xref: /openbmc/linux/net/sunrpc/svc.c (revision 8684014d)
1 /*
2  * linux/net/sunrpc/svc.c
3  *
4  * High-level RPC service routines
5  *
6  * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
7  *
8  * Multiple threads pools and NUMAisation
9  * Copyright (c) 2006 Silicon Graphics, Inc.
10  * by Greg Banks <gnb@melbourne.sgi.com>
11  */
12 
13 #include <linux/linkage.h>
14 #include <linux/sched.h>
15 #include <linux/errno.h>
16 #include <linux/net.h>
17 #include <linux/in.h>
18 #include <linux/mm.h>
19 #include <linux/interrupt.h>
20 #include <linux/module.h>
21 #include <linux/kthread.h>
22 #include <linux/slab.h>
23 
24 #include <linux/sunrpc/types.h>
25 #include <linux/sunrpc/xdr.h>
26 #include <linux/sunrpc/stats.h>
27 #include <linux/sunrpc/svcsock.h>
28 #include <linux/sunrpc/clnt.h>
29 #include <linux/sunrpc/bc_xprt.h>
30 
31 #include <trace/events/sunrpc.h>
32 
33 #define RPCDBG_FACILITY	RPCDBG_SVCDSP
34 
35 static void svc_unregister(const struct svc_serv *serv, struct net *net);
36 
37 #define svc_serv_is_pooled(serv)    ((serv)->sv_function)
38 
39 /*
40  * Mode for mapping cpus to pools.
41  */
42 enum {
43 	SVC_POOL_AUTO = -1,	/* choose one of the others */
44 	SVC_POOL_GLOBAL,	/* no mapping, just a single global pool
45 				 * (legacy & UP mode) */
46 	SVC_POOL_PERCPU,	/* one pool per cpu */
47 	SVC_POOL_PERNODE	/* one pool per numa node */
48 };
49 #define SVC_POOL_DEFAULT	SVC_POOL_GLOBAL
50 
51 /*
52  * Structure for mapping cpus to pools and vice versa.
53  * Setup once during sunrpc initialisation.
54  */
55 static struct svc_pool_map {
56 	int count;			/* How many svc_servs use us */
57 	int mode;			/* Note: int not enum to avoid
58 					 * warnings about "enumeration value
59 					 * not handled in switch" */
60 	unsigned int npools;
61 	unsigned int *pool_to;		/* maps pool id to cpu or node */
62 	unsigned int *to_pool;		/* maps cpu or node to pool id */
63 } svc_pool_map = {
64 	.count = 0,
65 	.mode = SVC_POOL_DEFAULT
66 };
67 static DEFINE_MUTEX(svc_pool_map_mutex);/* protects svc_pool_map.count only */
68 
69 static int
70 param_set_pool_mode(const char *val, struct kernel_param *kp)
71 {
72 	int *ip = (int *)kp->arg;
73 	struct svc_pool_map *m = &svc_pool_map;
74 	int err;
75 
76 	mutex_lock(&svc_pool_map_mutex);
77 
78 	err = -EBUSY;
79 	if (m->count)
80 		goto out;
81 
82 	err = 0;
83 	if (!strncmp(val, "auto", 4))
84 		*ip = SVC_POOL_AUTO;
85 	else if (!strncmp(val, "global", 6))
86 		*ip = SVC_POOL_GLOBAL;
87 	else if (!strncmp(val, "percpu", 6))
88 		*ip = SVC_POOL_PERCPU;
89 	else if (!strncmp(val, "pernode", 7))
90 		*ip = SVC_POOL_PERNODE;
91 	else
92 		err = -EINVAL;
93 
94 out:
95 	mutex_unlock(&svc_pool_map_mutex);
96 	return err;
97 }
98 
99 static int
100 param_get_pool_mode(char *buf, struct kernel_param *kp)
101 {
102 	int *ip = (int *)kp->arg;
103 
104 	switch (*ip)
105 	{
106 	case SVC_POOL_AUTO:
107 		return strlcpy(buf, "auto", 20);
108 	case SVC_POOL_GLOBAL:
109 		return strlcpy(buf, "global", 20);
110 	case SVC_POOL_PERCPU:
111 		return strlcpy(buf, "percpu", 20);
112 	case SVC_POOL_PERNODE:
113 		return strlcpy(buf, "pernode", 20);
114 	default:
115 		return sprintf(buf, "%d", *ip);
116 	}
117 }
118 
119 module_param_call(pool_mode, param_set_pool_mode, param_get_pool_mode,
120 		 &svc_pool_map.mode, 0644);
121 
122 /*
123  * Detect best pool mapping mode heuristically,
124  * according to the machine's topology.
125  */
126 static int
127 svc_pool_map_choose_mode(void)
128 {
129 	unsigned int node;
130 
131 	if (nr_online_nodes > 1) {
132 		/*
133 		 * Actually have multiple NUMA nodes,
134 		 * so split pools on NUMA node boundaries
135 		 */
136 		return SVC_POOL_PERNODE;
137 	}
138 
139 	node = first_online_node;
140 	if (nr_cpus_node(node) > 2) {
141 		/*
142 		 * Non-trivial SMP, or CONFIG_NUMA on
143 		 * non-NUMA hardware, e.g. with a generic
144 		 * x86_64 kernel on Xeons.  In this case we
145 		 * want to divide the pools on cpu boundaries.
146 		 */
147 		return SVC_POOL_PERCPU;
148 	}
149 
150 	/* default: one global pool */
151 	return SVC_POOL_GLOBAL;
152 }
153 
154 /*
155  * Allocate the to_pool[] and pool_to[] arrays.
156  * Returns 0 on success or an errno.
157  */
158 static int
159 svc_pool_map_alloc_arrays(struct svc_pool_map *m, unsigned int maxpools)
160 {
161 	m->to_pool = kcalloc(maxpools, sizeof(unsigned int), GFP_KERNEL);
162 	if (!m->to_pool)
163 		goto fail;
164 	m->pool_to = kcalloc(maxpools, sizeof(unsigned int), GFP_KERNEL);
165 	if (!m->pool_to)
166 		goto fail_free;
167 
168 	return 0;
169 
170 fail_free:
171 	kfree(m->to_pool);
172 	m->to_pool = NULL;
173 fail:
174 	return -ENOMEM;
175 }
176 
177 /*
178  * Initialise the pool map for SVC_POOL_PERCPU mode.
179  * Returns number of pools or <0 on error.
180  */
181 static int
182 svc_pool_map_init_percpu(struct svc_pool_map *m)
183 {
184 	unsigned int maxpools = nr_cpu_ids;
185 	unsigned int pidx = 0;
186 	unsigned int cpu;
187 	int err;
188 
189 	err = svc_pool_map_alloc_arrays(m, maxpools);
190 	if (err)
191 		return err;
192 
193 	for_each_online_cpu(cpu) {
194 		BUG_ON(pidx >= maxpools);
195 		m->to_pool[cpu] = pidx;
196 		m->pool_to[pidx] = cpu;
197 		pidx++;
198 	}
199 	/* cpus brought online later all get mapped to pool0, sorry */
200 
201 	return pidx;
202 };
203 
204 
205 /*
206  * Initialise the pool map for SVC_POOL_PERNODE mode.
207  * Returns number of pools or <0 on error.
208  */
209 static int
210 svc_pool_map_init_pernode(struct svc_pool_map *m)
211 {
212 	unsigned int maxpools = nr_node_ids;
213 	unsigned int pidx = 0;
214 	unsigned int node;
215 	int err;
216 
217 	err = svc_pool_map_alloc_arrays(m, maxpools);
218 	if (err)
219 		return err;
220 
221 	for_each_node_with_cpus(node) {
222 		/* some architectures (e.g. SN2) have cpuless nodes */
223 		BUG_ON(pidx > maxpools);
224 		m->to_pool[node] = pidx;
225 		m->pool_to[pidx] = node;
226 		pidx++;
227 	}
228 	/* nodes brought online later all get mapped to pool0, sorry */
229 
230 	return pidx;
231 }
232 
233 
234 /*
235  * Add a reference to the global map of cpus to pools (and
236  * vice versa).  Initialise the map if we're the first user.
237  * Returns the number of pools.
238  */
239 static unsigned int
240 svc_pool_map_get(void)
241 {
242 	struct svc_pool_map *m = &svc_pool_map;
243 	int npools = -1;
244 
245 	mutex_lock(&svc_pool_map_mutex);
246 
247 	if (m->count++) {
248 		mutex_unlock(&svc_pool_map_mutex);
249 		return m->npools;
250 	}
251 
252 	if (m->mode == SVC_POOL_AUTO)
253 		m->mode = svc_pool_map_choose_mode();
254 
255 	switch (m->mode) {
256 	case SVC_POOL_PERCPU:
257 		npools = svc_pool_map_init_percpu(m);
258 		break;
259 	case SVC_POOL_PERNODE:
260 		npools = svc_pool_map_init_pernode(m);
261 		break;
262 	}
263 
264 	if (npools < 0) {
265 		/* default, or memory allocation failure */
266 		npools = 1;
267 		m->mode = SVC_POOL_GLOBAL;
268 	}
269 	m->npools = npools;
270 
271 	mutex_unlock(&svc_pool_map_mutex);
272 	return m->npools;
273 }
274 
275 
276 /*
277  * Drop a reference to the global map of cpus to pools.
278  * When the last reference is dropped, the map data is
279  * freed; this allows the sysadmin to change the pool
280  * mode using the pool_mode module option without
281  * rebooting or re-loading sunrpc.ko.
282  */
283 static void
284 svc_pool_map_put(void)
285 {
286 	struct svc_pool_map *m = &svc_pool_map;
287 
288 	mutex_lock(&svc_pool_map_mutex);
289 
290 	if (!--m->count) {
291 		kfree(m->to_pool);
292 		m->to_pool = NULL;
293 		kfree(m->pool_to);
294 		m->pool_to = NULL;
295 		m->npools = 0;
296 	}
297 
298 	mutex_unlock(&svc_pool_map_mutex);
299 }
300 
301 
302 static int svc_pool_map_get_node(unsigned int pidx)
303 {
304 	const struct svc_pool_map *m = &svc_pool_map;
305 
306 	if (m->count) {
307 		if (m->mode == SVC_POOL_PERCPU)
308 			return cpu_to_node(m->pool_to[pidx]);
309 		if (m->mode == SVC_POOL_PERNODE)
310 			return m->pool_to[pidx];
311 	}
312 	return NUMA_NO_NODE;
313 }
314 /*
315  * Set the given thread's cpus_allowed mask so that it
316  * will only run on cpus in the given pool.
317  */
318 static inline void
319 svc_pool_map_set_cpumask(struct task_struct *task, unsigned int pidx)
320 {
321 	struct svc_pool_map *m = &svc_pool_map;
322 	unsigned int node = m->pool_to[pidx];
323 
324 	/*
325 	 * The caller checks for sv_nrpools > 1, which
326 	 * implies that we've been initialized.
327 	 */
328 	WARN_ON_ONCE(m->count == 0);
329 	if (m->count == 0)
330 		return;
331 
332 	switch (m->mode) {
333 	case SVC_POOL_PERCPU:
334 	{
335 		set_cpus_allowed_ptr(task, cpumask_of(node));
336 		break;
337 	}
338 	case SVC_POOL_PERNODE:
339 	{
340 		set_cpus_allowed_ptr(task, cpumask_of_node(node));
341 		break;
342 	}
343 	}
344 }
345 
346 /*
347  * Use the mapping mode to choose a pool for a given CPU.
348  * Used when enqueueing an incoming RPC.  Always returns
349  * a non-NULL pool pointer.
350  */
351 struct svc_pool *
352 svc_pool_for_cpu(struct svc_serv *serv, int cpu)
353 {
354 	struct svc_pool_map *m = &svc_pool_map;
355 	unsigned int pidx = 0;
356 
357 	/*
358 	 * An uninitialised map happens in a pure client when
359 	 * lockd is brought up, so silently treat it the
360 	 * same as SVC_POOL_GLOBAL.
361 	 */
362 	if (svc_serv_is_pooled(serv)) {
363 		switch (m->mode) {
364 		case SVC_POOL_PERCPU:
365 			pidx = m->to_pool[cpu];
366 			break;
367 		case SVC_POOL_PERNODE:
368 			pidx = m->to_pool[cpu_to_node(cpu)];
369 			break;
370 		}
371 	}
372 	return &serv->sv_pools[pidx % serv->sv_nrpools];
373 }
374 
375 int svc_rpcb_setup(struct svc_serv *serv, struct net *net)
376 {
377 	int err;
378 
379 	err = rpcb_create_local(net);
380 	if (err)
381 		return err;
382 
383 	/* Remove any stale portmap registrations */
384 	svc_unregister(serv, net);
385 	return 0;
386 }
387 EXPORT_SYMBOL_GPL(svc_rpcb_setup);
388 
389 void svc_rpcb_cleanup(struct svc_serv *serv, struct net *net)
390 {
391 	svc_unregister(serv, net);
392 	rpcb_put_local(net);
393 }
394 EXPORT_SYMBOL_GPL(svc_rpcb_cleanup);
395 
396 static int svc_uses_rpcbind(struct svc_serv *serv)
397 {
398 	struct svc_program	*progp;
399 	unsigned int		i;
400 
401 	for (progp = serv->sv_program; progp; progp = progp->pg_next) {
402 		for (i = 0; i < progp->pg_nvers; i++) {
403 			if (progp->pg_vers[i] == NULL)
404 				continue;
405 			if (progp->pg_vers[i]->vs_hidden == 0)
406 				return 1;
407 		}
408 	}
409 
410 	return 0;
411 }
412 
413 int svc_bind(struct svc_serv *serv, struct net *net)
414 {
415 	if (!svc_uses_rpcbind(serv))
416 		return 0;
417 	return svc_rpcb_setup(serv, net);
418 }
419 EXPORT_SYMBOL_GPL(svc_bind);
420 
421 /*
422  * Create an RPC service
423  */
424 static struct svc_serv *
425 __svc_create(struct svc_program *prog, unsigned int bufsize, int npools,
426 	     void (*shutdown)(struct svc_serv *serv, struct net *net))
427 {
428 	struct svc_serv	*serv;
429 	unsigned int vers;
430 	unsigned int xdrsize;
431 	unsigned int i;
432 
433 	if (!(serv = kzalloc(sizeof(*serv), GFP_KERNEL)))
434 		return NULL;
435 	serv->sv_name      = prog->pg_name;
436 	serv->sv_program   = prog;
437 	serv->sv_nrthreads = 1;
438 	serv->sv_stats     = prog->pg_stats;
439 	if (bufsize > RPCSVC_MAXPAYLOAD)
440 		bufsize = RPCSVC_MAXPAYLOAD;
441 	serv->sv_max_payload = bufsize? bufsize : 4096;
442 	serv->sv_max_mesg  = roundup(serv->sv_max_payload + PAGE_SIZE, PAGE_SIZE);
443 	serv->sv_shutdown  = shutdown;
444 	xdrsize = 0;
445 	while (prog) {
446 		prog->pg_lovers = prog->pg_nvers-1;
447 		for (vers=0; vers<prog->pg_nvers ; vers++)
448 			if (prog->pg_vers[vers]) {
449 				prog->pg_hivers = vers;
450 				if (prog->pg_lovers > vers)
451 					prog->pg_lovers = vers;
452 				if (prog->pg_vers[vers]->vs_xdrsize > xdrsize)
453 					xdrsize = prog->pg_vers[vers]->vs_xdrsize;
454 			}
455 		prog = prog->pg_next;
456 	}
457 	serv->sv_xdrsize   = xdrsize;
458 	INIT_LIST_HEAD(&serv->sv_tempsocks);
459 	INIT_LIST_HEAD(&serv->sv_permsocks);
460 	init_timer(&serv->sv_temptimer);
461 	spin_lock_init(&serv->sv_lock);
462 
463 	serv->sv_nrpools = npools;
464 	serv->sv_pools =
465 		kcalloc(serv->sv_nrpools, sizeof(struct svc_pool),
466 			GFP_KERNEL);
467 	if (!serv->sv_pools) {
468 		kfree(serv);
469 		return NULL;
470 	}
471 
472 	for (i = 0; i < serv->sv_nrpools; i++) {
473 		struct svc_pool *pool = &serv->sv_pools[i];
474 
475 		dprintk("svc: initialising pool %u for %s\n",
476 				i, serv->sv_name);
477 
478 		pool->sp_id = i;
479 		INIT_LIST_HEAD(&pool->sp_sockets);
480 		INIT_LIST_HEAD(&pool->sp_all_threads);
481 		spin_lock_init(&pool->sp_lock);
482 	}
483 
484 	return serv;
485 }
486 
487 struct svc_serv *
488 svc_create(struct svc_program *prog, unsigned int bufsize,
489 	   void (*shutdown)(struct svc_serv *serv, struct net *net))
490 {
491 	return __svc_create(prog, bufsize, /*npools*/1, shutdown);
492 }
493 EXPORT_SYMBOL_GPL(svc_create);
494 
495 struct svc_serv *
496 svc_create_pooled(struct svc_program *prog, unsigned int bufsize,
497 		  void (*shutdown)(struct svc_serv *serv, struct net *net),
498 		  svc_thread_fn func, struct module *mod)
499 {
500 	struct svc_serv *serv;
501 	unsigned int npools = svc_pool_map_get();
502 
503 	serv = __svc_create(prog, bufsize, npools, shutdown);
504 	if (!serv)
505 		goto out_err;
506 
507 	serv->sv_function = func;
508 	serv->sv_module = mod;
509 	return serv;
510 out_err:
511 	svc_pool_map_put();
512 	return NULL;
513 }
514 EXPORT_SYMBOL_GPL(svc_create_pooled);
515 
516 void svc_shutdown_net(struct svc_serv *serv, struct net *net)
517 {
518 	svc_close_net(serv, net);
519 
520 	if (serv->sv_shutdown)
521 		serv->sv_shutdown(serv, net);
522 }
523 EXPORT_SYMBOL_GPL(svc_shutdown_net);
524 
525 /*
526  * Destroy an RPC service. Should be called with appropriate locking to
527  * protect the sv_nrthreads, sv_permsocks and sv_tempsocks.
528  */
529 void
530 svc_destroy(struct svc_serv *serv)
531 {
532 	dprintk("svc: svc_destroy(%s, %d)\n",
533 				serv->sv_program->pg_name,
534 				serv->sv_nrthreads);
535 
536 	if (serv->sv_nrthreads) {
537 		if (--(serv->sv_nrthreads) != 0) {
538 			svc_sock_update_bufs(serv);
539 			return;
540 		}
541 	} else
542 		printk("svc_destroy: no threads for serv=%p!\n", serv);
543 
544 	del_timer_sync(&serv->sv_temptimer);
545 
546 	/*
547 	 * The last user is gone and thus all sockets have to be destroyed to
548 	 * the point. Check this.
549 	 */
550 	BUG_ON(!list_empty(&serv->sv_permsocks));
551 	BUG_ON(!list_empty(&serv->sv_tempsocks));
552 
553 	cache_clean_deferred(serv);
554 
555 	if (svc_serv_is_pooled(serv))
556 		svc_pool_map_put();
557 
558 	kfree(serv->sv_pools);
559 	kfree(serv);
560 }
561 EXPORT_SYMBOL_GPL(svc_destroy);
562 
563 /*
564  * Allocate an RPC server's buffer space.
565  * We allocate pages and place them in rq_argpages.
566  */
567 static int
568 svc_init_buffer(struct svc_rqst *rqstp, unsigned int size, int node)
569 {
570 	unsigned int pages, arghi;
571 
572 	/* bc_xprt uses fore channel allocated buffers */
573 	if (svc_is_backchannel(rqstp))
574 		return 1;
575 
576 	pages = size / PAGE_SIZE + 1; /* extra page as we hold both request and reply.
577 				       * We assume one is at most one page
578 				       */
579 	arghi = 0;
580 	WARN_ON_ONCE(pages > RPCSVC_MAXPAGES);
581 	if (pages > RPCSVC_MAXPAGES)
582 		pages = RPCSVC_MAXPAGES;
583 	while (pages) {
584 		struct page *p = alloc_pages_node(node, GFP_KERNEL, 0);
585 		if (!p)
586 			break;
587 		rqstp->rq_pages[arghi++] = p;
588 		pages--;
589 	}
590 	return pages == 0;
591 }
592 
593 /*
594  * Release an RPC server buffer
595  */
596 static void
597 svc_release_buffer(struct svc_rqst *rqstp)
598 {
599 	unsigned int i;
600 
601 	for (i = 0; i < ARRAY_SIZE(rqstp->rq_pages); i++)
602 		if (rqstp->rq_pages[i])
603 			put_page(rqstp->rq_pages[i]);
604 }
605 
606 struct svc_rqst *
607 svc_prepare_thread(struct svc_serv *serv, struct svc_pool *pool, int node)
608 {
609 	struct svc_rqst	*rqstp;
610 
611 	rqstp = kzalloc_node(sizeof(*rqstp), GFP_KERNEL, node);
612 	if (!rqstp)
613 		goto out_enomem;
614 
615 	serv->sv_nrthreads++;
616 	__set_bit(RQ_BUSY, &rqstp->rq_flags);
617 	spin_lock_init(&rqstp->rq_lock);
618 	rqstp->rq_server = serv;
619 	rqstp->rq_pool = pool;
620 	spin_lock_bh(&pool->sp_lock);
621 	pool->sp_nrthreads++;
622 	list_add_rcu(&rqstp->rq_all, &pool->sp_all_threads);
623 	spin_unlock_bh(&pool->sp_lock);
624 
625 	rqstp->rq_argp = kmalloc_node(serv->sv_xdrsize, GFP_KERNEL, node);
626 	if (!rqstp->rq_argp)
627 		goto out_thread;
628 
629 	rqstp->rq_resp = kmalloc_node(serv->sv_xdrsize, GFP_KERNEL, node);
630 	if (!rqstp->rq_resp)
631 		goto out_thread;
632 
633 	if (!svc_init_buffer(rqstp, serv->sv_max_mesg, node))
634 		goto out_thread;
635 
636 	return rqstp;
637 out_thread:
638 	svc_exit_thread(rqstp);
639 out_enomem:
640 	return ERR_PTR(-ENOMEM);
641 }
642 EXPORT_SYMBOL_GPL(svc_prepare_thread);
643 
644 /*
645  * Choose a pool in which to create a new thread, for svc_set_num_threads
646  */
647 static inline struct svc_pool *
648 choose_pool(struct svc_serv *serv, struct svc_pool *pool, unsigned int *state)
649 {
650 	if (pool != NULL)
651 		return pool;
652 
653 	return &serv->sv_pools[(*state)++ % serv->sv_nrpools];
654 }
655 
656 /*
657  * Choose a thread to kill, for svc_set_num_threads
658  */
659 static inline struct task_struct *
660 choose_victim(struct svc_serv *serv, struct svc_pool *pool, unsigned int *state)
661 {
662 	unsigned int i;
663 	struct task_struct *task = NULL;
664 
665 	if (pool != NULL) {
666 		spin_lock_bh(&pool->sp_lock);
667 	} else {
668 		/* choose a pool in round-robin fashion */
669 		for (i = 0; i < serv->sv_nrpools; i++) {
670 			pool = &serv->sv_pools[--(*state) % serv->sv_nrpools];
671 			spin_lock_bh(&pool->sp_lock);
672 			if (!list_empty(&pool->sp_all_threads))
673 				goto found_pool;
674 			spin_unlock_bh(&pool->sp_lock);
675 		}
676 		return NULL;
677 	}
678 
679 found_pool:
680 	if (!list_empty(&pool->sp_all_threads)) {
681 		struct svc_rqst *rqstp;
682 
683 		/*
684 		 * Remove from the pool->sp_all_threads list
685 		 * so we don't try to kill it again.
686 		 */
687 		rqstp = list_entry(pool->sp_all_threads.next, struct svc_rqst, rq_all);
688 		set_bit(RQ_VICTIM, &rqstp->rq_flags);
689 		list_del_rcu(&rqstp->rq_all);
690 		task = rqstp->rq_task;
691 	}
692 	spin_unlock_bh(&pool->sp_lock);
693 
694 	return task;
695 }
696 
697 /*
698  * Create or destroy enough new threads to make the number
699  * of threads the given number.  If `pool' is non-NULL, applies
700  * only to threads in that pool, otherwise round-robins between
701  * all pools.  Caller must ensure that mutual exclusion between this and
702  * server startup or shutdown.
703  *
704  * Destroying threads relies on the service threads filling in
705  * rqstp->rq_task, which only the nfs ones do.  Assumes the serv
706  * has been created using svc_create_pooled().
707  *
708  * Based on code that used to be in nfsd_svc() but tweaked
709  * to be pool-aware.
710  */
711 int
712 svc_set_num_threads(struct svc_serv *serv, struct svc_pool *pool, int nrservs)
713 {
714 	struct svc_rqst	*rqstp;
715 	struct task_struct *task;
716 	struct svc_pool *chosen_pool;
717 	int error = 0;
718 	unsigned int state = serv->sv_nrthreads-1;
719 	int node;
720 
721 	if (pool == NULL) {
722 		/* The -1 assumes caller has done a svc_get() */
723 		nrservs -= (serv->sv_nrthreads-1);
724 	} else {
725 		spin_lock_bh(&pool->sp_lock);
726 		nrservs -= pool->sp_nrthreads;
727 		spin_unlock_bh(&pool->sp_lock);
728 	}
729 
730 	/* create new threads */
731 	while (nrservs > 0) {
732 		nrservs--;
733 		chosen_pool = choose_pool(serv, pool, &state);
734 
735 		node = svc_pool_map_get_node(chosen_pool->sp_id);
736 		rqstp = svc_prepare_thread(serv, chosen_pool, node);
737 		if (IS_ERR(rqstp)) {
738 			error = PTR_ERR(rqstp);
739 			break;
740 		}
741 
742 		__module_get(serv->sv_module);
743 		task = kthread_create_on_node(serv->sv_function, rqstp,
744 					      node, "%s", serv->sv_name);
745 		if (IS_ERR(task)) {
746 			error = PTR_ERR(task);
747 			module_put(serv->sv_module);
748 			svc_exit_thread(rqstp);
749 			break;
750 		}
751 
752 		rqstp->rq_task = task;
753 		if (serv->sv_nrpools > 1)
754 			svc_pool_map_set_cpumask(task, chosen_pool->sp_id);
755 
756 		svc_sock_update_bufs(serv);
757 		wake_up_process(task);
758 	}
759 	/* destroy old threads */
760 	while (nrservs < 0 &&
761 	       (task = choose_victim(serv, pool, &state)) != NULL) {
762 		send_sig(SIGINT, task, 1);
763 		nrservs++;
764 	}
765 
766 	return error;
767 }
768 EXPORT_SYMBOL_GPL(svc_set_num_threads);
769 
770 /*
771  * Called from a server thread as it's exiting. Caller must hold the BKL or
772  * the "service mutex", whichever is appropriate for the service.
773  */
774 void
775 svc_exit_thread(struct svc_rqst *rqstp)
776 {
777 	struct svc_serv	*serv = rqstp->rq_server;
778 	struct svc_pool	*pool = rqstp->rq_pool;
779 
780 	svc_release_buffer(rqstp);
781 	kfree(rqstp->rq_resp);
782 	kfree(rqstp->rq_argp);
783 	kfree(rqstp->rq_auth_data);
784 
785 	spin_lock_bh(&pool->sp_lock);
786 	pool->sp_nrthreads--;
787 	if (!test_and_set_bit(RQ_VICTIM, &rqstp->rq_flags))
788 		list_del_rcu(&rqstp->rq_all);
789 	spin_unlock_bh(&pool->sp_lock);
790 
791 	kfree_rcu(rqstp, rq_rcu_head);
792 
793 	/* Release the server */
794 	if (serv)
795 		svc_destroy(serv);
796 }
797 EXPORT_SYMBOL_GPL(svc_exit_thread);
798 
799 /*
800  * Register an "inet" protocol family netid with the local
801  * rpcbind daemon via an rpcbind v4 SET request.
802  *
803  * No netconfig infrastructure is available in the kernel, so
804  * we map IP_ protocol numbers to netids by hand.
805  *
806  * Returns zero on success; a negative errno value is returned
807  * if any error occurs.
808  */
809 static int __svc_rpcb_register4(struct net *net, const u32 program,
810 				const u32 version,
811 				const unsigned short protocol,
812 				const unsigned short port)
813 {
814 	const struct sockaddr_in sin = {
815 		.sin_family		= AF_INET,
816 		.sin_addr.s_addr	= htonl(INADDR_ANY),
817 		.sin_port		= htons(port),
818 	};
819 	const char *netid;
820 	int error;
821 
822 	switch (protocol) {
823 	case IPPROTO_UDP:
824 		netid = RPCBIND_NETID_UDP;
825 		break;
826 	case IPPROTO_TCP:
827 		netid = RPCBIND_NETID_TCP;
828 		break;
829 	default:
830 		return -ENOPROTOOPT;
831 	}
832 
833 	error = rpcb_v4_register(net, program, version,
834 					(const struct sockaddr *)&sin, netid);
835 
836 	/*
837 	 * User space didn't support rpcbind v4, so retry this
838 	 * registration request with the legacy rpcbind v2 protocol.
839 	 */
840 	if (error == -EPROTONOSUPPORT)
841 		error = rpcb_register(net, program, version, protocol, port);
842 
843 	return error;
844 }
845 
846 #if IS_ENABLED(CONFIG_IPV6)
847 /*
848  * Register an "inet6" protocol family netid with the local
849  * rpcbind daemon via an rpcbind v4 SET request.
850  *
851  * No netconfig infrastructure is available in the kernel, so
852  * we map IP_ protocol numbers to netids by hand.
853  *
854  * Returns zero on success; a negative errno value is returned
855  * if any error occurs.
856  */
857 static int __svc_rpcb_register6(struct net *net, const u32 program,
858 				const u32 version,
859 				const unsigned short protocol,
860 				const unsigned short port)
861 {
862 	const struct sockaddr_in6 sin6 = {
863 		.sin6_family		= AF_INET6,
864 		.sin6_addr		= IN6ADDR_ANY_INIT,
865 		.sin6_port		= htons(port),
866 	};
867 	const char *netid;
868 	int error;
869 
870 	switch (protocol) {
871 	case IPPROTO_UDP:
872 		netid = RPCBIND_NETID_UDP6;
873 		break;
874 	case IPPROTO_TCP:
875 		netid = RPCBIND_NETID_TCP6;
876 		break;
877 	default:
878 		return -ENOPROTOOPT;
879 	}
880 
881 	error = rpcb_v4_register(net, program, version,
882 					(const struct sockaddr *)&sin6, netid);
883 
884 	/*
885 	 * User space didn't support rpcbind version 4, so we won't
886 	 * use a PF_INET6 listener.
887 	 */
888 	if (error == -EPROTONOSUPPORT)
889 		error = -EAFNOSUPPORT;
890 
891 	return error;
892 }
893 #endif	/* IS_ENABLED(CONFIG_IPV6) */
894 
895 /*
896  * Register a kernel RPC service via rpcbind version 4.
897  *
898  * Returns zero on success; a negative errno value is returned
899  * if any error occurs.
900  */
901 static int __svc_register(struct net *net, const char *progname,
902 			  const u32 program, const u32 version,
903 			  const int family,
904 			  const unsigned short protocol,
905 			  const unsigned short port)
906 {
907 	int error = -EAFNOSUPPORT;
908 
909 	switch (family) {
910 	case PF_INET:
911 		error = __svc_rpcb_register4(net, program, version,
912 						protocol, port);
913 		break;
914 #if IS_ENABLED(CONFIG_IPV6)
915 	case PF_INET6:
916 		error = __svc_rpcb_register6(net, program, version,
917 						protocol, port);
918 #endif
919 	}
920 
921 	return error;
922 }
923 
924 /**
925  * svc_register - register an RPC service with the local portmapper
926  * @serv: svc_serv struct for the service to register
927  * @net: net namespace for the service to register
928  * @family: protocol family of service's listener socket
929  * @proto: transport protocol number to advertise
930  * @port: port to advertise
931  *
932  * Service is registered for any address in the passed-in protocol family
933  */
934 int svc_register(const struct svc_serv *serv, struct net *net,
935 		 const int family, const unsigned short proto,
936 		 const unsigned short port)
937 {
938 	struct svc_program	*progp;
939 	struct svc_version	*vers;
940 	unsigned int		i;
941 	int			error = 0;
942 
943 	WARN_ON_ONCE(proto == 0 && port == 0);
944 	if (proto == 0 && port == 0)
945 		return -EINVAL;
946 
947 	for (progp = serv->sv_program; progp; progp = progp->pg_next) {
948 		for (i = 0; i < progp->pg_nvers; i++) {
949 			vers = progp->pg_vers[i];
950 			if (vers == NULL)
951 				continue;
952 
953 			dprintk("svc: svc_register(%sv%d, %s, %u, %u)%s\n",
954 					progp->pg_name,
955 					i,
956 					proto == IPPROTO_UDP?  "udp" : "tcp",
957 					port,
958 					family,
959 					vers->vs_hidden ?
960 					" (but not telling portmap)" : "");
961 
962 			if (vers->vs_hidden)
963 				continue;
964 
965 			error = __svc_register(net, progp->pg_name, progp->pg_prog,
966 						i, family, proto, port);
967 
968 			if (vers->vs_rpcb_optnl) {
969 				error = 0;
970 				continue;
971 			}
972 
973 			if (error < 0) {
974 				printk(KERN_WARNING "svc: failed to register "
975 					"%sv%u RPC service (errno %d).\n",
976 					progp->pg_name, i, -error);
977 				break;
978 			}
979 		}
980 	}
981 
982 	return error;
983 }
984 
985 /*
986  * If user space is running rpcbind, it should take the v4 UNSET
987  * and clear everything for this [program, version].  If user space
988  * is running portmap, it will reject the v4 UNSET, but won't have
989  * any "inet6" entries anyway.  So a PMAP_UNSET should be sufficient
990  * in this case to clear all existing entries for [program, version].
991  */
992 static void __svc_unregister(struct net *net, const u32 program, const u32 version,
993 			     const char *progname)
994 {
995 	int error;
996 
997 	error = rpcb_v4_register(net, program, version, NULL, "");
998 
999 	/*
1000 	 * User space didn't support rpcbind v4, so retry this
1001 	 * request with the legacy rpcbind v2 protocol.
1002 	 */
1003 	if (error == -EPROTONOSUPPORT)
1004 		error = rpcb_register(net, program, version, 0, 0);
1005 
1006 	dprintk("svc: %s(%sv%u), error %d\n",
1007 			__func__, progname, version, error);
1008 }
1009 
1010 /*
1011  * All netids, bind addresses and ports registered for [program, version]
1012  * are removed from the local rpcbind database (if the service is not
1013  * hidden) to make way for a new instance of the service.
1014  *
1015  * The result of unregistration is reported via dprintk for those who want
1016  * verification of the result, but is otherwise not important.
1017  */
1018 static void svc_unregister(const struct svc_serv *serv, struct net *net)
1019 {
1020 	struct svc_program *progp;
1021 	unsigned long flags;
1022 	unsigned int i;
1023 
1024 	clear_thread_flag(TIF_SIGPENDING);
1025 
1026 	for (progp = serv->sv_program; progp; progp = progp->pg_next) {
1027 		for (i = 0; i < progp->pg_nvers; i++) {
1028 			if (progp->pg_vers[i] == NULL)
1029 				continue;
1030 			if (progp->pg_vers[i]->vs_hidden)
1031 				continue;
1032 
1033 			dprintk("svc: attempting to unregister %sv%u\n",
1034 				progp->pg_name, i);
1035 			__svc_unregister(net, progp->pg_prog, i, progp->pg_name);
1036 		}
1037 	}
1038 
1039 	spin_lock_irqsave(&current->sighand->siglock, flags);
1040 	recalc_sigpending();
1041 	spin_unlock_irqrestore(&current->sighand->siglock, flags);
1042 }
1043 
1044 /*
1045  * dprintk the given error with the address of the client that caused it.
1046  */
1047 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
1048 static __printf(2, 3)
1049 void svc_printk(struct svc_rqst *rqstp, const char *fmt, ...)
1050 {
1051 	struct va_format vaf;
1052 	va_list args;
1053 	char 	buf[RPC_MAX_ADDRBUFLEN];
1054 
1055 	va_start(args, fmt);
1056 
1057 	vaf.fmt = fmt;
1058 	vaf.va = &args;
1059 
1060 	dprintk("svc: %s: %pV", svc_print_addr(rqstp, buf, sizeof(buf)), &vaf);
1061 
1062 	va_end(args);
1063 }
1064 #else
1065 static __printf(2,3) void svc_printk(struct svc_rqst *rqstp, const char *fmt, ...) {}
1066 #endif
1067 
1068 /*
1069  * Common routine for processing the RPC request.
1070  */
1071 static int
1072 svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
1073 {
1074 	struct svc_program	*progp;
1075 	struct svc_version	*versp = NULL;	/* compiler food */
1076 	struct svc_procedure	*procp = NULL;
1077 	struct svc_serv		*serv = rqstp->rq_server;
1078 	kxdrproc_t		xdr;
1079 	__be32			*statp;
1080 	u32			prog, vers, proc;
1081 	__be32			auth_stat, rpc_stat;
1082 	int			auth_res;
1083 	__be32			*reply_statp;
1084 
1085 	rpc_stat = rpc_success;
1086 
1087 	if (argv->iov_len < 6*4)
1088 		goto err_short_len;
1089 
1090 	/* Will be turned off only in gss privacy case: */
1091 	set_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
1092 	/* Will be turned off only when NFSv4 Sessions are used */
1093 	set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
1094 	clear_bit(RQ_DROPME, &rqstp->rq_flags);
1095 
1096 	/* Setup reply header */
1097 	rqstp->rq_xprt->xpt_ops->xpo_prep_reply_hdr(rqstp);
1098 
1099 	svc_putu32(resv, rqstp->rq_xid);
1100 
1101 	vers = svc_getnl(argv);
1102 
1103 	/* First words of reply: */
1104 	svc_putnl(resv, 1);		/* REPLY */
1105 
1106 	if (vers != 2)		/* RPC version number */
1107 		goto err_bad_rpc;
1108 
1109 	/* Save position in case we later decide to reject: */
1110 	reply_statp = resv->iov_base + resv->iov_len;
1111 
1112 	svc_putnl(resv, 0);		/* ACCEPT */
1113 
1114 	rqstp->rq_prog = prog = svc_getnl(argv);	/* program number */
1115 	rqstp->rq_vers = vers = svc_getnl(argv);	/* version number */
1116 	rqstp->rq_proc = proc = svc_getnl(argv);	/* procedure number */
1117 
1118 	for (progp = serv->sv_program; progp; progp = progp->pg_next)
1119 		if (prog == progp->pg_prog)
1120 			break;
1121 
1122 	/*
1123 	 * Decode auth data, and add verifier to reply buffer.
1124 	 * We do this before anything else in order to get a decent
1125 	 * auth verifier.
1126 	 */
1127 	auth_res = svc_authenticate(rqstp, &auth_stat);
1128 	/* Also give the program a chance to reject this call: */
1129 	if (auth_res == SVC_OK && progp) {
1130 		auth_stat = rpc_autherr_badcred;
1131 		auth_res = progp->pg_authenticate(rqstp);
1132 	}
1133 	switch (auth_res) {
1134 	case SVC_OK:
1135 		break;
1136 	case SVC_GARBAGE:
1137 		goto err_garbage;
1138 	case SVC_SYSERR:
1139 		rpc_stat = rpc_system_err;
1140 		goto err_bad;
1141 	case SVC_DENIED:
1142 		goto err_bad_auth;
1143 	case SVC_CLOSE:
1144 		if (test_bit(XPT_TEMP, &rqstp->rq_xprt->xpt_flags))
1145 			svc_close_xprt(rqstp->rq_xprt);
1146 	case SVC_DROP:
1147 		goto dropit;
1148 	case SVC_COMPLETE:
1149 		goto sendit;
1150 	}
1151 
1152 	if (progp == NULL)
1153 		goto err_bad_prog;
1154 
1155 	if (vers >= progp->pg_nvers ||
1156 	  !(versp = progp->pg_vers[vers]))
1157 		goto err_bad_vers;
1158 
1159 	procp = versp->vs_proc + proc;
1160 	if (proc >= versp->vs_nproc || !procp->pc_func)
1161 		goto err_bad_proc;
1162 	rqstp->rq_procinfo = procp;
1163 
1164 	/* Syntactic check complete */
1165 	serv->sv_stats->rpccnt++;
1166 
1167 	/* Build the reply header. */
1168 	statp = resv->iov_base +resv->iov_len;
1169 	svc_putnl(resv, RPC_SUCCESS);
1170 
1171 	/* Bump per-procedure stats counter */
1172 	procp->pc_count++;
1173 
1174 	/* Initialize storage for argp and resp */
1175 	memset(rqstp->rq_argp, 0, procp->pc_argsize);
1176 	memset(rqstp->rq_resp, 0, procp->pc_ressize);
1177 
1178 	/* un-reserve some of the out-queue now that we have a
1179 	 * better idea of reply size
1180 	 */
1181 	if (procp->pc_xdrressize)
1182 		svc_reserve_auth(rqstp, procp->pc_xdrressize<<2);
1183 
1184 	/* Call the function that processes the request. */
1185 	if (!versp->vs_dispatch) {
1186 		/* Decode arguments */
1187 		xdr = procp->pc_decode;
1188 		if (xdr && !xdr(rqstp, argv->iov_base, rqstp->rq_argp))
1189 			goto err_garbage;
1190 
1191 		*statp = procp->pc_func(rqstp, rqstp->rq_argp, rqstp->rq_resp);
1192 
1193 		/* Encode reply */
1194 		if (test_bit(RQ_DROPME, &rqstp->rq_flags)) {
1195 			if (procp->pc_release)
1196 				procp->pc_release(rqstp, NULL, rqstp->rq_resp);
1197 			goto dropit;
1198 		}
1199 		if (*statp == rpc_success &&
1200 		    (xdr = procp->pc_encode) &&
1201 		    !xdr(rqstp, resv->iov_base+resv->iov_len, rqstp->rq_resp)) {
1202 			dprintk("svc: failed to encode reply\n");
1203 			/* serv->sv_stats->rpcsystemerr++; */
1204 			*statp = rpc_system_err;
1205 		}
1206 	} else {
1207 		dprintk("svc: calling dispatcher\n");
1208 		if (!versp->vs_dispatch(rqstp, statp)) {
1209 			/* Release reply info */
1210 			if (procp->pc_release)
1211 				procp->pc_release(rqstp, NULL, rqstp->rq_resp);
1212 			goto dropit;
1213 		}
1214 	}
1215 
1216 	/* Check RPC status result */
1217 	if (*statp != rpc_success)
1218 		resv->iov_len = ((void*)statp)  - resv->iov_base + 4;
1219 
1220 	/* Release reply info */
1221 	if (procp->pc_release)
1222 		procp->pc_release(rqstp, NULL, rqstp->rq_resp);
1223 
1224 	if (procp->pc_encode == NULL)
1225 		goto dropit;
1226 
1227  sendit:
1228 	if (svc_authorise(rqstp))
1229 		goto dropit;
1230 	return 1;		/* Caller can now send it */
1231 
1232  dropit:
1233 	svc_authorise(rqstp);	/* doesn't hurt to call this twice */
1234 	dprintk("svc: svc_process dropit\n");
1235 	return 0;
1236 
1237 err_short_len:
1238 	svc_printk(rqstp, "short len %Zd, dropping request\n",
1239 			argv->iov_len);
1240 
1241 	goto dropit;			/* drop request */
1242 
1243 err_bad_rpc:
1244 	serv->sv_stats->rpcbadfmt++;
1245 	svc_putnl(resv, 1);	/* REJECT */
1246 	svc_putnl(resv, 0);	/* RPC_MISMATCH */
1247 	svc_putnl(resv, 2);	/* Only RPCv2 supported */
1248 	svc_putnl(resv, 2);
1249 	goto sendit;
1250 
1251 err_bad_auth:
1252 	dprintk("svc: authentication failed (%d)\n", ntohl(auth_stat));
1253 	serv->sv_stats->rpcbadauth++;
1254 	/* Restore write pointer to location of accept status: */
1255 	xdr_ressize_check(rqstp, reply_statp);
1256 	svc_putnl(resv, 1);	/* REJECT */
1257 	svc_putnl(resv, 1);	/* AUTH_ERROR */
1258 	svc_putnl(resv, ntohl(auth_stat));	/* status */
1259 	goto sendit;
1260 
1261 err_bad_prog:
1262 	dprintk("svc: unknown program %d\n", prog);
1263 	serv->sv_stats->rpcbadfmt++;
1264 	svc_putnl(resv, RPC_PROG_UNAVAIL);
1265 	goto sendit;
1266 
1267 err_bad_vers:
1268 	svc_printk(rqstp, "unknown version (%d for prog %d, %s)\n",
1269 		       vers, prog, progp->pg_name);
1270 
1271 	serv->sv_stats->rpcbadfmt++;
1272 	svc_putnl(resv, RPC_PROG_MISMATCH);
1273 	svc_putnl(resv, progp->pg_lovers);
1274 	svc_putnl(resv, progp->pg_hivers);
1275 	goto sendit;
1276 
1277 err_bad_proc:
1278 	svc_printk(rqstp, "unknown procedure (%d)\n", proc);
1279 
1280 	serv->sv_stats->rpcbadfmt++;
1281 	svc_putnl(resv, RPC_PROC_UNAVAIL);
1282 	goto sendit;
1283 
1284 err_garbage:
1285 	svc_printk(rqstp, "failed to decode args\n");
1286 
1287 	rpc_stat = rpc_garbage_args;
1288 err_bad:
1289 	serv->sv_stats->rpcbadfmt++;
1290 	svc_putnl(resv, ntohl(rpc_stat));
1291 	goto sendit;
1292 }
1293 EXPORT_SYMBOL_GPL(svc_process);
1294 
1295 /*
1296  * Process the RPC request.
1297  */
1298 int
1299 svc_process(struct svc_rqst *rqstp)
1300 {
1301 	struct kvec		*argv = &rqstp->rq_arg.head[0];
1302 	struct kvec		*resv = &rqstp->rq_res.head[0];
1303 	struct svc_serv		*serv = rqstp->rq_server;
1304 	u32			dir;
1305 
1306 	/*
1307 	 * Setup response xdr_buf.
1308 	 * Initially it has just one page
1309 	 */
1310 	rqstp->rq_next_page = &rqstp->rq_respages[1];
1311 	resv->iov_base = page_address(rqstp->rq_respages[0]);
1312 	resv->iov_len = 0;
1313 	rqstp->rq_res.pages = rqstp->rq_respages + 1;
1314 	rqstp->rq_res.len = 0;
1315 	rqstp->rq_res.page_base = 0;
1316 	rqstp->rq_res.page_len = 0;
1317 	rqstp->rq_res.buflen = PAGE_SIZE;
1318 	rqstp->rq_res.tail[0].iov_base = NULL;
1319 	rqstp->rq_res.tail[0].iov_len = 0;
1320 
1321 	dir  = svc_getnl(argv);
1322 	if (dir != 0) {
1323 		/* direction != CALL */
1324 		svc_printk(rqstp, "bad direction %d, dropping request\n", dir);
1325 		serv->sv_stats->rpcbadfmt++;
1326 		goto out_drop;
1327 	}
1328 
1329 	/* Returns 1 for send, 0 for drop */
1330 	if (likely(svc_process_common(rqstp, argv, resv))) {
1331 		int ret = svc_send(rqstp);
1332 
1333 		trace_svc_process(rqstp, ret);
1334 		return ret;
1335 	}
1336 out_drop:
1337 	trace_svc_process(rqstp, 0);
1338 	svc_drop(rqstp);
1339 	return 0;
1340 }
1341 
1342 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
1343 /*
1344  * Process a backchannel RPC request that arrived over an existing
1345  * outbound connection
1346  */
1347 int
1348 bc_svc_process(struct svc_serv *serv, struct rpc_rqst *req,
1349 	       struct svc_rqst *rqstp)
1350 {
1351 	struct kvec	*argv = &rqstp->rq_arg.head[0];
1352 	struct kvec	*resv = &rqstp->rq_res.head[0];
1353 
1354 	/* Build the svc_rqst used by the common processing routine */
1355 	rqstp->rq_xprt = serv->sv_bc_xprt;
1356 	rqstp->rq_xid = req->rq_xid;
1357 	rqstp->rq_prot = req->rq_xprt->prot;
1358 	rqstp->rq_server = serv;
1359 
1360 	rqstp->rq_addrlen = sizeof(req->rq_xprt->addr);
1361 	memcpy(&rqstp->rq_addr, &req->rq_xprt->addr, rqstp->rq_addrlen);
1362 	memcpy(&rqstp->rq_arg, &req->rq_rcv_buf, sizeof(rqstp->rq_arg));
1363 	memcpy(&rqstp->rq_res, &req->rq_snd_buf, sizeof(rqstp->rq_res));
1364 
1365 	/* reset result send buffer "put" position */
1366 	resv->iov_len = 0;
1367 
1368 	if (rqstp->rq_prot != IPPROTO_TCP) {
1369 		printk(KERN_ERR "No support for Non-TCP transports!\n");
1370 		BUG();
1371 	}
1372 
1373 	/*
1374 	 * Skip the next two words because they've already been
1375 	 * processed in the trasport
1376 	 */
1377 	svc_getu32(argv);	/* XID */
1378 	svc_getnl(argv);	/* CALLDIR */
1379 
1380 	/* Returns 1 for send, 0 for drop */
1381 	if (svc_process_common(rqstp, argv, resv)) {
1382 		memcpy(&req->rq_snd_buf, &rqstp->rq_res,
1383 						sizeof(req->rq_snd_buf));
1384 		return bc_send(req);
1385 	} else {
1386 		/* drop request */
1387 		xprt_free_bc_request(req);
1388 		return 0;
1389 	}
1390 }
1391 EXPORT_SYMBOL_GPL(bc_svc_process);
1392 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
1393 
1394 /*
1395  * Return (transport-specific) limit on the rpc payload.
1396  */
1397 u32 svc_max_payload(const struct svc_rqst *rqstp)
1398 {
1399 	u32 max = rqstp->rq_xprt->xpt_class->xcl_max_payload;
1400 
1401 	if (rqstp->rq_server->sv_max_payload < max)
1402 		max = rqstp->rq_server->sv_max_payload;
1403 	return max;
1404 }
1405 EXPORT_SYMBOL_GPL(svc_max_payload);
1406