Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
[linux-2.6-block.git] / net / sunrpc / xprtrdma / transport.c
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
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
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
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
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
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
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
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 *
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  */
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
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
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  */
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
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
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 = -ENOMEM;
514         xprt_add_backlog(xprt, task);
515 }
516
517 /**
518  * xprt_rdma_free_slot - release an rpc_rqst
519  * @xprt: controlling RPC transport
520  * @rqst: rpc_rqst to release
521  *
522  */
523 static void
524 xprt_rdma_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *rqst)
525 {
526         struct rpcrdma_xprt *r_xprt =
527                 container_of(xprt, struct rpcrdma_xprt, rx_xprt);
528
529         rpcrdma_reply_put(&r_xprt->rx_buf, rpcr_to_rdmar(rqst));
530         if (!xprt_wake_up_backlog(xprt, rqst)) {
531                 memset(rqst, 0, sizeof(*rqst));
532                 rpcrdma_buffer_put(&r_xprt->rx_buf, rpcr_to_rdmar(rqst));
533         }
534 }
535
536 static bool rpcrdma_check_regbuf(struct rpcrdma_xprt *r_xprt,
537                                  struct rpcrdma_regbuf *rb, size_t size,
538                                  gfp_t flags)
539 {
540         if (unlikely(rdmab_length(rb) < size)) {
541                 if (!rpcrdma_regbuf_realloc(rb, size, flags))
542                         return false;
543                 r_xprt->rx_stats.hardway_register_count += size;
544         }
545         return true;
546 }
547
548 /**
549  * xprt_rdma_allocate - allocate transport resources for an RPC
550  * @task: RPC task
551  *
552  * Return values:
553  *        0:    Success; rq_buffer points to RPC buffer to use
554  *   ENOMEM:    Out of memory, call again later
555  *      EIO:    A permanent error occurred, do not retry
556  */
557 static int
558 xprt_rdma_allocate(struct rpc_task *task)
559 {
560         struct rpc_rqst *rqst = task->tk_rqstp;
561         struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(rqst->rq_xprt);
562         struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
563         gfp_t flags = rpc_task_gfp_mask();
564
565         if (!rpcrdma_check_regbuf(r_xprt, req->rl_sendbuf, rqst->rq_callsize,
566                                   flags))
567                 goto out_fail;
568         if (!rpcrdma_check_regbuf(r_xprt, req->rl_recvbuf, rqst->rq_rcvsize,
569                                   flags))
570                 goto out_fail;
571
572         rqst->rq_buffer = rdmab_data(req->rl_sendbuf);
573         rqst->rq_rbuffer = rdmab_data(req->rl_recvbuf);
574         return 0;
575
576 out_fail:
577         return -ENOMEM;
578 }
579
580 /**
581  * xprt_rdma_free - release resources allocated by xprt_rdma_allocate
582  * @task: RPC task
583  *
584  * Caller guarantees rqst->rq_buffer is non-NULL.
585  */
586 static void
587 xprt_rdma_free(struct rpc_task *task)
588 {
589         struct rpc_rqst *rqst = task->tk_rqstp;
590         struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
591
592         if (unlikely(!list_empty(&req->rl_registered))) {
593                 trace_xprtrdma_mrs_zap(task);
594                 frwr_unmap_sync(rpcx_to_rdmax(rqst->rq_xprt), req);
595         }
596
597         /* XXX: If the RPC is completing because of a signal and
598          * not because a reply was received, we ought to ensure
599          * that the Send completion has fired, so that memory
600          * involved with the Send is not still visible to the NIC.
601          */
602 }
603
604 /**
605  * xprt_rdma_send_request - marshal and send an RPC request
606  * @rqst: RPC message in rq_snd_buf
607  *
608  * Caller holds the transport's write lock.
609  *
610  * Returns:
611  *      %0 if the RPC message has been sent
612  *      %-ENOTCONN if the caller should reconnect and call again
613  *      %-EAGAIN if the caller should call again
614  *      %-ENOBUFS if the caller should call again after a delay
615  *      %-EMSGSIZE if encoding ran out of buffer space. The request
616  *              was not sent. Do not try to send this message again.
617  *      %-EIO if an I/O error occurred. The request was not sent.
618  *              Do not try to send this message again.
619  */
620 static int
621 xprt_rdma_send_request(struct rpc_rqst *rqst)
622 {
623         struct rpc_xprt *xprt = rqst->rq_xprt;
624         struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
625         struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
626         int rc = 0;
627
628 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
629         if (unlikely(!rqst->rq_buffer))
630                 return xprt_rdma_bc_send_reply(rqst);
631 #endif  /* CONFIG_SUNRPC_BACKCHANNEL */
632
633         if (!xprt_connected(xprt))
634                 return -ENOTCONN;
635
636         if (!xprt_request_get_cong(xprt, rqst))
637                 return -EBADSLT;
638
639         rc = rpcrdma_marshal_req(r_xprt, rqst);
640         if (rc < 0)
641                 goto failed_marshal;
642
643         /* Must suppress retransmit to maintain credits */
644         if (rqst->rq_connect_cookie == xprt->connect_cookie)
645                 goto drop_connection;
646         rqst->rq_xtime = ktime_get();
647
648         if (frwr_send(r_xprt, req))
649                 goto drop_connection;
650
651         rqst->rq_xmit_bytes_sent += rqst->rq_snd_buf.len;
652
653         /* An RPC with no reply will throw off credit accounting,
654          * so drop the connection to reset the credit grant.
655          */
656         if (!rpc_reply_expected(rqst->rq_task))
657                 goto drop_connection;
658         return 0;
659
660 failed_marshal:
661         if (rc != -ENOTCONN)
662                 return rc;
663 drop_connection:
664         xprt_rdma_close(xprt);
665         return -ENOTCONN;
666 }
667
668 void xprt_rdma_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
669 {
670         struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
671         long idle_time = 0;
672
673         if (xprt_connected(xprt))
674                 idle_time = (long)(jiffies - xprt->last_used) / HZ;
675
676         seq_puts(seq, "\txprt:\trdma ");
677         seq_printf(seq, "%u %lu %lu %lu %ld %lu %lu %lu %llu %llu ",
678                    0,   /* need a local port? */
679                    xprt->stat.bind_count,
680                    xprt->stat.connect_count,
681                    xprt->stat.connect_time / HZ,
682                    idle_time,
683                    xprt->stat.sends,
684                    xprt->stat.recvs,
685                    xprt->stat.bad_xids,
686                    xprt->stat.req_u,
687                    xprt->stat.bklog_u);
688         seq_printf(seq, "%lu %lu %lu %llu %llu %llu %llu %lu %lu %lu %lu ",
689                    r_xprt->rx_stats.read_chunk_count,
690                    r_xprt->rx_stats.write_chunk_count,
691                    r_xprt->rx_stats.reply_chunk_count,
692                    r_xprt->rx_stats.total_rdma_request,
693                    r_xprt->rx_stats.total_rdma_reply,
694                    r_xprt->rx_stats.pullup_copy_count,
695                    r_xprt->rx_stats.fixup_copy_count,
696                    r_xprt->rx_stats.hardway_register_count,
697                    r_xprt->rx_stats.failed_marshal_count,
698                    r_xprt->rx_stats.bad_reply_count,
699                    r_xprt->rx_stats.nomsg_call_count);
700         seq_printf(seq, "%lu %lu %lu %lu %lu %lu\n",
701                    r_xprt->rx_stats.mrs_recycled,
702                    r_xprt->rx_stats.mrs_orphaned,
703                    r_xprt->rx_stats.mrs_allocated,
704                    r_xprt->rx_stats.local_inv_needed,
705                    r_xprt->rx_stats.empty_sendctx_q,
706                    r_xprt->rx_stats.reply_waits_for_send);
707 }
708
709 static int
710 xprt_rdma_enable_swap(struct rpc_xprt *xprt)
711 {
712         return 0;
713 }
714
715 static void
716 xprt_rdma_disable_swap(struct rpc_xprt *xprt)
717 {
718 }
719
720 /*
721  * Plumbing for rpc transport switch and kernel module
722  */
723
724 static const struct rpc_xprt_ops xprt_rdma_procs = {
725         .reserve_xprt           = xprt_reserve_xprt_cong,
726         .release_xprt           = xprt_release_xprt_cong, /* sunrpc/xprt.c */
727         .alloc_slot             = xprt_rdma_alloc_slot,
728         .free_slot              = xprt_rdma_free_slot,
729         .release_request        = xprt_release_rqst_cong,       /* ditto */
730         .wait_for_reply_request = xprt_wait_for_reply_request_def, /* ditto */
731         .timer                  = xprt_rdma_timer,
732         .rpcbind                = rpcb_getport_async,   /* sunrpc/rpcb_clnt.c */
733         .set_port               = xprt_rdma_set_port,
734         .connect                = xprt_rdma_connect,
735         .buf_alloc              = xprt_rdma_allocate,
736         .buf_free               = xprt_rdma_free,
737         .send_request           = xprt_rdma_send_request,
738         .close                  = xprt_rdma_close,
739         .destroy                = xprt_rdma_destroy,
740         .set_connect_timeout    = xprt_rdma_set_connect_timeout,
741         .print_stats            = xprt_rdma_print_stats,
742         .enable_swap            = xprt_rdma_enable_swap,
743         .disable_swap           = xprt_rdma_disable_swap,
744         .inject_disconnect      = xprt_rdma_inject_disconnect,
745 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
746         .bc_setup               = xprt_rdma_bc_setup,
747         .bc_maxpayload          = xprt_rdma_bc_maxpayload,
748         .bc_num_slots           = xprt_rdma_bc_max_slots,
749         .bc_free_rqst           = xprt_rdma_bc_free_rqst,
750         .bc_destroy             = xprt_rdma_bc_destroy,
751 #endif
752 };
753
754 static struct xprt_class xprt_rdma = {
755         .list                   = LIST_HEAD_INIT(xprt_rdma.list),
756         .name                   = "rdma",
757         .owner                  = THIS_MODULE,
758         .ident                  = XPRT_TRANSPORT_RDMA,
759         .setup                  = xprt_setup_rdma,
760         .netid                  = { "rdma", "rdma6", "" },
761 };
762
763 void xprt_rdma_cleanup(void)
764 {
765 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
766         if (sunrpc_table_header) {
767                 unregister_sysctl_table(sunrpc_table_header);
768                 sunrpc_table_header = NULL;
769         }
770 #endif
771
772         xprt_unregister_transport(&xprt_rdma);
773         xprt_unregister_transport(&xprt_rdma_bc);
774 }
775
776 int xprt_rdma_init(void)
777 {
778         int rc;
779
780         rc = xprt_register_transport(&xprt_rdma);
781         if (rc)
782                 return rc;
783
784         rc = xprt_register_transport(&xprt_rdma_bc);
785         if (rc) {
786                 xprt_unregister_transport(&xprt_rdma);
787                 return rc;
788         }
789
790 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
791         if (!sunrpc_table_header)
792                 sunrpc_table_header = register_sysctl("sunrpc", xr_tunables_table);
793 #endif
794         return 0;
795 }