xref: /linux/fs/nfsd/nfsctl.c (revision ca64d84e93762f4e587e040a44ad9f6089afc777)
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/gss_krb5_enctypes.h>
18 #include <linux/sunrpc/rpc_pipe_fs.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 
29 /*
30  *	We have a single directory with several nodes in it.
31  */
32 enum {
33 	NFSD_Root = 1,
34 	NFSD_List,
35 	NFSD_Export_features,
36 	NFSD_Fh,
37 	NFSD_FO_UnlockIP,
38 	NFSD_FO_UnlockFS,
39 	NFSD_Threads,
40 	NFSD_Pool_Threads,
41 	NFSD_Pool_Stats,
42 	NFSD_Reply_Cache_Stats,
43 	NFSD_Versions,
44 	NFSD_Ports,
45 	NFSD_MaxBlkSize,
46 	NFSD_MaxConnections,
47 	NFSD_SupportedEnctypes,
48 	/*
49 	 * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
50 	 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
51 	 */
52 #ifdef CONFIG_NFSD_V4
53 	NFSD_Leasetime,
54 	NFSD_Gracetime,
55 	NFSD_RecoveryDir,
56 	NFSD_V4EndGrace,
57 #endif
58 	NFSD_MaxReserved
59 };
60 
61 /*
62  * write() for these nodes.
63  */
64 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
65 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
66 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
67 static ssize_t write_threads(struct file *file, char *buf, size_t size);
68 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
69 static ssize_t write_versions(struct file *file, char *buf, size_t size);
70 static ssize_t write_ports(struct file *file, char *buf, size_t size);
71 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
72 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
73 #ifdef CONFIG_NFSD_V4
74 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
75 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
76 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
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 	[NFSD_RecoveryDir] = write_recoverydir,
94 	[NFSD_V4EndGrace] = write_v4_end_grace,
95 #endif
96 };
97 
98 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
99 {
100 	ino_t ino =  file_inode(file)->i_ino;
101 	char *data;
102 	ssize_t rv;
103 
104 	if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
105 		return -EINVAL;
106 
107 	data = simple_transaction_get(file, buf, size);
108 	if (IS_ERR(data))
109 		return PTR_ERR(data);
110 
111 	rv =  write_op[ino](file, data, size);
112 	if (rv >= 0) {
113 		simple_transaction_set(file, rv);
114 		rv = size;
115 	}
116 	return rv;
117 }
118 
119 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
120 {
121 	if (! file->private_data) {
122 		/* An attempt to read a transaction file without writing
123 		 * causes a 0-byte write so that the file can return
124 		 * state information
125 		 */
126 		ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
127 		if (rv < 0)
128 			return rv;
129 	}
130 	return simple_transaction_read(file, buf, size, pos);
131 }
132 
133 static const struct file_operations transaction_ops = {
134 	.write		= nfsctl_transaction_write,
135 	.read		= nfsctl_transaction_read,
136 	.release	= simple_transaction_release,
137 	.llseek		= default_llseek,
138 };
139 
140 static int exports_net_open(struct net *net, struct file *file)
141 {
142 	int err;
143 	struct seq_file *seq;
144 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
145 
146 	err = seq_open(file, &nfs_exports_op);
147 	if (err)
148 		return err;
149 
150 	seq = file->private_data;
151 	seq->private = nn->svc_export_cache;
152 	return 0;
153 }
154 
155 static int exports_proc_open(struct inode *inode, struct file *file)
156 {
157 	return exports_net_open(current->nsproxy->net_ns, file);
158 }
159 
160 static const struct proc_ops exports_proc_ops = {
161 	.proc_open	= exports_proc_open,
162 	.proc_read	= seq_read,
163 	.proc_lseek	= seq_lseek,
164 	.proc_release	= seq_release,
165 };
166 
167 static int exports_nfsd_open(struct inode *inode, struct file *file)
168 {
169 	return exports_net_open(inode->i_sb->s_fs_info, file);
170 }
171 
172 static const struct file_operations exports_nfsd_operations = {
173 	.open		= exports_nfsd_open,
174 	.read		= seq_read,
175 	.llseek		= seq_lseek,
176 	.release	= seq_release,
177 };
178 
179 static int export_features_show(struct seq_file *m, void *v)
180 {
181 	seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
182 	return 0;
183 }
184 
185 static int export_features_open(struct inode *inode, struct file *file)
186 {
187 	return single_open(file, export_features_show, NULL);
188 }
189 
190 static const struct file_operations export_features_operations = {
191 	.open		= export_features_open,
192 	.read		= seq_read,
193 	.llseek		= seq_lseek,
194 	.release	= single_release,
195 };
196 
197 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
198 static int supported_enctypes_show(struct seq_file *m, void *v)
199 {
200 	seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
201 	return 0;
202 }
203 
204 static int supported_enctypes_open(struct inode *inode, struct file *file)
205 {
206 	return single_open(file, supported_enctypes_show, NULL);
207 }
208 
209 static const struct file_operations supported_enctypes_ops = {
210 	.open		= supported_enctypes_open,
211 	.read		= seq_read,
212 	.llseek		= seq_lseek,
213 	.release	= single_release,
214 };
215 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
216 
217 static const struct file_operations pool_stats_operations = {
218 	.open		= nfsd_pool_stats_open,
219 	.read		= seq_read,
220 	.llseek		= seq_lseek,
221 	.release	= nfsd_pool_stats_release,
222 };
223 
224 static const struct file_operations reply_cache_stats_operations = {
225 	.open		= nfsd_reply_cache_stats_open,
226 	.read		= seq_read,
227 	.llseek		= seq_lseek,
228 	.release	= single_release,
229 };
230 
231 /*----------------------------------------------------------------------------*/
232 /*
233  * payload - write methods
234  */
235 
236 static inline struct net *netns(struct file *file)
237 {
238 	return file_inode(file)->i_sb->s_fs_info;
239 }
240 
241 /**
242  * write_unlock_ip - Release all locks used by a client
243  *
244  * Experimental.
245  *
246  * Input:
247  *			buf:	'\n'-terminated C string containing a
248  *				presentation format IP address
249  *			size:	length of C string in @buf
250  * Output:
251  *	On success:	returns zero if all specified locks were released;
252  *			returns one if one or more locks were not released
253  *	On error:	return code is negative errno value
254  */
255 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
256 {
257 	struct sockaddr_storage address;
258 	struct sockaddr *sap = (struct sockaddr *)&address;
259 	size_t salen = sizeof(address);
260 	char *fo_path;
261 	struct net *net = netns(file);
262 
263 	/* sanity check */
264 	if (size == 0)
265 		return -EINVAL;
266 
267 	if (buf[size-1] != '\n')
268 		return -EINVAL;
269 
270 	fo_path = buf;
271 	if (qword_get(&buf, fo_path, size) < 0)
272 		return -EINVAL;
273 
274 	if (rpc_pton(net, fo_path, size, sap, salen) == 0)
275 		return -EINVAL;
276 
277 	return nlmsvc_unlock_all_by_ip(sap);
278 }
279 
280 /**
281  * write_unlock_fs - Release all locks on a local file system
282  *
283  * Experimental.
284  *
285  * Input:
286  *			buf:	'\n'-terminated C string containing the
287  *				absolute pathname of a local file system
288  *			size:	length of C string in @buf
289  * Output:
290  *	On success:	returns zero if all specified locks were released;
291  *			returns one if one or more locks were not released
292  *	On error:	return code is negative errno value
293  */
294 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
295 {
296 	struct path path;
297 	char *fo_path;
298 	int error;
299 
300 	/* sanity check */
301 	if (size == 0)
302 		return -EINVAL;
303 
304 	if (buf[size-1] != '\n')
305 		return -EINVAL;
306 
307 	fo_path = buf;
308 	if (qword_get(&buf, fo_path, size) < 0)
309 		return -EINVAL;
310 
311 	error = kern_path(fo_path, 0, &path);
312 	if (error)
313 		return error;
314 
315 	/*
316 	 * XXX: Needs better sanity checking.  Otherwise we could end up
317 	 * releasing locks on the wrong file system.
318 	 *
319 	 * For example:
320 	 * 1.  Does the path refer to a directory?
321 	 * 2.  Is that directory a mount point, or
322 	 * 3.  Is that directory the root of an exported file system?
323 	 */
324 	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
325 
326 	path_put(&path);
327 	return error;
328 }
329 
330 /**
331  * write_filehandle - Get a variable-length NFS file handle by path
332  *
333  * On input, the buffer contains a '\n'-terminated C string comprised of
334  * three alphanumeric words separated by whitespace.  The string may
335  * contain escape sequences.
336  *
337  * Input:
338  *			buf:
339  *				domain:		client domain name
340  *				path:		export pathname
341  *				maxsize:	numeric maximum size of
342  *						@buf
343  *			size:	length of C string in @buf
344  * Output:
345  *	On success:	passed-in buffer filled with '\n'-terminated C
346  *			string containing a ASCII hex text version
347  *			of the NFS file handle;
348  *			return code is the size in bytes of the string
349  *	On error:	return code is negative errno value
350  */
351 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
352 {
353 	char *dname, *path;
354 	int uninitialized_var(maxsize);
355 	char *mesg = buf;
356 	int len;
357 	struct auth_domain *dom;
358 	struct knfsd_fh fh;
359 
360 	if (size == 0)
361 		return -EINVAL;
362 
363 	if (buf[size-1] != '\n')
364 		return -EINVAL;
365 	buf[size-1] = 0;
366 
367 	dname = mesg;
368 	len = qword_get(&mesg, dname, size);
369 	if (len <= 0)
370 		return -EINVAL;
371 
372 	path = dname+len+1;
373 	len = qword_get(&mesg, path, size);
374 	if (len <= 0)
375 		return -EINVAL;
376 
377 	len = get_int(&mesg, &maxsize);
378 	if (len)
379 		return len;
380 
381 	if (maxsize < NFS_FHSIZE)
382 		return -EINVAL;
383 	maxsize = min(maxsize, NFS3_FHSIZE);
384 
385 	if (qword_get(&mesg, mesg, size)>0)
386 		return -EINVAL;
387 
388 	/* we have all the words, they are in buf.. */
389 	dom = unix_domain_find(dname);
390 	if (!dom)
391 		return -ENOMEM;
392 
393 	len = exp_rootfh(netns(file), dom, path, &fh,  maxsize);
394 	auth_domain_put(dom);
395 	if (len)
396 		return len;
397 
398 	mesg = buf;
399 	len = SIMPLE_TRANSACTION_LIMIT;
400 	qword_addhex(&mesg, &len, (char*)&fh.fh_base, fh.fh_size);
401 	mesg[-1] = '\n';
402 	return mesg - buf;
403 }
404 
405 /**
406  * write_threads - Start NFSD, or report the current number of running threads
407  *
408  * Input:
409  *			buf:		ignored
410  *			size:		zero
411  * Output:
412  *	On success:	passed-in buffer filled with '\n'-terminated C
413  *			string numeric value representing the number of
414  *			running NFSD threads;
415  *			return code is the size in bytes of the string
416  *	On error:	return code is zero
417  *
418  * OR
419  *
420  * Input:
421  *			buf:		C string containing an unsigned
422  *					integer value representing the
423  *					number of NFSD threads to start
424  *			size:		non-zero length of C string in @buf
425  * Output:
426  *	On success:	NFS service is started;
427  *			passed-in buffer filled with '\n'-terminated C
428  *			string numeric value representing the number of
429  *			running NFSD threads;
430  *			return code is the size in bytes of the string
431  *	On error:	return code is zero or a negative errno value
432  */
433 static ssize_t write_threads(struct file *file, char *buf, size_t size)
434 {
435 	char *mesg = buf;
436 	int rv;
437 	struct net *net = netns(file);
438 
439 	if (size > 0) {
440 		int newthreads;
441 		rv = get_int(&mesg, &newthreads);
442 		if (rv)
443 			return rv;
444 		if (newthreads < 0)
445 			return -EINVAL;
446 		rv = nfsd_svc(newthreads, net, file->f_cred);
447 		if (rv < 0)
448 			return rv;
449 	} else
450 		rv = nfsd_nrthreads(net);
451 
452 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
453 }
454 
455 /**
456  * write_pool_threads - Set or report the current number of threads per pool
457  *
458  * Input:
459  *			buf:		ignored
460  *			size:		zero
461  *
462  * OR
463  *
464  * Input:
465  * 			buf:		C string containing whitespace-
466  * 					separated unsigned integer values
467  *					representing the number of NFSD
468  *					threads to start in each pool
469  *			size:		non-zero length of C string in @buf
470  * Output:
471  *	On success:	passed-in buffer filled with '\n'-terminated C
472  *			string containing integer values representing the
473  *			number of NFSD threads in each pool;
474  *			return code is the size in bytes of the string
475  *	On error:	return code is zero or a negative errno value
476  */
477 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
478 {
479 	/* if size > 0, look for an array of number of threads per node
480 	 * and apply them  then write out number of threads per node as reply
481 	 */
482 	char *mesg = buf;
483 	int i;
484 	int rv;
485 	int len;
486 	int npools;
487 	int *nthreads;
488 	struct net *net = netns(file);
489 
490 	mutex_lock(&nfsd_mutex);
491 	npools = nfsd_nrpools(net);
492 	if (npools == 0) {
493 		/*
494 		 * NFS is shut down.  The admin can start it by
495 		 * writing to the threads file but NOT the pool_threads
496 		 * file, sorry.  Report zero threads.
497 		 */
498 		mutex_unlock(&nfsd_mutex);
499 		strcpy(buf, "0\n");
500 		return strlen(buf);
501 	}
502 
503 	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
504 	rv = -ENOMEM;
505 	if (nthreads == NULL)
506 		goto out_free;
507 
508 	if (size > 0) {
509 		for (i = 0; i < npools; i++) {
510 			rv = get_int(&mesg, &nthreads[i]);
511 			if (rv == -ENOENT)
512 				break;		/* fewer numbers than pools */
513 			if (rv)
514 				goto out_free;	/* syntax error */
515 			rv = -EINVAL;
516 			if (nthreads[i] < 0)
517 				goto out_free;
518 		}
519 		rv = nfsd_set_nrthreads(i, nthreads, net);
520 		if (rv)
521 			goto out_free;
522 	}
523 
524 	rv = nfsd_get_nrthreads(npools, nthreads, net);
525 	if (rv)
526 		goto out_free;
527 
528 	mesg = buf;
529 	size = SIMPLE_TRANSACTION_LIMIT;
530 	for (i = 0; i < npools && size > 0; i++) {
531 		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
532 		len = strlen(mesg);
533 		size -= len;
534 		mesg += len;
535 	}
536 	rv = mesg - buf;
537 out_free:
538 	kfree(nthreads);
539 	mutex_unlock(&nfsd_mutex);
540 	return rv;
541 }
542 
543 static ssize_t
544 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
545 		const char *sep, unsigned vers, int minor)
546 {
547 	const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
548 	bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
549 
550 	if (vers == 4 && minor >= 0 &&
551 	    !nfsd_minorversion(nn, minor, NFSD_TEST))
552 		supported = false;
553 	if (minor == 0 && supported)
554 		/*
555 		 * special case for backward compatability.
556 		 * +4.0 is never reported, it is implied by
557 		 * +4, unless -4.0 is present.
558 		 */
559 		return 0;
560 	return snprintf(buf, remaining, format, sep,
561 			supported ? '+' : '-', vers, minor);
562 }
563 
564 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
565 {
566 	char *mesg = buf;
567 	char *vers, *minorp, sign;
568 	int len, num, remaining;
569 	ssize_t tlen = 0;
570 	char *sep;
571 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
572 
573 	if (size>0) {
574 		if (nn->nfsd_serv)
575 			/* Cannot change versions without updating
576 			 * nn->nfsd_serv->sv_xdrsize, and reallocing
577 			 * rq_argp and rq_resp
578 			 */
579 			return -EBUSY;
580 		if (buf[size-1] != '\n')
581 			return -EINVAL;
582 		buf[size-1] = 0;
583 
584 		vers = mesg;
585 		len = qword_get(&mesg, vers, size);
586 		if (len <= 0) return -EINVAL;
587 		do {
588 			enum vers_op cmd;
589 			unsigned minor;
590 			sign = *vers;
591 			if (sign == '+' || sign == '-')
592 				num = simple_strtol((vers+1), &minorp, 0);
593 			else
594 				num = simple_strtol(vers, &minorp, 0);
595 			if (*minorp == '.') {
596 				if (num != 4)
597 					return -EINVAL;
598 				if (kstrtouint(minorp+1, 0, &minor) < 0)
599 					return -EINVAL;
600 			}
601 
602 			cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
603 			switch(num) {
604 			case 2:
605 			case 3:
606 				nfsd_vers(nn, num, cmd);
607 				break;
608 			case 4:
609 				if (*minorp == '.') {
610 					if (nfsd_minorversion(nn, minor, cmd) < 0)
611 						return -EINVAL;
612 				} else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
613 					/*
614 					 * Either we have +4 and no minors are enabled,
615 					 * or we have -4 and at least one minor is enabled.
616 					 * In either case, propagate 'cmd' to all minors.
617 					 */
618 					minor = 0;
619 					while (nfsd_minorversion(nn, minor, cmd) >= 0)
620 						minor++;
621 				}
622 				break;
623 			default:
624 				return -EINVAL;
625 			}
626 			vers += len + 1;
627 		} while ((len = qword_get(&mesg, vers, size)) > 0);
628 		/* If all get turned off, turn them back on, as
629 		 * having no versions is BAD
630 		 */
631 		nfsd_reset_versions(nn);
632 	}
633 
634 	/* Now write current state into reply buffer */
635 	len = 0;
636 	sep = "";
637 	remaining = SIMPLE_TRANSACTION_LIMIT;
638 	for (num=2 ; num <= 4 ; num++) {
639 		int minor;
640 		if (!nfsd_vers(nn, num, NFSD_AVAIL))
641 			continue;
642 
643 		minor = -1;
644 		do {
645 			len = nfsd_print_version_support(nn, buf, remaining,
646 					sep, num, minor);
647 			if (len >= remaining)
648 				goto out;
649 			remaining -= len;
650 			buf += len;
651 			tlen += len;
652 			minor++;
653 			if (len)
654 				sep = " ";
655 		} while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
656 	}
657 out:
658 	len = snprintf(buf, remaining, "\n");
659 	if (len >= remaining)
660 		return -EINVAL;
661 	return tlen + len;
662 }
663 
664 /**
665  * write_versions - Set or report the available NFS protocol versions
666  *
667  * Input:
668  *			buf:		ignored
669  *			size:		zero
670  * Output:
671  *	On success:	passed-in buffer filled with '\n'-terminated C
672  *			string containing positive or negative integer
673  *			values representing the current status of each
674  *			protocol version;
675  *			return code is the size in bytes of the string
676  *	On error:	return code is zero or a negative errno value
677  *
678  * OR
679  *
680  * Input:
681  * 			buf:		C string containing whitespace-
682  * 					separated positive or negative
683  * 					integer values representing NFS
684  * 					protocol versions to enable ("+n")
685  * 					or disable ("-n")
686  *			size:		non-zero length of C string in @buf
687  * Output:
688  *	On success:	status of zero or more protocol versions has
689  *			been updated; passed-in buffer filled with
690  *			'\n'-terminated C string containing positive
691  *			or negative integer values representing the
692  *			current status of each protocol version;
693  *			return code is the size in bytes of the string
694  *	On error:	return code is zero or a negative errno value
695  */
696 static ssize_t write_versions(struct file *file, char *buf, size_t size)
697 {
698 	ssize_t rv;
699 
700 	mutex_lock(&nfsd_mutex);
701 	rv = __write_versions(file, buf, size);
702 	mutex_unlock(&nfsd_mutex);
703 	return rv;
704 }
705 
706 /*
707  * Zero-length write.  Return a list of NFSD's current listener
708  * transports.
709  */
710 static ssize_t __write_ports_names(char *buf, struct net *net)
711 {
712 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
713 
714 	if (nn->nfsd_serv == NULL)
715 		return 0;
716 	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
717 }
718 
719 /*
720  * A single 'fd' number was written, in which case it must be for
721  * a socket of a supported family/protocol, and we use it as an
722  * nfsd listener.
723  */
724 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
725 {
726 	char *mesg = buf;
727 	int fd, err;
728 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
729 
730 	err = get_int(&mesg, &fd);
731 	if (err != 0 || fd < 0)
732 		return -EINVAL;
733 
734 	if (svc_alien_sock(net, fd)) {
735 		printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
736 		return -EINVAL;
737 	}
738 
739 	err = nfsd_create_serv(net);
740 	if (err != 0)
741 		return err;
742 
743 	err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
744 	if (err < 0) {
745 		nfsd_destroy(net);
746 		return err;
747 	}
748 
749 	/* Decrease the count, but don't shut down the service */
750 	nn->nfsd_serv->sv_nrthreads--;
751 	return err;
752 }
753 
754 /*
755  * A transport listener is added by writing it's transport name and
756  * a port number.
757  */
758 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
759 {
760 	char transport[16];
761 	struct svc_xprt *xprt;
762 	int port, err;
763 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
764 
765 	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
766 		return -EINVAL;
767 
768 	if (port < 1 || port > USHRT_MAX)
769 		return -EINVAL;
770 
771 	err = nfsd_create_serv(net);
772 	if (err != 0)
773 		return err;
774 
775 	err = svc_create_xprt(nn->nfsd_serv, transport, net,
776 				PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
777 	if (err < 0)
778 		goto out_err;
779 
780 	err = svc_create_xprt(nn->nfsd_serv, transport, net,
781 				PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
782 	if (err < 0 && err != -EAFNOSUPPORT)
783 		goto out_close;
784 
785 	/* Decrease the count, but don't shut down the service */
786 	nn->nfsd_serv->sv_nrthreads--;
787 	return 0;
788 out_close:
789 	xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
790 	if (xprt != NULL) {
791 		svc_close_xprt(xprt);
792 		svc_xprt_put(xprt);
793 	}
794 out_err:
795 	nfsd_destroy(net);
796 	return err;
797 }
798 
799 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
800 			     struct net *net)
801 {
802 	if (size == 0)
803 		return __write_ports_names(buf, net);
804 
805 	if (isdigit(buf[0]))
806 		return __write_ports_addfd(buf, net, file->f_cred);
807 
808 	if (isalpha(buf[0]))
809 		return __write_ports_addxprt(buf, net, file->f_cred);
810 
811 	return -EINVAL;
812 }
813 
814 /**
815  * write_ports - Pass a socket file descriptor or transport name to listen on
816  *
817  * Input:
818  *			buf:		ignored
819  *			size:		zero
820  * Output:
821  *	On success:	passed-in buffer filled with a '\n'-terminated C
822  *			string containing a whitespace-separated list of
823  *			named NFSD listeners;
824  *			return code is the size in bytes of the string
825  *	On error:	return code is zero or a negative errno value
826  *
827  * OR
828  *
829  * Input:
830  *			buf:		C string containing an unsigned
831  *					integer value representing a bound
832  *					but unconnected socket that is to be
833  *					used as an NFSD listener; listen(3)
834  *					must be called for a SOCK_STREAM
835  *					socket, otherwise it is ignored
836  *			size:		non-zero length of C string in @buf
837  * Output:
838  *	On success:	NFS service is started;
839  *			passed-in buffer filled with a '\n'-terminated C
840  *			string containing a unique alphanumeric name of
841  *			the listener;
842  *			return code is the size in bytes of the string
843  *	On error:	return code is a negative errno value
844  *
845  * OR
846  *
847  * Input:
848  *			buf:		C string containing a transport
849  *					name and an unsigned integer value
850  *					representing the port to listen on,
851  *					separated by whitespace
852  *			size:		non-zero length of C string in @buf
853  * Output:
854  *	On success:	returns zero; NFS service is started
855  *	On error:	return code is a negative errno value
856  */
857 static ssize_t write_ports(struct file *file, char *buf, size_t size)
858 {
859 	ssize_t rv;
860 
861 	mutex_lock(&nfsd_mutex);
862 	rv = __write_ports(file, buf, size, netns(file));
863 	mutex_unlock(&nfsd_mutex);
864 	return rv;
865 }
866 
867 
868 int nfsd_max_blksize;
869 
870 /**
871  * write_maxblksize - Set or report the current NFS blksize
872  *
873  * Input:
874  *			buf:		ignored
875  *			size:		zero
876  *
877  * OR
878  *
879  * Input:
880  * 			buf:		C string containing an unsigned
881  * 					integer value representing the new
882  * 					NFS blksize
883  *			size:		non-zero length of C string in @buf
884  * Output:
885  *	On success:	passed-in buffer filled with '\n'-terminated C string
886  *			containing numeric value of the current NFS blksize
887  *			setting;
888  *			return code is the size in bytes of the string
889  *	On error:	return code is zero or a negative errno value
890  */
891 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
892 {
893 	char *mesg = buf;
894 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
895 
896 	if (size > 0) {
897 		int bsize;
898 		int rv = get_int(&mesg, &bsize);
899 		if (rv)
900 			return rv;
901 		/* force bsize into allowed range and
902 		 * required alignment.
903 		 */
904 		bsize = max_t(int, bsize, 1024);
905 		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
906 		bsize &= ~(1024-1);
907 		mutex_lock(&nfsd_mutex);
908 		if (nn->nfsd_serv) {
909 			mutex_unlock(&nfsd_mutex);
910 			return -EBUSY;
911 		}
912 		nfsd_max_blksize = bsize;
913 		mutex_unlock(&nfsd_mutex);
914 	}
915 
916 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
917 							nfsd_max_blksize);
918 }
919 
920 /**
921  * write_maxconn - Set or report the current max number of connections
922  *
923  * Input:
924  *			buf:		ignored
925  *			size:		zero
926  * OR
927  *
928  * Input:
929  * 			buf:		C string containing an unsigned
930  * 					integer value representing the new
931  * 					number of max connections
932  *			size:		non-zero length of C string in @buf
933  * Output:
934  *	On success:	passed-in buffer filled with '\n'-terminated C string
935  *			containing numeric value of max_connections setting
936  *			for this net namespace;
937  *			return code is the size in bytes of the string
938  *	On error:	return code is zero or a negative errno value
939  */
940 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
941 {
942 	char *mesg = buf;
943 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
944 	unsigned int maxconn = nn->max_connections;
945 
946 	if (size > 0) {
947 		int rv = get_uint(&mesg, &maxconn);
948 
949 		if (rv)
950 			return rv;
951 		nn->max_connections = maxconn;
952 	}
953 
954 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
955 }
956 
957 #ifdef CONFIG_NFSD_V4
958 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
959 				  time64_t *time, struct nfsd_net *nn)
960 {
961 	char *mesg = buf;
962 	int rv, i;
963 
964 	if (size > 0) {
965 		if (nn->nfsd_serv)
966 			return -EBUSY;
967 		rv = get_int(&mesg, &i);
968 		if (rv)
969 			return rv;
970 		/*
971 		 * Some sanity checking.  We don't have a reason for
972 		 * these particular numbers, but problems with the
973 		 * extremes are:
974 		 *	- Too short: the briefest network outage may
975 		 *	  cause clients to lose all their locks.  Also,
976 		 *	  the frequent polling may be wasteful.
977 		 *	- Too long: do you really want reboot recovery
978 		 *	  to take more than an hour?  Or to make other
979 		 *	  clients wait an hour before being able to
980 		 *	  revoke a dead client's locks?
981 		 */
982 		if (i < 10 || i > 3600)
983 			return -EINVAL;
984 		*time = i;
985 	}
986 
987 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
988 }
989 
990 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
991 				time64_t *time, struct nfsd_net *nn)
992 {
993 	ssize_t rv;
994 
995 	mutex_lock(&nfsd_mutex);
996 	rv = __nfsd4_write_time(file, buf, size, time, nn);
997 	mutex_unlock(&nfsd_mutex);
998 	return rv;
999 }
1000 
1001 /**
1002  * write_leasetime - Set or report the current NFSv4 lease time
1003  *
1004  * Input:
1005  *			buf:		ignored
1006  *			size:		zero
1007  *
1008  * OR
1009  *
1010  * Input:
1011  *			buf:		C string containing an unsigned
1012  *					integer value representing the new
1013  *					NFSv4 lease expiry time
1014  *			size:		non-zero length of C string in @buf
1015  * Output:
1016  *	On success:	passed-in buffer filled with '\n'-terminated C
1017  *			string containing unsigned integer value of the
1018  *			current lease expiry time;
1019  *			return code is the size in bytes of the string
1020  *	On error:	return code is zero or a negative errno value
1021  */
1022 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1023 {
1024 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1025 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1026 }
1027 
1028 /**
1029  * write_gracetime - Set or report current NFSv4 grace period time
1030  *
1031  * As above, but sets the time of the NFSv4 grace period.
1032  *
1033  * Note this should never be set to less than the *previous*
1034  * lease-period time, but we don't try to enforce this.  (In the common
1035  * case (a new boot), we don't know what the previous lease time was
1036  * anyway.)
1037  */
1038 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1039 {
1040 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1041 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1042 }
1043 
1044 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1045 				   struct nfsd_net *nn)
1046 {
1047 	char *mesg = buf;
1048 	char *recdir;
1049 	int len, status;
1050 
1051 	if (size > 0) {
1052 		if (nn->nfsd_serv)
1053 			return -EBUSY;
1054 		if (size > PATH_MAX || buf[size-1] != '\n')
1055 			return -EINVAL;
1056 		buf[size-1] = 0;
1057 
1058 		recdir = mesg;
1059 		len = qword_get(&mesg, recdir, size);
1060 		if (len <= 0)
1061 			return -EINVAL;
1062 
1063 		status = nfs4_reset_recoverydir(recdir);
1064 		if (status)
1065 			return status;
1066 	}
1067 
1068 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1069 							nfs4_recoverydir());
1070 }
1071 
1072 /**
1073  * write_recoverydir - Set or report the pathname of the recovery directory
1074  *
1075  * Input:
1076  *			buf:		ignored
1077  *			size:		zero
1078  *
1079  * OR
1080  *
1081  * Input:
1082  *			buf:		C string containing the pathname
1083  *					of the directory on a local file
1084  *					system containing permanent NFSv4
1085  *					recovery data
1086  *			size:		non-zero length of C string in @buf
1087  * Output:
1088  *	On success:	passed-in buffer filled with '\n'-terminated C string
1089  *			containing the current recovery pathname setting;
1090  *			return code is the size in bytes of the string
1091  *	On error:	return code is zero or a negative errno value
1092  */
1093 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1094 {
1095 	ssize_t rv;
1096 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1097 
1098 	mutex_lock(&nfsd_mutex);
1099 	rv = __write_recoverydir(file, buf, size, nn);
1100 	mutex_unlock(&nfsd_mutex);
1101 	return rv;
1102 }
1103 
1104 /**
1105  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1106  *
1107  * Input:
1108  *			buf:		ignored
1109  *			size:		zero
1110  * OR
1111  *
1112  * Input:
1113  * 			buf:		any value
1114  *			size:		non-zero length of C string in @buf
1115  * Output:
1116  *			passed-in buffer filled with "Y" or "N" with a newline
1117  *			and NULL-terminated C string. This indicates whether
1118  *			the grace period has ended in the current net
1119  *			namespace. Return code is the size in bytes of the
1120  *			string. Writing a string that starts with 'Y', 'y', or
1121  *			'1' to the file will end the grace period for nfsd's v4
1122  *			lock manager.
1123  */
1124 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1125 {
1126 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1127 
1128 	if (size > 0) {
1129 		switch(buf[0]) {
1130 		case 'Y':
1131 		case 'y':
1132 		case '1':
1133 			if (!nn->nfsd_serv)
1134 				return -EBUSY;
1135 			nfsd4_end_grace(nn);
1136 			break;
1137 		default:
1138 			return -EINVAL;
1139 		}
1140 	}
1141 
1142 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1143 			 nn->grace_ended ? 'Y' : 'N');
1144 }
1145 
1146 #endif
1147 
1148 /*----------------------------------------------------------------------------*/
1149 /*
1150  *	populating the filesystem.
1151  */
1152 
1153 /* Basically copying rpc_get_inode. */
1154 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1155 {
1156 	struct inode *inode = new_inode(sb);
1157 	if (!inode)
1158 		return NULL;
1159 	/* Following advice from simple_fill_super documentation: */
1160 	inode->i_ino = iunique(sb, NFSD_MaxReserved);
1161 	inode->i_mode = mode;
1162 	inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1163 	switch (mode & S_IFMT) {
1164 	case S_IFDIR:
1165 		inode->i_fop = &simple_dir_operations;
1166 		inode->i_op = &simple_dir_inode_operations;
1167 		inc_nlink(inode);
1168 	default:
1169 		break;
1170 	}
1171 	return inode;
1172 }
1173 
1174 static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1175 {
1176 	struct inode *inode;
1177 
1178 	inode = nfsd_get_inode(dir->i_sb, mode);
1179 	if (!inode)
1180 		return -ENOMEM;
1181 	if (ncl) {
1182 		inode->i_private = ncl;
1183 		kref_get(&ncl->cl_ref);
1184 	}
1185 	d_add(dentry, inode);
1186 	inc_nlink(dir);
1187 	fsnotify_mkdir(dir, dentry);
1188 	return 0;
1189 }
1190 
1191 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1192 {
1193 	struct inode *dir = parent->d_inode;
1194 	struct dentry *dentry;
1195 	int ret = -ENOMEM;
1196 
1197 	inode_lock(dir);
1198 	dentry = d_alloc_name(parent, name);
1199 	if (!dentry)
1200 		goto out_err;
1201 	ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1202 	if (ret)
1203 		goto out_err;
1204 out:
1205 	inode_unlock(dir);
1206 	return dentry;
1207 out_err:
1208 	dput(dentry);
1209 	dentry = ERR_PTR(ret);
1210 	goto out;
1211 }
1212 
1213 static void clear_ncl(struct inode *inode)
1214 {
1215 	struct nfsdfs_client *ncl = inode->i_private;
1216 
1217 	inode->i_private = NULL;
1218 	kref_put(&ncl->cl_ref, ncl->cl_release);
1219 }
1220 
1221 static struct nfsdfs_client *__get_nfsdfs_client(struct inode *inode)
1222 {
1223 	struct nfsdfs_client *nc = inode->i_private;
1224 
1225 	if (nc)
1226 		kref_get(&nc->cl_ref);
1227 	return nc;
1228 }
1229 
1230 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1231 {
1232 	struct nfsdfs_client *nc;
1233 
1234 	inode_lock_shared(inode);
1235 	nc = __get_nfsdfs_client(inode);
1236 	inode_unlock_shared(inode);
1237 	return nc;
1238 }
1239 /* from __rpc_unlink */
1240 static void nfsdfs_remove_file(struct inode *dir, struct dentry *dentry)
1241 {
1242 	int ret;
1243 
1244 	clear_ncl(d_inode(dentry));
1245 	dget(dentry);
1246 	ret = simple_unlink(dir, dentry);
1247 	d_delete(dentry);
1248 	dput(dentry);
1249 	WARN_ON_ONCE(ret);
1250 }
1251 
1252 static void nfsdfs_remove_files(struct dentry *root)
1253 {
1254 	struct dentry *dentry, *tmp;
1255 
1256 	list_for_each_entry_safe(dentry, tmp, &root->d_subdirs, d_child) {
1257 		if (!simple_positive(dentry)) {
1258 			WARN_ON_ONCE(1); /* I think this can't happen? */
1259 			continue;
1260 		}
1261 		nfsdfs_remove_file(d_inode(root), dentry);
1262 	}
1263 }
1264 
1265 /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1266  * code instead. */
1267 static  int nfsdfs_create_files(struct dentry *root,
1268 					const struct tree_descr *files)
1269 {
1270 	struct inode *dir = d_inode(root);
1271 	struct inode *inode;
1272 	struct dentry *dentry;
1273 	int i;
1274 
1275 	inode_lock(dir);
1276 	for (i = 0; files->name && files->name[0]; i++, files++) {
1277 		if (!files->name)
1278 			continue;
1279 		dentry = d_alloc_name(root, files->name);
1280 		if (!dentry)
1281 			goto out;
1282 		inode = nfsd_get_inode(d_inode(root)->i_sb,
1283 					S_IFREG | files->mode);
1284 		if (!inode) {
1285 			dput(dentry);
1286 			goto out;
1287 		}
1288 		inode->i_fop = files->ops;
1289 		inode->i_private = __get_nfsdfs_client(dir);
1290 		d_add(dentry, inode);
1291 		fsnotify_create(dir, dentry);
1292 	}
1293 	inode_unlock(dir);
1294 	return 0;
1295 out:
1296 	nfsdfs_remove_files(root);
1297 	inode_unlock(dir);
1298 	return -ENOMEM;
1299 }
1300 
1301 /* on success, returns positive number unique to that client. */
1302 struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1303 		struct nfsdfs_client *ncl, u32 id,
1304 		const struct tree_descr *files)
1305 {
1306 	struct dentry *dentry;
1307 	char name[11];
1308 	int ret;
1309 
1310 	sprintf(name, "%u", id);
1311 
1312 	dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1313 	if (IS_ERR(dentry)) /* XXX: tossing errors? */
1314 		return NULL;
1315 	ret = nfsdfs_create_files(dentry, files);
1316 	if (ret) {
1317 		nfsd_client_rmdir(dentry);
1318 		return NULL;
1319 	}
1320 	return dentry;
1321 }
1322 
1323 /* Taken from __rpc_rmdir: */
1324 void nfsd_client_rmdir(struct dentry *dentry)
1325 {
1326 	struct inode *dir = d_inode(dentry->d_parent);
1327 	struct inode *inode = d_inode(dentry);
1328 	int ret;
1329 
1330 	inode_lock(dir);
1331 	nfsdfs_remove_files(dentry);
1332 	clear_ncl(inode);
1333 	dget(dentry);
1334 	ret = simple_rmdir(dir, dentry);
1335 	WARN_ON_ONCE(ret);
1336 	fsnotify_rmdir(dir, dentry);
1337 	d_delete(dentry);
1338 	inode_unlock(dir);
1339 }
1340 
1341 static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1342 {
1343 	struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1344 							nfsd_net_id);
1345 	struct dentry *dentry;
1346 	int ret;
1347 
1348 	static const struct tree_descr nfsd_files[] = {
1349 		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1350 		[NFSD_Export_features] = {"export_features",
1351 					&export_features_operations, S_IRUGO},
1352 		[NFSD_FO_UnlockIP] = {"unlock_ip",
1353 					&transaction_ops, S_IWUSR|S_IRUSR},
1354 		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1355 					&transaction_ops, S_IWUSR|S_IRUSR},
1356 		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1357 		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1358 		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1359 		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1360 		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
1361 		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1362 		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1363 		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1364 		[NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1365 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1366 		[NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1367 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1368 #ifdef CONFIG_NFSD_V4
1369 		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1370 		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1371 		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1372 		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1373 #endif
1374 		/* last one */ {""}
1375 	};
1376 
1377 	ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1378 	if (ret)
1379 		return ret;
1380 	dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1381 	if (IS_ERR(dentry))
1382 		return PTR_ERR(dentry);
1383 	nn->nfsd_client_dir = dentry;
1384 	return 0;
1385 }
1386 
1387 static int nfsd_fs_get_tree(struct fs_context *fc)
1388 {
1389 	return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1390 }
1391 
1392 static void nfsd_fs_free_fc(struct fs_context *fc)
1393 {
1394 	if (fc->s_fs_info)
1395 		put_net(fc->s_fs_info);
1396 }
1397 
1398 static const struct fs_context_operations nfsd_fs_context_ops = {
1399 	.free		= nfsd_fs_free_fc,
1400 	.get_tree	= nfsd_fs_get_tree,
1401 };
1402 
1403 static int nfsd_init_fs_context(struct fs_context *fc)
1404 {
1405 	put_user_ns(fc->user_ns);
1406 	fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1407 	fc->ops = &nfsd_fs_context_ops;
1408 	return 0;
1409 }
1410 
1411 static void nfsd_umount(struct super_block *sb)
1412 {
1413 	struct net *net = sb->s_fs_info;
1414 
1415 	kill_litter_super(sb);
1416 	put_net(net);
1417 }
1418 
1419 static struct file_system_type nfsd_fs_type = {
1420 	.owner		= THIS_MODULE,
1421 	.name		= "nfsd",
1422 	.init_fs_context = nfsd_init_fs_context,
1423 	.kill_sb	= nfsd_umount,
1424 };
1425 MODULE_ALIAS_FS("nfsd");
1426 
1427 #ifdef CONFIG_PROC_FS
1428 static int create_proc_exports_entry(void)
1429 {
1430 	struct proc_dir_entry *entry;
1431 
1432 	entry = proc_mkdir("fs/nfs", NULL);
1433 	if (!entry)
1434 		return -ENOMEM;
1435 	entry = proc_create("exports", 0, entry, &exports_proc_ops);
1436 	if (!entry) {
1437 		remove_proc_entry("fs/nfs", NULL);
1438 		return -ENOMEM;
1439 	}
1440 	return 0;
1441 }
1442 #else /* CONFIG_PROC_FS */
1443 static int create_proc_exports_entry(void)
1444 {
1445 	return 0;
1446 }
1447 #endif
1448 
1449 unsigned int nfsd_net_id;
1450 
1451 static __net_init int nfsd_init_net(struct net *net)
1452 {
1453 	int retval;
1454 	struct vfsmount *mnt;
1455 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1456 
1457 	retval = nfsd_export_init(net);
1458 	if (retval)
1459 		goto out_export_error;
1460 	retval = nfsd_idmap_init(net);
1461 	if (retval)
1462 		goto out_idmap_error;
1463 	nn->nfsd_versions = NULL;
1464 	nn->nfsd4_minorversions = NULL;
1465 	retval = nfsd_reply_cache_init(nn);
1466 	if (retval)
1467 		goto out_drc_error;
1468 	nn->nfsd4_lease = 90;	/* default lease time */
1469 	nn->nfsd4_grace = 90;
1470 	nn->somebody_reclaimed = false;
1471 	nn->track_reclaim_completes = false;
1472 	nn->clverifier_counter = prandom_u32();
1473 	nn->clientid_base = prandom_u32();
1474 	nn->clientid_counter = nn->clientid_base + 1;
1475 	nn->s2s_cp_cl_id = nn->clientid_counter++;
1476 
1477 	atomic_set(&nn->ntf_refcnt, 0);
1478 	init_waitqueue_head(&nn->ntf_wq);
1479 	seqlock_init(&nn->boot_lock);
1480 
1481 	mnt =  vfs_kern_mount(&nfsd_fs_type, SB_KERNMOUNT, "nfsd", NULL);
1482 	if (IS_ERR(mnt)) {
1483 		retval = PTR_ERR(mnt);
1484 		goto out_mount_err;
1485 	}
1486 	nn->nfsd_mnt = mnt;
1487 	return 0;
1488 
1489 out_mount_err:
1490 	nfsd_reply_cache_shutdown(nn);
1491 out_drc_error:
1492 	nfsd_idmap_shutdown(net);
1493 out_idmap_error:
1494 	nfsd_export_shutdown(net);
1495 out_export_error:
1496 	return retval;
1497 }
1498 
1499 static __net_exit void nfsd_exit_net(struct net *net)
1500 {
1501 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1502 
1503 	mntput(nn->nfsd_mnt);
1504 	nfsd_reply_cache_shutdown(nn);
1505 	nfsd_idmap_shutdown(net);
1506 	nfsd_export_shutdown(net);
1507 	nfsd_netns_free_versions(net_generic(net, nfsd_net_id));
1508 }
1509 
1510 static struct pernet_operations nfsd_net_ops = {
1511 	.init = nfsd_init_net,
1512 	.exit = nfsd_exit_net,
1513 	.id   = &nfsd_net_id,
1514 	.size = sizeof(struct nfsd_net),
1515 };
1516 
1517 static int __init init_nfsd(void)
1518 {
1519 	int retval;
1520 	printk(KERN_INFO "Installing knfsd (copyright (C) 1996 okir@monad.swb.de).\n");
1521 
1522 	retval = register_pernet_subsys(&nfsd_net_ops);
1523 	if (retval < 0)
1524 		return retval;
1525 	retval = register_cld_notifier();
1526 	if (retval)
1527 		goto out_unregister_pernet;
1528 	retval = nfsd4_init_slabs();
1529 	if (retval)
1530 		goto out_unregister_notifier;
1531 	retval = nfsd4_init_pnfs();
1532 	if (retval)
1533 		goto out_free_slabs;
1534 	nfsd_fault_inject_init(); /* nfsd fault injection controls */
1535 	nfsd_stat_init();	/* Statistics */
1536 	nfsd_lockd_init();	/* lockd->nfsd callbacks */
1537 	retval = create_proc_exports_entry();
1538 	if (retval)
1539 		goto out_free_lockd;
1540 	retval = register_filesystem(&nfsd_fs_type);
1541 	if (retval)
1542 		goto out_free_all;
1543 	return 0;
1544 out_free_all:
1545 	remove_proc_entry("fs/nfs/exports", NULL);
1546 	remove_proc_entry("fs/nfs", NULL);
1547 out_free_lockd:
1548 	nfsd_lockd_shutdown();
1549 	nfsd_stat_shutdown();
1550 	nfsd_fault_inject_cleanup();
1551 	nfsd4_exit_pnfs();
1552 out_free_slabs:
1553 	nfsd4_free_slabs();
1554 out_unregister_notifier:
1555 	unregister_cld_notifier();
1556 out_unregister_pernet:
1557 	unregister_pernet_subsys(&nfsd_net_ops);
1558 	return retval;
1559 }
1560 
1561 static void __exit exit_nfsd(void)
1562 {
1563 	remove_proc_entry("fs/nfs/exports", NULL);
1564 	remove_proc_entry("fs/nfs", NULL);
1565 	nfsd_stat_shutdown();
1566 	nfsd_lockd_shutdown();
1567 	nfsd4_free_slabs();
1568 	nfsd4_exit_pnfs();
1569 	nfsd_fault_inject_cleanup();
1570 	unregister_filesystem(&nfsd_fs_type);
1571 	unregister_cld_notifier();
1572 	unregister_pernet_subsys(&nfsd_net_ops);
1573 }
1574 
1575 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1576 MODULE_LICENSE("GPL");
1577 module_init(init_nfsd)
1578 module_exit(exit_nfsd)
1579