xref: /linux/fs/nfsd/nfsctl.c (revision b88fe2b5dd018c2b856fd6c32b82f25033e908d4)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Syscall interface to knfsd.
4  *
5  * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
6  */
7 
8 #include <linux/slab.h>
9 #include <linux/namei.h>
10 #include <linux/ctype.h>
11 #include <linux/fs_context.h>
12 
13 #include <linux/sunrpc/svcsock.h>
14 #include <linux/lockd/lockd.h>
15 #include <linux/sunrpc/addr.h>
16 #include <linux/sunrpc/gss_api.h>
17 #include <linux/sunrpc/rpc_pipe_fs.h>
18 #include <linux/sunrpc/svc.h>
19 #include <linux/module.h>
20 #include <linux/fsnotify.h>
21 #include <linux/nfslocalio.h>
22 
23 #include "idmap.h"
24 #include "nfsd.h"
25 #include "cache.h"
26 #include "state.h"
27 #include "netns.h"
28 #include "pnfs.h"
29 #include "filecache.h"
30 #include "trace.h"
31 #include "netlink.h"
32 
33 /*
34  *	We have a single directory with several nodes in it.
35  */
36 enum {
37 	NFSD_Root = 1,
38 	NFSD_List,
39 	NFSD_Export_Stats,
40 	NFSD_Export_features,
41 	NFSD_Fh,
42 	NFSD_FO_UnlockIP,
43 	NFSD_FO_UnlockFS,
44 	NFSD_Threads,
45 	NFSD_Pool_Threads,
46 	NFSD_Pool_Stats,
47 	NFSD_Reply_Cache_Stats,
48 	NFSD_Versions,
49 	NFSD_Ports,
50 	NFSD_MaxBlkSize,
51 	NFSD_Filecache,
52 	NFSD_Leasetime,
53 	NFSD_Gracetime,
54 	NFSD_RecoveryDir,
55 	NFSD_V4EndGrace,
56 	NFSD_MaxReserved
57 };
58 
59 /*
60  * write() for these nodes.
61  */
62 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
63 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
64 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
65 static ssize_t write_threads(struct file *file, char *buf, size_t size);
66 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
67 static ssize_t write_versions(struct file *file, char *buf, size_t size);
68 static ssize_t write_ports(struct file *file, char *buf, size_t size);
69 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
70 #ifdef CONFIG_NFSD_V4
71 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
72 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
73 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
74 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
75 #endif
76 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
77 #endif
78 
79 static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
80 	[NFSD_Fh] = write_filehandle,
81 	[NFSD_FO_UnlockIP] = write_unlock_ip,
82 	[NFSD_FO_UnlockFS] = write_unlock_fs,
83 	[NFSD_Threads] = write_threads,
84 	[NFSD_Pool_Threads] = write_pool_threads,
85 	[NFSD_Versions] = write_versions,
86 	[NFSD_Ports] = write_ports,
87 	[NFSD_MaxBlkSize] = write_maxblksize,
88 #ifdef CONFIG_NFSD_V4
89 	[NFSD_Leasetime] = write_leasetime,
90 	[NFSD_Gracetime] = write_gracetime,
91 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
92 	[NFSD_RecoveryDir] = write_recoverydir,
93 #endif
94 	[NFSD_V4EndGrace] = write_v4_end_grace,
95 #endif
96 };
97 
98 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
99 {
100 	ino_t ino =  file_inode(file)->i_ino;
101 	char *data;
102 	ssize_t rv;
103 
104 	if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
105 		return -EINVAL;
106 
107 	data = simple_transaction_get(file, buf, size);
108 	if (IS_ERR(data))
109 		return PTR_ERR(data);
110 
111 	rv = write_op[ino](file, data, size);
112 	if (rv < 0)
113 		return rv;
114 
115 	simple_transaction_set(file, rv);
116 	return size;
117 }
118 
119 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
120 {
121 	if (! file->private_data) {
122 		/* An attempt to read a transaction file without writing
123 		 * causes a 0-byte write so that the file can return
124 		 * state information
125 		 */
126 		ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
127 		if (rv < 0)
128 			return rv;
129 	}
130 	return simple_transaction_read(file, buf, size, pos);
131 }
132 
133 static const struct file_operations transaction_ops = {
134 	.write		= nfsctl_transaction_write,
135 	.read		= nfsctl_transaction_read,
136 	.release	= simple_transaction_release,
137 	.llseek		= default_llseek,
138 };
139 
140 static int exports_net_open(struct net *net, struct file *file)
141 {
142 	int err;
143 	struct seq_file *seq;
144 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
145 
146 	err = seq_open(file, &nfs_exports_op);
147 	if (err)
148 		return err;
149 
150 	seq = file->private_data;
151 	seq->private = nn->svc_export_cache;
152 	return 0;
153 }
154 
155 static int exports_nfsd_open(struct inode *inode, struct file *file)
156 {
157 	return exports_net_open(inode->i_sb->s_fs_info, file);
158 }
159 
160 static const struct file_operations exports_nfsd_operations = {
161 	.open		= exports_nfsd_open,
162 	.read		= seq_read,
163 	.llseek		= seq_lseek,
164 	.release	= seq_release,
165 };
166 
167 static int export_features_show(struct seq_file *m, void *v)
168 {
169 	seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
170 	return 0;
171 }
172 
173 DEFINE_SHOW_ATTRIBUTE(export_features);
174 
175 static int nfsd_pool_stats_open(struct inode *inode, struct file *file)
176 {
177 	struct nfsd_net *nn = net_generic(inode->i_sb->s_fs_info, nfsd_net_id);
178 
179 	return svc_pool_stats_open(&nn->nfsd_info, file);
180 }
181 
182 static const struct file_operations pool_stats_operations = {
183 	.open		= nfsd_pool_stats_open,
184 	.read		= seq_read,
185 	.llseek		= seq_lseek,
186 	.release	= seq_release,
187 };
188 
189 DEFINE_SHOW_ATTRIBUTE(nfsd_reply_cache_stats);
190 
191 DEFINE_SHOW_ATTRIBUTE(nfsd_file_cache_stats);
192 
193 /*----------------------------------------------------------------------------*/
194 /*
195  * payload - write methods
196  */
197 
198 static inline struct net *netns(struct file *file)
199 {
200 	return file_inode(file)->i_sb->s_fs_info;
201 }
202 
203 /*
204  * write_unlock_ip - Release all locks used by a client
205  *
206  * Experimental.
207  *
208  * Input:
209  *			buf:	'\n'-terminated C string containing a
210  *				presentation format IP address
211  *			size:	length of C string in @buf
212  * Output:
213  *	On success:	returns zero if all specified locks were released;
214  *			returns one if one or more locks were not released
215  *	On error:	return code is negative errno value
216  */
217 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
218 {
219 	struct sockaddr_storage address;
220 	struct sockaddr *sap = (struct sockaddr *)&address;
221 	size_t salen = sizeof(address);
222 	char *fo_path;
223 	struct net *net = netns(file);
224 
225 	/* sanity check */
226 	if (size == 0)
227 		return -EINVAL;
228 
229 	if (buf[size-1] != '\n')
230 		return -EINVAL;
231 
232 	fo_path = buf;
233 	if (qword_get(&buf, fo_path, size) < 0)
234 		return -EINVAL;
235 
236 	if (rpc_pton(net, fo_path, size, sap, salen) == 0)
237 		return -EINVAL;
238 
239 	trace_nfsd_ctl_unlock_ip(net, buf);
240 	return nlmsvc_unlock_all_by_ip(sap);
241 }
242 
243 /*
244  * write_unlock_fs - Release all locks on a local file system
245  *
246  * Experimental.
247  *
248  * Input:
249  *			buf:	'\n'-terminated C string containing the
250  *				absolute pathname of a local file system
251  *			size:	length of C string in @buf
252  * Output:
253  *	On success:	returns zero if all specified locks were released;
254  *			returns one if one or more locks were not released
255  *	On error:	return code is negative errno value
256  */
257 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
258 {
259 	struct path path;
260 	char *fo_path;
261 	int error;
262 
263 	/* sanity check */
264 	if (size == 0)
265 		return -EINVAL;
266 
267 	if (buf[size-1] != '\n')
268 		return -EINVAL;
269 
270 	fo_path = buf;
271 	if (qword_get(&buf, fo_path, size) < 0)
272 		return -EINVAL;
273 	trace_nfsd_ctl_unlock_fs(netns(file), fo_path);
274 	error = kern_path(fo_path, 0, &path);
275 	if (error)
276 		return error;
277 
278 	/*
279 	 * XXX: Needs better sanity checking.  Otherwise we could end up
280 	 * releasing locks on the wrong file system.
281 	 *
282 	 * For example:
283 	 * 1.  Does the path refer to a directory?
284 	 * 2.  Is that directory a mount point, or
285 	 * 3.  Is that directory the root of an exported file system?
286 	 */
287 	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
288 	nfsd4_revoke_states(netns(file), path.dentry->d_sb);
289 
290 	path_put(&path);
291 	return error;
292 }
293 
294 /*
295  * write_filehandle - Get a variable-length NFS file handle by path
296  *
297  * On input, the buffer contains a '\n'-terminated C string comprised of
298  * three alphanumeric words separated by whitespace.  The string may
299  * contain escape sequences.
300  *
301  * Input:
302  *			buf:
303  *				domain:		client domain name
304  *				path:		export pathname
305  *				maxsize:	numeric maximum size of
306  *						@buf
307  *			size:	length of C string in @buf
308  * Output:
309  *	On success:	passed-in buffer filled with '\n'-terminated C
310  *			string containing a ASCII hex text version
311  *			of the NFS file handle;
312  *			return code is the size in bytes of the string
313  *	On error:	return code is negative errno value
314  */
315 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
316 {
317 	char *dname, *path;
318 	int maxsize;
319 	char *mesg = buf;
320 	int len;
321 	struct auth_domain *dom;
322 	struct knfsd_fh fh;
323 
324 	if (size == 0)
325 		return -EINVAL;
326 
327 	if (buf[size-1] != '\n')
328 		return -EINVAL;
329 	buf[size-1] = 0;
330 
331 	dname = mesg;
332 	len = qword_get(&mesg, dname, size);
333 	if (len <= 0)
334 		return -EINVAL;
335 
336 	path = dname+len+1;
337 	len = qword_get(&mesg, path, size);
338 	if (len <= 0)
339 		return -EINVAL;
340 
341 	len = get_int(&mesg, &maxsize);
342 	if (len)
343 		return len;
344 
345 	if (maxsize < NFS_FHSIZE)
346 		return -EINVAL;
347 	maxsize = min(maxsize, NFS3_FHSIZE);
348 
349 	if (qword_get(&mesg, mesg, size) > 0)
350 		return -EINVAL;
351 
352 	trace_nfsd_ctl_filehandle(netns(file), dname, path, maxsize);
353 
354 	/* we have all the words, they are in buf.. */
355 	dom = unix_domain_find(dname);
356 	if (!dom)
357 		return -ENOMEM;
358 
359 	len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
360 	auth_domain_put(dom);
361 	if (len)
362 		return len;
363 
364 	mesg = buf;
365 	len = SIMPLE_TRANSACTION_LIMIT;
366 	qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
367 	mesg[-1] = '\n';
368 	return mesg - buf;
369 }
370 
371 /*
372  * write_threads - Start NFSD, or report the current number of running threads
373  *
374  * Input:
375  *			buf:		ignored
376  *			size:		zero
377  * Output:
378  *	On success:	passed-in buffer filled with '\n'-terminated C
379  *			string numeric value representing the number of
380  *			running NFSD threads;
381  *			return code is the size in bytes of the string
382  *	On error:	return code is zero
383  *
384  * OR
385  *
386  * Input:
387  *			buf:		C string containing an unsigned
388  *					integer value representing the
389  *					number of NFSD threads to start
390  *			size:		non-zero length of C string in @buf
391  * Output:
392  *	On success:	NFS service is started;
393  *			passed-in buffer filled with '\n'-terminated C
394  *			string numeric value representing the number of
395  *			running NFSD threads;
396  *			return code is the size in bytes of the string
397  *	On error:	return code is zero or a negative errno value
398  */
399 static ssize_t write_threads(struct file *file, char *buf, size_t size)
400 {
401 	char *mesg = buf;
402 	int rv;
403 	struct net *net = netns(file);
404 
405 	if (size > 0) {
406 		int newthreads;
407 		rv = get_int(&mesg, &newthreads);
408 		if (rv)
409 			return rv;
410 		if (newthreads < 0)
411 			return -EINVAL;
412 		trace_nfsd_ctl_threads(net, newthreads);
413 		mutex_lock(&nfsd_mutex);
414 		rv = nfsd_svc(1, &newthreads, net, file->f_cred, NULL);
415 		mutex_unlock(&nfsd_mutex);
416 		if (rv < 0)
417 			return rv;
418 	} else
419 		rv = nfsd_nrthreads(net);
420 
421 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
422 }
423 
424 /*
425  * write_pool_threads - Set or report the current number of threads per pool
426  *
427  * Input:
428  *			buf:		ignored
429  *			size:		zero
430  *
431  * OR
432  *
433  * Input:
434  *			buf:		C string containing whitespace-
435  *					separated unsigned integer values
436  *					representing the number of NFSD
437  *					threads to start in each pool
438  *			size:		non-zero length of C string in @buf
439  * Output:
440  *	On success:	passed-in buffer filled with '\n'-terminated C
441  *			string containing integer values representing the
442  *			number of NFSD threads in each pool;
443  *			return code is the size in bytes of the string
444  *	On error:	return code is zero or a negative errno value
445  */
446 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
447 {
448 	/* if size > 0, look for an array of number of threads per node
449 	 * and apply them  then write out number of threads per node as reply
450 	 */
451 	char *mesg = buf;
452 	int i;
453 	int rv;
454 	int len;
455 	int npools;
456 	int *nthreads;
457 	struct net *net = netns(file);
458 
459 	mutex_lock(&nfsd_mutex);
460 	npools = nfsd_nrpools(net);
461 	if (npools == 0) {
462 		/*
463 		 * NFS is shut down.  The admin can start it by
464 		 * writing to the threads file but NOT the pool_threads
465 		 * file, sorry.  Report zero threads.
466 		 */
467 		mutex_unlock(&nfsd_mutex);
468 		strcpy(buf, "0\n");
469 		return strlen(buf);
470 	}
471 
472 	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
473 	rv = -ENOMEM;
474 	if (nthreads == NULL)
475 		goto out_free;
476 
477 	if (size > 0) {
478 		for (i = 0; i < npools; i++) {
479 			rv = get_int(&mesg, &nthreads[i]);
480 			if (rv == -ENOENT)
481 				break;		/* fewer numbers than pools */
482 			if (rv)
483 				goto out_free;	/* syntax error */
484 			rv = -EINVAL;
485 			if (nthreads[i] < 0)
486 				goto out_free;
487 			trace_nfsd_ctl_pool_threads(net, i, nthreads[i]);
488 		}
489 
490 		/*
491 		 * There must always be a thread in pool 0; the admin
492 		 * can't shut down NFS completely using pool_threads.
493 		 */
494 		if (nthreads[0] == 0)
495 			nthreads[0] = 1;
496 
497 		rv = nfsd_set_nrthreads(i, nthreads, net);
498 		if (rv)
499 			goto out_free;
500 	}
501 
502 	rv = nfsd_get_nrthreads(npools, nthreads, net);
503 	if (rv)
504 		goto out_free;
505 
506 	mesg = buf;
507 	size = SIMPLE_TRANSACTION_LIMIT;
508 	for (i = 0; i < npools && size > 0; i++) {
509 		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
510 		len = strlen(mesg);
511 		size -= len;
512 		mesg += len;
513 	}
514 	rv = mesg - buf;
515 out_free:
516 	kfree(nthreads);
517 	mutex_unlock(&nfsd_mutex);
518 	return rv;
519 }
520 
521 static ssize_t
522 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
523 		const char *sep, unsigned vers, int minor)
524 {
525 	const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
526 	bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
527 
528 	if (vers == 4 && minor >= 0 &&
529 	    !nfsd_minorversion(nn, minor, NFSD_TEST))
530 		supported = false;
531 	if (minor == 0 && supported)
532 		/*
533 		 * special case for backward compatability.
534 		 * +4.0 is never reported, it is implied by
535 		 * +4, unless -4.0 is present.
536 		 */
537 		return 0;
538 	return snprintf(buf, remaining, format, sep,
539 			supported ? '+' : '-', vers, minor);
540 }
541 
542 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
543 {
544 	char *mesg = buf;
545 	char *vers, *minorp, sign;
546 	int len, num, remaining;
547 	ssize_t tlen = 0;
548 	char *sep;
549 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
550 
551 	if (size > 0) {
552 		if (nn->nfsd_serv)
553 			/* Cannot change versions without updating
554 			 * nn->nfsd_serv->sv_xdrsize, and reallocing
555 			 * rq_argp and rq_resp
556 			 */
557 			return -EBUSY;
558 		if (buf[size-1] != '\n')
559 			return -EINVAL;
560 		buf[size-1] = 0;
561 		trace_nfsd_ctl_version(netns(file), buf);
562 
563 		vers = mesg;
564 		len = qword_get(&mesg, vers, size);
565 		if (len <= 0) return -EINVAL;
566 		do {
567 			enum vers_op cmd;
568 			unsigned minor;
569 			sign = *vers;
570 			if (sign == '+' || sign == '-')
571 				num = simple_strtol((vers+1), &minorp, 0);
572 			else
573 				num = simple_strtol(vers, &minorp, 0);
574 			if (*minorp == '.') {
575 				if (num != 4)
576 					return -EINVAL;
577 				if (kstrtouint(minorp+1, 0, &minor) < 0)
578 					return -EINVAL;
579 			}
580 
581 			cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
582 			switch(num) {
583 #ifdef CONFIG_NFSD_V2
584 			case 2:
585 #endif
586 			case 3:
587 				nfsd_vers(nn, num, cmd);
588 				break;
589 			case 4:
590 				if (*minorp == '.') {
591 					if (nfsd_minorversion(nn, minor, cmd) < 0)
592 						return -EINVAL;
593 				} else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
594 					/*
595 					 * Either we have +4 and no minors are enabled,
596 					 * or we have -4 and at least one minor is enabled.
597 					 * In either case, propagate 'cmd' to all minors.
598 					 */
599 					minor = 0;
600 					while (nfsd_minorversion(nn, minor, cmd) >= 0)
601 						minor++;
602 				}
603 				break;
604 			default:
605 				/* Ignore requests to disable non-existent versions */
606 				if (cmd == NFSD_SET)
607 					return -EINVAL;
608 			}
609 			vers += len + 1;
610 		} while ((len = qword_get(&mesg, vers, size)) > 0);
611 		/* If all get turned off, turn them back on, as
612 		 * having no versions is BAD
613 		 */
614 		nfsd_reset_versions(nn);
615 	}
616 
617 	/* Now write current state into reply buffer */
618 	sep = "";
619 	remaining = SIMPLE_TRANSACTION_LIMIT;
620 	for (num=2 ; num <= 4 ; num++) {
621 		int minor;
622 		if (!nfsd_vers(nn, num, NFSD_AVAIL))
623 			continue;
624 
625 		minor = -1;
626 		do {
627 			len = nfsd_print_version_support(nn, buf, remaining,
628 					sep, num, minor);
629 			if (len >= remaining)
630 				goto out;
631 			remaining -= len;
632 			buf += len;
633 			tlen += len;
634 			minor++;
635 			if (len)
636 				sep = " ";
637 		} while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
638 	}
639 out:
640 	len = snprintf(buf, remaining, "\n");
641 	if (len >= remaining)
642 		return -EINVAL;
643 	return tlen + len;
644 }
645 
646 /*
647  * write_versions - Set or report the available NFS protocol versions
648  *
649  * Input:
650  *			buf:		ignored
651  *			size:		zero
652  * Output:
653  *	On success:	passed-in buffer filled with '\n'-terminated C
654  *			string containing positive or negative integer
655  *			values representing the current status of each
656  *			protocol version;
657  *			return code is the size in bytes of the string
658  *	On error:	return code is zero or a negative errno value
659  *
660  * OR
661  *
662  * Input:
663  *			buf:		C string containing whitespace-
664  *					separated positive or negative
665  *					integer values representing NFS
666  *					protocol versions to enable ("+n")
667  *					or disable ("-n")
668  *			size:		non-zero length of C string in @buf
669  * Output:
670  *	On success:	status of zero or more protocol versions has
671  *			been updated; passed-in buffer filled with
672  *			'\n'-terminated C string containing positive
673  *			or negative integer values representing the
674  *			current status of each protocol version;
675  *			return code is the size in bytes of the string
676  *	On error:	return code is zero or a negative errno value
677  */
678 static ssize_t write_versions(struct file *file, char *buf, size_t size)
679 {
680 	ssize_t rv;
681 
682 	mutex_lock(&nfsd_mutex);
683 	rv = __write_versions(file, buf, size);
684 	mutex_unlock(&nfsd_mutex);
685 	return rv;
686 }
687 
688 /*
689  * Zero-length write.  Return a list of NFSD's current listener
690  * transports.
691  */
692 static ssize_t __write_ports_names(char *buf, struct net *net)
693 {
694 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
695 
696 	if (nn->nfsd_serv == NULL)
697 		return 0;
698 	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
699 }
700 
701 /*
702  * A single 'fd' number was written, in which case it must be for
703  * a socket of a supported family/protocol, and we use it as an
704  * nfsd listener.
705  */
706 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
707 {
708 	char *mesg = buf;
709 	int fd, err;
710 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
711 	struct svc_serv *serv;
712 
713 	err = get_int(&mesg, &fd);
714 	if (err != 0 || fd < 0)
715 		return -EINVAL;
716 	trace_nfsd_ctl_ports_addfd(net, fd);
717 
718 	err = nfsd_create_serv(net);
719 	if (err != 0)
720 		return err;
721 
722 	serv = nn->nfsd_serv;
723 	err = svc_addsock(serv, net, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
724 
725 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
726 		nfsd_destroy_serv(net);
727 
728 	return err;
729 }
730 
731 /*
732  * A transport listener is added by writing its transport name and
733  * a port number.
734  */
735 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
736 {
737 	char transport[16];
738 	struct svc_xprt *xprt;
739 	int port, err;
740 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
741 	struct svc_serv *serv;
742 
743 	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
744 		return -EINVAL;
745 
746 	if (port < 1 || port > USHRT_MAX)
747 		return -EINVAL;
748 	trace_nfsd_ctl_ports_addxprt(net, transport, port);
749 
750 	err = nfsd_create_serv(net);
751 	if (err != 0)
752 		return err;
753 
754 	serv = nn->nfsd_serv;
755 	err = svc_xprt_create(serv, transport, net,
756 			      PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
757 	if (err < 0)
758 		goto out_err;
759 
760 	err = svc_xprt_create(serv, transport, net,
761 			      PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
762 	if (err < 0 && err != -EAFNOSUPPORT)
763 		goto out_close;
764 
765 	return 0;
766 out_close:
767 	xprt = svc_find_xprt(serv, transport, net, PF_INET, port);
768 	if (xprt != NULL) {
769 		svc_xprt_close(xprt);
770 		svc_xprt_put(xprt);
771 	}
772 out_err:
773 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
774 		nfsd_destroy_serv(net);
775 
776 	return err;
777 }
778 
779 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
780 			     struct net *net)
781 {
782 	if (size == 0)
783 		return __write_ports_names(buf, net);
784 
785 	if (isdigit(buf[0]))
786 		return __write_ports_addfd(buf, net, file->f_cred);
787 
788 	if (isalpha(buf[0]))
789 		return __write_ports_addxprt(buf, net, file->f_cred);
790 
791 	return -EINVAL;
792 }
793 
794 /*
795  * write_ports - Pass a socket file descriptor or transport name to listen on
796  *
797  * Input:
798  *			buf:		ignored
799  *			size:		zero
800  * Output:
801  *	On success:	passed-in buffer filled with a '\n'-terminated C
802  *			string containing a whitespace-separated list of
803  *			named NFSD listeners;
804  *			return code is the size in bytes of the string
805  *	On error:	return code is zero or a negative errno value
806  *
807  * OR
808  *
809  * Input:
810  *			buf:		C string containing an unsigned
811  *					integer value representing a bound
812  *					but unconnected socket that is to be
813  *					used as an NFSD listener; listen(3)
814  *					must be called for a SOCK_STREAM
815  *					socket, otherwise it is ignored
816  *			size:		non-zero length of C string in @buf
817  * Output:
818  *	On success:	NFS service is started;
819  *			passed-in buffer filled with a '\n'-terminated C
820  *			string containing a unique alphanumeric name of
821  *			the listener;
822  *			return code is the size in bytes of the string
823  *	On error:	return code is a negative errno value
824  *
825  * OR
826  *
827  * Input:
828  *			buf:		C string containing a transport
829  *					name and an unsigned integer value
830  *					representing the port to listen on,
831  *					separated by whitespace
832  *			size:		non-zero length of C string in @buf
833  * Output:
834  *	On success:	returns zero; NFS service is started
835  *	On error:	return code is a negative errno value
836  */
837 static ssize_t write_ports(struct file *file, char *buf, size_t size)
838 {
839 	ssize_t rv;
840 
841 	mutex_lock(&nfsd_mutex);
842 	rv = __write_ports(file, buf, size, netns(file));
843 	mutex_unlock(&nfsd_mutex);
844 	return rv;
845 }
846 
847 
848 int nfsd_max_blksize;
849 
850 /*
851  * write_maxblksize - Set or report the current NFS blksize
852  *
853  * Input:
854  *			buf:		ignored
855  *			size:		zero
856  *
857  * OR
858  *
859  * Input:
860  *			buf:		C string containing an unsigned
861  *					integer value representing the new
862  *					NFS blksize
863  *			size:		non-zero length of C string in @buf
864  * Output:
865  *	On success:	passed-in buffer filled with '\n'-terminated C string
866  *			containing numeric value of the current NFS blksize
867  *			setting;
868  *			return code is the size in bytes of the string
869  *	On error:	return code is zero or a negative errno value
870  */
871 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
872 {
873 	char *mesg = buf;
874 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
875 
876 	if (size > 0) {
877 		int bsize;
878 		int rv = get_int(&mesg, &bsize);
879 		if (rv)
880 			return rv;
881 		trace_nfsd_ctl_maxblksize(netns(file), bsize);
882 
883 		/* force bsize into allowed range and
884 		 * required alignment.
885 		 */
886 		bsize = max_t(int, bsize, 1024);
887 		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
888 		bsize &= ~(1024-1);
889 		mutex_lock(&nfsd_mutex);
890 		if (nn->nfsd_serv) {
891 			mutex_unlock(&nfsd_mutex);
892 			return -EBUSY;
893 		}
894 		nfsd_max_blksize = bsize;
895 		mutex_unlock(&nfsd_mutex);
896 	}
897 
898 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
899 							nfsd_max_blksize);
900 }
901 
902 #ifdef CONFIG_NFSD_V4
903 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
904 				  time64_t *time, struct nfsd_net *nn)
905 {
906 	struct dentry *dentry = file_dentry(file);
907 	char *mesg = buf;
908 	int rv, i;
909 
910 	if (size > 0) {
911 		if (nn->nfsd_serv)
912 			return -EBUSY;
913 		rv = get_int(&mesg, &i);
914 		if (rv)
915 			return rv;
916 		trace_nfsd_ctl_time(netns(file), dentry->d_name.name,
917 				    dentry->d_name.len, i);
918 
919 		/*
920 		 * Some sanity checking.  We don't have a reason for
921 		 * these particular numbers, but problems with the
922 		 * extremes are:
923 		 *	- Too short: the briefest network outage may
924 		 *	  cause clients to lose all their locks.  Also,
925 		 *	  the frequent polling may be wasteful.
926 		 *	- Too long: do you really want reboot recovery
927 		 *	  to take more than an hour?  Or to make other
928 		 *	  clients wait an hour before being able to
929 		 *	  revoke a dead client's locks?
930 		 */
931 		if (i < 10 || i > 3600)
932 			return -EINVAL;
933 		*time = i;
934 	}
935 
936 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
937 }
938 
939 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
940 				time64_t *time, struct nfsd_net *nn)
941 {
942 	ssize_t rv;
943 
944 	mutex_lock(&nfsd_mutex);
945 	rv = __nfsd4_write_time(file, buf, size, time, nn);
946 	mutex_unlock(&nfsd_mutex);
947 	return rv;
948 }
949 
950 /*
951  * write_leasetime - Set or report the current NFSv4 lease time
952  *
953  * Input:
954  *			buf:		ignored
955  *			size:		zero
956  *
957  * OR
958  *
959  * Input:
960  *			buf:		C string containing an unsigned
961  *					integer value representing the new
962  *					NFSv4 lease expiry time
963  *			size:		non-zero length of C string in @buf
964  * Output:
965  *	On success:	passed-in buffer filled with '\n'-terminated C
966  *			string containing unsigned integer value of the
967  *			current lease expiry time;
968  *			return code is the size in bytes of the string
969  *	On error:	return code is zero or a negative errno value
970  */
971 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
972 {
973 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
974 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
975 }
976 
977 /*
978  * write_gracetime - Set or report current NFSv4 grace period time
979  *
980  * As above, but sets the time of the NFSv4 grace period.
981  *
982  * Note this should never be set to less than the *previous*
983  * lease-period time, but we don't try to enforce this.  (In the common
984  * case (a new boot), we don't know what the previous lease time was
985  * anyway.)
986  */
987 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
988 {
989 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
990 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
991 }
992 
993 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
994 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
995 				   struct nfsd_net *nn)
996 {
997 	char *mesg = buf;
998 	char *recdir;
999 	int len, status;
1000 
1001 	if (size > 0) {
1002 		if (nn->nfsd_serv)
1003 			return -EBUSY;
1004 		if (size > PATH_MAX || buf[size-1] != '\n')
1005 			return -EINVAL;
1006 		buf[size-1] = 0;
1007 
1008 		recdir = mesg;
1009 		len = qword_get(&mesg, recdir, size);
1010 		if (len <= 0)
1011 			return -EINVAL;
1012 		trace_nfsd_ctl_recoverydir(netns(file), recdir);
1013 
1014 		status = nfs4_reset_recoverydir(recdir);
1015 		if (status)
1016 			return status;
1017 	}
1018 
1019 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1020 							nfs4_recoverydir());
1021 }
1022 
1023 /*
1024  * write_recoverydir - Set or report the pathname of the recovery directory
1025  *
1026  * Input:
1027  *			buf:		ignored
1028  *			size:		zero
1029  *
1030  * OR
1031  *
1032  * Input:
1033  *			buf:		C string containing the pathname
1034  *					of the directory on a local file
1035  *					system containing permanent NFSv4
1036  *					recovery data
1037  *			size:		non-zero length of C string in @buf
1038  * Output:
1039  *	On success:	passed-in buffer filled with '\n'-terminated C string
1040  *			containing the current recovery pathname setting;
1041  *			return code is the size in bytes of the string
1042  *	On error:	return code is zero or a negative errno value
1043  */
1044 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1045 {
1046 	ssize_t rv;
1047 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1048 
1049 	mutex_lock(&nfsd_mutex);
1050 	rv = __write_recoverydir(file, buf, size, nn);
1051 	mutex_unlock(&nfsd_mutex);
1052 	return rv;
1053 }
1054 #endif
1055 
1056 /*
1057  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1058  *
1059  * Input:
1060  *			buf:		ignored
1061  *			size:		zero
1062  * OR
1063  *
1064  * Input:
1065  *			buf:		any value
1066  *			size:		non-zero length of C string in @buf
1067  * Output:
1068  *			passed-in buffer filled with "Y" or "N" with a newline
1069  *			and NULL-terminated C string. This indicates whether
1070  *			the grace period has ended in the current net
1071  *			namespace. Return code is the size in bytes of the
1072  *			string. Writing a string that starts with 'Y', 'y', or
1073  *			'1' to the file will end the grace period for nfsd's v4
1074  *			lock manager.
1075  */
1076 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1077 {
1078 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1079 
1080 	if (size > 0) {
1081 		switch(buf[0]) {
1082 		case 'Y':
1083 		case 'y':
1084 		case '1':
1085 			if (!nn->nfsd_serv)
1086 				return -EBUSY;
1087 			trace_nfsd_end_grace(netns(file));
1088 			nfsd4_end_grace(nn);
1089 			break;
1090 		default:
1091 			return -EINVAL;
1092 		}
1093 	}
1094 
1095 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1096 			 nn->grace_ended ? 'Y' : 'N');
1097 }
1098 
1099 #endif
1100 
1101 /*----------------------------------------------------------------------------*/
1102 /*
1103  *	populating the filesystem.
1104  */
1105 
1106 /* Basically copying rpc_get_inode. */
1107 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1108 {
1109 	struct inode *inode = new_inode(sb);
1110 	if (!inode)
1111 		return NULL;
1112 	/* Following advice from simple_fill_super documentation: */
1113 	inode->i_ino = iunique(sb, NFSD_MaxReserved);
1114 	inode->i_mode = mode;
1115 	simple_inode_init_ts(inode);
1116 	switch (mode & S_IFMT) {
1117 	case S_IFDIR:
1118 		inode->i_fop = &simple_dir_operations;
1119 		inode->i_op = &simple_dir_inode_operations;
1120 		inc_nlink(inode);
1121 		break;
1122 	case S_IFLNK:
1123 		inode->i_op = &simple_symlink_inode_operations;
1124 		break;
1125 	default:
1126 		break;
1127 	}
1128 	return inode;
1129 }
1130 
1131 static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1132 {
1133 	struct inode *inode;
1134 
1135 	inode = nfsd_get_inode(dir->i_sb, mode);
1136 	if (!inode)
1137 		return -ENOMEM;
1138 	if (ncl) {
1139 		inode->i_private = ncl;
1140 		kref_get(&ncl->cl_ref);
1141 	}
1142 	d_add(dentry, inode);
1143 	inc_nlink(dir);
1144 	fsnotify_mkdir(dir, dentry);
1145 	return 0;
1146 }
1147 
1148 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1149 {
1150 	struct inode *dir = parent->d_inode;
1151 	struct dentry *dentry;
1152 	int ret = -ENOMEM;
1153 
1154 	inode_lock(dir);
1155 	dentry = d_alloc_name(parent, name);
1156 	if (!dentry)
1157 		goto out_err;
1158 	ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1159 	if (ret)
1160 		goto out_err;
1161 out:
1162 	inode_unlock(dir);
1163 	return dentry;
1164 out_err:
1165 	dput(dentry);
1166 	dentry = ERR_PTR(ret);
1167 	goto out;
1168 }
1169 
1170 #if IS_ENABLED(CONFIG_SUNRPC_GSS)
1171 static int __nfsd_symlink(struct inode *dir, struct dentry *dentry,
1172 			  umode_t mode, const char *content)
1173 {
1174 	struct inode *inode;
1175 
1176 	inode = nfsd_get_inode(dir->i_sb, mode);
1177 	if (!inode)
1178 		return -ENOMEM;
1179 
1180 	inode->i_link = (char *)content;
1181 	inode->i_size = strlen(content);
1182 
1183 	d_add(dentry, inode);
1184 	inc_nlink(dir);
1185 	fsnotify_create(dir, dentry);
1186 	return 0;
1187 }
1188 
1189 /*
1190  * @content is assumed to be a NUL-terminated string that lives
1191  * longer than the symlink itself.
1192  */
1193 static void _nfsd_symlink(struct dentry *parent, const char *name,
1194 			  const char *content)
1195 {
1196 	struct inode *dir = parent->d_inode;
1197 	struct dentry *dentry;
1198 	int ret;
1199 
1200 	inode_lock(dir);
1201 	dentry = d_alloc_name(parent, name);
1202 	if (!dentry)
1203 		goto out;
1204 	ret = __nfsd_symlink(d_inode(parent), dentry, S_IFLNK | 0777, content);
1205 	if (ret)
1206 		dput(dentry);
1207 out:
1208 	inode_unlock(dir);
1209 }
1210 #else
1211 static inline void _nfsd_symlink(struct dentry *parent, const char *name,
1212 				 const char *content)
1213 {
1214 }
1215 
1216 #endif
1217 
1218 static void clear_ncl(struct dentry *dentry)
1219 {
1220 	struct inode *inode = d_inode(dentry);
1221 	struct nfsdfs_client *ncl = inode->i_private;
1222 
1223 	spin_lock(&inode->i_lock);
1224 	inode->i_private = NULL;
1225 	spin_unlock(&inode->i_lock);
1226 	kref_put(&ncl->cl_ref, ncl->cl_release);
1227 }
1228 
1229 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1230 {
1231 	struct nfsdfs_client *nc;
1232 
1233 	spin_lock(&inode->i_lock);
1234 	nc = inode->i_private;
1235 	if (nc)
1236 		kref_get(&nc->cl_ref);
1237 	spin_unlock(&inode->i_lock);
1238 	return nc;
1239 }
1240 
1241 /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1242  * code instead. */
1243 static  int nfsdfs_create_files(struct dentry *root,
1244 				const struct tree_descr *files,
1245 				struct nfsdfs_client *ncl,
1246 				struct dentry **fdentries)
1247 {
1248 	struct inode *dir = d_inode(root);
1249 	struct inode *inode;
1250 	struct dentry *dentry;
1251 	int i;
1252 
1253 	inode_lock(dir);
1254 	for (i = 0; files->name && files->name[0]; i++, files++) {
1255 		dentry = d_alloc_name(root, files->name);
1256 		if (!dentry)
1257 			goto out;
1258 		inode = nfsd_get_inode(d_inode(root)->i_sb,
1259 					S_IFREG | files->mode);
1260 		if (!inode) {
1261 			dput(dentry);
1262 			goto out;
1263 		}
1264 		kref_get(&ncl->cl_ref);
1265 		inode->i_fop = files->ops;
1266 		inode->i_private = ncl;
1267 		d_add(dentry, inode);
1268 		fsnotify_create(dir, dentry);
1269 		if (fdentries)
1270 			fdentries[i] = dentry;
1271 	}
1272 	inode_unlock(dir);
1273 	return 0;
1274 out:
1275 	inode_unlock(dir);
1276 	return -ENOMEM;
1277 }
1278 
1279 /* on success, returns positive number unique to that client. */
1280 struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1281 				 struct nfsdfs_client *ncl, u32 id,
1282 				 const struct tree_descr *files,
1283 				 struct dentry **fdentries)
1284 {
1285 	struct dentry *dentry;
1286 	char name[11];
1287 	int ret;
1288 
1289 	sprintf(name, "%u", id);
1290 
1291 	dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1292 	if (IS_ERR(dentry)) /* XXX: tossing errors? */
1293 		return NULL;
1294 	ret = nfsdfs_create_files(dentry, files, ncl, fdentries);
1295 	if (ret) {
1296 		nfsd_client_rmdir(dentry);
1297 		return NULL;
1298 	}
1299 	return dentry;
1300 }
1301 
1302 /* Taken from __rpc_rmdir: */
1303 void nfsd_client_rmdir(struct dentry *dentry)
1304 {
1305 	simple_recursive_removal(dentry, clear_ncl);
1306 }
1307 
1308 static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1309 {
1310 	struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1311 							nfsd_net_id);
1312 	struct dentry *dentry;
1313 	int ret;
1314 
1315 	static const struct tree_descr nfsd_files[] = {
1316 		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1317 		/* Per-export io stats use same ops as exports file */
1318 		[NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1319 		[NFSD_Export_features] = {"export_features",
1320 					&export_features_fops, S_IRUGO},
1321 		[NFSD_FO_UnlockIP] = {"unlock_ip",
1322 					&transaction_ops, S_IWUSR|S_IRUSR},
1323 		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1324 					&transaction_ops, S_IWUSR|S_IRUSR},
1325 		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1326 		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1327 		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1328 		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1329 		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1330 					&nfsd_reply_cache_stats_fops, S_IRUGO},
1331 		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1332 		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1333 		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1334 		[NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
1335 #ifdef CONFIG_NFSD_V4
1336 		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1337 		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1338 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1339 		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1340 #endif
1341 		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1342 #endif
1343 		/* last one */ {""}
1344 	};
1345 
1346 	ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1347 	if (ret)
1348 		return ret;
1349 	_nfsd_symlink(sb->s_root, "supported_krb5_enctypes",
1350 		      "/proc/net/rpc/gss_krb5_enctypes");
1351 	dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1352 	if (IS_ERR(dentry))
1353 		return PTR_ERR(dentry);
1354 	nn->nfsd_client_dir = dentry;
1355 	return 0;
1356 }
1357 
1358 static int nfsd_fs_get_tree(struct fs_context *fc)
1359 {
1360 	return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1361 }
1362 
1363 static void nfsd_fs_free_fc(struct fs_context *fc)
1364 {
1365 	if (fc->s_fs_info)
1366 		put_net(fc->s_fs_info);
1367 }
1368 
1369 static const struct fs_context_operations nfsd_fs_context_ops = {
1370 	.free		= nfsd_fs_free_fc,
1371 	.get_tree	= nfsd_fs_get_tree,
1372 };
1373 
1374 static int nfsd_init_fs_context(struct fs_context *fc)
1375 {
1376 	put_user_ns(fc->user_ns);
1377 	fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1378 	fc->ops = &nfsd_fs_context_ops;
1379 	return 0;
1380 }
1381 
1382 static void nfsd_umount(struct super_block *sb)
1383 {
1384 	struct net *net = sb->s_fs_info;
1385 
1386 	nfsd_shutdown_threads(net);
1387 
1388 	kill_litter_super(sb);
1389 	put_net(net);
1390 }
1391 
1392 static struct file_system_type nfsd_fs_type = {
1393 	.owner		= THIS_MODULE,
1394 	.name		= "nfsd",
1395 	.init_fs_context = nfsd_init_fs_context,
1396 	.kill_sb	= nfsd_umount,
1397 };
1398 MODULE_ALIAS_FS("nfsd");
1399 
1400 #ifdef CONFIG_PROC_FS
1401 
1402 static int exports_proc_open(struct inode *inode, struct file *file)
1403 {
1404 	return exports_net_open(current->nsproxy->net_ns, file);
1405 }
1406 
1407 static const struct proc_ops exports_proc_ops = {
1408 	.proc_open	= exports_proc_open,
1409 	.proc_read	= seq_read,
1410 	.proc_lseek	= seq_lseek,
1411 	.proc_release	= seq_release,
1412 };
1413 
1414 static int create_proc_exports_entry(void)
1415 {
1416 	struct proc_dir_entry *entry;
1417 
1418 	entry = proc_mkdir("fs/nfs", NULL);
1419 	if (!entry)
1420 		return -ENOMEM;
1421 	entry = proc_create("exports", 0, entry, &exports_proc_ops);
1422 	if (!entry) {
1423 		remove_proc_entry("fs/nfs", NULL);
1424 		return -ENOMEM;
1425 	}
1426 	return 0;
1427 }
1428 #else /* CONFIG_PROC_FS */
1429 static int create_proc_exports_entry(void)
1430 {
1431 	return 0;
1432 }
1433 #endif
1434 
1435 unsigned int nfsd_net_id;
1436 
1437 static int nfsd_genl_rpc_status_compose_msg(struct sk_buff *skb,
1438 					    struct netlink_callback *cb,
1439 					    struct nfsd_genl_rqstp *rqstp)
1440 {
1441 	void *hdr;
1442 	u32 i;
1443 
1444 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
1445 			  &nfsd_nl_family, 0, NFSD_CMD_RPC_STATUS_GET);
1446 	if (!hdr)
1447 		return -ENOBUFS;
1448 
1449 	if (nla_put_be32(skb, NFSD_A_RPC_STATUS_XID, rqstp->rq_xid) ||
1450 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_FLAGS, rqstp->rq_flags) ||
1451 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROG, rqstp->rq_prog) ||
1452 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROC, rqstp->rq_proc) ||
1453 	    nla_put_u8(skb, NFSD_A_RPC_STATUS_VERSION, rqstp->rq_vers) ||
1454 	    nla_put_s64(skb, NFSD_A_RPC_STATUS_SERVICE_TIME,
1455 			ktime_to_us(rqstp->rq_stime),
1456 			NFSD_A_RPC_STATUS_PAD))
1457 		return -ENOBUFS;
1458 
1459 	switch (rqstp->rq_saddr.sa_family) {
1460 	case AF_INET: {
1461 		const struct sockaddr_in *s_in, *d_in;
1462 
1463 		s_in = (const struct sockaddr_in *)&rqstp->rq_saddr;
1464 		d_in = (const struct sockaddr_in *)&rqstp->rq_daddr;
1465 		if (nla_put_in_addr(skb, NFSD_A_RPC_STATUS_SADDR4,
1466 				    s_in->sin_addr.s_addr) ||
1467 		    nla_put_in_addr(skb, NFSD_A_RPC_STATUS_DADDR4,
1468 				    d_in->sin_addr.s_addr) ||
1469 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1470 				 s_in->sin_port) ||
1471 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1472 				 d_in->sin_port))
1473 			return -ENOBUFS;
1474 		break;
1475 	}
1476 	case AF_INET6: {
1477 		const struct sockaddr_in6 *s_in, *d_in;
1478 
1479 		s_in = (const struct sockaddr_in6 *)&rqstp->rq_saddr;
1480 		d_in = (const struct sockaddr_in6 *)&rqstp->rq_daddr;
1481 		if (nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_SADDR6,
1482 				     &s_in->sin6_addr) ||
1483 		    nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_DADDR6,
1484 				     &d_in->sin6_addr) ||
1485 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1486 				 s_in->sin6_port) ||
1487 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1488 				 d_in->sin6_port))
1489 			return -ENOBUFS;
1490 		break;
1491 	}
1492 	}
1493 
1494 	for (i = 0; i < rqstp->rq_opcnt; i++)
1495 		if (nla_put_u32(skb, NFSD_A_RPC_STATUS_COMPOUND_OPS,
1496 				rqstp->rq_opnum[i]))
1497 			return -ENOBUFS;
1498 
1499 	genlmsg_end(skb, hdr);
1500 	return 0;
1501 }
1502 
1503 /**
1504  * nfsd_nl_rpc_status_get_dumpit - Handle rpc_status_get dumpit
1505  * @skb: reply buffer
1506  * @cb: netlink metadata and command arguments
1507  *
1508  * Returns the size of the reply or a negative errno.
1509  */
1510 int nfsd_nl_rpc_status_get_dumpit(struct sk_buff *skb,
1511 				  struct netlink_callback *cb)
1512 {
1513 	int i, ret, rqstp_index = 0;
1514 	struct nfsd_net *nn;
1515 
1516 	mutex_lock(&nfsd_mutex);
1517 
1518 	nn = net_generic(sock_net(skb->sk), nfsd_net_id);
1519 	if (!nn->nfsd_serv) {
1520 		ret = -ENODEV;
1521 		goto out_unlock;
1522 	}
1523 
1524 	rcu_read_lock();
1525 
1526 	for (i = 0; i < nn->nfsd_serv->sv_nrpools; i++) {
1527 		struct svc_rqst *rqstp;
1528 
1529 		if (i < cb->args[0]) /* already consumed */
1530 			continue;
1531 
1532 		rqstp_index = 0;
1533 		list_for_each_entry_rcu(rqstp,
1534 				&nn->nfsd_serv->sv_pools[i].sp_all_threads,
1535 				rq_all) {
1536 			struct nfsd_genl_rqstp genl_rqstp;
1537 			unsigned int status_counter;
1538 
1539 			if (rqstp_index++ < cb->args[1]) /* already consumed */
1540 				continue;
1541 			/*
1542 			 * Acquire rq_status_counter before parsing the rqst
1543 			 * fields. rq_status_counter is set to an odd value in
1544 			 * order to notify the consumers the rqstp fields are
1545 			 * meaningful.
1546 			 */
1547 			status_counter =
1548 				smp_load_acquire(&rqstp->rq_status_counter);
1549 			if (!(status_counter & 1))
1550 				continue;
1551 
1552 			genl_rqstp.rq_xid = rqstp->rq_xid;
1553 			genl_rqstp.rq_flags = rqstp->rq_flags;
1554 			genl_rqstp.rq_vers = rqstp->rq_vers;
1555 			genl_rqstp.rq_prog = rqstp->rq_prog;
1556 			genl_rqstp.rq_proc = rqstp->rq_proc;
1557 			genl_rqstp.rq_stime = rqstp->rq_stime;
1558 			genl_rqstp.rq_opcnt = 0;
1559 			memcpy(&genl_rqstp.rq_daddr, svc_daddr(rqstp),
1560 			       sizeof(struct sockaddr));
1561 			memcpy(&genl_rqstp.rq_saddr, svc_addr(rqstp),
1562 			       sizeof(struct sockaddr));
1563 
1564 #ifdef CONFIG_NFSD_V4
1565 			if (rqstp->rq_vers == NFS4_VERSION &&
1566 			    rqstp->rq_proc == NFSPROC4_COMPOUND) {
1567 				/* NFSv4 compound */
1568 				struct nfsd4_compoundargs *args;
1569 				int j;
1570 
1571 				args = rqstp->rq_argp;
1572 				genl_rqstp.rq_opcnt = args->opcnt;
1573 				for (j = 0; j < genl_rqstp.rq_opcnt; j++)
1574 					genl_rqstp.rq_opnum[j] =
1575 						args->ops[j].opnum;
1576 			}
1577 #endif /* CONFIG_NFSD_V4 */
1578 
1579 			/*
1580 			 * Acquire rq_status_counter before reporting the rqst
1581 			 * fields to the user.
1582 			 */
1583 			if (smp_load_acquire(&rqstp->rq_status_counter) !=
1584 			    status_counter)
1585 				continue;
1586 
1587 			ret = nfsd_genl_rpc_status_compose_msg(skb, cb,
1588 							       &genl_rqstp);
1589 			if (ret)
1590 				goto out;
1591 		}
1592 	}
1593 
1594 	cb->args[0] = i;
1595 	cb->args[1] = rqstp_index;
1596 	ret = skb->len;
1597 out:
1598 	rcu_read_unlock();
1599 out_unlock:
1600 	mutex_unlock(&nfsd_mutex);
1601 
1602 	return ret;
1603 }
1604 
1605 /**
1606  * nfsd_nl_threads_set_doit - set the number of running threads
1607  * @skb: reply buffer
1608  * @info: netlink metadata and command arguments
1609  *
1610  * Return 0 on success or a negative errno.
1611  */
1612 int nfsd_nl_threads_set_doit(struct sk_buff *skb, struct genl_info *info)
1613 {
1614 	int *nthreads, count = 0, nrpools, i, ret = -EOPNOTSUPP, rem;
1615 	struct net *net = genl_info_net(info);
1616 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1617 	const struct nlattr *attr;
1618 	const char *scope = NULL;
1619 
1620 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_SERVER_THREADS))
1621 		return -EINVAL;
1622 
1623 	/* count number of SERVER_THREADS values */
1624 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1625 		if (nla_type(attr) == NFSD_A_SERVER_THREADS)
1626 			count++;
1627 	}
1628 
1629 	mutex_lock(&nfsd_mutex);
1630 
1631 	nrpools = max(count, nfsd_nrpools(net));
1632 	nthreads = kcalloc(nrpools, sizeof(int), GFP_KERNEL);
1633 	if (!nthreads) {
1634 		ret = -ENOMEM;
1635 		goto out_unlock;
1636 	}
1637 
1638 	i = 0;
1639 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1640 		if (nla_type(attr) == NFSD_A_SERVER_THREADS) {
1641 			nthreads[i++] = nla_get_u32(attr);
1642 			if (i >= nrpools)
1643 				break;
1644 		}
1645 	}
1646 
1647 	if (info->attrs[NFSD_A_SERVER_GRACETIME] ||
1648 	    info->attrs[NFSD_A_SERVER_LEASETIME] ||
1649 	    info->attrs[NFSD_A_SERVER_SCOPE]) {
1650 		ret = -EBUSY;
1651 		if (nn->nfsd_serv && nn->nfsd_serv->sv_nrthreads)
1652 			goto out_unlock;
1653 
1654 		ret = -EINVAL;
1655 		attr = info->attrs[NFSD_A_SERVER_GRACETIME];
1656 		if (attr) {
1657 			u32 gracetime = nla_get_u32(attr);
1658 
1659 			if (gracetime < 10 || gracetime > 3600)
1660 				goto out_unlock;
1661 
1662 			nn->nfsd4_grace = gracetime;
1663 		}
1664 
1665 		attr = info->attrs[NFSD_A_SERVER_LEASETIME];
1666 		if (attr) {
1667 			u32 leasetime = nla_get_u32(attr);
1668 
1669 			if (leasetime < 10 || leasetime > 3600)
1670 				goto out_unlock;
1671 
1672 			nn->nfsd4_lease = leasetime;
1673 		}
1674 
1675 		attr = info->attrs[NFSD_A_SERVER_SCOPE];
1676 		if (attr)
1677 			scope = nla_data(attr);
1678 	}
1679 
1680 	ret = nfsd_svc(nrpools, nthreads, net, get_current_cred(), scope);
1681 	if (ret > 0)
1682 		ret = 0;
1683 out_unlock:
1684 	mutex_unlock(&nfsd_mutex);
1685 	kfree(nthreads);
1686 	return ret;
1687 }
1688 
1689 /**
1690  * nfsd_nl_threads_get_doit - get the number of running threads
1691  * @skb: reply buffer
1692  * @info: netlink metadata and command arguments
1693  *
1694  * Return 0 on success or a negative errno.
1695  */
1696 int nfsd_nl_threads_get_doit(struct sk_buff *skb, struct genl_info *info)
1697 {
1698 	struct net *net = genl_info_net(info);
1699 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1700 	void *hdr;
1701 	int err;
1702 
1703 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
1704 	if (!skb)
1705 		return -ENOMEM;
1706 
1707 	hdr = genlmsg_iput(skb, info);
1708 	if (!hdr) {
1709 		err = -EMSGSIZE;
1710 		goto err_free_msg;
1711 	}
1712 
1713 	mutex_lock(&nfsd_mutex);
1714 
1715 	err = nla_put_u32(skb, NFSD_A_SERVER_GRACETIME,
1716 			  nn->nfsd4_grace) ||
1717 	      nla_put_u32(skb, NFSD_A_SERVER_LEASETIME,
1718 			  nn->nfsd4_lease) ||
1719 	      nla_put_string(skb, NFSD_A_SERVER_SCOPE,
1720 			  nn->nfsd_name);
1721 	if (err)
1722 		goto err_unlock;
1723 
1724 	if (nn->nfsd_serv) {
1725 		int i;
1726 
1727 		for (i = 0; i < nfsd_nrpools(net); ++i) {
1728 			struct svc_pool *sp = &nn->nfsd_serv->sv_pools[i];
1729 
1730 			err = nla_put_u32(skb, NFSD_A_SERVER_THREADS,
1731 					  sp->sp_nrthreads);
1732 			if (err)
1733 				goto err_unlock;
1734 		}
1735 	} else {
1736 		err = nla_put_u32(skb, NFSD_A_SERVER_THREADS, 0);
1737 		if (err)
1738 			goto err_unlock;
1739 	}
1740 
1741 	mutex_unlock(&nfsd_mutex);
1742 
1743 	genlmsg_end(skb, hdr);
1744 
1745 	return genlmsg_reply(skb, info);
1746 
1747 err_unlock:
1748 	mutex_unlock(&nfsd_mutex);
1749 err_free_msg:
1750 	nlmsg_free(skb);
1751 
1752 	return err;
1753 }
1754 
1755 /**
1756  * nfsd_nl_version_set_doit - set the nfs enabled versions
1757  * @skb: reply buffer
1758  * @info: netlink metadata and command arguments
1759  *
1760  * Return 0 on success or a negative errno.
1761  */
1762 int nfsd_nl_version_set_doit(struct sk_buff *skb, struct genl_info *info)
1763 {
1764 	const struct nlattr *attr;
1765 	struct nfsd_net *nn;
1766 	int i, rem;
1767 
1768 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_SERVER_PROTO_VERSION))
1769 		return -EINVAL;
1770 
1771 	mutex_lock(&nfsd_mutex);
1772 
1773 	nn = net_generic(genl_info_net(info), nfsd_net_id);
1774 	if (nn->nfsd_serv) {
1775 		mutex_unlock(&nfsd_mutex);
1776 		return -EBUSY;
1777 	}
1778 
1779 	/* clear current supported versions. */
1780 	nfsd_vers(nn, 2, NFSD_CLEAR);
1781 	nfsd_vers(nn, 3, NFSD_CLEAR);
1782 	for (i = 0; i <= NFSD_SUPPORTED_MINOR_VERSION; i++)
1783 		nfsd_minorversion(nn, i, NFSD_CLEAR);
1784 
1785 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1786 		struct nlattr *tb[NFSD_A_VERSION_MAX + 1];
1787 		u32 major, minor = 0;
1788 		bool enabled;
1789 
1790 		if (nla_type(attr) != NFSD_A_SERVER_PROTO_VERSION)
1791 			continue;
1792 
1793 		if (nla_parse_nested(tb, NFSD_A_VERSION_MAX, attr,
1794 				     nfsd_version_nl_policy, info->extack) < 0)
1795 			continue;
1796 
1797 		if (!tb[NFSD_A_VERSION_MAJOR])
1798 			continue;
1799 
1800 		major = nla_get_u32(tb[NFSD_A_VERSION_MAJOR]);
1801 		if (tb[NFSD_A_VERSION_MINOR])
1802 			minor = nla_get_u32(tb[NFSD_A_VERSION_MINOR]);
1803 
1804 		enabled = nla_get_flag(tb[NFSD_A_VERSION_ENABLED]);
1805 
1806 		switch (major) {
1807 		case 4:
1808 			nfsd_minorversion(nn, minor, enabled ? NFSD_SET : NFSD_CLEAR);
1809 			break;
1810 		case 3:
1811 		case 2:
1812 			if (!minor)
1813 				nfsd_vers(nn, major, enabled ? NFSD_SET : NFSD_CLEAR);
1814 			break;
1815 		default:
1816 			break;
1817 		}
1818 	}
1819 
1820 	mutex_unlock(&nfsd_mutex);
1821 
1822 	return 0;
1823 }
1824 
1825 /**
1826  * nfsd_nl_version_get_doit - get the enabled status for all supported nfs versions
1827  * @skb: reply buffer
1828  * @info: netlink metadata and command arguments
1829  *
1830  * Return 0 on success or a negative errno.
1831  */
1832 int nfsd_nl_version_get_doit(struct sk_buff *skb, struct genl_info *info)
1833 {
1834 	struct nfsd_net *nn;
1835 	int i, err;
1836 	void *hdr;
1837 
1838 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
1839 	if (!skb)
1840 		return -ENOMEM;
1841 
1842 	hdr = genlmsg_iput(skb, info);
1843 	if (!hdr) {
1844 		err = -EMSGSIZE;
1845 		goto err_free_msg;
1846 	}
1847 
1848 	mutex_lock(&nfsd_mutex);
1849 	nn = net_generic(genl_info_net(info), nfsd_net_id);
1850 
1851 	for (i = 2; i <= 4; i++) {
1852 		int j;
1853 
1854 		for (j = 0; j <= NFSD_SUPPORTED_MINOR_VERSION; j++) {
1855 			struct nlattr *attr;
1856 
1857 			/* Don't record any versions the kernel doesn't have
1858 			 * compiled in
1859 			 */
1860 			if (!nfsd_support_version(i))
1861 				continue;
1862 
1863 			/* NFSv{2,3} does not support minor numbers */
1864 			if (i < 4 && j)
1865 				continue;
1866 
1867 			attr = nla_nest_start(skb,
1868 					      NFSD_A_SERVER_PROTO_VERSION);
1869 			if (!attr) {
1870 				err = -EINVAL;
1871 				goto err_nfsd_unlock;
1872 			}
1873 
1874 			if (nla_put_u32(skb, NFSD_A_VERSION_MAJOR, i) ||
1875 			    nla_put_u32(skb, NFSD_A_VERSION_MINOR, j)) {
1876 				err = -EINVAL;
1877 				goto err_nfsd_unlock;
1878 			}
1879 
1880 			/* Set the enabled flag if the version is enabled */
1881 			if (nfsd_vers(nn, i, NFSD_TEST) &&
1882 			    (i < 4 || nfsd_minorversion(nn, j, NFSD_TEST)) &&
1883 			    nla_put_flag(skb, NFSD_A_VERSION_ENABLED)) {
1884 				err = -EINVAL;
1885 				goto err_nfsd_unlock;
1886 			}
1887 
1888 			nla_nest_end(skb, attr);
1889 		}
1890 	}
1891 
1892 	mutex_unlock(&nfsd_mutex);
1893 	genlmsg_end(skb, hdr);
1894 
1895 	return genlmsg_reply(skb, info);
1896 
1897 err_nfsd_unlock:
1898 	mutex_unlock(&nfsd_mutex);
1899 err_free_msg:
1900 	nlmsg_free(skb);
1901 
1902 	return err;
1903 }
1904 
1905 /**
1906  * nfsd_nl_listener_set_doit - set the nfs running sockets
1907  * @skb: reply buffer
1908  * @info: netlink metadata and command arguments
1909  *
1910  * Return 0 on success or a negative errno.
1911  */
1912 int nfsd_nl_listener_set_doit(struct sk_buff *skb, struct genl_info *info)
1913 {
1914 	struct net *net = genl_info_net(info);
1915 	struct svc_xprt *xprt, *tmp;
1916 	const struct nlattr *attr;
1917 	struct svc_serv *serv;
1918 	LIST_HEAD(permsocks);
1919 	struct nfsd_net *nn;
1920 	int err, rem;
1921 
1922 	mutex_lock(&nfsd_mutex);
1923 
1924 	err = nfsd_create_serv(net);
1925 	if (err) {
1926 		mutex_unlock(&nfsd_mutex);
1927 		return err;
1928 	}
1929 
1930 	nn = net_generic(net, nfsd_net_id);
1931 	serv = nn->nfsd_serv;
1932 
1933 	spin_lock_bh(&serv->sv_lock);
1934 
1935 	/* Move all of the old listener sockets to a temp list */
1936 	list_splice_init(&serv->sv_permsocks, &permsocks);
1937 
1938 	/*
1939 	 * Walk the list of server_socks from userland and move any that match
1940 	 * back to sv_permsocks
1941 	 */
1942 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1943 		struct nlattr *tb[NFSD_A_SOCK_MAX + 1];
1944 		const char *xcl_name;
1945 		struct sockaddr *sa;
1946 
1947 		if (nla_type(attr) != NFSD_A_SERVER_SOCK_ADDR)
1948 			continue;
1949 
1950 		if (nla_parse_nested(tb, NFSD_A_SOCK_MAX, attr,
1951 				     nfsd_sock_nl_policy, info->extack) < 0)
1952 			continue;
1953 
1954 		if (!tb[NFSD_A_SOCK_ADDR] || !tb[NFSD_A_SOCK_TRANSPORT_NAME])
1955 			continue;
1956 
1957 		if (nla_len(tb[NFSD_A_SOCK_ADDR]) < sizeof(*sa))
1958 			continue;
1959 
1960 		xcl_name = nla_data(tb[NFSD_A_SOCK_TRANSPORT_NAME]);
1961 		sa = nla_data(tb[NFSD_A_SOCK_ADDR]);
1962 
1963 		/* Put back any matching sockets */
1964 		list_for_each_entry_safe(xprt, tmp, &permsocks, xpt_list) {
1965 			/* This shouldn't be possible */
1966 			if (WARN_ON_ONCE(xprt->xpt_net != net)) {
1967 				list_move(&xprt->xpt_list, &serv->sv_permsocks);
1968 				continue;
1969 			}
1970 
1971 			/* If everything matches, put it back */
1972 			if (!strcmp(xprt->xpt_class->xcl_name, xcl_name) &&
1973 			    rpc_cmp_addr_port(sa, (struct sockaddr *)&xprt->xpt_local)) {
1974 				list_move(&xprt->xpt_list, &serv->sv_permsocks);
1975 				break;
1976 			}
1977 		}
1978 	}
1979 
1980 	/* For now, no removing old sockets while server is running */
1981 	if (serv->sv_nrthreads && !list_empty(&permsocks)) {
1982 		list_splice_init(&permsocks, &serv->sv_permsocks);
1983 		spin_unlock_bh(&serv->sv_lock);
1984 		err = -EBUSY;
1985 		goto out_unlock_mtx;
1986 	}
1987 
1988 	/* Close the remaining sockets on the permsocks list */
1989 	while (!list_empty(&permsocks)) {
1990 		xprt = list_first_entry(&permsocks, struct svc_xprt, xpt_list);
1991 		list_move(&xprt->xpt_list, &serv->sv_permsocks);
1992 
1993 		/*
1994 		 * Newly-created sockets are born with the BUSY bit set. Clear
1995 		 * it if there are no threads, since nothing can pick it up
1996 		 * in that case.
1997 		 */
1998 		if (!serv->sv_nrthreads)
1999 			clear_bit(XPT_BUSY, &xprt->xpt_flags);
2000 
2001 		set_bit(XPT_CLOSE, &xprt->xpt_flags);
2002 		spin_unlock_bh(&serv->sv_lock);
2003 		svc_xprt_close(xprt);
2004 		spin_lock_bh(&serv->sv_lock);
2005 	}
2006 
2007 	spin_unlock_bh(&serv->sv_lock);
2008 
2009 	/* walk list of addrs again, open any that still don't exist */
2010 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
2011 		struct nlattr *tb[NFSD_A_SOCK_MAX + 1];
2012 		const char *xcl_name;
2013 		struct sockaddr *sa;
2014 		int ret;
2015 
2016 		if (nla_type(attr) != NFSD_A_SERVER_SOCK_ADDR)
2017 			continue;
2018 
2019 		if (nla_parse_nested(tb, NFSD_A_SOCK_MAX, attr,
2020 				     nfsd_sock_nl_policy, info->extack) < 0)
2021 			continue;
2022 
2023 		if (!tb[NFSD_A_SOCK_ADDR] || !tb[NFSD_A_SOCK_TRANSPORT_NAME])
2024 			continue;
2025 
2026 		if (nla_len(tb[NFSD_A_SOCK_ADDR]) < sizeof(*sa))
2027 			continue;
2028 
2029 		xcl_name = nla_data(tb[NFSD_A_SOCK_TRANSPORT_NAME]);
2030 		sa = nla_data(tb[NFSD_A_SOCK_ADDR]);
2031 
2032 		xprt = svc_find_listener(serv, xcl_name, net, sa);
2033 		if (xprt) {
2034 			svc_xprt_put(xprt);
2035 			continue;
2036 		}
2037 
2038 		ret = svc_xprt_create_from_sa(serv, xcl_name, net, sa, 0,
2039 					      get_current_cred());
2040 		/* always save the latest error */
2041 		if (ret < 0)
2042 			err = ret;
2043 	}
2044 
2045 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
2046 		nfsd_destroy_serv(net);
2047 
2048 out_unlock_mtx:
2049 	mutex_unlock(&nfsd_mutex);
2050 
2051 	return err;
2052 }
2053 
2054 /**
2055  * nfsd_nl_listener_get_doit - get the nfs running listeners
2056  * @skb: reply buffer
2057  * @info: netlink metadata and command arguments
2058  *
2059  * Return 0 on success or a negative errno.
2060  */
2061 int nfsd_nl_listener_get_doit(struct sk_buff *skb, struct genl_info *info)
2062 {
2063 	struct svc_xprt *xprt;
2064 	struct svc_serv *serv;
2065 	struct nfsd_net *nn;
2066 	void *hdr;
2067 	int err;
2068 
2069 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
2070 	if (!skb)
2071 		return -ENOMEM;
2072 
2073 	hdr = genlmsg_iput(skb, info);
2074 	if (!hdr) {
2075 		err = -EMSGSIZE;
2076 		goto err_free_msg;
2077 	}
2078 
2079 	mutex_lock(&nfsd_mutex);
2080 	nn = net_generic(genl_info_net(info), nfsd_net_id);
2081 
2082 	/* no nfs server? Just send empty socket list */
2083 	if (!nn->nfsd_serv)
2084 		goto out_unlock_mtx;
2085 
2086 	serv = nn->nfsd_serv;
2087 	spin_lock_bh(&serv->sv_lock);
2088 	list_for_each_entry(xprt, &serv->sv_permsocks, xpt_list) {
2089 		struct nlattr *attr;
2090 
2091 		attr = nla_nest_start(skb, NFSD_A_SERVER_SOCK_ADDR);
2092 		if (!attr) {
2093 			err = -EINVAL;
2094 			goto err_serv_unlock;
2095 		}
2096 
2097 		if (nla_put_string(skb, NFSD_A_SOCK_TRANSPORT_NAME,
2098 				   xprt->xpt_class->xcl_name) ||
2099 		    nla_put(skb, NFSD_A_SOCK_ADDR,
2100 			    sizeof(struct sockaddr_storage),
2101 			    &xprt->xpt_local)) {
2102 			err = -EINVAL;
2103 			goto err_serv_unlock;
2104 		}
2105 
2106 		nla_nest_end(skb, attr);
2107 	}
2108 	spin_unlock_bh(&serv->sv_lock);
2109 out_unlock_mtx:
2110 	mutex_unlock(&nfsd_mutex);
2111 	genlmsg_end(skb, hdr);
2112 
2113 	return genlmsg_reply(skb, info);
2114 
2115 err_serv_unlock:
2116 	spin_unlock_bh(&serv->sv_lock);
2117 	mutex_unlock(&nfsd_mutex);
2118 err_free_msg:
2119 	nlmsg_free(skb);
2120 
2121 	return err;
2122 }
2123 
2124 /**
2125  * nfsd_nl_pool_mode_set_doit - set the number of running threads
2126  * @skb: reply buffer
2127  * @info: netlink metadata and command arguments
2128  *
2129  * Return 0 on success or a negative errno.
2130  */
2131 int nfsd_nl_pool_mode_set_doit(struct sk_buff *skb, struct genl_info *info)
2132 {
2133 	const struct nlattr *attr;
2134 
2135 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_POOL_MODE_MODE))
2136 		return -EINVAL;
2137 
2138 	attr = info->attrs[NFSD_A_POOL_MODE_MODE];
2139 	return sunrpc_set_pool_mode(nla_data(attr));
2140 }
2141 
2142 /**
2143  * nfsd_nl_pool_mode_get_doit - get info about pool_mode
2144  * @skb: reply buffer
2145  * @info: netlink metadata and command arguments
2146  *
2147  * Return 0 on success or a negative errno.
2148  */
2149 int nfsd_nl_pool_mode_get_doit(struct sk_buff *skb, struct genl_info *info)
2150 {
2151 	struct net *net = genl_info_net(info);
2152 	char buf[16];
2153 	void *hdr;
2154 	int err;
2155 
2156 	if (sunrpc_get_pool_mode(buf, ARRAY_SIZE(buf)) >= ARRAY_SIZE(buf))
2157 		return -ERANGE;
2158 
2159 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
2160 	if (!skb)
2161 		return -ENOMEM;
2162 
2163 	err = -EMSGSIZE;
2164 	hdr = genlmsg_iput(skb, info);
2165 	if (!hdr)
2166 		goto err_free_msg;
2167 
2168 	err = nla_put_string(skb, NFSD_A_POOL_MODE_MODE, buf) |
2169 	      nla_put_u32(skb, NFSD_A_POOL_MODE_NPOOLS, nfsd_nrpools(net));
2170 	if (err)
2171 		goto err_free_msg;
2172 
2173 	genlmsg_end(skb, hdr);
2174 	return genlmsg_reply(skb, info);
2175 
2176 err_free_msg:
2177 	nlmsg_free(skb);
2178 	return err;
2179 }
2180 
2181 /**
2182  * nfsd_net_init - Prepare the nfsd_net portion of a new net namespace
2183  * @net: a freshly-created network namespace
2184  *
2185  * This information stays around as long as the network namespace is
2186  * alive whether or not there is an NFSD instance running in the
2187  * namespace.
2188  *
2189  * Returns zero on success, or a negative errno otherwise.
2190  */
2191 static __net_init int nfsd_net_init(struct net *net)
2192 {
2193 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2194 	int retval;
2195 	int i;
2196 
2197 	retval = nfsd_export_init(net);
2198 	if (retval)
2199 		goto out_export_error;
2200 	retval = nfsd_idmap_init(net);
2201 	if (retval)
2202 		goto out_idmap_error;
2203 	retval = percpu_counter_init_many(nn->counter, 0, GFP_KERNEL,
2204 					  NFSD_STATS_COUNTERS_NUM);
2205 	if (retval)
2206 		goto out_repcache_error;
2207 	memset(&nn->nfsd_svcstats, 0, sizeof(nn->nfsd_svcstats));
2208 	nn->nfsd_svcstats.program = &nfsd_programs[0];
2209 	for (i = 0; i < sizeof(nn->nfsd_versions); i++)
2210 		nn->nfsd_versions[i] = nfsd_support_version(i);
2211 	for (i = 0; i < sizeof(nn->nfsd4_minorversions); i++)
2212 		nn->nfsd4_minorversions[i] = nfsd_support_version(4);
2213 	nn->nfsd_info.mutex = &nfsd_mutex;
2214 	nn->nfsd_serv = NULL;
2215 	nfsd4_init_leases_net(nn);
2216 	get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
2217 	seqlock_init(&nn->writeverf_lock);
2218 	nfsd_proc_stat_init(net);
2219 #if IS_ENABLED(CONFIG_NFS_LOCALIO)
2220 	spin_lock_init(&nn->local_clients_lock);
2221 	INIT_LIST_HEAD(&nn->local_clients);
2222 #endif
2223 	return 0;
2224 
2225 out_repcache_error:
2226 	nfsd_idmap_shutdown(net);
2227 out_idmap_error:
2228 	nfsd_export_shutdown(net);
2229 out_export_error:
2230 	return retval;
2231 }
2232 
2233 #if IS_ENABLED(CONFIG_NFS_LOCALIO)
2234 /**
2235  * nfsd_net_pre_exit - Disconnect localio clients from net namespace
2236  * @net: a network namespace that is about to be destroyed
2237  *
2238  * This invalidates ->net pointers held by localio clients
2239  * while they can still safely access nn->counter.
2240  */
2241 static __net_exit void nfsd_net_pre_exit(struct net *net)
2242 {
2243 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2244 
2245 	nfs_localio_invalidate_clients(&nn->local_clients,
2246 				       &nn->local_clients_lock);
2247 }
2248 #endif
2249 
2250 /**
2251  * nfsd_net_exit - Release the nfsd_net portion of a net namespace
2252  * @net: a network namespace that is about to be destroyed
2253  *
2254  */
2255 static __net_exit void nfsd_net_exit(struct net *net)
2256 {
2257 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2258 
2259 	nfsd_proc_stat_shutdown(net);
2260 	percpu_counter_destroy_many(nn->counter, NFSD_STATS_COUNTERS_NUM);
2261 	nfsd_idmap_shutdown(net);
2262 	nfsd_export_shutdown(net);
2263 }
2264 
2265 static struct pernet_operations nfsd_net_ops = {
2266 	.init = nfsd_net_init,
2267 #if IS_ENABLED(CONFIG_NFS_LOCALIO)
2268 	.pre_exit = nfsd_net_pre_exit,
2269 #endif
2270 	.exit = nfsd_net_exit,
2271 	.id   = &nfsd_net_id,
2272 	.size = sizeof(struct nfsd_net),
2273 };
2274 
2275 static int __init init_nfsd(void)
2276 {
2277 	int retval;
2278 
2279 	retval = nfsd4_init_slabs();
2280 	if (retval)
2281 		return retval;
2282 	retval = nfsd4_init_pnfs();
2283 	if (retval)
2284 		goto out_free_slabs;
2285 	retval = nfsd_drc_slab_create();
2286 	if (retval)
2287 		goto out_free_pnfs;
2288 	nfsd_lockd_init();	/* lockd->nfsd callbacks */
2289 	retval = create_proc_exports_entry();
2290 	if (retval)
2291 		goto out_free_lockd;
2292 	retval = register_pernet_subsys(&nfsd_net_ops);
2293 	if (retval < 0)
2294 		goto out_free_exports;
2295 	retval = register_cld_notifier();
2296 	if (retval)
2297 		goto out_free_subsys;
2298 	retval = nfsd4_create_laundry_wq();
2299 	if (retval)
2300 		goto out_free_cld;
2301 	retval = register_filesystem(&nfsd_fs_type);
2302 	if (retval)
2303 		goto out_free_all;
2304 	retval = genl_register_family(&nfsd_nl_family);
2305 	if (retval)
2306 		goto out_free_all;
2307 	nfsd_localio_ops_init();
2308 
2309 	return 0;
2310 out_free_all:
2311 	nfsd4_destroy_laundry_wq();
2312 out_free_cld:
2313 	unregister_cld_notifier();
2314 out_free_subsys:
2315 	unregister_pernet_subsys(&nfsd_net_ops);
2316 out_free_exports:
2317 	remove_proc_entry("fs/nfs/exports", NULL);
2318 	remove_proc_entry("fs/nfs", NULL);
2319 out_free_lockd:
2320 	nfsd_lockd_shutdown();
2321 	nfsd_drc_slab_free();
2322 out_free_pnfs:
2323 	nfsd4_exit_pnfs();
2324 out_free_slabs:
2325 	nfsd4_free_slabs();
2326 	return retval;
2327 }
2328 
2329 static void __exit exit_nfsd(void)
2330 {
2331 	genl_unregister_family(&nfsd_nl_family);
2332 	unregister_filesystem(&nfsd_fs_type);
2333 	nfsd4_destroy_laundry_wq();
2334 	unregister_cld_notifier();
2335 	unregister_pernet_subsys(&nfsd_net_ops);
2336 	nfsd_drc_slab_free();
2337 	remove_proc_entry("fs/nfs/exports", NULL);
2338 	remove_proc_entry("fs/nfs", NULL);
2339 	nfsd_lockd_shutdown();
2340 	nfsd4_free_slabs();
2341 	nfsd4_exit_pnfs();
2342 }
2343 
2344 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2345 MODULE_DESCRIPTION("In-kernel NFS server");
2346 MODULE_LICENSE("GPL");
2347 module_init(init_nfsd)
2348 module_exit(exit_nfsd)
2349