2 * linux/net/sunrpc/xprt.c
4 * This is a generic RPC call interface supporting congestion avoidance,
5 * and asynchronous calls.
7 * The interface works like this:
9 * - When a process places a call, it allocates a request slot if
10 * one is available. Otherwise, it sleeps on the backlog queue
12 * - Next, the caller puts together the RPC message, stuffs it into
13 * the request struct, and calls xprt_transmit().
14 * - xprt_transmit sends the message and installs the caller on the
15 * transport's wait list. At the same time, if a reply is expected,
16 * it installs a timer that is run after the packet's timeout has
18 * - When a packet arrives, the data_ready handler walks the list of
19 * pending requests for that transport. If a matching XID is found, the
20 * caller is woken up, and the timer removed.
21 * - When no reply arrives within the timeout interval, the timer is
22 * fired by the kernel and runs xprt_timer(). It either adjusts the
23 * timeout values (minor timeout) or wakes up the caller with a status
25 * - When the caller receives a notification from RPC that a reply arrived,
26 * it should release the RPC slot, and process the reply.
27 * If the call timed out, it may choose to retry the operation by
28 * adjusting the initial timeout value, and simply calling rpc_call
31 * Support for async RPC is done through a set of RPC-specific scheduling
32 * primitives that `transparently' work for processes as well as async
33 * tasks that rely on callbacks.
35 * Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
37 * Transport switch API copyright (C) 2005, Chuck Lever <cel@netapp.com>
40 #include <linux/module.h>
42 #include <linux/types.h>
43 #include <linux/interrupt.h>
44 #include <linux/workqueue.h>
45 #include <linux/net.h>
46 #include <linux/ktime.h>
48 #include <linux/sunrpc/clnt.h>
49 #include <linux/sunrpc/metrics.h>
50 #include <linux/sunrpc/bc_xprt.h>
59 # define RPCDBG_FACILITY RPCDBG_XPRT
65 static void xprt_init(struct rpc_xprt
*xprt
, struct net
*net
);
66 static void xprt_request_init(struct rpc_task
*, struct rpc_xprt
*);
67 static void xprt_connect_status(struct rpc_task
*task
);
68 static int __xprt_get_cong(struct rpc_xprt
*, struct rpc_task
*);
69 static void xprt_destroy(struct rpc_xprt
*xprt
);
71 static DEFINE_SPINLOCK(xprt_list_lock
);
72 static LIST_HEAD(xprt_list
);
75 * The transport code maintains an estimate on the maximum number of out-
76 * standing RPC requests, using a smoothed version of the congestion
77 * avoidance implemented in 44BSD. This is basically the Van Jacobson
78 * congestion algorithm: If a retransmit occurs, the congestion window is
79 * halved; otherwise, it is incremented by 1/cwnd when
81 * - a reply is received and
82 * - a full number of requests are outstanding and
83 * - the congestion window hasn't been updated recently.
85 #define RPC_CWNDSHIFT (8U)
86 #define RPC_CWNDSCALE (1U << RPC_CWNDSHIFT)
87 #define RPC_INITCWND RPC_CWNDSCALE
88 #define RPC_MAXCWND(xprt) ((xprt)->max_reqs << RPC_CWNDSHIFT)
90 #define RPCXPRT_CONGESTED(xprt) ((xprt)->cong >= (xprt)->cwnd)
93 * xprt_register_transport - register a transport implementation
94 * @transport: transport to register
96 * If a transport implementation is loaded as a kernel module, it can
97 * call this interface to make itself known to the RPC client.
100 * 0: transport successfully registered
101 * -EEXIST: transport already registered
102 * -EINVAL: transport module being unloaded
104 int xprt_register_transport(struct xprt_class
*transport
)
106 struct xprt_class
*t
;
110 spin_lock(&xprt_list_lock
);
111 list_for_each_entry(t
, &xprt_list
, list
) {
112 /* don't register the same transport class twice */
113 if (t
->ident
== transport
->ident
)
117 list_add_tail(&transport
->list
, &xprt_list
);
118 printk(KERN_INFO
"RPC: Registered %s transport module.\n",
123 spin_unlock(&xprt_list_lock
);
126 EXPORT_SYMBOL_GPL(xprt_register_transport
);
129 * xprt_unregister_transport - unregister a transport implementation
130 * @transport: transport to unregister
133 * 0: transport successfully unregistered
134 * -ENOENT: transport never registered
136 int xprt_unregister_transport(struct xprt_class
*transport
)
138 struct xprt_class
*t
;
142 spin_lock(&xprt_list_lock
);
143 list_for_each_entry(t
, &xprt_list
, list
) {
144 if (t
== transport
) {
146 "RPC: Unregistered %s transport module.\n",
148 list_del_init(&transport
->list
);
155 spin_unlock(&xprt_list_lock
);
158 EXPORT_SYMBOL_GPL(xprt_unregister_transport
);
161 * xprt_load_transport - load a transport implementation
162 * @transport_name: transport to load
165 * 0: transport successfully loaded
166 * -ENOENT: transport module not available
168 int xprt_load_transport(const char *transport_name
)
170 struct xprt_class
*t
;
174 spin_lock(&xprt_list_lock
);
175 list_for_each_entry(t
, &xprt_list
, list
) {
176 if (strcmp(t
->name
, transport_name
) == 0) {
177 spin_unlock(&xprt_list_lock
);
181 spin_unlock(&xprt_list_lock
);
182 result
= request_module("xprt%s", transport_name
);
186 EXPORT_SYMBOL_GPL(xprt_load_transport
);
189 * xprt_reserve_xprt - serialize write access to transports
190 * @task: task that is requesting access to the transport
191 * @xprt: pointer to the target transport
193 * This prevents mixing the payload of separate requests, and prevents
194 * transport connects from colliding with writes. No congestion control
197 int xprt_reserve_xprt(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
199 struct rpc_rqst
*req
= task
->tk_rqstp
;
202 if (test_and_set_bit(XPRT_LOCKED
, &xprt
->state
)) {
203 if (task
== xprt
->snd_task
)
207 xprt
->snd_task
= task
;
214 dprintk("RPC: %5u failed to lock transport %p\n",
216 task
->tk_timeout
= 0;
217 task
->tk_status
= -EAGAIN
;
219 priority
= RPC_PRIORITY_LOW
;
220 else if (!req
->rq_ntrans
)
221 priority
= RPC_PRIORITY_NORMAL
;
223 priority
= RPC_PRIORITY_HIGH
;
224 rpc_sleep_on_priority(&xprt
->sending
, task
, NULL
, priority
);
227 EXPORT_SYMBOL_GPL(xprt_reserve_xprt
);
229 static void xprt_clear_locked(struct rpc_xprt
*xprt
)
231 xprt
->snd_task
= NULL
;
232 if (!test_bit(XPRT_CLOSE_WAIT
, &xprt
->state
)) {
233 smp_mb__before_clear_bit();
234 clear_bit(XPRT_LOCKED
, &xprt
->state
);
235 smp_mb__after_clear_bit();
237 queue_work(rpciod_workqueue
, &xprt
->task_cleanup
);
241 * xprt_reserve_xprt_cong - serialize write access to transports
242 * @task: task that is requesting access to the transport
244 * Same as xprt_reserve_xprt, but Van Jacobson congestion control is
245 * integrated into the decision of whether a request is allowed to be
246 * woken up and given access to the transport.
248 int xprt_reserve_xprt_cong(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
250 struct rpc_rqst
*req
= task
->tk_rqstp
;
253 if (test_and_set_bit(XPRT_LOCKED
, &xprt
->state
)) {
254 if (task
== xprt
->snd_task
)
259 xprt
->snd_task
= task
;
262 if (__xprt_get_cong(xprt
, task
)) {
263 xprt
->snd_task
= task
;
267 xprt_clear_locked(xprt
);
269 dprintk("RPC: %5u failed to lock transport %p\n", task
->tk_pid
, xprt
);
270 task
->tk_timeout
= 0;
271 task
->tk_status
= -EAGAIN
;
273 priority
= RPC_PRIORITY_LOW
;
274 else if (!req
->rq_ntrans
)
275 priority
= RPC_PRIORITY_NORMAL
;
277 priority
= RPC_PRIORITY_HIGH
;
278 rpc_sleep_on_priority(&xprt
->sending
, task
, NULL
, priority
);
281 EXPORT_SYMBOL_GPL(xprt_reserve_xprt_cong
);
283 static inline int xprt_lock_write(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
287 spin_lock_bh(&xprt
->transport_lock
);
288 retval
= xprt
->ops
->reserve_xprt(xprt
, task
);
289 spin_unlock_bh(&xprt
->transport_lock
);
293 static bool __xprt_lock_write_func(struct rpc_task
*task
, void *data
)
295 struct rpc_xprt
*xprt
= data
;
296 struct rpc_rqst
*req
;
298 req
= task
->tk_rqstp
;
299 xprt
->snd_task
= task
;
305 static void __xprt_lock_write_next(struct rpc_xprt
*xprt
)
307 if (test_and_set_bit(XPRT_LOCKED
, &xprt
->state
))
310 if (rpc_wake_up_first(&xprt
->sending
, __xprt_lock_write_func
, xprt
))
312 xprt_clear_locked(xprt
);
315 static bool __xprt_lock_write_cong_func(struct rpc_task
*task
, void *data
)
317 struct rpc_xprt
*xprt
= data
;
318 struct rpc_rqst
*req
;
320 req
= task
->tk_rqstp
;
322 xprt
->snd_task
= task
;
325 if (__xprt_get_cong(xprt
, task
)) {
326 xprt
->snd_task
= task
;
333 static void __xprt_lock_write_next_cong(struct rpc_xprt
*xprt
)
335 if (test_and_set_bit(XPRT_LOCKED
, &xprt
->state
))
337 if (RPCXPRT_CONGESTED(xprt
))
339 if (rpc_wake_up_first(&xprt
->sending
, __xprt_lock_write_cong_func
, xprt
))
342 xprt_clear_locked(xprt
);
346 * xprt_release_xprt - allow other requests to use a transport
347 * @xprt: transport with other tasks potentially waiting
348 * @task: task that is releasing access to the transport
350 * Note that "task" can be NULL. No congestion control is provided.
352 void xprt_release_xprt(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
354 if (xprt
->snd_task
== task
) {
356 struct rpc_rqst
*req
= task
->tk_rqstp
;
358 req
->rq_bytes_sent
= 0;
360 xprt_clear_locked(xprt
);
361 __xprt_lock_write_next(xprt
);
364 EXPORT_SYMBOL_GPL(xprt_release_xprt
);
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
371 * Note that "task" can be NULL. Another task is awoken to use the
372 * transport if the transport's congestion window allows it.
374 void xprt_release_xprt_cong(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
376 if (xprt
->snd_task
== task
) {
378 struct rpc_rqst
*req
= task
->tk_rqstp
;
380 req
->rq_bytes_sent
= 0;
382 xprt_clear_locked(xprt
);
383 __xprt_lock_write_next_cong(xprt
);
386 EXPORT_SYMBOL_GPL(xprt_release_xprt_cong
);
388 static inline void xprt_release_write(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
390 spin_lock_bh(&xprt
->transport_lock
);
391 xprt
->ops
->release_xprt(xprt
, task
);
392 spin_unlock_bh(&xprt
->transport_lock
);
396 * Van Jacobson congestion avoidance. Check if the congestion window
397 * overflowed. Put the task to sleep if this is the case.
400 __xprt_get_cong(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
402 struct rpc_rqst
*req
= task
->tk_rqstp
;
406 dprintk("RPC: %5u xprt_cwnd_limited cong = %lu cwnd = %lu\n",
407 task
->tk_pid
, xprt
->cong
, xprt
->cwnd
);
408 if (RPCXPRT_CONGESTED(xprt
))
411 xprt
->cong
+= RPC_CWNDSCALE
;
416 * Adjust the congestion window, and wake up the next task
417 * that has been sleeping due to congestion
420 __xprt_put_cong(struct rpc_xprt
*xprt
, struct rpc_rqst
*req
)
425 xprt
->cong
-= RPC_CWNDSCALE
;
426 __xprt_lock_write_next_cong(xprt
);
430 * xprt_release_rqst_cong - housekeeping when request is complete
431 * @task: RPC request that recently completed
433 * Useful for transports that require congestion control.
435 void xprt_release_rqst_cong(struct rpc_task
*task
)
437 struct rpc_rqst
*req
= task
->tk_rqstp
;
439 __xprt_put_cong(req
->rq_xprt
, req
);
441 EXPORT_SYMBOL_GPL(xprt_release_rqst_cong
);
444 * xprt_adjust_cwnd - adjust transport congestion window
445 * @xprt: pointer to xprt
446 * @task: recently completed RPC request used to adjust window
447 * @result: result code of completed RPC request
449 * We use a time-smoothed congestion estimator to avoid heavy oscillation.
451 void xprt_adjust_cwnd(struct rpc_xprt
*xprt
, struct rpc_task
*task
, int result
)
453 struct rpc_rqst
*req
= task
->tk_rqstp
;
454 unsigned long cwnd
= xprt
->cwnd
;
456 if (result
>= 0 && cwnd
<= xprt
->cong
) {
457 /* The (cwnd >> 1) term makes sure
458 * the result gets rounded properly. */
459 cwnd
+= (RPC_CWNDSCALE
* RPC_CWNDSCALE
+ (cwnd
>> 1)) / cwnd
;
460 if (cwnd
> RPC_MAXCWND(xprt
))
461 cwnd
= RPC_MAXCWND(xprt
);
462 __xprt_lock_write_next_cong(xprt
);
463 } else if (result
== -ETIMEDOUT
) {
465 if (cwnd
< RPC_CWNDSCALE
)
466 cwnd
= RPC_CWNDSCALE
;
468 dprintk("RPC: cong %ld, cwnd was %ld, now %ld\n",
469 xprt
->cong
, xprt
->cwnd
, cwnd
);
471 __xprt_put_cong(xprt
, req
);
473 EXPORT_SYMBOL_GPL(xprt_adjust_cwnd
);
476 * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue
477 * @xprt: transport with waiting tasks
478 * @status: result code to plant in each task before waking it
481 void xprt_wake_pending_tasks(struct rpc_xprt
*xprt
, int status
)
484 rpc_wake_up_status(&xprt
->pending
, status
);
486 rpc_wake_up(&xprt
->pending
);
488 EXPORT_SYMBOL_GPL(xprt_wake_pending_tasks
);
491 * xprt_wait_for_buffer_space - wait for transport output buffer to clear
492 * @task: task to be put to sleep
493 * @action: function pointer to be executed after wait
495 * Note that we only set the timer for the case of RPC_IS_SOFT(), since
496 * we don't in general want to force a socket disconnection due to
497 * an incomplete RPC call transmission.
499 void xprt_wait_for_buffer_space(struct rpc_task
*task
, rpc_action action
)
501 struct rpc_rqst
*req
= task
->tk_rqstp
;
502 struct rpc_xprt
*xprt
= req
->rq_xprt
;
504 task
->tk_timeout
= RPC_IS_SOFT(task
) ? req
->rq_timeout
: 0;
505 rpc_sleep_on(&xprt
->pending
, task
, action
);
507 EXPORT_SYMBOL_GPL(xprt_wait_for_buffer_space
);
510 * xprt_write_space - wake the task waiting for transport output buffer space
511 * @xprt: transport with waiting tasks
513 * Can be called in a soft IRQ context, so xprt_write_space never sleeps.
515 void xprt_write_space(struct rpc_xprt
*xprt
)
517 spin_lock_bh(&xprt
->transport_lock
);
518 if (xprt
->snd_task
) {
519 dprintk("RPC: write space: waking waiting task on "
521 rpc_wake_up_queued_task(&xprt
->pending
, xprt
->snd_task
);
523 spin_unlock_bh(&xprt
->transport_lock
);
525 EXPORT_SYMBOL_GPL(xprt_write_space
);
528 * xprt_set_retrans_timeout_def - set a request's retransmit timeout
529 * @task: task whose timeout is to be set
531 * Set a request's retransmit timeout based on the transport's
532 * default timeout parameters. Used by transports that don't adjust
533 * the retransmit timeout based on round-trip time estimation.
535 void xprt_set_retrans_timeout_def(struct rpc_task
*task
)
537 task
->tk_timeout
= task
->tk_rqstp
->rq_timeout
;
539 EXPORT_SYMBOL_GPL(xprt_set_retrans_timeout_def
);
542 * xprt_set_retrans_timeout_rtt - set a request's retransmit timeout
543 * @task: task whose timeout is to be set
545 * Set a request's retransmit timeout using the RTT estimator.
547 void xprt_set_retrans_timeout_rtt(struct rpc_task
*task
)
549 int timer
= task
->tk_msg
.rpc_proc
->p_timer
;
550 struct rpc_clnt
*clnt
= task
->tk_client
;
551 struct rpc_rtt
*rtt
= clnt
->cl_rtt
;
552 struct rpc_rqst
*req
= task
->tk_rqstp
;
553 unsigned long max_timeout
= clnt
->cl_timeout
->to_maxval
;
555 task
->tk_timeout
= rpc_calc_rto(rtt
, timer
);
556 task
->tk_timeout
<<= rpc_ntimeo(rtt
, timer
) + req
->rq_retries
;
557 if (task
->tk_timeout
> max_timeout
|| task
->tk_timeout
== 0)
558 task
->tk_timeout
= max_timeout
;
560 EXPORT_SYMBOL_GPL(xprt_set_retrans_timeout_rtt
);
562 static void xprt_reset_majortimeo(struct rpc_rqst
*req
)
564 const struct rpc_timeout
*to
= req
->rq_task
->tk_client
->cl_timeout
;
566 req
->rq_majortimeo
= req
->rq_timeout
;
567 if (to
->to_exponential
)
568 req
->rq_majortimeo
<<= to
->to_retries
;
570 req
->rq_majortimeo
+= to
->to_increment
* to
->to_retries
;
571 if (req
->rq_majortimeo
> to
->to_maxval
|| req
->rq_majortimeo
== 0)
572 req
->rq_majortimeo
= to
->to_maxval
;
573 req
->rq_majortimeo
+= jiffies
;
577 * xprt_adjust_timeout - adjust timeout values for next retransmit
578 * @req: RPC request containing parameters to use for the adjustment
581 int xprt_adjust_timeout(struct rpc_rqst
*req
)
583 struct rpc_xprt
*xprt
= req
->rq_xprt
;
584 const struct rpc_timeout
*to
= req
->rq_task
->tk_client
->cl_timeout
;
587 if (time_before(jiffies
, req
->rq_majortimeo
)) {
588 if (to
->to_exponential
)
589 req
->rq_timeout
<<= 1;
591 req
->rq_timeout
+= to
->to_increment
;
592 if (to
->to_maxval
&& req
->rq_timeout
>= to
->to_maxval
)
593 req
->rq_timeout
= to
->to_maxval
;
596 req
->rq_timeout
= to
->to_initval
;
598 xprt_reset_majortimeo(req
);
599 /* Reset the RTT counters == "slow start" */
600 spin_lock_bh(&xprt
->transport_lock
);
601 rpc_init_rtt(req
->rq_task
->tk_client
->cl_rtt
, to
->to_initval
);
602 spin_unlock_bh(&xprt
->transport_lock
);
606 if (req
->rq_timeout
== 0) {
607 printk(KERN_WARNING
"xprt_adjust_timeout: rq_timeout = 0!\n");
608 req
->rq_timeout
= 5 * HZ
;
613 static void xprt_autoclose(struct work_struct
*work
)
615 struct rpc_xprt
*xprt
=
616 container_of(work
, struct rpc_xprt
, task_cleanup
);
618 xprt
->ops
->close(xprt
);
619 clear_bit(XPRT_CLOSE_WAIT
, &xprt
->state
);
620 xprt_release_write(xprt
, NULL
);
624 * xprt_disconnect_done - mark a transport as disconnected
625 * @xprt: transport to flag for disconnect
628 void xprt_disconnect_done(struct rpc_xprt
*xprt
)
630 dprintk("RPC: disconnected transport %p\n", xprt
);
631 spin_lock_bh(&xprt
->transport_lock
);
632 xprt_clear_connected(xprt
);
633 xprt_wake_pending_tasks(xprt
, -EAGAIN
);
634 spin_unlock_bh(&xprt
->transport_lock
);
636 EXPORT_SYMBOL_GPL(xprt_disconnect_done
);
639 * xprt_force_disconnect - force a transport to disconnect
640 * @xprt: transport to disconnect
643 void xprt_force_disconnect(struct rpc_xprt
*xprt
)
645 /* Don't race with the test_bit() in xprt_clear_locked() */
646 spin_lock_bh(&xprt
->transport_lock
);
647 set_bit(XPRT_CLOSE_WAIT
, &xprt
->state
);
648 /* Try to schedule an autoclose RPC call */
649 if (test_and_set_bit(XPRT_LOCKED
, &xprt
->state
) == 0)
650 queue_work(rpciod_workqueue
, &xprt
->task_cleanup
);
651 xprt_wake_pending_tasks(xprt
, -EAGAIN
);
652 spin_unlock_bh(&xprt
->transport_lock
);
656 * xprt_conditional_disconnect - force a transport to disconnect
657 * @xprt: transport to disconnect
658 * @cookie: 'connection cookie'
660 * This attempts to break the connection if and only if 'cookie' matches
661 * the current transport 'connection cookie'. It ensures that we don't
662 * try to break the connection more than once when we need to retransmit
663 * a batch of RPC requests.
666 void xprt_conditional_disconnect(struct rpc_xprt
*xprt
, unsigned int cookie
)
668 /* Don't race with the test_bit() in xprt_clear_locked() */
669 spin_lock_bh(&xprt
->transport_lock
);
670 if (cookie
!= xprt
->connect_cookie
)
672 if (test_bit(XPRT_CLOSING
, &xprt
->state
) || !xprt_connected(xprt
))
674 set_bit(XPRT_CLOSE_WAIT
, &xprt
->state
);
675 /* Try to schedule an autoclose RPC call */
676 if (test_and_set_bit(XPRT_LOCKED
, &xprt
->state
) == 0)
677 queue_work(rpciod_workqueue
, &xprt
->task_cleanup
);
678 xprt_wake_pending_tasks(xprt
, -EAGAIN
);
680 spin_unlock_bh(&xprt
->transport_lock
);
684 xprt_init_autodisconnect(unsigned long data
)
686 struct rpc_xprt
*xprt
= (struct rpc_xprt
*)data
;
688 spin_lock(&xprt
->transport_lock
);
689 if (!list_empty(&xprt
->recv
))
691 if (test_and_set_bit(XPRT_LOCKED
, &xprt
->state
))
693 spin_unlock(&xprt
->transport_lock
);
694 set_bit(XPRT_CONNECTION_CLOSE
, &xprt
->state
);
695 queue_work(rpciod_workqueue
, &xprt
->task_cleanup
);
698 spin_unlock(&xprt
->transport_lock
);
702 * xprt_connect - schedule a transport connect operation
703 * @task: RPC task that is requesting the connect
706 void xprt_connect(struct rpc_task
*task
)
708 struct rpc_xprt
*xprt
= task
->tk_rqstp
->rq_xprt
;
710 dprintk("RPC: %5u xprt_connect xprt %p %s connected\n", task
->tk_pid
,
711 xprt
, (xprt_connected(xprt
) ? "is" : "is not"));
713 if (!xprt_bound(xprt
)) {
714 task
->tk_status
= -EAGAIN
;
717 if (!xprt_lock_write(xprt
, task
))
720 if (test_and_clear_bit(XPRT_CLOSE_WAIT
, &xprt
->state
))
721 xprt
->ops
->close(xprt
);
723 if (xprt_connected(xprt
))
724 xprt_release_write(xprt
, task
);
726 task
->tk_rqstp
->rq_bytes_sent
= 0;
727 task
->tk_timeout
= task
->tk_rqstp
->rq_timeout
;
728 rpc_sleep_on(&xprt
->pending
, task
, xprt_connect_status
);
730 if (test_bit(XPRT_CLOSING
, &xprt
->state
))
732 if (xprt_test_and_set_connecting(xprt
))
734 xprt
->stat
.connect_start
= jiffies
;
735 xprt
->ops
->connect(xprt
, task
);
739 static void xprt_connect_status(struct rpc_task
*task
)
741 struct rpc_xprt
*xprt
= task
->tk_rqstp
->rq_xprt
;
743 if (task
->tk_status
== 0) {
744 xprt
->stat
.connect_count
++;
745 xprt
->stat
.connect_time
+= (long)jiffies
- xprt
->stat
.connect_start
;
746 dprintk("RPC: %5u xprt_connect_status: connection established\n",
751 switch (task
->tk_status
) {
758 dprintk("RPC: %5u xprt_connect_status: retrying\n", task
->tk_pid
);
761 dprintk("RPC: %5u xprt_connect_status: connect attempt timed "
762 "out\n", task
->tk_pid
);
765 dprintk("RPC: %5u xprt_connect_status: error %d connecting to "
766 "server %s\n", task
->tk_pid
, -task
->tk_status
,
768 xprt_release_write(xprt
, task
);
769 task
->tk_status
= -EIO
;
774 * xprt_lookup_rqst - find an RPC request corresponding to an XID
775 * @xprt: transport on which the original request was transmitted
776 * @xid: RPC XID of incoming reply
779 struct rpc_rqst
*xprt_lookup_rqst(struct rpc_xprt
*xprt
, __be32 xid
)
781 struct rpc_rqst
*entry
;
783 list_for_each_entry(entry
, &xprt
->recv
, rq_list
)
784 if (entry
->rq_xid
== xid
)
787 dprintk("RPC: xprt_lookup_rqst did not find xid %08x\n",
789 xprt
->stat
.bad_xids
++;
792 EXPORT_SYMBOL_GPL(xprt_lookup_rqst
);
794 static void xprt_update_rtt(struct rpc_task
*task
)
796 struct rpc_rqst
*req
= task
->tk_rqstp
;
797 struct rpc_rtt
*rtt
= task
->tk_client
->cl_rtt
;
798 unsigned int timer
= task
->tk_msg
.rpc_proc
->p_timer
;
799 long m
= usecs_to_jiffies(ktime_to_us(req
->rq_rtt
));
802 if (req
->rq_ntrans
== 1)
803 rpc_update_rtt(rtt
, timer
, m
);
804 rpc_set_timeo(rtt
, timer
, req
->rq_ntrans
- 1);
809 * xprt_complete_rqst - called when reply processing is complete
810 * @task: RPC request that recently completed
811 * @copied: actual number of bytes received from the transport
813 * Caller holds transport lock.
815 void xprt_complete_rqst(struct rpc_task
*task
, int copied
)
817 struct rpc_rqst
*req
= task
->tk_rqstp
;
818 struct rpc_xprt
*xprt
= req
->rq_xprt
;
820 dprintk("RPC: %5u xid %08x complete (%d bytes received)\n",
821 task
->tk_pid
, ntohl(req
->rq_xid
), copied
);
824 req
->rq_rtt
= ktime_sub(ktime_get(), req
->rq_xtime
);
825 if (xprt
->ops
->timer
!= NULL
)
826 xprt_update_rtt(task
);
828 list_del_init(&req
->rq_list
);
829 req
->rq_private_buf
.len
= copied
;
830 /* Ensure all writes are done before we update */
831 /* req->rq_reply_bytes_recvd */
833 req
->rq_reply_bytes_recvd
= copied
;
834 rpc_wake_up_queued_task(&xprt
->pending
, task
);
836 EXPORT_SYMBOL_GPL(xprt_complete_rqst
);
838 static void xprt_timer(struct rpc_task
*task
)
840 struct rpc_rqst
*req
= task
->tk_rqstp
;
841 struct rpc_xprt
*xprt
= req
->rq_xprt
;
843 if (task
->tk_status
!= -ETIMEDOUT
)
845 dprintk("RPC: %5u xprt_timer\n", task
->tk_pid
);
847 spin_lock_bh(&xprt
->transport_lock
);
848 if (!req
->rq_reply_bytes_recvd
) {
849 if (xprt
->ops
->timer
)
850 xprt
->ops
->timer(xprt
, task
);
853 spin_unlock_bh(&xprt
->transport_lock
);
856 static inline int xprt_has_timer(struct rpc_xprt
*xprt
)
858 return xprt
->idle_timeout
!= 0;
862 * xprt_prepare_transmit - reserve the transport before sending a request
863 * @task: RPC task about to send a request
866 bool xprt_prepare_transmit(struct rpc_task
*task
)
868 struct rpc_rqst
*req
= task
->tk_rqstp
;
869 struct rpc_xprt
*xprt
= req
->rq_xprt
;
872 dprintk("RPC: %5u xprt_prepare_transmit\n", task
->tk_pid
);
874 spin_lock_bh(&xprt
->transport_lock
);
875 if (!req
->rq_bytes_sent
) {
876 if (req
->rq_reply_bytes_recvd
) {
877 task
->tk_status
= req
->rq_reply_bytes_recvd
;
880 if ((task
->tk_flags
& RPC_TASK_NO_RETRANS_TIMEOUT
)
881 && xprt_connected(xprt
)
882 && req
->rq_connect_cookie
== xprt
->connect_cookie
) {
883 xprt
->ops
->set_retrans_timeout(task
);
884 rpc_sleep_on(&xprt
->pending
, task
, xprt_timer
);
888 if (!xprt
->ops
->reserve_xprt(xprt
, task
)) {
889 task
->tk_status
= -EAGAIN
;
894 spin_unlock_bh(&xprt
->transport_lock
);
898 void xprt_end_transmit(struct rpc_task
*task
)
900 xprt_release_write(task
->tk_rqstp
->rq_xprt
, task
);
904 * xprt_transmit - send an RPC request on a transport
905 * @task: controlling RPC task
907 * We have to copy the iovec because sendmsg fiddles with its contents.
909 void xprt_transmit(struct rpc_task
*task
)
911 struct rpc_rqst
*req
= task
->tk_rqstp
;
912 struct rpc_xprt
*xprt
= req
->rq_xprt
;
915 dprintk("RPC: %5u xprt_transmit(%u)\n", task
->tk_pid
, req
->rq_slen
);
917 if (!req
->rq_reply_bytes_recvd
) {
918 if (list_empty(&req
->rq_list
) && rpc_reply_expected(task
)) {
920 * Add to the list only if we're expecting a reply
922 spin_lock_bh(&xprt
->transport_lock
);
923 /* Update the softirq receive buffer */
924 memcpy(&req
->rq_private_buf
, &req
->rq_rcv_buf
,
925 sizeof(req
->rq_private_buf
));
926 /* Add request to the receive list */
927 list_add_tail(&req
->rq_list
, &xprt
->recv
);
928 spin_unlock_bh(&xprt
->transport_lock
);
929 xprt_reset_majortimeo(req
);
930 /* Turn off autodisconnect */
931 del_singleshot_timer_sync(&xprt
->timer
);
933 } else if (!req
->rq_bytes_sent
)
936 req
->rq_xtime
= ktime_get();
937 status
= xprt
->ops
->send_request(task
);
939 task
->tk_status
= status
;
943 dprintk("RPC: %5u xmit complete\n", task
->tk_pid
);
944 task
->tk_flags
|= RPC_TASK_SENT
;
945 spin_lock_bh(&xprt
->transport_lock
);
947 xprt
->ops
->set_retrans_timeout(task
);
949 numreqs
= atomic_read(&xprt
->num_reqs
);
950 if (numreqs
> xprt
->stat
.max_slots
)
951 xprt
->stat
.max_slots
= numreqs
;
953 xprt
->stat
.req_u
+= xprt
->stat
.sends
- xprt
->stat
.recvs
;
954 xprt
->stat
.bklog_u
+= xprt
->backlog
.qlen
;
955 xprt
->stat
.sending_u
+= xprt
->sending
.qlen
;
956 xprt
->stat
.pending_u
+= xprt
->pending
.qlen
;
958 /* Don't race with disconnect */
959 if (!xprt_connected(xprt
))
960 task
->tk_status
= -ENOTCONN
;
963 * Sleep on the pending queue since
964 * we're expecting a reply.
966 if (!req
->rq_reply_bytes_recvd
&& rpc_reply_expected(task
))
967 rpc_sleep_on(&xprt
->pending
, task
, xprt_timer
);
968 req
->rq_connect_cookie
= xprt
->connect_cookie
;
970 spin_unlock_bh(&xprt
->transport_lock
);
973 static void xprt_add_backlog(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
975 set_bit(XPRT_CONGESTED
, &xprt
->state
);
976 rpc_sleep_on(&xprt
->backlog
, task
, NULL
);
979 static void xprt_wake_up_backlog(struct rpc_xprt
*xprt
)
981 if (rpc_wake_up_next(&xprt
->backlog
) == NULL
)
982 clear_bit(XPRT_CONGESTED
, &xprt
->state
);
985 static bool xprt_throttle_congested(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
989 if (!test_bit(XPRT_CONGESTED
, &xprt
->state
))
991 spin_lock(&xprt
->reserve_lock
);
992 if (test_bit(XPRT_CONGESTED
, &xprt
->state
)) {
993 rpc_sleep_on(&xprt
->backlog
, task
, NULL
);
996 spin_unlock(&xprt
->reserve_lock
);
1001 static struct rpc_rqst
*xprt_dynamic_alloc_slot(struct rpc_xprt
*xprt
, gfp_t gfp_flags
)
1003 struct rpc_rqst
*req
= ERR_PTR(-EAGAIN
);
1005 if (!atomic_add_unless(&xprt
->num_reqs
, 1, xprt
->max_reqs
))
1007 req
= kzalloc(sizeof(struct rpc_rqst
), gfp_flags
);
1010 atomic_dec(&xprt
->num_reqs
);
1011 req
= ERR_PTR(-ENOMEM
);
1016 static bool xprt_dynamic_free_slot(struct rpc_xprt
*xprt
, struct rpc_rqst
*req
)
1018 if (atomic_add_unless(&xprt
->num_reqs
, -1, xprt
->min_reqs
)) {
1025 void xprt_alloc_slot(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
1027 struct rpc_rqst
*req
;
1029 spin_lock(&xprt
->reserve_lock
);
1030 if (!list_empty(&xprt
->free
)) {
1031 req
= list_entry(xprt
->free
.next
, struct rpc_rqst
, rq_list
);
1032 list_del(&req
->rq_list
);
1035 req
= xprt_dynamic_alloc_slot(xprt
, GFP_NOWAIT
|__GFP_NOWARN
);
1038 switch (PTR_ERR(req
)) {
1040 dprintk("RPC: dynamic allocation of request slot "
1041 "failed! Retrying\n");
1042 task
->tk_status
= -ENOMEM
;
1045 xprt_add_backlog(xprt
, task
);
1046 dprintk("RPC: waiting for request slot\n");
1048 task
->tk_status
= -EAGAIN
;
1050 spin_unlock(&xprt
->reserve_lock
);
1053 task
->tk_status
= 0;
1054 task
->tk_rqstp
= req
;
1055 xprt_request_init(task
, xprt
);
1056 spin_unlock(&xprt
->reserve_lock
);
1058 EXPORT_SYMBOL_GPL(xprt_alloc_slot
);
1060 void xprt_lock_and_alloc_slot(struct rpc_xprt
*xprt
, struct rpc_task
*task
)
1062 /* Note: grabbing the xprt_lock_write() ensures that we throttle
1063 * new slot allocation if the transport is congested (i.e. when
1064 * reconnecting a stream transport or when out of socket write
1067 if (xprt_lock_write(xprt
, task
)) {
1068 xprt_alloc_slot(xprt
, task
);
1069 xprt_release_write(xprt
, task
);
1072 EXPORT_SYMBOL_GPL(xprt_lock_and_alloc_slot
);
1074 static void xprt_free_slot(struct rpc_xprt
*xprt
, struct rpc_rqst
*req
)
1076 spin_lock(&xprt
->reserve_lock
);
1077 if (!xprt_dynamic_free_slot(xprt
, req
)) {
1078 memset(req
, 0, sizeof(*req
)); /* mark unused */
1079 list_add(&req
->rq_list
, &xprt
->free
);
1081 xprt_wake_up_backlog(xprt
);
1082 spin_unlock(&xprt
->reserve_lock
);
1085 static void xprt_free_all_slots(struct rpc_xprt
*xprt
)
1087 struct rpc_rqst
*req
;
1088 while (!list_empty(&xprt
->free
)) {
1089 req
= list_first_entry(&xprt
->free
, struct rpc_rqst
, rq_list
);
1090 list_del(&req
->rq_list
);
1095 struct rpc_xprt
*xprt_alloc(struct net
*net
, size_t size
,
1096 unsigned int num_prealloc
,
1097 unsigned int max_alloc
)
1099 struct rpc_xprt
*xprt
;
1100 struct rpc_rqst
*req
;
1103 xprt
= kzalloc(size
, GFP_KERNEL
);
1107 xprt_init(xprt
, net
);
1109 for (i
= 0; i
< num_prealloc
; i
++) {
1110 req
= kzalloc(sizeof(struct rpc_rqst
), GFP_KERNEL
);
1113 list_add(&req
->rq_list
, &xprt
->free
);
1115 if (max_alloc
> num_prealloc
)
1116 xprt
->max_reqs
= max_alloc
;
1118 xprt
->max_reqs
= num_prealloc
;
1119 xprt
->min_reqs
= num_prealloc
;
1120 atomic_set(&xprt
->num_reqs
, num_prealloc
);
1129 EXPORT_SYMBOL_GPL(xprt_alloc
);
1131 void xprt_free(struct rpc_xprt
*xprt
)
1133 put_net(xprt
->xprt_net
);
1134 xprt_free_all_slots(xprt
);
1137 EXPORT_SYMBOL_GPL(xprt_free
);
1140 * xprt_reserve - allocate an RPC request slot
1141 * @task: RPC task requesting a slot allocation
1143 * If the transport is marked as being congested, or if no more
1144 * slots are available, place the task on the transport's
1147 void xprt_reserve(struct rpc_task
*task
)
1149 struct rpc_xprt
*xprt
;
1151 task
->tk_status
= 0;
1152 if (task
->tk_rqstp
!= NULL
)
1155 task
->tk_timeout
= 0;
1156 task
->tk_status
= -EAGAIN
;
1158 xprt
= rcu_dereference(task
->tk_client
->cl_xprt
);
1159 if (!xprt_throttle_congested(xprt
, task
))
1160 xprt
->ops
->alloc_slot(xprt
, task
);
1165 * xprt_retry_reserve - allocate an RPC request slot
1166 * @task: RPC task requesting a slot allocation
1168 * If no more slots are available, place the task on the transport's
1170 * Note that the only difference with xprt_reserve is that we now
1171 * ignore the value of the XPRT_CONGESTED flag.
1173 void xprt_retry_reserve(struct rpc_task
*task
)
1175 struct rpc_xprt
*xprt
;
1177 task
->tk_status
= 0;
1178 if (task
->tk_rqstp
!= NULL
)
1181 task
->tk_timeout
= 0;
1182 task
->tk_status
= -EAGAIN
;
1184 xprt
= rcu_dereference(task
->tk_client
->cl_xprt
);
1185 xprt
->ops
->alloc_slot(xprt
, task
);
1189 static inline __be32
xprt_alloc_xid(struct rpc_xprt
*xprt
)
1191 return (__force __be32
)xprt
->xid
++;
1194 static inline void xprt_init_xid(struct rpc_xprt
*xprt
)
1196 xprt
->xid
= prandom_u32();
1199 static void xprt_request_init(struct rpc_task
*task
, struct rpc_xprt
*xprt
)
1201 struct rpc_rqst
*req
= task
->tk_rqstp
;
1203 INIT_LIST_HEAD(&req
->rq_list
);
1204 req
->rq_timeout
= task
->tk_client
->cl_timeout
->to_initval
;
1205 req
->rq_task
= task
;
1206 req
->rq_xprt
= xprt
;
1207 req
->rq_buffer
= NULL
;
1208 req
->rq_xid
= xprt_alloc_xid(xprt
);
1209 req
->rq_connect_cookie
= xprt
->connect_cookie
- 1;
1210 req
->rq_bytes_sent
= 0;
1211 req
->rq_snd_buf
.len
= 0;
1212 req
->rq_snd_buf
.buflen
= 0;
1213 req
->rq_rcv_buf
.len
= 0;
1214 req
->rq_rcv_buf
.buflen
= 0;
1215 req
->rq_release_snd_buf
= NULL
;
1216 xprt_reset_majortimeo(req
);
1217 dprintk("RPC: %5u reserved req %p xid %08x\n", task
->tk_pid
,
1218 req
, ntohl(req
->rq_xid
));
1222 * xprt_release - release an RPC request slot
1223 * @task: task which is finished with the slot
1226 void xprt_release(struct rpc_task
*task
)
1228 struct rpc_xprt
*xprt
;
1229 struct rpc_rqst
*req
= task
->tk_rqstp
;
1232 if (task
->tk_client
) {
1234 xprt
= rcu_dereference(task
->tk_client
->cl_xprt
);
1235 if (xprt
->snd_task
== task
)
1236 xprt_release_write(xprt
, task
);
1242 xprt
= req
->rq_xprt
;
1243 if (task
->tk_ops
->rpc_count_stats
!= NULL
)
1244 task
->tk_ops
->rpc_count_stats(task
, task
->tk_calldata
);
1245 else if (task
->tk_client
)
1246 rpc_count_iostats(task
, task
->tk_client
->cl_metrics
);
1247 spin_lock_bh(&xprt
->transport_lock
);
1248 xprt
->ops
->release_xprt(xprt
, task
);
1249 if (xprt
->ops
->release_request
)
1250 xprt
->ops
->release_request(task
);
1251 if (!list_empty(&req
->rq_list
))
1252 list_del(&req
->rq_list
);
1253 xprt
->last_used
= jiffies
;
1254 if (list_empty(&xprt
->recv
) && xprt_has_timer(xprt
))
1255 mod_timer(&xprt
->timer
,
1256 xprt
->last_used
+ xprt
->idle_timeout
);
1257 spin_unlock_bh(&xprt
->transport_lock
);
1259 xprt
->ops
->buf_free(req
->rq_buffer
);
1260 if (req
->rq_cred
!= NULL
)
1261 put_rpccred(req
->rq_cred
);
1262 task
->tk_rqstp
= NULL
;
1263 if (req
->rq_release_snd_buf
)
1264 req
->rq_release_snd_buf(req
);
1266 dprintk("RPC: %5u release request %p\n", task
->tk_pid
, req
);
1267 if (likely(!bc_prealloc(req
)))
1268 xprt_free_slot(xprt
, req
);
1270 xprt_free_bc_request(req
);
1273 static void xprt_init(struct rpc_xprt
*xprt
, struct net
*net
)
1275 atomic_set(&xprt
->count
, 1);
1277 spin_lock_init(&xprt
->transport_lock
);
1278 spin_lock_init(&xprt
->reserve_lock
);
1280 INIT_LIST_HEAD(&xprt
->free
);
1281 INIT_LIST_HEAD(&xprt
->recv
);
1282 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
1283 spin_lock_init(&xprt
->bc_pa_lock
);
1284 INIT_LIST_HEAD(&xprt
->bc_pa_list
);
1285 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
1287 xprt
->last_used
= jiffies
;
1288 xprt
->cwnd
= RPC_INITCWND
;
1289 xprt
->bind_index
= 0;
1291 rpc_init_wait_queue(&xprt
->binding
, "xprt_binding");
1292 rpc_init_wait_queue(&xprt
->pending
, "xprt_pending");
1293 rpc_init_priority_wait_queue(&xprt
->sending
, "xprt_sending");
1294 rpc_init_priority_wait_queue(&xprt
->backlog
, "xprt_backlog");
1296 xprt_init_xid(xprt
);
1298 xprt
->xprt_net
= get_net(net
);
1302 * xprt_create_transport - create an RPC transport
1303 * @args: rpc transport creation arguments
1306 struct rpc_xprt
*xprt_create_transport(struct xprt_create
*args
)
1308 struct rpc_xprt
*xprt
;
1309 struct xprt_class
*t
;
1311 spin_lock(&xprt_list_lock
);
1312 list_for_each_entry(t
, &xprt_list
, list
) {
1313 if (t
->ident
== args
->ident
) {
1314 spin_unlock(&xprt_list_lock
);
1318 spin_unlock(&xprt_list_lock
);
1319 printk(KERN_ERR
"RPC: transport (%d) not supported\n", args
->ident
);
1320 return ERR_PTR(-EIO
);
1323 xprt
= t
->setup(args
);
1325 dprintk("RPC: xprt_create_transport: failed, %ld\n",
1329 if (args
->flags
& XPRT_CREATE_NO_IDLE_TIMEOUT
)
1330 xprt
->idle_timeout
= 0;
1331 INIT_WORK(&xprt
->task_cleanup
, xprt_autoclose
);
1332 if (xprt_has_timer(xprt
))
1333 setup_timer(&xprt
->timer
, xprt_init_autodisconnect
,
1334 (unsigned long)xprt
);
1336 init_timer(&xprt
->timer
);
1338 if (strlen(args
->servername
) > RPC_MAXNETNAMELEN
) {
1340 return ERR_PTR(-EINVAL
);
1342 xprt
->servername
= kstrdup(args
->servername
, GFP_KERNEL
);
1343 if (xprt
->servername
== NULL
) {
1345 return ERR_PTR(-ENOMEM
);
1348 dprintk("RPC: created transport %p with %u slots\n", xprt
,
1355 * xprt_destroy - destroy an RPC transport, killing off all requests.
1356 * @xprt: transport to destroy
1359 static void xprt_destroy(struct rpc_xprt
*xprt
)
1361 dprintk("RPC: destroying transport %p\n", xprt
);
1362 del_timer_sync(&xprt
->timer
);
1364 rpc_destroy_wait_queue(&xprt
->binding
);
1365 rpc_destroy_wait_queue(&xprt
->pending
);
1366 rpc_destroy_wait_queue(&xprt
->sending
);
1367 rpc_destroy_wait_queue(&xprt
->backlog
);
1368 cancel_work_sync(&xprt
->task_cleanup
);
1369 kfree(xprt
->servername
);
1371 * Tear down transport state and free the rpc_xprt
1373 xprt
->ops
->destroy(xprt
);
1377 * xprt_put - release a reference to an RPC transport.
1378 * @xprt: pointer to the transport
1381 void xprt_put(struct rpc_xprt
*xprt
)
1383 if (atomic_dec_and_test(&xprt
->count
))