xref: /openbmc/linux/fs/nfsd/nfsctl.c (revision 474be445555ba8f2e776b4b6458c310bc215f76b)
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  #include "filecache.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  	NFSD_SupportedEnctypes,
51  	/*
52  	 * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
53  	 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
54  	 */
55  #ifdef CONFIG_NFSD_V4
56  	NFSD_Leasetime,
57  	NFSD_Gracetime,
58  	NFSD_RecoveryDir,
59  	NFSD_V4EndGrace,
60  #endif
61  	NFSD_MaxReserved
62  };
63  
64  /*
65   * write() for these nodes.
66   */
67  static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
68  static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
69  static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
70  static ssize_t write_threads(struct file *file, char *buf, size_t size);
71  static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
72  static ssize_t write_versions(struct file *file, char *buf, size_t size);
73  static ssize_t write_ports(struct file *file, char *buf, size_t size);
74  static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
75  static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
76  #ifdef CONFIG_NFSD_V4
77  static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
78  static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
79  static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
80  static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
81  #endif
82  
83  static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
84  	[NFSD_Fh] = write_filehandle,
85  	[NFSD_FO_UnlockIP] = write_unlock_ip,
86  	[NFSD_FO_UnlockFS] = write_unlock_fs,
87  	[NFSD_Threads] = write_threads,
88  	[NFSD_Pool_Threads] = write_pool_threads,
89  	[NFSD_Versions] = write_versions,
90  	[NFSD_Ports] = write_ports,
91  	[NFSD_MaxBlkSize] = write_maxblksize,
92  	[NFSD_MaxConnections] = write_maxconn,
93  #ifdef CONFIG_NFSD_V4
94  	[NFSD_Leasetime] = write_leasetime,
95  	[NFSD_Gracetime] = write_gracetime,
96  	[NFSD_RecoveryDir] = write_recoverydir,
97  	[NFSD_V4EndGrace] = write_v4_end_grace,
98  #endif
99  };
100  
101  static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
102  {
103  	ino_t ino =  file_inode(file)->i_ino;
104  	char *data;
105  	ssize_t rv;
106  
107  	if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
108  		return -EINVAL;
109  
110  	data = simple_transaction_get(file, buf, size);
111  	if (IS_ERR(data))
112  		return PTR_ERR(data);
113  
114  	rv =  write_op[ino](file, data, size);
115  	if (rv >= 0) {
116  		simple_transaction_set(file, rv);
117  		rv = size;
118  	}
119  	return rv;
120  }
121  
122  static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
123  {
124  	if (! file->private_data) {
125  		/* An attempt to read a transaction file without writing
126  		 * causes a 0-byte write so that the file can return
127  		 * state information
128  		 */
129  		ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
130  		if (rv < 0)
131  			return rv;
132  	}
133  	return simple_transaction_read(file, buf, size, pos);
134  }
135  
136  static const struct file_operations transaction_ops = {
137  	.write		= nfsctl_transaction_write,
138  	.read		= nfsctl_transaction_read,
139  	.release	= simple_transaction_release,
140  	.llseek		= default_llseek,
141  };
142  
143  static int exports_net_open(struct net *net, struct file *file)
144  {
145  	int err;
146  	struct seq_file *seq;
147  	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
148  
149  	err = seq_open(file, &nfs_exports_op);
150  	if (err)
151  		return err;
152  
153  	seq = file->private_data;
154  	seq->private = nn->svc_export_cache;
155  	return 0;
156  }
157  
158  static int exports_proc_open(struct inode *inode, struct file *file)
159  {
160  	return exports_net_open(current->nsproxy->net_ns, file);
161  }
162  
163  static const struct proc_ops exports_proc_ops = {
164  	.proc_open	= exports_proc_open,
165  	.proc_read	= seq_read,
166  	.proc_lseek	= seq_lseek,
167  	.proc_release	= seq_release,
168  };
169  
170  static int exports_nfsd_open(struct inode *inode, struct file *file)
171  {
172  	return exports_net_open(inode->i_sb->s_fs_info, file);
173  }
174  
175  static const struct file_operations exports_nfsd_operations = {
176  	.open		= exports_nfsd_open,
177  	.read		= seq_read,
178  	.llseek		= seq_lseek,
179  	.release	= seq_release,
180  };
181  
182  static int export_features_show(struct seq_file *m, void *v)
183  {
184  	seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
185  	return 0;
186  }
187  
188  DEFINE_SHOW_ATTRIBUTE(export_features);
189  
190  #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
191  static int supported_enctypes_show(struct seq_file *m, void *v)
192  {
193  	seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
194  	return 0;
195  }
196  
197  DEFINE_SHOW_ATTRIBUTE(supported_enctypes);
198  #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
199  
200  static const struct file_operations pool_stats_operations = {
201  	.open		= nfsd_pool_stats_open,
202  	.read		= seq_read,
203  	.llseek		= seq_lseek,
204  	.release	= nfsd_pool_stats_release,
205  };
206  
207  DEFINE_SHOW_ATTRIBUTE(nfsd_reply_cache_stats);
208  
209  DEFINE_SHOW_ATTRIBUTE(nfsd_file_cache_stats);
210  
211  /*----------------------------------------------------------------------------*/
212  /*
213   * payload - write methods
214   */
215  
216  static inline struct net *netns(struct file *file)
217  {
218  	return file_inode(file)->i_sb->s_fs_info;
219  }
220  
221  /*
222   * write_unlock_ip - Release all locks used by a client
223   *
224   * Experimental.
225   *
226   * Input:
227   *			buf:	'\n'-terminated C string containing a
228   *				presentation format IP address
229   *			size:	length of C string in @buf
230   * Output:
231   *	On success:	returns zero if all specified locks were released;
232   *			returns one if one or more locks were not released
233   *	On error:	return code is negative errno value
234   */
235  static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
236  {
237  	struct sockaddr_storage address;
238  	struct sockaddr *sap = (struct sockaddr *)&address;
239  	size_t salen = sizeof(address);
240  	char *fo_path;
241  	struct net *net = netns(file);
242  
243  	/* sanity check */
244  	if (size == 0)
245  		return -EINVAL;
246  
247  	if (buf[size-1] != '\n')
248  		return -EINVAL;
249  
250  	fo_path = buf;
251  	if (qword_get(&buf, fo_path, size) < 0)
252  		return -EINVAL;
253  
254  	if (rpc_pton(net, fo_path, size, sap, salen) == 0)
255  		return -EINVAL;
256  
257  	return nlmsvc_unlock_all_by_ip(sap);
258  }
259  
260  /*
261   * write_unlock_fs - Release all locks on a local file system
262   *
263   * Experimental.
264   *
265   * Input:
266   *			buf:	'\n'-terminated C string containing the
267   *				absolute pathname of a local file system
268   *			size:	length of C string in @buf
269   * Output:
270   *	On success:	returns zero if all specified locks were released;
271   *			returns one if one or more locks were not released
272   *	On error:	return code is negative errno value
273   */
274  static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
275  {
276  	struct path path;
277  	char *fo_path;
278  	int error;
279  
280  	/* sanity check */
281  	if (size == 0)
282  		return -EINVAL;
283  
284  	if (buf[size-1] != '\n')
285  		return -EINVAL;
286  
287  	fo_path = buf;
288  	if (qword_get(&buf, fo_path, size) < 0)
289  		return -EINVAL;
290  
291  	error = kern_path(fo_path, 0, &path);
292  	if (error)
293  		return error;
294  
295  	/*
296  	 * XXX: Needs better sanity checking.  Otherwise we could end up
297  	 * releasing locks on the wrong file system.
298  	 *
299  	 * For example:
300  	 * 1.  Does the path refer to a directory?
301  	 * 2.  Is that directory a mount point, or
302  	 * 3.  Is that directory the root of an exported file system?
303  	 */
304  	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
305  
306  	path_put(&path);
307  	return error;
308  }
309  
310  /*
311   * write_filehandle - Get a variable-length NFS file handle by path
312   *
313   * On input, the buffer contains a '\n'-terminated C string comprised of
314   * three alphanumeric words separated by whitespace.  The string may
315   * contain escape sequences.
316   *
317   * Input:
318   *			buf:
319   *				domain:		client domain name
320   *				path:		export pathname
321   *				maxsize:	numeric maximum size of
322   *						@buf
323   *			size:	length of C string in @buf
324   * Output:
325   *	On success:	passed-in buffer filled with '\n'-terminated C
326   *			string containing a ASCII hex text version
327   *			of the NFS file handle;
328   *			return code is the size in bytes of the string
329   *	On error:	return code is negative errno value
330   */
331  static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
332  {
333  	char *dname, *path;
334  	int maxsize;
335  	char *mesg = buf;
336  	int len;
337  	struct auth_domain *dom;
338  	struct knfsd_fh fh;
339  
340  	if (size == 0)
341  		return -EINVAL;
342  
343  	if (buf[size-1] != '\n')
344  		return -EINVAL;
345  	buf[size-1] = 0;
346  
347  	dname = mesg;
348  	len = qword_get(&mesg, dname, size);
349  	if (len <= 0)
350  		return -EINVAL;
351  
352  	path = dname+len+1;
353  	len = qword_get(&mesg, path, size);
354  	if (len <= 0)
355  		return -EINVAL;
356  
357  	len = get_int(&mesg, &maxsize);
358  	if (len)
359  		return len;
360  
361  	if (maxsize < NFS_FHSIZE)
362  		return -EINVAL;
363  	maxsize = min(maxsize, NFS3_FHSIZE);
364  
365  	if (qword_get(&mesg, mesg, size)>0)
366  		return -EINVAL;
367  
368  	/* we have all the words, they are in buf.. */
369  	dom = unix_domain_find(dname);
370  	if (!dom)
371  		return -ENOMEM;
372  
373  	len = exp_rootfh(netns(file), dom, path, &fh,  maxsize);
374  	auth_domain_put(dom);
375  	if (len)
376  		return len;
377  
378  	mesg = buf;
379  	len = SIMPLE_TRANSACTION_LIMIT;
380  	qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
381  	mesg[-1] = '\n';
382  	return mesg - buf;
383  }
384  
385  /*
386   * write_threads - Start NFSD, or report the current number of running threads
387   *
388   * Input:
389   *			buf:		ignored
390   *			size:		zero
391   * Output:
392   *	On success:	passed-in buffer filled with '\n'-terminated C
393   *			string numeric value representing the number of
394   *			running NFSD threads;
395   *			return code is the size in bytes of the string
396   *	On error:	return code is zero
397   *
398   * OR
399   *
400   * Input:
401   *			buf:		C string containing an unsigned
402   *					integer value representing the
403   *					number of NFSD threads to start
404   *			size:		non-zero length of C string in @buf
405   * Output:
406   *	On success:	NFS service is started;
407   *			passed-in buffer filled with '\n'-terminated C
408   *			string numeric value representing the number of
409   *			running NFSD threads;
410   *			return code is the size in bytes of the string
411   *	On error:	return code is zero or a negative errno value
412   */
413  static ssize_t write_threads(struct file *file, char *buf, size_t size)
414  {
415  	char *mesg = buf;
416  	int rv;
417  	struct net *net = netns(file);
418  
419  	if (size > 0) {
420  		int newthreads;
421  		rv = get_int(&mesg, &newthreads);
422  		if (rv)
423  			return rv;
424  		if (newthreads < 0)
425  			return -EINVAL;
426  		rv = nfsd_svc(newthreads, net, file->f_cred);
427  		if (rv < 0)
428  			return rv;
429  	} else
430  		rv = nfsd_nrthreads(net);
431  
432  	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
433  }
434  
435  /*
436   * write_pool_threads - Set or report the current number of threads per pool
437   *
438   * Input:
439   *			buf:		ignored
440   *			size:		zero
441   *
442   * OR
443   *
444   * Input:
445   * 			buf:		C string containing whitespace-
446   * 					separated unsigned integer values
447   *					representing the number of NFSD
448   *					threads to start in each pool
449   *			size:		non-zero length of C string in @buf
450   * Output:
451   *	On success:	passed-in buffer filled with '\n'-terminated C
452   *			string containing integer values representing the
453   *			number of NFSD threads in each pool;
454   *			return code is the size in bytes of the string
455   *	On error:	return code is zero or a negative errno value
456   */
457  static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
458  {
459  	/* if size > 0, look for an array of number of threads per node
460  	 * and apply them  then write out number of threads per node as reply
461  	 */
462  	char *mesg = buf;
463  	int i;
464  	int rv;
465  	int len;
466  	int npools;
467  	int *nthreads;
468  	struct net *net = netns(file);
469  
470  	mutex_lock(&nfsd_mutex);
471  	npools = nfsd_nrpools(net);
472  	if (npools == 0) {
473  		/*
474  		 * NFS is shut down.  The admin can start it by
475  		 * writing to the threads file but NOT the pool_threads
476  		 * file, sorry.  Report zero threads.
477  		 */
478  		mutex_unlock(&nfsd_mutex);
479  		strcpy(buf, "0\n");
480  		return strlen(buf);
481  	}
482  
483  	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
484  	rv = -ENOMEM;
485  	if (nthreads == NULL)
486  		goto out_free;
487  
488  	if (size > 0) {
489  		for (i = 0; i < npools; i++) {
490  			rv = get_int(&mesg, &nthreads[i]);
491  			if (rv == -ENOENT)
492  				break;		/* fewer numbers than pools */
493  			if (rv)
494  				goto out_free;	/* syntax error */
495  			rv = -EINVAL;
496  			if (nthreads[i] < 0)
497  				goto out_free;
498  		}
499  		rv = nfsd_set_nrthreads(i, nthreads, net);
500  		if (rv)
501  			goto out_free;
502  	}
503  
504  	rv = nfsd_get_nrthreads(npools, nthreads, net);
505  	if (rv)
506  		goto out_free;
507  
508  	mesg = buf;
509  	size = SIMPLE_TRANSACTION_LIMIT;
510  	for (i = 0; i < npools && size > 0; i++) {
511  		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
512  		len = strlen(mesg);
513  		size -= len;
514  		mesg += len;
515  	}
516  	rv = mesg - buf;
517  out_free:
518  	kfree(nthreads);
519  	mutex_unlock(&nfsd_mutex);
520  	return rv;
521  }
522  
523  static ssize_t
524  nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
525  		const char *sep, unsigned vers, int minor)
526  {
527  	const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
528  	bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
529  
530  	if (vers == 4 && minor >= 0 &&
531  	    !nfsd_minorversion(nn, minor, NFSD_TEST))
532  		supported = false;
533  	if (minor == 0 && supported)
534  		/*
535  		 * special case for backward compatability.
536  		 * +4.0 is never reported, it is implied by
537  		 * +4, unless -4.0 is present.
538  		 */
539  		return 0;
540  	return snprintf(buf, remaining, format, sep,
541  			supported ? '+' : '-', vers, minor);
542  }
543  
544  static ssize_t __write_versions(struct file *file, char *buf, size_t size)
545  {
546  	char *mesg = buf;
547  	char *vers, *minorp, sign;
548  	int len, num, remaining;
549  	ssize_t tlen = 0;
550  	char *sep;
551  	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
552  
553  	if (size>0) {
554  		if (nn->nfsd_serv)
555  			/* Cannot change versions without updating
556  			 * nn->nfsd_serv->sv_xdrsize, and reallocing
557  			 * rq_argp and rq_resp
558  			 */
559  			return -EBUSY;
560  		if (buf[size-1] != '\n')
561  			return -EINVAL;
562  		buf[size-1] = 0;
563  
564  		vers = mesg;
565  		len = qword_get(&mesg, vers, size);
566  		if (len <= 0) return -EINVAL;
567  		do {
568  			enum vers_op cmd;
569  			unsigned minor;
570  			sign = *vers;
571  			if (sign == '+' || sign == '-')
572  				num = simple_strtol((vers+1), &minorp, 0);
573  			else
574  				num = simple_strtol(vers, &minorp, 0);
575  			if (*minorp == '.') {
576  				if (num != 4)
577  					return -EINVAL;
578  				if (kstrtouint(minorp+1, 0, &minor) < 0)
579  					return -EINVAL;
580  			}
581  
582  			cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
583  			switch(num) {
584  			case 2:
585  			case 3:
586  				nfsd_vers(nn, num, cmd);
587  				break;
588  			case 4:
589  				if (*minorp == '.') {
590  					if (nfsd_minorversion(nn, minor, cmd) < 0)
591  						return -EINVAL;
592  				} else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
593  					/*
594  					 * Either we have +4 and no minors are enabled,
595  					 * or we have -4 and at least one minor is enabled.
596  					 * In either case, propagate 'cmd' to all minors.
597  					 */
598  					minor = 0;
599  					while (nfsd_minorversion(nn, minor, cmd) >= 0)
600  						minor++;
601  				}
602  				break;
603  			default:
604  				return -EINVAL;
605  			}
606  			vers += len + 1;
607  		} while ((len = qword_get(&mesg, vers, size)) > 0);
608  		/* If all get turned off, turn them back on, as
609  		 * having no versions is BAD
610  		 */
611  		nfsd_reset_versions(nn);
612  	}
613  
614  	/* Now write current state into reply buffer */
615  	sep = "";
616  	remaining = SIMPLE_TRANSACTION_LIMIT;
617  	for (num=2 ; num <= 4 ; num++) {
618  		int minor;
619  		if (!nfsd_vers(nn, num, NFSD_AVAIL))
620  			continue;
621  
622  		minor = -1;
623  		do {
624  			len = nfsd_print_version_support(nn, buf, remaining,
625  					sep, num, minor);
626  			if (len >= remaining)
627  				goto out;
628  			remaining -= len;
629  			buf += len;
630  			tlen += len;
631  			minor++;
632  			if (len)
633  				sep = " ";
634  		} while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
635  	}
636  out:
637  	len = snprintf(buf, remaining, "\n");
638  	if (len >= remaining)
639  		return -EINVAL;
640  	return tlen + len;
641  }
642  
643  /*
644   * write_versions - Set or report the available NFS protocol versions
645   *
646   * Input:
647   *			buf:		ignored
648   *			size:		zero
649   * Output:
650   *	On success:	passed-in buffer filled with '\n'-terminated C
651   *			string containing positive or negative integer
652   *			values representing the current status of each
653   *			protocol version;
654   *			return code is the size in bytes of the string
655   *	On error:	return code is zero or a negative errno value
656   *
657   * OR
658   *
659   * Input:
660   * 			buf:		C string containing whitespace-
661   * 					separated positive or negative
662   * 					integer values representing NFS
663   * 					protocol versions to enable ("+n")
664   * 					or disable ("-n")
665   *			size:		non-zero length of C string in @buf
666   * Output:
667   *	On success:	status of zero or more protocol versions has
668   *			been updated; passed-in buffer filled with
669   *			'\n'-terminated C string containing positive
670   *			or negative integer values representing the
671   *			current status of each protocol version;
672   *			return code is the size in bytes of the string
673   *	On error:	return code is zero or a negative errno value
674   */
675  static ssize_t write_versions(struct file *file, char *buf, size_t size)
676  {
677  	ssize_t rv;
678  
679  	mutex_lock(&nfsd_mutex);
680  	rv = __write_versions(file, buf, size);
681  	mutex_unlock(&nfsd_mutex);
682  	return rv;
683  }
684  
685  /*
686   * Zero-length write.  Return a list of NFSD's current listener
687   * transports.
688   */
689  static ssize_t __write_ports_names(char *buf, struct net *net)
690  {
691  	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
692  
693  	if (nn->nfsd_serv == NULL)
694  		return 0;
695  	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
696  }
697  
698  /*
699   * A single 'fd' number was written, in which case it must be for
700   * a socket of a supported family/protocol, and we use it as an
701   * nfsd listener.
702   */
703  static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
704  {
705  	char *mesg = buf;
706  	int fd, err;
707  	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
708  
709  	err = get_int(&mesg, &fd);
710  	if (err != 0 || fd < 0)
711  		return -EINVAL;
712  
713  	if (svc_alien_sock(net, fd)) {
714  		printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
715  		return -EINVAL;
716  	}
717  
718  	err = nfsd_create_serv(net);
719  	if (err != 0)
720  		return err;
721  
722  	err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
723  
724  	if (err >= 0 &&
725  	    !nn->nfsd_serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
726  		svc_get(nn->nfsd_serv);
727  
728  	nfsd_put(net);
729  	return err;
730  }
731  
732  /*
733   * A transport listener is added by writing it's transport name and
734   * a port number.
735   */
736  static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
737  {
738  	char transport[16];
739  	struct svc_xprt *xprt;
740  	int port, err;
741  	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
742  
743  	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
744  		return -EINVAL;
745  
746  	if (port < 1 || port > USHRT_MAX)
747  		return -EINVAL;
748  
749  	err = nfsd_create_serv(net);
750  	if (err != 0)
751  		return err;
752  
753  	err = svc_xprt_create(nn->nfsd_serv, transport, net,
754  			      PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
755  	if (err < 0)
756  		goto out_err;
757  
758  	err = svc_xprt_create(nn->nfsd_serv, transport, net,
759  			      PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
760  	if (err < 0 && err != -EAFNOSUPPORT)
761  		goto out_close;
762  
763  	if (!nn->nfsd_serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
764  		svc_get(nn->nfsd_serv);
765  
766  	nfsd_put(net);
767  	return 0;
768  out_close:
769  	xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
770  	if (xprt != NULL) {
771  		svc_xprt_close(xprt);
772  		svc_xprt_put(xprt);
773  	}
774  out_err:
775  	nfsd_put(net);
776  	return err;
777  }
778  
779  static ssize_t __write_ports(struct file *file, char *buf, size_t size,
780  			     struct net *net)
781  {
782  	if (size == 0)
783  		return __write_ports_names(buf, net);
784  
785  	if (isdigit(buf[0]))
786  		return __write_ports_addfd(buf, net, file->f_cred);
787  
788  	if (isalpha(buf[0]))
789  		return __write_ports_addxprt(buf, net, file->f_cred);
790  
791  	return -EINVAL;
792  }
793  
794  /*
795   * write_ports - Pass a socket file descriptor or transport name to listen on
796   *
797   * Input:
798   *			buf:		ignored
799   *			size:		zero
800   * Output:
801   *	On success:	passed-in buffer filled with a '\n'-terminated C
802   *			string containing a whitespace-separated list of
803   *			named NFSD listeners;
804   *			return code is the size in bytes of the string
805   *	On error:	return code is zero or a negative errno value
806   *
807   * OR
808   *
809   * Input:
810   *			buf:		C string containing an unsigned
811   *					integer value representing a bound
812   *					but unconnected socket that is to be
813   *					used as an NFSD listener; listen(3)
814   *					must be called for a SOCK_STREAM
815   *					socket, otherwise it is ignored
816   *			size:		non-zero length of C string in @buf
817   * Output:
818   *	On success:	NFS service is started;
819   *			passed-in buffer filled with a '\n'-terminated C
820   *			string containing a unique alphanumeric name of
821   *			the listener;
822   *			return code is the size in bytes of the string
823   *	On error:	return code is a negative errno value
824   *
825   * OR
826   *
827   * Input:
828   *			buf:		C string containing a transport
829   *					name and an unsigned integer value
830   *					representing the port to listen on,
831   *					separated by whitespace
832   *			size:		non-zero length of C string in @buf
833   * Output:
834   *	On success:	returns zero; NFS service is started
835   *	On error:	return code is a negative errno value
836   */
837  static ssize_t write_ports(struct file *file, char *buf, size_t size)
838  {
839  	ssize_t rv;
840  
841  	mutex_lock(&nfsd_mutex);
842  	rv = __write_ports(file, buf, size, netns(file));
843  	mutex_unlock(&nfsd_mutex);
844  	return rv;
845  }
846  
847  
848  int nfsd_max_blksize;
849  
850  /*
851   * write_maxblksize - Set or report the current NFS blksize
852   *
853   * Input:
854   *			buf:		ignored
855   *			size:		zero
856   *
857   * OR
858   *
859   * Input:
860   * 			buf:		C string containing an unsigned
861   * 					integer value representing the new
862   * 					NFS blksize
863   *			size:		non-zero length of C string in @buf
864   * Output:
865   *	On success:	passed-in buffer filled with '\n'-terminated C string
866   *			containing numeric value of the current NFS blksize
867   *			setting;
868   *			return code is the size in bytes of the string
869   *	On error:	return code is zero or a negative errno value
870   */
871  static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
872  {
873  	char *mesg = buf;
874  	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
875  
876  	if (size > 0) {
877  		int bsize;
878  		int rv = get_int(&mesg, &bsize);
879  		if (rv)
880  			return rv;
881  		/* force bsize into allowed range and
882  		 * required alignment.
883  		 */
884  		bsize = max_t(int, bsize, 1024);
885  		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
886  		bsize &= ~(1024-1);
887  		mutex_lock(&nfsd_mutex);
888  		if (nn->nfsd_serv) {
889  			mutex_unlock(&nfsd_mutex);
890  			return -EBUSY;
891  		}
892  		nfsd_max_blksize = bsize;
893  		mutex_unlock(&nfsd_mutex);
894  	}
895  
896  	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
897  							nfsd_max_blksize);
898  }
899  
900  /*
901   * write_maxconn - Set or report the current max number of connections
902   *
903   * Input:
904   *			buf:		ignored
905   *			size:		zero
906   * OR
907   *
908   * Input:
909   * 			buf:		C string containing an unsigned
910   * 					integer value representing the new
911   * 					number of max connections
912   *			size:		non-zero length of C string in @buf
913   * Output:
914   *	On success:	passed-in buffer filled with '\n'-terminated C string
915   *			containing numeric value of max_connections setting
916   *			for this net namespace;
917   *			return code is the size in bytes of the string
918   *	On error:	return code is zero or a negative errno value
919   */
920  static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
921  {
922  	char *mesg = buf;
923  	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
924  	unsigned int maxconn = nn->max_connections;
925  
926  	if (size > 0) {
927  		int rv = get_uint(&mesg, &maxconn);
928  
929  		if (rv)
930  			return rv;
931  		nn->max_connections = maxconn;
932  	}
933  
934  	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
935  }
936  
937  #ifdef CONFIG_NFSD_V4
938  static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
939  				  time64_t *time, struct nfsd_net *nn)
940  {
941  	char *mesg = buf;
942  	int rv, i;
943  
944  	if (size > 0) {
945  		if (nn->nfsd_serv)
946  			return -EBUSY;
947  		rv = get_int(&mesg, &i);
948  		if (rv)
949  			return rv;
950  		/*
951  		 * Some sanity checking.  We don't have a reason for
952  		 * these particular numbers, but problems with the
953  		 * extremes are:
954  		 *	- Too short: the briefest network outage may
955  		 *	  cause clients to lose all their locks.  Also,
956  		 *	  the frequent polling may be wasteful.
957  		 *	- Too long: do you really want reboot recovery
958  		 *	  to take more than an hour?  Or to make other
959  		 *	  clients wait an hour before being able to
960  		 *	  revoke a dead client's locks?
961  		 */
962  		if (i < 10 || i > 3600)
963  			return -EINVAL;
964  		*time = i;
965  	}
966  
967  	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
968  }
969  
970  static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
971  				time64_t *time, struct nfsd_net *nn)
972  {
973  	ssize_t rv;
974  
975  	mutex_lock(&nfsd_mutex);
976  	rv = __nfsd4_write_time(file, buf, size, time, nn);
977  	mutex_unlock(&nfsd_mutex);
978  	return rv;
979  }
980  
981  /*
982   * write_leasetime - Set or report the current NFSv4 lease time
983   *
984   * Input:
985   *			buf:		ignored
986   *			size:		zero
987   *
988   * OR
989   *
990   * Input:
991   *			buf:		C string containing an unsigned
992   *					integer value representing the new
993   *					NFSv4 lease expiry time
994   *			size:		non-zero length of C string in @buf
995   * Output:
996   *	On success:	passed-in buffer filled with '\n'-terminated C
997   *			string containing unsigned integer value of the
998   *			current lease expiry time;
999   *			return code is the size in bytes of the string
1000   *	On error:	return code is zero or a negative errno value
1001   */
1002  static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1003  {
1004  	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1005  	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1006  }
1007  
1008  /*
1009   * write_gracetime - Set or report current NFSv4 grace period time
1010   *
1011   * As above, but sets the time of the NFSv4 grace period.
1012   *
1013   * Note this should never be set to less than the *previous*
1014   * lease-period time, but we don't try to enforce this.  (In the common
1015   * case (a new boot), we don't know what the previous lease time was
1016   * anyway.)
1017   */
1018  static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1019  {
1020  	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1021  	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1022  }
1023  
1024  static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1025  				   struct nfsd_net *nn)
1026  {
1027  	char *mesg = buf;
1028  	char *recdir;
1029  	int len, status;
1030  
1031  	if (size > 0) {
1032  		if (nn->nfsd_serv)
1033  			return -EBUSY;
1034  		if (size > PATH_MAX || buf[size-1] != '\n')
1035  			return -EINVAL;
1036  		buf[size-1] = 0;
1037  
1038  		recdir = mesg;
1039  		len = qword_get(&mesg, recdir, size);
1040  		if (len <= 0)
1041  			return -EINVAL;
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  			nfsd4_end_grace(nn);
1116  			break;
1117  		default:
1118  			return -EINVAL;
1119  		}
1120  	}
1121  
1122  	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1123  			 nn->grace_ended ? 'Y' : 'N');
1124  }
1125  
1126  #endif
1127  
1128  /*----------------------------------------------------------------------------*/
1129  /*
1130   *	populating the filesystem.
1131   */
1132  
1133  /* Basically copying rpc_get_inode. */
1134  static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1135  {
1136  	struct inode *inode = new_inode(sb);
1137  	if (!inode)
1138  		return NULL;
1139  	/* Following advice from simple_fill_super documentation: */
1140  	inode->i_ino = iunique(sb, NFSD_MaxReserved);
1141  	inode->i_mode = mode;
1142  	inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1143  	switch (mode & S_IFMT) {
1144  	case S_IFDIR:
1145  		inode->i_fop = &simple_dir_operations;
1146  		inode->i_op = &simple_dir_inode_operations;
1147  		inc_nlink(inode);
1148  		break;
1149  	default:
1150  		break;
1151  	}
1152  	return inode;
1153  }
1154  
1155  static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1156  {
1157  	struct inode *inode;
1158  
1159  	inode = nfsd_get_inode(dir->i_sb, mode);
1160  	if (!inode)
1161  		return -ENOMEM;
1162  	if (ncl) {
1163  		inode->i_private = ncl;
1164  		kref_get(&ncl->cl_ref);
1165  	}
1166  	d_add(dentry, inode);
1167  	inc_nlink(dir);
1168  	fsnotify_mkdir(dir, dentry);
1169  	return 0;
1170  }
1171  
1172  static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1173  {
1174  	struct inode *dir = parent->d_inode;
1175  	struct dentry *dentry;
1176  	int ret = -ENOMEM;
1177  
1178  	inode_lock(dir);
1179  	dentry = d_alloc_name(parent, name);
1180  	if (!dentry)
1181  		goto out_err;
1182  	ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1183  	if (ret)
1184  		goto out_err;
1185  out:
1186  	inode_unlock(dir);
1187  	return dentry;
1188  out_err:
1189  	dput(dentry);
1190  	dentry = ERR_PTR(ret);
1191  	goto out;
1192  }
1193  
1194  static void clear_ncl(struct inode *inode)
1195  {
1196  	struct nfsdfs_client *ncl = inode->i_private;
1197  
1198  	inode->i_private = NULL;
1199  	kref_put(&ncl->cl_ref, ncl->cl_release);
1200  }
1201  
1202  static struct nfsdfs_client *__get_nfsdfs_client(struct inode *inode)
1203  {
1204  	struct nfsdfs_client *nc = inode->i_private;
1205  
1206  	if (nc)
1207  		kref_get(&nc->cl_ref);
1208  	return nc;
1209  }
1210  
1211  struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1212  {
1213  	struct nfsdfs_client *nc;
1214  
1215  	inode_lock_shared(inode);
1216  	nc = __get_nfsdfs_client(inode);
1217  	inode_unlock_shared(inode);
1218  	return nc;
1219  }
1220  /* from __rpc_unlink */
1221  static void nfsdfs_remove_file(struct inode *dir, struct dentry *dentry)
1222  {
1223  	int ret;
1224  
1225  	clear_ncl(d_inode(dentry));
1226  	dget(dentry);
1227  	ret = simple_unlink(dir, dentry);
1228  	d_drop(dentry);
1229  	fsnotify_unlink(dir, dentry);
1230  	dput(dentry);
1231  	WARN_ON_ONCE(ret);
1232  }
1233  
1234  static void nfsdfs_remove_files(struct dentry *root)
1235  {
1236  	struct dentry *dentry, *tmp;
1237  
1238  	list_for_each_entry_safe(dentry, tmp, &root->d_subdirs, d_child) {
1239  		if (!simple_positive(dentry)) {
1240  			WARN_ON_ONCE(1); /* I think this can't happen? */
1241  			continue;
1242  		}
1243  		nfsdfs_remove_file(d_inode(root), dentry);
1244  	}
1245  }
1246  
1247  /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1248   * code instead. */
1249  static  int nfsdfs_create_files(struct dentry *root,
1250  				const struct tree_descr *files,
1251  				struct dentry **fdentries)
1252  {
1253  	struct inode *dir = d_inode(root);
1254  	struct inode *inode;
1255  	struct dentry *dentry;
1256  	int i;
1257  
1258  	inode_lock(dir);
1259  	for (i = 0; files->name && files->name[0]; i++, files++) {
1260  		dentry = d_alloc_name(root, files->name);
1261  		if (!dentry)
1262  			goto out;
1263  		inode = nfsd_get_inode(d_inode(root)->i_sb,
1264  					S_IFREG | files->mode);
1265  		if (!inode) {
1266  			dput(dentry);
1267  			goto out;
1268  		}
1269  		inode->i_fop = files->ops;
1270  		inode->i_private = __get_nfsdfs_client(dir);
1271  		d_add(dentry, inode);
1272  		fsnotify_create(dir, dentry);
1273  		if (fdentries)
1274  			fdentries[i] = dentry;
1275  	}
1276  	inode_unlock(dir);
1277  	return 0;
1278  out:
1279  	nfsdfs_remove_files(root);
1280  	inode_unlock(dir);
1281  	return -ENOMEM;
1282  }
1283  
1284  /* on success, returns positive number unique to that client. */
1285  struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1286  				 struct nfsdfs_client *ncl, u32 id,
1287  				 const struct tree_descr *files,
1288  				 struct dentry **fdentries)
1289  {
1290  	struct dentry *dentry;
1291  	char name[11];
1292  	int ret;
1293  
1294  	sprintf(name, "%u", id);
1295  
1296  	dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1297  	if (IS_ERR(dentry)) /* XXX: tossing errors? */
1298  		return NULL;
1299  	ret = nfsdfs_create_files(dentry, files, fdentries);
1300  	if (ret) {
1301  		nfsd_client_rmdir(dentry);
1302  		return NULL;
1303  	}
1304  	return dentry;
1305  }
1306  
1307  /* Taken from __rpc_rmdir: */
1308  void nfsd_client_rmdir(struct dentry *dentry)
1309  {
1310  	struct inode *dir = d_inode(dentry->d_parent);
1311  	struct inode *inode = d_inode(dentry);
1312  	int ret;
1313  
1314  	inode_lock(dir);
1315  	nfsdfs_remove_files(dentry);
1316  	clear_ncl(inode);
1317  	dget(dentry);
1318  	ret = simple_rmdir(dir, dentry);
1319  	WARN_ON_ONCE(ret);
1320  	d_drop(dentry);
1321  	fsnotify_rmdir(dir, dentry);
1322  	dput(dentry);
1323  	inode_unlock(dir);
1324  }
1325  
1326  static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1327  {
1328  	struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1329  							nfsd_net_id);
1330  	struct dentry *dentry;
1331  	int ret;
1332  
1333  	static const struct tree_descr nfsd_files[] = {
1334  		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1335  		/* Per-export io stats use same ops as exports file */
1336  		[NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1337  		[NFSD_Export_features] = {"export_features",
1338  					&export_features_fops, S_IRUGO},
1339  		[NFSD_FO_UnlockIP] = {"unlock_ip",
1340  					&transaction_ops, S_IWUSR|S_IRUSR},
1341  		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1342  					&transaction_ops, S_IWUSR|S_IRUSR},
1343  		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1344  		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1345  		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1346  		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1347  		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1348  					&nfsd_reply_cache_stats_fops, S_IRUGO},
1349  		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1350  		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1351  		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1352  		[NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1353  		[NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
1354  #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1355  		[NFSD_SupportedEnctypes] = {"supported_krb5_enctypes",
1356  					&supported_enctypes_fops, S_IRUGO},
1357  #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1358  #ifdef CONFIG_NFSD_V4
1359  		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1360  		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1361  		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1362  		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1363  #endif
1364  		/* last one */ {""}
1365  	};
1366  
1367  	ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1368  	if (ret)
1369  		return ret;
1370  	dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1371  	if (IS_ERR(dentry))
1372  		return PTR_ERR(dentry);
1373  	nn->nfsd_client_dir = dentry;
1374  	return 0;
1375  }
1376  
1377  static int nfsd_fs_get_tree(struct fs_context *fc)
1378  {
1379  	return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1380  }
1381  
1382  static void nfsd_fs_free_fc(struct fs_context *fc)
1383  {
1384  	if (fc->s_fs_info)
1385  		put_net(fc->s_fs_info);
1386  }
1387  
1388  static const struct fs_context_operations nfsd_fs_context_ops = {
1389  	.free		= nfsd_fs_free_fc,
1390  	.get_tree	= nfsd_fs_get_tree,
1391  };
1392  
1393  static int nfsd_init_fs_context(struct fs_context *fc)
1394  {
1395  	put_user_ns(fc->user_ns);
1396  	fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1397  	fc->ops = &nfsd_fs_context_ops;
1398  	return 0;
1399  }
1400  
1401  static void nfsd_umount(struct super_block *sb)
1402  {
1403  	struct net *net = sb->s_fs_info;
1404  
1405  	nfsd_shutdown_threads(net);
1406  
1407  	kill_litter_super(sb);
1408  	put_net(net);
1409  }
1410  
1411  static struct file_system_type nfsd_fs_type = {
1412  	.owner		= THIS_MODULE,
1413  	.name		= "nfsd",
1414  	.init_fs_context = nfsd_init_fs_context,
1415  	.kill_sb	= nfsd_umount,
1416  };
1417  MODULE_ALIAS_FS("nfsd");
1418  
1419  #ifdef CONFIG_PROC_FS
1420  static int create_proc_exports_entry(void)
1421  {
1422  	struct proc_dir_entry *entry;
1423  
1424  	entry = proc_mkdir("fs/nfs", NULL);
1425  	if (!entry)
1426  		return -ENOMEM;
1427  	entry = proc_create("exports", 0, entry, &exports_proc_ops);
1428  	if (!entry) {
1429  		remove_proc_entry("fs/nfs", NULL);
1430  		return -ENOMEM;
1431  	}
1432  	return 0;
1433  }
1434  #else /* CONFIG_PROC_FS */
1435  static int create_proc_exports_entry(void)
1436  {
1437  	return 0;
1438  }
1439  #endif
1440  
1441  unsigned int nfsd_net_id;
1442  
1443  static __net_init int nfsd_init_net(struct net *net)
1444  {
1445  	int retval;
1446  	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1447  
1448  	retval = nfsd_export_init(net);
1449  	if (retval)
1450  		goto out_export_error;
1451  	retval = nfsd_idmap_init(net);
1452  	if (retval)
1453  		goto out_idmap_error;
1454  	nn->nfsd_versions = NULL;
1455  	nn->nfsd4_minorversions = NULL;
1456  	retval = nfsd4_init_leases_net(nn);
1457  	if (retval)
1458  		goto out_drc_error;
1459  	retval = nfsd_reply_cache_init(nn);
1460  	if (retval)
1461  		goto out_cache_error;
1462  	get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
1463  	seqlock_init(&nn->writeverf_lock);
1464  
1465  	return 0;
1466  
1467  out_cache_error:
1468  	nfsd4_leases_net_shutdown(nn);
1469  out_drc_error:
1470  	nfsd_idmap_shutdown(net);
1471  out_idmap_error:
1472  	nfsd_export_shutdown(net);
1473  out_export_error:
1474  	return retval;
1475  }
1476  
1477  static __net_exit void nfsd_exit_net(struct net *net)
1478  {
1479  	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1480  
1481  	nfsd_reply_cache_shutdown(nn);
1482  	nfsd_idmap_shutdown(net);
1483  	nfsd_export_shutdown(net);
1484  	nfsd_netns_free_versions(net_generic(net, nfsd_net_id));
1485  	nfsd4_leases_net_shutdown(nn);
1486  }
1487  
1488  static struct pernet_operations nfsd_net_ops = {
1489  	.init = nfsd_init_net,
1490  	.exit = nfsd_exit_net,
1491  	.id   = &nfsd_net_id,
1492  	.size = sizeof(struct nfsd_net),
1493  };
1494  
1495  static int __init init_nfsd(void)
1496  {
1497  	int retval;
1498  
1499  	retval = nfsd4_init_slabs();
1500  	if (retval)
1501  		return retval;
1502  	retval = nfsd4_init_pnfs();
1503  	if (retval)
1504  		goto out_free_slabs;
1505  	retval = nfsd_stat_init();	/* Statistics */
1506  	if (retval)
1507  		goto out_free_pnfs;
1508  	retval = nfsd_drc_slab_create();
1509  	if (retval)
1510  		goto out_free_stat;
1511  	nfsd_lockd_init();	/* lockd->nfsd callbacks */
1512  	retval = create_proc_exports_entry();
1513  	if (retval)
1514  		goto out_free_lockd;
1515  	retval = register_pernet_subsys(&nfsd_net_ops);
1516  	if (retval < 0)
1517  		goto out_free_exports;
1518  	retval = register_cld_notifier();
1519  	if (retval)
1520  		goto out_free_subsys;
1521  	retval = nfsd4_create_laundry_wq();
1522  	if (retval)
1523  		goto out_free_cld;
1524  	retval = register_filesystem(&nfsd_fs_type);
1525  	if (retval)
1526  		goto out_free_all;
1527  	return 0;
1528  out_free_all:
1529  	nfsd4_destroy_laundry_wq();
1530  out_free_cld:
1531  	unregister_cld_notifier();
1532  out_free_subsys:
1533  	unregister_pernet_subsys(&nfsd_net_ops);
1534  out_free_exports:
1535  	remove_proc_entry("fs/nfs/exports", NULL);
1536  	remove_proc_entry("fs/nfs", NULL);
1537  out_free_lockd:
1538  	nfsd_lockd_shutdown();
1539  	nfsd_drc_slab_free();
1540  out_free_stat:
1541  	nfsd_stat_shutdown();
1542  out_free_pnfs:
1543  	nfsd4_exit_pnfs();
1544  out_free_slabs:
1545  	nfsd4_free_slabs();
1546  	return retval;
1547  }
1548  
1549  static void __exit exit_nfsd(void)
1550  {
1551  	unregister_filesystem(&nfsd_fs_type);
1552  	nfsd4_destroy_laundry_wq();
1553  	unregister_cld_notifier();
1554  	unregister_pernet_subsys(&nfsd_net_ops);
1555  	nfsd_drc_slab_free();
1556  	remove_proc_entry("fs/nfs/exports", NULL);
1557  	remove_proc_entry("fs/nfs", NULL);
1558  	nfsd_stat_shutdown();
1559  	nfsd_lockd_shutdown();
1560  	nfsd4_free_slabs();
1561  	nfsd4_exit_pnfs();
1562  }
1563  
1564  MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1565  MODULE_LICENSE("GPL");
1566  module_init(init_nfsd)
1567  module_exit(exit_nfsd)
1568