem_id];
1003 rtrs_clt_init_req(req, alive_sess, fail_req->conf, fail_req->permit,
1004 fail_req->priv, &vec, fail_req->usr_len,
1005 fail_req->sglist, fail_req->sg_cnt,
1006 fail_req->data_len, fail_req->dir);
1007 return req;
1008 }
1009
rtrs_post_rdma_write_sg(struct rtrs_clt_con * con,struct rtrs_clt_io_req * req,struct rtrs_rbuf * rbuf,bool fr_en,u32 size,u32 imm,struct ib_send_wr * wr,struct ib_send_wr * tail)1010 static int rtrs_post_rdma_write_sg(struct rtrs_clt_con *con,
1011 struct rtrs_clt_io_req *req,
1012 struct rtrs_rbuf *rbuf, bool fr_en,
1013 u32 size, u32 imm, struct ib_send_wr *wr,
1014 struct ib_send_wr *tail)
1015 {
1016 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
1017 struct ib_sge *sge = req->sge;
1018 enum ib_send_flags flags;
1019 struct scatterlist *sg;
1020 size_t num_sge;
1021 int i;
1022 struct ib_send_wr *ptail = NULL;
1023
1024 if (fr_en) {
1025 i = 0;
1026 sge[i].addr = req->mr->iova;
1027 sge[i].length = req->mr->length;
1028 sge[i].lkey = req->mr->lkey;
1029 i++;
1030 num_sge = 2;
1031 ptail = tail;
1032 } else {
1033 for_each_sg(req->sglist, sg, req->sg_cnt, i) {
1034 sge[i].addr = sg_dma_address(sg);
1035 sge[i].length = sg_dma_len(sg);
1036 sge[i].lkey = sess->s.dev->ib_pd->local_dma_lkey;
1037 }
1038 num_sge = 1 + req->sg_cnt;
1039 }
1040 sge[i].addr = req->iu->dma_addr;
1041 sge[i].length = size;
1042 sge[i].lkey = sess->s.dev->ib_pd->local_dma_lkey;
1043
1044 /*
1045 * From time to time we have to post signalled sends,
1046 * or send queue will fill up and only QP reset can help.
1047 */
1048 flags = atomic_inc_return(&con->c.wr_cnt) % sess->s.signal_interval ?
1049 0 : IB_SEND_SIGNALED;
1050
1051 ib_dma_sync_single_for_device(sess->s.dev->ib_dev, req->iu->dma_addr,
1052 size, DMA_TO_DEVICE);
1053
1054 return rtrs_iu_post_rdma_write_imm(&con->c, req->iu, sge, num_sge,
1055 rbuf->rkey, rbuf->addr, imm,
1056 flags, wr, ptail);
1057 }
1058
rtrs_map_sg_fr(struct rtrs_clt_io_req * req,size_t count)1059 static int rtrs_map_sg_fr(struct rtrs_clt_io_req *req, size_t count)
1060 {
1061 int nr;
1062
1063 /* Align the MR to a 4K page size to match the block virt boundary */
1064 nr = ib_map_mr_sg(req->mr, req->sglist, count, NULL, SZ_4K);
1065 if (nr < 0)
1066 return nr;
1067 if (nr < req->sg_cnt)
1068 return -EINVAL;
1069 ib_update_fast_reg_key(req->mr, ib_inc_rkey(req->mr->rkey));
1070
1071 return nr;
1072 }
1073
rtrs_clt_write_req(struct rtrs_clt_io_req * req)1074 static int rtrs_clt_write_req(struct rtrs_clt_io_req *req)
1075 {
1076 struct rtrs_clt_con *con = req->con;
1077 struct rtrs_sess *s = con->c.sess;
1078 struct rtrs_clt_sess *sess = to_clt_sess(s);
1079 struct rtrs_msg_rdma_write *msg;
1080
1081 struct rtrs_rbuf *rbuf;
1082 int ret, count = 0;
1083 u32 imm, buf_id;
1084 struct ib_reg_wr rwr;
1085 struct ib_send_wr inv_wr;
1086 struct ib_send_wr *wr = NULL;
1087 bool fr_en = false;
1088
1089 const size_t tsize = sizeof(*msg) + req->data_len + req->usr_len;
1090
1091 if (tsize > sess->chunk_size) {
1092 rtrs_wrn(s, "Write request failed, size too big %zu > %d\n",
1093 tsize, sess->chunk_size);
1094 return -EMSGSIZE;
1095 }
1096 if (req->sg_cnt) {
1097 count = ib_dma_map_sg(sess->s.dev->ib_dev, req->sglist,
1098 req->sg_cnt, req->dir);
1099 if (!count) {
1100 rtrs_wrn(s, "Write request failed, map failed\n");
1101 return -EINVAL;
1102 }
1103 }
1104 /* put rtrs msg after sg and user message */
1105 msg = req->iu->buf + req->usr_len;
1106 msg->type = cpu_to_le16(RTRS_MSG_WRITE);
1107 msg->usr_len = cpu_to_le16(req->usr_len);
1108
1109 /* rtrs message on server side will be after user data and message */
1110 imm = req->permit->mem_off + req->data_len + req->usr_len;
1111 imm = rtrs_to_io_req_imm(imm);
1112 buf_id = req->permit->mem_id;
1113 req->sg_size = tsize;
1114 rbuf = &sess->rbufs[buf_id];
1115
1116 if (count) {
1117 ret = rtrs_map_sg_fr(req, count);
1118 if (ret < 0) {
1119 rtrs_err_rl(s,
1120 "Write request failed, failed to map fast reg. data, err: %d\n",
1121 ret);
1122 ib_dma_unmap_sg(sess->s.dev->ib_dev, req->sglist,
1123 req->sg_cnt, req->dir);
1124 return ret;
1125 }
1126 inv_wr = (struct ib_send_wr) {
1127 .opcode = IB_WR_LOCAL_INV,
1128 .wr_cqe = &req->inv_cqe,
1129 .send_flags = IB_SEND_SIGNALED,
1130 .ex.invalidate_rkey = req->mr->rkey,
1131 };
1132 req->inv_cqe.done = rtrs_clt_inv_rkey_done;
1133 rwr = (struct ib_reg_wr) {
1134 .wr.opcode = IB_WR_REG_MR,
1135 .wr.wr_cqe = &fast_reg_cqe,
1136 .mr = req->mr,
1137 .key = req->mr->rkey,
1138 .access = (IB_ACCESS_LOCAL_WRITE),
1139 };
1140 wr = &rwr.wr;
1141 fr_en = true;
1142 refcount_inc(&req->ref);
1143 }
1144 /*
1145 * Update stats now, after request is successfully sent it is not
1146 * safe anymore to touch it.
1147 */
1148 rtrs_clt_update_all_stats(req, WRITE);
1149
1150 ret = rtrs_post_rdma_write_sg(req->con, req, rbuf, fr_en,
1151 req->usr_len + sizeof(*msg),
1152 imm, wr, &inv_wr);
1153 if (ret) {
1154 rtrs_err_rl(s,
1155 "Write request failed: error=%d path=%s [%s:%u]\n",
1156 ret, kobject_name(&sess->kobj), sess->hca_name,
1157 sess->hca_port);
1158 if (req->mp_policy == MP_POLICY_MIN_INFLIGHT)
1159 atomic_dec(&sess->stats->inflight);
1160 if (req->sg_cnt)
1161 ib_dma_unmap_sg(sess->s.dev->ib_dev, req->sglist,
1162 req->sg_cnt, req->dir);
1163 }
1164
1165 return ret;
1166 }
1167
rtrs_clt_read_req(struct rtrs_clt_io_req * req)1168 static int rtrs_clt_read_req(struct rtrs_clt_io_req *req)
1169 {
1170 struct rtrs_clt_con *con = req->con;
1171 struct rtrs_sess *s = con->c.sess;
1172 struct rtrs_clt_sess *sess = to_clt_sess(s);
1173 struct rtrs_msg_rdma_read *msg;
1174 struct rtrs_ib_dev *dev = sess->s.dev;
1175
1176 struct ib_reg_wr rwr;
1177 struct ib_send_wr *wr = NULL;
1178
1179 int ret, count = 0;
1180 u32 imm, buf_id;
1181
1182 const size_t tsize = sizeof(*msg) + req->data_len + req->usr_len;
1183
1184 if (tsize > sess->chunk_size) {
1185 rtrs_wrn(s,
1186 "Read request failed, message size is %zu, bigger than CHUNK_SIZE %d\n",
1187 tsize, sess->chunk_size);
1188 return -EMSGSIZE;
1189 }
1190
1191 if (req->sg_cnt) {
1192 count = ib_dma_map_sg(dev->ib_dev, req->sglist, req->sg_cnt,
1193 req->dir);
1194 if (!count) {
1195 rtrs_wrn(s,
1196 "Read request failed, dma map failed\n");
1197 return -EINVAL;
1198 }
1199 }
1200 /* put our message into req->buf after user message*/
1201 msg = req->iu->buf + req->usr_len;
1202 msg->type = cpu_to_le16(RTRS_MSG_READ);
1203 msg->usr_len = cpu_to_le16(req->usr_len);
1204
1205 if (count) {
1206 ret = rtrs_map_sg_fr(req, count);
1207 if (ret < 0) {
1208 rtrs_err_rl(s,
1209 "Read request failed, failed to map fast reg. data, err: %d\n",
1210 ret);
1211 ib_dma_unmap_sg(dev->ib_dev, req->sglist, req->sg_cnt,
1212 req->dir);
1213 return ret;
1214 }
1215 rwr = (struct ib_reg_wr) {
1216 .wr.opcode = IB_WR_REG_MR,
1217 .wr.wr_cqe = &fast_reg_cqe,
1218 .mr = req->mr,
1219 .key = req->mr->rkey,
1220 .access = (IB_ACCESS_LOCAL_WRITE |
1221 IB_ACCESS_REMOTE_WRITE),
1222 };
1223 wr = &rwr.wr;
1224
1225 msg->sg_cnt = cpu_to_le16(1);
1226 msg->flags = cpu_to_le16(RTRS_MSG_NEED_INVAL_F);
1227
1228 msg->desc[0].addr = cpu_to_le64(req->mr->iova);
1229 msg->desc[0].key = cpu_to_le32(req->mr->rkey);
1230 msg->desc[0].len = cpu_to_le32(req->mr->length);
1231
1232 /* Further invalidation is required */
1233 req->need_inv = !!RTRS_MSG_NEED_INVAL_F;
1234
1235 } else {
1236 msg->sg_cnt = 0;
1237 msg->flags = 0;
1238 }
1239 /*
1240 * rtrs message will be after the space reserved for disk data and
1241 * user message
1242 */
1243 imm = req->permit->mem_off + req->data_len + req->usr_len;
1244 imm = rtrs_to_io_req_imm(imm);
1245 buf_id = req->permit->mem_id;
1246
1247 req->sg_size = sizeof(*msg);
1248 req->sg_size += le16_to_cpu(msg->sg_cnt) * sizeof(struct rtrs_sg_desc);
1249 req->sg_size += req->usr_len;
1250
1251 /*
1252 * Update stats now, after request is successfully sent it is not
1253 * safe anymore to touch it.
1254 */
1255 rtrs_clt_update_all_stats(req, READ);
1256
1257 ret = rtrs_post_send_rdma(req->con, req, &sess->rbufs[buf_id],
1258 req->data_len, imm, wr);
1259 if (ret) {
1260 rtrs_err_rl(s,
1261 "Read request failed: error=%d path=%s [%s:%u]\n",
1262 ret, kobject_name(&sess->kobj), sess->hca_name,
1263 sess->hca_port);
1264 if (req->mp_policy == MP_POLICY_MIN_INFLIGHT)
1265 atomic_dec(&sess->stats->inflight);
1266 req->need_inv = false;
1267 if (req->sg_cnt)
1268 ib_dma_unmap_sg(dev->ib_dev, req->sglist,
1269 req->sg_cnt, req->dir);
1270 }
1271
1272 return ret;
1273 }
1274
1275 /**
1276 * rtrs_clt_failover_req() - Try to find an active path for a failed request
1277 * @clt: clt context
1278 * @fail_req: a failed io request.
1279 */
rtrs_clt_failover_req(struct rtrs_clt * clt,struct rtrs_clt_io_req * fail_req)1280 static int rtrs_clt_failover_req(struct rtrs_clt *clt,
1281 struct rtrs_clt_io_req *fail_req)
1282 {
1283 struct rtrs_clt_sess *alive_sess;
1284 struct rtrs_clt_io_req *req;
1285 int err = -ECONNABORTED;
1286 struct path_it it;
1287
1288 rcu_read_lock();
1289 for (path_it_init(&it, clt);
1290 (alive_sess = it.next_path(&it)) && it.i < it.clt->paths_num;
1291 it.i++) {
1292 if (READ_ONCE(alive_sess->state) != RTRS_CLT_CONNECTED)
1293 continue;
1294 req = rtrs_clt_get_copy_req(alive_sess, fail_req);
1295 if (req->dir == DMA_TO_DEVICE)
1296 err = rtrs_clt_write_req(req);
1297 else
1298 err = rtrs_clt_read_req(req);
1299 if (err) {
1300 req->in_use = false;
1301 continue;
1302 }
1303 /* Success path */
1304 rtrs_clt_inc_failover_cnt(alive_sess->stats);
1305 break;
1306 }
1307 path_it_deinit(&it);
1308 rcu_read_unlock();
1309
1310 return err;
1311 }
1312
fail_all_outstanding_reqs(struct rtrs_clt_sess * sess)1313 static void fail_all_outstanding_reqs(struct rtrs_clt_sess *sess)
1314 {
1315 struct rtrs_clt *clt = sess->clt;
1316 struct rtrs_clt_io_req *req;
1317 int i, err;
1318
1319 if (!sess->reqs)
1320 return;
1321 for (i = 0; i < sess->queue_depth; ++i) {
1322 req = &sess->reqs[i];
1323 if (!req->in_use)
1324 continue;
1325
1326 /*
1327 * Safely (without notification) complete failed request.
1328 * After completion this request is still useble and can
1329 * be failovered to another path.
1330 */
1331 complete_rdma_req(req, -ECONNABORTED, false, true);
1332
1333 err = rtrs_clt_failover_req(clt, req);
1334 if (err)
1335 /* Failover failed, notify anyway */
1336 req->conf(req->priv, err);
1337 }
1338 }
1339
free_sess_reqs(struct rtrs_clt_sess * sess)1340 static void free_sess_reqs(struct rtrs_clt_sess *sess)
1341 {
1342 struct rtrs_clt_io_req *req;
1343 int i;
1344
1345 if (!sess->reqs)
1346 return;
1347 for (i = 0; i < sess->queue_depth; ++i) {
1348 req = &sess->reqs[i];
1349 if (req->mr)
1350 ib_dereg_mr(req->mr);
1351 kfree(req->sge);
1352 rtrs_iu_free(req->iu, sess->s.dev->ib_dev, 1);
1353 }
1354 kfree(sess->reqs);
1355 sess->reqs = NULL;
1356 }
1357
alloc_sess_reqs(struct rtrs_clt_sess * sess)1358 static int alloc_sess_reqs(struct rtrs_clt_sess *sess)
1359 {
1360 struct rtrs_clt_io_req *req;
1361 int i, err = -ENOMEM;
1362
1363 sess->reqs = kcalloc(sess->queue_depth, sizeof(*sess->reqs),
1364 GFP_KERNEL);
1365 if (!sess->reqs)
1366 return -ENOMEM;
1367
1368 for (i = 0; i < sess->queue_depth; ++i) {
1369 req = &sess->reqs[i];
1370 req->iu = rtrs_iu_alloc(1, sess->max_hdr_size, GFP_KERNEL,
1371 sess->s.dev->ib_dev,
1372 DMA_TO_DEVICE,
1373 rtrs_clt_rdma_done);
1374 if (!req->iu)
1375 goto out;
1376
1377 req->sge = kcalloc(2, sizeof(*req->sge), GFP_KERNEL);
1378 if (!req->sge)
1379 goto out;
1380
1381 req->mr = ib_alloc_mr(sess->s.dev->ib_pd, IB_MR_TYPE_MEM_REG,
1382 sess->max_pages_per_mr);
1383 if (IS_ERR(req->mr)) {
1384 err = PTR_ERR(req->mr);
1385 req->mr = NULL;
1386 pr_err("Failed to alloc sess->max_pages_per_mr %d\n",
1387 sess->max_pages_per_mr);
1388 goto out;
1389 }
1390
1391 init_completion(&req->inv_comp);
1392 }
1393
1394 return 0;
1395
1396 out:
1397 free_sess_reqs(sess);
1398
1399 return err;
1400 }
1401
alloc_permits(struct rtrs_clt * clt)1402 static int alloc_permits(struct rtrs_clt *clt)
1403 {
1404 unsigned int chunk_bits;
1405 int err, i;
1406
1407 clt->permits_map = kcalloc(BITS_TO_LONGS(clt->queue_depth),
1408 sizeof(long), GFP_KERNEL);
1409 if (!clt->permits_map) {
1410 err = -ENOMEM;
1411 goto out_err;
1412 }
1413 clt->permits = kcalloc(clt->queue_depth, permit_size(clt), GFP_KERNEL);
1414 if (!clt->permits) {
1415 err = -ENOMEM;
1416 goto err_map;
1417 }
1418 chunk_bits = ilog2(clt->queue_depth - 1) + 1;
1419 for (i = 0; i < clt->queue_depth; i++) {
1420 struct rtrs_permit *permit;
1421
1422 permit = get_permit(clt, i);
1423 permit->mem_id = i;
1424 permit->mem_off = i << (MAX_IMM_PAYL_BITS - chunk_bits);
1425 }
1426
1427 return 0;
1428
1429 err_map:
1430 kfree(clt->permits_map);
1431 clt->permits_map = NULL;
1432 out_err:
1433 return err;
1434 }
1435
free_permits(struct rtrs_clt * clt)1436 static void free_permits(struct rtrs_clt *clt)
1437 {
1438 if (clt->permits_map) {
1439 size_t sz = clt->queue_depth;
1440
1441 wait_event(clt->permits_wait,
1442 find_first_bit(clt->permits_map, sz) >= sz);
1443 }
1444 kfree(clt->permits_map);
1445 clt->permits_map = NULL;
1446 kfree(clt->permits);
1447 clt->permits = NULL;
1448 }
1449
query_fast_reg_mode(struct rtrs_clt_sess * sess)1450 static void query_fast_reg_mode(struct rtrs_clt_sess *sess)
1451 {
1452 struct ib_device *ib_dev;
1453 u64 max_pages_per_mr;
1454 int mr_page_shift;
1455
1456 ib_dev = sess->s.dev->ib_dev;
1457
1458 /*
1459 * Use the smallest page size supported by the HCA, down to a
1460 * minimum of 4096 bytes. We're unlikely to build large sglists
1461 * out of smaller entries.
1462 */
1463 mr_page_shift = max(12, ffs(ib_dev->attrs.page_size_cap) - 1);
1464 max_pages_per_mr = ib_dev->attrs.max_mr_size;
1465 do_div(max_pages_per_mr, (1ull << mr_page_shift));
1466 sess->max_pages_per_mr =
1467 min3(sess->max_pages_per_mr, (u32)max_pages_per_mr,
1468 ib_dev->attrs.max_fast_reg_page_list_len);
1469 sess->clt->max_segments =
1470 min(sess->max_pages_per_mr, sess->clt->max_segments);
1471 }
1472
rtrs_clt_change_state_get_old(struct rtrs_clt_sess * sess,enum rtrs_clt_state new_state,enum rtrs_clt_state * old_state)1473 static bool rtrs_clt_change_state_get_old(struct rtrs_clt_sess *sess,
1474 enum rtrs_clt_state new_state,
1475 enum rtrs_clt_state *old_state)
1476 {
1477 bool changed;
1478
1479 spin_lock_irq(&sess->state_wq.lock);
1480 if (old_state)
1481 *old_state = sess->state;
1482 changed = rtrs_clt_change_state(sess, new_state);
1483 spin_unlock_irq(&sess->state_wq.lock);
1484
1485 return changed;
1486 }
1487
rtrs_clt_hb_err_handler(struct rtrs_con * c)1488 static void rtrs_clt_hb_err_handler(struct rtrs_con *c)
1489 {
1490 struct rtrs_clt_con *con = container_of(c, typeof(*con), c);
1491
1492 rtrs_rdma_error_recovery(con);
1493 }
1494
rtrs_clt_init_hb(struct rtrs_clt_sess * sess)1495 static void rtrs_clt_init_hb(struct rtrs_clt_sess *sess)
1496 {
1497 rtrs_init_hb(&sess->s, &io_comp_cqe,
1498 RTRS_HB_INTERVAL_MS,
1499 RTRS_HB_MISSED_MAX,
1500 rtrs_clt_hb_err_handler,
1501 rtrs_wq);
1502 }
1503
1504 static void rtrs_clt_reconnect_work(struct work_struct *work);
1505 static void rtrs_clt_close_work(struct work_struct *work);
1506
alloc_sess(struct rtrs_clt * clt,const struct rtrs_addr * path,size_t con_num,u32 nr_poll_queues)1507 static struct rtrs_clt_sess *alloc_sess(struct rtrs_clt *clt,
1508 const struct rtrs_addr *path,
1509 size_t con_num, u32 nr_poll_queues)
1510 {
1511 struct rtrs_clt_sess *sess;
1512 int err = -ENOMEM;
1513 int cpu;
1514 size_t total_con;
1515
1516 sess = kzalloc(sizeof(*sess), GFP_KERNEL);
1517 if (!sess)
1518 goto err;
1519
1520 /*
1521 * irqmode and poll
1522 * +1: Extra connection for user messages
1523 */
1524 total_con = con_num + nr_poll_queues + 1;
1525 sess->s.con = kcalloc(total_con, sizeof(*sess->s.con), GFP_KERNEL);
1526 if (!sess->s.con)
1527 goto err_free_sess;
1528
1529 sess->s.con_num = total_con;
1530 sess->s.irq_con_num = con_num + 1;
1531
1532 sess->stats = kzalloc(sizeof(*sess->stats), GFP_KERNEL);
1533 if (!sess->stats)
1534 goto err_free_con;
1535
1536 mutex_init(&sess->init_mutex);
1537 uuid_gen(&sess->s.uuid);
1538 memcpy(&sess->s.dst_addr, path->dst,
1539 rdma_addr_size((struct sockaddr *)path->dst));
1540
1541 /*
1542 * rdma_resolve_addr() passes src_addr to cma_bind_addr, which
1543 * checks the sa_family to be non-zero. If user passed src_addr=NULL
1544 * the sess->src_addr will contain only zeros, which is then fine.
1545 */
1546 if (path->src)
1547 memcpy(&sess->s.src_addr, path->src,
1548 rdma_addr_size((struct sockaddr *)path->src));
1549 strscpy(sess->s.sessname, clt->sessname, sizeof(sess->s.sessname));
1550 sess->clt = clt;
1551 sess->max_pages_per_mr = RTRS_MAX_SEGMENTS;
1552 init_waitqueue_head(&sess->state_wq);
1553 sess->state = RTRS_CLT_CONNECTING;
1554 atomic_set(&sess->connected_cnt, 0);
1555 INIT_WORK(&sess->close_work, rtrs_clt_close_work);
1556 INIT_DELAYED_WORK(&sess->reconnect_dwork, rtrs_clt_reconnect_work);
1557 rtrs_clt_init_hb(sess);
1558
1559 sess->mp_skip_entry = alloc_percpu(typeof(*sess->mp_skip_entry));
1560 if (!sess->mp_skip_entry)
1561 goto err_free_stats;
1562
1563 for_each_possible_cpu(cpu)
1564 INIT_LIST_HEAD(per_cpu_ptr(sess->mp_skip_entry, cpu));
1565
1566 err = rtrs_clt_init_stats(sess->stats);
1567 if (err)
1568 goto err_free_percpu;
1569
1570 return sess;
1571
1572 err_free_percpu:
1573 free_percpu(sess->mp_skip_entry);
1574 err_free_stats:
1575 kfree(sess->stats);
1576 err_free_con:
1577 kfree(sess->s.con);
1578 err_free_sess:
1579 kfree(sess);
1580 err:
1581 return ERR_PTR(err);
1582 }
1583
free_sess(struct rtrs_clt_sess * sess)1584 void free_sess(struct rtrs_clt_sess *sess)
1585 {
1586 free_percpu(sess->mp_skip_entry);
1587 mutex_destroy(&sess->init_mutex);
1588 kfree(sess->s.con);
1589 kfree(sess->rbufs);
1590 kfree(sess);
1591 }
1592
create_con(struct rtrs_clt_sess * sess,unsigned int cid)1593 static int create_con(struct rtrs_clt_sess *sess, unsigned int cid)
1594 {
1595 struct rtrs_clt_con *con;
1596
1597 con = kzalloc(sizeof(*con), GFP_KERNEL);
1598 if (!con)
1599 return -ENOMEM;
1600
1601 /* Map first two connections to the first CPU */
1602 con->cpu = (cid ? cid - 1 : 0) % nr_cpu_ids;
1603 con->c.cid = cid;
1604 con->c.sess = &sess->s;
1605 /* Align with srv, init as 1 */
1606 atomic_set(&con->c.wr_cnt, 1);
1607 mutex_init(&con->con_mutex);
1608
1609 sess->s.con[cid] = &con->c;
1610
1611 return 0;
1612 }
1613
destroy_con(struct rtrs_clt_con * con)1614 static void destroy_con(struct rtrs_clt_con *con)
1615 {
1616 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
1617
1618 sess->s.con[con->c.cid] = NULL;
1619 mutex_destroy(&con->con_mutex);
1620 kfree(con);
1621 }
1622
create_con_cq_qp(struct rtrs_clt_con * con)1623 static int create_con_cq_qp(struct rtrs_clt_con *con)
1624 {
1625 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
1626 u32 max_send_wr, max_recv_wr, cq_num, max_send_sge, wr_limit;
1627 int err, cq_vector;
1628 struct rtrs_msg_rkey_rsp *rsp;
1629
1630 lockdep_assert_held(&con->con_mutex);
1631 if (con->c.cid == 0) {
1632 max_send_sge = 1;
1633 /* We must be the first here */
1634 if (WARN_ON(sess->s.dev))
1635 return -EINVAL;
1636
1637 /*
1638 * The whole session uses device from user connection.
1639 * Be careful not to close user connection before ib dev
1640 * is gracefully put.
1641 */
1642 sess->s.dev = rtrs_ib_dev_find_or_add(con->c.cm_id->device,
1643 &dev_pd);
1644 if (!sess->s.dev) {
1645 rtrs_wrn(sess->clt,
1646 "rtrs_ib_dev_find_get_or_add(): no memory\n");
1647 return -ENOMEM;
1648 }
1649 sess->s.dev_ref = 1;
1650 query_fast_reg_mode(sess);
1651 wr_limit = sess->s.dev->ib_dev->attrs.max_qp_wr;
1652 /*
1653 * Two (request + registration) completion for send
1654 * Two for recv if always_invalidate is set on server
1655 * or one for recv.
1656 * + 2 for drain and heartbeat
1657 * in case qp gets into error state.
1658 */
1659 max_send_wr =
1660 min_t(int, wr_limit, SERVICE_CON_QUEUE_DEPTH * 2 + 2);
1661 max_recv_wr = max_send_wr;
1662 } else {
1663 /*
1664 * Here we assume that session members are correctly set.
1665 * This is always true if user connection (cid == 0) is
1666 * established first.
1667 */
1668 if (WARN_ON(!sess->s.dev))
1669 return -EINVAL;
1670 if (WARN_ON(!sess->queue_depth))
1671 return -EINVAL;
1672
1673 wr_limit = sess->s.dev->ib_dev->attrs.max_qp_wr;
1674 /* Shared between connections */
1675 sess->s.dev_ref++;
1676 max_send_wr = min_t(int, wr_limit,
1677 /* QD * (REQ + RSP + FR REGS or INVS) + drain */
1678 sess->queue_depth * 3 + 1);
1679 max_recv_wr = min_t(int, wr_limit,
1680 sess->queue_depth * 3 + 1);
1681 max_send_sge = 2;
1682 }
1683 atomic_set(&con->c.sq_wr_avail, max_send_wr);
1684 cq_num = max_send_wr + max_recv_wr;
1685 /* alloc iu to recv new rkey reply when server reports flags set */
1686 if (sess->flags & RTRS_MSG_NEW_RKEY_F || con->c.cid == 0) {
1687 con->rsp_ius = rtrs_iu_alloc(cq_num, sizeof(*rsp),
1688 GFP_KERNEL, sess->s.dev->ib_dev,
1689 DMA_FROM_DEVICE,
1690 rtrs_clt_rdma_done);
1691 if (!con->rsp_ius)
1692 return -ENOMEM;
1693 con->queue_num = cq_num;
1694 }
1695 cq_num = max_send_wr + max_recv_wr;
1696 cq_vector = con->cpu % sess->s.dev->ib_dev->num_comp_vectors;
1697 if (con->c.cid >= sess->s.irq_con_num)
1698 err = rtrs_cq_qp_create(&sess->s, &con->c, max_send_sge,
1699 cq_vector, cq_num, max_send_wr,
1700 max_recv_wr, IB_POLL_DIRECT);
1701 else
1702 err = rtrs_cq_qp_create(&sess->s, &con->c, max_send_sge,
1703 cq_vector, cq_num, max_send_wr,
1704 max_recv_wr, IB_POLL_SOFTIRQ);
1705 /*
1706 * In case of error we do not bother to clean previous allocations,
1707 * since destroy_con_cq_qp() must be called.
1708 */
1709 return err;
1710 }
1711
destroy_con_cq_qp(struct rtrs_clt_con * con)1712 static void destroy_con_cq_qp(struct rtrs_clt_con *con)
1713 {
1714 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
1715
1716 /*
1717 * Be careful here: destroy_con_cq_qp() can be called even
1718 * create_con_cq_qp() failed, see comments there.
1719 */
1720 lockdep_assert_held(&con->con_mutex);
1721 rtrs_cq_qp_destroy(&con->c);
1722 if (con->rsp_ius) {
1723 rtrs_iu_free(con->rsp_ius, sess->s.dev->ib_dev, con->queue_num);
1724 con->rsp_ius = NULL;
1725 con->queue_num = 0;
1726 }
1727 if (sess->s.dev_ref && !--sess->s.dev_ref) {
1728 rtrs_ib_dev_put(sess->s.dev);
1729 sess->s.dev = NULL;
1730 }
1731 }
1732
stop_cm(struct rtrs_clt_con * con)1733 static void stop_cm(struct rtrs_clt_con *con)
1734 {
1735 rdma_disconnect(con->c.cm_id);
1736 if (con->c.qp)
1737 ib_drain_qp(con->c.qp);
1738 }
1739
destroy_cm(struct rtrs_clt_con * con)1740 static void destroy_cm(struct rtrs_clt_con *con)
1741 {
1742 rdma_destroy_id(con->c.cm_id);
1743 con->c.cm_id = NULL;
1744 }
1745
rtrs_rdma_addr_resolved(struct rtrs_clt_con * con)1746 static int rtrs_rdma_addr_resolved(struct rtrs_clt_con *con)
1747 {
1748 struct rtrs_sess *s = con->c.sess;
1749 int err;
1750
1751 mutex_lock(&con->con_mutex);
1752 err = create_con_cq_qp(con);
1753 mutex_unlock(&con->con_mutex);
1754 if (err) {
1755 rtrs_err(s, "create_con_cq_qp(), err: %d\n", err);
1756 return err;
1757 }
1758 err = rdma_resolve_route(con->c.cm_id, RTRS_CONNECT_TIMEOUT_MS);
1759 if (err)
1760 rtrs_err(s, "Resolving route failed, err: %d\n", err);
1761
1762 return err;
1763 }
1764
rtrs_rdma_route_resolved(struct rtrs_clt_con * con)1765 static int rtrs_rdma_route_resolved(struct rtrs_clt_con *con)
1766 {
1767 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
1768 struct rtrs_clt *clt = sess->clt;
1769 struct rtrs_msg_conn_req msg;
1770 struct rdma_conn_param param;
1771
1772 int err;
1773
1774 param = (struct rdma_conn_param) {
1775 .retry_count = 7,
1776 .rnr_retry_count = 7,
1777 .private_data = &msg,
1778 .private_data_len = sizeof(msg),
1779 };
1780
1781 msg = (struct rtrs_msg_conn_req) {
1782 .magic = cpu_to_le16(RTRS_MAGIC),
1783 .version = cpu_to_le16(RTRS_PROTO_VER),
1784 .cid = cpu_to_le16(con->c.cid),
1785 .cid_num = cpu_to_le16(sess->s.con_num),
1786 .recon_cnt = cpu_to_le16(sess->s.recon_cnt),
1787 };
1788 msg.first_conn = sess->for_new_clt ? FIRST_CONN : 0;
1789 uuid_copy(&msg.sess_uuid, &sess->s.uuid);
1790 uuid_copy(&msg.paths_uuid, &clt->paths_uuid);
1791
1792 err = rdma_connect_locked(con->c.cm_id, ¶m);
1793 if (err)
1794 rtrs_err(clt, "rdma_connect_locked(): %d\n", err);
1795
1796 return err;
1797 }
1798
rtrs_rdma_conn_established(struct rtrs_clt_con * con,struct rdma_cm_event * ev)1799 static int rtrs_rdma_conn_established(struct rtrs_clt_con *con,
1800 struct rdma_cm_event *ev)
1801 {
1802 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
1803 struct rtrs_clt *clt = sess->clt;
1804 const struct rtrs_msg_conn_rsp *msg;
1805 u16 version, queue_depth;
1806 int errno;
1807 u8 len;
1808
1809 msg = ev->param.conn.private_data;
1810 len = ev->param.conn.private_data_len;
1811 if (len < sizeof(*msg)) {
1812 rtrs_err(clt, "Invalid RTRS connection response\n");
1813 return -ECONNRESET;
1814 }
1815 if (le16_to_cpu(msg->magic) != RTRS_MAGIC) {
1816 rtrs_err(clt, "Invalid RTRS magic\n");
1817 return -ECONNRESET;
1818 }
1819 version = le16_to_cpu(msg->version);
1820 if (version >> 8 != RTRS_PROTO_VER_MAJOR) {
1821 rtrs_err(clt, "Unsupported major RTRS version: %d, expected %d\n",
1822 version >> 8, RTRS_PROTO_VER_MAJOR);
1823 return -ECONNRESET;
1824 }
1825 errno = le16_to_cpu(msg->errno);
1826 if (errno) {
1827 rtrs_err(clt, "Invalid RTRS message: errno %d\n",
1828 errno);
1829 return -ECONNRESET;
1830 }
1831 if (con->c.cid == 0) {
1832 queue_depth = le16_to_cpu(msg->queue_depth);
1833
1834 if (sess->queue_depth > 0 && queue_depth != sess->queue_depth) {
1835 rtrs_err(clt, "Error: queue depth changed\n");
1836
1837 /*
1838 * Stop any more reconnection attempts
1839 */
1840 sess->reconnect_attempts = -1;
1841 rtrs_err(clt,
1842 "Disabling auto-reconnect. Trigger a manual reconnect after issue is resolved\n");
1843 return -ECONNRESET;
1844 }
1845
1846 if (!sess->rbufs) {
1847 sess->rbufs = kcalloc(queue_depth, sizeof(*sess->rbufs),
1848 GFP_KERNEL);
1849 if (!sess->rbufs)
1850 return -ENOMEM;
1851 }
1852 sess->queue_depth = queue_depth;
1853 sess->s.signal_interval = min_not_zero(queue_depth,
1854 (unsigned short) SERVICE_CON_QUEUE_DEPTH);
1855 sess->max_hdr_size = le32_to_cpu(msg->max_hdr_size);
1856 sess->max_io_size = le32_to_cpu(msg->max_io_size);
1857 sess->flags = le32_to_cpu(msg->flags);
1858 sess->chunk_size = sess->max_io_size + sess->max_hdr_size;
1859
1860 /*
1861 * Global IO size is always a minimum.
1862 * If while a reconnection server sends us a value a bit
1863 * higher - client does not care and uses cached minimum.
1864 *
1865 * Since we can have several sessions (paths) restablishing
1866 * connections in parallel, use lock.
1867 */
1868 mutex_lock(&clt->paths_mutex);
1869 clt->queue_depth = sess->queue_depth;
1870 clt->max_io_size = min_not_zero(sess->max_io_size,
1871 clt->max_io_size);
1872 mutex_unlock(&clt->paths_mutex);
1873
1874 /*
1875 * Cache the hca_port and hca_name for sysfs
1876 */
1877 sess->hca_port = con->c.cm_id->port_num;
1878 scnprintf(sess->hca_name, sizeof(sess->hca_name),
1879 sess->s.dev->ib_dev->name);
1880 sess->s.src_addr = con->c.cm_id->route.addr.src_addr;
1881 /* set for_new_clt, to allow future reconnect on any path */
1882 sess->for_new_clt = 1;
1883 }
1884
1885 return 0;
1886 }
1887
flag_success_on_conn(struct rtrs_clt_con * con)1888 static inline void flag_success_on_conn(struct rtrs_clt_con *con)
1889 {
1890 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
1891
1892 atomic_inc(&sess->connected_cnt);
1893 con->cm_err = 1;
1894 }
1895
rtrs_rdma_conn_rejected(struct rtrs_clt_con * con,struct rdma_cm_event * ev)1896 static int rtrs_rdma_conn_rejected(struct rtrs_clt_con *con,
1897 struct rdma_cm_event *ev)
1898 {
1899 struct rtrs_sess *s = con->c.sess;
1900 const struct rtrs_msg_conn_rsp *msg;
1901 const char *rej_msg;
1902 int status, errno;
1903 u8 data_len;
1904
1905 status = ev->status;
1906 rej_msg = rdma_reject_msg(con->c.cm_id, status);
1907 msg = rdma_consumer_reject_data(con->c.cm_id, ev, &data_len);
1908
1909 if (msg && data_len >= sizeof(*msg)) {
1910 errno = (int16_t)le16_to_cpu(msg->errno);
1911 if (errno == -EBUSY)
1912 rtrs_err(s,
1913 "Previous session is still exists on the server, please reconnect later\n");
1914 else
1915 rtrs_err(s,
1916 "Connect rejected: status %d (%s), rtrs errno %d\n",
1917 status, rej_msg, errno);
1918 } else {
1919 rtrs_err(s,
1920 "Connect rejected but with malformed message: status %d (%s)\n",
1921 status, rej_msg);
1922 }
1923
1924 return -ECONNRESET;
1925 }
1926
rtrs_clt_close_conns(struct rtrs_clt_sess * sess,bool wait)1927 void rtrs_clt_close_conns(struct rtrs_clt_sess *sess, bool wait)
1928 {
1929 if (rtrs_clt_change_state_get_old(sess, RTRS_CLT_CLOSING, NULL))
1930 queue_work(rtrs_wq, &sess->close_work);
1931 if (wait)
1932 flush_work(&sess->close_work);
1933 }
1934
flag_error_on_conn(struct rtrs_clt_con * con,int cm_err)1935 static inline void flag_error_on_conn(struct rtrs_clt_con *con, int cm_err)
1936 {
1937 if (con->cm_err == 1) {
1938 struct rtrs_clt_sess *sess;
1939
1940 sess = to_clt_sess(con->c.sess);
1941 if (atomic_dec_and_test(&sess->connected_cnt))
1942
1943 wake_up(&sess->state_wq);
1944 }
1945 con->cm_err = cm_err;
1946 }
1947
rtrs_clt_rdma_cm_handler(struct rdma_cm_id * cm_id,struct rdma_cm_event * ev)1948 static int rtrs_clt_rdma_cm_handler(struct rdma_cm_id *cm_id,
1949 struct rdma_cm_event *ev)
1950 {
1951 struct rtrs_clt_con *con = cm_id->context;
1952 struct rtrs_sess *s = con->c.sess;
1953 struct rtrs_clt_sess *sess = to_clt_sess(s);
1954 int cm_err = 0;
1955
1956 switch (ev->event) {
1957 case RDMA_CM_EVENT_ADDR_RESOLVED:
1958 cm_err = rtrs_rdma_addr_resolved(con);
1959 break;
1960 case RDMA_CM_EVENT_ROUTE_RESOLVED:
1961 cm_err = rtrs_rdma_route_resolved(con);
1962 break;
1963 case RDMA_CM_EVENT_ESTABLISHED:
1964 cm_err = rtrs_rdma_conn_established(con, ev);
1965 if (!cm_err) {
1966 /*
1967 * Report success and wake up. Here we abuse state_wq,
1968 * i.e. wake up without state change, but we set cm_err.
1969 */
1970 flag_success_on_conn(con);
1971 wake_up(&sess->state_wq);
1972 return 0;
1973 }
1974 break;
1975 case RDMA_CM_EVENT_REJECTED:
1976 cm_err = rtrs_rdma_conn_rejected(con, ev);
1977 break;
1978 case RDMA_CM_EVENT_DISCONNECTED:
1979 /* No message for disconnecting */
1980 cm_err = -ECONNRESET;
1981 break;
1982 case RDMA_CM_EVENT_CONNECT_ERROR:
1983 case RDMA_CM_EVENT_UNREACHABLE:
1984 case RDMA_CM_EVENT_ADDR_CHANGE:
1985 case RDMA_CM_EVENT_TIMEWAIT_EXIT:
1986 rtrs_wrn(s, "CM error (CM event: %s, err: %d)\n",
1987 rdma_event_msg(ev->event), ev->status);
1988 cm_err = -ECONNRESET;
1989 break;
1990 case RDMA_CM_EVENT_ADDR_ERROR:
1991 case RDMA_CM_EVENT_ROUTE_ERROR:
1992 rtrs_wrn(s, "CM error (CM event: %s, err: %d)\n",
1993 rdma_event_msg(ev->event), ev->status);
1994 cm_err = -EHOSTUNREACH;
1995 break;
1996 case RDMA_CM_EVENT_DEVICE_REMOVAL:
1997 /*
1998 * Device removal is a special case. Queue close and return 0.
1999 */
2000 rtrs_clt_close_conns(sess, false);
2001 return 0;
2002 default:
2003 rtrs_err(s, "Unexpected RDMA CM error (CM event: %s, err: %d)\n",
2004 rdma_event_msg(ev->event), ev->status);
2005 cm_err = -ECONNRESET;
2006 break;
2007 }
2008
2009 if (cm_err) {
2010 /*
2011 * cm error makes sense only on connection establishing,
2012 * in other cases we rely on normal procedure of reconnecting.
2013 */
2014 flag_error_on_conn(con, cm_err);
2015 rtrs_rdma_error_recovery(con);
2016 }
2017
2018 return 0;
2019 }
2020
create_cm(struct rtrs_clt_con * con)2021 static int create_cm(struct rtrs_clt_con *con)
2022 {
2023 struct rtrs_sess *s = con->c.sess;
2024 struct rtrs_clt_sess *sess = to_clt_sess(s);
2025 struct rdma_cm_id *cm_id;
2026 int err;
2027
2028 cm_id = rdma_create_id(&init_net, rtrs_clt_rdma_cm_handler, con,
2029 sess->s.dst_addr.ss_family == AF_IB ?
2030 RDMA_PS_IB : RDMA_PS_TCP, IB_QPT_RC);
2031 if (IS_ERR(cm_id)) {
2032 err = PTR_ERR(cm_id);
2033 rtrs_err(s, "Failed to create CM ID, err: %d\n", err);
2034
2035 return err;
2036 }
2037 con->c.cm_id = cm_id;
2038 con->cm_err = 0;
2039 /* allow the port to be reused */
2040 err = rdma_set_reuseaddr(cm_id, 1);
2041 if (err != 0) {
2042 rtrs_err(s, "Set address reuse failed, err: %d\n", err);
2043 goto destroy_cm;
2044 }
2045 err = rdma_resolve_addr(cm_id, (struct sockaddr *)&sess->s.src_addr,
2046 (struct sockaddr *)&sess->s.dst_addr,
2047 RTRS_CONNECT_TIMEOUT_MS);
2048 if (err) {
2049 rtrs_err(s, "Failed to resolve address, err: %d\n", err);
2050 goto destroy_cm;
2051 }
2052 /*
2053 * Combine connection status and session events. This is needed
2054 * for waiting two possible cases: cm_err has something meaningful
2055 * or session state was really changed to error by device removal.
2056 */
2057 err = wait_event_interruptible_timeout(
2058 sess->state_wq,
2059 con->cm_err || sess->state != RTRS_CLT_CONNECTING,
2060 msecs_to_jiffies(RTRS_CONNECT_TIMEOUT_MS));
2061 if (err == 0 || err == -ERESTARTSYS) {
2062 if (err == 0)
2063 err = -ETIMEDOUT;
2064 /* Timedout or interrupted */
2065 goto errr;
2066 }
2067 if (con->cm_err < 0) {
2068 err = con->cm_err;
2069 goto errr;
2070 }
2071 if (READ_ONCE(sess->state) != RTRS_CLT_CONNECTING) {
2072 /* Device removal */
2073 err = -ECONNABORTED;
2074 goto errr;
2075 }
2076
2077 return 0;
2078
2079 errr:
2080 stop_cm(con);
2081 mutex_lock(&con->con_mutex);
2082 destroy_con_cq_qp(con);
2083 mutex_unlock(&con->con_mutex);
2084 destroy_cm:
2085 destroy_cm(con);
2086
2087 return err;
2088 }
2089
rtrs_clt_sess_up(struct rtrs_clt_sess * sess)2090 static void rtrs_clt_sess_up(struct rtrs_clt_sess *sess)
2091 {
2092 struct rtrs_clt *clt = sess->clt;
2093 int up;
2094
2095 /*
2096 * We can fire RECONNECTED event only when all paths were
2097 * connected on rtrs_clt_open(), then each was disconnected
2098 * and the first one connected again. That's why this nasty
2099 * game with counter value.
2100 */
2101
2102 mutex_lock(&clt->paths_ev_mutex);
2103 up = ++clt->paths_up;
2104 /*
2105 * Here it is safe to access paths num directly since up counter
2106 * is greater than MAX_PATHS_NUM only while rtrs_clt_open() is
2107 * in progress, thus paths removals are impossible.
2108 */
2109 if (up > MAX_PATHS_NUM && up == MAX_PATHS_NUM + clt->paths_num)
2110 clt->paths_up = clt->paths_num;
2111 else if (up == 1)
2112 clt->link_ev(clt->priv, RTRS_CLT_LINK_EV_RECONNECTED);
2113 mutex_unlock(&clt->paths_ev_mutex);
2114
2115 /* Mark session as established */
2116 sess->established = true;
2117 sess->reconnect_attempts = 0;
2118 sess->stats->reconnects.successful_cnt++;
2119 }
2120
rtrs_clt_sess_down(struct rtrs_clt_sess * sess)2121 static void rtrs_clt_sess_down(struct rtrs_clt_sess *sess)
2122 {
2123 struct rtrs_clt *clt = sess->clt;
2124
2125 if (!sess->established)
2126 return;
2127
2128 sess->established = false;
2129 mutex_lock(&clt->paths_ev_mutex);
2130 WARN_ON(!clt->paths_up);
2131 if (--clt->paths_up == 0)
2132 clt->link_ev(clt->priv, RTRS_CLT_LINK_EV_DISCONNECTED);
2133 mutex_unlock(&clt->paths_ev_mutex);
2134 }
2135
rtrs_clt_stop_and_destroy_conns(struct rtrs_clt_sess * sess)2136 static void rtrs_clt_stop_and_destroy_conns(struct rtrs_clt_sess *sess)
2137 {
2138 struct rtrs_clt_con *con;
2139 unsigned int cid;
2140
2141 WARN_ON(READ_ONCE(sess->state) == RTRS_CLT_CONNECTED);
2142
2143 /*
2144 * Possible race with rtrs_clt_open(), when DEVICE_REMOVAL comes
2145 * exactly in between. Start destroying after it finishes.
2146 */
2147 mutex_lock(&sess->init_mutex);
2148 mutex_unlock(&sess->init_mutex);
2149
2150 /*
2151 * All IO paths must observe !CONNECTED state before we
2152 * free everything.
2153 */
2154 synchronize_rcu();
2155
2156 rtrs_stop_hb(&sess->s);
2157
2158 /*
2159 * The order it utterly crucial: firstly disconnect and complete all
2160 * rdma requests with error (thus set in_use=false for requests),
2161 * then fail outstanding requests checking in_use for each, and
2162 * eventually notify upper layer about session disconnection.
2163 */
2164
2165 for (cid = 0; cid < sess->s.con_num; cid++) {
2166 if (!sess->s.con[cid])
2167 break;
2168 con = to_clt_con(sess->s.con[cid]);
2169 stop_cm(con);
2170 }
2171 fail_all_outstanding_reqs(sess);
2172 free_sess_reqs(sess);
2173 rtrs_clt_sess_down(sess);
2174
2175 /*
2176 * Wait for graceful shutdown, namely when peer side invokes
2177 * rdma_disconnect(). 'connected_cnt' is decremented only on
2178 * CM events, thus if other side had crashed and hb has detected
2179 * something is wrong, here we will stuck for exactly timeout ms,
2180 * since CM does not fire anything. That is fine, we are not in
2181 * hurry.
2182 */
2183 wait_event_timeout(sess->state_wq, !atomic_read(&sess->connected_cnt),
2184 msecs_to_jiffies(RTRS_CONNECT_TIMEOUT_MS));
2185
2186 for (cid = 0; cid < sess->s.con_num; cid++) {
2187 if (!sess->s.con[cid])
2188 break;
2189 con = to_clt_con(sess->s.con[cid]);
2190 mutex_lock(&con->con_mutex);
2191 destroy_con_cq_qp(con);
2192 mutex_unlock(&con->con_mutex);
2193 destroy_cm(con);
2194 destroy_con(con);
2195 }
2196 }
2197
xchg_sessions(struct rtrs_clt_sess __rcu ** rcu_ppcpu_path,struct rtrs_clt_sess * sess,struct rtrs_clt_sess * next)2198 static inline bool xchg_sessions(struct rtrs_clt_sess __rcu **rcu_ppcpu_path,
2199 struct rtrs_clt_sess *sess,
2200 struct rtrs_clt_sess *next)
2201 {
2202 struct rtrs_clt_sess **ppcpu_path;
2203
2204 /* Call cmpxchg() without sparse warnings */
2205 ppcpu_path = (typeof(ppcpu_path))rcu_ppcpu_path;
2206 return sess == cmpxchg(ppcpu_path, sess, next);
2207 }
2208
rtrs_clt_remove_path_from_arr(struct rtrs_clt_sess * sess)2209 static void rtrs_clt_remove_path_from_arr(struct rtrs_clt_sess *sess)
2210 {
2211 struct rtrs_clt *clt = sess->clt;
2212 struct rtrs_clt_sess *next;
2213 bool wait_for_grace = false;
2214 int cpu;
2215
2216 mutex_lock(&clt->paths_mutex);
2217 list_del_rcu(&sess->s.entry);
2218
2219 /* Make sure everybody observes path removal. */
2220 synchronize_rcu();
2221
2222 /*
2223 * At this point nobody sees @sess in the list, but still we have
2224 * dangling pointer @pcpu_path which _can_ point to @sess. Since
2225 * nobody can observe @sess in the list, we guarantee that IO path
2226 * will not assign @sess to @pcpu_path, i.e. @pcpu_path can be equal
2227 * to @sess, but can never again become @sess.
2228 */
2229
2230 /*
2231 * Decrement paths number only after grace period, because
2232 * caller of do_each_path() must firstly observe list without
2233 * path and only then decremented paths number.
2234 *
2235 * Otherwise there can be the following situation:
2236 * o Two paths exist and IO is coming.
2237 * o One path is removed:
2238 * CPU#0 CPU#1
2239 * do_each_path(): rtrs_clt_remove_path_from_arr():
2240 * path = get_next_path()
2241 * ^^^ list_del_rcu(path)
2242 * [!CONNECTED path] clt->paths_num--
2243 * ^^^^^^^^^
2244 * load clt->paths_num from 2 to 1
2245 * ^^^^^^^^^
2246 * sees 1
2247 *
2248 * path is observed as !CONNECTED, but do_each_path() loop
2249 * ends, because expression i < clt->paths_num is false.
2250 */
2251 clt->paths_num--;
2252
2253 /*
2254 * Get @next connection from current @sess which is going to be
2255 * removed. If @sess is the last element, then @next is NULL.
2256 */
2257 rcu_read_lock();
2258 next = list_next_or_null_rr_rcu(&clt->paths_list, &sess->s.entry,
2259 typeof(*next), s.entry);
2260 rcu_read_unlock();
2261
2262 /*
2263 * @pcpu paths can still point to the path which is going to be
2264 * removed, so change the pointer manually.
2265 */
2266 for_each_possible_cpu(cpu) {
2267 struct rtrs_clt_sess __rcu **ppcpu_path;
2268
2269 ppcpu_path = per_cpu_ptr(clt->pcpu_path, cpu);
2270 if (rcu_dereference_protected(*ppcpu_path,
2271 lockdep_is_held(&clt->paths_mutex)) != sess)
2272 /*
2273 * synchronize_rcu() was called just after deleting
2274 * entry from the list, thus IO code path cannot
2275 * change pointer back to the pointer which is going
2276 * to be removed, we are safe here.
2277 */
2278 continue;
2279
2280 /*
2281 * We race with IO code path, which also changes pointer,
2282 * thus we have to be careful not to overwrite it.
2283 */
2284 if (xchg_sessions(ppcpu_path, sess, next))
2285 /*
2286 * @ppcpu_path was successfully replaced with @next,
2287 * that means that someone could also pick up the
2288 * @sess and dereferencing it right now, so wait for
2289 * a grace period is required.
2290 */
2291 wait_for_grace = true;
2292 }
2293 if (wait_for_grace)
2294 synchronize_rcu();
2295
2296 mutex_unlock(&clt->paths_mutex);
2297 }
2298
rtrs_clt_add_path_to_arr(struct rtrs_clt_sess * sess)2299 static void rtrs_clt_add_path_to_arr(struct rtrs_clt_sess *sess)
2300 {
2301 struct rtrs_clt *clt = sess->clt;
2302
2303 mutex_lock(&clt->paths_mutex);
2304 clt->paths_num++;
2305
2306 list_add_tail_rcu(&sess->s.entry, &clt->paths_list);
2307 mutex_unlock(&clt->paths_mutex);
2308 }
2309
rtrs_clt_close_work(struct work_struct * work)2310 static void rtrs_clt_close_work(struct work_struct *work)
2311 {
2312 struct rtrs_clt_sess *sess;
2313
2314 sess = container_of(work, struct rtrs_clt_sess, close_work);
2315
2316 cancel_delayed_work_sync(&sess->reconnect_dwork);
2317 rtrs_clt_stop_and_destroy_conns(sess);
2318 rtrs_clt_change_state_get_old(sess, RTRS_CLT_CLOSED, NULL);
2319 }
2320
init_conns(struct rtrs_clt_sess * sess)2321 static int init_conns(struct rtrs_clt_sess *sess)
2322 {
2323 unsigned int cid;
2324 int err;
2325
2326 /*
2327 * On every new session connections increase reconnect counter
2328 * to avoid clashes with previous sessions not yet closed
2329 * sessions on a server side.
2330 */
2331 sess->s.recon_cnt++;
2332
2333 /* Establish all RDMA connections */
2334 for (cid = 0; cid < sess->s.con_num; cid++) {
2335 err = create_con(sess, cid);
2336 if (err)
2337 goto destroy;
2338
2339 err = create_cm(to_clt_con(sess->s.con[cid]));
2340 if (err) {
2341 destroy_con(to_clt_con(sess->s.con[cid]));
2342 goto destroy;
2343 }
2344 }
2345 err = alloc_sess_reqs(sess);
2346 if (err)
2347 goto destroy;
2348
2349 rtrs_start_hb(&sess->s);
2350
2351 return 0;
2352
2353 destroy:
2354 while (cid--) {
2355 struct rtrs_clt_con *con = to_clt_con(sess->s.con[cid]);
2356
2357 stop_cm(con);
2358
2359 mutex_lock(&con->con_mutex);
2360 destroy_con_cq_qp(con);
2361 mutex_unlock(&con->con_mutex);
2362 destroy_cm(con);
2363 destroy_con(con);
2364 }
2365 /*
2366 * If we've never taken async path and got an error, say,
2367 * doing rdma_resolve_addr(), switch to CONNECTION_ERR state
2368 * manually to keep reconnecting.
2369 */
2370 rtrs_clt_change_state_get_old(sess, RTRS_CLT_CONNECTING_ERR, NULL);
2371
2372 return err;
2373 }
2374
rtrs_clt_info_req_done(struct ib_cq * cq,struct ib_wc * wc)2375 static void rtrs_clt_info_req_done(struct ib_cq *cq, struct ib_wc *wc)
2376 {
2377 struct rtrs_clt_con *con = to_clt_con(wc->qp->qp_context);
2378 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
2379 struct rtrs_iu *iu;
2380
2381 iu = container_of(wc->wr_cqe, struct rtrs_iu, cqe);
2382 rtrs_iu_free(iu, sess->s.dev->ib_dev, 1);
2383
2384 if (wc->status != IB_WC_SUCCESS) {
2385 rtrs_err(sess->clt, "Sess info request send failed: %s\n",
2386 ib_wc_status_msg(wc->status));
2387 rtrs_clt_change_state_get_old(sess, RTRS_CLT_CONNECTING_ERR, NULL);
2388 return;
2389 }
2390
2391 rtrs_clt_update_wc_stats(con);
2392 }
2393
process_info_rsp(struct rtrs_clt_sess * sess,const struct rtrs_msg_info_rsp * msg)2394 static int process_info_rsp(struct rtrs_clt_sess *sess,
2395 const struct rtrs_msg_info_rsp *msg)
2396 {
2397 unsigned int sg_cnt, total_len;
2398 int i, sgi;
2399
2400 sg_cnt = le16_to_cpu(msg->sg_cnt);
2401 if (!sg_cnt || (sess->queue_depth % sg_cnt)) {
2402 rtrs_err(sess->clt, "Incorrect sg_cnt %d, is not multiple\n",
2403 sg_cnt);
2404 return -EINVAL;
2405 }
2406
2407 /*
2408 * Check if IB immediate data size is enough to hold the mem_id and
2409 * the offset inside the memory chunk.
2410 */
2411 if ((ilog2(sg_cnt - 1) + 1) + (ilog2(sess->chunk_size - 1) + 1) >
2412 MAX_IMM_PAYL_BITS) {
2413 rtrs_err(sess->clt,
2414 "RDMA immediate size (%db) not enough to encode %d buffers of size %dB\n",
2415 MAX_IMM_PAYL_BITS, sg_cnt, sess->chunk_size);
2416 return -EINVAL;
2417 }
2418 total_len = 0;
2419 for (sgi = 0, i = 0; sgi < sg_cnt && i < sess->queue_depth; sgi++) {
2420 const struct rtrs_sg_desc *desc = &msg->desc[sgi];
2421 u32 len, rkey;
2422 u64 addr;
2423
2424 addr = le64_to_cpu(desc->addr);
2425 rkey = le32_to_cpu(desc->key);
2426 len = le32_to_cpu(desc->len);
2427
2428 total_len += len;
2429
2430 if (!len || (len % sess->chunk_size)) {
2431 rtrs_err(sess->clt, "Incorrect [%d].len %d\n", sgi,
2432 len);
2433 return -EINVAL;
2434 }
2435 for ( ; len && i < sess->queue_depth; i++) {
2436 sess->rbufs[i].addr = addr;
2437 sess->rbufs[i].rkey = rkey;
2438
2439 len -= sess->chunk_size;
2440 addr += sess->chunk_size;
2441 }
2442 }
2443 /* Sanity check */
2444 if (sgi != sg_cnt || i != sess->queue_depth) {
2445 rtrs_err(sess->clt, "Incorrect sg vector, not fully mapped\n");
2446 return -EINVAL;
2447 }
2448 if (total_len != sess->chunk_size * sess->queue_depth) {
2449 rtrs_err(sess->clt, "Incorrect total_len %d\n", total_len);
2450 return -EINVAL;
2451 }
2452
2453 return 0;
2454 }
2455
rtrs_clt_info_rsp_done(struct ib_cq * cq,struct ib_wc * wc)2456 static void rtrs_clt_info_rsp_done(struct ib_cq *cq, struct ib_wc *wc)
2457 {
2458 struct rtrs_clt_con *con = to_clt_con(wc->qp->qp_context);
2459 struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess);
2460 struct rtrs_msg_info_rsp *msg;
2461 enum rtrs_clt_state state;
2462 struct rtrs_iu *iu;
2463 size_t rx_sz;
2464 int err;
2465
2466 state = RTRS_CLT_CONNECTING_ERR;
2467
2468 WARN_ON(con->c.cid);
2469 iu = container_of(wc->wr_cqe, struct rtrs_iu, cqe);
2470 if (wc->status != IB_WC_SUCCESS) {
2471 rtrs_err(sess->clt, "Sess info response recv failed: %s\n",
2472 ib_wc_status_msg(wc->status));
2473 goto out;
2474 }
2475 WARN_ON(wc->opcode != IB_WC_RECV);
2476
2477 if (wc->byte_len < sizeof(*msg)) {
2478 rtrs_err(sess->clt, "Sess info response is malformed: size %d\n",
2479 wc->byte_len);
2480 goto out;
2481 }
2482 ib_dma_sync_single_for_cpu(sess->s.dev->ib_dev, iu->dma_addr,
2483 iu->size, DMA_FROM_DEVICE);
2484 msg = iu->buf;
2485 if (le16_to_cpu(msg->type) != RTRS_MSG_INFO_RSP) {
2486 rtrs_err(sess->clt, "Sess info response is malformed: type %d\n",
2487 le16_to_cpu(msg->type));
2488 goto out;
2489 }
2490 rx_sz = sizeof(*msg);
2491 rx_sz += sizeof(msg->desc[0]) * le16_to_cpu(msg->sg_cnt);
2492 if (wc->byte_len < rx_sz) {
2493 rtrs_err(sess->clt, "Sess info response is malformed: size %d\n",
2494 wc->byte_len);
2495 goto out;
2496 }
2497 err = process_info_rsp(sess, msg);
2498 if (err)
2499 goto out;
2500
2501 err = post_recv_sess(sess);
2502 if (err)
2503 goto out;
2504
2505 state = RTRS_CLT_CONNECTED;
2506
2507 out:
2508 rtrs_clt_update_wc_stats(con);
2509 rtrs_iu_free(iu, sess->s.dev->ib_dev, 1);
2510 rtrs_clt_change_state_get_old(sess, state, NULL);
2511 }
2512
rtrs_send_sess_info(struct rtrs_clt_sess * sess)2513 static int rtrs_send_sess_info(struct rtrs_clt_sess *sess)
2514 {
2515 struct rtrs_clt_con *usr_con = to_clt_con(sess->s.con[0]);
2516 struct rtrs_msg_info_req *msg;
2517 struct rtrs_iu *tx_iu, *rx_iu;
2518 size_t rx_sz;
2519 int err;
2520
2521 rx_sz = sizeof(struct rtrs_msg_info_rsp);
2522 rx_sz += sizeof(struct rtrs_sg_desc) * sess->queue_depth;
2523
2524 tx_iu = rtrs_iu_alloc(1, sizeof(struct rtrs_msg_info_req), GFP_KERNEL,
2525 sess->s.dev->ib_dev, DMA_TO_DEVICE,
2526 rtrs_clt_info_req_done);
2527 rx_iu = rtrs_iu_alloc(1, rx_sz, GFP_KERNEL, sess->s.dev->ib_dev,
2528 DMA_FROM_DEVICE, rtrs_clt_info_rsp_done);
2529 if (!tx_iu || !rx_iu) {
2530 err = -ENOMEM;
2531 goto out;
2532 }
2533 /* Prepare for getting info response */
2534 err = rtrs_iu_post_recv(&usr_con->c, rx_iu);
2535 if (err) {
2536 rtrs_err(sess->clt, "rtrs_iu_post_recv(), err: %d\n", err);
2537 goto out;
2538 }
2539 rx_iu = NULL;
2540
2541 msg = tx_iu->buf;
2542 msg->type = cpu_to_le16(RTRS_MSG_INFO_REQ);
2543 memcpy(msg->sessname, sess->s.sessname, sizeof(msg->sessname));
2544
2545 ib_dma_sync_single_for_device(sess->s.dev->ib_dev, tx_iu->dma_addr,
2546 tx_iu->size, DMA_TO_DEVICE);
2547
2548 /* Send info request */
2549 err = rtrs_iu_post_send(&usr_con->c, tx_iu, sizeof(*msg), NULL);
2550 if (err) {
2551 rtrs_err(sess->clt, "rtrs_iu_post_send(), err: %d\n", err);
2552 goto out;
2553 }
2554 tx_iu = NULL;
2555
2556 /* Wait for state change */
2557 wait_event_interruptible_timeout(sess->state_wq,
2558 sess->state != RTRS_CLT_CONNECTING,
2559 msecs_to_jiffies(
2560 RTRS_CONNECT_TIMEOUT_MS));
2561 if (READ_ONCE(sess->state) != RTRS_CLT_CONNECTED) {
2562 if (READ_ONCE(sess->state) == RTRS_CLT_CONNECTING_ERR)
2563 err = -ECONNRESET;
2564 else
2565 err = -ETIMEDOUT;
2566 }
2567
2568 out:
2569 if (tx_iu)
2570 rtrs_iu_free(tx_iu, sess->s.dev->ib_dev, 1);
2571 if (rx_iu)
2572 rtrs_iu_free(rx_iu, sess->s.dev->ib_dev, 1);
2573 if (err)
2574 /* If we've never taken async path because of malloc problems */
2575 rtrs_clt_change_state_get_old(sess, RTRS_CLT_CONNECTING_ERR, NULL);
2576
2577 return err;
2578 }
2579
2580 /**
2581 * init_sess() - establishes all session connections and does handshake
2582 * @sess: client session.
2583 * In case of error full close or reconnect procedure should be taken,
2584 * because reconnect or close async works can be started.
2585 */
init_sess(struct rtrs_clt_sess * sess)2586 static int init_sess(struct rtrs_clt_sess *sess)
2587 {
2588 int err;
2589 char str[NAME_MAX];
2590 struct rtrs_addr path = {
2591 .src = &sess->s.src_addr,
2592 .dst = &sess->s.dst_addr,
2593 };
2594
2595 rtrs_addr_to_str(&path, str, sizeof(str));
2596
2597 mutex_lock(&sess->init_mutex);
2598 err = init_conns(sess);
2599 if (err) {
2600 rtrs_err(sess->clt,
2601 "init_conns() failed: err=%d path=%s [%s:%u]\n", err,
2602 str, sess->hca_name, sess->hca_port);
2603 goto out;
2604 }
2605 err = rtrs_send_sess_info(sess);
2606 if (err) {
2607 rtrs_err(
2608 sess->clt,
2609 "rtrs_send_sess_info() failed: err=%d path=%s [%s:%u]\n",
2610 err, str, sess->hca_name, sess->hca_port);
2611 goto out;
2612 }
2613 rtrs_clt_sess_up(sess);
2614 out:
2615 mutex_unlock(&sess->init_mutex);
2616
2617 return err;
2618 }
2619
rtrs_clt_reconnect_work(struct work_struct * work)2620 static void rtrs_clt_reconnect_work(struct work_struct *work)
2621 {
2622 struct rtrs_clt_sess *sess;
2623 struct rtrs_clt *clt;
2624 unsigned int delay_ms;
2625 int err;
2626
2627 sess = container_of(to_delayed_work(work), struct rtrs_clt_sess,
2628 reconnect_dwork);
2629 clt = sess->clt;
2630
2631 if (READ_ONCE(sess->state) != RTRS_CLT_RECONNECTING)
2632 return;
2633
2634 if (sess->reconnect_attempts >= clt->max_reconnect_attempts) {
2635 /* Close a session completely if max attempts is reached */
2636 rtrs_clt_close_conns(sess, false);
2637 return;
2638 }
2639 sess->reconnect_attempts++;
2640
2641 /* Stop everything */
2642 rtrs_clt_stop_and_destroy_conns(sess);
2643 msleep(RTRS_RECONNECT_BACKOFF);
2644 if (rtrs_clt_change_state_get_old(sess, RTRS_CLT_CONNECTING, NULL)) {
2645 err = init_sess(sess);
2646 if (err)
2647 goto reconnect_again;
2648 }
2649
2650 return;
2651
2652 reconnect_again:
2653 if (rtrs_clt_change_state_get_old(sess, RTRS_CLT_RECONNECTING, NULL)) {
2654 sess->stats->reconnects.fail_cnt++;
2655 delay_ms = clt->reconnect_delay_sec * 1000;
2656 queue_delayed_work(rtrs_wq, &sess->reconnect_dwork,
2657 msecs_to_jiffies(delay_ms +
2658 prandom_u32() %
2659 RTRS_RECONNECT_SEED));
2660 }
2661 }
2662
rtrs_clt_dev_release(struct device * dev)2663 static void rtrs_clt_dev_release(struct device *dev)
2664 {
2665 struct rtrs_clt *clt = container_of(dev, struct rtrs_clt, dev);
2666
2667 kfree(clt);
2668 }
2669
alloc_clt(const char * sessname,size_t paths_num,u16 port,size_t pdu_sz,void * priv,void (* link_ev)(void * priv,enum rtrs_clt_link_ev ev),unsigned int reconnect_delay_sec,unsigned int max_reconnect_attempts)2670 static struct rtrs_clt *alloc_clt(const char *sessname, size_t paths_num,
2671 u16 port, size_t pdu_sz, void *priv,
2672 void (*link_ev)(void *priv,
2673 enum rtrs_clt_link_ev ev),
2674 unsigned int reconnect_delay_sec,
2675 unsigned int max_reconnect_attempts)
2676 {
2677 struct rtrs_clt *clt;
2678 int err;
2679
2680 if (!paths_num || paths_num > MAX_PATHS_NUM)
2681 return ERR_PTR(-EINVAL);
2682
2683 if (strlen(sessname) >= sizeof(clt->sessname))
2684 return ERR_PTR(-EINVAL);
2685
2686 clt = kzalloc(sizeof(*clt), GFP_KERNEL);
2687 if (!clt)
2688 return ERR_PTR(-ENOMEM);
2689
2690 clt->pcpu_path = alloc_percpu(typeof(*clt->pcpu_path));
2691 if (!clt->pcpu_path) {
2692 kfree(clt);
2693 return ERR_PTR(-ENOMEM);
2694 }
2695
2696 uuid_gen(&clt->paths_uuid);
2697 INIT_LIST_HEAD_RCU(&clt->paths_list);
2698 clt->paths_num = paths_num;
2699 clt->paths_up = MAX_PATHS_NUM;
2700 clt->port = port;
2701 clt->pdu_sz = pdu_sz;
2702 clt->max_segments = RTRS_MAX_SEGMENTS;
2703 clt->reconnect_delay_sec = reconnect_delay_sec;
2704 clt->max_reconnect_attempts = max_reconnect_attempts;
2705 clt->priv = priv;
2706 clt->link_ev = link_ev;
2707 clt->mp_policy = MP_POLICY_MIN_INFLIGHT;
2708 strscpy(clt->sessname, sessname, sizeof(clt->sessname));
2709 init_waitqueue_head(&clt->permits_wait);
2710 mutex_init(&clt->paths_ev_mutex);
2711 mutex_init(&clt->paths_mutex);
2712
2713 clt->dev.class = rtrs_clt_dev_class;
2714 clt->dev.release = rtrs_clt_dev_release;
2715 err = dev_set_name(&clt->dev, "%s", sessname);
2716 if (err)
2717 goto err;
2718 /*
2719 * Suppress user space notification until
2720 * sysfs files are created
2721 */
2722 dev_set_uevent_suppress(&clt->dev, true);
2723 err = device_register(&clt->dev);
2724 if (err) {
2725 put_device(&clt->dev);
2726 goto err;
2727 }
2728
2729 clt->kobj_paths = kobject_create_and_add("paths", &clt->dev.kobj);
2730 if (!clt->kobj_paths) {
2731 err = -ENOMEM;
2732 goto err_dev;
2733 }
2734 err = rtrs_clt_create_sysfs_root_files(clt);
2735 if (err) {
2736 kobject_del(clt->kobj_paths);
2737 kobject_put(clt->kobj_paths);
2738 goto err_dev;
2739 }
2740 dev_set_uevent_suppress(&clt->dev, false);
2741 kobject_uevent(&clt->dev.kobj, KOBJ_ADD);
2742
2743 return clt;
2744 err_dev:
2745 device_unregister(&clt->dev);
2746 err:
2747 free_percpu(clt->pcpu_path);
2748 kfree(clt);
2749 return ERR_PTR(err);
2750 }
2751
free_clt(struct rtrs_clt * clt)2752 static void free_clt(struct rtrs_clt *clt)
2753 {
2754 free_permits(clt);
2755 free_percpu(clt->pcpu_path);
2756 mutex_destroy(&clt->paths_ev_mutex);
2757 mutex_destroy(&clt->paths_mutex);
2758 /* release callback will free clt in last put */
2759 device_unregister(&clt->dev);
2760 }
2761
2762 /**
2763 * rtrs_clt_open() - Open a session to an RTRS server
2764 * @ops: holds the link event callback and the private pointer.
2765 * @sessname: name of the session
2766 * @paths: Paths to be established defined by their src and dst addresses
2767 * @paths_num: Number of elements in the @paths array
2768 * @port: port to be used by the RTRS session
2769 * @pdu_sz: Size of extra payload which can be accessed after permit allocation.
2770 * @reconnect_delay_sec: time between reconnect tries
2771 * @max_reconnect_attempts: Number of times to reconnect on error before giving
2772 * up, 0 for * disabled, -1 for forever
2773 * @nr_poll_queues: number of polling mode connection using IB_POLL_DIRECT flag
2774 *
2775 * Starts session establishment with the rtrs_server. The function can block
2776 * up to ~2000ms before it returns.
2777 *
2778 * Return a valid pointer on success otherwise PTR_ERR.
2779 */
rtrs_clt_open(struct rtrs_clt_ops * ops,const char * sessname,const struct rtrs_addr * paths,size_t paths_num,u16 port,size_t pdu_sz,u8 reconnect_delay_sec,s16 max_reconnect_attempts,u32 nr_poll_queues)2780 struct rtrs_clt *rtrs_clt_open(struct rtrs_clt_ops *ops,
2781 const char *sessname,
2782 const struct rtrs_addr *paths,
2783 size_t paths_num, u16 port,
2784 size_t pdu_sz, u8 reconnect_delay_sec,
2785 s16 max_reconnect_attempts, u32 nr_poll_queues)
2786 {
2787 struct rtrs_clt_sess *sess, *tmp;
2788 struct rtrs_clt *clt;
2789 int err, i;
2790
2791 clt = alloc_clt(sessname, paths_num, port, pdu_sz, ops->priv,
2792 ops->link_ev,
2793 reconnect_delay_sec,
2794 max_reconnect_attempts);
2795 if (IS_ERR(clt)) {
2796 err = PTR_ERR(clt);
2797 goto out;
2798 }
2799 for (i = 0; i < paths_num; i++) {
2800 struct rtrs_clt_sess *sess;
2801
2802 sess = alloc_sess(clt, &paths[i], nr_cpu_ids,
2803 nr_poll_queues);
2804 if (IS_ERR(sess)) {
2805 err = PTR_ERR(sess);
2806 goto close_all_sess;
2807 }
2808 if (!i)
2809 sess->for_new_clt = 1;
2810 list_add_tail_rcu(&sess->s.entry, &clt->paths_list);
2811
2812 err = init_sess(sess);
2813 if (err) {
2814 list_del_rcu(&sess->s.entry);
2815 rtrs_clt_close_conns(sess, true);
2816 free_percpu(sess->stats->pcpu_stats);
2817 kfree(sess->stats);
2818 free_sess(sess);
2819 goto close_all_sess;
2820 }
2821
2822 err = rtrs_clt_create_sess_files(sess);
2823 if (err) {
2824 list_del_rcu(&sess->s.entry);
2825 rtrs_clt_close_conns(sess, true);
2826 free_percpu(sess->stats->pcpu_stats);
2827 kfree(sess->stats);
2828 free_sess(sess);
2829 goto close_all_sess;
2830 }
2831 }
2832 err = alloc_permits(clt);
2833 if (err)
2834 goto close_all_sess;
2835
2836 return clt;
2837
2838 close_all_sess:
2839 list_for_each_entry_safe(sess, tmp, &clt->paths_list, s.entry) {
2840 rtrs_clt_destroy_sess_files(sess, NULL);
2841 rtrs_clt_close_conns(sess, true);
2842 kobject_put(&sess->kobj);
2843 }
2844 rtrs_clt_destroy_sysfs_root(clt);
2845 free_clt(clt);
2846
2847 out:
2848 return ERR_PTR(err);
2849 }
2850 EXPORT_SYMBOL(rtrs_clt_open);
2851
2852 /**
2853 * rtrs_clt_close() - Close a session
2854 * @clt: Session handle. Session is freed upon return.
2855 */
rtrs_clt_close(struct rtrs_clt * clt)2856 void rtrs_clt_close(struct rtrs_clt *clt)
2857 {
2858 struct rtrs_clt_sess *sess, *tmp;
2859
2860 /* Firstly forbid sysfs access */
2861 rtrs_clt_destroy_sysfs_root(clt);
2862
2863 /* Now it is safe to iterate over all paths without locks */
2864 list_for_each_entry_safe(sess, tmp, &clt->paths_list, s.entry) {
2865 rtrs_clt_close_conns(sess, true);
2866 rtrs_clt_destroy_sess_files(sess, NULL);
2867 kobject_put(&sess->kobj);
2868 }
2869 free_clt(clt);
2870 }
2871 EXPORT_SYMBOL(rtrs_clt_close);
2872
rtrs_clt_reconnect_from_sysfs(struct rtrs_clt_sess * sess)2873 int rtrs_clt_reconnect_from_sysfs(struct rtrs_clt_sess *sess)
2874 {
2875 enum rtrs_clt_state old_state;
2876 int err = -EBUSY;
2877 bool changed;
2878
2879 changed = rtrs_clt_change_state_get_old(sess, RTRS_CLT_RECONNECTING,
2880 &old_state);
2881 if (changed) {
2882 sess->reconnect_attempts = 0;
2883 queue_delayed_work(rtrs_wq, &sess->reconnect_dwork, 0);
2884 }
2885 if (changed || old_state == RTRS_CLT_RECONNECTING) {
2886 /*
2887 * flush_delayed_work() queues pending work for immediate
2888 * execution, so do the flush if we have queued something
2889 * right now or work is pending.
2890 */
2891 flush_delayed_work(&sess->reconnect_dwork);
2892 err = (READ_ONCE(sess->state) ==
2893 RTRS_CLT_CONNECTED ? 0 : -ENOTCONN);
2894 }
2895
2896 return err;
2897 }
2898
rtrs_clt_remove_path_from_sysfs(struct rtrs_clt_sess * sess,const struct attribute * sysfs_self)2899 int rtrs_clt_remove_path_from_sysfs(struct rtrs_clt_sess *sess,
2900 const struct attribute *sysfs_self)
2901 {
2902 enum rtrs_clt_state old_state;
2903 bool changed;
2904
2905 /*
2906 * Continue stopping path till state was changed to DEAD or
2907 * state was observed as DEAD:
2908 * 1. State was changed to DEAD - we were fast and nobody
2909 * invoked rtrs_clt_reconnect(), which can again start
2910 * reconnecting.
2911 * 2. State was observed as DEAD - we have someone in parallel
2912 * removing the path.
2913 */
2914 do {
2915 rtrs_clt_close_conns(sess, true);
2916 changed = rtrs_clt_change_state_get_old(sess,
2917 RTRS_CLT_DEAD,
2918 &old_state);
2919 } while (!changed && old_state != RTRS_CLT_DEAD);
2920
2921 if (changed) {
2922 rtrs_clt_remove_path_from_arr(sess);
2923 rtrs_clt_destroy_sess_files(sess, sysfs_self);
2924 kobject_put(&sess->kobj);
2925 }
2926
2927 return 0;
2928 }
2929
rtrs_clt_set_max_reconnect_attempts(struct rtrs_clt * clt,int value)2930 void rtrs_clt_set_max_reconnect_attempts(struct rtrs_clt *clt, int value)
2931 {
2932 clt->max_reconnect_attempts = (unsigned int)value;
2933 }
2934
rtrs_clt_get_max_reconnect_attempts(const struct rtrs_clt * clt)2935 int rtrs_clt_get_max_reconnect_attempts(const struct rtrs_clt *clt)
2936 {
2937 return (int)clt->max_reconnect_attempts;
2938 }
2939
2940 /**
2941 * rtrs_clt_request() - Request data transfer to/from server via RDMA.
2942 *
2943 * @dir: READ/WRITE
2944 * @ops: callback function to be called as confirmation, and the pointer.
2945 * @clt: Session
2946 * @permit: Preallocated permit
2947 * @vec: Message that is sent to server together with the request.
2948 * Sum of len of all @vec elements limited to <= IO_MSG_SIZE.
2949 * Since the msg is copied internally it can be allocated on stack.
2950 * @nr: Number of elements in @vec.
2951 * @data_len: length of data sent to/from server
2952 * @sg: Pages to be sent/received to/from server.
2953 * @sg_cnt: Number of elements in the @sg
2954 *
2955 * Return:
2956 * 0: Success
2957 * <0: Error
2958 *
2959 * On dir=READ rtrs client will request a data transfer from Server to client.
2960 * The data that the server will respond with will be stored in @sg when
2961 * the user receives an %RTRS_CLT_RDMA_EV_RDMA_REQUEST_WRITE_COMPL event.
2962 * On dir=WRITE rtrs client will rdma write data in sg to server side.
2963 */
rtrs_clt_request(int dir,struct rtrs_clt_req_ops * ops,struct rtrs_clt * clt,struct rtrs_permit * permit,const struct kvec * vec,size_t nr,size_t data_len,struct scatterlist * sg,unsigned int sg_cnt)2964 int rtrs_clt_request(int dir, struct rtrs_clt_req_ops *ops,
2965 struct rtrs_clt *clt, struct rtrs_permit *permit,
2966 const struct kvec *vec, size_t nr, size_t data_len,
2967 struct scatterlist *sg, unsigned int sg_cnt)
2968 {
2969 struct rtrs_clt_io_req *req;
2970 struct rtrs_clt_sess *sess;
2971
2972 enum dma_data_direction dma_dir;
2973 int err = -ECONNABORTED, i;
2974 size_t usr_len, hdr_len;
2975 struct path_it it;
2976
2977 /* Get kvec length */
2978 for (i = 0, usr_len = 0; i < nr; i++)
2979 usr_len += vec[i].iov_len;
2980
2981 if (dir == READ) {
2982 hdr_len = sizeof(struct rtrs_msg_rdma_read) +
2983 sg_cnt * sizeof(struct rtrs_sg_desc);
2984 dma_dir = DMA_FROM_DEVICE;
2985 } else {
2986 hdr_len = sizeof(struct rtrs_msg_rdma_write);
2987 dma_dir = DMA_TO_DEVICE;
2988 }
2989
2990 rcu_read_lock();
2991 for (path_it_init(&it, clt);
2992 (sess = it.next_path(&it)) && it.i < it.clt->paths_num; it.i++) {
2993 if (READ_ONCE(sess->state) != RTRS_CLT_CONNECTED)
2994 continue;
2995
2996 if (usr_len + hdr_len > sess->max_hdr_size) {
2997 rtrs_wrn_rl(sess->clt,
2998 "%s request failed, user message size is %zu and header length %zu, but max size is %u\n",
2999 dir == READ ? "Read" : "Write",
3000 usr_len, hdr_len, sess->max_hdr_size);
3001 err = -EMSGSIZE;
3002 break;
3003 }
3004 req = rtrs_clt_get_req(sess, ops->conf_fn, permit, ops->priv,
3005 vec, usr_len, sg, sg_cnt, data_len,
3006 dma_dir);
3007 if (dir == READ)
3008 err = rtrs_clt_read_req(req);
3009 else
3010 err = rtrs_clt_write_req(req);
3011 if (err) {
3012 req->in_use = false;
3013 continue;
3014 }
3015 /* Success path */
3016 break;
3017 }
3018 path_it_deinit(&it);
3019 rcu_read_unlock();
3020
3021 return err;
3022 }
3023 EXPORT_SYMBOL(rtrs_clt_request);
3024
rtrs_clt_rdma_cq_direct(struct rtrs_clt * clt,unsigned int index)3025 int rtrs_clt_rdma_cq_direct(struct rtrs_clt *clt, unsigned int index)
3026 {
3027 /* If no path, return -1 for block layer not to try again */
3028 int cnt = -1;
3029 struct rtrs_con *con;
3030 struct rtrs_clt_sess *sess;
3031 struct path_it it;
3032
3033 rcu_read_lock();
3034 for (path_it_init(&it, clt);
3035 (sess = it.next_path(&it)) && it.i < it.clt->paths_num; it.i++) {
3036 if (READ_ONCE(sess->state) != RTRS_CLT_CONNECTED)
3037 continue;
3038
3039 con = sess->s.con[index + 1];
3040 cnt = ib_process_cq_direct(con->cq, -1);
3041 if (cnt)
3042 break;
3043 }
3044 path_it_deinit(&it);
3045 rcu_read_unlock();
3046
3047 return cnt;
3048 }
3049 EXPORT_SYMBOL(rtrs_clt_rdma_cq_direct);
3050
3051 /**
3052 * rtrs_clt_query() - queries RTRS session attributes
3053 *@clt: session pointer
3054 *@attr: query results for session attributes.
3055 * Returns:
3056 * 0 on success
3057 * -ECOMM no connection to the server
3058 */
rtrs_clt_query(struct rtrs_clt * clt,struct rtrs_attrs * attr)3059 int rtrs_clt_query(struct rtrs_clt *clt, struct rtrs_attrs *attr)
3060 {
3061 if (!rtrs_clt_is_connected(clt))
3062 return -ECOMM;
3063
3064 attr->queue_depth = clt->queue_depth;
3065 attr->max_segments = clt->max_segments;
3066 /* Cap max_io_size to min of remote buffer size and the fr pages */
3067 attr->max_io_size = min_t(int, clt->max_io_size,
3068 clt->max_segments * SZ_4K);
3069
3070 return 0;
3071 }
3072 EXPORT_SYMBOL(rtrs_clt_query);
3073
rtrs_clt_create_path_from_sysfs(struct rtrs_clt * clt,struct rtrs_addr * addr)3074 int rtrs_clt_create_path_from_sysfs(struct rtrs_clt *clt,
3075 struct rtrs_addr *addr)
3076 {
3077 struct rtrs_clt_sess *sess;
3078 int err;
3079
3080 sess = alloc_sess(clt, addr, nr_cpu_ids, 0);
3081 if (IS_ERR(sess))
3082 return PTR_ERR(sess);
3083
3084 mutex_lock(&clt->paths_mutex);
3085 if (clt->paths_num == 0) {
3086 /*
3087 * When all the paths are removed for a session,
3088 * the addition of the first path is like a new session for
3089 * the storage server
3090 */
3091 sess->for_new_clt = 1;
3092 }
3093
3094 mutex_unlock(&clt->paths_mutex);
3095
3096 /*
3097 * It is totally safe to add path in CONNECTING state: coming
3098 * IO will never grab it. Also it is very important to add
3099 * path before init, since init fires LINK_CONNECTED event.
3100 */
3101 rtrs_clt_add_path_to_arr(sess);
3102
3103 err = init_sess(sess);
3104 if (err)
3105 goto close_sess;
3106
3107 err = rtrs_clt_create_sess_files(sess);
3108 if (err)
3109 goto close_sess;
3110
3111 return 0;
3112
3113 close_sess:
3114 rtrs_clt_remove_path_from_arr(sess);
3115 rtrs_clt_close_conns(sess, true);
3116 free_percpu(sess->stats->pcpu_stats);
3117 kfree(sess->stats);
3118 free_sess(sess);
3119
3120 return err;
3121 }
3122
rtrs_clt_ib_dev_init(struct rtrs_ib_dev * dev)3123 static int rtrs_clt_ib_dev_init(struct rtrs_ib_dev *dev)
3124 {
3125 if (!(dev->ib_dev->attrs.device_cap_flags &
3126 IB_DEVICE_MEM_MGT_EXTENSIONS)) {
3127 pr_err("Memory registrations not supported.\n");
3128 return -ENOTSUPP;
3129 }
3130
3131 return 0;
3132 }
3133
3134 static const struct rtrs_rdma_dev_pd_ops dev_pd_ops = {
3135 .init = rtrs_clt_ib_dev_init
3136 };
3137
rtrs_client_init(void)3138 static int __init rtrs_client_init(void)
3139 {
3140 rtrs_rdma_dev_pd_init(0, &dev_pd);
3141
3142 rtrs_clt_dev_class = class_create(THIS_MODULE, "rtrs-client");
3143 if (IS_ERR(rtrs_clt_dev_class)) {
3144 pr_err("Failed to create rtrs-client dev class\n");
3145 return PTR_ERR(rtrs_clt_dev_class);
3146 }
3147 rtrs_wq = alloc_workqueue("rtrs_client_wq", 0, 0);
3148 if (!rtrs_wq) {
3149 class_destroy(rtrs_clt_dev_class);
3150 return -ENOMEM;
3151 }
3152
3153 return 0;
3154 }
3155
rtrs_client_exit(void)3156 static void __exit rtrs_client_exit(void)
3157 {
3158 destroy_workqueue(rtrs_wq);
3159 class_destroy(rtrs_clt_dev_class);
3160 rtrs_rdma_dev_pd_deinit(&dev_pd);
3161 }
3162
3163 module_init(rtrs_client_init);
3164 module_exit(rtrs_client_exit);
3165
