xref: /linux/fs/nfsd/nfsctl.c (revision 4b132aacb0768ac1e652cf517097ea6f237214b9)
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 int nfsd_pool_stats_open(struct inode *inode, struct file *file)
178 {
179 	struct nfsd_net *nn = net_generic(inode->i_sb->s_fs_info, nfsd_net_id);
180 
181 	return svc_pool_stats_open(&nn->nfsd_info, file);
182 }
183 
184 static const struct file_operations pool_stats_operations = {
185 	.open		= nfsd_pool_stats_open,
186 	.read		= seq_read,
187 	.llseek		= seq_lseek,
188 	.release	= seq_release,
189 };
190 
191 DEFINE_SHOW_ATTRIBUTE(nfsd_reply_cache_stats);
192 
193 DEFINE_SHOW_ATTRIBUTE(nfsd_file_cache_stats);
194 
195 /*----------------------------------------------------------------------------*/
196 /*
197  * payload - write methods
198  */
199 
200 static inline struct net *netns(struct file *file)
201 {
202 	return file_inode(file)->i_sb->s_fs_info;
203 }
204 
205 /*
206  * write_unlock_ip - Release all locks used by a client
207  *
208  * Experimental.
209  *
210  * Input:
211  *			buf:	'\n'-terminated C string containing a
212  *				presentation format IP address
213  *			size:	length of C string in @buf
214  * Output:
215  *	On success:	returns zero if all specified locks were released;
216  *			returns one if one or more locks were not released
217  *	On error:	return code is negative errno value
218  */
219 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
220 {
221 	struct sockaddr_storage address;
222 	struct sockaddr *sap = (struct sockaddr *)&address;
223 	size_t salen = sizeof(address);
224 	char *fo_path;
225 	struct net *net = netns(file);
226 
227 	/* sanity check */
228 	if (size == 0)
229 		return -EINVAL;
230 
231 	if (buf[size-1] != '\n')
232 		return -EINVAL;
233 
234 	fo_path = buf;
235 	if (qword_get(&buf, fo_path, size) < 0)
236 		return -EINVAL;
237 
238 	if (rpc_pton(net, fo_path, size, sap, salen) == 0)
239 		return -EINVAL;
240 
241 	trace_nfsd_ctl_unlock_ip(net, buf);
242 	return nlmsvc_unlock_all_by_ip(sap);
243 }
244 
245 /*
246  * write_unlock_fs - Release all locks on a local file system
247  *
248  * Experimental.
249  *
250  * Input:
251  *			buf:	'\n'-terminated C string containing the
252  *				absolute pathname of a local file system
253  *			size:	length of C string in @buf
254  * Output:
255  *	On success:	returns zero if all specified locks were released;
256  *			returns one if one or more locks were not released
257  *	On error:	return code is negative errno value
258  */
259 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
260 {
261 	struct path path;
262 	char *fo_path;
263 	int error;
264 
265 	/* sanity check */
266 	if (size == 0)
267 		return -EINVAL;
268 
269 	if (buf[size-1] != '\n')
270 		return -EINVAL;
271 
272 	fo_path = buf;
273 	if (qword_get(&buf, fo_path, size) < 0)
274 		return -EINVAL;
275 	trace_nfsd_ctl_unlock_fs(netns(file), fo_path);
276 	error = kern_path(fo_path, 0, &path);
277 	if (error)
278 		return error;
279 
280 	/*
281 	 * XXX: Needs better sanity checking.  Otherwise we could end up
282 	 * releasing locks on the wrong file system.
283 	 *
284 	 * For example:
285 	 * 1.  Does the path refer to a directory?
286 	 * 2.  Is that directory a mount point, or
287 	 * 3.  Is that directory the root of an exported file system?
288 	 */
289 	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
290 	nfsd4_revoke_states(netns(file), path.dentry->d_sb);
291 
292 	path_put(&path);
293 	return error;
294 }
295 
296 /*
297  * write_filehandle - Get a variable-length NFS file handle by path
298  *
299  * On input, the buffer contains a '\n'-terminated C string comprised of
300  * three alphanumeric words separated by whitespace.  The string may
301  * contain escape sequences.
302  *
303  * Input:
304  *			buf:
305  *				domain:		client domain name
306  *				path:		export pathname
307  *				maxsize:	numeric maximum size of
308  *						@buf
309  *			size:	length of C string in @buf
310  * Output:
311  *	On success:	passed-in buffer filled with '\n'-terminated C
312  *			string containing a ASCII hex text version
313  *			of the NFS file handle;
314  *			return code is the size in bytes of the string
315  *	On error:	return code is negative errno value
316  */
317 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
318 {
319 	char *dname, *path;
320 	int maxsize;
321 	char *mesg = buf;
322 	int len;
323 	struct auth_domain *dom;
324 	struct knfsd_fh fh;
325 
326 	if (size == 0)
327 		return -EINVAL;
328 
329 	if (buf[size-1] != '\n')
330 		return -EINVAL;
331 	buf[size-1] = 0;
332 
333 	dname = mesg;
334 	len = qword_get(&mesg, dname, size);
335 	if (len <= 0)
336 		return -EINVAL;
337 
338 	path = dname+len+1;
339 	len = qword_get(&mesg, path, size);
340 	if (len <= 0)
341 		return -EINVAL;
342 
343 	len = get_int(&mesg, &maxsize);
344 	if (len)
345 		return len;
346 
347 	if (maxsize < NFS_FHSIZE)
348 		return -EINVAL;
349 	maxsize = min(maxsize, NFS3_FHSIZE);
350 
351 	if (qword_get(&mesg, mesg, size) > 0)
352 		return -EINVAL;
353 
354 	trace_nfsd_ctl_filehandle(netns(file), dname, path, maxsize);
355 
356 	/* we have all the words, they are in buf.. */
357 	dom = unix_domain_find(dname);
358 	if (!dom)
359 		return -ENOMEM;
360 
361 	len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
362 	auth_domain_put(dom);
363 	if (len)
364 		return len;
365 
366 	mesg = buf;
367 	len = SIMPLE_TRANSACTION_LIMIT;
368 	qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
369 	mesg[-1] = '\n';
370 	return mesg - buf;
371 }
372 
373 /*
374  * write_threads - Start NFSD, or report the current number of running threads
375  *
376  * Input:
377  *			buf:		ignored
378  *			size:		zero
379  * Output:
380  *	On success:	passed-in buffer filled with '\n'-terminated C
381  *			string numeric value representing the number of
382  *			running NFSD threads;
383  *			return code is the size in bytes of the string
384  *	On error:	return code is zero
385  *
386  * OR
387  *
388  * Input:
389  *			buf:		C string containing an unsigned
390  *					integer value representing the
391  *					number of NFSD threads to start
392  *			size:		non-zero length of C string in @buf
393  * Output:
394  *	On success:	NFS service is started;
395  *			passed-in buffer filled with '\n'-terminated C
396  *			string numeric value representing the number of
397  *			running NFSD threads;
398  *			return code is the size in bytes of the string
399  *	On error:	return code is zero or a negative errno value
400  */
401 static ssize_t write_threads(struct file *file, char *buf, size_t size)
402 {
403 	char *mesg = buf;
404 	int rv;
405 	struct net *net = netns(file);
406 
407 	if (size > 0) {
408 		int newthreads;
409 		rv = get_int(&mesg, &newthreads);
410 		if (rv)
411 			return rv;
412 		if (newthreads < 0)
413 			return -EINVAL;
414 		trace_nfsd_ctl_threads(net, newthreads);
415 		mutex_lock(&nfsd_mutex);
416 		rv = nfsd_svc(1, &newthreads, net, file->f_cred, NULL);
417 		mutex_unlock(&nfsd_mutex);
418 		if (rv < 0)
419 			return rv;
420 	} else
421 		rv = nfsd_nrthreads(net);
422 
423 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
424 }
425 
426 /*
427  * write_pool_threads - Set or report the current number of threads per pool
428  *
429  * Input:
430  *			buf:		ignored
431  *			size:		zero
432  *
433  * OR
434  *
435  * Input:
436  *			buf:		C string containing whitespace-
437  *					separated unsigned integer values
438  *					representing the number of NFSD
439  *					threads to start in each pool
440  *			size:		non-zero length of C string in @buf
441  * Output:
442  *	On success:	passed-in buffer filled with '\n'-terminated C
443  *			string containing integer values representing the
444  *			number of NFSD threads in each pool;
445  *			return code is the size in bytes of the string
446  *	On error:	return code is zero or a negative errno value
447  */
448 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
449 {
450 	/* if size > 0, look for an array of number of threads per node
451 	 * and apply them  then write out number of threads per node as reply
452 	 */
453 	char *mesg = buf;
454 	int i;
455 	int rv;
456 	int len;
457 	int npools;
458 	int *nthreads;
459 	struct net *net = netns(file);
460 
461 	mutex_lock(&nfsd_mutex);
462 	npools = nfsd_nrpools(net);
463 	if (npools == 0) {
464 		/*
465 		 * NFS is shut down.  The admin can start it by
466 		 * writing to the threads file but NOT the pool_threads
467 		 * file, sorry.  Report zero threads.
468 		 */
469 		mutex_unlock(&nfsd_mutex);
470 		strcpy(buf, "0\n");
471 		return strlen(buf);
472 	}
473 
474 	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
475 	rv = -ENOMEM;
476 	if (nthreads == NULL)
477 		goto out_free;
478 
479 	if (size > 0) {
480 		for (i = 0; i < npools; i++) {
481 			rv = get_int(&mesg, &nthreads[i]);
482 			if (rv == -ENOENT)
483 				break;		/* fewer numbers than pools */
484 			if (rv)
485 				goto out_free;	/* syntax error */
486 			rv = -EINVAL;
487 			if (nthreads[i] < 0)
488 				goto out_free;
489 			trace_nfsd_ctl_pool_threads(net, i, nthreads[i]);
490 		}
491 
492 		/*
493 		 * There must always be a thread in pool 0; the admin
494 		 * can't shut down NFS completely using pool_threads.
495 		 */
496 		if (nthreads[0] == 0)
497 			nthreads[0] = 1;
498 
499 		rv = nfsd_set_nrthreads(i, nthreads, net);
500 		if (rv)
501 			goto out_free;
502 	}
503 
504 	rv = nfsd_get_nrthreads(npools, nthreads, net);
505 	if (rv)
506 		goto out_free;
507 
508 	mesg = buf;
509 	size = SIMPLE_TRANSACTION_LIMIT;
510 	for (i = 0; i < npools && size > 0; i++) {
511 		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
512 		len = strlen(mesg);
513 		size -= len;
514 		mesg += len;
515 	}
516 	rv = mesg - buf;
517 out_free:
518 	kfree(nthreads);
519 	mutex_unlock(&nfsd_mutex);
520 	return rv;
521 }
522 
523 static ssize_t
524 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
525 		const char *sep, unsigned vers, int minor)
526 {
527 	const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
528 	bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
529 
530 	if (vers == 4 && minor >= 0 &&
531 	    !nfsd_minorversion(nn, minor, NFSD_TEST))
532 		supported = false;
533 	if (minor == 0 && supported)
534 		/*
535 		 * special case for backward compatability.
536 		 * +4.0 is never reported, it is implied by
537 		 * +4, unless -4.0 is present.
538 		 */
539 		return 0;
540 	return snprintf(buf, remaining, format, sep,
541 			supported ? '+' : '-', vers, minor);
542 }
543 
544 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
545 {
546 	char *mesg = buf;
547 	char *vers, *minorp, sign;
548 	int len, num, remaining;
549 	ssize_t tlen = 0;
550 	char *sep;
551 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
552 
553 	if (size > 0) {
554 		if (nn->nfsd_serv)
555 			/* Cannot change versions without updating
556 			 * nn->nfsd_serv->sv_xdrsize, and reallocing
557 			 * rq_argp and rq_resp
558 			 */
559 			return -EBUSY;
560 		if (buf[size-1] != '\n')
561 			return -EINVAL;
562 		buf[size-1] = 0;
563 		trace_nfsd_ctl_version(netns(file), buf);
564 
565 		vers = mesg;
566 		len = qword_get(&mesg, vers, size);
567 		if (len <= 0) return -EINVAL;
568 		do {
569 			enum vers_op cmd;
570 			unsigned minor;
571 			sign = *vers;
572 			if (sign == '+' || sign == '-')
573 				num = simple_strtol((vers+1), &minorp, 0);
574 			else
575 				num = simple_strtol(vers, &minorp, 0);
576 			if (*minorp == '.') {
577 				if (num != 4)
578 					return -EINVAL;
579 				if (kstrtouint(minorp+1, 0, &minor) < 0)
580 					return -EINVAL;
581 			}
582 
583 			cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
584 			switch(num) {
585 #ifdef CONFIG_NFSD_V2
586 			case 2:
587 #endif
588 			case 3:
589 				nfsd_vers(nn, num, cmd);
590 				break;
591 			case 4:
592 				if (*minorp == '.') {
593 					if (nfsd_minorversion(nn, minor, cmd) < 0)
594 						return -EINVAL;
595 				} else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
596 					/*
597 					 * Either we have +4 and no minors are enabled,
598 					 * or we have -4 and at least one minor is enabled.
599 					 * In either case, propagate 'cmd' to all minors.
600 					 */
601 					minor = 0;
602 					while (nfsd_minorversion(nn, minor, cmd) >= 0)
603 						minor++;
604 				}
605 				break;
606 			default:
607 				/* Ignore requests to disable non-existent versions */
608 				if (cmd == NFSD_SET)
609 					return -EINVAL;
610 			}
611 			vers += len + 1;
612 		} while ((len = qword_get(&mesg, vers, size)) > 0);
613 		/* If all get turned off, turn them back on, as
614 		 * having no versions is BAD
615 		 */
616 		nfsd_reset_versions(nn);
617 	}
618 
619 	/* Now write current state into reply buffer */
620 	sep = "";
621 	remaining = SIMPLE_TRANSACTION_LIMIT;
622 	for (num=2 ; num <= 4 ; num++) {
623 		int minor;
624 		if (!nfsd_vers(nn, num, NFSD_AVAIL))
625 			continue;
626 
627 		minor = -1;
628 		do {
629 			len = nfsd_print_version_support(nn, buf, remaining,
630 					sep, num, minor);
631 			if (len >= remaining)
632 				goto out;
633 			remaining -= len;
634 			buf += len;
635 			tlen += len;
636 			minor++;
637 			if (len)
638 				sep = " ";
639 		} while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
640 	}
641 out:
642 	len = snprintf(buf, remaining, "\n");
643 	if (len >= remaining)
644 		return -EINVAL;
645 	return tlen + len;
646 }
647 
648 /*
649  * write_versions - Set or report the available NFS protocol versions
650  *
651  * Input:
652  *			buf:		ignored
653  *			size:		zero
654  * Output:
655  *	On success:	passed-in buffer filled with '\n'-terminated C
656  *			string containing positive or negative integer
657  *			values representing the current status of each
658  *			protocol version;
659  *			return code is the size in bytes of the string
660  *	On error:	return code is zero or a negative errno value
661  *
662  * OR
663  *
664  * Input:
665  *			buf:		C string containing whitespace-
666  *					separated positive or negative
667  *					integer values representing NFS
668  *					protocol versions to enable ("+n")
669  *					or disable ("-n")
670  *			size:		non-zero length of C string in @buf
671  * Output:
672  *	On success:	status of zero or more protocol versions has
673  *			been updated; passed-in buffer filled with
674  *			'\n'-terminated C string containing positive
675  *			or negative integer values representing the
676  *			current status of each protocol version;
677  *			return code is the size in bytes of the string
678  *	On error:	return code is zero or a negative errno value
679  */
680 static ssize_t write_versions(struct file *file, char *buf, size_t size)
681 {
682 	ssize_t rv;
683 
684 	mutex_lock(&nfsd_mutex);
685 	rv = __write_versions(file, buf, size);
686 	mutex_unlock(&nfsd_mutex);
687 	return rv;
688 }
689 
690 /*
691  * Zero-length write.  Return a list of NFSD's current listener
692  * transports.
693  */
694 static ssize_t __write_ports_names(char *buf, struct net *net)
695 {
696 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
697 
698 	if (nn->nfsd_serv == NULL)
699 		return 0;
700 	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
701 }
702 
703 /*
704  * A single 'fd' number was written, in which case it must be for
705  * a socket of a supported family/protocol, and we use it as an
706  * nfsd listener.
707  */
708 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
709 {
710 	char *mesg = buf;
711 	int fd, err;
712 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
713 	struct svc_serv *serv;
714 
715 	err = get_int(&mesg, &fd);
716 	if (err != 0 || fd < 0)
717 		return -EINVAL;
718 	trace_nfsd_ctl_ports_addfd(net, fd);
719 
720 	err = nfsd_create_serv(net);
721 	if (err != 0)
722 		return err;
723 
724 	serv = nn->nfsd_serv;
725 	err = svc_addsock(serv, net, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
726 
727 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
728 		nfsd_destroy_serv(net);
729 
730 	return err;
731 }
732 
733 /*
734  * A transport listener is added by writing its transport name and
735  * a port number.
736  */
737 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
738 {
739 	char transport[16];
740 	struct svc_xprt *xprt;
741 	int port, err;
742 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
743 	struct svc_serv *serv;
744 
745 	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
746 		return -EINVAL;
747 
748 	if (port < 1 || port > USHRT_MAX)
749 		return -EINVAL;
750 	trace_nfsd_ctl_ports_addxprt(net, transport, port);
751 
752 	err = nfsd_create_serv(net);
753 	if (err != 0)
754 		return err;
755 
756 	serv = nn->nfsd_serv;
757 	err = svc_xprt_create(serv, transport, net,
758 			      PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
759 	if (err < 0)
760 		goto out_err;
761 
762 	err = svc_xprt_create(serv, transport, net,
763 			      PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
764 	if (err < 0 && err != -EAFNOSUPPORT)
765 		goto out_close;
766 
767 	return 0;
768 out_close:
769 	xprt = svc_find_xprt(serv, transport, net, PF_INET, port);
770 	if (xprt != NULL) {
771 		svc_xprt_close(xprt);
772 		svc_xprt_put(xprt);
773 	}
774 out_err:
775 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
776 		nfsd_destroy_serv(net);
777 
778 	return err;
779 }
780 
781 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
782 			     struct net *net)
783 {
784 	if (size == 0)
785 		return __write_ports_names(buf, net);
786 
787 	if (isdigit(buf[0]))
788 		return __write_ports_addfd(buf, net, file->f_cred);
789 
790 	if (isalpha(buf[0]))
791 		return __write_ports_addxprt(buf, net, file->f_cred);
792 
793 	return -EINVAL;
794 }
795 
796 /*
797  * write_ports - Pass a socket file descriptor or transport name to listen on
798  *
799  * Input:
800  *			buf:		ignored
801  *			size:		zero
802  * Output:
803  *	On success:	passed-in buffer filled with a '\n'-terminated C
804  *			string containing a whitespace-separated list of
805  *			named NFSD listeners;
806  *			return code is the size in bytes of the string
807  *	On error:	return code is zero or a negative errno value
808  *
809  * OR
810  *
811  * Input:
812  *			buf:		C string containing an unsigned
813  *					integer value representing a bound
814  *					but unconnected socket that is to be
815  *					used as an NFSD listener; listen(3)
816  *					must be called for a SOCK_STREAM
817  *					socket, otherwise it is ignored
818  *			size:		non-zero length of C string in @buf
819  * Output:
820  *	On success:	NFS service is started;
821  *			passed-in buffer filled with a '\n'-terminated C
822  *			string containing a unique alphanumeric name of
823  *			the listener;
824  *			return code is the size in bytes of the string
825  *	On error:	return code is a negative errno value
826  *
827  * OR
828  *
829  * Input:
830  *			buf:		C string containing a transport
831  *					name and an unsigned integer value
832  *					representing the port to listen on,
833  *					separated by whitespace
834  *			size:		non-zero length of C string in @buf
835  * Output:
836  *	On success:	returns zero; NFS service is started
837  *	On error:	return code is a negative errno value
838  */
839 static ssize_t write_ports(struct file *file, char *buf, size_t size)
840 {
841 	ssize_t rv;
842 
843 	mutex_lock(&nfsd_mutex);
844 	rv = __write_ports(file, buf, size, netns(file));
845 	mutex_unlock(&nfsd_mutex);
846 	return rv;
847 }
848 
849 
850 int nfsd_max_blksize;
851 
852 /*
853  * write_maxblksize - Set or report the current NFS blksize
854  *
855  * Input:
856  *			buf:		ignored
857  *			size:		zero
858  *
859  * OR
860  *
861  * Input:
862  *			buf:		C string containing an unsigned
863  *					integer value representing the new
864  *					NFS blksize
865  *			size:		non-zero length of C string in @buf
866  * Output:
867  *	On success:	passed-in buffer filled with '\n'-terminated C string
868  *			containing numeric value of the current NFS blksize
869  *			setting;
870  *			return code is the size in bytes of the string
871  *	On error:	return code is zero or a negative errno value
872  */
873 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
874 {
875 	char *mesg = buf;
876 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
877 
878 	if (size > 0) {
879 		int bsize;
880 		int rv = get_int(&mesg, &bsize);
881 		if (rv)
882 			return rv;
883 		trace_nfsd_ctl_maxblksize(netns(file), bsize);
884 
885 		/* force bsize into allowed range and
886 		 * required alignment.
887 		 */
888 		bsize = max_t(int, bsize, 1024);
889 		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
890 		bsize &= ~(1024-1);
891 		mutex_lock(&nfsd_mutex);
892 		if (nn->nfsd_serv) {
893 			mutex_unlock(&nfsd_mutex);
894 			return -EBUSY;
895 		}
896 		nfsd_max_blksize = bsize;
897 		mutex_unlock(&nfsd_mutex);
898 	}
899 
900 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
901 							nfsd_max_blksize);
902 }
903 
904 /*
905  * write_maxconn - Set or report the current max number of connections
906  *
907  * Input:
908  *			buf:		ignored
909  *			size:		zero
910  * OR
911  *
912  * Input:
913  *			buf:		C string containing an unsigned
914  *					integer value representing the new
915  *					number of max connections
916  *			size:		non-zero length of C string in @buf
917  * Output:
918  *	On success:	passed-in buffer filled with '\n'-terminated C string
919  *			containing numeric value of max_connections setting
920  *			for this net namespace;
921  *			return code is the size in bytes of the string
922  *	On error:	return code is zero or a negative errno value
923  */
924 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
925 {
926 	char *mesg = buf;
927 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
928 	unsigned int maxconn = nn->max_connections;
929 
930 	if (size > 0) {
931 		int rv = get_uint(&mesg, &maxconn);
932 
933 		if (rv)
934 			return rv;
935 		trace_nfsd_ctl_maxconn(netns(file), maxconn);
936 		nn->max_connections = maxconn;
937 	}
938 
939 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
940 }
941 
942 #ifdef CONFIG_NFSD_V4
943 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
944 				  time64_t *time, struct nfsd_net *nn)
945 {
946 	struct dentry *dentry = file_dentry(file);
947 	char *mesg = buf;
948 	int rv, i;
949 
950 	if (size > 0) {
951 		if (nn->nfsd_serv)
952 			return -EBUSY;
953 		rv = get_int(&mesg, &i);
954 		if (rv)
955 			return rv;
956 		trace_nfsd_ctl_time(netns(file), dentry->d_name.name,
957 				    dentry->d_name.len, i);
958 
959 		/*
960 		 * Some sanity checking.  We don't have a reason for
961 		 * these particular numbers, but problems with the
962 		 * extremes are:
963 		 *	- Too short: the briefest network outage may
964 		 *	  cause clients to lose all their locks.  Also,
965 		 *	  the frequent polling may be wasteful.
966 		 *	- Too long: do you really want reboot recovery
967 		 *	  to take more than an hour?  Or to make other
968 		 *	  clients wait an hour before being able to
969 		 *	  revoke a dead client's locks?
970 		 */
971 		if (i < 10 || i > 3600)
972 			return -EINVAL;
973 		*time = i;
974 	}
975 
976 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
977 }
978 
979 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
980 				time64_t *time, struct nfsd_net *nn)
981 {
982 	ssize_t rv;
983 
984 	mutex_lock(&nfsd_mutex);
985 	rv = __nfsd4_write_time(file, buf, size, time, nn);
986 	mutex_unlock(&nfsd_mutex);
987 	return rv;
988 }
989 
990 /*
991  * write_leasetime - Set or report the current NFSv4 lease time
992  *
993  * Input:
994  *			buf:		ignored
995  *			size:		zero
996  *
997  * OR
998  *
999  * Input:
1000  *			buf:		C string containing an unsigned
1001  *					integer value representing the new
1002  *					NFSv4 lease expiry time
1003  *			size:		non-zero length of C string in @buf
1004  * Output:
1005  *	On success:	passed-in buffer filled with '\n'-terminated C
1006  *			string containing unsigned integer value of the
1007  *			current lease expiry time;
1008  *			return code is the size in bytes of the string
1009  *	On error:	return code is zero or a negative errno value
1010  */
1011 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1012 {
1013 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1014 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1015 }
1016 
1017 /*
1018  * write_gracetime - Set or report current NFSv4 grace period time
1019  *
1020  * As above, but sets the time of the NFSv4 grace period.
1021  *
1022  * Note this should never be set to less than the *previous*
1023  * lease-period time, but we don't try to enforce this.  (In the common
1024  * case (a new boot), we don't know what the previous lease time was
1025  * anyway.)
1026  */
1027 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1028 {
1029 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1030 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1031 }
1032 
1033 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1034 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1035 				   struct nfsd_net *nn)
1036 {
1037 	char *mesg = buf;
1038 	char *recdir;
1039 	int len, status;
1040 
1041 	if (size > 0) {
1042 		if (nn->nfsd_serv)
1043 			return -EBUSY;
1044 		if (size > PATH_MAX || buf[size-1] != '\n')
1045 			return -EINVAL;
1046 		buf[size-1] = 0;
1047 
1048 		recdir = mesg;
1049 		len = qword_get(&mesg, recdir, size);
1050 		if (len <= 0)
1051 			return -EINVAL;
1052 		trace_nfsd_ctl_recoverydir(netns(file), recdir);
1053 
1054 		status = nfs4_reset_recoverydir(recdir);
1055 		if (status)
1056 			return status;
1057 	}
1058 
1059 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1060 							nfs4_recoverydir());
1061 }
1062 
1063 /*
1064  * write_recoverydir - Set or report the pathname of the recovery directory
1065  *
1066  * Input:
1067  *			buf:		ignored
1068  *			size:		zero
1069  *
1070  * OR
1071  *
1072  * Input:
1073  *			buf:		C string containing the pathname
1074  *					of the directory on a local file
1075  *					system containing permanent NFSv4
1076  *					recovery data
1077  *			size:		non-zero length of C string in @buf
1078  * Output:
1079  *	On success:	passed-in buffer filled with '\n'-terminated C string
1080  *			containing the current recovery pathname setting;
1081  *			return code is the size in bytes of the string
1082  *	On error:	return code is zero or a negative errno value
1083  */
1084 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1085 {
1086 	ssize_t rv;
1087 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1088 
1089 	mutex_lock(&nfsd_mutex);
1090 	rv = __write_recoverydir(file, buf, size, nn);
1091 	mutex_unlock(&nfsd_mutex);
1092 	return rv;
1093 }
1094 #endif
1095 
1096 /*
1097  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1098  *
1099  * Input:
1100  *			buf:		ignored
1101  *			size:		zero
1102  * OR
1103  *
1104  * Input:
1105  *			buf:		any value
1106  *			size:		non-zero length of C string in @buf
1107  * Output:
1108  *			passed-in buffer filled with "Y" or "N" with a newline
1109  *			and NULL-terminated C string. This indicates whether
1110  *			the grace period has ended in the current net
1111  *			namespace. Return code is the size in bytes of the
1112  *			string. Writing a string that starts with 'Y', 'y', or
1113  *			'1' to the file will end the grace period for nfsd's v4
1114  *			lock manager.
1115  */
1116 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1117 {
1118 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1119 
1120 	if (size > 0) {
1121 		switch(buf[0]) {
1122 		case 'Y':
1123 		case 'y':
1124 		case '1':
1125 			if (!nn->nfsd_serv)
1126 				return -EBUSY;
1127 			trace_nfsd_end_grace(netns(file));
1128 			nfsd4_end_grace(nn);
1129 			break;
1130 		default:
1131 			return -EINVAL;
1132 		}
1133 	}
1134 
1135 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1136 			 nn->grace_ended ? 'Y' : 'N');
1137 }
1138 
1139 #endif
1140 
1141 /*----------------------------------------------------------------------------*/
1142 /*
1143  *	populating the filesystem.
1144  */
1145 
1146 /* Basically copying rpc_get_inode. */
1147 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1148 {
1149 	struct inode *inode = new_inode(sb);
1150 	if (!inode)
1151 		return NULL;
1152 	/* Following advice from simple_fill_super documentation: */
1153 	inode->i_ino = iunique(sb, NFSD_MaxReserved);
1154 	inode->i_mode = mode;
1155 	simple_inode_init_ts(inode);
1156 	switch (mode & S_IFMT) {
1157 	case S_IFDIR:
1158 		inode->i_fop = &simple_dir_operations;
1159 		inode->i_op = &simple_dir_inode_operations;
1160 		inc_nlink(inode);
1161 		break;
1162 	case S_IFLNK:
1163 		inode->i_op = &simple_symlink_inode_operations;
1164 		break;
1165 	default:
1166 		break;
1167 	}
1168 	return inode;
1169 }
1170 
1171 static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1172 {
1173 	struct inode *inode;
1174 
1175 	inode = nfsd_get_inode(dir->i_sb, mode);
1176 	if (!inode)
1177 		return -ENOMEM;
1178 	if (ncl) {
1179 		inode->i_private = ncl;
1180 		kref_get(&ncl->cl_ref);
1181 	}
1182 	d_add(dentry, inode);
1183 	inc_nlink(dir);
1184 	fsnotify_mkdir(dir, dentry);
1185 	return 0;
1186 }
1187 
1188 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1189 {
1190 	struct inode *dir = parent->d_inode;
1191 	struct dentry *dentry;
1192 	int ret = -ENOMEM;
1193 
1194 	inode_lock(dir);
1195 	dentry = d_alloc_name(parent, name);
1196 	if (!dentry)
1197 		goto out_err;
1198 	ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1199 	if (ret)
1200 		goto out_err;
1201 out:
1202 	inode_unlock(dir);
1203 	return dentry;
1204 out_err:
1205 	dput(dentry);
1206 	dentry = ERR_PTR(ret);
1207 	goto out;
1208 }
1209 
1210 #if IS_ENABLED(CONFIG_SUNRPC_GSS)
1211 static int __nfsd_symlink(struct inode *dir, struct dentry *dentry,
1212 			  umode_t mode, const char *content)
1213 {
1214 	struct inode *inode;
1215 
1216 	inode = nfsd_get_inode(dir->i_sb, mode);
1217 	if (!inode)
1218 		return -ENOMEM;
1219 
1220 	inode->i_link = (char *)content;
1221 	inode->i_size = strlen(content);
1222 
1223 	d_add(dentry, inode);
1224 	inc_nlink(dir);
1225 	fsnotify_create(dir, dentry);
1226 	return 0;
1227 }
1228 
1229 /*
1230  * @content is assumed to be a NUL-terminated string that lives
1231  * longer than the symlink itself.
1232  */
1233 static void _nfsd_symlink(struct dentry *parent, const char *name,
1234 			  const char *content)
1235 {
1236 	struct inode *dir = parent->d_inode;
1237 	struct dentry *dentry;
1238 	int ret;
1239 
1240 	inode_lock(dir);
1241 	dentry = d_alloc_name(parent, name);
1242 	if (!dentry)
1243 		goto out;
1244 	ret = __nfsd_symlink(d_inode(parent), dentry, S_IFLNK | 0777, content);
1245 	if (ret)
1246 		dput(dentry);
1247 out:
1248 	inode_unlock(dir);
1249 }
1250 #else
1251 static inline void _nfsd_symlink(struct dentry *parent, const char *name,
1252 				 const char *content)
1253 {
1254 }
1255 
1256 #endif
1257 
1258 static void clear_ncl(struct dentry *dentry)
1259 {
1260 	struct inode *inode = d_inode(dentry);
1261 	struct nfsdfs_client *ncl = inode->i_private;
1262 
1263 	spin_lock(&inode->i_lock);
1264 	inode->i_private = NULL;
1265 	spin_unlock(&inode->i_lock);
1266 	kref_put(&ncl->cl_ref, ncl->cl_release);
1267 }
1268 
1269 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1270 {
1271 	struct nfsdfs_client *nc;
1272 
1273 	spin_lock(&inode->i_lock);
1274 	nc = inode->i_private;
1275 	if (nc)
1276 		kref_get(&nc->cl_ref);
1277 	spin_unlock(&inode->i_lock);
1278 	return nc;
1279 }
1280 
1281 /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1282  * code instead. */
1283 static  int nfsdfs_create_files(struct dentry *root,
1284 				const struct tree_descr *files,
1285 				struct nfsdfs_client *ncl,
1286 				struct dentry **fdentries)
1287 {
1288 	struct inode *dir = d_inode(root);
1289 	struct inode *inode;
1290 	struct dentry *dentry;
1291 	int i;
1292 
1293 	inode_lock(dir);
1294 	for (i = 0; files->name && files->name[0]; i++, files++) {
1295 		dentry = d_alloc_name(root, files->name);
1296 		if (!dentry)
1297 			goto out;
1298 		inode = nfsd_get_inode(d_inode(root)->i_sb,
1299 					S_IFREG | files->mode);
1300 		if (!inode) {
1301 			dput(dentry);
1302 			goto out;
1303 		}
1304 		kref_get(&ncl->cl_ref);
1305 		inode->i_fop = files->ops;
1306 		inode->i_private = ncl;
1307 		d_add(dentry, inode);
1308 		fsnotify_create(dir, dentry);
1309 		if (fdentries)
1310 			fdentries[i] = dentry;
1311 	}
1312 	inode_unlock(dir);
1313 	return 0;
1314 out:
1315 	inode_unlock(dir);
1316 	return -ENOMEM;
1317 }
1318 
1319 /* on success, returns positive number unique to that client. */
1320 struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1321 				 struct nfsdfs_client *ncl, u32 id,
1322 				 const struct tree_descr *files,
1323 				 struct dentry **fdentries)
1324 {
1325 	struct dentry *dentry;
1326 	char name[11];
1327 	int ret;
1328 
1329 	sprintf(name, "%u", id);
1330 
1331 	dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1332 	if (IS_ERR(dentry)) /* XXX: tossing errors? */
1333 		return NULL;
1334 	ret = nfsdfs_create_files(dentry, files, ncl, fdentries);
1335 	if (ret) {
1336 		nfsd_client_rmdir(dentry);
1337 		return NULL;
1338 	}
1339 	return dentry;
1340 }
1341 
1342 /* Taken from __rpc_rmdir: */
1343 void nfsd_client_rmdir(struct dentry *dentry)
1344 {
1345 	simple_recursive_removal(dentry, clear_ncl);
1346 }
1347 
1348 static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1349 {
1350 	struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1351 							nfsd_net_id);
1352 	struct dentry *dentry;
1353 	int ret;
1354 
1355 	static const struct tree_descr nfsd_files[] = {
1356 		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1357 		/* Per-export io stats use same ops as exports file */
1358 		[NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1359 		[NFSD_Export_features] = {"export_features",
1360 					&export_features_fops, S_IRUGO},
1361 		[NFSD_FO_UnlockIP] = {"unlock_ip",
1362 					&transaction_ops, S_IWUSR|S_IRUSR},
1363 		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1364 					&transaction_ops, S_IWUSR|S_IRUSR},
1365 		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1366 		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1367 		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1368 		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1369 		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1370 					&nfsd_reply_cache_stats_fops, S_IRUGO},
1371 		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1372 		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1373 		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1374 		[NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1375 		[NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
1376 #ifdef CONFIG_NFSD_V4
1377 		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1378 		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1379 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1380 		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1381 #endif
1382 		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1383 #endif
1384 		/* last one */ {""}
1385 	};
1386 
1387 	ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1388 	if (ret)
1389 		return ret;
1390 	_nfsd_symlink(sb->s_root, "supported_krb5_enctypes",
1391 		      "/proc/net/rpc/gss_krb5_enctypes");
1392 	dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1393 	if (IS_ERR(dentry))
1394 		return PTR_ERR(dentry);
1395 	nn->nfsd_client_dir = dentry;
1396 	return 0;
1397 }
1398 
1399 static int nfsd_fs_get_tree(struct fs_context *fc)
1400 {
1401 	return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1402 }
1403 
1404 static void nfsd_fs_free_fc(struct fs_context *fc)
1405 {
1406 	if (fc->s_fs_info)
1407 		put_net(fc->s_fs_info);
1408 }
1409 
1410 static const struct fs_context_operations nfsd_fs_context_ops = {
1411 	.free		= nfsd_fs_free_fc,
1412 	.get_tree	= nfsd_fs_get_tree,
1413 };
1414 
1415 static int nfsd_init_fs_context(struct fs_context *fc)
1416 {
1417 	put_user_ns(fc->user_ns);
1418 	fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1419 	fc->ops = &nfsd_fs_context_ops;
1420 	return 0;
1421 }
1422 
1423 static void nfsd_umount(struct super_block *sb)
1424 {
1425 	struct net *net = sb->s_fs_info;
1426 
1427 	nfsd_shutdown_threads(net);
1428 
1429 	kill_litter_super(sb);
1430 	put_net(net);
1431 }
1432 
1433 static struct file_system_type nfsd_fs_type = {
1434 	.owner		= THIS_MODULE,
1435 	.name		= "nfsd",
1436 	.init_fs_context = nfsd_init_fs_context,
1437 	.kill_sb	= nfsd_umount,
1438 };
1439 MODULE_ALIAS_FS("nfsd");
1440 
1441 #ifdef CONFIG_PROC_FS
1442 
1443 static int exports_proc_open(struct inode *inode, struct file *file)
1444 {
1445 	return exports_net_open(current->nsproxy->net_ns, file);
1446 }
1447 
1448 static const struct proc_ops exports_proc_ops = {
1449 	.proc_open	= exports_proc_open,
1450 	.proc_read	= seq_read,
1451 	.proc_lseek	= seq_lseek,
1452 	.proc_release	= seq_release,
1453 };
1454 
1455 static int create_proc_exports_entry(void)
1456 {
1457 	struct proc_dir_entry *entry;
1458 
1459 	entry = proc_mkdir("fs/nfs", NULL);
1460 	if (!entry)
1461 		return -ENOMEM;
1462 	entry = proc_create("exports", 0, entry, &exports_proc_ops);
1463 	if (!entry) {
1464 		remove_proc_entry("fs/nfs", NULL);
1465 		return -ENOMEM;
1466 	}
1467 	return 0;
1468 }
1469 #else /* CONFIG_PROC_FS */
1470 static int create_proc_exports_entry(void)
1471 {
1472 	return 0;
1473 }
1474 #endif
1475 
1476 unsigned int nfsd_net_id;
1477 
1478 static int nfsd_genl_rpc_status_compose_msg(struct sk_buff *skb,
1479 					    struct netlink_callback *cb,
1480 					    struct nfsd_genl_rqstp *rqstp)
1481 {
1482 	void *hdr;
1483 	u32 i;
1484 
1485 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
1486 			  &nfsd_nl_family, 0, NFSD_CMD_RPC_STATUS_GET);
1487 	if (!hdr)
1488 		return -ENOBUFS;
1489 
1490 	if (nla_put_be32(skb, NFSD_A_RPC_STATUS_XID, rqstp->rq_xid) ||
1491 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_FLAGS, rqstp->rq_flags) ||
1492 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROG, rqstp->rq_prog) ||
1493 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROC, rqstp->rq_proc) ||
1494 	    nla_put_u8(skb, NFSD_A_RPC_STATUS_VERSION, rqstp->rq_vers) ||
1495 	    nla_put_s64(skb, NFSD_A_RPC_STATUS_SERVICE_TIME,
1496 			ktime_to_us(rqstp->rq_stime),
1497 			NFSD_A_RPC_STATUS_PAD))
1498 		return -ENOBUFS;
1499 
1500 	switch (rqstp->rq_saddr.sa_family) {
1501 	case AF_INET: {
1502 		const struct sockaddr_in *s_in, *d_in;
1503 
1504 		s_in = (const struct sockaddr_in *)&rqstp->rq_saddr;
1505 		d_in = (const struct sockaddr_in *)&rqstp->rq_daddr;
1506 		if (nla_put_in_addr(skb, NFSD_A_RPC_STATUS_SADDR4,
1507 				    s_in->sin_addr.s_addr) ||
1508 		    nla_put_in_addr(skb, NFSD_A_RPC_STATUS_DADDR4,
1509 				    d_in->sin_addr.s_addr) ||
1510 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1511 				 s_in->sin_port) ||
1512 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1513 				 d_in->sin_port))
1514 			return -ENOBUFS;
1515 		break;
1516 	}
1517 	case AF_INET6: {
1518 		const struct sockaddr_in6 *s_in, *d_in;
1519 
1520 		s_in = (const struct sockaddr_in6 *)&rqstp->rq_saddr;
1521 		d_in = (const struct sockaddr_in6 *)&rqstp->rq_daddr;
1522 		if (nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_SADDR6,
1523 				     &s_in->sin6_addr) ||
1524 		    nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_DADDR6,
1525 				     &d_in->sin6_addr) ||
1526 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1527 				 s_in->sin6_port) ||
1528 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1529 				 d_in->sin6_port))
1530 			return -ENOBUFS;
1531 		break;
1532 	}
1533 	}
1534 
1535 	for (i = 0; i < rqstp->rq_opcnt; i++)
1536 		if (nla_put_u32(skb, NFSD_A_RPC_STATUS_COMPOUND_OPS,
1537 				rqstp->rq_opnum[i]))
1538 			return -ENOBUFS;
1539 
1540 	genlmsg_end(skb, hdr);
1541 	return 0;
1542 }
1543 
1544 /**
1545  * nfsd_nl_rpc_status_get_dumpit - Handle rpc_status_get dumpit
1546  * @skb: reply buffer
1547  * @cb: netlink metadata and command arguments
1548  *
1549  * Returns the size of the reply or a negative errno.
1550  */
1551 int nfsd_nl_rpc_status_get_dumpit(struct sk_buff *skb,
1552 				  struct netlink_callback *cb)
1553 {
1554 	int i, ret, rqstp_index = 0;
1555 	struct nfsd_net *nn;
1556 
1557 	mutex_lock(&nfsd_mutex);
1558 
1559 	nn = net_generic(sock_net(skb->sk), nfsd_net_id);
1560 	if (!nn->nfsd_serv) {
1561 		ret = -ENODEV;
1562 		goto out_unlock;
1563 	}
1564 
1565 	rcu_read_lock();
1566 
1567 	for (i = 0; i < nn->nfsd_serv->sv_nrpools; i++) {
1568 		struct svc_rqst *rqstp;
1569 
1570 		if (i < cb->args[0]) /* already consumed */
1571 			continue;
1572 
1573 		rqstp_index = 0;
1574 		list_for_each_entry_rcu(rqstp,
1575 				&nn->nfsd_serv->sv_pools[i].sp_all_threads,
1576 				rq_all) {
1577 			struct nfsd_genl_rqstp genl_rqstp;
1578 			unsigned int status_counter;
1579 
1580 			if (rqstp_index++ < cb->args[1]) /* already consumed */
1581 				continue;
1582 			/*
1583 			 * Acquire rq_status_counter before parsing the rqst
1584 			 * fields. rq_status_counter is set to an odd value in
1585 			 * order to notify the consumers the rqstp fields are
1586 			 * meaningful.
1587 			 */
1588 			status_counter =
1589 				smp_load_acquire(&rqstp->rq_status_counter);
1590 			if (!(status_counter & 1))
1591 				continue;
1592 
1593 			genl_rqstp.rq_xid = rqstp->rq_xid;
1594 			genl_rqstp.rq_flags = rqstp->rq_flags;
1595 			genl_rqstp.rq_vers = rqstp->rq_vers;
1596 			genl_rqstp.rq_prog = rqstp->rq_prog;
1597 			genl_rqstp.rq_proc = rqstp->rq_proc;
1598 			genl_rqstp.rq_stime = rqstp->rq_stime;
1599 			genl_rqstp.rq_opcnt = 0;
1600 			memcpy(&genl_rqstp.rq_daddr, svc_daddr(rqstp),
1601 			       sizeof(struct sockaddr));
1602 			memcpy(&genl_rqstp.rq_saddr, svc_addr(rqstp),
1603 			       sizeof(struct sockaddr));
1604 
1605 #ifdef CONFIG_NFSD_V4
1606 			if (rqstp->rq_vers == NFS4_VERSION &&
1607 			    rqstp->rq_proc == NFSPROC4_COMPOUND) {
1608 				/* NFSv4 compound */
1609 				struct nfsd4_compoundargs *args;
1610 				int j;
1611 
1612 				args = rqstp->rq_argp;
1613 				genl_rqstp.rq_opcnt = args->opcnt;
1614 				for (j = 0; j < genl_rqstp.rq_opcnt; j++)
1615 					genl_rqstp.rq_opnum[j] =
1616 						args->ops[j].opnum;
1617 			}
1618 #endif /* CONFIG_NFSD_V4 */
1619 
1620 			/*
1621 			 * Acquire rq_status_counter before reporting the rqst
1622 			 * fields to the user.
1623 			 */
1624 			if (smp_load_acquire(&rqstp->rq_status_counter) !=
1625 			    status_counter)
1626 				continue;
1627 
1628 			ret = nfsd_genl_rpc_status_compose_msg(skb, cb,
1629 							       &genl_rqstp);
1630 			if (ret)
1631 				goto out;
1632 		}
1633 	}
1634 
1635 	cb->args[0] = i;
1636 	cb->args[1] = rqstp_index;
1637 	ret = skb->len;
1638 out:
1639 	rcu_read_unlock();
1640 out_unlock:
1641 	mutex_unlock(&nfsd_mutex);
1642 
1643 	return ret;
1644 }
1645 
1646 /**
1647  * nfsd_nl_threads_set_doit - set the number of running threads
1648  * @skb: reply buffer
1649  * @info: netlink metadata and command arguments
1650  *
1651  * Return 0 on success or a negative errno.
1652  */
1653 int nfsd_nl_threads_set_doit(struct sk_buff *skb, struct genl_info *info)
1654 {
1655 	int *nthreads, count = 0, nrpools, i, ret = -EOPNOTSUPP, rem;
1656 	struct net *net = genl_info_net(info);
1657 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1658 	const struct nlattr *attr;
1659 	const char *scope = NULL;
1660 
1661 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_SERVER_THREADS))
1662 		return -EINVAL;
1663 
1664 	/* count number of SERVER_THREADS values */
1665 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1666 		if (nla_type(attr) == NFSD_A_SERVER_THREADS)
1667 			count++;
1668 	}
1669 
1670 	mutex_lock(&nfsd_mutex);
1671 
1672 	nrpools = max(count, nfsd_nrpools(net));
1673 	nthreads = kcalloc(nrpools, sizeof(int), GFP_KERNEL);
1674 	if (!nthreads) {
1675 		ret = -ENOMEM;
1676 		goto out_unlock;
1677 	}
1678 
1679 	i = 0;
1680 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1681 		if (nla_type(attr) == NFSD_A_SERVER_THREADS) {
1682 			nthreads[i++] = nla_get_u32(attr);
1683 			if (i >= nrpools)
1684 				break;
1685 		}
1686 	}
1687 
1688 	if (info->attrs[NFSD_A_SERVER_GRACETIME] ||
1689 	    info->attrs[NFSD_A_SERVER_LEASETIME] ||
1690 	    info->attrs[NFSD_A_SERVER_SCOPE]) {
1691 		ret = -EBUSY;
1692 		if (nn->nfsd_serv && nn->nfsd_serv->sv_nrthreads)
1693 			goto out_unlock;
1694 
1695 		ret = -EINVAL;
1696 		attr = info->attrs[NFSD_A_SERVER_GRACETIME];
1697 		if (attr) {
1698 			u32 gracetime = nla_get_u32(attr);
1699 
1700 			if (gracetime < 10 || gracetime > 3600)
1701 				goto out_unlock;
1702 
1703 			nn->nfsd4_grace = gracetime;
1704 		}
1705 
1706 		attr = info->attrs[NFSD_A_SERVER_LEASETIME];
1707 		if (attr) {
1708 			u32 leasetime = nla_get_u32(attr);
1709 
1710 			if (leasetime < 10 || leasetime > 3600)
1711 				goto out_unlock;
1712 
1713 			nn->nfsd4_lease = leasetime;
1714 		}
1715 
1716 		attr = info->attrs[NFSD_A_SERVER_SCOPE];
1717 		if (attr)
1718 			scope = nla_data(attr);
1719 	}
1720 
1721 	ret = nfsd_svc(nrpools, nthreads, net, get_current_cred(), scope);
1722 	if (ret > 0)
1723 		ret = 0;
1724 out_unlock:
1725 	mutex_unlock(&nfsd_mutex);
1726 	kfree(nthreads);
1727 	return ret;
1728 }
1729 
1730 /**
1731  * nfsd_nl_threads_get_doit - get the number of running threads
1732  * @skb: reply buffer
1733  * @info: netlink metadata and command arguments
1734  *
1735  * Return 0 on success or a negative errno.
1736  */
1737 int nfsd_nl_threads_get_doit(struct sk_buff *skb, struct genl_info *info)
1738 {
1739 	struct net *net = genl_info_net(info);
1740 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1741 	void *hdr;
1742 	int err;
1743 
1744 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
1745 	if (!skb)
1746 		return -ENOMEM;
1747 
1748 	hdr = genlmsg_iput(skb, info);
1749 	if (!hdr) {
1750 		err = -EMSGSIZE;
1751 		goto err_free_msg;
1752 	}
1753 
1754 	mutex_lock(&nfsd_mutex);
1755 
1756 	err = nla_put_u32(skb, NFSD_A_SERVER_GRACETIME,
1757 			  nn->nfsd4_grace) ||
1758 	      nla_put_u32(skb, NFSD_A_SERVER_LEASETIME,
1759 			  nn->nfsd4_lease) ||
1760 	      nla_put_string(skb, NFSD_A_SERVER_SCOPE,
1761 			  nn->nfsd_name);
1762 	if (err)
1763 		goto err_unlock;
1764 
1765 	if (nn->nfsd_serv) {
1766 		int i;
1767 
1768 		for (i = 0; i < nfsd_nrpools(net); ++i) {
1769 			struct svc_pool *sp = &nn->nfsd_serv->sv_pools[i];
1770 
1771 			err = nla_put_u32(skb, NFSD_A_SERVER_THREADS,
1772 					  sp->sp_nrthreads);
1773 			if (err)
1774 				goto err_unlock;
1775 		}
1776 	} else {
1777 		err = nla_put_u32(skb, NFSD_A_SERVER_THREADS, 0);
1778 		if (err)
1779 			goto err_unlock;
1780 	}
1781 
1782 	mutex_unlock(&nfsd_mutex);
1783 
1784 	genlmsg_end(skb, hdr);
1785 
1786 	return genlmsg_reply(skb, info);
1787 
1788 err_unlock:
1789 	mutex_unlock(&nfsd_mutex);
1790 err_free_msg:
1791 	nlmsg_free(skb);
1792 
1793 	return err;
1794 }
1795 
1796 /**
1797  * nfsd_nl_version_set_doit - set the nfs enabled versions
1798  * @skb: reply buffer
1799  * @info: netlink metadata and command arguments
1800  *
1801  * Return 0 on success or a negative errno.
1802  */
1803 int nfsd_nl_version_set_doit(struct sk_buff *skb, struct genl_info *info)
1804 {
1805 	const struct nlattr *attr;
1806 	struct nfsd_net *nn;
1807 	int i, rem;
1808 
1809 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_SERVER_PROTO_VERSION))
1810 		return -EINVAL;
1811 
1812 	mutex_lock(&nfsd_mutex);
1813 
1814 	nn = net_generic(genl_info_net(info), nfsd_net_id);
1815 	if (nn->nfsd_serv) {
1816 		mutex_unlock(&nfsd_mutex);
1817 		return -EBUSY;
1818 	}
1819 
1820 	/* clear current supported versions. */
1821 	nfsd_vers(nn, 2, NFSD_CLEAR);
1822 	nfsd_vers(nn, 3, NFSD_CLEAR);
1823 	for (i = 0; i <= NFSD_SUPPORTED_MINOR_VERSION; i++)
1824 		nfsd_minorversion(nn, i, NFSD_CLEAR);
1825 
1826 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1827 		struct nlattr *tb[NFSD_A_VERSION_MAX + 1];
1828 		u32 major, minor = 0;
1829 		bool enabled;
1830 
1831 		if (nla_type(attr) != NFSD_A_SERVER_PROTO_VERSION)
1832 			continue;
1833 
1834 		if (nla_parse_nested(tb, NFSD_A_VERSION_MAX, attr,
1835 				     nfsd_version_nl_policy, info->extack) < 0)
1836 			continue;
1837 
1838 		if (!tb[NFSD_A_VERSION_MAJOR])
1839 			continue;
1840 
1841 		major = nla_get_u32(tb[NFSD_A_VERSION_MAJOR]);
1842 		if (tb[NFSD_A_VERSION_MINOR])
1843 			minor = nla_get_u32(tb[NFSD_A_VERSION_MINOR]);
1844 
1845 		enabled = nla_get_flag(tb[NFSD_A_VERSION_ENABLED]);
1846 
1847 		switch (major) {
1848 		case 4:
1849 			nfsd_minorversion(nn, minor, enabled ? NFSD_SET : NFSD_CLEAR);
1850 			break;
1851 		case 3:
1852 		case 2:
1853 			if (!minor)
1854 				nfsd_vers(nn, major, enabled ? NFSD_SET : NFSD_CLEAR);
1855 			break;
1856 		default:
1857 			break;
1858 		}
1859 	}
1860 
1861 	mutex_unlock(&nfsd_mutex);
1862 
1863 	return 0;
1864 }
1865 
1866 /**
1867  * nfsd_nl_version_get_doit - get the enabled status for all supported nfs versions
1868  * @skb: reply buffer
1869  * @info: netlink metadata and command arguments
1870  *
1871  * Return 0 on success or a negative errno.
1872  */
1873 int nfsd_nl_version_get_doit(struct sk_buff *skb, struct genl_info *info)
1874 {
1875 	struct nfsd_net *nn;
1876 	int i, err;
1877 	void *hdr;
1878 
1879 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
1880 	if (!skb)
1881 		return -ENOMEM;
1882 
1883 	hdr = genlmsg_iput(skb, info);
1884 	if (!hdr) {
1885 		err = -EMSGSIZE;
1886 		goto err_free_msg;
1887 	}
1888 
1889 	mutex_lock(&nfsd_mutex);
1890 	nn = net_generic(genl_info_net(info), nfsd_net_id);
1891 
1892 	for (i = 2; i <= 4; i++) {
1893 		int j;
1894 
1895 		for (j = 0; j <= NFSD_SUPPORTED_MINOR_VERSION; j++) {
1896 			struct nlattr *attr;
1897 
1898 			/* Don't record any versions the kernel doesn't have
1899 			 * compiled in
1900 			 */
1901 			if (!nfsd_support_version(i))
1902 				continue;
1903 
1904 			/* NFSv{2,3} does not support minor numbers */
1905 			if (i < 4 && j)
1906 				continue;
1907 
1908 			attr = nla_nest_start(skb,
1909 					      NFSD_A_SERVER_PROTO_VERSION);
1910 			if (!attr) {
1911 				err = -EINVAL;
1912 				goto err_nfsd_unlock;
1913 			}
1914 
1915 			if (nla_put_u32(skb, NFSD_A_VERSION_MAJOR, i) ||
1916 			    nla_put_u32(skb, NFSD_A_VERSION_MINOR, j)) {
1917 				err = -EINVAL;
1918 				goto err_nfsd_unlock;
1919 			}
1920 
1921 			/* Set the enabled flag if the version is enabled */
1922 			if (nfsd_vers(nn, i, NFSD_TEST) &&
1923 			    (i < 4 || nfsd_minorversion(nn, j, NFSD_TEST)) &&
1924 			    nla_put_flag(skb, NFSD_A_VERSION_ENABLED)) {
1925 				err = -EINVAL;
1926 				goto err_nfsd_unlock;
1927 			}
1928 
1929 			nla_nest_end(skb, attr);
1930 		}
1931 	}
1932 
1933 	mutex_unlock(&nfsd_mutex);
1934 	genlmsg_end(skb, hdr);
1935 
1936 	return genlmsg_reply(skb, info);
1937 
1938 err_nfsd_unlock:
1939 	mutex_unlock(&nfsd_mutex);
1940 err_free_msg:
1941 	nlmsg_free(skb);
1942 
1943 	return err;
1944 }
1945 
1946 /**
1947  * nfsd_nl_listener_set_doit - set the nfs running sockets
1948  * @skb: reply buffer
1949  * @info: netlink metadata and command arguments
1950  *
1951  * Return 0 on success or a negative errno.
1952  */
1953 int nfsd_nl_listener_set_doit(struct sk_buff *skb, struct genl_info *info)
1954 {
1955 	struct net *net = genl_info_net(info);
1956 	struct svc_xprt *xprt, *tmp;
1957 	const struct nlattr *attr;
1958 	struct svc_serv *serv;
1959 	LIST_HEAD(permsocks);
1960 	struct nfsd_net *nn;
1961 	int err, rem;
1962 
1963 	mutex_lock(&nfsd_mutex);
1964 
1965 	err = nfsd_create_serv(net);
1966 	if (err) {
1967 		mutex_unlock(&nfsd_mutex);
1968 		return err;
1969 	}
1970 
1971 	nn = net_generic(net, nfsd_net_id);
1972 	serv = nn->nfsd_serv;
1973 
1974 	spin_lock_bh(&serv->sv_lock);
1975 
1976 	/* Move all of the old listener sockets to a temp list */
1977 	list_splice_init(&serv->sv_permsocks, &permsocks);
1978 
1979 	/*
1980 	 * Walk the list of server_socks from userland and move any that match
1981 	 * back to sv_permsocks
1982 	 */
1983 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
1984 		struct nlattr *tb[NFSD_A_SOCK_MAX + 1];
1985 		const char *xcl_name;
1986 		struct sockaddr *sa;
1987 
1988 		if (nla_type(attr) != NFSD_A_SERVER_SOCK_ADDR)
1989 			continue;
1990 
1991 		if (nla_parse_nested(tb, NFSD_A_SOCK_MAX, attr,
1992 				     nfsd_sock_nl_policy, info->extack) < 0)
1993 			continue;
1994 
1995 		if (!tb[NFSD_A_SOCK_ADDR] || !tb[NFSD_A_SOCK_TRANSPORT_NAME])
1996 			continue;
1997 
1998 		if (nla_len(tb[NFSD_A_SOCK_ADDR]) < sizeof(*sa))
1999 			continue;
2000 
2001 		xcl_name = nla_data(tb[NFSD_A_SOCK_TRANSPORT_NAME]);
2002 		sa = nla_data(tb[NFSD_A_SOCK_ADDR]);
2003 
2004 		/* Put back any matching sockets */
2005 		list_for_each_entry_safe(xprt, tmp, &permsocks, xpt_list) {
2006 			/* This shouldn't be possible */
2007 			if (WARN_ON_ONCE(xprt->xpt_net != net)) {
2008 				list_move(&xprt->xpt_list, &serv->sv_permsocks);
2009 				continue;
2010 			}
2011 
2012 			/* If everything matches, put it back */
2013 			if (!strcmp(xprt->xpt_class->xcl_name, xcl_name) &&
2014 			    rpc_cmp_addr_port(sa, (struct sockaddr *)&xprt->xpt_local)) {
2015 				list_move(&xprt->xpt_list, &serv->sv_permsocks);
2016 				break;
2017 			}
2018 		}
2019 	}
2020 
2021 	/* For now, no removing old sockets while server is running */
2022 	if (serv->sv_nrthreads && !list_empty(&permsocks)) {
2023 		list_splice_init(&permsocks, &serv->sv_permsocks);
2024 		spin_unlock_bh(&serv->sv_lock);
2025 		err = -EBUSY;
2026 		goto out_unlock_mtx;
2027 	}
2028 
2029 	/* Close the remaining sockets on the permsocks list */
2030 	while (!list_empty(&permsocks)) {
2031 		xprt = list_first_entry(&permsocks, struct svc_xprt, xpt_list);
2032 		list_move(&xprt->xpt_list, &serv->sv_permsocks);
2033 
2034 		/*
2035 		 * Newly-created sockets are born with the BUSY bit set. Clear
2036 		 * it if there are no threads, since nothing can pick it up
2037 		 * in that case.
2038 		 */
2039 		if (!serv->sv_nrthreads)
2040 			clear_bit(XPT_BUSY, &xprt->xpt_flags);
2041 
2042 		set_bit(XPT_CLOSE, &xprt->xpt_flags);
2043 		spin_unlock_bh(&serv->sv_lock);
2044 		svc_xprt_close(xprt);
2045 		spin_lock_bh(&serv->sv_lock);
2046 	}
2047 
2048 	spin_unlock_bh(&serv->sv_lock);
2049 
2050 	/* walk list of addrs again, open any that still don't exist */
2051 	nlmsg_for_each_attr(attr, info->nlhdr, GENL_HDRLEN, rem) {
2052 		struct nlattr *tb[NFSD_A_SOCK_MAX + 1];
2053 		const char *xcl_name;
2054 		struct sockaddr *sa;
2055 		int ret;
2056 
2057 		if (nla_type(attr) != NFSD_A_SERVER_SOCK_ADDR)
2058 			continue;
2059 
2060 		if (nla_parse_nested(tb, NFSD_A_SOCK_MAX, attr,
2061 				     nfsd_sock_nl_policy, info->extack) < 0)
2062 			continue;
2063 
2064 		if (!tb[NFSD_A_SOCK_ADDR] || !tb[NFSD_A_SOCK_TRANSPORT_NAME])
2065 			continue;
2066 
2067 		if (nla_len(tb[NFSD_A_SOCK_ADDR]) < sizeof(*sa))
2068 			continue;
2069 
2070 		xcl_name = nla_data(tb[NFSD_A_SOCK_TRANSPORT_NAME]);
2071 		sa = nla_data(tb[NFSD_A_SOCK_ADDR]);
2072 
2073 		xprt = svc_find_listener(serv, xcl_name, net, sa);
2074 		if (xprt) {
2075 			svc_xprt_put(xprt);
2076 			continue;
2077 		}
2078 
2079 		ret = svc_xprt_create_from_sa(serv, xcl_name, net, sa, 0,
2080 					      get_current_cred());
2081 		/* always save the latest error */
2082 		if (ret < 0)
2083 			err = ret;
2084 	}
2085 
2086 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
2087 		nfsd_destroy_serv(net);
2088 
2089 out_unlock_mtx:
2090 	mutex_unlock(&nfsd_mutex);
2091 
2092 	return err;
2093 }
2094 
2095 /**
2096  * nfsd_nl_listener_get_doit - get the nfs running listeners
2097  * @skb: reply buffer
2098  * @info: netlink metadata and command arguments
2099  *
2100  * Return 0 on success or a negative errno.
2101  */
2102 int nfsd_nl_listener_get_doit(struct sk_buff *skb, struct genl_info *info)
2103 {
2104 	struct svc_xprt *xprt;
2105 	struct svc_serv *serv;
2106 	struct nfsd_net *nn;
2107 	void *hdr;
2108 	int err;
2109 
2110 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
2111 	if (!skb)
2112 		return -ENOMEM;
2113 
2114 	hdr = genlmsg_iput(skb, info);
2115 	if (!hdr) {
2116 		err = -EMSGSIZE;
2117 		goto err_free_msg;
2118 	}
2119 
2120 	mutex_lock(&nfsd_mutex);
2121 	nn = net_generic(genl_info_net(info), nfsd_net_id);
2122 
2123 	/* no nfs server? Just send empty socket list */
2124 	if (!nn->nfsd_serv)
2125 		goto out_unlock_mtx;
2126 
2127 	serv = nn->nfsd_serv;
2128 	spin_lock_bh(&serv->sv_lock);
2129 	list_for_each_entry(xprt, &serv->sv_permsocks, xpt_list) {
2130 		struct nlattr *attr;
2131 
2132 		attr = nla_nest_start(skb, NFSD_A_SERVER_SOCK_ADDR);
2133 		if (!attr) {
2134 			err = -EINVAL;
2135 			goto err_serv_unlock;
2136 		}
2137 
2138 		if (nla_put_string(skb, NFSD_A_SOCK_TRANSPORT_NAME,
2139 				   xprt->xpt_class->xcl_name) ||
2140 		    nla_put(skb, NFSD_A_SOCK_ADDR,
2141 			    sizeof(struct sockaddr_storage),
2142 			    &xprt->xpt_local)) {
2143 			err = -EINVAL;
2144 			goto err_serv_unlock;
2145 		}
2146 
2147 		nla_nest_end(skb, attr);
2148 	}
2149 	spin_unlock_bh(&serv->sv_lock);
2150 out_unlock_mtx:
2151 	mutex_unlock(&nfsd_mutex);
2152 	genlmsg_end(skb, hdr);
2153 
2154 	return genlmsg_reply(skb, info);
2155 
2156 err_serv_unlock:
2157 	spin_unlock_bh(&serv->sv_lock);
2158 	mutex_unlock(&nfsd_mutex);
2159 err_free_msg:
2160 	nlmsg_free(skb);
2161 
2162 	return err;
2163 }
2164 
2165 /**
2166  * nfsd_nl_pool_mode_set_doit - set the number of running threads
2167  * @skb: reply buffer
2168  * @info: netlink metadata and command arguments
2169  *
2170  * Return 0 on success or a negative errno.
2171  */
2172 int nfsd_nl_pool_mode_set_doit(struct sk_buff *skb, struct genl_info *info)
2173 {
2174 	const struct nlattr *attr;
2175 
2176 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_POOL_MODE_MODE))
2177 		return -EINVAL;
2178 
2179 	attr = info->attrs[NFSD_A_POOL_MODE_MODE];
2180 	return sunrpc_set_pool_mode(nla_data(attr));
2181 }
2182 
2183 /**
2184  * nfsd_nl_pool_mode_get_doit - get info about pool_mode
2185  * @skb: reply buffer
2186  * @info: netlink metadata and command arguments
2187  *
2188  * Return 0 on success or a negative errno.
2189  */
2190 int nfsd_nl_pool_mode_get_doit(struct sk_buff *skb, struct genl_info *info)
2191 {
2192 	struct net *net = genl_info_net(info);
2193 	char buf[16];
2194 	void *hdr;
2195 	int err;
2196 
2197 	if (sunrpc_get_pool_mode(buf, ARRAY_SIZE(buf)) >= ARRAY_SIZE(buf))
2198 		return -ERANGE;
2199 
2200 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
2201 	if (!skb)
2202 		return -ENOMEM;
2203 
2204 	err = -EMSGSIZE;
2205 	hdr = genlmsg_iput(skb, info);
2206 	if (!hdr)
2207 		goto err_free_msg;
2208 
2209 	err = nla_put_string(skb, NFSD_A_POOL_MODE_MODE, buf) |
2210 	      nla_put_u32(skb, NFSD_A_POOL_MODE_NPOOLS, nfsd_nrpools(net));
2211 	if (err)
2212 		goto err_free_msg;
2213 
2214 	genlmsg_end(skb, hdr);
2215 	return genlmsg_reply(skb, info);
2216 
2217 err_free_msg:
2218 	nlmsg_free(skb);
2219 	return err;
2220 }
2221 
2222 /**
2223  * nfsd_net_init - Prepare the nfsd_net portion of a new net namespace
2224  * @net: a freshly-created network namespace
2225  *
2226  * This information stays around as long as the network namespace is
2227  * alive whether or not there is an NFSD instance running in the
2228  * namespace.
2229  *
2230  * Returns zero on success, or a negative errno otherwise.
2231  */
2232 static __net_init int nfsd_net_init(struct net *net)
2233 {
2234 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2235 	int retval;
2236 	int i;
2237 
2238 	retval = nfsd_export_init(net);
2239 	if (retval)
2240 		goto out_export_error;
2241 	retval = nfsd_idmap_init(net);
2242 	if (retval)
2243 		goto out_idmap_error;
2244 	retval = percpu_counter_init_many(nn->counter, 0, GFP_KERNEL,
2245 					  NFSD_STATS_COUNTERS_NUM);
2246 	if (retval)
2247 		goto out_repcache_error;
2248 	memset(&nn->nfsd_svcstats, 0, sizeof(nn->nfsd_svcstats));
2249 	nn->nfsd_svcstats.program = &nfsd_program;
2250 	for (i = 0; i < sizeof(nn->nfsd_versions); i++)
2251 		nn->nfsd_versions[i] = nfsd_support_version(i);
2252 	for (i = 0; i < sizeof(nn->nfsd4_minorversions); i++)
2253 		nn->nfsd4_minorversions[i] = nfsd_support_version(4);
2254 	nn->nfsd_info.mutex = &nfsd_mutex;
2255 	nn->nfsd_serv = NULL;
2256 	nfsd4_init_leases_net(nn);
2257 	get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
2258 	seqlock_init(&nn->writeverf_lock);
2259 	nfsd_proc_stat_init(net);
2260 
2261 	return 0;
2262 
2263 out_repcache_error:
2264 	nfsd_idmap_shutdown(net);
2265 out_idmap_error:
2266 	nfsd_export_shutdown(net);
2267 out_export_error:
2268 	return retval;
2269 }
2270 
2271 /**
2272  * nfsd_net_exit - Release the nfsd_net portion of a net namespace
2273  * @net: a network namespace that is about to be destroyed
2274  *
2275  */
2276 static __net_exit void nfsd_net_exit(struct net *net)
2277 {
2278 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2279 
2280 	nfsd_proc_stat_shutdown(net);
2281 	percpu_counter_destroy_many(nn->counter, NFSD_STATS_COUNTERS_NUM);
2282 	nfsd_idmap_shutdown(net);
2283 	nfsd_export_shutdown(net);
2284 }
2285 
2286 static struct pernet_operations nfsd_net_ops = {
2287 	.init = nfsd_net_init,
2288 	.exit = nfsd_net_exit,
2289 	.id   = &nfsd_net_id,
2290 	.size = sizeof(struct nfsd_net),
2291 };
2292 
2293 static int __init init_nfsd(void)
2294 {
2295 	int retval;
2296 
2297 	retval = nfsd4_init_slabs();
2298 	if (retval)
2299 		return retval;
2300 	retval = nfsd4_init_pnfs();
2301 	if (retval)
2302 		goto out_free_slabs;
2303 	retval = nfsd_drc_slab_create();
2304 	if (retval)
2305 		goto out_free_pnfs;
2306 	nfsd_lockd_init();	/* lockd->nfsd callbacks */
2307 	retval = create_proc_exports_entry();
2308 	if (retval)
2309 		goto out_free_lockd;
2310 	retval = register_pernet_subsys(&nfsd_net_ops);
2311 	if (retval < 0)
2312 		goto out_free_exports;
2313 	retval = register_cld_notifier();
2314 	if (retval)
2315 		goto out_free_subsys;
2316 	retval = nfsd4_create_laundry_wq();
2317 	if (retval)
2318 		goto out_free_cld;
2319 	retval = register_filesystem(&nfsd_fs_type);
2320 	if (retval)
2321 		goto out_free_all;
2322 	retval = genl_register_family(&nfsd_nl_family);
2323 	if (retval)
2324 		goto out_free_all;
2325 
2326 	return 0;
2327 out_free_all:
2328 	nfsd4_destroy_laundry_wq();
2329 out_free_cld:
2330 	unregister_cld_notifier();
2331 out_free_subsys:
2332 	unregister_pernet_subsys(&nfsd_net_ops);
2333 out_free_exports:
2334 	remove_proc_entry("fs/nfs/exports", NULL);
2335 	remove_proc_entry("fs/nfs", NULL);
2336 out_free_lockd:
2337 	nfsd_lockd_shutdown();
2338 	nfsd_drc_slab_free();
2339 out_free_pnfs:
2340 	nfsd4_exit_pnfs();
2341 out_free_slabs:
2342 	nfsd4_free_slabs();
2343 	return retval;
2344 }
2345 
2346 static void __exit exit_nfsd(void)
2347 {
2348 	genl_unregister_family(&nfsd_nl_family);
2349 	unregister_filesystem(&nfsd_fs_type);
2350 	nfsd4_destroy_laundry_wq();
2351 	unregister_cld_notifier();
2352 	unregister_pernet_subsys(&nfsd_net_ops);
2353 	nfsd_drc_slab_free();
2354 	remove_proc_entry("fs/nfs/exports", NULL);
2355 	remove_proc_entry("fs/nfs", NULL);
2356 	nfsd_lockd_shutdown();
2357 	nfsd4_free_slabs();
2358 	nfsd4_exit_pnfs();
2359 }
2360 
2361 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2362 MODULE_DESCRIPTION("In-kernel NFS server");
2363 MODULE_LICENSE("GPL");
2364 module_init(init_nfsd)
2365 module_exit(exit_nfsd)
2366