xref: /linux/fs/nfsd/nfsctl.c (revision bf36793fa260cb68cc817f311f1f683788261796)
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 
22 #include "idmap.h"
23 #include "nfsd.h"
24 #include "cache.h"
25 #include "state.h"
26 #include "netns.h"
27 #include "pnfs.h"
28 #include "filecache.h"
29 #include "trace.h"
30 #include "netlink.h"
31 
32 /*
33  *	We have a single directory with several nodes in it.
34  */
35 enum {
36 	NFSD_Root = 1,
37 	NFSD_List,
38 	NFSD_Export_Stats,
39 	NFSD_Export_features,
40 	NFSD_Fh,
41 	NFSD_FO_UnlockIP,
42 	NFSD_FO_UnlockFS,
43 	NFSD_Threads,
44 	NFSD_Pool_Threads,
45 	NFSD_Pool_Stats,
46 	NFSD_Reply_Cache_Stats,
47 	NFSD_Versions,
48 	NFSD_Ports,
49 	NFSD_MaxBlkSize,
50 	NFSD_MaxConnections,
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 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
71 #ifdef CONFIG_NFSD_V4
72 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
73 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
74 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
75 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
76 #endif
77 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
78 #endif
79 
80 static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
81 	[NFSD_Fh] = write_filehandle,
82 	[NFSD_FO_UnlockIP] = write_unlock_ip,
83 	[NFSD_FO_UnlockFS] = write_unlock_fs,
84 	[NFSD_Threads] = write_threads,
85 	[NFSD_Pool_Threads] = write_pool_threads,
86 	[NFSD_Versions] = write_versions,
87 	[NFSD_Ports] = write_ports,
88 	[NFSD_MaxBlkSize] = write_maxblksize,
89 	[NFSD_MaxConnections] = write_maxconn,
90 #ifdef CONFIG_NFSD_V4
91 	[NFSD_Leasetime] = write_leasetime,
92 	[NFSD_Gracetime] = write_gracetime,
93 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
94 	[NFSD_RecoveryDir] = write_recoverydir,
95 #endif
96 	[NFSD_V4EndGrace] = write_v4_end_grace,
97 #endif
98 };
99 
100 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
101 {
102 	ino_t ino =  file_inode(file)->i_ino;
103 	char *data;
104 	ssize_t rv;
105 
106 	if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
107 		return -EINVAL;
108 
109 	data = simple_transaction_get(file, buf, size);
110 	if (IS_ERR(data))
111 		return PTR_ERR(data);
112 
113 	rv = write_op[ino](file, data, size);
114 	if (rv < 0)
115 		return rv;
116 
117 	simple_transaction_set(file, rv);
118 	return size;
119 }
120 
121 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
122 {
123 	if (! file->private_data) {
124 		/* An attempt to read a transaction file without writing
125 		 * causes a 0-byte write so that the file can return
126 		 * state information
127 		 */
128 		ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
129 		if (rv < 0)
130 			return rv;
131 	}
132 	return simple_transaction_read(file, buf, size, pos);
133 }
134 
135 static const struct file_operations transaction_ops = {
136 	.write		= nfsctl_transaction_write,
137 	.read		= nfsctl_transaction_read,
138 	.release	= simple_transaction_release,
139 	.llseek		= default_llseek,
140 };
141 
142 static int exports_net_open(struct net *net, struct file *file)
143 {
144 	int err;
145 	struct seq_file *seq;
146 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
147 
148 	err = seq_open(file, &nfs_exports_op);
149 	if (err)
150 		return err;
151 
152 	seq = file->private_data;
153 	seq->private = nn->svc_export_cache;
154 	return 0;
155 }
156 
157 static int exports_nfsd_open(struct inode *inode, struct file *file)
158 {
159 	return exports_net_open(inode->i_sb->s_fs_info, file);
160 }
161 
162 static const struct file_operations exports_nfsd_operations = {
163 	.open		= exports_nfsd_open,
164 	.read		= seq_read,
165 	.llseek		= seq_lseek,
166 	.release	= seq_release,
167 };
168 
169 static int export_features_show(struct seq_file *m, void *v)
170 {
171 	seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
172 	return 0;
173 }
174 
175 DEFINE_SHOW_ATTRIBUTE(export_features);
176 
177 static const struct file_operations pool_stats_operations = {
178 	.open		= nfsd_pool_stats_open,
179 	.read		= seq_read,
180 	.llseek		= seq_lseek,
181 	.release	= seq_release,
182 };
183 
184 DEFINE_SHOW_ATTRIBUTE(nfsd_reply_cache_stats);
185 
186 DEFINE_SHOW_ATTRIBUTE(nfsd_file_cache_stats);
187 
188 /*----------------------------------------------------------------------------*/
189 /*
190  * payload - write methods
191  */
192 
193 static inline struct net *netns(struct file *file)
194 {
195 	return file_inode(file)->i_sb->s_fs_info;
196 }
197 
198 /*
199  * write_unlock_ip - Release all locks used by a client
200  *
201  * Experimental.
202  *
203  * Input:
204  *			buf:	'\n'-terminated C string containing a
205  *				presentation format IP address
206  *			size:	length of C string in @buf
207  * Output:
208  *	On success:	returns zero if all specified locks were released;
209  *			returns one if one or more locks were not released
210  *	On error:	return code is negative errno value
211  */
212 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
213 {
214 	struct sockaddr_storage address;
215 	struct sockaddr *sap = (struct sockaddr *)&address;
216 	size_t salen = sizeof(address);
217 	char *fo_path;
218 	struct net *net = netns(file);
219 
220 	/* sanity check */
221 	if (size == 0)
222 		return -EINVAL;
223 
224 	if (buf[size-1] != '\n')
225 		return -EINVAL;
226 
227 	fo_path = buf;
228 	if (qword_get(&buf, fo_path, size) < 0)
229 		return -EINVAL;
230 
231 	if (rpc_pton(net, fo_path, size, sap, salen) == 0)
232 		return -EINVAL;
233 
234 	trace_nfsd_ctl_unlock_ip(net, buf);
235 	return nlmsvc_unlock_all_by_ip(sap);
236 }
237 
238 /*
239  * write_unlock_fs - Release all locks on a local file system
240  *
241  * Experimental.
242  *
243  * Input:
244  *			buf:	'\n'-terminated C string containing the
245  *				absolute pathname of a local file system
246  *			size:	length of C string in @buf
247  * Output:
248  *	On success:	returns zero if all specified locks were released;
249  *			returns one if one or more locks were not released
250  *	On error:	return code is negative errno value
251  */
252 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
253 {
254 	struct path path;
255 	char *fo_path;
256 	int error;
257 
258 	/* sanity check */
259 	if (size == 0)
260 		return -EINVAL;
261 
262 	if (buf[size-1] != '\n')
263 		return -EINVAL;
264 
265 	fo_path = buf;
266 	if (qword_get(&buf, fo_path, size) < 0)
267 		return -EINVAL;
268 	trace_nfsd_ctl_unlock_fs(netns(file), fo_path);
269 	error = kern_path(fo_path, 0, &path);
270 	if (error)
271 		return error;
272 
273 	/*
274 	 * XXX: Needs better sanity checking.  Otherwise we could end up
275 	 * releasing locks on the wrong file system.
276 	 *
277 	 * For example:
278 	 * 1.  Does the path refer to a directory?
279 	 * 2.  Is that directory a mount point, or
280 	 * 3.  Is that directory the root of an exported file system?
281 	 */
282 	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
283 	nfsd4_revoke_states(netns(file), path.dentry->d_sb);
284 
285 	path_put(&path);
286 	return error;
287 }
288 
289 /*
290  * write_filehandle - Get a variable-length NFS file handle by path
291  *
292  * On input, the buffer contains a '\n'-terminated C string comprised of
293  * three alphanumeric words separated by whitespace.  The string may
294  * contain escape sequences.
295  *
296  * Input:
297  *			buf:
298  *				domain:		client domain name
299  *				path:		export pathname
300  *				maxsize:	numeric maximum size of
301  *						@buf
302  *			size:	length of C string in @buf
303  * Output:
304  *	On success:	passed-in buffer filled with '\n'-terminated C
305  *			string containing a ASCII hex text version
306  *			of the NFS file handle;
307  *			return code is the size in bytes of the string
308  *	On error:	return code is negative errno value
309  */
310 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
311 {
312 	char *dname, *path;
313 	int maxsize;
314 	char *mesg = buf;
315 	int len;
316 	struct auth_domain *dom;
317 	struct knfsd_fh fh;
318 
319 	if (size == 0)
320 		return -EINVAL;
321 
322 	if (buf[size-1] != '\n')
323 		return -EINVAL;
324 	buf[size-1] = 0;
325 
326 	dname = mesg;
327 	len = qword_get(&mesg, dname, size);
328 	if (len <= 0)
329 		return -EINVAL;
330 
331 	path = dname+len+1;
332 	len = qword_get(&mesg, path, size);
333 	if (len <= 0)
334 		return -EINVAL;
335 
336 	len = get_int(&mesg, &maxsize);
337 	if (len)
338 		return len;
339 
340 	if (maxsize < NFS_FHSIZE)
341 		return -EINVAL;
342 	maxsize = min(maxsize, NFS3_FHSIZE);
343 
344 	if (qword_get(&mesg, mesg, size) > 0)
345 		return -EINVAL;
346 
347 	trace_nfsd_ctl_filehandle(netns(file), dname, path, maxsize);
348 
349 	/* we have all the words, they are in buf.. */
350 	dom = unix_domain_find(dname);
351 	if (!dom)
352 		return -ENOMEM;
353 
354 	len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
355 	auth_domain_put(dom);
356 	if (len)
357 		return len;
358 
359 	mesg = buf;
360 	len = SIMPLE_TRANSACTION_LIMIT;
361 	qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
362 	mesg[-1] = '\n';
363 	return mesg - buf;
364 }
365 
366 /*
367  * write_threads - Start NFSD, or report the current number of running threads
368  *
369  * Input:
370  *			buf:		ignored
371  *			size:		zero
372  * Output:
373  *	On success:	passed-in buffer filled with '\n'-terminated C
374  *			string numeric value representing the number of
375  *			running NFSD threads;
376  *			return code is the size in bytes of the string
377  *	On error:	return code is zero
378  *
379  * OR
380  *
381  * Input:
382  *			buf:		C string containing an unsigned
383  *					integer value representing the
384  *					number of NFSD threads to start
385  *			size:		non-zero length of C string in @buf
386  * Output:
387  *	On success:	NFS service is started;
388  *			passed-in buffer filled with '\n'-terminated C
389  *			string numeric value representing the number of
390  *			running NFSD threads;
391  *			return code is the size in bytes of the string
392  *	On error:	return code is zero or a negative errno value
393  */
394 static ssize_t write_threads(struct file *file, char *buf, size_t size)
395 {
396 	char *mesg = buf;
397 	int rv;
398 	struct net *net = netns(file);
399 
400 	if (size > 0) {
401 		int newthreads;
402 		rv = get_int(&mesg, &newthreads);
403 		if (rv)
404 			return rv;
405 		if (newthreads < 0)
406 			return -EINVAL;
407 		trace_nfsd_ctl_threads(net, newthreads);
408 		mutex_lock(&nfsd_mutex);
409 		rv = nfsd_svc(1, &newthreads, net, file->f_cred, NULL);
410 		mutex_unlock(&nfsd_mutex);
411 		if (rv < 0)
412 			return rv;
413 	} else
414 		rv = nfsd_nrthreads(net);
415 
416 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
417 }
418 
419 /*
420  * write_pool_threads - Set or report the current number of threads per pool
421  *
422  * Input:
423  *			buf:		ignored
424  *			size:		zero
425  *
426  * OR
427  *
428  * Input:
429  *			buf:		C string containing whitespace-
430  *					separated unsigned integer values
431  *					representing the number of NFSD
432  *					threads to start in each pool
433  *			size:		non-zero length of C string in @buf
434  * Output:
435  *	On success:	passed-in buffer filled with '\n'-terminated C
436  *			string containing integer values representing the
437  *			number of NFSD threads in each pool;
438  *			return code is the size in bytes of the string
439  *	On error:	return code is zero or a negative errno value
440  */
441 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
442 {
443 	/* if size > 0, look for an array of number of threads per node
444 	 * and apply them  then write out number of threads per node as reply
445 	 */
446 	char *mesg = buf;
447 	int i;
448 	int rv;
449 	int len;
450 	int npools;
451 	int *nthreads;
452 	struct net *net = netns(file);
453 
454 	mutex_lock(&nfsd_mutex);
455 	npools = nfsd_nrpools(net);
456 	if (npools == 0) {
457 		/*
458 		 * NFS is shut down.  The admin can start it by
459 		 * writing to the threads file but NOT the pool_threads
460 		 * file, sorry.  Report zero threads.
461 		 */
462 		mutex_unlock(&nfsd_mutex);
463 		strcpy(buf, "0\n");
464 		return strlen(buf);
465 	}
466 
467 	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
468 	rv = -ENOMEM;
469 	if (nthreads == NULL)
470 		goto out_free;
471 
472 	if (size > 0) {
473 		for (i = 0; i < npools; i++) {
474 			rv = get_int(&mesg, &nthreads[i]);
475 			if (rv == -ENOENT)
476 				break;		/* fewer numbers than pools */
477 			if (rv)
478 				goto out_free;	/* syntax error */
479 			rv = -EINVAL;
480 			if (nthreads[i] < 0)
481 				goto out_free;
482 			trace_nfsd_ctl_pool_threads(net, i, nthreads[i]);
483 		}
484 
485 		/*
486 		 * There must always be a thread in pool 0; the admin
487 		 * can't shut down NFS completely using pool_threads.
488 		 */
489 		if (nthreads[0] == 0)
490 			nthreads[0] = 1;
491 
492 		rv = nfsd_set_nrthreads(i, nthreads, net);
493 		if (rv)
494 			goto out_free;
495 	}
496 
497 	rv = nfsd_get_nrthreads(npools, nthreads, net);
498 	if (rv)
499 		goto out_free;
500 
501 	mesg = buf;
502 	size = SIMPLE_TRANSACTION_LIMIT;
503 	for (i = 0; i < npools && size > 0; i++) {
504 		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
505 		len = strlen(mesg);
506 		size -= len;
507 		mesg += len;
508 	}
509 	rv = mesg - buf;
510 out_free:
511 	kfree(nthreads);
512 	mutex_unlock(&nfsd_mutex);
513 	return rv;
514 }
515 
516 static ssize_t
517 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
518 		const char *sep, unsigned vers, int minor)
519 {
520 	const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
521 	bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
522 
523 	if (vers == 4 && minor >= 0 &&
524 	    !nfsd_minorversion(nn, minor, NFSD_TEST))
525 		supported = false;
526 	if (minor == 0 && supported)
527 		/*
528 		 * special case for backward compatability.
529 		 * +4.0 is never reported, it is implied by
530 		 * +4, unless -4.0 is present.
531 		 */
532 		return 0;
533 	return snprintf(buf, remaining, format, sep,
534 			supported ? '+' : '-', vers, minor);
535 }
536 
537 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
538 {
539 	char *mesg = buf;
540 	char *vers, *minorp, sign;
541 	int len, num, remaining;
542 	ssize_t tlen = 0;
543 	char *sep;
544 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
545 
546 	if (size > 0) {
547 		if (nn->nfsd_serv)
548 			/* Cannot change versions without updating
549 			 * nn->nfsd_serv->sv_xdrsize, and reallocing
550 			 * rq_argp and rq_resp
551 			 */
552 			return -EBUSY;
553 		if (buf[size-1] != '\n')
554 			return -EINVAL;
555 		buf[size-1] = 0;
556 		trace_nfsd_ctl_version(netns(file), buf);
557 
558 		vers = mesg;
559 		len = qword_get(&mesg, vers, size);
560 		if (len <= 0) return -EINVAL;
561 		do {
562 			enum vers_op cmd;
563 			unsigned minor;
564 			sign = *vers;
565 			if (sign == '+' || sign == '-')
566 				num = simple_strtol((vers+1), &minorp, 0);
567 			else
568 				num = simple_strtol(vers, &minorp, 0);
569 			if (*minorp == '.') {
570 				if (num != 4)
571 					return -EINVAL;
572 				if (kstrtouint(minorp+1, 0, &minor) < 0)
573 					return -EINVAL;
574 			}
575 
576 			cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
577 			switch(num) {
578 #ifdef CONFIG_NFSD_V2
579 			case 2:
580 #endif
581 			case 3:
582 				nfsd_vers(nn, num, cmd);
583 				break;
584 			case 4:
585 				if (*minorp == '.') {
586 					if (nfsd_minorversion(nn, minor, cmd) < 0)
587 						return -EINVAL;
588 				} else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
589 					/*
590 					 * Either we have +4 and no minors are enabled,
591 					 * or we have -4 and at least one minor is enabled.
592 					 * In either case, propagate 'cmd' to all minors.
593 					 */
594 					minor = 0;
595 					while (nfsd_minorversion(nn, minor, cmd) >= 0)
596 						minor++;
597 				}
598 				break;
599 			default:
600 				/* Ignore requests to disable non-existent versions */
601 				if (cmd == NFSD_SET)
602 					return -EINVAL;
603 			}
604 			vers += len + 1;
605 		} while ((len = qword_get(&mesg, vers, size)) > 0);
606 		/* If all get turned off, turn them back on, as
607 		 * having no versions is BAD
608 		 */
609 		nfsd_reset_versions(nn);
610 	}
611 
612 	/* Now write current state into reply buffer */
613 	sep = "";
614 	remaining = SIMPLE_TRANSACTION_LIMIT;
615 	for (num=2 ; num <= 4 ; num++) {
616 		int minor;
617 		if (!nfsd_vers(nn, num, NFSD_AVAIL))
618 			continue;
619 
620 		minor = -1;
621 		do {
622 			len = nfsd_print_version_support(nn, buf, remaining,
623 					sep, num, minor);
624 			if (len >= remaining)
625 				goto out;
626 			remaining -= len;
627 			buf += len;
628 			tlen += len;
629 			minor++;
630 			if (len)
631 				sep = " ";
632 		} while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
633 	}
634 out:
635 	len = snprintf(buf, remaining, "\n");
636 	if (len >= remaining)
637 		return -EINVAL;
638 	return tlen + len;
639 }
640 
641 /*
642  * write_versions - Set or report the available NFS protocol versions
643  *
644  * Input:
645  *			buf:		ignored
646  *			size:		zero
647  * Output:
648  *	On success:	passed-in buffer filled with '\n'-terminated C
649  *			string containing positive or negative integer
650  *			values representing the current status of each
651  *			protocol version;
652  *			return code is the size in bytes of the string
653  *	On error:	return code is zero or a negative errno value
654  *
655  * OR
656  *
657  * Input:
658  *			buf:		C string containing whitespace-
659  *					separated positive or negative
660  *					integer values representing NFS
661  *					protocol versions to enable ("+n")
662  *					or disable ("-n")
663  *			size:		non-zero length of C string in @buf
664  * Output:
665  *	On success:	status of zero or more protocol versions has
666  *			been updated; passed-in buffer filled with
667  *			'\n'-terminated C string containing positive
668  *			or negative integer values representing the
669  *			current status of each protocol version;
670  *			return code is the size in bytes of the string
671  *	On error:	return code is zero or a negative errno value
672  */
673 static ssize_t write_versions(struct file *file, char *buf, size_t size)
674 {
675 	ssize_t rv;
676 
677 	mutex_lock(&nfsd_mutex);
678 	rv = __write_versions(file, buf, size);
679 	mutex_unlock(&nfsd_mutex);
680 	return rv;
681 }
682 
683 /*
684  * Zero-length write.  Return a list of NFSD's current listener
685  * transports.
686  */
687 static ssize_t __write_ports_names(char *buf, struct net *net)
688 {
689 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
690 
691 	if (nn->nfsd_serv == NULL)
692 		return 0;
693 	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
694 }
695 
696 /*
697  * A single 'fd' number was written, in which case it must be for
698  * a socket of a supported family/protocol, and we use it as an
699  * nfsd listener.
700  */
701 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
702 {
703 	char *mesg = buf;
704 	int fd, err;
705 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
706 	struct svc_serv *serv;
707 
708 	err = get_int(&mesg, &fd);
709 	if (err != 0 || fd < 0)
710 		return -EINVAL;
711 	trace_nfsd_ctl_ports_addfd(net, fd);
712 
713 	err = nfsd_create_serv(net);
714 	if (err != 0)
715 		return err;
716 
717 	serv = nn->nfsd_serv;
718 	err = svc_addsock(serv, net, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
719 
720 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
721 		nfsd_destroy_serv(net);
722 
723 	return err;
724 }
725 
726 /*
727  * A transport listener is added by writing its transport name and
728  * a port number.
729  */
730 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
731 {
732 	char transport[16];
733 	struct svc_xprt *xprt;
734 	int port, err;
735 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
736 	struct svc_serv *serv;
737 
738 	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
739 		return -EINVAL;
740 
741 	if (port < 1 || port > USHRT_MAX)
742 		return -EINVAL;
743 	trace_nfsd_ctl_ports_addxprt(net, transport, port);
744 
745 	err = nfsd_create_serv(net);
746 	if (err != 0)
747 		return err;
748 
749 	serv = nn->nfsd_serv;
750 	err = svc_xprt_create(serv, transport, net,
751 			      PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
752 	if (err < 0)
753 		goto out_err;
754 
755 	err = svc_xprt_create(serv, transport, net,
756 			      PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
757 	if (err < 0 && err != -EAFNOSUPPORT)
758 		goto out_close;
759 
760 	return 0;
761 out_close:
762 	xprt = svc_find_xprt(serv, transport, net, PF_INET, port);
763 	if (xprt != NULL) {
764 		svc_xprt_close(xprt);
765 		svc_xprt_put(xprt);
766 	}
767 out_err:
768 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
769 		nfsd_destroy_serv(net);
770 
771 	return err;
772 }
773 
774 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
775 			     struct net *net)
776 {
777 	if (size == 0)
778 		return __write_ports_names(buf, net);
779 
780 	if (isdigit(buf[0]))
781 		return __write_ports_addfd(buf, net, file->f_cred);
782 
783 	if (isalpha(buf[0]))
784 		return __write_ports_addxprt(buf, net, file->f_cred);
785 
786 	return -EINVAL;
787 }
788 
789 /*
790  * write_ports - Pass a socket file descriptor or transport name to listen on
791  *
792  * Input:
793  *			buf:		ignored
794  *			size:		zero
795  * Output:
796  *	On success:	passed-in buffer filled with a '\n'-terminated C
797  *			string containing a whitespace-separated list of
798  *			named NFSD listeners;
799  *			return code is the size in bytes of the string
800  *	On error:	return code is zero or a negative errno value
801  *
802  * OR
803  *
804  * Input:
805  *			buf:		C string containing an unsigned
806  *					integer value representing a bound
807  *					but unconnected socket that is to be
808  *					used as an NFSD listener; listen(3)
809  *					must be called for a SOCK_STREAM
810  *					socket, otherwise it is ignored
811  *			size:		non-zero length of C string in @buf
812  * Output:
813  *	On success:	NFS service is started;
814  *			passed-in buffer filled with a '\n'-terminated C
815  *			string containing a unique alphanumeric name of
816  *			the listener;
817  *			return code is the size in bytes of the string
818  *	On error:	return code is a negative errno value
819  *
820  * OR
821  *
822  * Input:
823  *			buf:		C string containing a transport
824  *					name and an unsigned integer value
825  *					representing the port to listen on,
826  *					separated by whitespace
827  *			size:		non-zero length of C string in @buf
828  * Output:
829  *	On success:	returns zero; NFS service is started
830  *	On error:	return code is a negative errno value
831  */
832 static ssize_t write_ports(struct file *file, char *buf, size_t size)
833 {
834 	ssize_t rv;
835 
836 	mutex_lock(&nfsd_mutex);
837 	rv = __write_ports(file, buf, size, netns(file));
838 	mutex_unlock(&nfsd_mutex);
839 	return rv;
840 }
841 
842 
843 int nfsd_max_blksize;
844 
845 /*
846  * write_maxblksize - Set or report the current NFS blksize
847  *
848  * Input:
849  *			buf:		ignored
850  *			size:		zero
851  *
852  * OR
853  *
854  * Input:
855  *			buf:		C string containing an unsigned
856  *					integer value representing the new
857  *					NFS blksize
858  *			size:		non-zero length of C string in @buf
859  * Output:
860  *	On success:	passed-in buffer filled with '\n'-terminated C string
861  *			containing numeric value of the current NFS blksize
862  *			setting;
863  *			return code is the size in bytes of the string
864  *	On error:	return code is zero or a negative errno value
865  */
866 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
867 {
868 	char *mesg = buf;
869 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
870 
871 	if (size > 0) {
872 		int bsize;
873 		int rv = get_int(&mesg, &bsize);
874 		if (rv)
875 			return rv;
876 		trace_nfsd_ctl_maxblksize(netns(file), bsize);
877 
878 		/* force bsize into allowed range and
879 		 * required alignment.
880 		 */
881 		bsize = max_t(int, bsize, 1024);
882 		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
883 		bsize &= ~(1024-1);
884 		mutex_lock(&nfsd_mutex);
885 		if (nn->nfsd_serv) {
886 			mutex_unlock(&nfsd_mutex);
887 			return -EBUSY;
888 		}
889 		nfsd_max_blksize = bsize;
890 		mutex_unlock(&nfsd_mutex);
891 	}
892 
893 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
894 							nfsd_max_blksize);
895 }
896 
897 /*
898  * write_maxconn - Set or report the current max number of connections
899  *
900  * Input:
901  *			buf:		ignored
902  *			size:		zero
903  * OR
904  *
905  * Input:
906  *			buf:		C string containing an unsigned
907  *					integer value representing the new
908  *					number of max connections
909  *			size:		non-zero length of C string in @buf
910  * Output:
911  *	On success:	passed-in buffer filled with '\n'-terminated C string
912  *			containing numeric value of max_connections setting
913  *			for this net namespace;
914  *			return code is the size in bytes of the string
915  *	On error:	return code is zero or a negative errno value
916  */
917 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
918 {
919 	char *mesg = buf;
920 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
921 	unsigned int maxconn = nn->max_connections;
922 
923 	if (size > 0) {
924 		int rv = get_uint(&mesg, &maxconn);
925 
926 		if (rv)
927 			return rv;
928 		trace_nfsd_ctl_maxconn(netns(file), maxconn);
929 		nn->max_connections = maxconn;
930 	}
931 
932 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
933 }
934 
935 #ifdef CONFIG_NFSD_V4
936 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
937 				  time64_t *time, struct nfsd_net *nn)
938 {
939 	struct dentry *dentry = file_dentry(file);
940 	char *mesg = buf;
941 	int rv, i;
942 
943 	if (size > 0) {
944 		if (nn->nfsd_serv)
945 			return -EBUSY;
946 		rv = get_int(&mesg, &i);
947 		if (rv)
948 			return rv;
949 		trace_nfsd_ctl_time(netns(file), dentry->d_name.name,
950 				    dentry->d_name.len, i);
951 
952 		/*
953 		 * Some sanity checking.  We don't have a reason for
954 		 * these particular numbers, but problems with the
955 		 * extremes are:
956 		 *	- Too short: the briefest network outage may
957 		 *	  cause clients to lose all their locks.  Also,
958 		 *	  the frequent polling may be wasteful.
959 		 *	- Too long: do you really want reboot recovery
960 		 *	  to take more than an hour?  Or to make other
961 		 *	  clients wait an hour before being able to
962 		 *	  revoke a dead client's locks?
963 		 */
964 		if (i < 10 || i > 3600)
965 			return -EINVAL;
966 		*time = i;
967 	}
968 
969 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
970 }
971 
972 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
973 				time64_t *time, struct nfsd_net *nn)
974 {
975 	ssize_t rv;
976 
977 	mutex_lock(&nfsd_mutex);
978 	rv = __nfsd4_write_time(file, buf, size, time, nn);
979 	mutex_unlock(&nfsd_mutex);
980 	return rv;
981 }
982 
983 /*
984  * write_leasetime - Set or report the current NFSv4 lease time
985  *
986  * Input:
987  *			buf:		ignored
988  *			size:		zero
989  *
990  * OR
991  *
992  * Input:
993  *			buf:		C string containing an unsigned
994  *					integer value representing the new
995  *					NFSv4 lease expiry time
996  *			size:		non-zero length of C string in @buf
997  * Output:
998  *	On success:	passed-in buffer filled with '\n'-terminated C
999  *			string containing unsigned integer value of the
1000  *			current lease expiry time;
1001  *			return code is the size in bytes of the string
1002  *	On error:	return code is zero or a negative errno value
1003  */
1004 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1005 {
1006 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1007 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1008 }
1009 
1010 /*
1011  * write_gracetime - Set or report current NFSv4 grace period time
1012  *
1013  * As above, but sets the time of the NFSv4 grace period.
1014  *
1015  * Note this should never be set to less than the *previous*
1016  * lease-period time, but we don't try to enforce this.  (In the common
1017  * case (a new boot), we don't know what the previous lease time was
1018  * anyway.)
1019  */
1020 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1021 {
1022 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1023 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1024 }
1025 
1026 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1027 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1028 				   struct nfsd_net *nn)
1029 {
1030 	char *mesg = buf;
1031 	char *recdir;
1032 	int len, status;
1033 
1034 	if (size > 0) {
1035 		if (nn->nfsd_serv)
1036 			return -EBUSY;
1037 		if (size > PATH_MAX || buf[size-1] != '\n')
1038 			return -EINVAL;
1039 		buf[size-1] = 0;
1040 
1041 		recdir = mesg;
1042 		len = qword_get(&mesg, recdir, size);
1043 		if (len <= 0)
1044 			return -EINVAL;
1045 		trace_nfsd_ctl_recoverydir(netns(file), recdir);
1046 
1047 		status = nfs4_reset_recoverydir(recdir);
1048 		if (status)
1049 			return status;
1050 	}
1051 
1052 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1053 							nfs4_recoverydir());
1054 }
1055 
1056 /*
1057  * write_recoverydir - Set or report the pathname of the recovery directory
1058  *
1059  * Input:
1060  *			buf:		ignored
1061  *			size:		zero
1062  *
1063  * OR
1064  *
1065  * Input:
1066  *			buf:		C string containing the pathname
1067  *					of the directory on a local file
1068  *					system containing permanent NFSv4
1069  *					recovery data
1070  *			size:		non-zero length of C string in @buf
1071  * Output:
1072  *	On success:	passed-in buffer filled with '\n'-terminated C string
1073  *			containing the current recovery pathname setting;
1074  *			return code is the size in bytes of the string
1075  *	On error:	return code is zero or a negative errno value
1076  */
1077 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1078 {
1079 	ssize_t rv;
1080 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1081 
1082 	mutex_lock(&nfsd_mutex);
1083 	rv = __write_recoverydir(file, buf, size, nn);
1084 	mutex_unlock(&nfsd_mutex);
1085 	return rv;
1086 }
1087 #endif
1088 
1089 /*
1090  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1091  *
1092  * Input:
1093  *			buf:		ignored
1094  *			size:		zero
1095  * OR
1096  *
1097  * Input:
1098  *			buf:		any value
1099  *			size:		non-zero length of C string in @buf
1100  * Output:
1101  *			passed-in buffer filled with "Y" or "N" with a newline
1102  *			and NULL-terminated C string. This indicates whether
1103  *			the grace period has ended in the current net
1104  *			namespace. Return code is the size in bytes of the
1105  *			string. Writing a string that starts with 'Y', 'y', or
1106  *			'1' to the file will end the grace period for nfsd's v4
1107  *			lock manager.
1108  */
1109 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1110 {
1111 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1112 
1113 	if (size > 0) {
1114 		switch(buf[0]) {
1115 		case 'Y':
1116 		case 'y':
1117 		case '1':
1118 			if (!nn->nfsd_serv)
1119 				return -EBUSY;
1120 			trace_nfsd_end_grace(netns(file));
1121 			nfsd4_end_grace(nn);
1122 			break;
1123 		default:
1124 			return -EINVAL;
1125 		}
1126 	}
1127 
1128 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1129 			 nn->grace_ended ? 'Y' : 'N');
1130 }
1131 
1132 #endif
1133 
1134 /*----------------------------------------------------------------------------*/
1135 /*
1136  *	populating the filesystem.
1137  */
1138 
1139 /* Basically copying rpc_get_inode. */
1140 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1141 {
1142 	struct inode *inode = new_inode(sb);
1143 	if (!inode)
1144 		return NULL;
1145 	/* Following advice from simple_fill_super documentation: */
1146 	inode->i_ino = iunique(sb, NFSD_MaxReserved);
1147 	inode->i_mode = mode;
1148 	simple_inode_init_ts(inode);
1149 	switch (mode & S_IFMT) {
1150 	case S_IFDIR:
1151 		inode->i_fop = &simple_dir_operations;
1152 		inode->i_op = &simple_dir_inode_operations;
1153 		inc_nlink(inode);
1154 		break;
1155 	case S_IFLNK:
1156 		inode->i_op = &simple_symlink_inode_operations;
1157 		break;
1158 	default:
1159 		break;
1160 	}
1161 	return inode;
1162 }
1163 
1164 static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1165 {
1166 	struct inode *inode;
1167 
1168 	inode = nfsd_get_inode(dir->i_sb, mode);
1169 	if (!inode)
1170 		return -ENOMEM;
1171 	if (ncl) {
1172 		inode->i_private = ncl;
1173 		kref_get(&ncl->cl_ref);
1174 	}
1175 	d_add(dentry, inode);
1176 	inc_nlink(dir);
1177 	fsnotify_mkdir(dir, dentry);
1178 	return 0;
1179 }
1180 
1181 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1182 {
1183 	struct inode *dir = parent->d_inode;
1184 	struct dentry *dentry;
1185 	int ret = -ENOMEM;
1186 
1187 	inode_lock(dir);
1188 	dentry = d_alloc_name(parent, name);
1189 	if (!dentry)
1190 		goto out_err;
1191 	ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1192 	if (ret)
1193 		goto out_err;
1194 out:
1195 	inode_unlock(dir);
1196 	return dentry;
1197 out_err:
1198 	dput(dentry);
1199 	dentry = ERR_PTR(ret);
1200 	goto out;
1201 }
1202 
1203 #if IS_ENABLED(CONFIG_SUNRPC_GSS)
1204 static int __nfsd_symlink(struct inode *dir, struct dentry *dentry,
1205 			  umode_t mode, const char *content)
1206 {
1207 	struct inode *inode;
1208 
1209 	inode = nfsd_get_inode(dir->i_sb, mode);
1210 	if (!inode)
1211 		return -ENOMEM;
1212 
1213 	inode->i_link = (char *)content;
1214 	inode->i_size = strlen(content);
1215 
1216 	d_add(dentry, inode);
1217 	inc_nlink(dir);
1218 	fsnotify_create(dir, dentry);
1219 	return 0;
1220 }
1221 
1222 /*
1223  * @content is assumed to be a NUL-terminated string that lives
1224  * longer than the symlink itself.
1225  */
1226 static void _nfsd_symlink(struct dentry *parent, const char *name,
1227 			  const char *content)
1228 {
1229 	struct inode *dir = parent->d_inode;
1230 	struct dentry *dentry;
1231 	int ret;
1232 
1233 	inode_lock(dir);
1234 	dentry = d_alloc_name(parent, name);
1235 	if (!dentry)
1236 		goto out;
1237 	ret = __nfsd_symlink(d_inode(parent), dentry, S_IFLNK | 0777, content);
1238 	if (ret)
1239 		dput(dentry);
1240 out:
1241 	inode_unlock(dir);
1242 }
1243 #else
1244 static inline void _nfsd_symlink(struct dentry *parent, const char *name,
1245 				 const char *content)
1246 {
1247 }
1248 
1249 #endif
1250 
1251 static void clear_ncl(struct dentry *dentry)
1252 {
1253 	struct inode *inode = d_inode(dentry);
1254 	struct nfsdfs_client *ncl = inode->i_private;
1255 
1256 	spin_lock(&inode->i_lock);
1257 	inode->i_private = NULL;
1258 	spin_unlock(&inode->i_lock);
1259 	kref_put(&ncl->cl_ref, ncl->cl_release);
1260 }
1261 
1262 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1263 {
1264 	struct nfsdfs_client *nc;
1265 
1266 	spin_lock(&inode->i_lock);
1267 	nc = inode->i_private;
1268 	if (nc)
1269 		kref_get(&nc->cl_ref);
1270 	spin_unlock(&inode->i_lock);
1271 	return nc;
1272 }
1273 
1274 /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1275  * code instead. */
1276 static  int nfsdfs_create_files(struct dentry *root,
1277 				const struct tree_descr *files,
1278 				struct nfsdfs_client *ncl,
1279 				struct dentry **fdentries)
1280 {
1281 	struct inode *dir = d_inode(root);
1282 	struct inode *inode;
1283 	struct dentry *dentry;
1284 	int i;
1285 
1286 	inode_lock(dir);
1287 	for (i = 0; files->name && files->name[0]; i++, files++) {
1288 		dentry = d_alloc_name(root, files->name);
1289 		if (!dentry)
1290 			goto out;
1291 		inode = nfsd_get_inode(d_inode(root)->i_sb,
1292 					S_IFREG | files->mode);
1293 		if (!inode) {
1294 			dput(dentry);
1295 			goto out;
1296 		}
1297 		kref_get(&ncl->cl_ref);
1298 		inode->i_fop = files->ops;
1299 		inode->i_private = ncl;
1300 		d_add(dentry, inode);
1301 		fsnotify_create(dir, dentry);
1302 		if (fdentries)
1303 			fdentries[i] = dentry;
1304 	}
1305 	inode_unlock(dir);
1306 	return 0;
1307 out:
1308 	inode_unlock(dir);
1309 	return -ENOMEM;
1310 }
1311 
1312 /* on success, returns positive number unique to that client. */
1313 struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1314 				 struct nfsdfs_client *ncl, u32 id,
1315 				 const struct tree_descr *files,
1316 				 struct dentry **fdentries)
1317 {
1318 	struct dentry *dentry;
1319 	char name[11];
1320 	int ret;
1321 
1322 	sprintf(name, "%u", id);
1323 
1324 	dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1325 	if (IS_ERR(dentry)) /* XXX: tossing errors? */
1326 		return NULL;
1327 	ret = nfsdfs_create_files(dentry, files, ncl, fdentries);
1328 	if (ret) {
1329 		nfsd_client_rmdir(dentry);
1330 		return NULL;
1331 	}
1332 	return dentry;
1333 }
1334 
1335 /* Taken from __rpc_rmdir: */
1336 void nfsd_client_rmdir(struct dentry *dentry)
1337 {
1338 	simple_recursive_removal(dentry, clear_ncl);
1339 }
1340 
1341 static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1342 {
1343 	struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1344 							nfsd_net_id);
1345 	struct dentry *dentry;
1346 	int ret;
1347 
1348 	static const struct tree_descr nfsd_files[] = {
1349 		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1350 		/* Per-export io stats use same ops as exports file */
1351 		[NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1352 		[NFSD_Export_features] = {"export_features",
1353 					&export_features_fops, S_IRUGO},
1354 		[NFSD_FO_UnlockIP] = {"unlock_ip",
1355 					&transaction_ops, S_IWUSR|S_IRUSR},
1356 		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1357 					&transaction_ops, S_IWUSR|S_IRUSR},
1358 		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1359 		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1360 		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1361 		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1362 		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1363 					&nfsd_reply_cache_stats_fops, S_IRUGO},
1364 		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1365 		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1366 		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1367 		[NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1368 		[NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
1369 #ifdef CONFIG_NFSD_V4
1370 		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1371 		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1372 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1373 		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1374 #endif
1375 		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1376 #endif
1377 		/* last one */ {""}
1378 	};
1379 
1380 	ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1381 	if (ret)
1382 		return ret;
1383 	_nfsd_symlink(sb->s_root, "supported_krb5_enctypes",
1384 		      "/proc/net/rpc/gss_krb5_enctypes");
1385 	dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1386 	if (IS_ERR(dentry))
1387 		return PTR_ERR(dentry);
1388 	nn->nfsd_client_dir = dentry;
1389 	return 0;
1390 }
1391 
1392 static int nfsd_fs_get_tree(struct fs_context *fc)
1393 {
1394 	return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1395 }
1396 
1397 static void nfsd_fs_free_fc(struct fs_context *fc)
1398 {
1399 	if (fc->s_fs_info)
1400 		put_net(fc->s_fs_info);
1401 }
1402 
1403 static const struct fs_context_operations nfsd_fs_context_ops = {
1404 	.free		= nfsd_fs_free_fc,
1405 	.get_tree	= nfsd_fs_get_tree,
1406 };
1407 
1408 static int nfsd_init_fs_context(struct fs_context *fc)
1409 {
1410 	put_user_ns(fc->user_ns);
1411 	fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1412 	fc->ops = &nfsd_fs_context_ops;
1413 	return 0;
1414 }
1415 
1416 static void nfsd_umount(struct super_block *sb)
1417 {
1418 	struct net *net = sb->s_fs_info;
1419 
1420 	nfsd_shutdown_threads(net);
1421 
1422 	kill_litter_super(sb);
1423 	put_net(net);
1424 }
1425 
1426 static struct file_system_type nfsd_fs_type = {
1427 	.owner		= THIS_MODULE,
1428 	.name		= "nfsd",
1429 	.init_fs_context = nfsd_init_fs_context,
1430 	.kill_sb	= nfsd_umount,
1431 };
1432 MODULE_ALIAS_FS("nfsd");
1433 
1434 #ifdef CONFIG_PROC_FS
1435 
1436 static int exports_proc_open(struct inode *inode, struct file *file)
1437 {
1438 	return exports_net_open(current->nsproxy->net_ns, file);
1439 }
1440 
1441 static const struct proc_ops exports_proc_ops = {
1442 	.proc_open	= exports_proc_open,
1443 	.proc_read	= seq_read,
1444 	.proc_lseek	= seq_lseek,
1445 	.proc_release	= seq_release,
1446 };
1447 
1448 static int create_proc_exports_entry(void)
1449 {
1450 	struct proc_dir_entry *entry;
1451 
1452 	entry = proc_mkdir("fs/nfs", NULL);
1453 	if (!entry)
1454 		return -ENOMEM;
1455 	entry = proc_create("exports", 0, entry, &exports_proc_ops);
1456 	if (!entry) {
1457 		remove_proc_entry("fs/nfs", NULL);
1458 		return -ENOMEM;
1459 	}
1460 	return 0;
1461 }
1462 #else /* CONFIG_PROC_FS */
1463 static int create_proc_exports_entry(void)
1464 {
1465 	return 0;
1466 }
1467 #endif
1468 
1469 unsigned int nfsd_net_id;
1470 
1471 static int nfsd_genl_rpc_status_compose_msg(struct sk_buff *skb,
1472 					    struct netlink_callback *cb,
1473 					    struct nfsd_genl_rqstp *rqstp)
1474 {
1475 	void *hdr;
1476 	u32 i;
1477 
1478 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
1479 			  &nfsd_nl_family, 0, NFSD_CMD_RPC_STATUS_GET);
1480 	if (!hdr)
1481 		return -ENOBUFS;
1482 
1483 	if (nla_put_be32(skb, NFSD_A_RPC_STATUS_XID, rqstp->rq_xid) ||
1484 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_FLAGS, rqstp->rq_flags) ||
1485 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROG, rqstp->rq_prog) ||
1486 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROC, rqstp->rq_proc) ||
1487 	    nla_put_u8(skb, NFSD_A_RPC_STATUS_VERSION, rqstp->rq_vers) ||
1488 	    nla_put_s64(skb, NFSD_A_RPC_STATUS_SERVICE_TIME,
1489 			ktime_to_us(rqstp->rq_stime),
1490 			NFSD_A_RPC_STATUS_PAD))
1491 		return -ENOBUFS;
1492 
1493 	switch (rqstp->rq_saddr.sa_family) {
1494 	case AF_INET: {
1495 		const struct sockaddr_in *s_in, *d_in;
1496 
1497 		s_in = (const struct sockaddr_in *)&rqstp->rq_saddr;
1498 		d_in = (const struct sockaddr_in *)&rqstp->rq_daddr;
1499 		if (nla_put_in_addr(skb, NFSD_A_RPC_STATUS_SADDR4,
1500 				    s_in->sin_addr.s_addr) ||
1501 		    nla_put_in_addr(skb, NFSD_A_RPC_STATUS_DADDR4,
1502 				    d_in->sin_addr.s_addr) ||
1503 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1504 				 s_in->sin_port) ||
1505 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1506 				 d_in->sin_port))
1507 			return -ENOBUFS;
1508 		break;
1509 	}
1510 	case AF_INET6: {
1511 		const struct sockaddr_in6 *s_in, *d_in;
1512 
1513 		s_in = (const struct sockaddr_in6 *)&rqstp->rq_saddr;
1514 		d_in = (const struct sockaddr_in6 *)&rqstp->rq_daddr;
1515 		if (nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_SADDR6,
1516 				     &s_in->sin6_addr) ||
1517 		    nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_DADDR6,
1518 				     &d_in->sin6_addr) ||
1519 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1520 				 s_in->sin6_port) ||
1521 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1522 				 d_in->sin6_port))
1523 			return -ENOBUFS;
1524 		break;
1525 	}
1526 	}
1527 
1528 	for (i = 0; i < rqstp->rq_opcnt; i++)
1529 		if (nla_put_u32(skb, NFSD_A_RPC_STATUS_COMPOUND_OPS,
1530 				rqstp->rq_opnum[i]))
1531 			return -ENOBUFS;
1532 
1533 	genlmsg_end(skb, hdr);
1534 	return 0;
1535 }
1536 
1537 /**
1538  * nfsd_nl_rpc_status_get_dumpit - Handle rpc_status_get dumpit
1539  * @skb: reply buffer
1540  * @cb: netlink metadata and command arguments
1541  *
1542  * Returns the size of the reply or a negative errno.
1543  */
1544 int nfsd_nl_rpc_status_get_dumpit(struct sk_buff *skb,
1545 				  struct netlink_callback *cb)
1546 {
1547 	int i, ret, rqstp_index = 0;
1548 	struct nfsd_net *nn;
1549 
1550 	mutex_lock(&nfsd_mutex);
1551 
1552 	nn = net_generic(sock_net(skb->sk), nfsd_net_id);
1553 	if (!nn->nfsd_serv) {
1554 		ret = -ENODEV;
1555 		goto out_unlock;
1556 	}
1557 
1558 	rcu_read_lock();
1559 
1560 	for (i = 0; i < nn->nfsd_serv->sv_nrpools; i++) {
1561 		struct svc_rqst *rqstp;
1562 
1563 		if (i < cb->args[0]) /* already consumed */
1564 			continue;
1565 
1566 		rqstp_index = 0;
1567 		list_for_each_entry_rcu(rqstp,
1568 				&nn->nfsd_serv->sv_pools[i].sp_all_threads,
1569 				rq_all) {
1570 			struct nfsd_genl_rqstp genl_rqstp;
1571 			unsigned int status_counter;
1572 
1573 			if (rqstp_index++ < cb->args[1]) /* already consumed */
1574 				continue;
1575 			/*
1576 			 * Acquire rq_status_counter before parsing the rqst
1577 			 * fields. rq_status_counter is set to an odd value in
1578 			 * order to notify the consumers the rqstp fields are
1579 			 * meaningful.
1580 			 */
1581 			status_counter =
1582 				smp_load_acquire(&rqstp->rq_status_counter);
1583 			if (!(status_counter & 1))
1584 				continue;
1585 
1586 			genl_rqstp.rq_xid = rqstp->rq_xid;
1587 			genl_rqstp.rq_flags = rqstp->rq_flags;
1588 			genl_rqstp.rq_vers = rqstp->rq_vers;
1589 			genl_rqstp.rq_prog = rqstp->rq_prog;
1590 			genl_rqstp.rq_proc = rqstp->rq_proc;
1591 			genl_rqstp.rq_stime = rqstp->rq_stime;
1592 			genl_rqstp.rq_opcnt = 0;
1593 			memcpy(&genl_rqstp.rq_daddr, svc_daddr(rqstp),
1594 			       sizeof(struct sockaddr));
1595 			memcpy(&genl_rqstp.rq_saddr, svc_addr(rqstp),
1596 			       sizeof(struct sockaddr));
1597 
1598 #ifdef CONFIG_NFSD_V4
1599 			if (rqstp->rq_vers == NFS4_VERSION &&
1600 			    rqstp->rq_proc == NFSPROC4_COMPOUND) {
1601 				/* NFSv4 compound */
1602 				struct nfsd4_compoundargs *args;
1603 				int j;
1604 
1605 				args = rqstp->rq_argp;
1606 				genl_rqstp.rq_opcnt = args->opcnt;
1607 				for (j = 0; j < genl_rqstp.rq_opcnt; j++)
1608 					genl_rqstp.rq_opnum[j] =
1609 						args->ops[j].opnum;
1610 			}
1611 #endif /* CONFIG_NFSD_V4 */
1612 
1613 			/*
1614 			 * Acquire rq_status_counter before reporting the rqst
1615 			 * fields to the user.
1616 			 */
1617 			if (smp_load_acquire(&rqstp->rq_status_counter) !=
1618 			    status_counter)
1619 				continue;
1620 
1621 			ret = nfsd_genl_rpc_status_compose_msg(skb, cb,
1622 							       &genl_rqstp);
1623 			if (ret)
1624 				goto out;
1625 		}
1626 	}
1627 
1628 	cb->args[0] = i;
1629 	cb->args[1] = rqstp_index;
1630 	ret = skb->len;
1631 out:
1632 	rcu_read_unlock();
1633 out_unlock:
1634 	mutex_unlock(&nfsd_mutex);
1635 
1636 	return ret;
1637 }
1638 
1639 /**
1640  * nfsd_nl_threads_set_doit - set the number of running threads
1641  * @skb: reply buffer
1642  * @info: netlink metadata and command arguments
1643  *
1644  * Return 0 on success or a negative errno.
1645  */
1646 int nfsd_nl_threads_set_doit(struct sk_buff *skb, struct genl_info *info)
1647 {
1648 	int *nthreads, count = 0, nrpools, i, ret = -EOPNOTSUPP, rem;
1649 	struct net *net = genl_info_net(info);
1650 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1651 	const struct nlattr *attr;
1652 	const char *scope = NULL;
1653 
1654 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_SERVER_THREADS))
1655 		return -EINVAL;
1656 
1657 	/* count number of SERVER_THREADS values */
1658 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1659 		if (nla_type(attr) == NFSD_A_SERVER_THREADS)
1660 			count++;
1661 	}
1662 
1663 	mutex_lock(&nfsd_mutex);
1664 
1665 	nrpools = max(count, nfsd_nrpools(net));
1666 	nthreads = kcalloc(nrpools, sizeof(int), GFP_KERNEL);
1667 	if (!nthreads) {
1668 		ret = -ENOMEM;
1669 		goto out_unlock;
1670 	}
1671 
1672 	i = 0;
1673 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1674 		if (nla_type(attr) == NFSD_A_SERVER_THREADS) {
1675 			nthreads[i++] = nla_get_u32(attr);
1676 			if (i >= nrpools)
1677 				break;
1678 		}
1679 	}
1680 
1681 	if (info->attrs[NFSD_A_SERVER_GRACETIME] ||
1682 	    info->attrs[NFSD_A_SERVER_LEASETIME] ||
1683 	    info->attrs[NFSD_A_SERVER_SCOPE]) {
1684 		ret = -EBUSY;
1685 		if (nn->nfsd_serv && nn->nfsd_serv->sv_nrthreads)
1686 			goto out_unlock;
1687 
1688 		ret = -EINVAL;
1689 		attr = info->attrs[NFSD_A_SERVER_GRACETIME];
1690 		if (attr) {
1691 			u32 gracetime = nla_get_u32(attr);
1692 
1693 			if (gracetime < 10 || gracetime > 3600)
1694 				goto out_unlock;
1695 
1696 			nn->nfsd4_grace = gracetime;
1697 		}
1698 
1699 		attr = info->attrs[NFSD_A_SERVER_LEASETIME];
1700 		if (attr) {
1701 			u32 leasetime = nla_get_u32(attr);
1702 
1703 			if (leasetime < 10 || leasetime > 3600)
1704 				goto out_unlock;
1705 
1706 			nn->nfsd4_lease = leasetime;
1707 		}
1708 
1709 		attr = info->attrs[NFSD_A_SERVER_SCOPE];
1710 		if (attr)
1711 			scope = nla_data(attr);
1712 	}
1713 
1714 	ret = nfsd_svc(nrpools, nthreads, net, get_current_cred(), scope);
1715 	if (ret > 0)
1716 		ret = 0;
1717 out_unlock:
1718 	mutex_unlock(&nfsd_mutex);
1719 	kfree(nthreads);
1720 	return ret;
1721 }
1722 
1723 /**
1724  * nfsd_nl_threads_get_doit - get the number of running threads
1725  * @skb: reply buffer
1726  * @info: netlink metadata and command arguments
1727  *
1728  * Return 0 on success or a negative errno.
1729  */
1730 int nfsd_nl_threads_get_doit(struct sk_buff *skb, struct genl_info *info)
1731 {
1732 	struct net *net = genl_info_net(info);
1733 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1734 	void *hdr;
1735 	int err;
1736 
1737 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
1738 	if (!skb)
1739 		return -ENOMEM;
1740 
1741 	hdr = genlmsg_iput(skb, info);
1742 	if (!hdr) {
1743 		err = -EMSGSIZE;
1744 		goto err_free_msg;
1745 	}
1746 
1747 	mutex_lock(&nfsd_mutex);
1748 
1749 	err = nla_put_u32(skb, NFSD_A_SERVER_GRACETIME,
1750 			  nn->nfsd4_grace) ||
1751 	      nla_put_u32(skb, NFSD_A_SERVER_LEASETIME,
1752 			  nn->nfsd4_lease) ||
1753 	      nla_put_string(skb, NFSD_A_SERVER_SCOPE,
1754 			  nn->nfsd_name);
1755 	if (err)
1756 		goto err_unlock;
1757 
1758 	if (nn->nfsd_serv) {
1759 		int i;
1760 
1761 		for (i = 0; i < nfsd_nrpools(net); ++i) {
1762 			struct svc_pool *sp = &nn->nfsd_serv->sv_pools[i];
1763 
1764 			err = nla_put_u32(skb, NFSD_A_SERVER_THREADS,
1765 					  atomic_read(&sp->sp_nrthreads));
1766 			if (err)
1767 				goto err_unlock;
1768 		}
1769 	} else {
1770 		err = nla_put_u32(skb, NFSD_A_SERVER_THREADS, 0);
1771 		if (err)
1772 			goto err_unlock;
1773 	}
1774 
1775 	mutex_unlock(&nfsd_mutex);
1776 
1777 	genlmsg_end(skb, hdr);
1778 
1779 	return genlmsg_reply(skb, info);
1780 
1781 err_unlock:
1782 	mutex_unlock(&nfsd_mutex);
1783 err_free_msg:
1784 	nlmsg_free(skb);
1785 
1786 	return err;
1787 }
1788 
1789 /**
1790  * nfsd_nl_version_set_doit - set the nfs enabled versions
1791  * @skb: reply buffer
1792  * @info: netlink metadata and command arguments
1793  *
1794  * Return 0 on success or a negative errno.
1795  */
1796 int nfsd_nl_version_set_doit(struct sk_buff *skb, struct genl_info *info)
1797 {
1798 	const struct nlattr *attr;
1799 	struct nfsd_net *nn;
1800 	int i, rem;
1801 
1802 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_SERVER_PROTO_VERSION))
1803 		return -EINVAL;
1804 
1805 	mutex_lock(&nfsd_mutex);
1806 
1807 	nn = net_generic(genl_info_net(info), nfsd_net_id);
1808 	if (nn->nfsd_serv) {
1809 		mutex_unlock(&nfsd_mutex);
1810 		return -EBUSY;
1811 	}
1812 
1813 	/* clear current supported versions. */
1814 	nfsd_vers(nn, 2, NFSD_CLEAR);
1815 	nfsd_vers(nn, 3, NFSD_CLEAR);
1816 	for (i = 0; i <= NFSD_SUPPORTED_MINOR_VERSION; i++)
1817 		nfsd_minorversion(nn, i, NFSD_CLEAR);
1818 
1819 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1820 		struct nlattr *tb[NFSD_A_VERSION_MAX + 1];
1821 		u32 major, minor = 0;
1822 		bool enabled;
1823 
1824 		if (nla_type(attr) != NFSD_A_SERVER_PROTO_VERSION)
1825 			continue;
1826 
1827 		if (nla_parse_nested(tb, NFSD_A_VERSION_MAX, attr,
1828 				     nfsd_version_nl_policy, info->extack) < 0)
1829 			continue;
1830 
1831 		if (!tb[NFSD_A_VERSION_MAJOR])
1832 			continue;
1833 
1834 		major = nla_get_u32(tb[NFSD_A_VERSION_MAJOR]);
1835 		if (tb[NFSD_A_VERSION_MINOR])
1836 			minor = nla_get_u32(tb[NFSD_A_VERSION_MINOR]);
1837 
1838 		enabled = nla_get_flag(tb[NFSD_A_VERSION_ENABLED]);
1839 
1840 		switch (major) {
1841 		case 4:
1842 			nfsd_minorversion(nn, minor, enabled ? NFSD_SET : NFSD_CLEAR);
1843 			break;
1844 		case 3:
1845 		case 2:
1846 			if (!minor)
1847 				nfsd_vers(nn, major, enabled ? NFSD_SET : NFSD_CLEAR);
1848 			break;
1849 		default:
1850 			break;
1851 		}
1852 	}
1853 
1854 	mutex_unlock(&nfsd_mutex);
1855 
1856 	return 0;
1857 }
1858 
1859 /**
1860  * nfsd_nl_version_get_doit - get the enabled status for all supported nfs versions
1861  * @skb: reply buffer
1862  * @info: netlink metadata and command arguments
1863  *
1864  * Return 0 on success or a negative errno.
1865  */
1866 int nfsd_nl_version_get_doit(struct sk_buff *skb, struct genl_info *info)
1867 {
1868 	struct nfsd_net *nn;
1869 	int i, err;
1870 	void *hdr;
1871 
1872 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
1873 	if (!skb)
1874 		return -ENOMEM;
1875 
1876 	hdr = genlmsg_iput(skb, info);
1877 	if (!hdr) {
1878 		err = -EMSGSIZE;
1879 		goto err_free_msg;
1880 	}
1881 
1882 	mutex_lock(&nfsd_mutex);
1883 	nn = net_generic(genl_info_net(info), nfsd_net_id);
1884 
1885 	for (i = 2; i <= 4; i++) {
1886 		int j;
1887 
1888 		for (j = 0; j <= NFSD_SUPPORTED_MINOR_VERSION; j++) {
1889 			struct nlattr *attr;
1890 
1891 			/* Don't record any versions the kernel doesn't have
1892 			 * compiled in
1893 			 */
1894 			if (!nfsd_support_version(i))
1895 				continue;
1896 
1897 			/* NFSv{2,3} does not support minor numbers */
1898 			if (i < 4 && j)
1899 				continue;
1900 
1901 			attr = nla_nest_start(skb,
1902 					      NFSD_A_SERVER_PROTO_VERSION);
1903 			if (!attr) {
1904 				err = -EINVAL;
1905 				goto err_nfsd_unlock;
1906 			}
1907 
1908 			if (nla_put_u32(skb, NFSD_A_VERSION_MAJOR, i) ||
1909 			    nla_put_u32(skb, NFSD_A_VERSION_MINOR, j)) {
1910 				err = -EINVAL;
1911 				goto err_nfsd_unlock;
1912 			}
1913 
1914 			/* Set the enabled flag if the version is enabled */
1915 			if (nfsd_vers(nn, i, NFSD_TEST) &&
1916 			    (i < 4 || nfsd_minorversion(nn, j, NFSD_TEST)) &&
1917 			    nla_put_flag(skb, NFSD_A_VERSION_ENABLED)) {
1918 				err = -EINVAL;
1919 				goto err_nfsd_unlock;
1920 			}
1921 
1922 			nla_nest_end(skb, attr);
1923 		}
1924 	}
1925 
1926 	mutex_unlock(&nfsd_mutex);
1927 	genlmsg_end(skb, hdr);
1928 
1929 	return genlmsg_reply(skb, info);
1930 
1931 err_nfsd_unlock:
1932 	mutex_unlock(&nfsd_mutex);
1933 err_free_msg:
1934 	nlmsg_free(skb);
1935 
1936 	return err;
1937 }
1938 
1939 /**
1940  * nfsd_nl_listener_set_doit - set the nfs running sockets
1941  * @skb: reply buffer
1942  * @info: netlink metadata and command arguments
1943  *
1944  * Return 0 on success or a negative errno.
1945  */
1946 int nfsd_nl_listener_set_doit(struct sk_buff *skb, struct genl_info *info)
1947 {
1948 	struct net *net = genl_info_net(info);
1949 	struct svc_xprt *xprt, *tmp;
1950 	const struct nlattr *attr;
1951 	struct svc_serv *serv;
1952 	LIST_HEAD(permsocks);
1953 	struct nfsd_net *nn;
1954 	int err, rem;
1955 
1956 	mutex_lock(&nfsd_mutex);
1957 
1958 	err = nfsd_create_serv(net);
1959 	if (err) {
1960 		mutex_unlock(&nfsd_mutex);
1961 		return err;
1962 	}
1963 
1964 	nn = net_generic(net, nfsd_net_id);
1965 	serv = nn->nfsd_serv;
1966 
1967 	spin_lock_bh(&serv->sv_lock);
1968 
1969 	/* Move all of the old listener sockets to a temp list */
1970 	list_splice_init(&serv->sv_permsocks, &permsocks);
1971 
1972 	/*
1973 	 * Walk the list of server_socks from userland and move any that match
1974 	 * back to sv_permsocks
1975 	 */
1976 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1977 		struct nlattr *tb[NFSD_A_SOCK_MAX + 1];
1978 		const char *xcl_name;
1979 		struct sockaddr *sa;
1980 
1981 		if (nla_type(attr) != NFSD_A_SERVER_SOCK_ADDR)
1982 			continue;
1983 
1984 		if (nla_parse_nested(tb, NFSD_A_SOCK_MAX, attr,
1985 				     nfsd_sock_nl_policy, info->extack) < 0)
1986 			continue;
1987 
1988 		if (!tb[NFSD_A_SOCK_ADDR] || !tb[NFSD_A_SOCK_TRANSPORT_NAME])
1989 			continue;
1990 
1991 		if (nla_len(tb[NFSD_A_SOCK_ADDR]) < sizeof(*sa))
1992 			continue;
1993 
1994 		xcl_name = nla_data(tb[NFSD_A_SOCK_TRANSPORT_NAME]);
1995 		sa = nla_data(tb[NFSD_A_SOCK_ADDR]);
1996 
1997 		/* Put back any matching sockets */
1998 		list_for_each_entry_safe(xprt, tmp, &permsocks, xpt_list) {
1999 			/* This shouldn't be possible */
2000 			if (WARN_ON_ONCE(xprt->xpt_net != net)) {
2001 				list_move(&xprt->xpt_list, &serv->sv_permsocks);
2002 				continue;
2003 			}
2004 
2005 			/* If everything matches, put it back */
2006 			if (!strcmp(xprt->xpt_class->xcl_name, xcl_name) &&
2007 			    rpc_cmp_addr_port(sa, (struct sockaddr *)&xprt->xpt_local)) {
2008 				list_move(&xprt->xpt_list, &serv->sv_permsocks);
2009 				break;
2010 			}
2011 		}
2012 	}
2013 
2014 	/* For now, no removing old sockets while server is running */
2015 	if (serv->sv_nrthreads && !list_empty(&permsocks)) {
2016 		list_splice_init(&permsocks, &serv->sv_permsocks);
2017 		spin_unlock_bh(&serv->sv_lock);
2018 		err = -EBUSY;
2019 		goto out_unlock_mtx;
2020 	}
2021 
2022 	/* Close the remaining sockets on the permsocks list */
2023 	while (!list_empty(&permsocks)) {
2024 		xprt = list_first_entry(&permsocks, struct svc_xprt, xpt_list);
2025 		list_move(&xprt->xpt_list, &serv->sv_permsocks);
2026 
2027 		/*
2028 		 * Newly-created sockets are born with the BUSY bit set. Clear
2029 		 * it if there are no threads, since nothing can pick it up
2030 		 * in that case.
2031 		 */
2032 		if (!serv->sv_nrthreads)
2033 			clear_bit(XPT_BUSY, &xprt->xpt_flags);
2034 
2035 		set_bit(XPT_CLOSE, &xprt->xpt_flags);
2036 		spin_unlock_bh(&serv->sv_lock);
2037 		svc_xprt_close(xprt);
2038 		spin_lock_bh(&serv->sv_lock);
2039 	}
2040 
2041 	spin_unlock_bh(&serv->sv_lock);
2042 
2043 	/* walk list of addrs again, open any that still don't exist */
2044 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
2045 		struct nlattr *tb[NFSD_A_SOCK_MAX + 1];
2046 		const char *xcl_name;
2047 		struct sockaddr *sa;
2048 		int ret;
2049 
2050 		if (nla_type(attr) != NFSD_A_SERVER_SOCK_ADDR)
2051 			continue;
2052 
2053 		if (nla_parse_nested(tb, NFSD_A_SOCK_MAX, attr,
2054 				     nfsd_sock_nl_policy, info->extack) < 0)
2055 			continue;
2056 
2057 		if (!tb[NFSD_A_SOCK_ADDR] || !tb[NFSD_A_SOCK_TRANSPORT_NAME])
2058 			continue;
2059 
2060 		if (nla_len(tb[NFSD_A_SOCK_ADDR]) < sizeof(*sa))
2061 			continue;
2062 
2063 		xcl_name = nla_data(tb[NFSD_A_SOCK_TRANSPORT_NAME]);
2064 		sa = nla_data(tb[NFSD_A_SOCK_ADDR]);
2065 
2066 		xprt = svc_find_listener(serv, xcl_name, net, sa);
2067 		if (xprt) {
2068 			svc_xprt_put(xprt);
2069 			continue;
2070 		}
2071 
2072 		ret = svc_xprt_create_from_sa(serv, xcl_name, net, sa, 0,
2073 					      get_current_cred());
2074 		/* always save the latest error */
2075 		if (ret < 0)
2076 			err = ret;
2077 	}
2078 
2079 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
2080 		nfsd_destroy_serv(net);
2081 
2082 out_unlock_mtx:
2083 	mutex_unlock(&nfsd_mutex);
2084 
2085 	return err;
2086 }
2087 
2088 /**
2089  * nfsd_nl_listener_get_doit - get the nfs running listeners
2090  * @skb: reply buffer
2091  * @info: netlink metadata and command arguments
2092  *
2093  * Return 0 on success or a negative errno.
2094  */
2095 int nfsd_nl_listener_get_doit(struct sk_buff *skb, struct genl_info *info)
2096 {
2097 	struct svc_xprt *xprt;
2098 	struct svc_serv *serv;
2099 	struct nfsd_net *nn;
2100 	void *hdr;
2101 	int err;
2102 
2103 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
2104 	if (!skb)
2105 		return -ENOMEM;
2106 
2107 	hdr = genlmsg_iput(skb, info);
2108 	if (!hdr) {
2109 		err = -EMSGSIZE;
2110 		goto err_free_msg;
2111 	}
2112 
2113 	mutex_lock(&nfsd_mutex);
2114 	nn = net_generic(genl_info_net(info), nfsd_net_id);
2115 
2116 	/* no nfs server? Just send empty socket list */
2117 	if (!nn->nfsd_serv)
2118 		goto out_unlock_mtx;
2119 
2120 	serv = nn->nfsd_serv;
2121 	spin_lock_bh(&serv->sv_lock);
2122 	list_for_each_entry(xprt, &serv->sv_permsocks, xpt_list) {
2123 		struct nlattr *attr;
2124 
2125 		attr = nla_nest_start(skb, NFSD_A_SERVER_SOCK_ADDR);
2126 		if (!attr) {
2127 			err = -EINVAL;
2128 			goto err_serv_unlock;
2129 		}
2130 
2131 		if (nla_put_string(skb, NFSD_A_SOCK_TRANSPORT_NAME,
2132 				   xprt->xpt_class->xcl_name) ||
2133 		    nla_put(skb, NFSD_A_SOCK_ADDR,
2134 			    sizeof(struct sockaddr_storage),
2135 			    &xprt->xpt_local)) {
2136 			err = -EINVAL;
2137 			goto err_serv_unlock;
2138 		}
2139 
2140 		nla_nest_end(skb, attr);
2141 	}
2142 	spin_unlock_bh(&serv->sv_lock);
2143 out_unlock_mtx:
2144 	mutex_unlock(&nfsd_mutex);
2145 	genlmsg_end(skb, hdr);
2146 
2147 	return genlmsg_reply(skb, info);
2148 
2149 err_serv_unlock:
2150 	spin_unlock_bh(&serv->sv_lock);
2151 	mutex_unlock(&nfsd_mutex);
2152 err_free_msg:
2153 	nlmsg_free(skb);
2154 
2155 	return err;
2156 }
2157 
2158 /**
2159  * nfsd_nl_pool_mode_set_doit - set the number of running threads
2160  * @skb: reply buffer
2161  * @info: netlink metadata and command arguments
2162  *
2163  * Return 0 on success or a negative errno.
2164  */
2165 int nfsd_nl_pool_mode_set_doit(struct sk_buff *skb, struct genl_info *info)
2166 {
2167 	const struct nlattr *attr;
2168 
2169 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_POOL_MODE_MODE))
2170 		return -EINVAL;
2171 
2172 	attr = info->attrs[NFSD_A_POOL_MODE_MODE];
2173 	return sunrpc_set_pool_mode(nla_data(attr));
2174 }
2175 
2176 /**
2177  * nfsd_nl_pool_mode_get_doit - get info about pool_mode
2178  * @skb: reply buffer
2179  * @info: netlink metadata and command arguments
2180  *
2181  * Return 0 on success or a negative errno.
2182  */
2183 int nfsd_nl_pool_mode_get_doit(struct sk_buff *skb, struct genl_info *info)
2184 {
2185 	struct net *net = genl_info_net(info);
2186 	char buf[16];
2187 	void *hdr;
2188 	int err;
2189 
2190 	if (sunrpc_get_pool_mode(buf, ARRAY_SIZE(buf)) >= ARRAY_SIZE(buf))
2191 		return -ERANGE;
2192 
2193 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
2194 	if (!skb)
2195 		return -ENOMEM;
2196 
2197 	err = -EMSGSIZE;
2198 	hdr = genlmsg_iput(skb, info);
2199 	if (!hdr)
2200 		goto err_free_msg;
2201 
2202 	err = nla_put_string(skb, NFSD_A_POOL_MODE_MODE, buf) |
2203 	      nla_put_u32(skb, NFSD_A_POOL_MODE_NPOOLS, nfsd_nrpools(net));
2204 	if (err)
2205 		goto err_free_msg;
2206 
2207 	genlmsg_end(skb, hdr);
2208 	return genlmsg_reply(skb, info);
2209 
2210 err_free_msg:
2211 	nlmsg_free(skb);
2212 	return err;
2213 }
2214 
2215 /**
2216  * nfsd_net_init - Prepare the nfsd_net portion of a new net namespace
2217  * @net: a freshly-created network namespace
2218  *
2219  * This information stays around as long as the network namespace is
2220  * alive whether or not there is an NFSD instance running in the
2221  * namespace.
2222  *
2223  * Returns zero on success, or a negative errno otherwise.
2224  */
2225 static __net_init int nfsd_net_init(struct net *net)
2226 {
2227 	int retval;
2228 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2229 
2230 	retval = nfsd_export_init(net);
2231 	if (retval)
2232 		goto out_export_error;
2233 	retval = nfsd_idmap_init(net);
2234 	if (retval)
2235 		goto out_idmap_error;
2236 	retval = percpu_counter_init_many(nn->counter, 0, GFP_KERNEL,
2237 					  NFSD_STATS_COUNTERS_NUM);
2238 	if (retval)
2239 		goto out_repcache_error;
2240 	memset(&nn->nfsd_svcstats, 0, sizeof(nn->nfsd_svcstats));
2241 	nn->nfsd_svcstats.program = &nfsd_program;
2242 	nn->nfsd_versions = NULL;
2243 	nn->nfsd4_minorversions = NULL;
2244 	nn->nfsd_info.mutex = &nfsd_mutex;
2245 	nn->nfsd_serv = NULL;
2246 	nfsd4_init_leases_net(nn);
2247 	get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
2248 	seqlock_init(&nn->writeverf_lock);
2249 	nfsd_proc_stat_init(net);
2250 
2251 	return 0;
2252 
2253 out_repcache_error:
2254 	nfsd_idmap_shutdown(net);
2255 out_idmap_error:
2256 	nfsd_export_shutdown(net);
2257 out_export_error:
2258 	return retval;
2259 }
2260 
2261 /**
2262  * nfsd_net_exit - Release the nfsd_net portion of a net namespace
2263  * @net: a network namespace that is about to be destroyed
2264  *
2265  */
2266 static __net_exit void nfsd_net_exit(struct net *net)
2267 {
2268 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2269 
2270 	nfsd_proc_stat_shutdown(net);
2271 	percpu_counter_destroy_many(nn->counter, NFSD_STATS_COUNTERS_NUM);
2272 	nfsd_idmap_shutdown(net);
2273 	nfsd_export_shutdown(net);
2274 	nfsd_netns_free_versions(nn);
2275 }
2276 
2277 static struct pernet_operations nfsd_net_ops = {
2278 	.init = nfsd_net_init,
2279 	.exit = nfsd_net_exit,
2280 	.id   = &nfsd_net_id,
2281 	.size = sizeof(struct nfsd_net),
2282 };
2283 
2284 static int __init init_nfsd(void)
2285 {
2286 	int retval;
2287 
2288 	retval = nfsd4_init_slabs();
2289 	if (retval)
2290 		return retval;
2291 	retval = nfsd4_init_pnfs();
2292 	if (retval)
2293 		goto out_free_slabs;
2294 	retval = nfsd_drc_slab_create();
2295 	if (retval)
2296 		goto out_free_pnfs;
2297 	nfsd_lockd_init();	/* lockd->nfsd callbacks */
2298 	retval = create_proc_exports_entry();
2299 	if (retval)
2300 		goto out_free_lockd;
2301 	retval = register_pernet_subsys(&nfsd_net_ops);
2302 	if (retval < 0)
2303 		goto out_free_exports;
2304 	retval = register_cld_notifier();
2305 	if (retval)
2306 		goto out_free_subsys;
2307 	retval = nfsd4_create_laundry_wq();
2308 	if (retval)
2309 		goto out_free_cld;
2310 	retval = register_filesystem(&nfsd_fs_type);
2311 	if (retval)
2312 		goto out_free_all;
2313 	retval = genl_register_family(&nfsd_nl_family);
2314 	if (retval)
2315 		goto out_free_all;
2316 
2317 	return 0;
2318 out_free_all:
2319 	nfsd4_destroy_laundry_wq();
2320 out_free_cld:
2321 	unregister_cld_notifier();
2322 out_free_subsys:
2323 	unregister_pernet_subsys(&nfsd_net_ops);
2324 out_free_exports:
2325 	remove_proc_entry("fs/nfs/exports", NULL);
2326 	remove_proc_entry("fs/nfs", NULL);
2327 out_free_lockd:
2328 	nfsd_lockd_shutdown();
2329 	nfsd_drc_slab_free();
2330 out_free_pnfs:
2331 	nfsd4_exit_pnfs();
2332 out_free_slabs:
2333 	nfsd4_free_slabs();
2334 	return retval;
2335 }
2336 
2337 static void __exit exit_nfsd(void)
2338 {
2339 	genl_unregister_family(&nfsd_nl_family);
2340 	unregister_filesystem(&nfsd_fs_type);
2341 	nfsd4_destroy_laundry_wq();
2342 	unregister_cld_notifier();
2343 	unregister_pernet_subsys(&nfsd_net_ops);
2344 	nfsd_drc_slab_free();
2345 	remove_proc_entry("fs/nfs/exports", NULL);
2346 	remove_proc_entry("fs/nfs", NULL);
2347 	nfsd_lockd_shutdown();
2348 	nfsd4_free_slabs();
2349 	nfsd4_exit_pnfs();
2350 }
2351 
2352 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2353 MODULE_DESCRIPTION("In-kernel NFS server");
2354 MODULE_LICENSE("GPL");
2355 module_init(init_nfsd)
2356 module_exit(exit_nfsd)
2357