xprt.c 21.3 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12
/*
 *  linux/net/sunrpc/xprt.c
 *
 *  This is a generic RPC call interface supporting congestion avoidance,
 *  and asynchronous calls.
 *
 *  The interface works like this:
 *
 *  -	When a process places a call, it allocates a request slot if
 *	one is available. Otherwise, it sleeps on the backlog queue
 *	(xprt_reserve).
 *  -	Next, the caller puts together the RPC message, stuffs it into
13 14 15
 *	the request struct, and calls xprt_transmit().
 *  -	xprt_transmit sends the message and installs the caller on the
 *	transport's wait list. At the same time, it installs a timer that
L
Linus Torvalds 已提交
16 17
 *	is run after the packet's timeout has expired.
 *  -	When a packet arrives, the data_ready handler walks the list of
18
 *	pending requests for that transport. If a matching XID is found, the
L
Linus Torvalds 已提交
19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
 *	caller is woken up, and the timer removed.
 *  -	When no reply arrives within the timeout interval, the timer is
 *	fired by the kernel and runs xprt_timer(). It either adjusts the
 *	timeout values (minor timeout) or wakes up the caller with a status
 *	of -ETIMEDOUT.
 *  -	When the caller receives a notification from RPC that a reply arrived,
 *	it should release the RPC slot, and process the reply.
 *	If the call timed out, it may choose to retry the operation by
 *	adjusting the initial timeout value, and simply calling rpc_call
 *	again.
 *
 *  Support for async RPC is done through a set of RPC-specific scheduling
 *  primitives that `transparently' work for processes as well as async
 *  tasks that rely on callbacks.
 *
 *  Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
35 36
 *
 *  Transport switch API copyright (C) 2005, Chuck Lever <cel@netapp.com>
L
Linus Torvalds 已提交
37 38
 */

39 40
#include <linux/module.h>

L
Linus Torvalds 已提交
41
#include <linux/types.h>
42
#include <linux/interrupt.h>
L
Linus Torvalds 已提交
43 44 45
#include <linux/workqueue.h>
#include <linux/random.h>

46
#include <linux/sunrpc/clnt.h>
L
Linus Torvalds 已提交
47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67

/*
 * Local variables
 */

#ifdef RPC_DEBUG
# undef  RPC_DEBUG_DATA
# define RPCDBG_FACILITY	RPCDBG_XPRT
#endif

/*
 * Local functions
 */
static void	xprt_request_init(struct rpc_task *, struct rpc_xprt *);
static inline void	do_xprt_reserve(struct rpc_task *);
static void	xprt_connect_status(struct rpc_task *task);
static int      __xprt_get_cong(struct rpc_xprt *, struct rpc_task *);

static int	xprt_clear_backlog(struct rpc_xprt *xprt);

/*
68
 * Serialize write access to transports, in order to prevent different
L
Linus Torvalds 已提交
69
 * requests from interfering with each other.
70
 * Also prevents transport connects from colliding with writes.
L
Linus Torvalds 已提交
71 72 73 74 75 76
 */
static int
__xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
{
	struct rpc_rqst *req = task->tk_rqstp;

77
	if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
L
Linus Torvalds 已提交
78 79 80 81 82 83 84 85 86 87 88 89 90
		if (task == xprt->snd_task)
			return 1;
		goto out_sleep;
	}
	if (xprt->nocong || __xprt_get_cong(xprt, task)) {
		xprt->snd_task = task;
		if (req) {
			req->rq_bytes_sent = 0;
			req->rq_ntrans++;
		}
		return 1;
	}
	smp_mb__before_clear_bit();
91
	clear_bit(XPRT_LOCKED, &xprt->state);
L
Linus Torvalds 已提交
92 93
	smp_mb__after_clear_bit();
out_sleep:
94
	dprintk("RPC: %4d failed to lock transport %p\n", task->tk_pid, xprt);
L
Linus Torvalds 已提交
95 96 97 98 99 100 101 102 103 104 105 106 107 108
	task->tk_timeout = 0;
	task->tk_status = -EAGAIN;
	if (req && req->rq_ntrans)
		rpc_sleep_on(&xprt->resend, task, NULL, NULL);
	else
		rpc_sleep_on(&xprt->sending, task, NULL, NULL);
	return 0;
}

static inline int
xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
{
	int retval;

C
Chuck Lever 已提交
109
	spin_lock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
110
	retval = __xprt_lock_write(xprt, task);
C
Chuck Lever 已提交
111
	spin_unlock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
112 113 114 115 116 117 118 119 120
	return retval;
}


static void
__xprt_lock_write_next(struct rpc_xprt *xprt)
{
	struct rpc_task *task;

121
	if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
L
Linus Torvalds 已提交
122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141
		return;
	if (!xprt->nocong && RPCXPRT_CONGESTED(xprt))
		goto out_unlock;
	task = rpc_wake_up_next(&xprt->resend);
	if (!task) {
		task = rpc_wake_up_next(&xprt->sending);
		if (!task)
			goto out_unlock;
	}
	if (xprt->nocong || __xprt_get_cong(xprt, task)) {
		struct rpc_rqst *req = task->tk_rqstp;
		xprt->snd_task = task;
		if (req) {
			req->rq_bytes_sent = 0;
			req->rq_ntrans++;
		}
		return;
	}
out_unlock:
	smp_mb__before_clear_bit();
142
	clear_bit(XPRT_LOCKED, &xprt->state);
L
Linus Torvalds 已提交
143 144 145 146
	smp_mb__after_clear_bit();
}

/*
147
 * Releases the transport for use by other requests.
L
Linus Torvalds 已提交
148 149 150 151 152 153 154
 */
static void
__xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
{
	if (xprt->snd_task == task) {
		xprt->snd_task = NULL;
		smp_mb__before_clear_bit();
155
		clear_bit(XPRT_LOCKED, &xprt->state);
L
Linus Torvalds 已提交
156 157 158 159 160 161 162 163
		smp_mb__after_clear_bit();
		__xprt_lock_write_next(xprt);
	}
}

static inline void
xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
{
C
Chuck Lever 已提交
164
	spin_lock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
165
	__xprt_release_write(xprt, task);
C
Chuck Lever 已提交
166
	spin_unlock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229
}

/*
 * Van Jacobson congestion avoidance. Check if the congestion window
 * overflowed. Put the task to sleep if this is the case.
 */
static int
__xprt_get_cong(struct rpc_xprt *xprt, struct rpc_task *task)
{
	struct rpc_rqst *req = task->tk_rqstp;

	if (req->rq_cong)
		return 1;
	dprintk("RPC: %4d xprt_cwnd_limited cong = %ld cwnd = %ld\n",
			task->tk_pid, xprt->cong, xprt->cwnd);
	if (RPCXPRT_CONGESTED(xprt))
		return 0;
	req->rq_cong = 1;
	xprt->cong += RPC_CWNDSCALE;
	return 1;
}

/*
 * Adjust the congestion window, and wake up the next task
 * that has been sleeping due to congestion
 */
static void
__xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
{
	if (!req->rq_cong)
		return;
	req->rq_cong = 0;
	xprt->cong -= RPC_CWNDSCALE;
	__xprt_lock_write_next(xprt);
}

/*
 * Adjust RPC congestion window
 * We use a time-smoothed congestion estimator to avoid heavy oscillation.
 */
static void
xprt_adjust_cwnd(struct rpc_xprt *xprt, int result)
{
	unsigned long	cwnd;

	cwnd = xprt->cwnd;
	if (result >= 0 && cwnd <= xprt->cong) {
		/* The (cwnd >> 1) term makes sure
		 * the result gets rounded properly. */
		cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
		if (cwnd > RPC_MAXCWND(xprt))
			cwnd = RPC_MAXCWND(xprt);
		__xprt_lock_write_next(xprt);
	} else if (result == -ETIMEDOUT) {
		cwnd >>= 1;
		if (cwnd < RPC_CWNDSCALE)
			cwnd = RPC_CWNDSCALE;
	}
	dprintk("RPC:      cong %ld, cwnd was %ld, now %ld\n",
			xprt->cong, xprt->cwnd, cwnd);
	xprt->cwnd = cwnd;
}

230 231 232 233 234 235 236 237 238 239 240 241 242 243
/**
 * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue
 * @xprt: transport with waiting tasks
 * @status: result code to plant in each task before waking it
 *
 */
void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status)
{
	if (status < 0)
		rpc_wake_up_status(&xprt->pending, status);
	else
		rpc_wake_up(&xprt->pending);
}

L
Linus Torvalds 已提交
244 245 246 247 248 249 250 251 252 253 254 255 256 257
static void xprt_reset_majortimeo(struct rpc_rqst *req)
{
	struct rpc_timeout *to = &req->rq_xprt->timeout;

	req->rq_majortimeo = req->rq_timeout;
	if (to->to_exponential)
		req->rq_majortimeo <<= to->to_retries;
	else
		req->rq_majortimeo += to->to_increment * to->to_retries;
	if (req->rq_majortimeo > to->to_maxval || req->rq_majortimeo == 0)
		req->rq_majortimeo = to->to_maxval;
	req->rq_majortimeo += jiffies;
}

258 259 260 261
/**
 * xprt_adjust_timeout - adjust timeout values for next retransmit
 * @req: RPC request containing parameters to use for the adjustment
 *
L
Linus Torvalds 已提交
262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282
 */
int xprt_adjust_timeout(struct rpc_rqst *req)
{
	struct rpc_xprt *xprt = req->rq_xprt;
	struct rpc_timeout *to = &xprt->timeout;
	int status = 0;

	if (time_before(jiffies, req->rq_majortimeo)) {
		if (to->to_exponential)
			req->rq_timeout <<= 1;
		else
			req->rq_timeout += to->to_increment;
		if (to->to_maxval && req->rq_timeout >= to->to_maxval)
			req->rq_timeout = to->to_maxval;
		req->rq_retries++;
		pprintk("RPC: %lu retrans\n", jiffies);
	} else {
		req->rq_timeout = to->to_initval;
		req->rq_retries = 0;
		xprt_reset_majortimeo(req);
		/* Reset the RTT counters == "slow start" */
C
Chuck Lever 已提交
283
		spin_lock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
284
		rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval);
C
Chuck Lever 已提交
285
		spin_unlock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
286 287 288 289 290 291 292 293 294 295 296
		pprintk("RPC: %lu timeout\n", jiffies);
		status = -ETIMEDOUT;
	}

	if (req->rq_timeout == 0) {
		printk(KERN_WARNING "xprt_adjust_timeout: rq_timeout = 0!\n");
		req->rq_timeout = 5 * HZ;
	}
	return status;
}

297
static void xprt_autoclose(void *args)
L
Linus Torvalds 已提交
298 299 300 301
{
	struct rpc_xprt *xprt = (struct rpc_xprt *)args;

	xprt_disconnect(xprt);
302
	xprt->ops->close(xprt);
L
Linus Torvalds 已提交
303 304 305
	xprt_release_write(xprt, NULL);
}

306 307 308 309
/**
 * xprt_disconnect - mark a transport as disconnected
 * @xprt: transport to flag for disconnect
 *
L
Linus Torvalds 已提交
310
 */
311
void xprt_disconnect(struct rpc_xprt *xprt)
L
Linus Torvalds 已提交
312 313
{
	dprintk("RPC:      disconnected transport %p\n", xprt);
C
Chuck Lever 已提交
314
	spin_lock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
315
	xprt_clear_connected(xprt);
316
	xprt_wake_pending_tasks(xprt, -ENOTCONN);
C
Chuck Lever 已提交
317
	spin_unlock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
318 319 320 321 322 323 324
}

static void
xprt_init_autodisconnect(unsigned long data)
{
	struct rpc_xprt *xprt = (struct rpc_xprt *)data;

C
Chuck Lever 已提交
325
	spin_lock(&xprt->transport_lock);
L
Linus Torvalds 已提交
326 327
	if (!list_empty(&xprt->recv) || xprt->shutdown)
		goto out_abort;
328
	if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
L
Linus Torvalds 已提交
329
		goto out_abort;
C
Chuck Lever 已提交
330
	spin_unlock(&xprt->transport_lock);
331
	if (xprt_connecting(xprt))
L
Linus Torvalds 已提交
332 333 334 335 336
		xprt_release_write(xprt, NULL);
	else
		schedule_work(&xprt->task_cleanup);
	return;
out_abort:
C
Chuck Lever 已提交
337
	spin_unlock(&xprt->transport_lock);
L
Linus Torvalds 已提交
338 339
}

340 341 342
/**
 * xprt_connect - schedule a transport connect operation
 * @task: RPC task that is requesting the connect
L
Linus Torvalds 已提交
343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362
 *
 */
void xprt_connect(struct rpc_task *task)
{
	struct rpc_xprt	*xprt = task->tk_xprt;

	dprintk("RPC: %4d xprt_connect xprt %p %s connected\n", task->tk_pid,
			xprt, (xprt_connected(xprt) ? "is" : "is not"));

	if (xprt->shutdown) {
		task->tk_status = -EIO;
		return;
	}
	if (!xprt->addr.sin_port) {
		task->tk_status = -EIO;
		return;
	}
	if (!xprt_lock_write(xprt, task))
		return;
	if (xprt_connected(xprt))
363 364 365 366
		xprt_release_write(xprt, task);
	else {
		if (task->tk_rqstp)
			task->tk_rqstp->rq_bytes_sent = 0;
L
Linus Torvalds 已提交
367

368 369 370
		task->tk_timeout = RPC_CONNECT_TIMEOUT;
		rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL);
		xprt->ops->connect(task);
L
Linus Torvalds 已提交
371 372 373 374
	}
	return;
}

375
static void xprt_connect_status(struct rpc_task *task)
L
Linus Torvalds 已提交
376 377 378 379 380 381 382 383 384 385 386 387
{
	struct rpc_xprt	*xprt = task->tk_xprt;

	if (task->tk_status >= 0) {
		dprintk("RPC: %4d xprt_connect_status: connection established\n",
				task->tk_pid);
		return;
	}

	switch (task->tk_status) {
	case -ECONNREFUSED:
	case -ECONNRESET:
388 389 390
		dprintk("RPC: %4d xprt_connect_status: server %s refused connection\n",
				task->tk_pid, task->tk_client->cl_server);
		break;
L
Linus Torvalds 已提交
391
	case -ENOTCONN:
392 393 394
		dprintk("RPC: %4d xprt_connect_status: connection broken\n",
				task->tk_pid);
		break;
L
Linus Torvalds 已提交
395
	case -ETIMEDOUT:
396
		dprintk("RPC: %4d xprt_connect_status: connect attempt timed out\n",
L
Linus Torvalds 已提交
397 398 399
				task->tk_pid);
		break;
	default:
400 401 402 403 404 405 406 407 408 409 410
		dprintk("RPC: %4d xprt_connect_status: error %d connecting to server %s\n",
				task->tk_pid, -task->tk_status, task->tk_client->cl_server);
		xprt_release_write(xprt, task);
		task->tk_status = -EIO;
		return;
	}

	/* if soft mounted, just cause this RPC to fail */
	if (RPC_IS_SOFT(task)) {
		xprt_release_write(xprt, task);
		task->tk_status = -EIO;
L
Linus Torvalds 已提交
411 412 413
	}
}

414 415 416 417 418
/**
 * xprt_lookup_rqst - find an RPC request corresponding to an XID
 * @xprt: transport on which the original request was transmitted
 * @xid: RPC XID of incoming reply
 *
L
Linus Torvalds 已提交
419
 */
420
struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
L
Linus Torvalds 已提交
421 422 423 424 425 426 427 428 429 430 431 432 433 434
{
	struct list_head *pos;
	struct rpc_rqst	*req = NULL;

	list_for_each(pos, &xprt->recv) {
		struct rpc_rqst *entry = list_entry(pos, struct rpc_rqst, rq_list);
		if (entry->rq_xid == xid) {
			req = entry;
			break;
		}
	}
	return req;
}

435 436 437 438 439 440
/**
 * xprt_complete_rqst - called when reply processing is complete
 * @xprt: controlling transport
 * @req: RPC request that just completed
 * @copied: actual number of bytes received from the transport
 *
L
Linus Torvalds 已提交
441
 */
442
void xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied)
L
Linus Torvalds 已提交
443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496
{
	struct rpc_task	*task = req->rq_task;
	struct rpc_clnt *clnt = task->tk_client;

	/* Adjust congestion window */
	if (!xprt->nocong) {
		unsigned timer = task->tk_msg.rpc_proc->p_timer;
		xprt_adjust_cwnd(xprt, copied);
		__xprt_put_cong(xprt, req);
		if (timer) {
			if (req->rq_ntrans == 1)
				rpc_update_rtt(clnt->cl_rtt, timer,
						(long)jiffies - req->rq_xtime);
			rpc_set_timeo(clnt->cl_rtt, timer, req->rq_ntrans - 1);
		}
	}

#ifdef RPC_PROFILE
	/* Profile only reads for now */
	if (copied > 1024) {
		static unsigned long	nextstat;
		static unsigned long	pkt_rtt, pkt_len, pkt_cnt;

		pkt_cnt++;
		pkt_len += req->rq_slen + copied;
		pkt_rtt += jiffies - req->rq_xtime;
		if (time_before(nextstat, jiffies)) {
			printk("RPC: %lu %ld cwnd\n", jiffies, xprt->cwnd);
			printk("RPC: %ld %ld %ld %ld stat\n",
					jiffies, pkt_cnt, pkt_len, pkt_rtt);
			pkt_rtt = pkt_len = pkt_cnt = 0;
			nextstat = jiffies + 5 * HZ;
		}
	}
#endif

	dprintk("RPC: %4d has input (%d bytes)\n", task->tk_pid, copied);
	list_del_init(&req->rq_list);
	req->rq_received = req->rq_private_buf.len = copied;

	/* ... and wake up the process. */
	rpc_wake_up_task(task);
	return;
}

/*
 * RPC receive timeout handler.
 */
static void
xprt_timer(struct rpc_task *task)
{
	struct rpc_rqst	*req = task->tk_rqstp;
	struct rpc_xprt *xprt = req->rq_xprt;

C
Chuck Lever 已提交
497
	spin_lock(&xprt->transport_lock);
L
Linus Torvalds 已提交
498 499 500 501 502 503 504 505 506 507 508 509 510
	if (req->rq_received)
		goto out;

	xprt_adjust_cwnd(req->rq_xprt, -ETIMEDOUT);
	__xprt_put_cong(xprt, req);

	dprintk("RPC: %4d xprt_timer (%s request)\n",
		task->tk_pid, req ? "pending" : "backlogged");

	task->tk_status  = -ETIMEDOUT;
out:
	task->tk_timeout = 0;
	rpc_wake_up_task(task);
C
Chuck Lever 已提交
511
	spin_unlock(&xprt->transport_lock);
L
Linus Torvalds 已提交
512 513
}

514 515 516 517
/**
 * xprt_prepare_transmit - reserve the transport before sending a request
 * @task: RPC task about to send a request
 *
L
Linus Torvalds 已提交
518
 */
519
int xprt_prepare_transmit(struct rpc_task *task)
L
Linus Torvalds 已提交
520 521 522 523 524 525 526 527 528 529
{
	struct rpc_rqst	*req = task->tk_rqstp;
	struct rpc_xprt	*xprt = req->rq_xprt;
	int err = 0;

	dprintk("RPC: %4d xprt_prepare_transmit\n", task->tk_pid);

	if (xprt->shutdown)
		return -EIO;

C
Chuck Lever 已提交
530
	spin_lock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
531 532 533 534 535 536 537 538 539 540 541 542 543 544
	if (req->rq_received && !req->rq_bytes_sent) {
		err = req->rq_received;
		goto out_unlock;
	}
	if (!__xprt_lock_write(xprt, task)) {
		err = -EAGAIN;
		goto out_unlock;
	}

	if (!xprt_connected(xprt)) {
		err = -ENOTCONN;
		goto out_unlock;
	}
out_unlock:
C
Chuck Lever 已提交
545
	spin_unlock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
546 547 548
	return err;
}

549 550 551 552 553 554 555
/**
 * xprt_transmit - send an RPC request on a transport
 * @task: controlling RPC task
 *
 * We have to copy the iovec because sendmsg fiddles with its contents.
 */
void xprt_transmit(struct rpc_task *task)
L
Linus Torvalds 已提交
556 557 558 559
{
	struct rpc_clnt *clnt = task->tk_client;
	struct rpc_rqst	*req = task->tk_rqstp;
	struct rpc_xprt	*xprt = req->rq_xprt;
560
	int status;
L
Linus Torvalds 已提交
561 562 563 564 565 566

	dprintk("RPC: %4d xprt_transmit(%u)\n", task->tk_pid, req->rq_slen);

	smp_rmb();
	if (!req->rq_received) {
		if (list_empty(&req->rq_list)) {
C
Chuck Lever 已提交
567
			spin_lock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
568 569 570 571 572
			/* Update the softirq receive buffer */
			memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
					sizeof(req->rq_private_buf));
			/* Add request to the receive list */
			list_add_tail(&req->rq_list, &xprt->recv);
C
Chuck Lever 已提交
573
			spin_unlock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
574
			xprt_reset_majortimeo(req);
575 576
			/* Turn off autodisconnect */
			del_singleshot_timer_sync(&xprt->timer);
L
Linus Torvalds 已提交
577 578 579 580
		}
	} else if (!req->rq_bytes_sent)
		return;

581 582 583
	status = xprt->ops->send_request(task);
	if (!status)
		goto out_receive;
L
Linus Torvalds 已提交
584 585 586 587 588 589 590 591 592 593 594

	/* Note: at this point, task->tk_sleeping has not yet been set,
	 *	 hence there is no danger of the waking up task being put on
	 *	 schedq, and being picked up by a parallel run of rpciod().
	 */
	task->tk_status = status;

	switch (status) {
	case -ECONNREFUSED:
		task->tk_timeout = RPC_REESTABLISH_TIMEOUT;
		rpc_sleep_on(&xprt->sending, task, NULL, NULL);
595
	case -EAGAIN:
L
Linus Torvalds 已提交
596 597 598 599 600 601 602 603 604 605 606
	case -ENOTCONN:
		return;
	default:
		if (xprt->stream)
			xprt_disconnect(xprt);
	}
	xprt_release_write(xprt, task);
	return;
 out_receive:
	dprintk("RPC: %4d xmit complete\n", task->tk_pid);
	/* Set the task's receive timeout value */
C
Chuck Lever 已提交
607
	spin_lock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
608 609 610 611 612 613 614 615 616 617 618 619 620 621
	if (!xprt->nocong) {
		int timer = task->tk_msg.rpc_proc->p_timer;
		task->tk_timeout = rpc_calc_rto(clnt->cl_rtt, timer);
		task->tk_timeout <<= rpc_ntimeo(clnt->cl_rtt, timer) + req->rq_retries;
		if (task->tk_timeout > xprt->timeout.to_maxval || task->tk_timeout == 0)
			task->tk_timeout = xprt->timeout.to_maxval;
	} else
		task->tk_timeout = req->rq_timeout;
	/* Don't race with disconnect */
	if (!xprt_connected(xprt))
		task->tk_status = -ENOTCONN;
	else if (!req->rq_received)
		rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer);
	__xprt_release_write(xprt, task);
C
Chuck Lever 已提交
622
	spin_unlock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
623 624
}

625
static inline void do_xprt_reserve(struct rpc_task *task)
L
Linus Torvalds 已提交
626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644
{
	struct rpc_xprt	*xprt = task->tk_xprt;

	task->tk_status = 0;
	if (task->tk_rqstp)
		return;
	if (!list_empty(&xprt->free)) {
		struct rpc_rqst	*req = list_entry(xprt->free.next, struct rpc_rqst, rq_list);
		list_del_init(&req->rq_list);
		task->tk_rqstp = req;
		xprt_request_init(task, xprt);
		return;
	}
	dprintk("RPC:      waiting for request slot\n");
	task->tk_status = -EAGAIN;
	task->tk_timeout = 0;
	rpc_sleep_on(&xprt->backlog, task, NULL, NULL);
}

645 646 647 648 649 650 651 652
/**
 * xprt_reserve - allocate an RPC request slot
 * @task: RPC task requesting a slot allocation
 *
 * If no more slots are available, place the task on the transport's
 * backlog queue.
 */
void xprt_reserve(struct rpc_task *task)
L
Linus Torvalds 已提交
653 654 655 656 657
{
	struct rpc_xprt	*xprt = task->tk_xprt;

	task->tk_status = -EIO;
	if (!xprt->shutdown) {
C
Chuck Lever 已提交
658
		spin_lock(&xprt->reserve_lock);
L
Linus Torvalds 已提交
659
		do_xprt_reserve(task);
C
Chuck Lever 已提交
660
		spin_unlock(&xprt->reserve_lock);
L
Linus Torvalds 已提交
661 662 663 664 665 666 667 668 669 670 671 672 673
	}
}

static inline u32 xprt_alloc_xid(struct rpc_xprt *xprt)
{
	return xprt->xid++;
}

static inline void xprt_init_xid(struct rpc_xprt *xprt)
{
	get_random_bytes(&xprt->xid, sizeof(xprt->xid));
}

674
static void xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt)
L
Linus Torvalds 已提交
675 676 677 678 679 680 681 682 683 684 685
{
	struct rpc_rqst	*req = task->tk_rqstp;

	req->rq_timeout = xprt->timeout.to_initval;
	req->rq_task	= task;
	req->rq_xprt    = xprt;
	req->rq_xid     = xprt_alloc_xid(xprt);
	dprintk("RPC: %4d reserved req %p xid %08x\n", task->tk_pid,
			req, ntohl(req->rq_xid));
}

686 687 688 689
/**
 * xprt_release - release an RPC request slot
 * @task: task which is finished with the slot
 *
L
Linus Torvalds 已提交
690
 */
691
void xprt_release(struct rpc_task *task)
L
Linus Torvalds 已提交
692 693 694 695 696 697
{
	struct rpc_xprt	*xprt = task->tk_xprt;
	struct rpc_rqst	*req;

	if (!(req = task->tk_rqstp))
		return;
C
Chuck Lever 已提交
698
	spin_lock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
699 700 701 702 703 704
	__xprt_release_write(xprt, task);
	__xprt_put_cong(xprt, req);
	if (!list_empty(&req->rq_list))
		list_del(&req->rq_list);
	xprt->last_used = jiffies;
	if (list_empty(&xprt->recv) && !xprt->shutdown)
705 706
		mod_timer(&xprt->timer,
				xprt->last_used + RPC_IDLE_DISCONNECT_TIMEOUT);
C
Chuck Lever 已提交
707
	spin_unlock_bh(&xprt->transport_lock);
L
Linus Torvalds 已提交
708 709 710 711 712
	task->tk_rqstp = NULL;
	memset(req, 0, sizeof(*req));	/* mark unused */

	dprintk("RPC: %4d release request %p\n", task->tk_pid, req);

C
Chuck Lever 已提交
713
	spin_lock(&xprt->reserve_lock);
L
Linus Torvalds 已提交
714 715
	list_add(&req->rq_list, &xprt->free);
	xprt_clear_backlog(xprt);
C
Chuck Lever 已提交
716
	spin_unlock(&xprt->reserve_lock);
L
Linus Torvalds 已提交
717 718
}

719 720 721 722 723 724
/**
 * xprt_set_timeout - set constant RPC timeout
 * @to: RPC timeout parameters to set up
 * @retr: number of retries
 * @incr: amount of increase after each retry
 *
L
Linus Torvalds 已提交
725
 */
726
void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr)
L
Linus Torvalds 已提交
727 728 729
{
	to->to_initval   = 
	to->to_increment = incr;
730
	to->to_maxval    = to->to_initval + (incr * retr);
L
Linus Torvalds 已提交
731 732 733 734
	to->to_retries   = retr;
	to->to_exponential = 0;
}

735
static struct rpc_xprt *xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to)
L
Linus Torvalds 已提交
736
{
737
	int result;
L
Linus Torvalds 已提交
738 739 740 741 742 743 744 745
	struct rpc_xprt	*xprt;
	struct rpc_rqst	*req;

	if ((xprt = kmalloc(sizeof(struct rpc_xprt), GFP_KERNEL)) == NULL)
		return ERR_PTR(-ENOMEM);
	memset(xprt, 0, sizeof(*xprt)); /* Nnnngh! */

	xprt->addr = *ap;
746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762

	switch (proto) {
	case IPPROTO_UDP:
		result = xs_setup_udp(xprt, to);
		break;
	case IPPROTO_TCP:
		result = xs_setup_tcp(xprt, to);
		break;
	default:
		printk(KERN_ERR "RPC: unrecognized transport protocol: %d\n",
				proto);
		result = -EIO;
		break;
	}
	if (result) {
		kfree(xprt);
		return ERR_PTR(result);
L
Linus Torvalds 已提交
763
	}
764

C
Chuck Lever 已提交
765
	spin_lock_init(&xprt->transport_lock);
C
Chuck Lever 已提交
766
	spin_lock_init(&xprt->reserve_lock);
L
Linus Torvalds 已提交
767 768 769 770
	init_waitqueue_head(&xprt->cong_wait);

	INIT_LIST_HEAD(&xprt->free);
	INIT_LIST_HEAD(&xprt->recv);
771
	INIT_WORK(&xprt->task_cleanup, xprt_autoclose, xprt);
L
Linus Torvalds 已提交
772 773 774 775 776 777 778 779 780 781 782
	init_timer(&xprt->timer);
	xprt->timer.function = xprt_init_autodisconnect;
	xprt->timer.data = (unsigned long) xprt;
	xprt->last_used = jiffies;

	rpc_init_wait_queue(&xprt->pending, "xprt_pending");
	rpc_init_wait_queue(&xprt->sending, "xprt_sending");
	rpc_init_wait_queue(&xprt->resend, "xprt_resend");
	rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");

	/* initialize free list */
783
	for (req = &xprt->slot[xprt->max_reqs-1]; req >= &xprt->slot[0]; req--)
L
Linus Torvalds 已提交
784 785 786 787 788 789 790 791 792 793
		list_add(&req->rq_list, &xprt->free);

	xprt_init_xid(xprt);

	dprintk("RPC:      created transport %p with %u slots\n", xprt,
			xprt->max_reqs);
	
	return xprt;
}

794 795 796 797 798 799
/**
 * xprt_create_proto - create an RPC client transport
 * @proto: requested transport protocol
 * @sap: remote peer's address
 * @to: timeout parameters for new transport
 *
L
Linus Torvalds 已提交
800
 */
801
struct rpc_xprt *xprt_create_proto(int proto, struct sockaddr_in *sap, struct rpc_timeout *to)
L
Linus Torvalds 已提交
802 803 804 805 806 807 808 809 810 811 812
{
	struct rpc_xprt	*xprt;

	xprt = xprt_setup(proto, sap, to);
	if (IS_ERR(xprt))
		dprintk("RPC:      xprt_create_proto failed\n");
	else
		dprintk("RPC:      xprt_create_proto created xprt %p\n", xprt);
	return xprt;
}

813
static void xprt_shutdown(struct rpc_xprt *xprt)
L
Linus Torvalds 已提交
814 815 816 817
{
	xprt->shutdown = 1;
	rpc_wake_up(&xprt->sending);
	rpc_wake_up(&xprt->resend);
818
	xprt_wake_pending_tasks(xprt, -EIO);
L
Linus Torvalds 已提交
819 820 821 822 823
	rpc_wake_up(&xprt->backlog);
	wake_up(&xprt->cong_wait);
	del_timer_sync(&xprt->timer);
}

824
static int xprt_clear_backlog(struct rpc_xprt *xprt) {
L
Linus Torvalds 已提交
825 826 827 828 829
	rpc_wake_up_next(&xprt->backlog);
	wake_up(&xprt->cong_wait);
	return 1;
}

830 831 832 833
/**
 * xprt_destroy - destroy an RPC transport, killing off all requests.
 * @xprt: transport to destroy
 *
L
Linus Torvalds 已提交
834
 */
835
int xprt_destroy(struct rpc_xprt *xprt)
L
Linus Torvalds 已提交
836 837 838
{
	dprintk("RPC:      destroying transport %p\n", xprt);
	xprt_shutdown(xprt);
839
	xprt->ops->destroy(xprt);
L
Linus Torvalds 已提交
840 841 842 843
	kfree(xprt);

	return 0;
}