1 // SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
2 /*
3 * Copyright (c) 2014-2017 Oracle. All rights reserved.
4 * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved.
5 *
6 * This software is available to you under a choice of one of two
7 * licenses. You may choose to be licensed under the terms of the GNU
8 * General Public License (GPL) Version 2, available from the file
9 * COPYING in the main directory of this source tree, or the BSD-type
10 * license below:
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 *
16 * Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 *
19 * Redistributions in binary form must reproduce the above
20 * copyright notice, this list of conditions and the following
21 * disclaimer in the documentation and/or other materials provided
22 * with the distribution.
23 *
24 * Neither the name of the Network Appliance, Inc. nor the names of
25 * its contributors may be used to endorse or promote products
26 * derived from this software without specific prior written
27 * permission.
28 *
29 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
30 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
31 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
32 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
33 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
34 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
35 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
36 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
37 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
38 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
39 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
40 */
41
42 /*
43 * transport.c
44 *
45 * This file contains the top-level implementation of an RPC RDMA
46 * transport.
47 *
48 * Naming convention: functions beginning with xprt_ are part of the
49 * transport switch. All others are RPC RDMA internal.
50 */
51
52 #include <linux/module.h>
53 #include <linux/slab.h>
54 #include <linux/seq_file.h>
55 #include <linux/smp.h>
56
57 #include <linux/sunrpc/addr.h>
58 #include <linux/sunrpc/svc_rdma.h>
59
60 #include "xprt_rdma.h"
61 #include <trace/events/rpcrdma.h>
62
63 /*
64 * tunables
65 */
66
67 static unsigned int xprt_rdma_slot_table_entries = RPCRDMA_DEF_SLOT_TABLE;
68 unsigned int xprt_rdma_max_inline_read = RPCRDMA_DEF_INLINE;
69 unsigned int xprt_rdma_max_inline_write = RPCRDMA_DEF_INLINE;
70 unsigned int xprt_rdma_memreg_strategy = RPCRDMA_FRWR;
71 int xprt_rdma_pad_optimize;
72 static struct xprt_class xprt_rdma;
73
74 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
75
76 static unsigned int min_slot_table_size = RPCRDMA_MIN_SLOT_TABLE;
77 static unsigned int max_slot_table_size = RPCRDMA_MAX_SLOT_TABLE;
78 static unsigned int min_inline_size = RPCRDMA_MIN_INLINE;
79 static unsigned int max_inline_size = RPCRDMA_MAX_INLINE;
80 static unsigned int max_padding = PAGE_SIZE;
81 static unsigned int min_memreg = RPCRDMA_BOUNCEBUFFERS;
82 static unsigned int max_memreg = RPCRDMA_LAST - 1;
83 static unsigned int dummy;
84
85 static struct ctl_table_header *sunrpc_table_header;
86
87 static struct ctl_table xr_tunables_table[] = {
88 {
89 .procname = "rdma_slot_table_entries",
90 .data = &xprt_rdma_slot_table_entries,
91 .maxlen = sizeof(unsigned int),
92 .mode = 0644,
93 .proc_handler = proc_dointvec_minmax,
94 .extra1 = &min_slot_table_size,
95 .extra2 = &max_slot_table_size
96 },
97 {
98 .procname = "rdma_max_inline_read",
99 .data = &xprt_rdma_max_inline_read,
100 .maxlen = sizeof(unsigned int),
101 .mode = 0644,
102 .proc_handler = proc_dointvec_minmax,
103 .extra1 = &min_inline_size,
104 .extra2 = &max_inline_size,
105 },
106 {
107 .procname = "rdma_max_inline_write",
108 .data = &xprt_rdma_max_inline_write,
109 .maxlen = sizeof(unsigned int),
110 .mode = 0644,
111 .proc_handler = proc_dointvec_minmax,
112 .extra1 = &min_inline_size,
113 .extra2 = &max_inline_size,
114 },
115 {
116 .procname = "rdma_inline_write_padding",
117 .data = &dummy,
118 .maxlen = sizeof(unsigned int),
119 .mode = 0644,
120 .proc_handler = proc_dointvec_minmax,
121 .extra1 = SYSCTL_ZERO,
122 .extra2 = &max_padding,
123 },
124 {
125 .procname = "rdma_memreg_strategy",
126 .data = &xprt_rdma_memreg_strategy,
127 .maxlen = sizeof(unsigned int),
128 .mode = 0644,
129 .proc_handler = proc_dointvec_minmax,
130 .extra1 = &min_memreg,
131 .extra2 = &max_memreg,
132 },
133 {
134 .procname = "rdma_pad_optimize",
135 .data = &xprt_rdma_pad_optimize,
136 .maxlen = sizeof(unsigned int),
137 .mode = 0644,
138 .proc_handler = proc_dointvec,
139 },
140 };
141
142 #endif
143
144 static const struct rpc_xprt_ops xprt_rdma_procs;
145
146 static void
xprt_rdma_format_addresses4(struct rpc_xprt * xprt,struct sockaddr * sap)147 xprt_rdma_format_addresses4(struct rpc_xprt *xprt, struct sockaddr *sap)
148 {
149 struct sockaddr_in *sin = (struct sockaddr_in *)sap;
150 char buf[20];
151
152 snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr));
153 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL);
154
155 xprt->address_strings[RPC_DISPLAY_NETID] = RPCBIND_NETID_RDMA;
156 }
157
158 static void
xprt_rdma_format_addresses6(struct rpc_xprt * xprt,struct sockaddr * sap)159 xprt_rdma_format_addresses6(struct rpc_xprt *xprt, struct sockaddr *sap)
160 {
161 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sap;
162 char buf[40];
163
164 snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr);
165 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL);
166
167 xprt->address_strings[RPC_DISPLAY_NETID] = RPCBIND_NETID_RDMA6;
168 }
169
170 void
xprt_rdma_format_addresses(struct rpc_xprt * xprt,struct sockaddr * sap)171 xprt_rdma_format_addresses(struct rpc_xprt *xprt, struct sockaddr *sap)
172 {
173 char buf[128];
174
175 switch (sap->sa_family) {
176 case AF_INET:
177 xprt_rdma_format_addresses4(xprt, sap);
178 break;
179 case AF_INET6:
180 xprt_rdma_format_addresses6(xprt, sap);
181 break;
182 default:
183 pr_err("rpcrdma: Unrecognized address family\n");
184 return;
185 }
186
187 (void)rpc_ntop(sap, buf, sizeof(buf));
188 xprt->address_strings[RPC_DISPLAY_ADDR] = kstrdup(buf, GFP_KERNEL);
189
190 snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap));
191 xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL);
192
193 snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap));
194 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
195
196 xprt->address_strings[RPC_DISPLAY_PROTO] = "rdma";
197 }
198
199 void
xprt_rdma_free_addresses(struct rpc_xprt * xprt)200 xprt_rdma_free_addresses(struct rpc_xprt *xprt)
201 {
202 unsigned int i;
203
204 for (i = 0; i < RPC_DISPLAY_MAX; i++)
205 switch (i) {
206 case RPC_DISPLAY_PROTO:
207 case RPC_DISPLAY_NETID:
208 continue;
209 default:
210 kfree(xprt->address_strings[i]);
211 }
212 }
213
214 /**
215 * xprt_rdma_connect_worker - establish connection in the background
216 * @work: worker thread context
217 *
218 * Requester holds the xprt's send lock to prevent activity on this
219 * transport while a fresh connection is being established. RPC tasks
220 * sleep on the xprt's pending queue waiting for connect to complete.
221 */
222 static void
xprt_rdma_connect_worker(struct work_struct * work)223 xprt_rdma_connect_worker(struct work_struct *work)
224 {
225 struct rpcrdma_xprt *r_xprt = container_of(work, struct rpcrdma_xprt,
226 rx_connect_worker.work);
227 struct rpc_xprt *xprt = &r_xprt->rx_xprt;
228 unsigned int pflags = current->flags;
229 int rc;
230
231 if (atomic_read(&xprt->swapper))
232 current->flags |= PF_MEMALLOC;
233 rc = rpcrdma_xprt_connect(r_xprt);
234 xprt_clear_connecting(xprt);
235 if (!rc) {
236 xprt->connect_cookie++;
237 xprt->stat.connect_count++;
238 xprt->stat.connect_time += (long)jiffies -
239 xprt->stat.connect_start;
240 xprt_set_connected(xprt);
241 rc = -EAGAIN;
242 } else
243 rpcrdma_xprt_disconnect(r_xprt);
244 xprt_unlock_connect(xprt, r_xprt);
245 xprt_wake_pending_tasks(xprt, rc);
246 current_restore_flags(pflags, PF_MEMALLOC);
247 }
248
249 /**
250 * xprt_rdma_inject_disconnect - inject a connection fault
251 * @xprt: transport context
252 *
253 * If @xprt is connected, disconnect it to simulate spurious
254 * connection loss. Caller must hold @xprt's send lock to
255 * ensure that data structures and hardware resources are
256 * stable during the rdma_disconnect() call.
257 */
258 static void
xprt_rdma_inject_disconnect(struct rpc_xprt * xprt)259 xprt_rdma_inject_disconnect(struct rpc_xprt *xprt)
260 {
261 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
262
263 trace_xprtrdma_op_inject_dsc(r_xprt);
264 rdma_disconnect(r_xprt->rx_ep->re_id);
265 }
266
267 /**
268 * xprt_rdma_destroy - Full tear down of transport
269 * @xprt: doomed transport context
270 *
271 * Caller guarantees there will be no more calls to us with
272 * this @xprt.
273 */
274 static void
xprt_rdma_destroy(struct rpc_xprt * xprt)275 xprt_rdma_destroy(struct rpc_xprt *xprt)
276 {
277 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
278
279 cancel_delayed_work_sync(&r_xprt->rx_connect_worker);
280
281 rpcrdma_xprt_disconnect(r_xprt);
282 rpcrdma_buffer_destroy(&r_xprt->rx_buf);
283
284 xprt_rdma_free_addresses(xprt);
285 xprt_free(xprt);
286
287 module_put(THIS_MODULE);
288 }
289
290 /* 60 second timeout, no retries */
291 static const struct rpc_timeout xprt_rdma_default_timeout = {
292 .to_initval = 60 * HZ,
293 .to_maxval = 60 * HZ,
294 };
295
296 /**
297 * xprt_setup_rdma - Set up transport to use RDMA
298 *
299 * @args: rpc transport arguments
300 */
301 static struct rpc_xprt *
xprt_setup_rdma(struct xprt_create * args)302 xprt_setup_rdma(struct xprt_create *args)
303 {
304 struct rpc_xprt *xprt;
305 struct rpcrdma_xprt *new_xprt;
306 struct sockaddr *sap;
307 int rc;
308
309 if (args->addrlen > sizeof(xprt->addr))
310 return ERR_PTR(-EBADF);
311
312 if (!try_module_get(THIS_MODULE))
313 return ERR_PTR(-EIO);
314
315 xprt = xprt_alloc(args->net, sizeof(struct rpcrdma_xprt), 0,
316 xprt_rdma_slot_table_entries);
317 if (!xprt) {
318 module_put(THIS_MODULE);
319 return ERR_PTR(-ENOMEM);
320 }
321
322 xprt->timeout = &xprt_rdma_default_timeout;
323 xprt->connect_timeout = xprt->timeout->to_initval;
324 xprt->max_reconnect_timeout = xprt->timeout->to_maxval;
325 xprt->bind_timeout = RPCRDMA_BIND_TO;
326 xprt->reestablish_timeout = RPCRDMA_INIT_REEST_TO;
327 xprt->idle_timeout = RPCRDMA_IDLE_DISC_TO;
328
329 xprt->resvport = 0; /* privileged port not needed */
330 xprt->ops = &xprt_rdma_procs;
331
332 /*
333 * Set up RDMA-specific connect data.
334 */
335 sap = args->dstaddr;
336
337 /* Ensure xprt->addr holds valid server TCP (not RDMA)
338 * address, for any side protocols which peek at it */
339 xprt->prot = IPPROTO_TCP;
340 xprt->xprt_class = &xprt_rdma;
341 xprt->addrlen = args->addrlen;
342 memcpy(&xprt->addr, sap, xprt->addrlen);
343
344 if (rpc_get_port(sap))
345 xprt_set_bound(xprt);
346 xprt_rdma_format_addresses(xprt, sap);
347
348 new_xprt = rpcx_to_rdmax(xprt);
349 rc = rpcrdma_buffer_create(new_xprt);
350 if (rc) {
351 xprt_rdma_free_addresses(xprt);
352 xprt_free(xprt);
353 module_put(THIS_MODULE);
354 return ERR_PTR(rc);
355 }
356
357 INIT_DELAYED_WORK(&new_xprt->rx_connect_worker,
358 xprt_rdma_connect_worker);
359
360 xprt->max_payload = RPCRDMA_MAX_DATA_SEGS << PAGE_SHIFT;
361
362 return xprt;
363 }
364
365 /**
366 * xprt_rdma_close - close a transport connection
367 * @xprt: transport context
368 *
369 * Called during autoclose or device removal.
370 *
371 * Caller holds @xprt's send lock to prevent activity on this
372 * transport while the connection is torn down.
373 */
xprt_rdma_close(struct rpc_xprt * xprt)374 void xprt_rdma_close(struct rpc_xprt *xprt)
375 {
376 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
377
378 rpcrdma_xprt_disconnect(r_xprt);
379
380 xprt->reestablish_timeout = 0;
381 ++xprt->connect_cookie;
382 xprt_disconnect_done(xprt);
383 }
384
385 /**
386 * xprt_rdma_set_port - update server port with rpcbind result
387 * @xprt: controlling RPC transport
388 * @port: new port value
389 *
390 * Transport connect status is unchanged.
391 */
392 static void
xprt_rdma_set_port(struct rpc_xprt * xprt,u16 port)393 xprt_rdma_set_port(struct rpc_xprt *xprt, u16 port)
394 {
395 struct sockaddr *sap = (struct sockaddr *)&xprt->addr;
396 char buf[8];
397
398 rpc_set_port(sap, port);
399
400 kfree(xprt->address_strings[RPC_DISPLAY_PORT]);
401 snprintf(buf, sizeof(buf), "%u", port);
402 xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL);
403
404 kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]);
405 snprintf(buf, sizeof(buf), "%4hx", port);
406 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
407 }
408
409 /**
410 * xprt_rdma_timer - invoked when an RPC times out
411 * @xprt: controlling RPC transport
412 * @task: RPC task that timed out
413 *
414 * Invoked when the transport is still connected, but an RPC
415 * retransmit timeout occurs.
416 *
417 * Since RDMA connections don't have a keep-alive, forcibly
418 * disconnect and retry to connect. This drives full
419 * detection of the network path, and retransmissions of
420 * all pending RPCs.
421 */
422 static void
xprt_rdma_timer(struct rpc_xprt * xprt,struct rpc_task * task)423 xprt_rdma_timer(struct rpc_xprt *xprt, struct rpc_task *task)
424 {
425 xprt_force_disconnect(xprt);
426 }
427
428 /**
429 * xprt_rdma_set_connect_timeout - set timeouts for establishing a connection
430 * @xprt: controlling transport instance
431 * @connect_timeout: reconnect timeout after client disconnects
432 * @reconnect_timeout: reconnect timeout after server disconnects
433 *
434 */
xprt_rdma_set_connect_timeout(struct rpc_xprt * xprt,unsigned long connect_timeout,unsigned long reconnect_timeout)435 static void xprt_rdma_set_connect_timeout(struct rpc_xprt *xprt,
436 unsigned long connect_timeout,
437 unsigned long reconnect_timeout)
438 {
439 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
440
441 trace_xprtrdma_op_set_cto(r_xprt, connect_timeout, reconnect_timeout);
442
443 spin_lock(&xprt->transport_lock);
444
445 if (connect_timeout < xprt->connect_timeout) {
446 struct rpc_timeout to;
447 unsigned long initval;
448
449 to = *xprt->timeout;
450 initval = connect_timeout;
451 if (initval < RPCRDMA_INIT_REEST_TO << 1)
452 initval = RPCRDMA_INIT_REEST_TO << 1;
453 to.to_initval = initval;
454 to.to_maxval = initval;
455 r_xprt->rx_timeout = to;
456 xprt->timeout = &r_xprt->rx_timeout;
457 xprt->connect_timeout = connect_timeout;
458 }
459
460 if (reconnect_timeout < xprt->max_reconnect_timeout)
461 xprt->max_reconnect_timeout = reconnect_timeout;
462
463 spin_unlock(&xprt->transport_lock);
464 }
465
466 /**
467 * xprt_rdma_connect - schedule an attempt to reconnect
468 * @xprt: transport state
469 * @task: RPC scheduler context (unused)
470 *
471 */
472 static void
xprt_rdma_connect(struct rpc_xprt * xprt,struct rpc_task * task)473 xprt_rdma_connect(struct rpc_xprt *xprt, struct rpc_task *task)
474 {
475 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
476 struct rpcrdma_ep *ep = r_xprt->rx_ep;
477 unsigned long delay;
478
479 WARN_ON_ONCE(!xprt_lock_connect(xprt, task, r_xprt));
480
481 delay = 0;
482 if (ep && ep->re_connect_status != 0) {
483 delay = xprt_reconnect_delay(xprt);
484 xprt_reconnect_backoff(xprt, RPCRDMA_INIT_REEST_TO);
485 }
486 trace_xprtrdma_op_connect(r_xprt, delay);
487 queue_delayed_work(system_long_wq, &r_xprt->rx_connect_worker, delay);
488 }
489
490 /**
491 * xprt_rdma_alloc_slot - allocate an rpc_rqst
492 * @xprt: controlling RPC transport
493 * @task: RPC task requesting a fresh rpc_rqst
494 *
495 * tk_status values:
496 * %0 if task->tk_rqstp points to a fresh rpc_rqst
497 * %-EAGAIN if no rpc_rqst is available; queued on backlog
498 */
499 static void
xprt_rdma_alloc_slot(struct rpc_xprt * xprt,struct rpc_task * task)500 xprt_rdma_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
501 {
502 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
503 struct rpcrdma_req *req;
504
505 req = rpcrdma_buffer_get(&r_xprt->rx_buf);
506 if (!req)
507 goto out_sleep;
508 task->tk_rqstp = &req->rl_slot;
509 task->tk_status = 0;
510 return;
511
512 out_sleep:
513 task->tk_status = -EAGAIN;
514 xprt_add_backlog_noncongested(xprt, task);
515 /* A buffer freed between buffer_get and rpc_sleep_on
516 * goes back to the pool with no waiter to wake.
517 * Re-check after joining the backlog to close that gap.
518 */
519 req = rpcrdma_buffer_get(&r_xprt->rx_buf);
520 if (req) {
521 struct rpc_rqst *rqst = &req->rl_slot;
522
523 if (!xprt_wake_up_backlog(xprt, rqst)) {
524 memset(rqst, 0, sizeof(*rqst));
525 rpcrdma_buffer_put(&r_xprt->rx_buf, req);
526 }
527 }
528 }
529
530 /**
531 * xprt_rdma_free_slot - release an rpc_rqst
532 * @xprt: controlling RPC transport
533 * @rqst: rpc_rqst to release
534 *
535 */
536 static void
xprt_rdma_free_slot(struct rpc_xprt * xprt,struct rpc_rqst * rqst)537 xprt_rdma_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *rqst)
538 {
539 struct rpcrdma_xprt *r_xprt =
540 container_of(xprt, struct rpcrdma_xprt, rx_xprt);
541
542 rpcrdma_reply_put(&r_xprt->rx_buf, rpcr_to_rdmar(rqst));
543 if (!xprt_wake_up_backlog(xprt, rqst)) {
544 memset(rqst, 0, sizeof(*rqst));
545 rpcrdma_buffer_put(&r_xprt->rx_buf, rpcr_to_rdmar(rqst));
546 }
547 }
548
rpcrdma_check_regbuf(struct rpcrdma_xprt * r_xprt,struct rpcrdma_regbuf * rb,size_t size,gfp_t flags)549 static bool rpcrdma_check_regbuf(struct rpcrdma_xprt *r_xprt,
550 struct rpcrdma_regbuf *rb, size_t size,
551 gfp_t flags)
552 {
553 if (unlikely(rdmab_length(rb) < size)) {
554 if (!rpcrdma_regbuf_realloc(rb, size, flags))
555 return false;
556 r_xprt->rx_stats.hardway_register_count += size;
557 }
558 return true;
559 }
560
561 /**
562 * xprt_rdma_allocate - allocate transport resources for an RPC
563 * @task: RPC task
564 *
565 * Return values:
566 * 0: Success; rq_buffer points to RPC buffer to use
567 * ENOMEM: Out of memory, call again later
568 * EIO: A permanent error occurred, do not retry
569 */
570 static int
xprt_rdma_allocate(struct rpc_task * task)571 xprt_rdma_allocate(struct rpc_task *task)
572 {
573 struct rpc_rqst *rqst = task->tk_rqstp;
574 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(rqst->rq_xprt);
575 struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
576 gfp_t flags = rpc_task_gfp_mask();
577
578 if (!rpcrdma_check_regbuf(r_xprt, req->rl_sendbuf, rqst->rq_callsize,
579 flags))
580 goto out_fail;
581 if (!rpcrdma_check_regbuf(r_xprt, req->rl_recvbuf, rqst->rq_rcvsize,
582 flags))
583 goto out_fail;
584
585 rqst->rq_buffer = rdmab_data(req->rl_sendbuf);
586 rqst->rq_rbuffer = rdmab_data(req->rl_recvbuf);
587 return 0;
588
589 out_fail:
590 return -ENOMEM;
591 }
592
593 /**
594 * xprt_rdma_free - release resources allocated by xprt_rdma_allocate
595 * @task: RPC task
596 *
597 * Caller guarantees rqst->rq_buffer is non-NULL.
598 */
599 static void
xprt_rdma_free(struct rpc_task * task)600 xprt_rdma_free(struct rpc_task *task)
601 {
602 struct rpc_rqst *rqst = task->tk_rqstp;
603 struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
604
605 if (unlikely(!list_empty(&req->rl_registered))) {
606 trace_xprtrdma_mrs_zap(task);
607 frwr_unmap_sync(rpcx_to_rdmax(rqst->rq_xprt), req);
608 }
609
610 /* XXX: If the RPC is completing because of a signal and
611 * not because a reply was received, we ought to ensure
612 * that the Send completion has fired, so that memory
613 * involved with the Send is not still visible to the NIC.
614 */
615 }
616
617 /**
618 * xprt_rdma_send_request - marshal and send an RPC request
619 * @rqst: RPC message in rq_snd_buf
620 *
621 * Caller holds the transport's write lock.
622 *
623 * Returns:
624 * %0 if the RPC message has been sent
625 * %-ENOTCONN if the caller should reconnect and call again
626 * %-EAGAIN if the caller should call again
627 * %-ENOBUFS if the caller should call again after a delay
628 * %-EMSGSIZE if encoding ran out of buffer space. The request
629 * was not sent. Do not try to send this message again.
630 * %-EIO if an I/O error occurred. The request was not sent.
631 * Do not try to send this message again.
632 */
633 static int
xprt_rdma_send_request(struct rpc_rqst * rqst)634 xprt_rdma_send_request(struct rpc_rqst *rqst)
635 {
636 struct rpc_xprt *xprt = rqst->rq_xprt;
637 struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
638 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
639 int rc = 0;
640
641 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
642 if (unlikely(!rqst->rq_buffer))
643 return xprt_rdma_bc_send_reply(rqst);
644 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
645
646 if (!xprt_connected(xprt))
647 return -ENOTCONN;
648
649 if (!xprt_request_get_cong(xprt, rqst))
650 return -EBADSLT;
651
652 rc = rpcrdma_marshal_req(r_xprt, rqst);
653 if (rc < 0)
654 goto failed_marshal;
655
656 /* Must suppress retransmit to maintain credits */
657 if (rqst->rq_connect_cookie == xprt->connect_cookie)
658 goto drop_connection;
659 rqst->rq_xtime = ktime_get();
660
661 if (frwr_send(r_xprt, req))
662 goto drop_connection;
663
664 rqst->rq_xmit_bytes_sent += rqst->rq_snd_buf.len;
665
666 /* An RPC with no reply will throw off credit accounting,
667 * so drop the connection to reset the credit grant.
668 */
669 if (!rpc_reply_expected(rqst->rq_task))
670 goto drop_connection;
671 return 0;
672
673 failed_marshal:
674 if (rc != -ENOTCONN)
675 return rc;
676 drop_connection:
677 xprt_rdma_close(xprt);
678 return -ENOTCONN;
679 }
680
xprt_rdma_print_stats(struct rpc_xprt * xprt,struct seq_file * seq)681 void xprt_rdma_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
682 {
683 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
684 long idle_time = 0;
685
686 if (xprt_connected(xprt))
687 idle_time = (long)(jiffies - xprt->last_used) / HZ;
688
689 seq_puts(seq, "\txprt:\trdma ");
690 seq_printf(seq, "%u %lu %lu %lu %ld %lu %lu %lu %llu %llu ",
691 0, /* need a local port? */
692 xprt->stat.bind_count,
693 xprt->stat.connect_count,
694 xprt->stat.connect_time / HZ,
695 idle_time,
696 xprt->stat.sends,
697 xprt->stat.recvs,
698 xprt->stat.bad_xids,
699 xprt->stat.req_u,
700 xprt->stat.bklog_u);
701 seq_printf(seq, "%lu %lu %lu %llu %llu %llu %llu %lu %lu %lu %lu ",
702 r_xprt->rx_stats.read_chunk_count,
703 r_xprt->rx_stats.write_chunk_count,
704 r_xprt->rx_stats.reply_chunk_count,
705 r_xprt->rx_stats.total_rdma_request,
706 r_xprt->rx_stats.total_rdma_reply,
707 r_xprt->rx_stats.pullup_copy_count,
708 r_xprt->rx_stats.fixup_copy_count,
709 r_xprt->rx_stats.hardway_register_count,
710 r_xprt->rx_stats.failed_marshal_count,
711 r_xprt->rx_stats.bad_reply_count,
712 r_xprt->rx_stats.nomsg_call_count);
713 seq_printf(seq, "%lu %lu %lu %lu %lu %lu\n",
714 r_xprt->rx_stats.mrs_recycled,
715 r_xprt->rx_stats.mrs_orphaned,
716 r_xprt->rx_stats.mrs_allocated,
717 r_xprt->rx_stats.local_inv_needed,
718 r_xprt->rx_stats.empty_sendctx_q,
719 r_xprt->rx_stats.reply_waits_for_send);
720 }
721
722 static int
xprt_rdma_enable_swap(struct rpc_xprt * xprt)723 xprt_rdma_enable_swap(struct rpc_xprt *xprt)
724 {
725 return 0;
726 }
727
728 static void
xprt_rdma_disable_swap(struct rpc_xprt * xprt)729 xprt_rdma_disable_swap(struct rpc_xprt *xprt)
730 {
731 }
732
733 /*
734 * Plumbing for rpc transport switch and kernel module
735 */
736
737 static const struct rpc_xprt_ops xprt_rdma_procs = {
738 .reserve_xprt = xprt_reserve_xprt_cong,
739 .release_xprt = xprt_release_xprt_cong, /* sunrpc/xprt.c */
740 .alloc_slot = xprt_rdma_alloc_slot,
741 .free_slot = xprt_rdma_free_slot,
742 .release_request = xprt_release_rqst_cong, /* ditto */
743 .wait_for_reply_request = xprt_wait_for_reply_request_def, /* ditto */
744 .timer = xprt_rdma_timer,
745 .rpcbind = rpcb_getport_async, /* sunrpc/rpcb_clnt.c */
746 .set_port = xprt_rdma_set_port,
747 .connect = xprt_rdma_connect,
748 .buf_alloc = xprt_rdma_allocate,
749 .buf_free = xprt_rdma_free,
750 .send_request = xprt_rdma_send_request,
751 .close = xprt_rdma_close,
752 .destroy = xprt_rdma_destroy,
753 .set_connect_timeout = xprt_rdma_set_connect_timeout,
754 .print_stats = xprt_rdma_print_stats,
755 .enable_swap = xprt_rdma_enable_swap,
756 .disable_swap = xprt_rdma_disable_swap,
757 .inject_disconnect = xprt_rdma_inject_disconnect,
758 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
759 .bc_setup = xprt_rdma_bc_setup,
760 .bc_maxpayload = xprt_rdma_bc_maxpayload,
761 .bc_num_slots = xprt_rdma_bc_max_slots,
762 .bc_free_rqst = xprt_rdma_bc_free_rqst,
763 .bc_destroy = xprt_rdma_bc_destroy,
764 #endif
765 };
766
767 static struct xprt_class xprt_rdma = {
768 .list = LIST_HEAD_INIT(xprt_rdma.list),
769 .name = "rdma",
770 .owner = THIS_MODULE,
771 .ident = XPRT_TRANSPORT_RDMA,
772 .setup = xprt_setup_rdma,
773 .netid = { "rdma", "rdma6", "" },
774 };
775
xprt_rdma_cleanup(void)776 void xprt_rdma_cleanup(void)
777 {
778 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
779 if (sunrpc_table_header) {
780 unregister_sysctl_table(sunrpc_table_header);
781 sunrpc_table_header = NULL;
782 }
783 #endif
784
785 xprt_unregister_transport(&xprt_rdma);
786 xprt_unregister_transport(&xprt_rdma_bc);
787 }
788
xprt_rdma_init(void)789 int xprt_rdma_init(void)
790 {
791 int rc;
792
793 rc = xprt_register_transport(&xprt_rdma);
794 if (rc)
795 return rc;
796
797 rc = xprt_register_transport(&xprt_rdma_bc);
798 if (rc) {
799 xprt_unregister_transport(&xprt_rdma);
800 return rc;
801 }
802
803 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
804 if (!sunrpc_table_header)
805 sunrpc_table_header = register_sysctl("sunrpc", xr_tunables_table);
806 #endif
807 return 0;
808 }
809