xref: /openbmc/linux/fs/nfsd/nfsctl.c (revision adb19164)
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/module.h>
19 #include <linux/fsnotify.h>
20 
21 #include "idmap.h"
22 #include "nfsd.h"
23 #include "cache.h"
24 #include "state.h"
25 #include "netns.h"
26 #include "pnfs.h"
27 #include "filecache.h"
28 #include "trace.h"
29 
30 /*
31  *	We have a single directory with several nodes in it.
32  */
33 enum {
34 	NFSD_Root = 1,
35 	NFSD_List,
36 	NFSD_Export_Stats,
37 	NFSD_Export_features,
38 	NFSD_Fh,
39 	NFSD_FO_UnlockIP,
40 	NFSD_FO_UnlockFS,
41 	NFSD_Threads,
42 	NFSD_Pool_Threads,
43 	NFSD_Pool_Stats,
44 	NFSD_Reply_Cache_Stats,
45 	NFSD_Versions,
46 	NFSD_Ports,
47 	NFSD_MaxBlkSize,
48 	NFSD_MaxConnections,
49 	NFSD_Filecache,
50 	/*
51 	 * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
52 	 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
53 	 */
54 #ifdef CONFIG_NFSD_V4
55 	NFSD_Leasetime,
56 	NFSD_Gracetime,
57 	NFSD_RecoveryDir,
58 	NFSD_V4EndGrace,
59 #endif
60 	NFSD_MaxReserved
61 };
62 
63 /*
64  * write() for these nodes.
65  */
66 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
67 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
68 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
69 static ssize_t write_threads(struct file *file, char *buf, size_t size);
70 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
71 static ssize_t write_versions(struct file *file, char *buf, size_t size);
72 static ssize_t write_ports(struct file *file, char *buf, size_t size);
73 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
74 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
75 #ifdef CONFIG_NFSD_V4
76 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
77 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
78 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
79 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
80 #endif
81 
82 static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
83 	[NFSD_Fh] = write_filehandle,
84 	[NFSD_FO_UnlockIP] = write_unlock_ip,
85 	[NFSD_FO_UnlockFS] = write_unlock_fs,
86 	[NFSD_Threads] = write_threads,
87 	[NFSD_Pool_Threads] = write_pool_threads,
88 	[NFSD_Versions] = write_versions,
89 	[NFSD_Ports] = write_ports,
90 	[NFSD_MaxBlkSize] = write_maxblksize,
91 	[NFSD_MaxConnections] = write_maxconn,
92 #ifdef CONFIG_NFSD_V4
93 	[NFSD_Leasetime] = write_leasetime,
94 	[NFSD_Gracetime] = write_gracetime,
95 	[NFSD_RecoveryDir] = write_recoverydir,
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	= nfsd_pool_stats_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 
284 	path_put(&path);
285 	return error;
286 }
287 
288 /*
289  * write_filehandle - Get a variable-length NFS file handle by path
290  *
291  * On input, the buffer contains a '\n'-terminated C string comprised of
292  * three alphanumeric words separated by whitespace.  The string may
293  * contain escape sequences.
294  *
295  * Input:
296  *			buf:
297  *				domain:		client domain name
298  *				path:		export pathname
299  *				maxsize:	numeric maximum size of
300  *						@buf
301  *			size:	length of C string in @buf
302  * Output:
303  *	On success:	passed-in buffer filled with '\n'-terminated C
304  *			string containing a ASCII hex text version
305  *			of the NFS file handle;
306  *			return code is the size in bytes of the string
307  *	On error:	return code is negative errno value
308  */
309 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
310 {
311 	char *dname, *path;
312 	int maxsize;
313 	char *mesg = buf;
314 	int len;
315 	struct auth_domain *dom;
316 	struct knfsd_fh fh;
317 
318 	if (size == 0)
319 		return -EINVAL;
320 
321 	if (buf[size-1] != '\n')
322 		return -EINVAL;
323 	buf[size-1] = 0;
324 
325 	dname = mesg;
326 	len = qword_get(&mesg, dname, size);
327 	if (len <= 0)
328 		return -EINVAL;
329 
330 	path = dname+len+1;
331 	len = qword_get(&mesg, path, size);
332 	if (len <= 0)
333 		return -EINVAL;
334 
335 	len = get_int(&mesg, &maxsize);
336 	if (len)
337 		return len;
338 
339 	if (maxsize < NFS_FHSIZE)
340 		return -EINVAL;
341 	maxsize = min(maxsize, NFS3_FHSIZE);
342 
343 	if (qword_get(&mesg, mesg, size) > 0)
344 		return -EINVAL;
345 
346 	trace_nfsd_ctl_filehandle(netns(file), dname, path, maxsize);
347 
348 	/* we have all the words, they are in buf.. */
349 	dom = unix_domain_find(dname);
350 	if (!dom)
351 		return -ENOMEM;
352 
353 	len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
354 	auth_domain_put(dom);
355 	if (len)
356 		return len;
357 
358 	mesg = buf;
359 	len = SIMPLE_TRANSACTION_LIMIT;
360 	qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
361 	mesg[-1] = '\n';
362 	return mesg - buf;
363 }
364 
365 /*
366  * write_threads - Start NFSD, or report the current number of running threads
367  *
368  * Input:
369  *			buf:		ignored
370  *			size:		zero
371  * Output:
372  *	On success:	passed-in buffer filled with '\n'-terminated C
373  *			string numeric value representing the number of
374  *			running NFSD threads;
375  *			return code is the size in bytes of the string
376  *	On error:	return code is zero
377  *
378  * OR
379  *
380  * Input:
381  *			buf:		C string containing an unsigned
382  *					integer value representing the
383  *					number of NFSD threads to start
384  *			size:		non-zero length of C string in @buf
385  * Output:
386  *	On success:	NFS service is started;
387  *			passed-in buffer filled with '\n'-terminated C
388  *			string numeric value representing the number of
389  *			running NFSD threads;
390  *			return code is the size in bytes of the string
391  *	On error:	return code is zero or a negative errno value
392  */
393 static ssize_t write_threads(struct file *file, char *buf, size_t size)
394 {
395 	char *mesg = buf;
396 	int rv;
397 	struct net *net = netns(file);
398 
399 	if (size > 0) {
400 		int newthreads;
401 		rv = get_int(&mesg, &newthreads);
402 		if (rv)
403 			return rv;
404 		if (newthreads < 0)
405 			return -EINVAL;
406 		trace_nfsd_ctl_threads(net, newthreads);
407 		rv = nfsd_svc(newthreads, net, file->f_cred);
408 		if (rv < 0)
409 			return rv;
410 	} else
411 		rv = nfsd_nrthreads(net);
412 
413 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
414 }
415 
416 /*
417  * write_pool_threads - Set or report the current number of threads per pool
418  *
419  * Input:
420  *			buf:		ignored
421  *			size:		zero
422  *
423  * OR
424  *
425  * Input:
426  *			buf:		C string containing whitespace-
427  *					separated unsigned integer values
428  *					representing the number of NFSD
429  *					threads to start in each pool
430  *			size:		non-zero length of C string in @buf
431  * Output:
432  *	On success:	passed-in buffer filled with '\n'-terminated C
433  *			string containing integer values representing the
434  *			number of NFSD threads in each pool;
435  *			return code is the size in bytes of the string
436  *	On error:	return code is zero or a negative errno value
437  */
438 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
439 {
440 	/* if size > 0, look for an array of number of threads per node
441 	 * and apply them  then write out number of threads per node as reply
442 	 */
443 	char *mesg = buf;
444 	int i;
445 	int rv;
446 	int len;
447 	int npools;
448 	int *nthreads;
449 	struct net *net = netns(file);
450 
451 	mutex_lock(&nfsd_mutex);
452 	npools = nfsd_nrpools(net);
453 	if (npools == 0) {
454 		/*
455 		 * NFS is shut down.  The admin can start it by
456 		 * writing to the threads file but NOT the pool_threads
457 		 * file, sorry.  Report zero threads.
458 		 */
459 		mutex_unlock(&nfsd_mutex);
460 		strcpy(buf, "0\n");
461 		return strlen(buf);
462 	}
463 
464 	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
465 	rv = -ENOMEM;
466 	if (nthreads == NULL)
467 		goto out_free;
468 
469 	if (size > 0) {
470 		for (i = 0; i < npools; i++) {
471 			rv = get_int(&mesg, &nthreads[i]);
472 			if (rv == -ENOENT)
473 				break;		/* fewer numbers than pools */
474 			if (rv)
475 				goto out_free;	/* syntax error */
476 			rv = -EINVAL;
477 			if (nthreads[i] < 0)
478 				goto out_free;
479 			trace_nfsd_ctl_pool_threads(net, i, nthreads[i]);
480 		}
481 		rv = nfsd_set_nrthreads(i, nthreads, net);
482 		if (rv)
483 			goto out_free;
484 	}
485 
486 	rv = nfsd_get_nrthreads(npools, nthreads, net);
487 	if (rv)
488 		goto out_free;
489 
490 	mesg = buf;
491 	size = SIMPLE_TRANSACTION_LIMIT;
492 	for (i = 0; i < npools && size > 0; i++) {
493 		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
494 		len = strlen(mesg);
495 		size -= len;
496 		mesg += len;
497 	}
498 	rv = mesg - buf;
499 out_free:
500 	kfree(nthreads);
501 	mutex_unlock(&nfsd_mutex);
502 	return rv;
503 }
504 
505 static ssize_t
506 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
507 		const char *sep, unsigned vers, int minor)
508 {
509 	const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
510 	bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
511 
512 	if (vers == 4 && minor >= 0 &&
513 	    !nfsd_minorversion(nn, minor, NFSD_TEST))
514 		supported = false;
515 	if (minor == 0 && supported)
516 		/*
517 		 * special case for backward compatability.
518 		 * +4.0 is never reported, it is implied by
519 		 * +4, unless -4.0 is present.
520 		 */
521 		return 0;
522 	return snprintf(buf, remaining, format, sep,
523 			supported ? '+' : '-', vers, minor);
524 }
525 
526 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
527 {
528 	char *mesg = buf;
529 	char *vers, *minorp, sign;
530 	int len, num, remaining;
531 	ssize_t tlen = 0;
532 	char *sep;
533 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
534 
535 	if (size > 0) {
536 		if (nn->nfsd_serv)
537 			/* Cannot change versions without updating
538 			 * nn->nfsd_serv->sv_xdrsize, and reallocing
539 			 * rq_argp and rq_resp
540 			 */
541 			return -EBUSY;
542 		if (buf[size-1] != '\n')
543 			return -EINVAL;
544 		buf[size-1] = 0;
545 		trace_nfsd_ctl_version(netns(file), buf);
546 
547 		vers = mesg;
548 		len = qword_get(&mesg, vers, size);
549 		if (len <= 0) return -EINVAL;
550 		do {
551 			enum vers_op cmd;
552 			unsigned minor;
553 			sign = *vers;
554 			if (sign == '+' || sign == '-')
555 				num = simple_strtol((vers+1), &minorp, 0);
556 			else
557 				num = simple_strtol(vers, &minorp, 0);
558 			if (*minorp == '.') {
559 				if (num != 4)
560 					return -EINVAL;
561 				if (kstrtouint(minorp+1, 0, &minor) < 0)
562 					return -EINVAL;
563 			}
564 
565 			cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
566 			switch(num) {
567 #ifdef CONFIG_NFSD_V2
568 			case 2:
569 #endif
570 			case 3:
571 				nfsd_vers(nn, num, cmd);
572 				break;
573 			case 4:
574 				if (*minorp == '.') {
575 					if (nfsd_minorversion(nn, minor, cmd) < 0)
576 						return -EINVAL;
577 				} else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
578 					/*
579 					 * Either we have +4 and no minors are enabled,
580 					 * or we have -4 and at least one minor is enabled.
581 					 * In either case, propagate 'cmd' to all minors.
582 					 */
583 					minor = 0;
584 					while (nfsd_minorversion(nn, minor, cmd) >= 0)
585 						minor++;
586 				}
587 				break;
588 			default:
589 				/* Ignore requests to disable non-existent versions */
590 				if (cmd == NFSD_SET)
591 					return -EINVAL;
592 			}
593 			vers += len + 1;
594 		} while ((len = qword_get(&mesg, vers, size)) > 0);
595 		/* If all get turned off, turn them back on, as
596 		 * having no versions is BAD
597 		 */
598 		nfsd_reset_versions(nn);
599 	}
600 
601 	/* Now write current state into reply buffer */
602 	sep = "";
603 	remaining = SIMPLE_TRANSACTION_LIMIT;
604 	for (num=2 ; num <= 4 ; num++) {
605 		int minor;
606 		if (!nfsd_vers(nn, num, NFSD_AVAIL))
607 			continue;
608 
609 		minor = -1;
610 		do {
611 			len = nfsd_print_version_support(nn, buf, remaining,
612 					sep, num, minor);
613 			if (len >= remaining)
614 				goto out;
615 			remaining -= len;
616 			buf += len;
617 			tlen += len;
618 			minor++;
619 			if (len)
620 				sep = " ";
621 		} while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
622 	}
623 out:
624 	len = snprintf(buf, remaining, "\n");
625 	if (len >= remaining)
626 		return -EINVAL;
627 	return tlen + len;
628 }
629 
630 /*
631  * write_versions - Set or report the available NFS protocol versions
632  *
633  * Input:
634  *			buf:		ignored
635  *			size:		zero
636  * Output:
637  *	On success:	passed-in buffer filled with '\n'-terminated C
638  *			string containing positive or negative integer
639  *			values representing the current status of each
640  *			protocol version;
641  *			return code is the size in bytes of the string
642  *	On error:	return code is zero or a negative errno value
643  *
644  * OR
645  *
646  * Input:
647  *			buf:		C string containing whitespace-
648  *					separated positive or negative
649  *					integer values representing NFS
650  *					protocol versions to enable ("+n")
651  *					or disable ("-n")
652  *			size:		non-zero length of C string in @buf
653  * Output:
654  *	On success:	status of zero or more protocol versions has
655  *			been updated; passed-in buffer filled with
656  *			'\n'-terminated C string containing positive
657  *			or negative integer values representing the
658  *			current status of each 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 static ssize_t write_versions(struct file *file, char *buf, size_t size)
663 {
664 	ssize_t rv;
665 
666 	mutex_lock(&nfsd_mutex);
667 	rv = __write_versions(file, buf, size);
668 	mutex_unlock(&nfsd_mutex);
669 	return rv;
670 }
671 
672 /*
673  * Zero-length write.  Return a list of NFSD's current listener
674  * transports.
675  */
676 static ssize_t __write_ports_names(char *buf, struct net *net)
677 {
678 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
679 
680 	if (nn->nfsd_serv == NULL)
681 		return 0;
682 	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
683 }
684 
685 /*
686  * A single 'fd' number was written, in which case it must be for
687  * a socket of a supported family/protocol, and we use it as an
688  * nfsd listener.
689  */
690 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
691 {
692 	char *mesg = buf;
693 	int fd, err;
694 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
695 	struct svc_serv *serv;
696 
697 	err = get_int(&mesg, &fd);
698 	if (err != 0 || fd < 0)
699 		return -EINVAL;
700 	trace_nfsd_ctl_ports_addfd(net, fd);
701 
702 	err = nfsd_create_serv(net);
703 	if (err != 0)
704 		return err;
705 
706 	serv = nn->nfsd_serv;
707 	err = svc_addsock(serv, net, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
708 
709 	if (err < 0 && !serv->sv_nrthreads && !nn->keep_active)
710 		nfsd_last_thread(net);
711 	else if (err >= 0 && !serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
712 		svc_get(serv);
713 
714 	svc_put(serv);
715 	return err;
716 }
717 
718 /*
719  * A transport listener is added by writing its transport name and
720  * a port number.
721  */
722 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
723 {
724 	char transport[16];
725 	struct svc_xprt *xprt;
726 	int port, err;
727 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
728 	struct svc_serv *serv;
729 
730 	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
731 		return -EINVAL;
732 
733 	if (port < 1 || port > USHRT_MAX)
734 		return -EINVAL;
735 	trace_nfsd_ctl_ports_addxprt(net, transport, port);
736 
737 	err = nfsd_create_serv(net);
738 	if (err != 0)
739 		return err;
740 
741 	serv = nn->nfsd_serv;
742 	err = svc_xprt_create(serv, transport, net,
743 			      PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
744 	if (err < 0)
745 		goto out_err;
746 
747 	err = svc_xprt_create(serv, transport, net,
748 			      PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
749 	if (err < 0 && err != -EAFNOSUPPORT)
750 		goto out_close;
751 
752 	if (!serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
753 		svc_get(serv);
754 
755 	svc_put(serv);
756 	return 0;
757 out_close:
758 	xprt = svc_find_xprt(serv, transport, net, PF_INET, port);
759 	if (xprt != NULL) {
760 		svc_xprt_close(xprt);
761 		svc_xprt_put(xprt);
762 	}
763 out_err:
764 	if (!serv->sv_nrthreads && !nn->keep_active)
765 		nfsd_last_thread(net);
766 
767 	svc_put(serv);
768 	return err;
769 }
770 
771 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
772 			     struct net *net)
773 {
774 	if (size == 0)
775 		return __write_ports_names(buf, net);
776 
777 	if (isdigit(buf[0]))
778 		return __write_ports_addfd(buf, net, file->f_cred);
779 
780 	if (isalpha(buf[0]))
781 		return __write_ports_addxprt(buf, net, file->f_cred);
782 
783 	return -EINVAL;
784 }
785 
786 /*
787  * write_ports - Pass a socket file descriptor or transport name to listen on
788  *
789  * Input:
790  *			buf:		ignored
791  *			size:		zero
792  * Output:
793  *	On success:	passed-in buffer filled with a '\n'-terminated C
794  *			string containing a whitespace-separated list of
795  *			named NFSD listeners;
796  *			return code is the size in bytes of the string
797  *	On error:	return code is zero or a negative errno value
798  *
799  * OR
800  *
801  * Input:
802  *			buf:		C string containing an unsigned
803  *					integer value representing a bound
804  *					but unconnected socket that is to be
805  *					used as an NFSD listener; listen(3)
806  *					must be called for a SOCK_STREAM
807  *					socket, otherwise it is ignored
808  *			size:		non-zero length of C string in @buf
809  * Output:
810  *	On success:	NFS service is started;
811  *			passed-in buffer filled with a '\n'-terminated C
812  *			string containing a unique alphanumeric name of
813  *			the listener;
814  *			return code is the size in bytes of the string
815  *	On error:	return code is a negative errno value
816  *
817  * OR
818  *
819  * Input:
820  *			buf:		C string containing a transport
821  *					name and an unsigned integer value
822  *					representing the port to listen on,
823  *					separated by whitespace
824  *			size:		non-zero length of C string in @buf
825  * Output:
826  *	On success:	returns zero; NFS service is started
827  *	On error:	return code is a negative errno value
828  */
829 static ssize_t write_ports(struct file *file, char *buf, size_t size)
830 {
831 	ssize_t rv;
832 
833 	mutex_lock(&nfsd_mutex);
834 	rv = __write_ports(file, buf, size, netns(file));
835 	mutex_unlock(&nfsd_mutex);
836 	return rv;
837 }
838 
839 
840 int nfsd_max_blksize;
841 
842 /*
843  * write_maxblksize - Set or report the current NFS blksize
844  *
845  * Input:
846  *			buf:		ignored
847  *			size:		zero
848  *
849  * OR
850  *
851  * Input:
852  *			buf:		C string containing an unsigned
853  *					integer value representing the new
854  *					NFS blksize
855  *			size:		non-zero length of C string in @buf
856  * Output:
857  *	On success:	passed-in buffer filled with '\n'-terminated C string
858  *			containing numeric value of the current NFS blksize
859  *			setting;
860  *			return code is the size in bytes of the string
861  *	On error:	return code is zero or a negative errno value
862  */
863 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
864 {
865 	char *mesg = buf;
866 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
867 
868 	if (size > 0) {
869 		int bsize;
870 		int rv = get_int(&mesg, &bsize);
871 		if (rv)
872 			return rv;
873 		trace_nfsd_ctl_maxblksize(netns(file), bsize);
874 
875 		/* force bsize into allowed range and
876 		 * required alignment.
877 		 */
878 		bsize = max_t(int, bsize, 1024);
879 		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
880 		bsize &= ~(1024-1);
881 		mutex_lock(&nfsd_mutex);
882 		if (nn->nfsd_serv) {
883 			mutex_unlock(&nfsd_mutex);
884 			return -EBUSY;
885 		}
886 		nfsd_max_blksize = bsize;
887 		mutex_unlock(&nfsd_mutex);
888 	}
889 
890 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
891 							nfsd_max_blksize);
892 }
893 
894 /*
895  * write_maxconn - Set or report the current max number of connections
896  *
897  * Input:
898  *			buf:		ignored
899  *			size:		zero
900  * OR
901  *
902  * Input:
903  *			buf:		C string containing an unsigned
904  *					integer value representing the new
905  *					number of max connections
906  *			size:		non-zero length of C string in @buf
907  * Output:
908  *	On success:	passed-in buffer filled with '\n'-terminated C string
909  *			containing numeric value of max_connections setting
910  *			for this net namespace;
911  *			return code is the size in bytes of the string
912  *	On error:	return code is zero or a negative errno value
913  */
914 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
915 {
916 	char *mesg = buf;
917 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
918 	unsigned int maxconn = nn->max_connections;
919 
920 	if (size > 0) {
921 		int rv = get_uint(&mesg, &maxconn);
922 
923 		if (rv)
924 			return rv;
925 		trace_nfsd_ctl_maxconn(netns(file), maxconn);
926 		nn->max_connections = maxconn;
927 	}
928 
929 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
930 }
931 
932 #ifdef CONFIG_NFSD_V4
933 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
934 				  time64_t *time, struct nfsd_net *nn)
935 {
936 	struct dentry *dentry = file_dentry(file);
937 	char *mesg = buf;
938 	int rv, i;
939 
940 	if (size > 0) {
941 		if (nn->nfsd_serv)
942 			return -EBUSY;
943 		rv = get_int(&mesg, &i);
944 		if (rv)
945 			return rv;
946 		trace_nfsd_ctl_time(netns(file), dentry->d_name.name,
947 				    dentry->d_name.len, i);
948 
949 		/*
950 		 * Some sanity checking.  We don't have a reason for
951 		 * these particular numbers, but problems with the
952 		 * extremes are:
953 		 *	- Too short: the briefest network outage may
954 		 *	  cause clients to lose all their locks.  Also,
955 		 *	  the frequent polling may be wasteful.
956 		 *	- Too long: do you really want reboot recovery
957 		 *	  to take more than an hour?  Or to make other
958 		 *	  clients wait an hour before being able to
959 		 *	  revoke a dead client's locks?
960 		 */
961 		if (i < 10 || i > 3600)
962 			return -EINVAL;
963 		*time = i;
964 	}
965 
966 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
967 }
968 
969 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
970 				time64_t *time, struct nfsd_net *nn)
971 {
972 	ssize_t rv;
973 
974 	mutex_lock(&nfsd_mutex);
975 	rv = __nfsd4_write_time(file, buf, size, time, nn);
976 	mutex_unlock(&nfsd_mutex);
977 	return rv;
978 }
979 
980 /*
981  * write_leasetime - Set or report the current NFSv4 lease time
982  *
983  * Input:
984  *			buf:		ignored
985  *			size:		zero
986  *
987  * OR
988  *
989  * Input:
990  *			buf:		C string containing an unsigned
991  *					integer value representing the new
992  *					NFSv4 lease expiry time
993  *			size:		non-zero length of C string in @buf
994  * Output:
995  *	On success:	passed-in buffer filled with '\n'-terminated C
996  *			string containing unsigned integer value of the
997  *			current lease expiry time;
998  *			return code is the size in bytes of the string
999  *	On error:	return code is zero or a negative errno value
1000  */
1001 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1002 {
1003 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1004 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1005 }
1006 
1007 /*
1008  * write_gracetime - Set or report current NFSv4 grace period time
1009  *
1010  * As above, but sets the time of the NFSv4 grace period.
1011  *
1012  * Note this should never be set to less than the *previous*
1013  * lease-period time, but we don't try to enforce this.  (In the common
1014  * case (a new boot), we don't know what the previous lease time was
1015  * anyway.)
1016  */
1017 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1018 {
1019 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1020 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1021 }
1022 
1023 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1024 				   struct nfsd_net *nn)
1025 {
1026 	char *mesg = buf;
1027 	char *recdir;
1028 	int len, status;
1029 
1030 	if (size > 0) {
1031 		if (nn->nfsd_serv)
1032 			return -EBUSY;
1033 		if (size > PATH_MAX || buf[size-1] != '\n')
1034 			return -EINVAL;
1035 		buf[size-1] = 0;
1036 
1037 		recdir = mesg;
1038 		len = qword_get(&mesg, recdir, size);
1039 		if (len <= 0)
1040 			return -EINVAL;
1041 		trace_nfsd_ctl_recoverydir(netns(file), recdir);
1042 
1043 		status = nfs4_reset_recoverydir(recdir);
1044 		if (status)
1045 			return status;
1046 	}
1047 
1048 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1049 							nfs4_recoverydir());
1050 }
1051 
1052 /*
1053  * write_recoverydir - Set or report the pathname of the recovery directory
1054  *
1055  * Input:
1056  *			buf:		ignored
1057  *			size:		zero
1058  *
1059  * OR
1060  *
1061  * Input:
1062  *			buf:		C string containing the pathname
1063  *					of the directory on a local file
1064  *					system containing permanent NFSv4
1065  *					recovery data
1066  *			size:		non-zero length of C string in @buf
1067  * Output:
1068  *	On success:	passed-in buffer filled with '\n'-terminated C string
1069  *			containing the current recovery pathname setting;
1070  *			return code is the size in bytes of the string
1071  *	On error:	return code is zero or a negative errno value
1072  */
1073 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1074 {
1075 	ssize_t rv;
1076 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1077 
1078 	mutex_lock(&nfsd_mutex);
1079 	rv = __write_recoverydir(file, buf, size, nn);
1080 	mutex_unlock(&nfsd_mutex);
1081 	return rv;
1082 }
1083 
1084 /*
1085  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1086  *
1087  * Input:
1088  *			buf:		ignored
1089  *			size:		zero
1090  * OR
1091  *
1092  * Input:
1093  *			buf:		any value
1094  *			size:		non-zero length of C string in @buf
1095  * Output:
1096  *			passed-in buffer filled with "Y" or "N" with a newline
1097  *			and NULL-terminated C string. This indicates whether
1098  *			the grace period has ended in the current net
1099  *			namespace. Return code is the size in bytes of the
1100  *			string. Writing a string that starts with 'Y', 'y', or
1101  *			'1' to the file will end the grace period for nfsd's v4
1102  *			lock manager.
1103  */
1104 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1105 {
1106 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1107 
1108 	if (size > 0) {
1109 		switch(buf[0]) {
1110 		case 'Y':
1111 		case 'y':
1112 		case '1':
1113 			if (!nn->nfsd_serv)
1114 				return -EBUSY;
1115 			trace_nfsd_end_grace(netns(file));
1116 			nfsd4_end_grace(nn);
1117 			break;
1118 		default:
1119 			return -EINVAL;
1120 		}
1121 	}
1122 
1123 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1124 			 nn->grace_ended ? 'Y' : 'N');
1125 }
1126 
1127 #endif
1128 
1129 /*----------------------------------------------------------------------------*/
1130 /*
1131  *	populating the filesystem.
1132  */
1133 
1134 /* Basically copying rpc_get_inode. */
1135 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1136 {
1137 	struct inode *inode = new_inode(sb);
1138 	if (!inode)
1139 		return NULL;
1140 	/* Following advice from simple_fill_super documentation: */
1141 	inode->i_ino = iunique(sb, NFSD_MaxReserved);
1142 	inode->i_mode = mode;
1143 	inode->i_atime = inode->i_mtime = inode_set_ctime_current(inode);
1144 	switch (mode & S_IFMT) {
1145 	case S_IFDIR:
1146 		inode->i_fop = &simple_dir_operations;
1147 		inode->i_op = &simple_dir_inode_operations;
1148 		inc_nlink(inode);
1149 		break;
1150 	case S_IFLNK:
1151 		inode->i_op = &simple_symlink_inode_operations;
1152 		break;
1153 	default:
1154 		break;
1155 	}
1156 	return inode;
1157 }
1158 
1159 static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1160 {
1161 	struct inode *inode;
1162 
1163 	inode = nfsd_get_inode(dir->i_sb, mode);
1164 	if (!inode)
1165 		return -ENOMEM;
1166 	if (ncl) {
1167 		inode->i_private = ncl;
1168 		kref_get(&ncl->cl_ref);
1169 	}
1170 	d_add(dentry, inode);
1171 	inc_nlink(dir);
1172 	fsnotify_mkdir(dir, dentry);
1173 	return 0;
1174 }
1175 
1176 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1177 {
1178 	struct inode *dir = parent->d_inode;
1179 	struct dentry *dentry;
1180 	int ret = -ENOMEM;
1181 
1182 	inode_lock(dir);
1183 	dentry = d_alloc_name(parent, name);
1184 	if (!dentry)
1185 		goto out_err;
1186 	ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1187 	if (ret)
1188 		goto out_err;
1189 out:
1190 	inode_unlock(dir);
1191 	return dentry;
1192 out_err:
1193 	dput(dentry);
1194 	dentry = ERR_PTR(ret);
1195 	goto out;
1196 }
1197 
1198 #if IS_ENABLED(CONFIG_SUNRPC_GSS)
1199 static int __nfsd_symlink(struct inode *dir, struct dentry *dentry,
1200 			  umode_t mode, const char *content)
1201 {
1202 	struct inode *inode;
1203 
1204 	inode = nfsd_get_inode(dir->i_sb, mode);
1205 	if (!inode)
1206 		return -ENOMEM;
1207 
1208 	inode->i_link = (char *)content;
1209 	inode->i_size = strlen(content);
1210 
1211 	d_add(dentry, inode);
1212 	inc_nlink(dir);
1213 	fsnotify_create(dir, dentry);
1214 	return 0;
1215 }
1216 
1217 /*
1218  * @content is assumed to be a NUL-terminated string that lives
1219  * longer than the symlink itself.
1220  */
1221 static void _nfsd_symlink(struct dentry *parent, const char *name,
1222 			  const char *content)
1223 {
1224 	struct inode *dir = parent->d_inode;
1225 	struct dentry *dentry;
1226 	int ret;
1227 
1228 	inode_lock(dir);
1229 	dentry = d_alloc_name(parent, name);
1230 	if (!dentry)
1231 		goto out;
1232 	ret = __nfsd_symlink(d_inode(parent), dentry, S_IFLNK | 0777, content);
1233 	if (ret)
1234 		dput(dentry);
1235 out:
1236 	inode_unlock(dir);
1237 }
1238 #else
1239 static inline void _nfsd_symlink(struct dentry *parent, const char *name,
1240 				 const char *content)
1241 {
1242 }
1243 
1244 #endif
1245 
1246 static void clear_ncl(struct inode *inode)
1247 {
1248 	struct nfsdfs_client *ncl = inode->i_private;
1249 
1250 	inode->i_private = NULL;
1251 	kref_put(&ncl->cl_ref, ncl->cl_release);
1252 }
1253 
1254 static struct nfsdfs_client *__get_nfsdfs_client(struct inode *inode)
1255 {
1256 	struct nfsdfs_client *nc = inode->i_private;
1257 
1258 	if (nc)
1259 		kref_get(&nc->cl_ref);
1260 	return nc;
1261 }
1262 
1263 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1264 {
1265 	struct nfsdfs_client *nc;
1266 
1267 	inode_lock_shared(inode);
1268 	nc = __get_nfsdfs_client(inode);
1269 	inode_unlock_shared(inode);
1270 	return nc;
1271 }
1272 /* from __rpc_unlink */
1273 static void nfsdfs_remove_file(struct inode *dir, struct dentry *dentry)
1274 {
1275 	int ret;
1276 
1277 	clear_ncl(d_inode(dentry));
1278 	dget(dentry);
1279 	ret = simple_unlink(dir, dentry);
1280 	d_drop(dentry);
1281 	fsnotify_unlink(dir, dentry);
1282 	dput(dentry);
1283 	WARN_ON_ONCE(ret);
1284 }
1285 
1286 static void nfsdfs_remove_files(struct dentry *root)
1287 {
1288 	struct dentry *dentry, *tmp;
1289 
1290 	list_for_each_entry_safe(dentry, tmp, &root->d_subdirs, d_child) {
1291 		if (!simple_positive(dentry)) {
1292 			WARN_ON_ONCE(1); /* I think this can't happen? */
1293 			continue;
1294 		}
1295 		nfsdfs_remove_file(d_inode(root), dentry);
1296 	}
1297 }
1298 
1299 /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1300  * code instead. */
1301 static  int nfsdfs_create_files(struct dentry *root,
1302 				const struct tree_descr *files,
1303 				struct dentry **fdentries)
1304 {
1305 	struct inode *dir = d_inode(root);
1306 	struct inode *inode;
1307 	struct dentry *dentry;
1308 	int i;
1309 
1310 	inode_lock(dir);
1311 	for (i = 0; files->name && files->name[0]; i++, files++) {
1312 		dentry = d_alloc_name(root, files->name);
1313 		if (!dentry)
1314 			goto out;
1315 		inode = nfsd_get_inode(d_inode(root)->i_sb,
1316 					S_IFREG | files->mode);
1317 		if (!inode) {
1318 			dput(dentry);
1319 			goto out;
1320 		}
1321 		inode->i_fop = files->ops;
1322 		inode->i_private = __get_nfsdfs_client(dir);
1323 		d_add(dentry, inode);
1324 		fsnotify_create(dir, dentry);
1325 		if (fdentries)
1326 			fdentries[i] = dentry;
1327 	}
1328 	inode_unlock(dir);
1329 	return 0;
1330 out:
1331 	nfsdfs_remove_files(root);
1332 	inode_unlock(dir);
1333 	return -ENOMEM;
1334 }
1335 
1336 /* on success, returns positive number unique to that client. */
1337 struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1338 				 struct nfsdfs_client *ncl, u32 id,
1339 				 const struct tree_descr *files,
1340 				 struct dentry **fdentries)
1341 {
1342 	struct dentry *dentry;
1343 	char name[11];
1344 	int ret;
1345 
1346 	sprintf(name, "%u", id);
1347 
1348 	dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1349 	if (IS_ERR(dentry)) /* XXX: tossing errors? */
1350 		return NULL;
1351 	ret = nfsdfs_create_files(dentry, files, fdentries);
1352 	if (ret) {
1353 		nfsd_client_rmdir(dentry);
1354 		return NULL;
1355 	}
1356 	return dentry;
1357 }
1358 
1359 /* Taken from __rpc_rmdir: */
1360 void nfsd_client_rmdir(struct dentry *dentry)
1361 {
1362 	struct inode *dir = d_inode(dentry->d_parent);
1363 	struct inode *inode = d_inode(dentry);
1364 	int ret;
1365 
1366 	inode_lock(dir);
1367 	nfsdfs_remove_files(dentry);
1368 	clear_ncl(inode);
1369 	dget(dentry);
1370 	ret = simple_rmdir(dir, dentry);
1371 	WARN_ON_ONCE(ret);
1372 	d_drop(dentry);
1373 	fsnotify_rmdir(dir, dentry);
1374 	dput(dentry);
1375 	inode_unlock(dir);
1376 }
1377 
1378 static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1379 {
1380 	struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1381 							nfsd_net_id);
1382 	struct dentry *dentry;
1383 	int ret;
1384 
1385 	static const struct tree_descr nfsd_files[] = {
1386 		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1387 		/* Per-export io stats use same ops as exports file */
1388 		[NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1389 		[NFSD_Export_features] = {"export_features",
1390 					&export_features_fops, S_IRUGO},
1391 		[NFSD_FO_UnlockIP] = {"unlock_ip",
1392 					&transaction_ops, S_IWUSR|S_IRUSR},
1393 		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1394 					&transaction_ops, S_IWUSR|S_IRUSR},
1395 		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1396 		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1397 		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1398 		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1399 		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1400 					&nfsd_reply_cache_stats_fops, S_IRUGO},
1401 		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1402 		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1403 		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1404 		[NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1405 		[NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
1406 #ifdef CONFIG_NFSD_V4
1407 		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1408 		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1409 		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1410 		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1411 #endif
1412 		/* last one */ {""}
1413 	};
1414 
1415 	ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1416 	if (ret)
1417 		return ret;
1418 	_nfsd_symlink(sb->s_root, "supported_krb5_enctypes",
1419 		      "/proc/net/rpc/gss_krb5_enctypes");
1420 	dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1421 	if (IS_ERR(dentry))
1422 		return PTR_ERR(dentry);
1423 	nn->nfsd_client_dir = dentry;
1424 	return 0;
1425 }
1426 
1427 static int nfsd_fs_get_tree(struct fs_context *fc)
1428 {
1429 	return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1430 }
1431 
1432 static void nfsd_fs_free_fc(struct fs_context *fc)
1433 {
1434 	if (fc->s_fs_info)
1435 		put_net(fc->s_fs_info);
1436 }
1437 
1438 static const struct fs_context_operations nfsd_fs_context_ops = {
1439 	.free		= nfsd_fs_free_fc,
1440 	.get_tree	= nfsd_fs_get_tree,
1441 };
1442 
1443 static int nfsd_init_fs_context(struct fs_context *fc)
1444 {
1445 	put_user_ns(fc->user_ns);
1446 	fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1447 	fc->ops = &nfsd_fs_context_ops;
1448 	return 0;
1449 }
1450 
1451 static void nfsd_umount(struct super_block *sb)
1452 {
1453 	struct net *net = sb->s_fs_info;
1454 
1455 	nfsd_shutdown_threads(net);
1456 
1457 	kill_litter_super(sb);
1458 	put_net(net);
1459 }
1460 
1461 static struct file_system_type nfsd_fs_type = {
1462 	.owner		= THIS_MODULE,
1463 	.name		= "nfsd",
1464 	.init_fs_context = nfsd_init_fs_context,
1465 	.kill_sb	= nfsd_umount,
1466 };
1467 MODULE_ALIAS_FS("nfsd");
1468 
1469 #ifdef CONFIG_PROC_FS
1470 
1471 static int exports_proc_open(struct inode *inode, struct file *file)
1472 {
1473 	return exports_net_open(current->nsproxy->net_ns, file);
1474 }
1475 
1476 static const struct proc_ops exports_proc_ops = {
1477 	.proc_open	= exports_proc_open,
1478 	.proc_read	= seq_read,
1479 	.proc_lseek	= seq_lseek,
1480 	.proc_release	= seq_release,
1481 };
1482 
1483 static int create_proc_exports_entry(void)
1484 {
1485 	struct proc_dir_entry *entry;
1486 
1487 	entry = proc_mkdir("fs/nfs", NULL);
1488 	if (!entry)
1489 		return -ENOMEM;
1490 	entry = proc_create("exports", 0, entry, &exports_proc_ops);
1491 	if (!entry) {
1492 		remove_proc_entry("fs/nfs", NULL);
1493 		return -ENOMEM;
1494 	}
1495 	return 0;
1496 }
1497 #else /* CONFIG_PROC_FS */
1498 static int create_proc_exports_entry(void)
1499 {
1500 	return 0;
1501 }
1502 #endif
1503 
1504 unsigned int nfsd_net_id;
1505 
1506 /**
1507  * nfsd_net_init - Prepare the nfsd_net portion of a new net namespace
1508  * @net: a freshly-created network namespace
1509  *
1510  * This information stays around as long as the network namespace is
1511  * alive whether or not there is an NFSD instance running in the
1512  * namespace.
1513  *
1514  * Returns zero on success, or a negative errno otherwise.
1515  */
1516 static __net_init int nfsd_net_init(struct net *net)
1517 {
1518 	int retval;
1519 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1520 
1521 	retval = nfsd_export_init(net);
1522 	if (retval)
1523 		goto out_export_error;
1524 	retval = nfsd_idmap_init(net);
1525 	if (retval)
1526 		goto out_idmap_error;
1527 	retval = nfsd_net_reply_cache_init(nn);
1528 	if (retval)
1529 		goto out_repcache_error;
1530 	nn->nfsd_versions = NULL;
1531 	nn->nfsd4_minorversions = NULL;
1532 	nfsd4_init_leases_net(nn);
1533 	get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
1534 	seqlock_init(&nn->writeverf_lock);
1535 
1536 	return 0;
1537 
1538 out_repcache_error:
1539 	nfsd_idmap_shutdown(net);
1540 out_idmap_error:
1541 	nfsd_export_shutdown(net);
1542 out_export_error:
1543 	return retval;
1544 }
1545 
1546 /**
1547  * nfsd_net_exit - Release the nfsd_net portion of a net namespace
1548  * @net: a network namespace that is about to be destroyed
1549  *
1550  */
1551 static __net_exit void nfsd_net_exit(struct net *net)
1552 {
1553 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1554 
1555 	nfsd_net_reply_cache_destroy(nn);
1556 	nfsd_idmap_shutdown(net);
1557 	nfsd_export_shutdown(net);
1558 	nfsd_netns_free_versions(nn);
1559 }
1560 
1561 static struct pernet_operations nfsd_net_ops = {
1562 	.init = nfsd_net_init,
1563 	.exit = nfsd_net_exit,
1564 	.id   = &nfsd_net_id,
1565 	.size = sizeof(struct nfsd_net),
1566 };
1567 
1568 static int __init init_nfsd(void)
1569 {
1570 	int retval;
1571 
1572 	retval = nfsd4_init_slabs();
1573 	if (retval)
1574 		return retval;
1575 	retval = nfsd4_init_pnfs();
1576 	if (retval)
1577 		goto out_free_slabs;
1578 	retval = nfsd_stat_init();	/* Statistics */
1579 	if (retval)
1580 		goto out_free_pnfs;
1581 	retval = nfsd_drc_slab_create();
1582 	if (retval)
1583 		goto out_free_stat;
1584 	nfsd_lockd_init();	/* lockd->nfsd callbacks */
1585 	retval = create_proc_exports_entry();
1586 	if (retval)
1587 		goto out_free_lockd;
1588 	retval = register_pernet_subsys(&nfsd_net_ops);
1589 	if (retval < 0)
1590 		goto out_free_exports;
1591 	retval = register_cld_notifier();
1592 	if (retval)
1593 		goto out_free_subsys;
1594 	retval = nfsd4_create_laundry_wq();
1595 	if (retval)
1596 		goto out_free_cld;
1597 	retval = register_filesystem(&nfsd_fs_type);
1598 	if (retval)
1599 		goto out_free_all;
1600 	return 0;
1601 out_free_all:
1602 	nfsd4_destroy_laundry_wq();
1603 out_free_cld:
1604 	unregister_cld_notifier();
1605 out_free_subsys:
1606 	unregister_pernet_subsys(&nfsd_net_ops);
1607 out_free_exports:
1608 	remove_proc_entry("fs/nfs/exports", NULL);
1609 	remove_proc_entry("fs/nfs", NULL);
1610 out_free_lockd:
1611 	nfsd_lockd_shutdown();
1612 	nfsd_drc_slab_free();
1613 out_free_stat:
1614 	nfsd_stat_shutdown();
1615 out_free_pnfs:
1616 	nfsd4_exit_pnfs();
1617 out_free_slabs:
1618 	nfsd4_free_slabs();
1619 	return retval;
1620 }
1621 
1622 static void __exit exit_nfsd(void)
1623 {
1624 	unregister_filesystem(&nfsd_fs_type);
1625 	nfsd4_destroy_laundry_wq();
1626 	unregister_cld_notifier();
1627 	unregister_pernet_subsys(&nfsd_net_ops);
1628 	nfsd_drc_slab_free();
1629 	remove_proc_entry("fs/nfs/exports", NULL);
1630 	remove_proc_entry("fs/nfs", NULL);
1631 	nfsd_stat_shutdown();
1632 	nfsd_lockd_shutdown();
1633 	nfsd4_free_slabs();
1634 	nfsd4_exit_pnfs();
1635 }
1636 
1637 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1638 MODULE_DESCRIPTION("In-kernel NFS server");
1639 MODULE_LICENSE("GPL");
1640 module_init(init_nfsd)
1641 module_exit(exit_nfsd)
1642