xref: /openbmc/linux/net/sunrpc/xprt.c (revision 825c7f4a)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/net/sunrpc/xprt.c
4  *
5  *  This is a generic RPC call interface supporting congestion avoidance,
6  *  and asynchronous calls.
7  *
8  *  The interface works like this:
9  *
10  *  -	When a process places a call, it allocates a request slot if
11  *	one is available. Otherwise, it sleeps on the backlog queue
12  *	(xprt_reserve).
13  *  -	Next, the caller puts together the RPC message, stuffs it into
14  *	the request struct, and calls xprt_transmit().
15  *  -	xprt_transmit sends the message and installs the caller on the
16  *	transport's wait list. At the same time, if a reply is expected,
17  *	it installs a timer that is run after the packet's timeout has
18  *	expired.
19  *  -	When a packet arrives, the data_ready handler walks the list of
20  *	pending requests for that transport. If a matching XID is found, the
21  *	caller is woken up, and the timer removed.
22  *  -	When no reply arrives within the timeout interval, the timer is
23  *	fired by the kernel and runs xprt_timer(). It either adjusts the
24  *	timeout values (minor timeout) or wakes up the caller with a status
25  *	of -ETIMEDOUT.
26  *  -	When the caller receives a notification from RPC that a reply arrived,
27  *	it should release the RPC slot, and process the reply.
28  *	If the call timed out, it may choose to retry the operation by
29  *	adjusting the initial timeout value, and simply calling rpc_call
30  *	again.
31  *
32  *  Support for async RPC is done through a set of RPC-specific scheduling
33  *  primitives that `transparently' work for processes as well as async
34  *  tasks that rely on callbacks.
35  *
36  *  Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
37  *
38  *  Transport switch API copyright (C) 2005, Chuck Lever <cel@netapp.com>
39  */
40 
41 #include <linux/module.h>
42 
43 #include <linux/types.h>
44 #include <linux/interrupt.h>
45 #include <linux/workqueue.h>
46 #include <linux/net.h>
47 #include <linux/ktime.h>
48 
49 #include <linux/sunrpc/clnt.h>
50 #include <linux/sunrpc/metrics.h>
51 #include <linux/sunrpc/bc_xprt.h>
52 #include <linux/rcupdate.h>
53 #include <linux/sched/mm.h>
54 
55 #include <trace/events/sunrpc.h>
56 
57 #include "sunrpc.h"
58 
59 /*
60  * Local variables
61  */
62 
63 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
64 # define RPCDBG_FACILITY	RPCDBG_XPRT
65 #endif
66 
67 /*
68  * Local functions
69  */
70 static void	 xprt_init(struct rpc_xprt *xprt, struct net *net);
71 static __be32	xprt_alloc_xid(struct rpc_xprt *xprt);
72 static void	 xprt_destroy(struct rpc_xprt *xprt);
73 
74 static DEFINE_SPINLOCK(xprt_list_lock);
75 static LIST_HEAD(xprt_list);
76 
77 static unsigned long xprt_request_timeout(const struct rpc_rqst *req)
78 {
79 	unsigned long timeout = jiffies + req->rq_timeout;
80 
81 	if (time_before(timeout, req->rq_majortimeo))
82 		return timeout;
83 	return req->rq_majortimeo;
84 }
85 
86 /**
87  * xprt_register_transport - register a transport implementation
88  * @transport: transport to register
89  *
90  * If a transport implementation is loaded as a kernel module, it can
91  * call this interface to make itself known to the RPC client.
92  *
93  * Returns:
94  * 0:		transport successfully registered
95  * -EEXIST:	transport already registered
96  * -EINVAL:	transport module being unloaded
97  */
98 int xprt_register_transport(struct xprt_class *transport)
99 {
100 	struct xprt_class *t;
101 	int result;
102 
103 	result = -EEXIST;
104 	spin_lock(&xprt_list_lock);
105 	list_for_each_entry(t, &xprt_list, list) {
106 		/* don't register the same transport class twice */
107 		if (t->ident == transport->ident)
108 			goto out;
109 	}
110 
111 	list_add_tail(&transport->list, &xprt_list);
112 	printk(KERN_INFO "RPC: Registered %s transport module.\n",
113 	       transport->name);
114 	result = 0;
115 
116 out:
117 	spin_unlock(&xprt_list_lock);
118 	return result;
119 }
120 EXPORT_SYMBOL_GPL(xprt_register_transport);
121 
122 /**
123  * xprt_unregister_transport - unregister a transport implementation
124  * @transport: transport to unregister
125  *
126  * Returns:
127  * 0:		transport successfully unregistered
128  * -ENOENT:	transport never registered
129  */
130 int xprt_unregister_transport(struct xprt_class *transport)
131 {
132 	struct xprt_class *t;
133 	int result;
134 
135 	result = 0;
136 	spin_lock(&xprt_list_lock);
137 	list_for_each_entry(t, &xprt_list, list) {
138 		if (t == transport) {
139 			printk(KERN_INFO
140 				"RPC: Unregistered %s transport module.\n",
141 				transport->name);
142 			list_del_init(&transport->list);
143 			goto out;
144 		}
145 	}
146 	result = -ENOENT;
147 
148 out:
149 	spin_unlock(&xprt_list_lock);
150 	return result;
151 }
152 EXPORT_SYMBOL_GPL(xprt_unregister_transport);
153 
154 /**
155  * xprt_load_transport - load a transport implementation
156  * @transport_name: transport to load
157  *
158  * Returns:
159  * 0:		transport successfully loaded
160  * -ENOENT:	transport module not available
161  */
162 int xprt_load_transport(const char *transport_name)
163 {
164 	struct xprt_class *t;
165 	int result;
166 
167 	result = 0;
168 	spin_lock(&xprt_list_lock);
169 	list_for_each_entry(t, &xprt_list, list) {
170 		if (strcmp(t->name, transport_name) == 0) {
171 			spin_unlock(&xprt_list_lock);
172 			goto out;
173 		}
174 	}
175 	spin_unlock(&xprt_list_lock);
176 	result = request_module("xprt%s", transport_name);
177 out:
178 	return result;
179 }
180 EXPORT_SYMBOL_GPL(xprt_load_transport);
181 
182 static void xprt_clear_locked(struct rpc_xprt *xprt)
183 {
184 	xprt->snd_task = NULL;
185 	if (!test_bit(XPRT_CLOSE_WAIT, &xprt->state)) {
186 		smp_mb__before_atomic();
187 		clear_bit(XPRT_LOCKED, &xprt->state);
188 		smp_mb__after_atomic();
189 	} else
190 		queue_work(xprtiod_workqueue, &xprt->task_cleanup);
191 }
192 
193 /**
194  * xprt_reserve_xprt - serialize write access to transports
195  * @task: task that is requesting access to the transport
196  * @xprt: pointer to the target transport
197  *
198  * This prevents mixing the payload of separate requests, and prevents
199  * transport connects from colliding with writes.  No congestion control
200  * is provided.
201  */
202 int xprt_reserve_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
203 {
204 	struct rpc_rqst *req = task->tk_rqstp;
205 
206 	if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
207 		if (task == xprt->snd_task)
208 			goto out_locked;
209 		goto out_sleep;
210 	}
211 	if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
212 		goto out_unlock;
213 	xprt->snd_task = task;
214 
215 out_locked:
216 	trace_xprt_reserve_xprt(xprt, task);
217 	return 1;
218 
219 out_unlock:
220 	xprt_clear_locked(xprt);
221 out_sleep:
222 	task->tk_status = -EAGAIN;
223 	if  (RPC_IS_SOFT(task))
224 		rpc_sleep_on_timeout(&xprt->sending, task, NULL,
225 				xprt_request_timeout(req));
226 	else
227 		rpc_sleep_on(&xprt->sending, task, NULL);
228 	return 0;
229 }
230 EXPORT_SYMBOL_GPL(xprt_reserve_xprt);
231 
232 static bool
233 xprt_need_congestion_window_wait(struct rpc_xprt *xprt)
234 {
235 	return test_bit(XPRT_CWND_WAIT, &xprt->state);
236 }
237 
238 static void
239 xprt_set_congestion_window_wait(struct rpc_xprt *xprt)
240 {
241 	if (!list_empty(&xprt->xmit_queue)) {
242 		/* Peek at head of queue to see if it can make progress */
243 		if (list_first_entry(&xprt->xmit_queue, struct rpc_rqst,
244 					rq_xmit)->rq_cong)
245 			return;
246 	}
247 	set_bit(XPRT_CWND_WAIT, &xprt->state);
248 }
249 
250 static void
251 xprt_test_and_clear_congestion_window_wait(struct rpc_xprt *xprt)
252 {
253 	if (!RPCXPRT_CONGESTED(xprt))
254 		clear_bit(XPRT_CWND_WAIT, &xprt->state);
255 }
256 
257 /*
258  * xprt_reserve_xprt_cong - serialize write access to transports
259  * @task: task that is requesting access to the transport
260  *
261  * Same as xprt_reserve_xprt, but Van Jacobson congestion control is
262  * integrated into the decision of whether a request is allowed to be
263  * woken up and given access to the transport.
264  * Note that the lock is only granted if we know there are free slots.
265  */
266 int xprt_reserve_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
267 {
268 	struct rpc_rqst *req = task->tk_rqstp;
269 
270 	if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
271 		if (task == xprt->snd_task)
272 			goto out_locked;
273 		goto out_sleep;
274 	}
275 	if (req == NULL) {
276 		xprt->snd_task = task;
277 		goto out_locked;
278 	}
279 	if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
280 		goto out_unlock;
281 	if (!xprt_need_congestion_window_wait(xprt)) {
282 		xprt->snd_task = task;
283 		goto out_locked;
284 	}
285 out_unlock:
286 	xprt_clear_locked(xprt);
287 out_sleep:
288 	task->tk_status = -EAGAIN;
289 	if (RPC_IS_SOFT(task))
290 		rpc_sleep_on_timeout(&xprt->sending, task, NULL,
291 				xprt_request_timeout(req));
292 	else
293 		rpc_sleep_on(&xprt->sending, task, NULL);
294 	return 0;
295 out_locked:
296 	trace_xprt_reserve_cong(xprt, task);
297 	return 1;
298 }
299 EXPORT_SYMBOL_GPL(xprt_reserve_xprt_cong);
300 
301 static inline int xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
302 {
303 	int retval;
304 
305 	if (test_bit(XPRT_LOCKED, &xprt->state) && xprt->snd_task == task)
306 		return 1;
307 	spin_lock(&xprt->transport_lock);
308 	retval = xprt->ops->reserve_xprt(xprt, task);
309 	spin_unlock(&xprt->transport_lock);
310 	return retval;
311 }
312 
313 static bool __xprt_lock_write_func(struct rpc_task *task, void *data)
314 {
315 	struct rpc_xprt *xprt = data;
316 
317 	xprt->snd_task = task;
318 	return true;
319 }
320 
321 static void __xprt_lock_write_next(struct rpc_xprt *xprt)
322 {
323 	if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
324 		return;
325 	if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
326 		goto out_unlock;
327 	if (rpc_wake_up_first_on_wq(xprtiod_workqueue, &xprt->sending,
328 				__xprt_lock_write_func, xprt))
329 		return;
330 out_unlock:
331 	xprt_clear_locked(xprt);
332 }
333 
334 static void __xprt_lock_write_next_cong(struct rpc_xprt *xprt)
335 {
336 	if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
337 		return;
338 	if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
339 		goto out_unlock;
340 	if (xprt_need_congestion_window_wait(xprt))
341 		goto out_unlock;
342 	if (rpc_wake_up_first_on_wq(xprtiod_workqueue, &xprt->sending,
343 				__xprt_lock_write_func, xprt))
344 		return;
345 out_unlock:
346 	xprt_clear_locked(xprt);
347 }
348 
349 /**
350  * xprt_release_xprt - allow other requests to use a transport
351  * @xprt: transport with other tasks potentially waiting
352  * @task: task that is releasing access to the transport
353  *
354  * Note that "task" can be NULL.  No congestion control is provided.
355  */
356 void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
357 {
358 	if (xprt->snd_task == task) {
359 		xprt_clear_locked(xprt);
360 		__xprt_lock_write_next(xprt);
361 	}
362 	trace_xprt_release_xprt(xprt, task);
363 }
364 EXPORT_SYMBOL_GPL(xprt_release_xprt);
365 
366 /**
367  * xprt_release_xprt_cong - allow other requests to use a transport
368  * @xprt: transport with other tasks potentially waiting
369  * @task: task that is releasing access to the transport
370  *
371  * Note that "task" can be NULL.  Another task is awoken to use the
372  * transport if the transport's congestion window allows it.
373  */
374 void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
375 {
376 	if (xprt->snd_task == task) {
377 		xprt_clear_locked(xprt);
378 		__xprt_lock_write_next_cong(xprt);
379 	}
380 	trace_xprt_release_cong(xprt, task);
381 }
382 EXPORT_SYMBOL_GPL(xprt_release_xprt_cong);
383 
384 static inline void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
385 {
386 	if (xprt->snd_task != task)
387 		return;
388 	spin_lock(&xprt->transport_lock);
389 	xprt->ops->release_xprt(xprt, task);
390 	spin_unlock(&xprt->transport_lock);
391 }
392 
393 /*
394  * Van Jacobson congestion avoidance. Check if the congestion window
395  * overflowed. Put the task to sleep if this is the case.
396  */
397 static int
398 __xprt_get_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
399 {
400 	if (req->rq_cong)
401 		return 1;
402 	trace_xprt_get_cong(xprt, req->rq_task);
403 	if (RPCXPRT_CONGESTED(xprt)) {
404 		xprt_set_congestion_window_wait(xprt);
405 		return 0;
406 	}
407 	req->rq_cong = 1;
408 	xprt->cong += RPC_CWNDSCALE;
409 	return 1;
410 }
411 
412 /*
413  * Adjust the congestion window, and wake up the next task
414  * that has been sleeping due to congestion
415  */
416 static void
417 __xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
418 {
419 	if (!req->rq_cong)
420 		return;
421 	req->rq_cong = 0;
422 	xprt->cong -= RPC_CWNDSCALE;
423 	xprt_test_and_clear_congestion_window_wait(xprt);
424 	trace_xprt_put_cong(xprt, req->rq_task);
425 	__xprt_lock_write_next_cong(xprt);
426 }
427 
428 /**
429  * xprt_request_get_cong - Request congestion control credits
430  * @xprt: pointer to transport
431  * @req: pointer to RPC request
432  *
433  * Useful for transports that require congestion control.
434  */
435 bool
436 xprt_request_get_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
437 {
438 	bool ret = false;
439 
440 	if (req->rq_cong)
441 		return true;
442 	spin_lock(&xprt->transport_lock);
443 	ret = __xprt_get_cong(xprt, req) != 0;
444 	spin_unlock(&xprt->transport_lock);
445 	return ret;
446 }
447 EXPORT_SYMBOL_GPL(xprt_request_get_cong);
448 
449 /**
450  * xprt_release_rqst_cong - housekeeping when request is complete
451  * @task: RPC request that recently completed
452  *
453  * Useful for transports that require congestion control.
454  */
455 void xprt_release_rqst_cong(struct rpc_task *task)
456 {
457 	struct rpc_rqst *req = task->tk_rqstp;
458 
459 	__xprt_put_cong(req->rq_xprt, req);
460 }
461 EXPORT_SYMBOL_GPL(xprt_release_rqst_cong);
462 
463 static void xprt_clear_congestion_window_wait_locked(struct rpc_xprt *xprt)
464 {
465 	if (test_and_clear_bit(XPRT_CWND_WAIT, &xprt->state))
466 		__xprt_lock_write_next_cong(xprt);
467 }
468 
469 /*
470  * Clear the congestion window wait flag and wake up the next
471  * entry on xprt->sending
472  */
473 static void
474 xprt_clear_congestion_window_wait(struct rpc_xprt *xprt)
475 {
476 	if (test_and_clear_bit(XPRT_CWND_WAIT, &xprt->state)) {
477 		spin_lock(&xprt->transport_lock);
478 		__xprt_lock_write_next_cong(xprt);
479 		spin_unlock(&xprt->transport_lock);
480 	}
481 }
482 
483 /**
484  * xprt_adjust_cwnd - adjust transport congestion window
485  * @xprt: pointer to xprt
486  * @task: recently completed RPC request used to adjust window
487  * @result: result code of completed RPC request
488  *
489  * The transport code maintains an estimate on the maximum number of out-
490  * standing RPC requests, using a smoothed version of the congestion
491  * avoidance implemented in 44BSD. This is basically the Van Jacobson
492  * congestion algorithm: If a retransmit occurs, the congestion window is
493  * halved; otherwise, it is incremented by 1/cwnd when
494  *
495  *	-	a reply is received and
496  *	-	a full number of requests are outstanding and
497  *	-	the congestion window hasn't been updated recently.
498  */
499 void xprt_adjust_cwnd(struct rpc_xprt *xprt, struct rpc_task *task, int result)
500 {
501 	struct rpc_rqst *req = task->tk_rqstp;
502 	unsigned long cwnd = xprt->cwnd;
503 
504 	if (result >= 0 && cwnd <= xprt->cong) {
505 		/* The (cwnd >> 1) term makes sure
506 		 * the result gets rounded properly. */
507 		cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
508 		if (cwnd > RPC_MAXCWND(xprt))
509 			cwnd = RPC_MAXCWND(xprt);
510 		__xprt_lock_write_next_cong(xprt);
511 	} else if (result == -ETIMEDOUT) {
512 		cwnd >>= 1;
513 		if (cwnd < RPC_CWNDSCALE)
514 			cwnd = RPC_CWNDSCALE;
515 	}
516 	dprintk("RPC:       cong %ld, cwnd was %ld, now %ld\n",
517 			xprt->cong, xprt->cwnd, cwnd);
518 	xprt->cwnd = cwnd;
519 	__xprt_put_cong(xprt, req);
520 }
521 EXPORT_SYMBOL_GPL(xprt_adjust_cwnd);
522 
523 /**
524  * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue
525  * @xprt: transport with waiting tasks
526  * @status: result code to plant in each task before waking it
527  *
528  */
529 void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status)
530 {
531 	if (status < 0)
532 		rpc_wake_up_status(&xprt->pending, status);
533 	else
534 		rpc_wake_up(&xprt->pending);
535 }
536 EXPORT_SYMBOL_GPL(xprt_wake_pending_tasks);
537 
538 /**
539  * xprt_wait_for_buffer_space - wait for transport output buffer to clear
540  * @xprt: transport
541  *
542  * Note that we only set the timer for the case of RPC_IS_SOFT(), since
543  * we don't in general want to force a socket disconnection due to
544  * an incomplete RPC call transmission.
545  */
546 void xprt_wait_for_buffer_space(struct rpc_xprt *xprt)
547 {
548 	set_bit(XPRT_WRITE_SPACE, &xprt->state);
549 }
550 EXPORT_SYMBOL_GPL(xprt_wait_for_buffer_space);
551 
552 static bool
553 xprt_clear_write_space_locked(struct rpc_xprt *xprt)
554 {
555 	if (test_and_clear_bit(XPRT_WRITE_SPACE, &xprt->state)) {
556 		__xprt_lock_write_next(xprt);
557 		dprintk("RPC:       write space: waking waiting task on "
558 				"xprt %p\n", xprt);
559 		return true;
560 	}
561 	return false;
562 }
563 
564 /**
565  * xprt_write_space - wake the task waiting for transport output buffer space
566  * @xprt: transport with waiting tasks
567  *
568  * Can be called in a soft IRQ context, so xprt_write_space never sleeps.
569  */
570 bool xprt_write_space(struct rpc_xprt *xprt)
571 {
572 	bool ret;
573 
574 	if (!test_bit(XPRT_WRITE_SPACE, &xprt->state))
575 		return false;
576 	spin_lock(&xprt->transport_lock);
577 	ret = xprt_clear_write_space_locked(xprt);
578 	spin_unlock(&xprt->transport_lock);
579 	return ret;
580 }
581 EXPORT_SYMBOL_GPL(xprt_write_space);
582 
583 static unsigned long xprt_abs_ktime_to_jiffies(ktime_t abstime)
584 {
585 	s64 delta = ktime_to_ns(ktime_get() - abstime);
586 	return likely(delta >= 0) ?
587 		jiffies - nsecs_to_jiffies(delta) :
588 		jiffies + nsecs_to_jiffies(-delta);
589 }
590 
591 static unsigned long xprt_calc_majortimeo(struct rpc_rqst *req)
592 {
593 	const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
594 	unsigned long majortimeo = req->rq_timeout;
595 
596 	if (to->to_exponential)
597 		majortimeo <<= to->to_retries;
598 	else
599 		majortimeo += to->to_increment * to->to_retries;
600 	if (majortimeo > to->to_maxval || majortimeo == 0)
601 		majortimeo = to->to_maxval;
602 	return majortimeo;
603 }
604 
605 static void xprt_reset_majortimeo(struct rpc_rqst *req)
606 {
607 	req->rq_majortimeo += xprt_calc_majortimeo(req);
608 }
609 
610 static void xprt_reset_minortimeo(struct rpc_rqst *req)
611 {
612 	req->rq_minortimeo += req->rq_timeout;
613 }
614 
615 static void xprt_init_majortimeo(struct rpc_task *task, struct rpc_rqst *req)
616 {
617 	unsigned long time_init;
618 	struct rpc_xprt *xprt = req->rq_xprt;
619 
620 	if (likely(xprt && xprt_connected(xprt)))
621 		time_init = jiffies;
622 	else
623 		time_init = xprt_abs_ktime_to_jiffies(task->tk_start);
624 	req->rq_timeout = task->tk_client->cl_timeout->to_initval;
625 	req->rq_majortimeo = time_init + xprt_calc_majortimeo(req);
626 	req->rq_minortimeo = time_init + req->rq_timeout;
627 }
628 
629 /**
630  * xprt_adjust_timeout - adjust timeout values for next retransmit
631  * @req: RPC request containing parameters to use for the adjustment
632  *
633  */
634 int xprt_adjust_timeout(struct rpc_rqst *req)
635 {
636 	struct rpc_xprt *xprt = req->rq_xprt;
637 	const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
638 	int status = 0;
639 
640 	if (time_before(jiffies, req->rq_minortimeo))
641 		return status;
642 	if (time_before(jiffies, req->rq_majortimeo)) {
643 		if (to->to_exponential)
644 			req->rq_timeout <<= 1;
645 		else
646 			req->rq_timeout += to->to_increment;
647 		if (to->to_maxval && req->rq_timeout >= to->to_maxval)
648 			req->rq_timeout = to->to_maxval;
649 		req->rq_retries++;
650 	} else {
651 		req->rq_timeout = to->to_initval;
652 		req->rq_retries = 0;
653 		xprt_reset_majortimeo(req);
654 		/* Reset the RTT counters == "slow start" */
655 		spin_lock(&xprt->transport_lock);
656 		rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval);
657 		spin_unlock(&xprt->transport_lock);
658 		status = -ETIMEDOUT;
659 	}
660 	xprt_reset_minortimeo(req);
661 
662 	if (req->rq_timeout == 0) {
663 		printk(KERN_WARNING "xprt_adjust_timeout: rq_timeout = 0!\n");
664 		req->rq_timeout = 5 * HZ;
665 	}
666 	return status;
667 }
668 
669 static void xprt_autoclose(struct work_struct *work)
670 {
671 	struct rpc_xprt *xprt =
672 		container_of(work, struct rpc_xprt, task_cleanup);
673 	unsigned int pflags = memalloc_nofs_save();
674 
675 	trace_xprt_disconnect_auto(xprt);
676 	clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
677 	xprt->ops->close(xprt);
678 	xprt_release_write(xprt, NULL);
679 	wake_up_bit(&xprt->state, XPRT_LOCKED);
680 	memalloc_nofs_restore(pflags);
681 }
682 
683 /**
684  * xprt_disconnect_done - mark a transport as disconnected
685  * @xprt: transport to flag for disconnect
686  *
687  */
688 void xprt_disconnect_done(struct rpc_xprt *xprt)
689 {
690 	trace_xprt_disconnect_done(xprt);
691 	spin_lock(&xprt->transport_lock);
692 	xprt_clear_connected(xprt);
693 	xprt_clear_write_space_locked(xprt);
694 	xprt_clear_congestion_window_wait_locked(xprt);
695 	xprt_wake_pending_tasks(xprt, -ENOTCONN);
696 	spin_unlock(&xprt->transport_lock);
697 }
698 EXPORT_SYMBOL_GPL(xprt_disconnect_done);
699 
700 /**
701  * xprt_force_disconnect - force a transport to disconnect
702  * @xprt: transport to disconnect
703  *
704  */
705 void xprt_force_disconnect(struct rpc_xprt *xprt)
706 {
707 	trace_xprt_disconnect_force(xprt);
708 
709 	/* Don't race with the test_bit() in xprt_clear_locked() */
710 	spin_lock(&xprt->transport_lock);
711 	set_bit(XPRT_CLOSE_WAIT, &xprt->state);
712 	/* Try to schedule an autoclose RPC call */
713 	if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
714 		queue_work(xprtiod_workqueue, &xprt->task_cleanup);
715 	else if (xprt->snd_task)
716 		rpc_wake_up_queued_task_set_status(&xprt->pending,
717 				xprt->snd_task, -ENOTCONN);
718 	spin_unlock(&xprt->transport_lock);
719 }
720 EXPORT_SYMBOL_GPL(xprt_force_disconnect);
721 
722 static unsigned int
723 xprt_connect_cookie(struct rpc_xprt *xprt)
724 {
725 	return READ_ONCE(xprt->connect_cookie);
726 }
727 
728 static bool
729 xprt_request_retransmit_after_disconnect(struct rpc_task *task)
730 {
731 	struct rpc_rqst *req = task->tk_rqstp;
732 	struct rpc_xprt *xprt = req->rq_xprt;
733 
734 	return req->rq_connect_cookie != xprt_connect_cookie(xprt) ||
735 		!xprt_connected(xprt);
736 }
737 
738 /**
739  * xprt_conditional_disconnect - force a transport to disconnect
740  * @xprt: transport to disconnect
741  * @cookie: 'connection cookie'
742  *
743  * This attempts to break the connection if and only if 'cookie' matches
744  * the current transport 'connection cookie'. It ensures that we don't
745  * try to break the connection more than once when we need to retransmit
746  * a batch of RPC requests.
747  *
748  */
749 void xprt_conditional_disconnect(struct rpc_xprt *xprt, unsigned int cookie)
750 {
751 	/* Don't race with the test_bit() in xprt_clear_locked() */
752 	spin_lock(&xprt->transport_lock);
753 	if (cookie != xprt->connect_cookie)
754 		goto out;
755 	if (test_bit(XPRT_CLOSING, &xprt->state))
756 		goto out;
757 	set_bit(XPRT_CLOSE_WAIT, &xprt->state);
758 	/* Try to schedule an autoclose RPC call */
759 	if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
760 		queue_work(xprtiod_workqueue, &xprt->task_cleanup);
761 	xprt_wake_pending_tasks(xprt, -EAGAIN);
762 out:
763 	spin_unlock(&xprt->transport_lock);
764 }
765 
766 static bool
767 xprt_has_timer(const struct rpc_xprt *xprt)
768 {
769 	return xprt->idle_timeout != 0;
770 }
771 
772 static void
773 xprt_schedule_autodisconnect(struct rpc_xprt *xprt)
774 	__must_hold(&xprt->transport_lock)
775 {
776 	xprt->last_used = jiffies;
777 	if (RB_EMPTY_ROOT(&xprt->recv_queue) && xprt_has_timer(xprt))
778 		mod_timer(&xprt->timer, xprt->last_used + xprt->idle_timeout);
779 }
780 
781 static void
782 xprt_init_autodisconnect(struct timer_list *t)
783 {
784 	struct rpc_xprt *xprt = from_timer(xprt, t, timer);
785 
786 	if (!RB_EMPTY_ROOT(&xprt->recv_queue))
787 		return;
788 	/* Reset xprt->last_used to avoid connect/autodisconnect cycling */
789 	xprt->last_used = jiffies;
790 	if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
791 		return;
792 	queue_work(xprtiod_workqueue, &xprt->task_cleanup);
793 }
794 
795 bool xprt_lock_connect(struct rpc_xprt *xprt,
796 		struct rpc_task *task,
797 		void *cookie)
798 {
799 	bool ret = false;
800 
801 	spin_lock(&xprt->transport_lock);
802 	if (!test_bit(XPRT_LOCKED, &xprt->state))
803 		goto out;
804 	if (xprt->snd_task != task)
805 		goto out;
806 	xprt->snd_task = cookie;
807 	ret = true;
808 out:
809 	spin_unlock(&xprt->transport_lock);
810 	return ret;
811 }
812 
813 void xprt_unlock_connect(struct rpc_xprt *xprt, void *cookie)
814 {
815 	spin_lock(&xprt->transport_lock);
816 	if (xprt->snd_task != cookie)
817 		goto out;
818 	if (!test_bit(XPRT_LOCKED, &xprt->state))
819 		goto out;
820 	xprt->snd_task =NULL;
821 	xprt->ops->release_xprt(xprt, NULL);
822 	xprt_schedule_autodisconnect(xprt);
823 out:
824 	spin_unlock(&xprt->transport_lock);
825 	wake_up_bit(&xprt->state, XPRT_LOCKED);
826 }
827 
828 /**
829  * xprt_connect - schedule a transport connect operation
830  * @task: RPC task that is requesting the connect
831  *
832  */
833 void xprt_connect(struct rpc_task *task)
834 {
835 	struct rpc_xprt	*xprt = task->tk_rqstp->rq_xprt;
836 
837 	trace_xprt_connect(xprt);
838 
839 	if (!xprt_bound(xprt)) {
840 		task->tk_status = -EAGAIN;
841 		return;
842 	}
843 	if (!xprt_lock_write(xprt, task))
844 		return;
845 
846 	if (test_and_clear_bit(XPRT_CLOSE_WAIT, &xprt->state)) {
847 		trace_xprt_disconnect_cleanup(xprt);
848 		xprt->ops->close(xprt);
849 	}
850 
851 	if (!xprt_connected(xprt)) {
852 		task->tk_rqstp->rq_connect_cookie = xprt->connect_cookie;
853 		rpc_sleep_on_timeout(&xprt->pending, task, NULL,
854 				xprt_request_timeout(task->tk_rqstp));
855 
856 		if (test_bit(XPRT_CLOSING, &xprt->state))
857 			return;
858 		if (xprt_test_and_set_connecting(xprt))
859 			return;
860 		/* Race breaker */
861 		if (!xprt_connected(xprt)) {
862 			xprt->stat.connect_start = jiffies;
863 			xprt->ops->connect(xprt, task);
864 		} else {
865 			xprt_clear_connecting(xprt);
866 			task->tk_status = 0;
867 			rpc_wake_up_queued_task(&xprt->pending, task);
868 		}
869 	}
870 	xprt_release_write(xprt, task);
871 }
872 
873 /**
874  * xprt_reconnect_delay - compute the wait before scheduling a connect
875  * @xprt: transport instance
876  *
877  */
878 unsigned long xprt_reconnect_delay(const struct rpc_xprt *xprt)
879 {
880 	unsigned long start, now = jiffies;
881 
882 	start = xprt->stat.connect_start + xprt->reestablish_timeout;
883 	if (time_after(start, now))
884 		return start - now;
885 	return 0;
886 }
887 EXPORT_SYMBOL_GPL(xprt_reconnect_delay);
888 
889 /**
890  * xprt_reconnect_backoff - compute the new re-establish timeout
891  * @xprt: transport instance
892  * @init_to: initial reestablish timeout
893  *
894  */
895 void xprt_reconnect_backoff(struct rpc_xprt *xprt, unsigned long init_to)
896 {
897 	xprt->reestablish_timeout <<= 1;
898 	if (xprt->reestablish_timeout > xprt->max_reconnect_timeout)
899 		xprt->reestablish_timeout = xprt->max_reconnect_timeout;
900 	if (xprt->reestablish_timeout < init_to)
901 		xprt->reestablish_timeout = init_to;
902 }
903 EXPORT_SYMBOL_GPL(xprt_reconnect_backoff);
904 
905 enum xprt_xid_rb_cmp {
906 	XID_RB_EQUAL,
907 	XID_RB_LEFT,
908 	XID_RB_RIGHT,
909 };
910 static enum xprt_xid_rb_cmp
911 xprt_xid_cmp(__be32 xid1, __be32 xid2)
912 {
913 	if (xid1 == xid2)
914 		return XID_RB_EQUAL;
915 	if ((__force u32)xid1 < (__force u32)xid2)
916 		return XID_RB_LEFT;
917 	return XID_RB_RIGHT;
918 }
919 
920 static struct rpc_rqst *
921 xprt_request_rb_find(struct rpc_xprt *xprt, __be32 xid)
922 {
923 	struct rb_node *n = xprt->recv_queue.rb_node;
924 	struct rpc_rqst *req;
925 
926 	while (n != NULL) {
927 		req = rb_entry(n, struct rpc_rqst, rq_recv);
928 		switch (xprt_xid_cmp(xid, req->rq_xid)) {
929 		case XID_RB_LEFT:
930 			n = n->rb_left;
931 			break;
932 		case XID_RB_RIGHT:
933 			n = n->rb_right;
934 			break;
935 		case XID_RB_EQUAL:
936 			return req;
937 		}
938 	}
939 	return NULL;
940 }
941 
942 static void
943 xprt_request_rb_insert(struct rpc_xprt *xprt, struct rpc_rqst *new)
944 {
945 	struct rb_node **p = &xprt->recv_queue.rb_node;
946 	struct rb_node *n = NULL;
947 	struct rpc_rqst *req;
948 
949 	while (*p != NULL) {
950 		n = *p;
951 		req = rb_entry(n, struct rpc_rqst, rq_recv);
952 		switch(xprt_xid_cmp(new->rq_xid, req->rq_xid)) {
953 		case XID_RB_LEFT:
954 			p = &n->rb_left;
955 			break;
956 		case XID_RB_RIGHT:
957 			p = &n->rb_right;
958 			break;
959 		case XID_RB_EQUAL:
960 			WARN_ON_ONCE(new != req);
961 			return;
962 		}
963 	}
964 	rb_link_node(&new->rq_recv, n, p);
965 	rb_insert_color(&new->rq_recv, &xprt->recv_queue);
966 }
967 
968 static void
969 xprt_request_rb_remove(struct rpc_xprt *xprt, struct rpc_rqst *req)
970 {
971 	rb_erase(&req->rq_recv, &xprt->recv_queue);
972 }
973 
974 /**
975  * xprt_lookup_rqst - find an RPC request corresponding to an XID
976  * @xprt: transport on which the original request was transmitted
977  * @xid: RPC XID of incoming reply
978  *
979  * Caller holds xprt->queue_lock.
980  */
981 struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid)
982 {
983 	struct rpc_rqst *entry;
984 
985 	entry = xprt_request_rb_find(xprt, xid);
986 	if (entry != NULL) {
987 		trace_xprt_lookup_rqst(xprt, xid, 0);
988 		entry->rq_rtt = ktime_sub(ktime_get(), entry->rq_xtime);
989 		return entry;
990 	}
991 
992 	dprintk("RPC:       xprt_lookup_rqst did not find xid %08x\n",
993 			ntohl(xid));
994 	trace_xprt_lookup_rqst(xprt, xid, -ENOENT);
995 	xprt->stat.bad_xids++;
996 	return NULL;
997 }
998 EXPORT_SYMBOL_GPL(xprt_lookup_rqst);
999 
1000 static bool
1001 xprt_is_pinned_rqst(struct rpc_rqst *req)
1002 {
1003 	return atomic_read(&req->rq_pin) != 0;
1004 }
1005 
1006 /**
1007  * xprt_pin_rqst - Pin a request on the transport receive list
1008  * @req: Request to pin
1009  *
1010  * Caller must ensure this is atomic with the call to xprt_lookup_rqst()
1011  * so should be holding xprt->queue_lock.
1012  */
1013 void xprt_pin_rqst(struct rpc_rqst *req)
1014 {
1015 	atomic_inc(&req->rq_pin);
1016 }
1017 EXPORT_SYMBOL_GPL(xprt_pin_rqst);
1018 
1019 /**
1020  * xprt_unpin_rqst - Unpin a request on the transport receive list
1021  * @req: Request to pin
1022  *
1023  * Caller should be holding xprt->queue_lock.
1024  */
1025 void xprt_unpin_rqst(struct rpc_rqst *req)
1026 {
1027 	if (!test_bit(RPC_TASK_MSG_PIN_WAIT, &req->rq_task->tk_runstate)) {
1028 		atomic_dec(&req->rq_pin);
1029 		return;
1030 	}
1031 	if (atomic_dec_and_test(&req->rq_pin))
1032 		wake_up_var(&req->rq_pin);
1033 }
1034 EXPORT_SYMBOL_GPL(xprt_unpin_rqst);
1035 
1036 static void xprt_wait_on_pinned_rqst(struct rpc_rqst *req)
1037 {
1038 	wait_var_event(&req->rq_pin, !xprt_is_pinned_rqst(req));
1039 }
1040 
1041 static bool
1042 xprt_request_data_received(struct rpc_task *task)
1043 {
1044 	return !test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) &&
1045 		READ_ONCE(task->tk_rqstp->rq_reply_bytes_recvd) != 0;
1046 }
1047 
1048 static bool
1049 xprt_request_need_enqueue_receive(struct rpc_task *task, struct rpc_rqst *req)
1050 {
1051 	return !test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) &&
1052 		READ_ONCE(task->tk_rqstp->rq_reply_bytes_recvd) == 0;
1053 }
1054 
1055 /**
1056  * xprt_request_enqueue_receive - Add an request to the receive queue
1057  * @task: RPC task
1058  *
1059  */
1060 void
1061 xprt_request_enqueue_receive(struct rpc_task *task)
1062 {
1063 	struct rpc_rqst *req = task->tk_rqstp;
1064 	struct rpc_xprt *xprt = req->rq_xprt;
1065 
1066 	if (!xprt_request_need_enqueue_receive(task, req))
1067 		return;
1068 
1069 	xprt_request_prepare(task->tk_rqstp);
1070 	spin_lock(&xprt->queue_lock);
1071 
1072 	/* Update the softirq receive buffer */
1073 	memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
1074 			sizeof(req->rq_private_buf));
1075 
1076 	/* Add request to the receive list */
1077 	xprt_request_rb_insert(xprt, req);
1078 	set_bit(RPC_TASK_NEED_RECV, &task->tk_runstate);
1079 	spin_unlock(&xprt->queue_lock);
1080 
1081 	/* Turn off autodisconnect */
1082 	del_singleshot_timer_sync(&xprt->timer);
1083 }
1084 
1085 /**
1086  * xprt_request_dequeue_receive_locked - Remove a request from the receive queue
1087  * @task: RPC task
1088  *
1089  * Caller must hold xprt->queue_lock.
1090  */
1091 static void
1092 xprt_request_dequeue_receive_locked(struct rpc_task *task)
1093 {
1094 	struct rpc_rqst *req = task->tk_rqstp;
1095 
1096 	if (test_and_clear_bit(RPC_TASK_NEED_RECV, &task->tk_runstate))
1097 		xprt_request_rb_remove(req->rq_xprt, req);
1098 }
1099 
1100 /**
1101  * xprt_update_rtt - Update RPC RTT statistics
1102  * @task: RPC request that recently completed
1103  *
1104  * Caller holds xprt->queue_lock.
1105  */
1106 void xprt_update_rtt(struct rpc_task *task)
1107 {
1108 	struct rpc_rqst *req = task->tk_rqstp;
1109 	struct rpc_rtt *rtt = task->tk_client->cl_rtt;
1110 	unsigned int timer = task->tk_msg.rpc_proc->p_timer;
1111 	long m = usecs_to_jiffies(ktime_to_us(req->rq_rtt));
1112 
1113 	if (timer) {
1114 		if (req->rq_ntrans == 1)
1115 			rpc_update_rtt(rtt, timer, m);
1116 		rpc_set_timeo(rtt, timer, req->rq_ntrans - 1);
1117 	}
1118 }
1119 EXPORT_SYMBOL_GPL(xprt_update_rtt);
1120 
1121 /**
1122  * xprt_complete_rqst - called when reply processing is complete
1123  * @task: RPC request that recently completed
1124  * @copied: actual number of bytes received from the transport
1125  *
1126  * Caller holds xprt->queue_lock.
1127  */
1128 void xprt_complete_rqst(struct rpc_task *task, int copied)
1129 {
1130 	struct rpc_rqst *req = task->tk_rqstp;
1131 	struct rpc_xprt *xprt = req->rq_xprt;
1132 
1133 	xprt->stat.recvs++;
1134 
1135 	req->rq_private_buf.len = copied;
1136 	/* Ensure all writes are done before we update */
1137 	/* req->rq_reply_bytes_recvd */
1138 	smp_wmb();
1139 	req->rq_reply_bytes_recvd = copied;
1140 	xprt_request_dequeue_receive_locked(task);
1141 	rpc_wake_up_queued_task(&xprt->pending, task);
1142 }
1143 EXPORT_SYMBOL_GPL(xprt_complete_rqst);
1144 
1145 static void xprt_timer(struct rpc_task *task)
1146 {
1147 	struct rpc_rqst *req = task->tk_rqstp;
1148 	struct rpc_xprt *xprt = req->rq_xprt;
1149 
1150 	if (task->tk_status != -ETIMEDOUT)
1151 		return;
1152 
1153 	trace_xprt_timer(xprt, req->rq_xid, task->tk_status);
1154 	if (!req->rq_reply_bytes_recvd) {
1155 		if (xprt->ops->timer)
1156 			xprt->ops->timer(xprt, task);
1157 	} else
1158 		task->tk_status = 0;
1159 }
1160 
1161 /**
1162  * xprt_wait_for_reply_request_def - wait for reply
1163  * @task: pointer to rpc_task
1164  *
1165  * Set a request's retransmit timeout based on the transport's
1166  * default timeout parameters.  Used by transports that don't adjust
1167  * the retransmit timeout based on round-trip time estimation,
1168  * and put the task to sleep on the pending queue.
1169  */
1170 void xprt_wait_for_reply_request_def(struct rpc_task *task)
1171 {
1172 	struct rpc_rqst *req = task->tk_rqstp;
1173 
1174 	rpc_sleep_on_timeout(&req->rq_xprt->pending, task, xprt_timer,
1175 			xprt_request_timeout(req));
1176 }
1177 EXPORT_SYMBOL_GPL(xprt_wait_for_reply_request_def);
1178 
1179 /**
1180  * xprt_wait_for_reply_request_rtt - wait for reply using RTT estimator
1181  * @task: pointer to rpc_task
1182  *
1183  * Set a request's retransmit timeout using the RTT estimator,
1184  * and put the task to sleep on the pending queue.
1185  */
1186 void xprt_wait_for_reply_request_rtt(struct rpc_task *task)
1187 {
1188 	int timer = task->tk_msg.rpc_proc->p_timer;
1189 	struct rpc_clnt *clnt = task->tk_client;
1190 	struct rpc_rtt *rtt = clnt->cl_rtt;
1191 	struct rpc_rqst *req = task->tk_rqstp;
1192 	unsigned long max_timeout = clnt->cl_timeout->to_maxval;
1193 	unsigned long timeout;
1194 
1195 	timeout = rpc_calc_rto(rtt, timer);
1196 	timeout <<= rpc_ntimeo(rtt, timer) + req->rq_retries;
1197 	if (timeout > max_timeout || timeout == 0)
1198 		timeout = max_timeout;
1199 	rpc_sleep_on_timeout(&req->rq_xprt->pending, task, xprt_timer,
1200 			jiffies + timeout);
1201 }
1202 EXPORT_SYMBOL_GPL(xprt_wait_for_reply_request_rtt);
1203 
1204 /**
1205  * xprt_request_wait_receive - wait for the reply to an RPC request
1206  * @task: RPC task about to send a request
1207  *
1208  */
1209 void xprt_request_wait_receive(struct rpc_task *task)
1210 {
1211 	struct rpc_rqst *req = task->tk_rqstp;
1212 	struct rpc_xprt *xprt = req->rq_xprt;
1213 
1214 	if (!test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate))
1215 		return;
1216 	/*
1217 	 * Sleep on the pending queue if we're expecting a reply.
1218 	 * The spinlock ensures atomicity between the test of
1219 	 * req->rq_reply_bytes_recvd, and the call to rpc_sleep_on().
1220 	 */
1221 	spin_lock(&xprt->queue_lock);
1222 	if (test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate)) {
1223 		xprt->ops->wait_for_reply_request(task);
1224 		/*
1225 		 * Send an extra queue wakeup call if the
1226 		 * connection was dropped in case the call to
1227 		 * rpc_sleep_on() raced.
1228 		 */
1229 		if (xprt_request_retransmit_after_disconnect(task))
1230 			rpc_wake_up_queued_task_set_status(&xprt->pending,
1231 					task, -ENOTCONN);
1232 	}
1233 	spin_unlock(&xprt->queue_lock);
1234 }
1235 
1236 static bool
1237 xprt_request_need_enqueue_transmit(struct rpc_task *task, struct rpc_rqst *req)
1238 {
1239 	return !test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate);
1240 }
1241 
1242 /**
1243  * xprt_request_enqueue_transmit - queue a task for transmission
1244  * @task: pointer to rpc_task
1245  *
1246  * Add a task to the transmission queue.
1247  */
1248 void
1249 xprt_request_enqueue_transmit(struct rpc_task *task)
1250 {
1251 	struct rpc_rqst *pos, *req = task->tk_rqstp;
1252 	struct rpc_xprt *xprt = req->rq_xprt;
1253 
1254 	if (xprt_request_need_enqueue_transmit(task, req)) {
1255 		req->rq_bytes_sent = 0;
1256 		spin_lock(&xprt->queue_lock);
1257 		/*
1258 		 * Requests that carry congestion control credits are added
1259 		 * to the head of the list to avoid starvation issues.
1260 		 */
1261 		if (req->rq_cong) {
1262 			xprt_clear_congestion_window_wait(xprt);
1263 			list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1264 				if (pos->rq_cong)
1265 					continue;
1266 				/* Note: req is added _before_ pos */
1267 				list_add_tail(&req->rq_xmit, &pos->rq_xmit);
1268 				INIT_LIST_HEAD(&req->rq_xmit2);
1269 				goto out;
1270 			}
1271 		} else if (RPC_IS_SWAPPER(task)) {
1272 			list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1273 				if (pos->rq_cong || pos->rq_bytes_sent)
1274 					continue;
1275 				if (RPC_IS_SWAPPER(pos->rq_task))
1276 					continue;
1277 				/* Note: req is added _before_ pos */
1278 				list_add_tail(&req->rq_xmit, &pos->rq_xmit);
1279 				INIT_LIST_HEAD(&req->rq_xmit2);
1280 				goto out;
1281 			}
1282 		} else if (!req->rq_seqno) {
1283 			list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1284 				if (pos->rq_task->tk_owner != task->tk_owner)
1285 					continue;
1286 				list_add_tail(&req->rq_xmit2, &pos->rq_xmit2);
1287 				INIT_LIST_HEAD(&req->rq_xmit);
1288 				goto out;
1289 			}
1290 		}
1291 		list_add_tail(&req->rq_xmit, &xprt->xmit_queue);
1292 		INIT_LIST_HEAD(&req->rq_xmit2);
1293 out:
1294 		set_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate);
1295 		spin_unlock(&xprt->queue_lock);
1296 	}
1297 }
1298 
1299 /**
1300  * xprt_request_dequeue_transmit_locked - remove a task from the transmission queue
1301  * @task: pointer to rpc_task
1302  *
1303  * Remove a task from the transmission queue
1304  * Caller must hold xprt->queue_lock
1305  */
1306 static void
1307 xprt_request_dequeue_transmit_locked(struct rpc_task *task)
1308 {
1309 	struct rpc_rqst *req = task->tk_rqstp;
1310 
1311 	if (!test_and_clear_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1312 		return;
1313 	if (!list_empty(&req->rq_xmit)) {
1314 		list_del(&req->rq_xmit);
1315 		if (!list_empty(&req->rq_xmit2)) {
1316 			struct rpc_rqst *next = list_first_entry(&req->rq_xmit2,
1317 					struct rpc_rqst, rq_xmit2);
1318 			list_del(&req->rq_xmit2);
1319 			list_add_tail(&next->rq_xmit, &next->rq_xprt->xmit_queue);
1320 		}
1321 	} else
1322 		list_del(&req->rq_xmit2);
1323 }
1324 
1325 /**
1326  * xprt_request_dequeue_transmit - remove a task from the transmission queue
1327  * @task: pointer to rpc_task
1328  *
1329  * Remove a task from the transmission queue
1330  */
1331 static void
1332 xprt_request_dequeue_transmit(struct rpc_task *task)
1333 {
1334 	struct rpc_rqst *req = task->tk_rqstp;
1335 	struct rpc_xprt *xprt = req->rq_xprt;
1336 
1337 	spin_lock(&xprt->queue_lock);
1338 	xprt_request_dequeue_transmit_locked(task);
1339 	spin_unlock(&xprt->queue_lock);
1340 }
1341 
1342 /**
1343  * xprt_request_dequeue_xprt - remove a task from the transmit+receive queue
1344  * @task: pointer to rpc_task
1345  *
1346  * Remove a task from the transmit and receive queues, and ensure that
1347  * it is not pinned by the receive work item.
1348  */
1349 void
1350 xprt_request_dequeue_xprt(struct rpc_task *task)
1351 {
1352 	struct rpc_rqst	*req = task->tk_rqstp;
1353 	struct rpc_xprt *xprt = req->rq_xprt;
1354 
1355 	if (test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate) ||
1356 	    test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) ||
1357 	    xprt_is_pinned_rqst(req)) {
1358 		spin_lock(&xprt->queue_lock);
1359 		xprt_request_dequeue_transmit_locked(task);
1360 		xprt_request_dequeue_receive_locked(task);
1361 		while (xprt_is_pinned_rqst(req)) {
1362 			set_bit(RPC_TASK_MSG_PIN_WAIT, &task->tk_runstate);
1363 			spin_unlock(&xprt->queue_lock);
1364 			xprt_wait_on_pinned_rqst(req);
1365 			spin_lock(&xprt->queue_lock);
1366 			clear_bit(RPC_TASK_MSG_PIN_WAIT, &task->tk_runstate);
1367 		}
1368 		spin_unlock(&xprt->queue_lock);
1369 	}
1370 }
1371 
1372 /**
1373  * xprt_request_prepare - prepare an encoded request for transport
1374  * @req: pointer to rpc_rqst
1375  *
1376  * Calls into the transport layer to do whatever is needed to prepare
1377  * the request for transmission or receive.
1378  */
1379 void
1380 xprt_request_prepare(struct rpc_rqst *req)
1381 {
1382 	struct rpc_xprt *xprt = req->rq_xprt;
1383 
1384 	if (xprt->ops->prepare_request)
1385 		xprt->ops->prepare_request(req);
1386 }
1387 
1388 /**
1389  * xprt_request_need_retransmit - Test if a task needs retransmission
1390  * @task: pointer to rpc_task
1391  *
1392  * Test for whether a connection breakage requires the task to retransmit
1393  */
1394 bool
1395 xprt_request_need_retransmit(struct rpc_task *task)
1396 {
1397 	return xprt_request_retransmit_after_disconnect(task);
1398 }
1399 
1400 /**
1401  * xprt_prepare_transmit - reserve the transport before sending a request
1402  * @task: RPC task about to send a request
1403  *
1404  */
1405 bool xprt_prepare_transmit(struct rpc_task *task)
1406 {
1407 	struct rpc_rqst	*req = task->tk_rqstp;
1408 	struct rpc_xprt	*xprt = req->rq_xprt;
1409 
1410 	if (!xprt_lock_write(xprt, task)) {
1411 		trace_xprt_transmit_queued(xprt, task);
1412 
1413 		/* Race breaker: someone may have transmitted us */
1414 		if (!test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1415 			rpc_wake_up_queued_task_set_status(&xprt->sending,
1416 					task, 0);
1417 		return false;
1418 
1419 	}
1420 	return true;
1421 }
1422 
1423 void xprt_end_transmit(struct rpc_task *task)
1424 {
1425 	xprt_release_write(task->tk_rqstp->rq_xprt, task);
1426 }
1427 
1428 /**
1429  * xprt_request_transmit - send an RPC request on a transport
1430  * @req: pointer to request to transmit
1431  * @snd_task: RPC task that owns the transport lock
1432  *
1433  * This performs the transmission of a single request.
1434  * Note that if the request is not the same as snd_task, then it
1435  * does need to be pinned.
1436  * Returns '0' on success.
1437  */
1438 static int
1439 xprt_request_transmit(struct rpc_rqst *req, struct rpc_task *snd_task)
1440 {
1441 	struct rpc_xprt *xprt = req->rq_xprt;
1442 	struct rpc_task *task = req->rq_task;
1443 	unsigned int connect_cookie;
1444 	int is_retrans = RPC_WAS_SENT(task);
1445 	int status;
1446 
1447 	if (!req->rq_bytes_sent) {
1448 		if (xprt_request_data_received(task)) {
1449 			status = 0;
1450 			goto out_dequeue;
1451 		}
1452 		/* Verify that our message lies in the RPCSEC_GSS window */
1453 		if (rpcauth_xmit_need_reencode(task)) {
1454 			status = -EBADMSG;
1455 			goto out_dequeue;
1456 		}
1457 		if (RPC_SIGNALLED(task)) {
1458 			status = -ERESTARTSYS;
1459 			goto out_dequeue;
1460 		}
1461 	}
1462 
1463 	/*
1464 	 * Update req->rq_ntrans before transmitting to avoid races with
1465 	 * xprt_update_rtt(), which needs to know that it is recording a
1466 	 * reply to the first transmission.
1467 	 */
1468 	req->rq_ntrans++;
1469 
1470 	trace_rpc_xdr_sendto(task, &req->rq_snd_buf);
1471 	connect_cookie = xprt->connect_cookie;
1472 	status = xprt->ops->send_request(req);
1473 	if (status != 0) {
1474 		req->rq_ntrans--;
1475 		trace_xprt_transmit(req, status);
1476 		return status;
1477 	}
1478 
1479 	if (is_retrans)
1480 		task->tk_client->cl_stats->rpcretrans++;
1481 
1482 	xprt_inject_disconnect(xprt);
1483 
1484 	task->tk_flags |= RPC_TASK_SENT;
1485 	spin_lock(&xprt->transport_lock);
1486 
1487 	xprt->stat.sends++;
1488 	xprt->stat.req_u += xprt->stat.sends - xprt->stat.recvs;
1489 	xprt->stat.bklog_u += xprt->backlog.qlen;
1490 	xprt->stat.sending_u += xprt->sending.qlen;
1491 	xprt->stat.pending_u += xprt->pending.qlen;
1492 	spin_unlock(&xprt->transport_lock);
1493 
1494 	req->rq_connect_cookie = connect_cookie;
1495 out_dequeue:
1496 	trace_xprt_transmit(req, status);
1497 	xprt_request_dequeue_transmit(task);
1498 	rpc_wake_up_queued_task_set_status(&xprt->sending, task, status);
1499 	return status;
1500 }
1501 
1502 /**
1503  * xprt_transmit - send an RPC request on a transport
1504  * @task: controlling RPC task
1505  *
1506  * Attempts to drain the transmit queue. On exit, either the transport
1507  * signalled an error that needs to be handled before transmission can
1508  * resume, or @task finished transmitting, and detected that it already
1509  * received a reply.
1510  */
1511 void
1512 xprt_transmit(struct rpc_task *task)
1513 {
1514 	struct rpc_rqst *next, *req = task->tk_rqstp;
1515 	struct rpc_xprt	*xprt = req->rq_xprt;
1516 	int counter, status;
1517 
1518 	spin_lock(&xprt->queue_lock);
1519 	counter = 0;
1520 	while (!list_empty(&xprt->xmit_queue)) {
1521 		if (++counter == 20)
1522 			break;
1523 		next = list_first_entry(&xprt->xmit_queue,
1524 				struct rpc_rqst, rq_xmit);
1525 		xprt_pin_rqst(next);
1526 		spin_unlock(&xprt->queue_lock);
1527 		status = xprt_request_transmit(next, task);
1528 		if (status == -EBADMSG && next != req)
1529 			status = 0;
1530 		spin_lock(&xprt->queue_lock);
1531 		xprt_unpin_rqst(next);
1532 		if (status == 0) {
1533 			if (!xprt_request_data_received(task) ||
1534 			    test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1535 				continue;
1536 		} else if (test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1537 			task->tk_status = status;
1538 		break;
1539 	}
1540 	spin_unlock(&xprt->queue_lock);
1541 }
1542 
1543 static void xprt_add_backlog(struct rpc_xprt *xprt, struct rpc_task *task)
1544 {
1545 	set_bit(XPRT_CONGESTED, &xprt->state);
1546 	rpc_sleep_on(&xprt->backlog, task, NULL);
1547 }
1548 
1549 static void xprt_wake_up_backlog(struct rpc_xprt *xprt)
1550 {
1551 	if (rpc_wake_up_next(&xprt->backlog) == NULL)
1552 		clear_bit(XPRT_CONGESTED, &xprt->state);
1553 }
1554 
1555 static bool xprt_throttle_congested(struct rpc_xprt *xprt, struct rpc_task *task)
1556 {
1557 	bool ret = false;
1558 
1559 	if (!test_bit(XPRT_CONGESTED, &xprt->state))
1560 		goto out;
1561 	spin_lock(&xprt->reserve_lock);
1562 	if (test_bit(XPRT_CONGESTED, &xprt->state)) {
1563 		rpc_sleep_on(&xprt->backlog, task, NULL);
1564 		ret = true;
1565 	}
1566 	spin_unlock(&xprt->reserve_lock);
1567 out:
1568 	return ret;
1569 }
1570 
1571 static struct rpc_rqst *xprt_dynamic_alloc_slot(struct rpc_xprt *xprt)
1572 {
1573 	struct rpc_rqst *req = ERR_PTR(-EAGAIN);
1574 
1575 	if (xprt->num_reqs >= xprt->max_reqs)
1576 		goto out;
1577 	++xprt->num_reqs;
1578 	spin_unlock(&xprt->reserve_lock);
1579 	req = kzalloc(sizeof(struct rpc_rqst), GFP_NOFS);
1580 	spin_lock(&xprt->reserve_lock);
1581 	if (req != NULL)
1582 		goto out;
1583 	--xprt->num_reqs;
1584 	req = ERR_PTR(-ENOMEM);
1585 out:
1586 	return req;
1587 }
1588 
1589 static bool xprt_dynamic_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
1590 {
1591 	if (xprt->num_reqs > xprt->min_reqs) {
1592 		--xprt->num_reqs;
1593 		kfree(req);
1594 		return true;
1595 	}
1596 	return false;
1597 }
1598 
1599 void xprt_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
1600 {
1601 	struct rpc_rqst *req;
1602 
1603 	spin_lock(&xprt->reserve_lock);
1604 	if (!list_empty(&xprt->free)) {
1605 		req = list_entry(xprt->free.next, struct rpc_rqst, rq_list);
1606 		list_del(&req->rq_list);
1607 		goto out_init_req;
1608 	}
1609 	req = xprt_dynamic_alloc_slot(xprt);
1610 	if (!IS_ERR(req))
1611 		goto out_init_req;
1612 	switch (PTR_ERR(req)) {
1613 	case -ENOMEM:
1614 		dprintk("RPC:       dynamic allocation of request slot "
1615 				"failed! Retrying\n");
1616 		task->tk_status = -ENOMEM;
1617 		break;
1618 	case -EAGAIN:
1619 		xprt_add_backlog(xprt, task);
1620 		dprintk("RPC:       waiting for request slot\n");
1621 		fallthrough;
1622 	default:
1623 		task->tk_status = -EAGAIN;
1624 	}
1625 	spin_unlock(&xprt->reserve_lock);
1626 	return;
1627 out_init_req:
1628 	xprt->stat.max_slots = max_t(unsigned int, xprt->stat.max_slots,
1629 				     xprt->num_reqs);
1630 	spin_unlock(&xprt->reserve_lock);
1631 
1632 	task->tk_status = 0;
1633 	task->tk_rqstp = req;
1634 }
1635 EXPORT_SYMBOL_GPL(xprt_alloc_slot);
1636 
1637 void xprt_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
1638 {
1639 	spin_lock(&xprt->reserve_lock);
1640 	if (!xprt_dynamic_free_slot(xprt, req)) {
1641 		memset(req, 0, sizeof(*req));	/* mark unused */
1642 		list_add(&req->rq_list, &xprt->free);
1643 	}
1644 	xprt_wake_up_backlog(xprt);
1645 	spin_unlock(&xprt->reserve_lock);
1646 }
1647 EXPORT_SYMBOL_GPL(xprt_free_slot);
1648 
1649 static void xprt_free_all_slots(struct rpc_xprt *xprt)
1650 {
1651 	struct rpc_rqst *req;
1652 	while (!list_empty(&xprt->free)) {
1653 		req = list_first_entry(&xprt->free, struct rpc_rqst, rq_list);
1654 		list_del(&req->rq_list);
1655 		kfree(req);
1656 	}
1657 }
1658 
1659 struct rpc_xprt *xprt_alloc(struct net *net, size_t size,
1660 		unsigned int num_prealloc,
1661 		unsigned int max_alloc)
1662 {
1663 	struct rpc_xprt *xprt;
1664 	struct rpc_rqst *req;
1665 	int i;
1666 
1667 	xprt = kzalloc(size, GFP_KERNEL);
1668 	if (xprt == NULL)
1669 		goto out;
1670 
1671 	xprt_init(xprt, net);
1672 
1673 	for (i = 0; i < num_prealloc; i++) {
1674 		req = kzalloc(sizeof(struct rpc_rqst), GFP_KERNEL);
1675 		if (!req)
1676 			goto out_free;
1677 		list_add(&req->rq_list, &xprt->free);
1678 	}
1679 	if (max_alloc > num_prealloc)
1680 		xprt->max_reqs = max_alloc;
1681 	else
1682 		xprt->max_reqs = num_prealloc;
1683 	xprt->min_reqs = num_prealloc;
1684 	xprt->num_reqs = num_prealloc;
1685 
1686 	return xprt;
1687 
1688 out_free:
1689 	xprt_free(xprt);
1690 out:
1691 	return NULL;
1692 }
1693 EXPORT_SYMBOL_GPL(xprt_alloc);
1694 
1695 void xprt_free(struct rpc_xprt *xprt)
1696 {
1697 	put_net(xprt->xprt_net);
1698 	xprt_free_all_slots(xprt);
1699 	kfree_rcu(xprt, rcu);
1700 }
1701 EXPORT_SYMBOL_GPL(xprt_free);
1702 
1703 static void
1704 xprt_init_connect_cookie(struct rpc_rqst *req, struct rpc_xprt *xprt)
1705 {
1706 	req->rq_connect_cookie = xprt_connect_cookie(xprt) - 1;
1707 }
1708 
1709 static __be32
1710 xprt_alloc_xid(struct rpc_xprt *xprt)
1711 {
1712 	__be32 xid;
1713 
1714 	spin_lock(&xprt->reserve_lock);
1715 	xid = (__force __be32)xprt->xid++;
1716 	spin_unlock(&xprt->reserve_lock);
1717 	return xid;
1718 }
1719 
1720 static void
1721 xprt_init_xid(struct rpc_xprt *xprt)
1722 {
1723 	xprt->xid = prandom_u32();
1724 }
1725 
1726 static void
1727 xprt_request_init(struct rpc_task *task)
1728 {
1729 	struct rpc_xprt *xprt = task->tk_xprt;
1730 	struct rpc_rqst	*req = task->tk_rqstp;
1731 
1732 	req->rq_task	= task;
1733 	req->rq_xprt    = xprt;
1734 	req->rq_buffer  = NULL;
1735 	req->rq_xid	= xprt_alloc_xid(xprt);
1736 	xprt_init_connect_cookie(req, xprt);
1737 	req->rq_snd_buf.len = 0;
1738 	req->rq_snd_buf.buflen = 0;
1739 	req->rq_rcv_buf.len = 0;
1740 	req->rq_rcv_buf.buflen = 0;
1741 	req->rq_snd_buf.bvec = NULL;
1742 	req->rq_rcv_buf.bvec = NULL;
1743 	req->rq_release_snd_buf = NULL;
1744 	xprt_init_majortimeo(task, req);
1745 
1746 	trace_xprt_reserve(req);
1747 }
1748 
1749 static void
1750 xprt_do_reserve(struct rpc_xprt *xprt, struct rpc_task *task)
1751 {
1752 	xprt->ops->alloc_slot(xprt, task);
1753 	if (task->tk_rqstp != NULL)
1754 		xprt_request_init(task);
1755 }
1756 
1757 /**
1758  * xprt_reserve - allocate an RPC request slot
1759  * @task: RPC task requesting a slot allocation
1760  *
1761  * If the transport is marked as being congested, or if no more
1762  * slots are available, place the task on the transport's
1763  * backlog queue.
1764  */
1765 void xprt_reserve(struct rpc_task *task)
1766 {
1767 	struct rpc_xprt *xprt = task->tk_xprt;
1768 
1769 	task->tk_status = 0;
1770 	if (task->tk_rqstp != NULL)
1771 		return;
1772 
1773 	task->tk_status = -EAGAIN;
1774 	if (!xprt_throttle_congested(xprt, task))
1775 		xprt_do_reserve(xprt, task);
1776 }
1777 
1778 /**
1779  * xprt_retry_reserve - allocate an RPC request slot
1780  * @task: RPC task requesting a slot allocation
1781  *
1782  * If no more slots are available, place the task on the transport's
1783  * backlog queue.
1784  * Note that the only difference with xprt_reserve is that we now
1785  * ignore the value of the XPRT_CONGESTED flag.
1786  */
1787 void xprt_retry_reserve(struct rpc_task *task)
1788 {
1789 	struct rpc_xprt *xprt = task->tk_xprt;
1790 
1791 	task->tk_status = 0;
1792 	if (task->tk_rqstp != NULL)
1793 		return;
1794 
1795 	task->tk_status = -EAGAIN;
1796 	xprt_do_reserve(xprt, task);
1797 }
1798 
1799 /**
1800  * xprt_release - release an RPC request slot
1801  * @task: task which is finished with the slot
1802  *
1803  */
1804 void xprt_release(struct rpc_task *task)
1805 {
1806 	struct rpc_xprt	*xprt;
1807 	struct rpc_rqst	*req = task->tk_rqstp;
1808 
1809 	if (req == NULL) {
1810 		if (task->tk_client) {
1811 			xprt = task->tk_xprt;
1812 			xprt_release_write(xprt, task);
1813 		}
1814 		return;
1815 	}
1816 
1817 	xprt = req->rq_xprt;
1818 	xprt_request_dequeue_xprt(task);
1819 	spin_lock(&xprt->transport_lock);
1820 	xprt->ops->release_xprt(xprt, task);
1821 	if (xprt->ops->release_request)
1822 		xprt->ops->release_request(task);
1823 	xprt_schedule_autodisconnect(xprt);
1824 	spin_unlock(&xprt->transport_lock);
1825 	if (req->rq_buffer)
1826 		xprt->ops->buf_free(task);
1827 	xprt_inject_disconnect(xprt);
1828 	xdr_free_bvec(&req->rq_rcv_buf);
1829 	xdr_free_bvec(&req->rq_snd_buf);
1830 	if (req->rq_cred != NULL)
1831 		put_rpccred(req->rq_cred);
1832 	task->tk_rqstp = NULL;
1833 	if (req->rq_release_snd_buf)
1834 		req->rq_release_snd_buf(req);
1835 
1836 	if (likely(!bc_prealloc(req)))
1837 		xprt->ops->free_slot(xprt, req);
1838 	else
1839 		xprt_free_bc_request(req);
1840 }
1841 
1842 #ifdef CONFIG_SUNRPC_BACKCHANNEL
1843 void
1844 xprt_init_bc_request(struct rpc_rqst *req, struct rpc_task *task)
1845 {
1846 	struct xdr_buf *xbufp = &req->rq_snd_buf;
1847 
1848 	task->tk_rqstp = req;
1849 	req->rq_task = task;
1850 	xprt_init_connect_cookie(req, req->rq_xprt);
1851 	/*
1852 	 * Set up the xdr_buf length.
1853 	 * This also indicates that the buffer is XDR encoded already.
1854 	 */
1855 	xbufp->len = xbufp->head[0].iov_len + xbufp->page_len +
1856 		xbufp->tail[0].iov_len;
1857 }
1858 #endif
1859 
1860 static void xprt_init(struct rpc_xprt *xprt, struct net *net)
1861 {
1862 	kref_init(&xprt->kref);
1863 
1864 	spin_lock_init(&xprt->transport_lock);
1865 	spin_lock_init(&xprt->reserve_lock);
1866 	spin_lock_init(&xprt->queue_lock);
1867 
1868 	INIT_LIST_HEAD(&xprt->free);
1869 	xprt->recv_queue = RB_ROOT;
1870 	INIT_LIST_HEAD(&xprt->xmit_queue);
1871 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
1872 	spin_lock_init(&xprt->bc_pa_lock);
1873 	INIT_LIST_HEAD(&xprt->bc_pa_list);
1874 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
1875 	INIT_LIST_HEAD(&xprt->xprt_switch);
1876 
1877 	xprt->last_used = jiffies;
1878 	xprt->cwnd = RPC_INITCWND;
1879 	xprt->bind_index = 0;
1880 
1881 	rpc_init_wait_queue(&xprt->binding, "xprt_binding");
1882 	rpc_init_wait_queue(&xprt->pending, "xprt_pending");
1883 	rpc_init_wait_queue(&xprt->sending, "xprt_sending");
1884 	rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");
1885 
1886 	xprt_init_xid(xprt);
1887 
1888 	xprt->xprt_net = get_net(net);
1889 }
1890 
1891 /**
1892  * xprt_create_transport - create an RPC transport
1893  * @args: rpc transport creation arguments
1894  *
1895  */
1896 struct rpc_xprt *xprt_create_transport(struct xprt_create *args)
1897 {
1898 	struct rpc_xprt	*xprt;
1899 	struct xprt_class *t;
1900 
1901 	spin_lock(&xprt_list_lock);
1902 	list_for_each_entry(t, &xprt_list, list) {
1903 		if (t->ident == args->ident) {
1904 			spin_unlock(&xprt_list_lock);
1905 			goto found;
1906 		}
1907 	}
1908 	spin_unlock(&xprt_list_lock);
1909 	dprintk("RPC: transport (%d) not supported\n", args->ident);
1910 	return ERR_PTR(-EIO);
1911 
1912 found:
1913 	xprt = t->setup(args);
1914 	if (IS_ERR(xprt))
1915 		goto out;
1916 	if (args->flags & XPRT_CREATE_NO_IDLE_TIMEOUT)
1917 		xprt->idle_timeout = 0;
1918 	INIT_WORK(&xprt->task_cleanup, xprt_autoclose);
1919 	if (xprt_has_timer(xprt))
1920 		timer_setup(&xprt->timer, xprt_init_autodisconnect, 0);
1921 	else
1922 		timer_setup(&xprt->timer, NULL, 0);
1923 
1924 	if (strlen(args->servername) > RPC_MAXNETNAMELEN) {
1925 		xprt_destroy(xprt);
1926 		return ERR_PTR(-EINVAL);
1927 	}
1928 	xprt->servername = kstrdup(args->servername, GFP_KERNEL);
1929 	if (xprt->servername == NULL) {
1930 		xprt_destroy(xprt);
1931 		return ERR_PTR(-ENOMEM);
1932 	}
1933 
1934 	rpc_xprt_debugfs_register(xprt);
1935 
1936 	trace_xprt_create(xprt);
1937 out:
1938 	return xprt;
1939 }
1940 
1941 static void xprt_destroy_cb(struct work_struct *work)
1942 {
1943 	struct rpc_xprt *xprt =
1944 		container_of(work, struct rpc_xprt, task_cleanup);
1945 
1946 	trace_xprt_destroy(xprt);
1947 
1948 	rpc_xprt_debugfs_unregister(xprt);
1949 	rpc_destroy_wait_queue(&xprt->binding);
1950 	rpc_destroy_wait_queue(&xprt->pending);
1951 	rpc_destroy_wait_queue(&xprt->sending);
1952 	rpc_destroy_wait_queue(&xprt->backlog);
1953 	kfree(xprt->servername);
1954 	/*
1955 	 * Destroy any existing back channel
1956 	 */
1957 	xprt_destroy_backchannel(xprt, UINT_MAX);
1958 
1959 	/*
1960 	 * Tear down transport state and free the rpc_xprt
1961 	 */
1962 	xprt->ops->destroy(xprt);
1963 }
1964 
1965 /**
1966  * xprt_destroy - destroy an RPC transport, killing off all requests.
1967  * @xprt: transport to destroy
1968  *
1969  */
1970 static void xprt_destroy(struct rpc_xprt *xprt)
1971 {
1972 	/*
1973 	 * Exclude transport connect/disconnect handlers and autoclose
1974 	 */
1975 	wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_UNINTERRUPTIBLE);
1976 
1977 	del_timer_sync(&xprt->timer);
1978 
1979 	/*
1980 	 * Destroy sockets etc from the system workqueue so they can
1981 	 * safely flush receive work running on rpciod.
1982 	 */
1983 	INIT_WORK(&xprt->task_cleanup, xprt_destroy_cb);
1984 	schedule_work(&xprt->task_cleanup);
1985 }
1986 
1987 static void xprt_destroy_kref(struct kref *kref)
1988 {
1989 	xprt_destroy(container_of(kref, struct rpc_xprt, kref));
1990 }
1991 
1992 /**
1993  * xprt_get - return a reference to an RPC transport.
1994  * @xprt: pointer to the transport
1995  *
1996  */
1997 struct rpc_xprt *xprt_get(struct rpc_xprt *xprt)
1998 {
1999 	if (xprt != NULL && kref_get_unless_zero(&xprt->kref))
2000 		return xprt;
2001 	return NULL;
2002 }
2003 EXPORT_SYMBOL_GPL(xprt_get);
2004 
2005 /**
2006  * xprt_put - release a reference to an RPC transport.
2007  * @xprt: pointer to the transport
2008  *
2009  */
2010 void xprt_put(struct rpc_xprt *xprt)
2011 {
2012 	if (xprt != NULL)
2013 		kref_put(&xprt->kref, xprt_destroy_kref);
2014 }
2015 EXPORT_SYMBOL_GPL(xprt_put);
2016