xref: /linux/fs/nfsd/nfsctl.c (revision e21f9e2e862e9eb3dd64eaddb6256b3e5098660f)
1 /*
2  * Syscall interface to knfsd.
3  *
4  * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
5  */
6 
7 #include <linux/slab.h>
8 #include <linux/namei.h>
9 #include <linux/ctype.h>
10 
11 #include <linux/sunrpc/svcsock.h>
12 #include <linux/lockd/lockd.h>
13 #include <linux/sunrpc/addr.h>
14 #include <linux/sunrpc/gss_api.h>
15 #include <linux/sunrpc/gss_krb5_enctypes.h>
16 #include <linux/sunrpc/rpc_pipe_fs.h>
17 #include <linux/module.h>
18 
19 #include "idmap.h"
20 #include "nfsd.h"
21 #include "cache.h"
22 #include "state.h"
23 #include "netns.h"
24 #include "pnfs.h"
25 
26 /*
27  *	We have a single directory with several nodes in it.
28  */
29 enum {
30 	NFSD_Root = 1,
31 	NFSD_List,
32 	NFSD_Export_features,
33 	NFSD_Fh,
34 	NFSD_FO_UnlockIP,
35 	NFSD_FO_UnlockFS,
36 	NFSD_Threads,
37 	NFSD_Pool_Threads,
38 	NFSD_Pool_Stats,
39 	NFSD_Reply_Cache_Stats,
40 	NFSD_Versions,
41 	NFSD_Ports,
42 	NFSD_MaxBlkSize,
43 	NFSD_MaxConnections,
44 	NFSD_SupportedEnctypes,
45 	/*
46 	 * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
47 	 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
48 	 */
49 #ifdef CONFIG_NFSD_V4
50 	NFSD_Leasetime,
51 	NFSD_Gracetime,
52 	NFSD_RecoveryDir,
53 	NFSD_V4EndGrace,
54 #endif
55 };
56 
57 /*
58  * write() for these nodes.
59  */
60 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
61 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
62 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
63 static ssize_t write_threads(struct file *file, char *buf, size_t size);
64 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
65 static ssize_t write_versions(struct file *file, char *buf, size_t size);
66 static ssize_t write_ports(struct file *file, char *buf, size_t size);
67 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
68 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
69 #ifdef CONFIG_NFSD_V4
70 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
71 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
72 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
73 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
74 #endif
75 
76 static ssize_t (*write_op[])(struct file *, char *, size_t) = {
77 	[NFSD_Fh] = write_filehandle,
78 	[NFSD_FO_UnlockIP] = write_unlock_ip,
79 	[NFSD_FO_UnlockFS] = write_unlock_fs,
80 	[NFSD_Threads] = write_threads,
81 	[NFSD_Pool_Threads] = write_pool_threads,
82 	[NFSD_Versions] = write_versions,
83 	[NFSD_Ports] = write_ports,
84 	[NFSD_MaxBlkSize] = write_maxblksize,
85 	[NFSD_MaxConnections] = write_maxconn,
86 #ifdef CONFIG_NFSD_V4
87 	[NFSD_Leasetime] = write_leasetime,
88 	[NFSD_Gracetime] = write_gracetime,
89 	[NFSD_RecoveryDir] = write_recoverydir,
90 	[NFSD_V4EndGrace] = write_v4_end_grace,
91 #endif
92 };
93 
94 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
95 {
96 	ino_t ino =  file_inode(file)->i_ino;
97 	char *data;
98 	ssize_t rv;
99 
100 	if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
101 		return -EINVAL;
102 
103 	data = simple_transaction_get(file, buf, size);
104 	if (IS_ERR(data))
105 		return PTR_ERR(data);
106 
107 	rv =  write_op[ino](file, data, size);
108 	if (rv >= 0) {
109 		simple_transaction_set(file, rv);
110 		rv = size;
111 	}
112 	return rv;
113 }
114 
115 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
116 {
117 	if (! file->private_data) {
118 		/* An attempt to read a transaction file without writing
119 		 * causes a 0-byte write so that the file can return
120 		 * state information
121 		 */
122 		ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
123 		if (rv < 0)
124 			return rv;
125 	}
126 	return simple_transaction_read(file, buf, size, pos);
127 }
128 
129 static const struct file_operations transaction_ops = {
130 	.write		= nfsctl_transaction_write,
131 	.read		= nfsctl_transaction_read,
132 	.release	= simple_transaction_release,
133 	.llseek		= default_llseek,
134 };
135 
136 static int exports_net_open(struct net *net, struct file *file)
137 {
138 	int err;
139 	struct seq_file *seq;
140 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
141 
142 	err = seq_open(file, &nfs_exports_op);
143 	if (err)
144 		return err;
145 
146 	seq = file->private_data;
147 	seq->private = nn->svc_export_cache;
148 	return 0;
149 }
150 
151 static int exports_proc_open(struct inode *inode, struct file *file)
152 {
153 	return exports_net_open(current->nsproxy->net_ns, file);
154 }
155 
156 static const struct file_operations exports_proc_operations = {
157 	.open		= exports_proc_open,
158 	.read		= seq_read,
159 	.llseek		= seq_lseek,
160 	.release	= seq_release,
161 };
162 
163 static int exports_nfsd_open(struct inode *inode, struct file *file)
164 {
165 	return exports_net_open(inode->i_sb->s_fs_info, file);
166 }
167 
168 static const struct file_operations exports_nfsd_operations = {
169 	.open		= exports_nfsd_open,
170 	.read		= seq_read,
171 	.llseek		= seq_lseek,
172 	.release	= seq_release,
173 };
174 
175 static int export_features_show(struct seq_file *m, void *v)
176 {
177 	seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
178 	return 0;
179 }
180 
181 static int export_features_open(struct inode *inode, struct file *file)
182 {
183 	return single_open(file, export_features_show, NULL);
184 }
185 
186 static const struct file_operations export_features_operations = {
187 	.open		= export_features_open,
188 	.read		= seq_read,
189 	.llseek		= seq_lseek,
190 	.release	= single_release,
191 };
192 
193 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
194 static int supported_enctypes_show(struct seq_file *m, void *v)
195 {
196 	seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
197 	return 0;
198 }
199 
200 static int supported_enctypes_open(struct inode *inode, struct file *file)
201 {
202 	return single_open(file, supported_enctypes_show, NULL);
203 }
204 
205 static const struct file_operations supported_enctypes_ops = {
206 	.open		= supported_enctypes_open,
207 	.read		= seq_read,
208 	.llseek		= seq_lseek,
209 	.release	= single_release,
210 };
211 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
212 
213 static const struct file_operations pool_stats_operations = {
214 	.open		= nfsd_pool_stats_open,
215 	.read		= seq_read,
216 	.llseek		= seq_lseek,
217 	.release	= nfsd_pool_stats_release,
218 };
219 
220 static const struct file_operations reply_cache_stats_operations = {
221 	.open		= nfsd_reply_cache_stats_open,
222 	.read		= seq_read,
223 	.llseek		= seq_lseek,
224 	.release	= single_release,
225 };
226 
227 /*----------------------------------------------------------------------------*/
228 /*
229  * payload - write methods
230  */
231 
232 static inline struct net *netns(struct file *file)
233 {
234 	return file_inode(file)->i_sb->s_fs_info;
235 }
236 
237 /**
238  * write_unlock_ip - Release all locks used by a client
239  *
240  * Experimental.
241  *
242  * Input:
243  *			buf:	'\n'-terminated C string containing a
244  *				presentation format IP address
245  *			size:	length of C string in @buf
246  * Output:
247  *	On success:	returns zero if all specified locks were released;
248  *			returns one if one or more locks were not released
249  *	On error:	return code is negative errno value
250  */
251 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
252 {
253 	struct sockaddr_storage address;
254 	struct sockaddr *sap = (struct sockaddr *)&address;
255 	size_t salen = sizeof(address);
256 	char *fo_path;
257 	struct net *net = netns(file);
258 
259 	/* sanity check */
260 	if (size == 0)
261 		return -EINVAL;
262 
263 	if (buf[size-1] != '\n')
264 		return -EINVAL;
265 
266 	fo_path = buf;
267 	if (qword_get(&buf, fo_path, size) < 0)
268 		return -EINVAL;
269 
270 	if (rpc_pton(net, fo_path, size, sap, salen) == 0)
271 		return -EINVAL;
272 
273 	return nlmsvc_unlock_all_by_ip(sap);
274 }
275 
276 /**
277  * write_unlock_fs - Release all locks on a local file system
278  *
279  * Experimental.
280  *
281  * Input:
282  *			buf:	'\n'-terminated C string containing the
283  *				absolute pathname of a local file system
284  *			size:	length of C string in @buf
285  * Output:
286  *	On success:	returns zero if all specified locks were released;
287  *			returns one if one or more locks were not released
288  *	On error:	return code is negative errno value
289  */
290 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
291 {
292 	struct path path;
293 	char *fo_path;
294 	int error;
295 
296 	/* sanity check */
297 	if (size == 0)
298 		return -EINVAL;
299 
300 	if (buf[size-1] != '\n')
301 		return -EINVAL;
302 
303 	fo_path = buf;
304 	if (qword_get(&buf, fo_path, size) < 0)
305 		return -EINVAL;
306 
307 	error = kern_path(fo_path, 0, &path);
308 	if (error)
309 		return error;
310 
311 	/*
312 	 * XXX: Needs better sanity checking.  Otherwise we could end up
313 	 * releasing locks on the wrong file system.
314 	 *
315 	 * For example:
316 	 * 1.  Does the path refer to a directory?
317 	 * 2.  Is that directory a mount point, or
318 	 * 3.  Is that directory the root of an exported file system?
319 	 */
320 	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
321 
322 	path_put(&path);
323 	return error;
324 }
325 
326 /**
327  * write_filehandle - Get a variable-length NFS file handle by path
328  *
329  * On input, the buffer contains a '\n'-terminated C string comprised of
330  * three alphanumeric words separated by whitespace.  The string may
331  * contain escape sequences.
332  *
333  * Input:
334  *			buf:
335  *				domain:		client domain name
336  *				path:		export pathname
337  *				maxsize:	numeric maximum size of
338  *						@buf
339  *			size:	length of C string in @buf
340  * Output:
341  *	On success:	passed-in buffer filled with '\n'-terminated C
342  *			string containing a ASCII hex text version
343  *			of the NFS file handle;
344  *			return code is the size in bytes of the string
345  *	On error:	return code is negative errno value
346  */
347 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
348 {
349 	char *dname, *path;
350 	int uninitialized_var(maxsize);
351 	char *mesg = buf;
352 	int len;
353 	struct auth_domain *dom;
354 	struct knfsd_fh fh;
355 
356 	if (size == 0)
357 		return -EINVAL;
358 
359 	if (buf[size-1] != '\n')
360 		return -EINVAL;
361 	buf[size-1] = 0;
362 
363 	dname = mesg;
364 	len = qword_get(&mesg, dname, size);
365 	if (len <= 0)
366 		return -EINVAL;
367 
368 	path = dname+len+1;
369 	len = qword_get(&mesg, path, size);
370 	if (len <= 0)
371 		return -EINVAL;
372 
373 	len = get_int(&mesg, &maxsize);
374 	if (len)
375 		return len;
376 
377 	if (maxsize < NFS_FHSIZE)
378 		return -EINVAL;
379 	maxsize = min(maxsize, NFS3_FHSIZE);
380 
381 	if (qword_get(&mesg, mesg, size)>0)
382 		return -EINVAL;
383 
384 	/* we have all the words, they are in buf.. */
385 	dom = unix_domain_find(dname);
386 	if (!dom)
387 		return -ENOMEM;
388 
389 	len = exp_rootfh(netns(file), dom, path, &fh,  maxsize);
390 	auth_domain_put(dom);
391 	if (len)
392 		return len;
393 
394 	mesg = buf;
395 	len = SIMPLE_TRANSACTION_LIMIT;
396 	qword_addhex(&mesg, &len, (char*)&fh.fh_base, fh.fh_size);
397 	mesg[-1] = '\n';
398 	return mesg - buf;
399 }
400 
401 /**
402  * write_threads - Start NFSD, or report the current number of running threads
403  *
404  * Input:
405  *			buf:		ignored
406  *			size:		zero
407  * Output:
408  *	On success:	passed-in buffer filled with '\n'-terminated C
409  *			string numeric value representing the number of
410  *			running NFSD threads;
411  *			return code is the size in bytes of the string
412  *	On error:	return code is zero
413  *
414  * OR
415  *
416  * Input:
417  *			buf:		C string containing an unsigned
418  *					integer value representing the
419  *					number of NFSD threads to start
420  *			size:		non-zero length of C string in @buf
421  * Output:
422  *	On success:	NFS service is started;
423  *			passed-in buffer filled with '\n'-terminated C
424  *			string numeric value representing the number of
425  *			running NFSD threads;
426  *			return code is the size in bytes of the string
427  *	On error:	return code is zero or a negative errno value
428  */
429 static ssize_t write_threads(struct file *file, char *buf, size_t size)
430 {
431 	char *mesg = buf;
432 	int rv;
433 	struct net *net = netns(file);
434 
435 	if (size > 0) {
436 		int newthreads;
437 		rv = get_int(&mesg, &newthreads);
438 		if (rv)
439 			return rv;
440 		if (newthreads < 0)
441 			return -EINVAL;
442 		rv = nfsd_svc(newthreads, net);
443 		if (rv < 0)
444 			return rv;
445 	} else
446 		rv = nfsd_nrthreads(net);
447 
448 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
449 }
450 
451 /**
452  * write_pool_threads - Set or report the current number of threads per pool
453  *
454  * Input:
455  *			buf:		ignored
456  *			size:		zero
457  *
458  * OR
459  *
460  * Input:
461  * 			buf:		C string containing whitespace-
462  * 					separated unsigned integer values
463  *					representing the number of NFSD
464  *					threads to start in each pool
465  *			size:		non-zero length of C string in @buf
466  * Output:
467  *	On success:	passed-in buffer filled with '\n'-terminated C
468  *			string containing integer values representing the
469  *			number of NFSD threads in each pool;
470  *			return code is the size in bytes of the string
471  *	On error:	return code is zero or a negative errno value
472  */
473 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
474 {
475 	/* if size > 0, look for an array of number of threads per node
476 	 * and apply them  then write out number of threads per node as reply
477 	 */
478 	char *mesg = buf;
479 	int i;
480 	int rv;
481 	int len;
482 	int npools;
483 	int *nthreads;
484 	struct net *net = netns(file);
485 
486 	mutex_lock(&nfsd_mutex);
487 	npools = nfsd_nrpools(net);
488 	if (npools == 0) {
489 		/*
490 		 * NFS is shut down.  The admin can start it by
491 		 * writing to the threads file but NOT the pool_threads
492 		 * file, sorry.  Report zero threads.
493 		 */
494 		mutex_unlock(&nfsd_mutex);
495 		strcpy(buf, "0\n");
496 		return strlen(buf);
497 	}
498 
499 	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
500 	rv = -ENOMEM;
501 	if (nthreads == NULL)
502 		goto out_free;
503 
504 	if (size > 0) {
505 		for (i = 0; i < npools; i++) {
506 			rv = get_int(&mesg, &nthreads[i]);
507 			if (rv == -ENOENT)
508 				break;		/* fewer numbers than pools */
509 			if (rv)
510 				goto out_free;	/* syntax error */
511 			rv = -EINVAL;
512 			if (nthreads[i] < 0)
513 				goto out_free;
514 		}
515 		rv = nfsd_set_nrthreads(i, nthreads, net);
516 		if (rv)
517 			goto out_free;
518 	}
519 
520 	rv = nfsd_get_nrthreads(npools, nthreads, net);
521 	if (rv)
522 		goto out_free;
523 
524 	mesg = buf;
525 	size = SIMPLE_TRANSACTION_LIMIT;
526 	for (i = 0; i < npools && size > 0; i++) {
527 		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
528 		len = strlen(mesg);
529 		size -= len;
530 		mesg += len;
531 	}
532 	rv = mesg - buf;
533 out_free:
534 	kfree(nthreads);
535 	mutex_unlock(&nfsd_mutex);
536 	return rv;
537 }
538 
539 static ssize_t
540 nfsd_print_version_support(char *buf, int remaining, const char *sep,
541 		unsigned vers, int minor)
542 {
543 	const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
544 	bool supported = !!nfsd_vers(vers, NFSD_TEST);
545 
546 	if (vers == 4 && minor >= 0 &&
547 	    !nfsd_minorversion(minor, NFSD_TEST))
548 		supported = false;
549 	if (minor == 0 && supported)
550 		/*
551 		 * special case for backward compatability.
552 		 * +4.0 is never reported, it is implied by
553 		 * +4, unless -4.0 is present.
554 		 */
555 		return 0;
556 	return snprintf(buf, remaining, format, sep,
557 			supported ? '+' : '-', vers, minor);
558 }
559 
560 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
561 {
562 	char *mesg = buf;
563 	char *vers, *minorp, sign;
564 	int len, num, remaining;
565 	ssize_t tlen = 0;
566 	char *sep;
567 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
568 
569 	if (size>0) {
570 		if (nn->nfsd_serv)
571 			/* Cannot change versions without updating
572 			 * nn->nfsd_serv->sv_xdrsize, and reallocing
573 			 * rq_argp and rq_resp
574 			 */
575 			return -EBUSY;
576 		if (buf[size-1] != '\n')
577 			return -EINVAL;
578 		buf[size-1] = 0;
579 
580 		vers = mesg;
581 		len = qword_get(&mesg, vers, size);
582 		if (len <= 0) return -EINVAL;
583 		do {
584 			enum vers_op cmd;
585 			unsigned minor;
586 			sign = *vers;
587 			if (sign == '+' || sign == '-')
588 				num = simple_strtol((vers+1), &minorp, 0);
589 			else
590 				num = simple_strtol(vers, &minorp, 0);
591 			if (*minorp == '.') {
592 				if (num != 4)
593 					return -EINVAL;
594 				if (kstrtouint(minorp+1, 0, &minor) < 0)
595 					return -EINVAL;
596 			}
597 
598 			cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
599 			switch(num) {
600 			case 2:
601 			case 3:
602 				nfsd_vers(num, cmd);
603 				break;
604 			case 4:
605 				if (*minorp == '.') {
606 					if (nfsd_minorversion(minor, cmd) < 0)
607 						return -EINVAL;
608 				} else if ((cmd == NFSD_SET) != nfsd_vers(num, NFSD_TEST)) {
609 					/*
610 					 * Either we have +4 and no minors are enabled,
611 					 * or we have -4 and at least one minor is enabled.
612 					 * In either case, propagate 'cmd' to all minors.
613 					 */
614 					minor = 0;
615 					while (nfsd_minorversion(minor, cmd) >= 0)
616 						minor++;
617 				}
618 				break;
619 			default:
620 				return -EINVAL;
621 			}
622 			vers += len + 1;
623 		} while ((len = qword_get(&mesg, vers, size)) > 0);
624 		/* If all get turned off, turn them back on, as
625 		 * having no versions is BAD
626 		 */
627 		nfsd_reset_versions();
628 	}
629 
630 	/* Now write current state into reply buffer */
631 	len = 0;
632 	sep = "";
633 	remaining = SIMPLE_TRANSACTION_LIMIT;
634 	for (num=2 ; num <= 4 ; num++) {
635 		int minor;
636 		if (!nfsd_vers(num, NFSD_AVAIL))
637 			continue;
638 
639 		minor = -1;
640 		do {
641 			len = nfsd_print_version_support(buf, remaining,
642 					sep, num, minor);
643 			if (len >= remaining)
644 				goto out;
645 			remaining -= len;
646 			buf += len;
647 			tlen += len;
648 			minor++;
649 			if (len)
650 				sep = " ";
651 		} while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
652 	}
653 out:
654 	len = snprintf(buf, remaining, "\n");
655 	if (len >= remaining)
656 		return -EINVAL;
657 	return tlen + len;
658 }
659 
660 /**
661  * write_versions - Set or report the available NFS protocol versions
662  *
663  * Input:
664  *			buf:		ignored
665  *			size:		zero
666  * Output:
667  *	On success:	passed-in buffer filled with '\n'-terminated C
668  *			string containing positive or negative integer
669  *			values representing the current status of each
670  *			protocol version;
671  *			return code is the size in bytes of the string
672  *	On error:	return code is zero or a negative errno value
673  *
674  * OR
675  *
676  * Input:
677  * 			buf:		C string containing whitespace-
678  * 					separated positive or negative
679  * 					integer values representing NFS
680  * 					protocol versions to enable ("+n")
681  * 					or disable ("-n")
682  *			size:		non-zero length of C string in @buf
683  * Output:
684  *	On success:	status of zero or more protocol versions has
685  *			been updated; passed-in buffer filled with
686  *			'\n'-terminated C string containing positive
687  *			or negative integer values representing the
688  *			current status of each protocol version;
689  *			return code is the size in bytes of the string
690  *	On error:	return code is zero or a negative errno value
691  */
692 static ssize_t write_versions(struct file *file, char *buf, size_t size)
693 {
694 	ssize_t rv;
695 
696 	mutex_lock(&nfsd_mutex);
697 	rv = __write_versions(file, buf, size);
698 	mutex_unlock(&nfsd_mutex);
699 	return rv;
700 }
701 
702 /*
703  * Zero-length write.  Return a list of NFSD's current listener
704  * transports.
705  */
706 static ssize_t __write_ports_names(char *buf, struct net *net)
707 {
708 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
709 
710 	if (nn->nfsd_serv == NULL)
711 		return 0;
712 	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
713 }
714 
715 /*
716  * A single 'fd' number was written, in which case it must be for
717  * a socket of a supported family/protocol, and we use it as an
718  * nfsd listener.
719  */
720 static ssize_t __write_ports_addfd(char *buf, struct net *net)
721 {
722 	char *mesg = buf;
723 	int fd, err;
724 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
725 
726 	err = get_int(&mesg, &fd);
727 	if (err != 0 || fd < 0)
728 		return -EINVAL;
729 
730 	if (svc_alien_sock(net, fd)) {
731 		printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
732 		return -EINVAL;
733 	}
734 
735 	err = nfsd_create_serv(net);
736 	if (err != 0)
737 		return err;
738 
739 	err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT);
740 	if (err < 0) {
741 		nfsd_destroy(net);
742 		return err;
743 	}
744 
745 	/* Decrease the count, but don't shut down the service */
746 	nn->nfsd_serv->sv_nrthreads--;
747 	return err;
748 }
749 
750 /*
751  * A transport listener is added by writing it's transport name and
752  * a port number.
753  */
754 static ssize_t __write_ports_addxprt(char *buf, struct net *net)
755 {
756 	char transport[16];
757 	struct svc_xprt *xprt;
758 	int port, err;
759 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
760 
761 	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
762 		return -EINVAL;
763 
764 	if (port < 1 || port > USHRT_MAX)
765 		return -EINVAL;
766 
767 	err = nfsd_create_serv(net);
768 	if (err != 0)
769 		return err;
770 
771 	err = svc_create_xprt(nn->nfsd_serv, transport, net,
772 				PF_INET, port, SVC_SOCK_ANONYMOUS);
773 	if (err < 0)
774 		goto out_err;
775 
776 	err = svc_create_xprt(nn->nfsd_serv, transport, net,
777 				PF_INET6, port, SVC_SOCK_ANONYMOUS);
778 	if (err < 0 && err != -EAFNOSUPPORT)
779 		goto out_close;
780 
781 	/* Decrease the count, but don't shut down the service */
782 	nn->nfsd_serv->sv_nrthreads--;
783 	return 0;
784 out_close:
785 	xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
786 	if (xprt != NULL) {
787 		svc_close_xprt(xprt);
788 		svc_xprt_put(xprt);
789 	}
790 out_err:
791 	nfsd_destroy(net);
792 	return err;
793 }
794 
795 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
796 			     struct net *net)
797 {
798 	if (size == 0)
799 		return __write_ports_names(buf, net);
800 
801 	if (isdigit(buf[0]))
802 		return __write_ports_addfd(buf, net);
803 
804 	if (isalpha(buf[0]))
805 		return __write_ports_addxprt(buf, net);
806 
807 	return -EINVAL;
808 }
809 
810 /**
811  * write_ports - Pass a socket file descriptor or transport name to listen on
812  *
813  * Input:
814  *			buf:		ignored
815  *			size:		zero
816  * Output:
817  *	On success:	passed-in buffer filled with a '\n'-terminated C
818  *			string containing a whitespace-separated list of
819  *			named NFSD listeners;
820  *			return code is the size in bytes of the string
821  *	On error:	return code is zero or a negative errno value
822  *
823  * OR
824  *
825  * Input:
826  *			buf:		C string containing an unsigned
827  *					integer value representing a bound
828  *					but unconnected socket that is to be
829  *					used as an NFSD listener; listen(3)
830  *					must be called for a SOCK_STREAM
831  *					socket, otherwise it is ignored
832  *			size:		non-zero length of C string in @buf
833  * Output:
834  *	On success:	NFS service is started;
835  *			passed-in buffer filled with a '\n'-terminated C
836  *			string containing a unique alphanumeric name of
837  *			the listener;
838  *			return code is the size in bytes of the string
839  *	On error:	return code is a negative errno value
840  *
841  * OR
842  *
843  * Input:
844  *			buf:		C string containing a transport
845  *					name and an unsigned integer value
846  *					representing the port to listen on,
847  *					separated by whitespace
848  *			size:		non-zero length of C string in @buf
849  * Output:
850  *	On success:	returns zero; NFS service is started
851  *	On error:	return code is a negative errno value
852  */
853 static ssize_t write_ports(struct file *file, char *buf, size_t size)
854 {
855 	ssize_t rv;
856 
857 	mutex_lock(&nfsd_mutex);
858 	rv = __write_ports(file, buf, size, netns(file));
859 	mutex_unlock(&nfsd_mutex);
860 	return rv;
861 }
862 
863 
864 int nfsd_max_blksize;
865 
866 /**
867  * write_maxblksize - Set or report the current NFS blksize
868  *
869  * Input:
870  *			buf:		ignored
871  *			size:		zero
872  *
873  * OR
874  *
875  * Input:
876  * 			buf:		C string containing an unsigned
877  * 					integer value representing the new
878  * 					NFS blksize
879  *			size:		non-zero length of C string in @buf
880  * Output:
881  *	On success:	passed-in buffer filled with '\n'-terminated C string
882  *			containing numeric value of the current NFS blksize
883  *			setting;
884  *			return code is the size in bytes of the string
885  *	On error:	return code is zero or a negative errno value
886  */
887 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
888 {
889 	char *mesg = buf;
890 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
891 
892 	if (size > 0) {
893 		int bsize;
894 		int rv = get_int(&mesg, &bsize);
895 		if (rv)
896 			return rv;
897 		/* force bsize into allowed range and
898 		 * required alignment.
899 		 */
900 		bsize = max_t(int, bsize, 1024);
901 		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
902 		bsize &= ~(1024-1);
903 		mutex_lock(&nfsd_mutex);
904 		if (nn->nfsd_serv) {
905 			mutex_unlock(&nfsd_mutex);
906 			return -EBUSY;
907 		}
908 		nfsd_max_blksize = bsize;
909 		mutex_unlock(&nfsd_mutex);
910 	}
911 
912 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
913 							nfsd_max_blksize);
914 }
915 
916 /**
917  * write_maxconn - Set or report the current max number of connections
918  *
919  * Input:
920  *			buf:		ignored
921  *			size:		zero
922  * OR
923  *
924  * Input:
925  * 			buf:		C string containing an unsigned
926  * 					integer value representing the new
927  * 					number of max connections
928  *			size:		non-zero length of C string in @buf
929  * Output:
930  *	On success:	passed-in buffer filled with '\n'-terminated C string
931  *			containing numeric value of max_connections setting
932  *			for this net namespace;
933  *			return code is the size in bytes of the string
934  *	On error:	return code is zero or a negative errno value
935  */
936 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
937 {
938 	char *mesg = buf;
939 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
940 	unsigned int maxconn = nn->max_connections;
941 
942 	if (size > 0) {
943 		int rv = get_uint(&mesg, &maxconn);
944 
945 		if (rv)
946 			return rv;
947 		nn->max_connections = maxconn;
948 	}
949 
950 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
951 }
952 
953 #ifdef CONFIG_NFSD_V4
954 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
955 				  time_t *time, struct nfsd_net *nn)
956 {
957 	char *mesg = buf;
958 	int rv, i;
959 
960 	if (size > 0) {
961 		if (nn->nfsd_serv)
962 			return -EBUSY;
963 		rv = get_int(&mesg, &i);
964 		if (rv)
965 			return rv;
966 		/*
967 		 * Some sanity checking.  We don't have a reason for
968 		 * these particular numbers, but problems with the
969 		 * extremes are:
970 		 *	- Too short: the briefest network outage may
971 		 *	  cause clients to lose all their locks.  Also,
972 		 *	  the frequent polling may be wasteful.
973 		 *	- Too long: do you really want reboot recovery
974 		 *	  to take more than an hour?  Or to make other
975 		 *	  clients wait an hour before being able to
976 		 *	  revoke a dead client's locks?
977 		 */
978 		if (i < 10 || i > 3600)
979 			return -EINVAL;
980 		*time = i;
981 	}
982 
983 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%ld\n", *time);
984 }
985 
986 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
987 				time_t *time, struct nfsd_net *nn)
988 {
989 	ssize_t rv;
990 
991 	mutex_lock(&nfsd_mutex);
992 	rv = __nfsd4_write_time(file, buf, size, time, nn);
993 	mutex_unlock(&nfsd_mutex);
994 	return rv;
995 }
996 
997 /**
998  * write_leasetime - Set or report the current NFSv4 lease time
999  *
1000  * Input:
1001  *			buf:		ignored
1002  *			size:		zero
1003  *
1004  * OR
1005  *
1006  * Input:
1007  *			buf:		C string containing an unsigned
1008  *					integer value representing the new
1009  *					NFSv4 lease expiry time
1010  *			size:		non-zero length of C string in @buf
1011  * Output:
1012  *	On success:	passed-in buffer filled with '\n'-terminated C
1013  *			string containing unsigned integer value of the
1014  *			current lease expiry time;
1015  *			return code is the size in bytes of the string
1016  *	On error:	return code is zero or a negative errno value
1017  */
1018 static ssize_t write_leasetime(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_lease, nn);
1022 }
1023 
1024 /**
1025  * write_gracetime - Set or report current NFSv4 grace period time
1026  *
1027  * As above, but sets the time of the NFSv4 grace period.
1028  *
1029  * Note this should never be set to less than the *previous*
1030  * lease-period time, but we don't try to enforce this.  (In the common
1031  * case (a new boot), we don't know what the previous lease time was
1032  * anyway.)
1033  */
1034 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1035 {
1036 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1037 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1038 }
1039 
1040 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1041 				   struct nfsd_net *nn)
1042 {
1043 	char *mesg = buf;
1044 	char *recdir;
1045 	int len, status;
1046 
1047 	if (size > 0) {
1048 		if (nn->nfsd_serv)
1049 			return -EBUSY;
1050 		if (size > PATH_MAX || buf[size-1] != '\n')
1051 			return -EINVAL;
1052 		buf[size-1] = 0;
1053 
1054 		recdir = mesg;
1055 		len = qword_get(&mesg, recdir, size);
1056 		if (len <= 0)
1057 			return -EINVAL;
1058 
1059 		status = nfs4_reset_recoverydir(recdir);
1060 		if (status)
1061 			return status;
1062 	}
1063 
1064 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1065 							nfs4_recoverydir());
1066 }
1067 
1068 /**
1069  * write_recoverydir - Set or report the pathname of the recovery directory
1070  *
1071  * Input:
1072  *			buf:		ignored
1073  *			size:		zero
1074  *
1075  * OR
1076  *
1077  * Input:
1078  *			buf:		C string containing the pathname
1079  *					of the directory on a local file
1080  *					system containing permanent NFSv4
1081  *					recovery data
1082  *			size:		non-zero length of C string in @buf
1083  * Output:
1084  *	On success:	passed-in buffer filled with '\n'-terminated C string
1085  *			containing the current recovery pathname setting;
1086  *			return code is the size in bytes of the string
1087  *	On error:	return code is zero or a negative errno value
1088  */
1089 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1090 {
1091 	ssize_t rv;
1092 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1093 
1094 	mutex_lock(&nfsd_mutex);
1095 	rv = __write_recoverydir(file, buf, size, nn);
1096 	mutex_unlock(&nfsd_mutex);
1097 	return rv;
1098 }
1099 
1100 /**
1101  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1102  *
1103  * Input:
1104  *			buf:		ignored
1105  *			size:		zero
1106  * OR
1107  *
1108  * Input:
1109  * 			buf:		any value
1110  *			size:		non-zero length of C string in @buf
1111  * Output:
1112  *			passed-in buffer filled with "Y" or "N" with a newline
1113  *			and NULL-terminated C string. This indicates whether
1114  *			the grace period has ended in the current net
1115  *			namespace. Return code is the size in bytes of the
1116  *			string. Writing a string that starts with 'Y', 'y', or
1117  *			'1' to the file will end the grace period for nfsd's v4
1118  *			lock manager.
1119  */
1120 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1121 {
1122 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1123 
1124 	if (size > 0) {
1125 		switch(buf[0]) {
1126 		case 'Y':
1127 		case 'y':
1128 		case '1':
1129 			nfsd4_end_grace(nn);
1130 			break;
1131 		default:
1132 			return -EINVAL;
1133 		}
1134 	}
1135 
1136 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1137 			 nn->grace_ended ? 'Y' : 'N');
1138 }
1139 
1140 #endif
1141 
1142 /*----------------------------------------------------------------------------*/
1143 /*
1144  *	populating the filesystem.
1145  */
1146 
1147 static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
1148 {
1149 	static const struct tree_descr nfsd_files[] = {
1150 		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1151 		[NFSD_Export_features] = {"export_features",
1152 					&export_features_operations, S_IRUGO},
1153 		[NFSD_FO_UnlockIP] = {"unlock_ip",
1154 					&transaction_ops, S_IWUSR|S_IRUSR},
1155 		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1156 					&transaction_ops, S_IWUSR|S_IRUSR},
1157 		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1158 		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1159 		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1160 		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1161 		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
1162 		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1163 		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1164 		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1165 		[NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1166 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1167 		[NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1168 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1169 #ifdef CONFIG_NFSD_V4
1170 		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1171 		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1172 		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1173 		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1174 #endif
1175 		/* last one */ {""}
1176 	};
1177 	get_net(sb->s_fs_info);
1178 	return simple_fill_super(sb, 0x6e667364, nfsd_files);
1179 }
1180 
1181 static struct dentry *nfsd_mount(struct file_system_type *fs_type,
1182 	int flags, const char *dev_name, void *data)
1183 {
1184 	struct net *net = current->nsproxy->net_ns;
1185 	return mount_ns(fs_type, flags, data, net, net->user_ns, nfsd_fill_super);
1186 }
1187 
1188 static void nfsd_umount(struct super_block *sb)
1189 {
1190 	struct net *net = sb->s_fs_info;
1191 
1192 	kill_litter_super(sb);
1193 	put_net(net);
1194 }
1195 
1196 static struct file_system_type nfsd_fs_type = {
1197 	.owner		= THIS_MODULE,
1198 	.name		= "nfsd",
1199 	.mount		= nfsd_mount,
1200 	.kill_sb	= nfsd_umount,
1201 };
1202 MODULE_ALIAS_FS("nfsd");
1203 
1204 #ifdef CONFIG_PROC_FS
1205 static int create_proc_exports_entry(void)
1206 {
1207 	struct proc_dir_entry *entry;
1208 
1209 	entry = proc_mkdir("fs/nfs", NULL);
1210 	if (!entry)
1211 		return -ENOMEM;
1212 	entry = proc_create("exports", 0, entry,
1213 				 &exports_proc_operations);
1214 	if (!entry) {
1215 		remove_proc_entry("fs/nfs", NULL);
1216 		return -ENOMEM;
1217 	}
1218 	return 0;
1219 }
1220 #else /* CONFIG_PROC_FS */
1221 static int create_proc_exports_entry(void)
1222 {
1223 	return 0;
1224 }
1225 #endif
1226 
1227 unsigned int nfsd_net_id;
1228 
1229 static __net_init int nfsd_init_net(struct net *net)
1230 {
1231 	int retval;
1232 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1233 
1234 	retval = nfsd_export_init(net);
1235 	if (retval)
1236 		goto out_export_error;
1237 	retval = nfsd_idmap_init(net);
1238 	if (retval)
1239 		goto out_idmap_error;
1240 	nn->nfsd4_lease = 90;	/* default lease time */
1241 	nn->nfsd4_grace = 90;
1242 	nn->clverifier_counter = prandom_u32();
1243 	nn->clientid_counter = prandom_u32();
1244 
1245 	atomic_set(&nn->ntf_refcnt, 0);
1246 	init_waitqueue_head(&nn->ntf_wq);
1247 	return 0;
1248 
1249 out_idmap_error:
1250 	nfsd_export_shutdown(net);
1251 out_export_error:
1252 	return retval;
1253 }
1254 
1255 static __net_exit void nfsd_exit_net(struct net *net)
1256 {
1257 	nfsd_idmap_shutdown(net);
1258 	nfsd_export_shutdown(net);
1259 }
1260 
1261 static struct pernet_operations nfsd_net_ops = {
1262 	.init = nfsd_init_net,
1263 	.exit = nfsd_exit_net,
1264 	.id   = &nfsd_net_id,
1265 	.size = sizeof(struct nfsd_net),
1266 };
1267 
1268 static int __init init_nfsd(void)
1269 {
1270 	int retval;
1271 	printk(KERN_INFO "Installing knfsd (copyright (C) 1996 okir@monad.swb.de).\n");
1272 
1273 	retval = register_pernet_subsys(&nfsd_net_ops);
1274 	if (retval < 0)
1275 		return retval;
1276 	retval = register_cld_notifier();
1277 	if (retval)
1278 		goto out_unregister_pernet;
1279 	retval = nfsd4_init_slabs();
1280 	if (retval)
1281 		goto out_unregister_notifier;
1282 	retval = nfsd4_init_pnfs();
1283 	if (retval)
1284 		goto out_free_slabs;
1285 	retval = nfsd_fault_inject_init(); /* nfsd fault injection controls */
1286 	if (retval)
1287 		goto out_exit_pnfs;
1288 	nfsd_stat_init();	/* Statistics */
1289 	retval = nfsd_reply_cache_init();
1290 	if (retval)
1291 		goto out_free_stat;
1292 	nfsd_lockd_init();	/* lockd->nfsd callbacks */
1293 	retval = create_proc_exports_entry();
1294 	if (retval)
1295 		goto out_free_lockd;
1296 	retval = register_filesystem(&nfsd_fs_type);
1297 	if (retval)
1298 		goto out_free_all;
1299 	return 0;
1300 out_free_all:
1301 	remove_proc_entry("fs/nfs/exports", NULL);
1302 	remove_proc_entry("fs/nfs", NULL);
1303 out_free_lockd:
1304 	nfsd_lockd_shutdown();
1305 	nfsd_reply_cache_shutdown();
1306 out_free_stat:
1307 	nfsd_stat_shutdown();
1308 	nfsd_fault_inject_cleanup();
1309 out_exit_pnfs:
1310 	nfsd4_exit_pnfs();
1311 out_free_slabs:
1312 	nfsd4_free_slabs();
1313 out_unregister_notifier:
1314 	unregister_cld_notifier();
1315 out_unregister_pernet:
1316 	unregister_pernet_subsys(&nfsd_net_ops);
1317 	return retval;
1318 }
1319 
1320 static void __exit exit_nfsd(void)
1321 {
1322 	nfsd_reply_cache_shutdown();
1323 	remove_proc_entry("fs/nfs/exports", NULL);
1324 	remove_proc_entry("fs/nfs", NULL);
1325 	nfsd_stat_shutdown();
1326 	nfsd_lockd_shutdown();
1327 	nfsd4_free_slabs();
1328 	nfsd4_exit_pnfs();
1329 	nfsd_fault_inject_cleanup();
1330 	unregister_filesystem(&nfsd_fs_type);
1331 	unregister_cld_notifier();
1332 	unregister_pernet_subsys(&nfsd_net_ops);
1333 }
1334 
1335 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1336 MODULE_LICENSE("GPL");
1337 module_init(init_nfsd)
1338 module_exit(exit_nfsd)
1339