Merge branch 'stable-3.2' into pandora-3.2
[pandora-kernel.git] / net / sunrpc / xprt.c
1 /*
2  *  linux/net/sunrpc/xprt.c
3  *
4  *  This is a generic RPC call interface supporting congestion avoidance,
5  *  and asynchronous calls.
6  *
7  *  The interface works like this:
8  *
9  *  -   When a process places a call, it allocates a request slot if
10  *      one is available. Otherwise, it sleeps on the backlog queue
11  *      (xprt_reserve).
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
17  *      expired.
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
24  *      of -ETIMEDOUT.
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
29  *      again.
30  *
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.
34  *
35  *  Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
36  *
37  *  Transport switch API copyright (C) 2005, Chuck Lever <cel@netapp.com>
38  */
39
40 #include <linux/module.h>
41
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>
47
48 #include <linux/sunrpc/clnt.h>
49 #include <linux/sunrpc/metrics.h>
50 #include <linux/sunrpc/bc_xprt.h>
51
52 #include "sunrpc.h"
53
54 /*
55  * Local variables
56  */
57
58 #ifdef RPC_DEBUG
59 # define RPCDBG_FACILITY        RPCDBG_XPRT
60 #endif
61
62 /*
63  * Local functions
64  */
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
70 static DEFINE_SPINLOCK(xprt_list_lock);
71 static LIST_HEAD(xprt_list);
72
73 /*
74  * The transport code maintains an estimate on the maximum number of out-
75  * standing RPC requests, using a smoothed version of the congestion
76  * avoidance implemented in 44BSD. This is basically the Van Jacobson
77  * congestion algorithm: If a retransmit occurs, the congestion window is
78  * halved; otherwise, it is incremented by 1/cwnd when
79  *
80  *      -       a reply is received and
81  *      -       a full number of requests are outstanding and
82  *      -       the congestion window hasn't been updated recently.
83  */
84 #define RPC_CWNDSHIFT           (8U)
85 #define RPC_CWNDSCALE           (1U << RPC_CWNDSHIFT)
86 #define RPC_INITCWND            RPC_CWNDSCALE
87 #define RPC_MAXCWND(xprt)       ((xprt)->max_reqs << RPC_CWNDSHIFT)
88
89 #define RPCXPRT_CONGESTED(xprt) ((xprt)->cong >= (xprt)->cwnd)
90
91 /**
92  * xprt_register_transport - register a transport implementation
93  * @transport: transport to register
94  *
95  * If a transport implementation is loaded as a kernel module, it can
96  * call this interface to make itself known to the RPC client.
97  *
98  * Returns:
99  * 0:           transport successfully registered
100  * -EEXIST:     transport already registered
101  * -EINVAL:     transport module being unloaded
102  */
103 int xprt_register_transport(struct xprt_class *transport)
104 {
105         struct xprt_class *t;
106         int result;
107
108         result = -EEXIST;
109         spin_lock(&xprt_list_lock);
110         list_for_each_entry(t, &xprt_list, list) {
111                 /* don't register the same transport class twice */
112                 if (t->ident == transport->ident)
113                         goto out;
114         }
115
116         list_add_tail(&transport->list, &xprt_list);
117         printk(KERN_INFO "RPC: Registered %s transport module.\n",
118                transport->name);
119         result = 0;
120
121 out:
122         spin_unlock(&xprt_list_lock);
123         return result;
124 }
125 EXPORT_SYMBOL_GPL(xprt_register_transport);
126
127 /**
128  * xprt_unregister_transport - unregister a transport implementation
129  * @transport: transport to unregister
130  *
131  * Returns:
132  * 0:           transport successfully unregistered
133  * -ENOENT:     transport never registered
134  */
135 int xprt_unregister_transport(struct xprt_class *transport)
136 {
137         struct xprt_class *t;
138         int result;
139
140         result = 0;
141         spin_lock(&xprt_list_lock);
142         list_for_each_entry(t, &xprt_list, list) {
143                 if (t == transport) {
144                         printk(KERN_INFO
145                                 "RPC: Unregistered %s transport module.\n",
146                                 transport->name);
147                         list_del_init(&transport->list);
148                         goto out;
149                 }
150         }
151         result = -ENOENT;
152
153 out:
154         spin_unlock(&xprt_list_lock);
155         return result;
156 }
157 EXPORT_SYMBOL_GPL(xprt_unregister_transport);
158
159 /**
160  * xprt_load_transport - load a transport implementation
161  * @transport_name: transport to load
162  *
163  * Returns:
164  * 0:           transport successfully loaded
165  * -ENOENT:     transport module not available
166  */
167 int xprt_load_transport(const char *transport_name)
168 {
169         struct xprt_class *t;
170         int result;
171
172         result = 0;
173         spin_lock(&xprt_list_lock);
174         list_for_each_entry(t, &xprt_list, list) {
175                 if (strcmp(t->name, transport_name) == 0) {
176                         spin_unlock(&xprt_list_lock);
177                         goto out;
178                 }
179         }
180         spin_unlock(&xprt_list_lock);
181         result = request_module("xprt%s", transport_name);
182 out:
183         return result;
184 }
185 EXPORT_SYMBOL_GPL(xprt_load_transport);
186
187 /**
188  * xprt_reserve_xprt - serialize write access to transports
189  * @task: task that is requesting access to the transport
190  * @xprt: pointer to the target transport
191  *
192  * This prevents mixing the payload of separate requests, and prevents
193  * transport connects from colliding with writes.  No congestion control
194  * is provided.
195  */
196 int xprt_reserve_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
197 {
198         struct rpc_rqst *req = task->tk_rqstp;
199         int priority;
200
201         if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
202                 if (task == xprt->snd_task)
203                         return 1;
204                 goto out_sleep;
205         }
206         xprt->snd_task = task;
207         if (req != NULL) {
208                 req->rq_bytes_sent = 0;
209                 req->rq_ntrans++;
210         }
211
212         return 1;
213
214 out_sleep:
215         dprintk("RPC: %5u failed to lock transport %p\n",
216                         task->tk_pid, xprt);
217         task->tk_timeout = 0;
218         task->tk_status = -EAGAIN;
219         if (req == NULL)
220                 priority = RPC_PRIORITY_LOW;
221         else if (!req->rq_ntrans)
222                 priority = RPC_PRIORITY_NORMAL;
223         else
224                 priority = RPC_PRIORITY_HIGH;
225         rpc_sleep_on_priority(&xprt->sending, task, NULL, priority);
226         return 0;
227 }
228 EXPORT_SYMBOL_GPL(xprt_reserve_xprt);
229
230 static void xprt_clear_locked(struct rpc_xprt *xprt)
231 {
232         xprt->snd_task = NULL;
233         if (!test_bit(XPRT_CLOSE_WAIT, &xprt->state) || xprt->shutdown) {
234                 smp_mb__before_clear_bit();
235                 clear_bit(XPRT_LOCKED, &xprt->state);
236                 smp_mb__after_clear_bit();
237         } else
238                 queue_work(rpciod_workqueue, &xprt->task_cleanup);
239 }
240
241 /*
242  * xprt_reserve_xprt_cong - serialize write access to transports
243  * @task: task that is requesting access to the transport
244  *
245  * Same as xprt_reserve_xprt, but Van Jacobson congestion control is
246  * integrated into the decision of whether a request is allowed to be
247  * woken up and given access to the transport.
248  */
249 int xprt_reserve_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
250 {
251         struct rpc_rqst *req = task->tk_rqstp;
252         int priority;
253
254         if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
255                 if (task == xprt->snd_task)
256                         return 1;
257                 goto out_sleep;
258         }
259         if (req == NULL) {
260                 xprt->snd_task = task;
261                 return 1;
262         }
263         if (__xprt_get_cong(xprt, task)) {
264                 xprt->snd_task = task;
265                 req->rq_bytes_sent = 0;
266                 req->rq_ntrans++;
267                 return 1;
268         }
269         xprt_clear_locked(xprt);
270 out_sleep:
271         dprintk("RPC: %5u failed to lock transport %p\n", task->tk_pid, xprt);
272         task->tk_timeout = 0;
273         task->tk_status = -EAGAIN;
274         if (req == NULL)
275                 priority = RPC_PRIORITY_LOW;
276         else if (!req->rq_ntrans)
277                 priority = RPC_PRIORITY_NORMAL;
278         else
279                 priority = RPC_PRIORITY_HIGH;
280         rpc_sleep_on_priority(&xprt->sending, task, NULL, priority);
281         return 0;
282 }
283 EXPORT_SYMBOL_GPL(xprt_reserve_xprt_cong);
284
285 static inline int xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
286 {
287         int retval;
288
289         spin_lock_bh(&xprt->transport_lock);
290         retval = xprt->ops->reserve_xprt(xprt, task);
291         spin_unlock_bh(&xprt->transport_lock);
292         return retval;
293 }
294
295 static void __xprt_lock_write_next(struct rpc_xprt *xprt)
296 {
297         struct rpc_task *task;
298         struct rpc_rqst *req;
299
300         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
301                 return;
302
303         task = rpc_wake_up_next(&xprt->sending);
304         if (task == NULL)
305                 goto out_unlock;
306
307         req = task->tk_rqstp;
308         xprt->snd_task = task;
309         if (req) {
310                 req->rq_bytes_sent = 0;
311                 req->rq_ntrans++;
312         }
313         return;
314
315 out_unlock:
316         xprt_clear_locked(xprt);
317 }
318
319 static void __xprt_lock_write_next_cong(struct rpc_xprt *xprt)
320 {
321         struct rpc_task *task;
322         struct rpc_rqst *req;
323
324         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
325                 return;
326         if (RPCXPRT_CONGESTED(xprt))
327                 goto out_unlock;
328         task = rpc_wake_up_next(&xprt->sending);
329         if (task == NULL)
330                 goto out_unlock;
331
332         req = task->tk_rqstp;
333         if (req == NULL) {
334                 xprt->snd_task = task;
335                 return;
336         }
337         if (__xprt_get_cong(xprt, task)) {
338                 xprt->snd_task = task;
339                 req->rq_bytes_sent = 0;
340                 req->rq_ntrans++;
341                 return;
342         }
343 out_unlock:
344         xprt_clear_locked(xprt);
345 }
346
347 /**
348  * xprt_release_xprt - allow other requests to use a transport
349  * @xprt: transport with other tasks potentially waiting
350  * @task: task that is releasing access to the transport
351  *
352  * Note that "task" can be NULL.  No congestion control is provided.
353  */
354 void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
355 {
356         if (xprt->snd_task == task) {
357                 xprt_clear_locked(xprt);
358                 __xprt_lock_write_next(xprt);
359         }
360 }
361 EXPORT_SYMBOL_GPL(xprt_release_xprt);
362
363 /**
364  * xprt_release_xprt_cong - allow other requests to use a transport
365  * @xprt: transport with other tasks potentially waiting
366  * @task: task that is releasing access to the transport
367  *
368  * Note that "task" can be NULL.  Another task is awoken to use the
369  * transport if the transport's congestion window allows it.
370  */
371 void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
372 {
373         if (xprt->snd_task == task) {
374                 xprt_clear_locked(xprt);
375                 __xprt_lock_write_next_cong(xprt);
376         }
377 }
378 EXPORT_SYMBOL_GPL(xprt_release_xprt_cong);
379
380 static inline void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
381 {
382         spin_lock_bh(&xprt->transport_lock);
383         xprt->ops->release_xprt(xprt, task);
384         spin_unlock_bh(&xprt->transport_lock);
385 }
386
387 /*
388  * Van Jacobson congestion avoidance. Check if the congestion window
389  * overflowed. Put the task to sleep if this is the case.
390  */
391 static int
392 __xprt_get_cong(struct rpc_xprt *xprt, struct rpc_task *task)
393 {
394         struct rpc_rqst *req = task->tk_rqstp;
395
396         if (req->rq_cong)
397                 return 1;
398         dprintk("RPC: %5u xprt_cwnd_limited cong = %lu cwnd = %lu\n",
399                         task->tk_pid, xprt->cong, xprt->cwnd);
400         if (RPCXPRT_CONGESTED(xprt))
401                 return 0;
402         req->rq_cong = 1;
403         xprt->cong += RPC_CWNDSCALE;
404         return 1;
405 }
406
407 /*
408  * Adjust the congestion window, and wake up the next task
409  * that has been sleeping due to congestion
410  */
411 static void
412 __xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
413 {
414         if (!req->rq_cong)
415                 return;
416         req->rq_cong = 0;
417         xprt->cong -= RPC_CWNDSCALE;
418         __xprt_lock_write_next_cong(xprt);
419 }
420
421 /**
422  * xprt_release_rqst_cong - housekeeping when request is complete
423  * @task: RPC request that recently completed
424  *
425  * Useful for transports that require congestion control.
426  */
427 void xprt_release_rqst_cong(struct rpc_task *task)
428 {
429         __xprt_put_cong(task->tk_xprt, task->tk_rqstp);
430 }
431 EXPORT_SYMBOL_GPL(xprt_release_rqst_cong);
432
433 /**
434  * xprt_adjust_cwnd - adjust transport congestion window
435  * @task: recently completed RPC request used to adjust window
436  * @result: result code of completed RPC request
437  *
438  * We use a time-smoothed congestion estimator to avoid heavy oscillation.
439  */
440 void xprt_adjust_cwnd(struct rpc_task *task, int result)
441 {
442         struct rpc_rqst *req = task->tk_rqstp;
443         struct rpc_xprt *xprt = task->tk_xprt;
444         unsigned long cwnd = xprt->cwnd;
445
446         if (result >= 0 && cwnd <= xprt->cong) {
447                 /* The (cwnd >> 1) term makes sure
448                  * the result gets rounded properly. */
449                 cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
450                 if (cwnd > RPC_MAXCWND(xprt))
451                         cwnd = RPC_MAXCWND(xprt);
452                 __xprt_lock_write_next_cong(xprt);
453         } else if (result == -ETIMEDOUT) {
454                 cwnd >>= 1;
455                 if (cwnd < RPC_CWNDSCALE)
456                         cwnd = RPC_CWNDSCALE;
457         }
458         dprintk("RPC:       cong %ld, cwnd was %ld, now %ld\n",
459                         xprt->cong, xprt->cwnd, cwnd);
460         xprt->cwnd = cwnd;
461         __xprt_put_cong(xprt, req);
462 }
463 EXPORT_SYMBOL_GPL(xprt_adjust_cwnd);
464
465 /**
466  * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue
467  * @xprt: transport with waiting tasks
468  * @status: result code to plant in each task before waking it
469  *
470  */
471 void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status)
472 {
473         if (status < 0)
474                 rpc_wake_up_status(&xprt->pending, status);
475         else
476                 rpc_wake_up(&xprt->pending);
477 }
478 EXPORT_SYMBOL_GPL(xprt_wake_pending_tasks);
479
480 /**
481  * xprt_wait_for_buffer_space - wait for transport output buffer to clear
482  * @task: task to be put to sleep
483  * @action: function pointer to be executed after wait
484  */
485 void xprt_wait_for_buffer_space(struct rpc_task *task, rpc_action action)
486 {
487         struct rpc_rqst *req = task->tk_rqstp;
488         struct rpc_xprt *xprt = req->rq_xprt;
489
490         task->tk_timeout = req->rq_timeout;
491         rpc_sleep_on(&xprt->pending, task, action);
492 }
493 EXPORT_SYMBOL_GPL(xprt_wait_for_buffer_space);
494
495 /**
496  * xprt_write_space - wake the task waiting for transport output buffer space
497  * @xprt: transport with waiting tasks
498  *
499  * Can be called in a soft IRQ context, so xprt_write_space never sleeps.
500  */
501 void xprt_write_space(struct rpc_xprt *xprt)
502 {
503         if (unlikely(xprt->shutdown))
504                 return;
505
506         spin_lock_bh(&xprt->transport_lock);
507         if (xprt->snd_task) {
508                 dprintk("RPC:       write space: waking waiting task on "
509                                 "xprt %p\n", xprt);
510                 rpc_wake_up_queued_task(&xprt->pending, xprt->snd_task);
511         }
512         spin_unlock_bh(&xprt->transport_lock);
513 }
514 EXPORT_SYMBOL_GPL(xprt_write_space);
515
516 /**
517  * xprt_set_retrans_timeout_def - set a request's retransmit timeout
518  * @task: task whose timeout is to be set
519  *
520  * Set a request's retransmit timeout based on the transport's
521  * default timeout parameters.  Used by transports that don't adjust
522  * the retransmit timeout based on round-trip time estimation.
523  */
524 void xprt_set_retrans_timeout_def(struct rpc_task *task)
525 {
526         task->tk_timeout = task->tk_rqstp->rq_timeout;
527 }
528 EXPORT_SYMBOL_GPL(xprt_set_retrans_timeout_def);
529
530 /*
531  * xprt_set_retrans_timeout_rtt - set a request's retransmit timeout
532  * @task: task whose timeout is to be set
533  *
534  * Set a request's retransmit timeout using the RTT estimator.
535  */
536 void xprt_set_retrans_timeout_rtt(struct rpc_task *task)
537 {
538         int timer = task->tk_msg.rpc_proc->p_timer;
539         struct rpc_clnt *clnt = task->tk_client;
540         struct rpc_rtt *rtt = clnt->cl_rtt;
541         struct rpc_rqst *req = task->tk_rqstp;
542         unsigned long max_timeout = clnt->cl_timeout->to_maxval;
543
544         task->tk_timeout = rpc_calc_rto(rtt, timer);
545         task->tk_timeout <<= rpc_ntimeo(rtt, timer) + req->rq_retries;
546         if (task->tk_timeout > max_timeout || task->tk_timeout == 0)
547                 task->tk_timeout = max_timeout;
548 }
549 EXPORT_SYMBOL_GPL(xprt_set_retrans_timeout_rtt);
550
551 static void xprt_reset_majortimeo(struct rpc_rqst *req)
552 {
553         const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
554
555         req->rq_majortimeo = req->rq_timeout;
556         if (to->to_exponential)
557                 req->rq_majortimeo <<= to->to_retries;
558         else
559                 req->rq_majortimeo += to->to_increment * to->to_retries;
560         if (req->rq_majortimeo > to->to_maxval || req->rq_majortimeo == 0)
561                 req->rq_majortimeo = to->to_maxval;
562         req->rq_majortimeo += jiffies;
563 }
564
565 /**
566  * xprt_adjust_timeout - adjust timeout values for next retransmit
567  * @req: RPC request containing parameters to use for the adjustment
568  *
569  */
570 int xprt_adjust_timeout(struct rpc_rqst *req)
571 {
572         struct rpc_xprt *xprt = req->rq_xprt;
573         const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
574         int status = 0;
575
576         if (time_before(jiffies, req->rq_majortimeo)) {
577                 if (to->to_exponential)
578                         req->rq_timeout <<= 1;
579                 else
580                         req->rq_timeout += to->to_increment;
581                 if (to->to_maxval && req->rq_timeout >= to->to_maxval)
582                         req->rq_timeout = to->to_maxval;
583                 req->rq_retries++;
584         } else {
585                 req->rq_timeout = to->to_initval;
586                 req->rq_retries = 0;
587                 xprt_reset_majortimeo(req);
588                 /* Reset the RTT counters == "slow start" */
589                 spin_lock_bh(&xprt->transport_lock);
590                 rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval);
591                 spin_unlock_bh(&xprt->transport_lock);
592                 status = -ETIMEDOUT;
593         }
594
595         if (req->rq_timeout == 0) {
596                 printk(KERN_WARNING "xprt_adjust_timeout: rq_timeout = 0!\n");
597                 req->rq_timeout = 5 * HZ;
598         }
599         return status;
600 }
601
602 static void xprt_autoclose(struct work_struct *work)
603 {
604         struct rpc_xprt *xprt =
605                 container_of(work, struct rpc_xprt, task_cleanup);
606
607         xprt->ops->close(xprt);
608         clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
609         xprt_release_write(xprt, NULL);
610 }
611
612 /**
613  * xprt_disconnect_done - mark a transport as disconnected
614  * @xprt: transport to flag for disconnect
615  *
616  */
617 void xprt_disconnect_done(struct rpc_xprt *xprt)
618 {
619         dprintk("RPC:       disconnected transport %p\n", xprt);
620         spin_lock_bh(&xprt->transport_lock);
621         xprt_clear_connected(xprt);
622         xprt_wake_pending_tasks(xprt, -EAGAIN);
623         spin_unlock_bh(&xprt->transport_lock);
624 }
625 EXPORT_SYMBOL_GPL(xprt_disconnect_done);
626
627 /**
628  * xprt_force_disconnect - force a transport to disconnect
629  * @xprt: transport to disconnect
630  *
631  */
632 void xprt_force_disconnect(struct rpc_xprt *xprt)
633 {
634         /* Don't race with the test_bit() in xprt_clear_locked() */
635         spin_lock_bh(&xprt->transport_lock);
636         set_bit(XPRT_CLOSE_WAIT, &xprt->state);
637         /* Try to schedule an autoclose RPC call */
638         if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
639                 queue_work(rpciod_workqueue, &xprt->task_cleanup);
640         xprt_wake_pending_tasks(xprt, -EAGAIN);
641         spin_unlock_bh(&xprt->transport_lock);
642 }
643
644 /**
645  * xprt_conditional_disconnect - force a transport to disconnect
646  * @xprt: transport to disconnect
647  * @cookie: 'connection cookie'
648  *
649  * This attempts to break the connection if and only if 'cookie' matches
650  * the current transport 'connection cookie'. It ensures that we don't
651  * try to break the connection more than once when we need to retransmit
652  * a batch of RPC requests.
653  *
654  */
655 void xprt_conditional_disconnect(struct rpc_xprt *xprt, unsigned int cookie)
656 {
657         /* Don't race with the test_bit() in xprt_clear_locked() */
658         spin_lock_bh(&xprt->transport_lock);
659         if (cookie != xprt->connect_cookie)
660                 goto out;
661         if (test_bit(XPRT_CLOSING, &xprt->state) || !xprt_connected(xprt))
662                 goto out;
663         set_bit(XPRT_CLOSE_WAIT, &xprt->state);
664         /* Try to schedule an autoclose RPC call */
665         if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
666                 queue_work(rpciod_workqueue, &xprt->task_cleanup);
667         xprt_wake_pending_tasks(xprt, -EAGAIN);
668 out:
669         spin_unlock_bh(&xprt->transport_lock);
670 }
671
672 static void
673 xprt_init_autodisconnect(unsigned long data)
674 {
675         struct rpc_xprt *xprt = (struct rpc_xprt *)data;
676
677         spin_lock(&xprt->transport_lock);
678         if (!list_empty(&xprt->recv) || xprt->shutdown)
679                 goto out_abort;
680         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
681                 goto out_abort;
682         spin_unlock(&xprt->transport_lock);
683         set_bit(XPRT_CONNECTION_CLOSE, &xprt->state);
684         queue_work(rpciod_workqueue, &xprt->task_cleanup);
685         return;
686 out_abort:
687         spin_unlock(&xprt->transport_lock);
688 }
689
690 /**
691  * xprt_connect - schedule a transport connect operation
692  * @task: RPC task that is requesting the connect
693  *
694  */
695 void xprt_connect(struct rpc_task *task)
696 {
697         struct rpc_xprt *xprt = task->tk_xprt;
698
699         dprintk("RPC: %5u xprt_connect xprt %p %s connected\n", task->tk_pid,
700                         xprt, (xprt_connected(xprt) ? "is" : "is not"));
701
702         if (!xprt_bound(xprt)) {
703                 task->tk_status = -EAGAIN;
704                 return;
705         }
706         if (!xprt_lock_write(xprt, task))
707                 return;
708
709         if (test_and_clear_bit(XPRT_CLOSE_WAIT, &xprt->state))
710                 xprt->ops->close(xprt);
711
712         if (xprt_connected(xprt))
713                 xprt_release_write(xprt, task);
714         else {
715                 if (task->tk_rqstp)
716                         task->tk_rqstp->rq_bytes_sent = 0;
717
718                 task->tk_timeout = task->tk_rqstp->rq_timeout;
719                 rpc_sleep_on(&xprt->pending, task, xprt_connect_status);
720
721                 if (test_bit(XPRT_CLOSING, &xprt->state))
722                         return;
723                 if (xprt_test_and_set_connecting(xprt))
724                         return;
725                 xprt->stat.connect_start = jiffies;
726                 xprt->ops->connect(task);
727         }
728 }
729
730 static void xprt_connect_status(struct rpc_task *task)
731 {
732         struct rpc_xprt *xprt = task->tk_xprt;
733
734         if (task->tk_status == 0) {
735                 xprt->stat.connect_count++;
736                 xprt->stat.connect_time += (long)jiffies - xprt->stat.connect_start;
737                 dprintk("RPC: %5u xprt_connect_status: connection established\n",
738                                 task->tk_pid);
739                 return;
740         }
741
742         switch (task->tk_status) {
743         case -EAGAIN:
744                 dprintk("RPC: %5u xprt_connect_status: retrying\n", task->tk_pid);
745                 break;
746         case -ETIMEDOUT:
747                 dprintk("RPC: %5u xprt_connect_status: connect attempt timed "
748                                 "out\n", task->tk_pid);
749                 break;
750         default:
751                 dprintk("RPC: %5u xprt_connect_status: error %d connecting to "
752                                 "server %s\n", task->tk_pid, -task->tk_status,
753                                 task->tk_client->cl_server);
754                 xprt_release_write(xprt, task);
755                 task->tk_status = -EIO;
756         }
757 }
758
759 /**
760  * xprt_lookup_rqst - find an RPC request corresponding to an XID
761  * @xprt: transport on which the original request was transmitted
762  * @xid: RPC XID of incoming reply
763  *
764  */
765 struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid)
766 {
767         struct rpc_rqst *entry;
768
769         list_for_each_entry(entry, &xprt->recv, rq_list)
770                 if (entry->rq_xid == xid)
771                         return entry;
772
773         dprintk("RPC:       xprt_lookup_rqst did not find xid %08x\n",
774                         ntohl(xid));
775         xprt->stat.bad_xids++;
776         return NULL;
777 }
778 EXPORT_SYMBOL_GPL(xprt_lookup_rqst);
779
780 static void xprt_update_rtt(struct rpc_task *task)
781 {
782         struct rpc_rqst *req = task->tk_rqstp;
783         struct rpc_rtt *rtt = task->tk_client->cl_rtt;
784         unsigned timer = task->tk_msg.rpc_proc->p_timer;
785         long m = usecs_to_jiffies(ktime_to_us(req->rq_rtt));
786
787         if (timer) {
788                 if (req->rq_ntrans == 1)
789                         rpc_update_rtt(rtt, timer, m);
790                 rpc_set_timeo(rtt, timer, req->rq_ntrans - 1);
791         }
792 }
793
794 /**
795  * xprt_complete_rqst - called when reply processing is complete
796  * @task: RPC request that recently completed
797  * @copied: actual number of bytes received from the transport
798  *
799  * Caller holds transport lock.
800  */
801 void xprt_complete_rqst(struct rpc_task *task, int copied)
802 {
803         struct rpc_rqst *req = task->tk_rqstp;
804         struct rpc_xprt *xprt = req->rq_xprt;
805
806         dprintk("RPC: %5u xid %08x complete (%d bytes received)\n",
807                         task->tk_pid, ntohl(req->rq_xid), copied);
808
809         xprt->stat.recvs++;
810         req->rq_rtt = ktime_sub(ktime_get(), req->rq_xtime);
811         if (xprt->ops->timer != NULL)
812                 xprt_update_rtt(task);
813
814         list_del_init(&req->rq_list);
815         req->rq_private_buf.len = copied;
816         /* Ensure all writes are done before we update */
817         /* req->rq_reply_bytes_recvd */
818         smp_wmb();
819         req->rq_reply_bytes_recvd = copied;
820         rpc_wake_up_queued_task(&xprt->pending, task);
821 }
822 EXPORT_SYMBOL_GPL(xprt_complete_rqst);
823
824 static void xprt_timer(struct rpc_task *task)
825 {
826         struct rpc_rqst *req = task->tk_rqstp;
827         struct rpc_xprt *xprt = req->rq_xprt;
828
829         if (task->tk_status != -ETIMEDOUT)
830                 return;
831         dprintk("RPC: %5u xprt_timer\n", task->tk_pid);
832
833         spin_lock_bh(&xprt->transport_lock);
834         if (!req->rq_reply_bytes_recvd) {
835                 if (xprt->ops->timer)
836                         xprt->ops->timer(task);
837         } else
838                 task->tk_status = 0;
839         spin_unlock_bh(&xprt->transport_lock);
840 }
841
842 static inline int xprt_has_timer(struct rpc_xprt *xprt)
843 {
844         return xprt->idle_timeout != 0;
845 }
846
847 /**
848  * xprt_prepare_transmit - reserve the transport before sending a request
849  * @task: RPC task about to send a request
850  *
851  */
852 int xprt_prepare_transmit(struct rpc_task *task)
853 {
854         struct rpc_rqst *req = task->tk_rqstp;
855         struct rpc_xprt *xprt = req->rq_xprt;
856         int err = 0;
857
858         dprintk("RPC: %5u xprt_prepare_transmit\n", task->tk_pid);
859
860         spin_lock_bh(&xprt->transport_lock);
861         if (req->rq_reply_bytes_recvd && !req->rq_bytes_sent) {
862                 err = req->rq_reply_bytes_recvd;
863                 goto out_unlock;
864         }
865         if (!xprt->ops->reserve_xprt(xprt, task))
866                 err = -EAGAIN;
867 out_unlock:
868         spin_unlock_bh(&xprt->transport_lock);
869         return err;
870 }
871
872 void xprt_end_transmit(struct rpc_task *task)
873 {
874         xprt_release_write(task->tk_rqstp->rq_xprt, task);
875 }
876
877 /**
878  * xprt_transmit - send an RPC request on a transport
879  * @task: controlling RPC task
880  *
881  * We have to copy the iovec because sendmsg fiddles with its contents.
882  */
883 void xprt_transmit(struct rpc_task *task)
884 {
885         struct rpc_rqst *req = task->tk_rqstp;
886         struct rpc_xprt *xprt = req->rq_xprt;
887         int status;
888
889         dprintk("RPC: %5u xprt_transmit(%u)\n", task->tk_pid, req->rq_slen);
890
891         if (!req->rq_reply_bytes_recvd) {
892                 if (list_empty(&req->rq_list) && rpc_reply_expected(task)) {
893                         /*
894                          * Add to the list only if we're expecting a reply
895                          */
896                         spin_lock_bh(&xprt->transport_lock);
897                         /* Update the softirq receive buffer */
898                         memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
899                                         sizeof(req->rq_private_buf));
900                         /* Add request to the receive list */
901                         list_add_tail(&req->rq_list, &xprt->recv);
902                         spin_unlock_bh(&xprt->transport_lock);
903                         xprt_reset_majortimeo(req);
904                         /* Turn off autodisconnect */
905                         del_singleshot_timer_sync(&xprt->timer);
906                 }
907         } else if (!req->rq_bytes_sent)
908                 return;
909
910         req->rq_connect_cookie = xprt->connect_cookie;
911         req->rq_xtime = ktime_get();
912         status = xprt->ops->send_request(task);
913         if (status != 0) {
914                 task->tk_status = status;
915                 return;
916         }
917
918         dprintk("RPC: %5u xmit complete\n", task->tk_pid);
919         task->tk_flags |= RPC_TASK_SENT;
920         spin_lock_bh(&xprt->transport_lock);
921
922         xprt->ops->set_retrans_timeout(task);
923
924         xprt->stat.sends++;
925         xprt->stat.req_u += xprt->stat.sends - xprt->stat.recvs;
926         xprt->stat.bklog_u += xprt->backlog.qlen;
927
928         /* Don't race with disconnect */
929         if (!xprt_connected(xprt))
930                 task->tk_status = -ENOTCONN;
931         else if (!req->rq_reply_bytes_recvd && rpc_reply_expected(task)) {
932                 /*
933                  * Sleep on the pending queue since
934                  * we're expecting a reply.
935                  */
936                 rpc_sleep_on(&xprt->pending, task, xprt_timer);
937         }
938         spin_unlock_bh(&xprt->transport_lock);
939 }
940
941 static struct rpc_rqst *xprt_dynamic_alloc_slot(struct rpc_xprt *xprt, gfp_t gfp_flags)
942 {
943         struct rpc_rqst *req = ERR_PTR(-EAGAIN);
944
945         if (!atomic_add_unless(&xprt->num_reqs, 1, xprt->max_reqs))
946                 goto out;
947         req = kzalloc(sizeof(struct rpc_rqst), gfp_flags);
948         if (req != NULL)
949                 goto out;
950         atomic_dec(&xprt->num_reqs);
951         req = ERR_PTR(-ENOMEM);
952 out:
953         return req;
954 }
955
956 static bool xprt_dynamic_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
957 {
958         if (atomic_add_unless(&xprt->num_reqs, -1, xprt->min_reqs)) {
959                 kfree(req);
960                 return true;
961         }
962         return false;
963 }
964
965 void xprt_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
966 {
967         struct rpc_rqst *req;
968
969         spin_lock(&xprt->reserve_lock);
970         if (!list_empty(&xprt->free)) {
971                 req = list_entry(xprt->free.next, struct rpc_rqst, rq_list);
972                 list_del(&req->rq_list);
973                 goto out_init_req;
974         }
975         req = xprt_dynamic_alloc_slot(xprt, GFP_NOWAIT);
976         if (!IS_ERR(req))
977                 goto out_init_req;
978         switch (PTR_ERR(req)) {
979         case -ENOMEM:
980                 dprintk("RPC:       dynamic allocation of request slot "
981                                 "failed! Retrying\n");
982                 task->tk_status = -ENOMEM;
983                 break;
984         case -EAGAIN:
985                 rpc_sleep_on(&xprt->backlog, task, NULL);
986                 dprintk("RPC:       waiting for request slot\n");
987         default:
988                 task->tk_status = -EAGAIN;
989         }
990         spin_unlock(&xprt->reserve_lock);
991         return;
992 out_init_req:
993         task->tk_status = 0;
994         task->tk_rqstp = req;
995         xprt_request_init(task, xprt);
996         spin_unlock(&xprt->reserve_lock);
997 }
998 EXPORT_SYMBOL_GPL(xprt_alloc_slot);
999
1000 void xprt_lock_and_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
1001 {
1002         /* Note: grabbing the xprt_lock_write() ensures that we throttle
1003          * new slot allocation if the transport is congested (i.e. when
1004          * reconnecting a stream transport or when out of socket write
1005          * buffer space).
1006          */
1007         if (xprt_lock_write(xprt, task)) {
1008                 xprt_alloc_slot(xprt, task);
1009                 xprt_release_write(xprt, task);
1010         }
1011 }
1012 EXPORT_SYMBOL_GPL(xprt_lock_and_alloc_slot);
1013
1014 static void xprt_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
1015 {
1016         spin_lock(&xprt->reserve_lock);
1017         if (!xprt_dynamic_free_slot(xprt, req)) {
1018                 memset(req, 0, sizeof(*req));   /* mark unused */
1019                 list_add(&req->rq_list, &xprt->free);
1020         }
1021         rpc_wake_up_next(&xprt->backlog);
1022         spin_unlock(&xprt->reserve_lock);
1023 }
1024
1025 static void xprt_free_all_slots(struct rpc_xprt *xprt)
1026 {
1027         struct rpc_rqst *req;
1028         while (!list_empty(&xprt->free)) {
1029                 req = list_first_entry(&xprt->free, struct rpc_rqst, rq_list);
1030                 list_del(&req->rq_list);
1031                 kfree(req);
1032         }
1033 }
1034
1035 struct rpc_xprt *xprt_alloc(struct net *net, size_t size,
1036                 unsigned int num_prealloc,
1037                 unsigned int max_alloc)
1038 {
1039         struct rpc_xprt *xprt;
1040         struct rpc_rqst *req;
1041         int i;
1042
1043         xprt = kzalloc(size, GFP_KERNEL);
1044         if (xprt == NULL)
1045                 goto out;
1046
1047         xprt_init(xprt, net);
1048
1049         for (i = 0; i < num_prealloc; i++) {
1050                 req = kzalloc(sizeof(struct rpc_rqst), GFP_KERNEL);
1051                 if (!req)
1052                         break;
1053                 list_add(&req->rq_list, &xprt->free);
1054         }
1055         if (i < num_prealloc)
1056                 goto out_free;
1057         if (max_alloc > num_prealloc)
1058                 xprt->max_reqs = max_alloc;
1059         else
1060                 xprt->max_reqs = num_prealloc;
1061         xprt->min_reqs = num_prealloc;
1062         atomic_set(&xprt->num_reqs, num_prealloc);
1063
1064         return xprt;
1065
1066 out_free:
1067         xprt_free(xprt);
1068 out:
1069         return NULL;
1070 }
1071 EXPORT_SYMBOL_GPL(xprt_alloc);
1072
1073 void xprt_free(struct rpc_xprt *xprt)
1074 {
1075         put_net(xprt->xprt_net);
1076         xprt_free_all_slots(xprt);
1077         kfree(xprt);
1078 }
1079 EXPORT_SYMBOL_GPL(xprt_free);
1080
1081 /**
1082  * xprt_reserve - allocate an RPC request slot
1083  * @task: RPC task requesting a slot allocation
1084  *
1085  * If no more slots are available, place the task on the transport's
1086  * backlog queue.
1087  */
1088 void xprt_reserve(struct rpc_task *task)
1089 {
1090         struct rpc_xprt *xprt = task->tk_xprt;
1091
1092         task->tk_status = 0;
1093         if (task->tk_rqstp != NULL)
1094                 return;
1095
1096         task->tk_timeout = 0;
1097         task->tk_status = -EAGAIN;
1098         xprt->ops->alloc_slot(xprt, task);
1099 }
1100
1101 static inline __be32 xprt_alloc_xid(struct rpc_xprt *xprt)
1102 {
1103         return (__force __be32)xprt->xid++;
1104 }
1105
1106 static inline void xprt_init_xid(struct rpc_xprt *xprt)
1107 {
1108         xprt->xid = net_random();
1109 }
1110
1111 static void xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt)
1112 {
1113         struct rpc_rqst *req = task->tk_rqstp;
1114
1115         INIT_LIST_HEAD(&req->rq_list);
1116         req->rq_timeout = task->tk_client->cl_timeout->to_initval;
1117         req->rq_task    = task;
1118         req->rq_xprt    = xprt;
1119         req->rq_buffer  = NULL;
1120         req->rq_xid     = xprt_alloc_xid(xprt);
1121         req->rq_release_snd_buf = NULL;
1122         xprt_reset_majortimeo(req);
1123         dprintk("RPC: %5u reserved req %p xid %08x\n", task->tk_pid,
1124                         req, ntohl(req->rq_xid));
1125 }
1126
1127 /**
1128  * xprt_release - release an RPC request slot
1129  * @task: task which is finished with the slot
1130  *
1131  */
1132 void xprt_release(struct rpc_task *task)
1133 {
1134         struct rpc_xprt *xprt;
1135         struct rpc_rqst *req;
1136
1137         if (!(req = task->tk_rqstp))
1138                 return;
1139
1140         xprt = req->rq_xprt;
1141         rpc_count_iostats(task);
1142         spin_lock_bh(&xprt->transport_lock);
1143         xprt->ops->release_xprt(xprt, task);
1144         if (xprt->ops->release_request)
1145                 xprt->ops->release_request(task);
1146         if (!list_empty(&req->rq_list))
1147                 list_del(&req->rq_list);
1148         xprt->last_used = jiffies;
1149         if (list_empty(&xprt->recv) && xprt_has_timer(xprt))
1150                 mod_timer(&xprt->timer,
1151                                 xprt->last_used + xprt->idle_timeout);
1152         spin_unlock_bh(&xprt->transport_lock);
1153         if (req->rq_buffer)
1154                 xprt->ops->buf_free(req->rq_buffer);
1155         if (req->rq_cred != NULL)
1156                 put_rpccred(req->rq_cred);
1157         task->tk_rqstp = NULL;
1158         if (req->rq_release_snd_buf)
1159                 req->rq_release_snd_buf(req);
1160
1161         dprintk("RPC: %5u release request %p\n", task->tk_pid, req);
1162         if (likely(!bc_prealloc(req)))
1163                 xprt_free_slot(xprt, req);
1164         else
1165                 xprt_free_bc_request(req);
1166 }
1167
1168 static void xprt_init(struct rpc_xprt *xprt, struct net *net)
1169 {
1170         atomic_set(&xprt->count, 1);
1171
1172         spin_lock_init(&xprt->transport_lock);
1173         spin_lock_init(&xprt->reserve_lock);
1174
1175         INIT_LIST_HEAD(&xprt->free);
1176         INIT_LIST_HEAD(&xprt->recv);
1177 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
1178         spin_lock_init(&xprt->bc_pa_lock);
1179         INIT_LIST_HEAD(&xprt->bc_pa_list);
1180 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
1181
1182         xprt->last_used = jiffies;
1183         xprt->cwnd = RPC_INITCWND;
1184         xprt->bind_index = 0;
1185
1186         rpc_init_wait_queue(&xprt->binding, "xprt_binding");
1187         rpc_init_wait_queue(&xprt->pending, "xprt_pending");
1188         rpc_init_priority_wait_queue(&xprt->sending, "xprt_sending");
1189         rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");
1190
1191         xprt_init_xid(xprt);
1192
1193         xprt->xprt_net = get_net(net);
1194 }
1195
1196 /**
1197  * xprt_create_transport - create an RPC transport
1198  * @args: rpc transport creation arguments
1199  *
1200  */
1201 struct rpc_xprt *xprt_create_transport(struct xprt_create *args)
1202 {
1203         struct rpc_xprt *xprt;
1204         struct xprt_class *t;
1205
1206         spin_lock(&xprt_list_lock);
1207         list_for_each_entry(t, &xprt_list, list) {
1208                 if (t->ident == args->ident) {
1209                         spin_unlock(&xprt_list_lock);
1210                         goto found;
1211                 }
1212         }
1213         spin_unlock(&xprt_list_lock);
1214         printk(KERN_ERR "RPC: transport (%d) not supported\n", args->ident);
1215         return ERR_PTR(-EIO);
1216
1217 found:
1218         xprt = t->setup(args);
1219         if (IS_ERR(xprt)) {
1220                 dprintk("RPC:       xprt_create_transport: failed, %ld\n",
1221                                 -PTR_ERR(xprt));
1222                 goto out;
1223         }
1224         INIT_WORK(&xprt->task_cleanup, xprt_autoclose);
1225         if (xprt_has_timer(xprt))
1226                 setup_timer(&xprt->timer, xprt_init_autodisconnect,
1227                             (unsigned long)xprt);
1228         else
1229                 init_timer(&xprt->timer);
1230         dprintk("RPC:       created transport %p with %u slots\n", xprt,
1231                         xprt->max_reqs);
1232 out:
1233         return xprt;
1234 }
1235
1236 /**
1237  * xprt_destroy - destroy an RPC transport, killing off all requests.
1238  * @xprt: transport to destroy
1239  *
1240  */
1241 static void xprt_destroy(struct rpc_xprt *xprt)
1242 {
1243         dprintk("RPC:       destroying transport %p\n", xprt);
1244         xprt->shutdown = 1;
1245         del_timer_sync(&xprt->timer);
1246
1247         rpc_destroy_wait_queue(&xprt->binding);
1248         rpc_destroy_wait_queue(&xprt->pending);
1249         rpc_destroy_wait_queue(&xprt->sending);
1250         rpc_destroy_wait_queue(&xprt->backlog);
1251         cancel_work_sync(&xprt->task_cleanup);
1252         /*
1253          * Tear down transport state and free the rpc_xprt
1254          */
1255         xprt->ops->destroy(xprt);
1256 }
1257
1258 /**
1259  * xprt_put - release a reference to an RPC transport.
1260  * @xprt: pointer to the transport
1261  *
1262  */
1263 void xprt_put(struct rpc_xprt *xprt)
1264 {
1265         if (atomic_dec_and_test(&xprt->count))
1266                 xprt_destroy(xprt);
1267 }
1268
1269 /**
1270  * xprt_get - return a reference to an RPC transport.
1271  * @xprt: pointer to the transport
1272  *
1273  */
1274 struct rpc_xprt *xprt_get(struct rpc_xprt *xprt)
1275 {
1276         if (atomic_inc_not_zero(&xprt->count))
1277                 return xprt;
1278         return NULL;
1279 }