xref: /openbmc/linux/fs/nfsd/nfsctl.c (revision 3c6a73cc)
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 
25 /*
26  *	We have a single directory with several nodes in it.
27  */
28 enum {
29 	NFSD_Root = 1,
30 	NFSD_List,
31 	NFSD_Export_features,
32 	NFSD_Fh,
33 	NFSD_FO_UnlockIP,
34 	NFSD_FO_UnlockFS,
35 	NFSD_Threads,
36 	NFSD_Pool_Threads,
37 	NFSD_Pool_Stats,
38 	NFSD_Reply_Cache_Stats,
39 	NFSD_Versions,
40 	NFSD_Ports,
41 	NFSD_MaxBlkSize,
42 	NFSD_MaxConnections,
43 	NFSD_SupportedEnctypes,
44 	/*
45 	 * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
46 	 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
47 	 */
48 #ifdef CONFIG_NFSD_V4
49 	NFSD_Leasetime,
50 	NFSD_Gracetime,
51 	NFSD_RecoveryDir,
52 	NFSD_V4EndGrace,
53 #endif
54 };
55 
56 /*
57  * write() for these nodes.
58  */
59 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
60 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
61 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
62 static ssize_t write_threads(struct file *file, char *buf, size_t size);
63 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
64 static ssize_t write_versions(struct file *file, char *buf, size_t size);
65 static ssize_t write_ports(struct file *file, char *buf, size_t size);
66 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
67 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
68 #ifdef CONFIG_NFSD_V4
69 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
70 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
71 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
72 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
73 #endif
74 
75 static ssize_t (*write_op[])(struct file *, char *, size_t) = {
76 	[NFSD_Fh] = write_filehandle,
77 	[NFSD_FO_UnlockIP] = write_unlock_ip,
78 	[NFSD_FO_UnlockFS] = write_unlock_fs,
79 	[NFSD_Threads] = write_threads,
80 	[NFSD_Pool_Threads] = write_pool_threads,
81 	[NFSD_Versions] = write_versions,
82 	[NFSD_Ports] = write_ports,
83 	[NFSD_MaxBlkSize] = write_maxblksize,
84 	[NFSD_MaxConnections] = write_maxconn,
85 #ifdef CONFIG_NFSD_V4
86 	[NFSD_Leasetime] = write_leasetime,
87 	[NFSD_Gracetime] = write_gracetime,
88 	[NFSD_RecoveryDir] = write_recoverydir,
89 	[NFSD_V4EndGrace] = write_v4_end_grace,
90 #endif
91 };
92 
93 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
94 {
95 	ino_t ino =  file_inode(file)->i_ino;
96 	char *data;
97 	ssize_t rv;
98 
99 	if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
100 		return -EINVAL;
101 
102 	data = simple_transaction_get(file, buf, size);
103 	if (IS_ERR(data))
104 		return PTR_ERR(data);
105 
106 	rv =  write_op[ino](file, data, size);
107 	if (rv >= 0) {
108 		simple_transaction_set(file, rv);
109 		rv = size;
110 	}
111 	return rv;
112 }
113 
114 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
115 {
116 	if (! file->private_data) {
117 		/* An attempt to read a transaction file without writing
118 		 * causes a 0-byte write so that the file can return
119 		 * state information
120 		 */
121 		ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
122 		if (rv < 0)
123 			return rv;
124 	}
125 	return simple_transaction_read(file, buf, size, pos);
126 }
127 
128 static const struct file_operations transaction_ops = {
129 	.write		= nfsctl_transaction_write,
130 	.read		= nfsctl_transaction_read,
131 	.release	= simple_transaction_release,
132 	.llseek		= default_llseek,
133 };
134 
135 static int exports_net_open(struct net *net, struct file *file)
136 {
137 	int err;
138 	struct seq_file *seq;
139 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
140 
141 	err = seq_open(file, &nfs_exports_op);
142 	if (err)
143 		return err;
144 
145 	seq = file->private_data;
146 	seq->private = nn->svc_export_cache;
147 	return 0;
148 }
149 
150 static int exports_proc_open(struct inode *inode, struct file *file)
151 {
152 	return exports_net_open(current->nsproxy->net_ns, file);
153 }
154 
155 static const struct file_operations exports_proc_operations = {
156 	.open		= exports_proc_open,
157 	.read		= seq_read,
158 	.llseek		= seq_lseek,
159 	.release	= seq_release,
160 	.owner		= THIS_MODULE,
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 	.owner		= THIS_MODULE,
174 };
175 
176 static int export_features_show(struct seq_file *m, void *v)
177 {
178 	seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
179 	return 0;
180 }
181 
182 static int export_features_open(struct inode *inode, struct file *file)
183 {
184 	return single_open(file, export_features_show, NULL);
185 }
186 
187 static const struct file_operations export_features_operations = {
188 	.open		= export_features_open,
189 	.read		= seq_read,
190 	.llseek		= seq_lseek,
191 	.release	= single_release,
192 };
193 
194 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
195 static int supported_enctypes_show(struct seq_file *m, void *v)
196 {
197 	seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
198 	return 0;
199 }
200 
201 static int supported_enctypes_open(struct inode *inode, struct file *file)
202 {
203 	return single_open(file, supported_enctypes_show, NULL);
204 }
205 
206 static const struct file_operations supported_enctypes_ops = {
207 	.open		= supported_enctypes_open,
208 	.read		= seq_read,
209 	.llseek		= seq_lseek,
210 	.release	= single_release,
211 };
212 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
213 
214 static const struct file_operations pool_stats_operations = {
215 	.open		= nfsd_pool_stats_open,
216 	.read		= seq_read,
217 	.llseek		= seq_lseek,
218 	.release	= nfsd_pool_stats_release,
219 	.owner		= THIS_MODULE,
220 };
221 
222 static struct file_operations reply_cache_stats_operations = {
223 	.open		= nfsd_reply_cache_stats_open,
224 	.read		= seq_read,
225 	.llseek		= seq_lseek,
226 	.release	= single_release,
227 };
228 
229 /*----------------------------------------------------------------------------*/
230 /*
231  * payload - write methods
232  */
233 
234 
235 /**
236  * write_unlock_ip - Release all locks used by a client
237  *
238  * Experimental.
239  *
240  * Input:
241  *			buf:	'\n'-terminated C string containing a
242  *				presentation format IP address
243  *			size:	length of C string in @buf
244  * Output:
245  *	On success:	returns zero if all specified locks were released;
246  *			returns one if one or more locks were not released
247  *	On error:	return code is negative errno value
248  */
249 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
250 {
251 	struct sockaddr_storage address;
252 	struct sockaddr *sap = (struct sockaddr *)&address;
253 	size_t salen = sizeof(address);
254 	char *fo_path;
255 	struct net *net = file->f_dentry->d_sb->s_fs_info;
256 
257 	/* sanity check */
258 	if (size == 0)
259 		return -EINVAL;
260 
261 	if (buf[size-1] != '\n')
262 		return -EINVAL;
263 
264 	fo_path = buf;
265 	if (qword_get(&buf, fo_path, size) < 0)
266 		return -EINVAL;
267 
268 	if (rpc_pton(net, fo_path, size, sap, salen) == 0)
269 		return -EINVAL;
270 
271 	return nlmsvc_unlock_all_by_ip(sap);
272 }
273 
274 /**
275  * write_unlock_fs - Release all locks on a local file system
276  *
277  * Experimental.
278  *
279  * Input:
280  *			buf:	'\n'-terminated C string containing the
281  *				absolute pathname of a local file system
282  *			size:	length of C string in @buf
283  * Output:
284  *	On success:	returns zero if all specified locks were released;
285  *			returns one if one or more locks were not released
286  *	On error:	return code is negative errno value
287  */
288 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
289 {
290 	struct path path;
291 	char *fo_path;
292 	int error;
293 
294 	/* sanity check */
295 	if (size == 0)
296 		return -EINVAL;
297 
298 	if (buf[size-1] != '\n')
299 		return -EINVAL;
300 
301 	fo_path = buf;
302 	if (qword_get(&buf, fo_path, size) < 0)
303 		return -EINVAL;
304 
305 	error = kern_path(fo_path, 0, &path);
306 	if (error)
307 		return error;
308 
309 	/*
310 	 * XXX: Needs better sanity checking.  Otherwise we could end up
311 	 * releasing locks on the wrong file system.
312 	 *
313 	 * For example:
314 	 * 1.  Does the path refer to a directory?
315 	 * 2.  Is that directory a mount point, or
316 	 * 3.  Is that directory the root of an exported file system?
317 	 */
318 	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
319 
320 	path_put(&path);
321 	return error;
322 }
323 
324 /**
325  * write_filehandle - Get a variable-length NFS file handle by path
326  *
327  * On input, the buffer contains a '\n'-terminated C string comprised of
328  * three alphanumeric words separated by whitespace.  The string may
329  * contain escape sequences.
330  *
331  * Input:
332  *			buf:
333  *				domain:		client domain name
334  *				path:		export pathname
335  *				maxsize:	numeric maximum size of
336  *						@buf
337  *			size:	length of C string in @buf
338  * Output:
339  *	On success:	passed-in buffer filled with '\n'-terminated C
340  *			string containing a ASCII hex text version
341  *			of the NFS file handle;
342  *			return code is the size in bytes of the string
343  *	On error:	return code is negative errno value
344  */
345 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
346 {
347 	char *dname, *path;
348 	int uninitialized_var(maxsize);
349 	char *mesg = buf;
350 	int len;
351 	struct auth_domain *dom;
352 	struct knfsd_fh fh;
353 	struct net *net = file->f_dentry->d_sb->s_fs_info;
354 
355 	if (size == 0)
356 		return -EINVAL;
357 
358 	if (buf[size-1] != '\n')
359 		return -EINVAL;
360 	buf[size-1] = 0;
361 
362 	dname = mesg;
363 	len = qword_get(&mesg, dname, size);
364 	if (len <= 0)
365 		return -EINVAL;
366 
367 	path = dname+len+1;
368 	len = qword_get(&mesg, path, size);
369 	if (len <= 0)
370 		return -EINVAL;
371 
372 	len = get_int(&mesg, &maxsize);
373 	if (len)
374 		return len;
375 
376 	if (maxsize < NFS_FHSIZE)
377 		return -EINVAL;
378 	maxsize = min(maxsize, NFS3_FHSIZE);
379 
380 	if (qword_get(&mesg, mesg, size)>0)
381 		return -EINVAL;
382 
383 	/* we have all the words, they are in buf.. */
384 	dom = unix_domain_find(dname);
385 	if (!dom)
386 		return -ENOMEM;
387 
388 	len = exp_rootfh(net, dom, path, &fh,  maxsize);
389 	auth_domain_put(dom);
390 	if (len)
391 		return len;
392 
393 	mesg = buf;
394 	len = SIMPLE_TRANSACTION_LIMIT;
395 	qword_addhex(&mesg, &len, (char*)&fh.fh_base, fh.fh_size);
396 	mesg[-1] = '\n';
397 	return mesg - buf;
398 }
399 
400 /**
401  * write_threads - Start NFSD, or report the current number of running threads
402  *
403  * Input:
404  *			buf:		ignored
405  *			size:		zero
406  * Output:
407  *	On success:	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
412  *
413  * OR
414  *
415  * Input:
416  *			buf:		C string containing an unsigned
417  *					integer value representing the
418  *					number of NFSD threads to start
419  *			size:		non-zero length of C string in @buf
420  * Output:
421  *	On success:	NFS service is started;
422  *			passed-in buffer filled with '\n'-terminated C
423  *			string numeric value representing the number of
424  *			running NFSD threads;
425  *			return code is the size in bytes of the string
426  *	On error:	return code is zero or a negative errno value
427  */
428 static ssize_t write_threads(struct file *file, char *buf, size_t size)
429 {
430 	char *mesg = buf;
431 	int rv;
432 	struct net *net = file->f_dentry->d_sb->s_fs_info;
433 
434 	if (size > 0) {
435 		int newthreads;
436 		rv = get_int(&mesg, &newthreads);
437 		if (rv)
438 			return rv;
439 		if (newthreads < 0)
440 			return -EINVAL;
441 		rv = nfsd_svc(newthreads, net);
442 		if (rv < 0)
443 			return rv;
444 	} else
445 		rv = nfsd_nrthreads(net);
446 
447 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
448 }
449 
450 /**
451  * write_pool_threads - Set or report the current number of threads per pool
452  *
453  * Input:
454  *			buf:		ignored
455  *			size:		zero
456  *
457  * OR
458  *
459  * Input:
460  * 			buf:		C string containing whitespace-
461  * 					separated unsigned integer values
462  *					representing the number of NFSD
463  *					threads to start in each pool
464  *			size:		non-zero length of C string in @buf
465  * Output:
466  *	On success:	passed-in buffer filled with '\n'-terminated C
467  *			string containing integer values representing the
468  *			number of NFSD threads in each pool;
469  *			return code is the size in bytes of the string
470  *	On error:	return code is zero or a negative errno value
471  */
472 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
473 {
474 	/* if size > 0, look for an array of number of threads per node
475 	 * and apply them  then write out number of threads per node as reply
476 	 */
477 	char *mesg = buf;
478 	int i;
479 	int rv;
480 	int len;
481 	int npools;
482 	int *nthreads;
483 	struct net *net = file->f_dentry->d_sb->s_fs_info;
484 
485 	mutex_lock(&nfsd_mutex);
486 	npools = nfsd_nrpools(net);
487 	if (npools == 0) {
488 		/*
489 		 * NFS is shut down.  The admin can start it by
490 		 * writing to the threads file but NOT the pool_threads
491 		 * file, sorry.  Report zero threads.
492 		 */
493 		mutex_unlock(&nfsd_mutex);
494 		strcpy(buf, "0\n");
495 		return strlen(buf);
496 	}
497 
498 	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
499 	rv = -ENOMEM;
500 	if (nthreads == NULL)
501 		goto out_free;
502 
503 	if (size > 0) {
504 		for (i = 0; i < npools; i++) {
505 			rv = get_int(&mesg, &nthreads[i]);
506 			if (rv == -ENOENT)
507 				break;		/* fewer numbers than pools */
508 			if (rv)
509 				goto out_free;	/* syntax error */
510 			rv = -EINVAL;
511 			if (nthreads[i] < 0)
512 				goto out_free;
513 		}
514 		rv = nfsd_set_nrthreads(i, nthreads, net);
515 		if (rv)
516 			goto out_free;
517 	}
518 
519 	rv = nfsd_get_nrthreads(npools, nthreads, net);
520 	if (rv)
521 		goto out_free;
522 
523 	mesg = buf;
524 	size = SIMPLE_TRANSACTION_LIMIT;
525 	for (i = 0; i < npools && size > 0; i++) {
526 		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
527 		len = strlen(mesg);
528 		size -= len;
529 		mesg += len;
530 	}
531 	rv = mesg - buf;
532 out_free:
533 	kfree(nthreads);
534 	mutex_unlock(&nfsd_mutex);
535 	return rv;
536 }
537 
538 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
539 {
540 	char *mesg = buf;
541 	char *vers, *minorp, sign;
542 	int len, num, remaining;
543 	unsigned minor;
544 	ssize_t tlen = 0;
545 	char *sep;
546 	struct net *net = file->f_dentry->d_sb->s_fs_info;
547 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
548 
549 	if (size>0) {
550 		if (nn->nfsd_serv)
551 			/* Cannot change versions without updating
552 			 * nn->nfsd_serv->sv_xdrsize, and reallocing
553 			 * rq_argp and rq_resp
554 			 */
555 			return -EBUSY;
556 		if (buf[size-1] != '\n')
557 			return -EINVAL;
558 		buf[size-1] = 0;
559 
560 		vers = mesg;
561 		len = qword_get(&mesg, vers, size);
562 		if (len <= 0) return -EINVAL;
563 		do {
564 			sign = *vers;
565 			if (sign == '+' || sign == '-')
566 				num = simple_strtol((vers+1), &minorp, 0);
567 			else
568 				num = simple_strtol(vers, &minorp, 0);
569 			if (*minorp == '.') {
570 				if (num != 4)
571 					return -EINVAL;
572 				minor = simple_strtoul(minorp+1, NULL, 0);
573 				if (minor == 0)
574 					return -EINVAL;
575 				if (nfsd_minorversion(minor, sign == '-' ?
576 						     NFSD_CLEAR : NFSD_SET) < 0)
577 					return -EINVAL;
578 				goto next;
579 			}
580 			switch(num) {
581 			case 2:
582 			case 3:
583 			case 4:
584 				nfsd_vers(num, sign == '-' ? NFSD_CLEAR : NFSD_SET);
585 				break;
586 			default:
587 				return -EINVAL;
588 			}
589 		next:
590 			vers += len + 1;
591 		} while ((len = qword_get(&mesg, vers, size)) > 0);
592 		/* If all get turned off, turn them back on, as
593 		 * having no versions is BAD
594 		 */
595 		nfsd_reset_versions();
596 	}
597 
598 	/* Now write current state into reply buffer */
599 	len = 0;
600 	sep = "";
601 	remaining = SIMPLE_TRANSACTION_LIMIT;
602 	for (num=2 ; num <= 4 ; num++)
603 		if (nfsd_vers(num, NFSD_AVAIL)) {
604 			len = snprintf(buf, remaining, "%s%c%d", sep,
605 				       nfsd_vers(num, NFSD_TEST)?'+':'-',
606 				       num);
607 			sep = " ";
608 
609 			if (len > remaining)
610 				break;
611 			remaining -= len;
612 			buf += len;
613 			tlen += len;
614 		}
615 	if (nfsd_vers(4, NFSD_AVAIL))
616 		for (minor = 1; minor <= NFSD_SUPPORTED_MINOR_VERSION;
617 		     minor++) {
618 			len = snprintf(buf, remaining, " %c4.%u",
619 					(nfsd_vers(4, NFSD_TEST) &&
620 					 nfsd_minorversion(minor, NFSD_TEST)) ?
621 						'+' : '-',
622 					minor);
623 
624 			if (len > remaining)
625 				break;
626 			remaining -= len;
627 			buf += len;
628 			tlen += len;
629 		}
630 
631 	len = snprintf(buf, remaining, "\n");
632 	if (len > remaining)
633 		return -EINVAL;
634 	return tlen + len;
635 }
636 
637 /**
638  * write_versions - Set or report the available NFS protocol versions
639  *
640  * Input:
641  *			buf:		ignored
642  *			size:		zero
643  * Output:
644  *	On success:	passed-in buffer filled with '\n'-terminated C
645  *			string containing positive or negative integer
646  *			values representing the current status of each
647  *			protocol version;
648  *			return code is the size in bytes of the string
649  *	On error:	return code is zero or a negative errno value
650  *
651  * OR
652  *
653  * Input:
654  * 			buf:		C string containing whitespace-
655  * 					separated positive or negative
656  * 					integer values representing NFS
657  * 					protocol versions to enable ("+n")
658  * 					or disable ("-n")
659  *			size:		non-zero length of C string in @buf
660  * Output:
661  *	On success:	status of zero or more protocol versions has
662  *			been updated; passed-in buffer filled with
663  *			'\n'-terminated C string containing positive
664  *			or negative integer values representing the
665  *			current status of each protocol version;
666  *			return code is the size in bytes of the string
667  *	On error:	return code is zero or a negative errno value
668  */
669 static ssize_t write_versions(struct file *file, char *buf, size_t size)
670 {
671 	ssize_t rv;
672 
673 	mutex_lock(&nfsd_mutex);
674 	rv = __write_versions(file, buf, size);
675 	mutex_unlock(&nfsd_mutex);
676 	return rv;
677 }
678 
679 /*
680  * Zero-length write.  Return a list of NFSD's current listener
681  * transports.
682  */
683 static ssize_t __write_ports_names(char *buf, struct net *net)
684 {
685 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
686 
687 	if (nn->nfsd_serv == NULL)
688 		return 0;
689 	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
690 }
691 
692 /*
693  * A single 'fd' number was written, in which case it must be for
694  * a socket of a supported family/protocol, and we use it as an
695  * nfsd listener.
696  */
697 static ssize_t __write_ports_addfd(char *buf, struct net *net)
698 {
699 	char *mesg = buf;
700 	int fd, err;
701 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
702 
703 	err = get_int(&mesg, &fd);
704 	if (err != 0 || fd < 0)
705 		return -EINVAL;
706 
707 	if (svc_alien_sock(net, fd)) {
708 		printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
709 		return -EINVAL;
710 	}
711 
712 	err = nfsd_create_serv(net);
713 	if (err != 0)
714 		return err;
715 
716 	err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT);
717 	if (err < 0) {
718 		nfsd_destroy(net);
719 		return err;
720 	}
721 
722 	/* Decrease the count, but don't shut down the service */
723 	nn->nfsd_serv->sv_nrthreads--;
724 	return err;
725 }
726 
727 /*
728  * A transport listener is added by writing it's transport name and
729  * a port number.
730  */
731 static ssize_t __write_ports_addxprt(char *buf, struct net *net)
732 {
733 	char transport[16];
734 	struct svc_xprt *xprt;
735 	int port, err;
736 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
737 
738 	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
739 		return -EINVAL;
740 
741 	if (port < 1 || port > USHRT_MAX)
742 		return -EINVAL;
743 
744 	err = nfsd_create_serv(net);
745 	if (err != 0)
746 		return err;
747 
748 	err = svc_create_xprt(nn->nfsd_serv, transport, net,
749 				PF_INET, port, SVC_SOCK_ANONYMOUS);
750 	if (err < 0)
751 		goto out_err;
752 
753 	err = svc_create_xprt(nn->nfsd_serv, transport, net,
754 				PF_INET6, port, SVC_SOCK_ANONYMOUS);
755 	if (err < 0 && err != -EAFNOSUPPORT)
756 		goto out_close;
757 
758 	/* Decrease the count, but don't shut down the service */
759 	nn->nfsd_serv->sv_nrthreads--;
760 	return 0;
761 out_close:
762 	xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
763 	if (xprt != NULL) {
764 		svc_close_xprt(xprt);
765 		svc_xprt_put(xprt);
766 	}
767 out_err:
768 	nfsd_destroy(net);
769 	return err;
770 }
771 
772 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
773 			     struct net *net)
774 {
775 	if (size == 0)
776 		return __write_ports_names(buf, net);
777 
778 	if (isdigit(buf[0]))
779 		return __write_ports_addfd(buf, net);
780 
781 	if (isalpha(buf[0]))
782 		return __write_ports_addxprt(buf, net);
783 
784 	return -EINVAL;
785 }
786 
787 /**
788  * write_ports - Pass a socket file descriptor or transport name to listen on
789  *
790  * Input:
791  *			buf:		ignored
792  *			size:		zero
793  * Output:
794  *	On success:	passed-in buffer filled with a '\n'-terminated C
795  *			string containing a whitespace-separated list of
796  *			named NFSD listeners;
797  *			return code is the size in bytes of the string
798  *	On error:	return code is zero or a negative errno value
799  *
800  * OR
801  *
802  * Input:
803  *			buf:		C string containing an unsigned
804  *					integer value representing a bound
805  *					but unconnected socket that is to be
806  *					used as an NFSD listener; listen(3)
807  *					must be called for a SOCK_STREAM
808  *					socket, otherwise it is ignored
809  *			size:		non-zero length of C string in @buf
810  * Output:
811  *	On success:	NFS service is started;
812  *			passed-in buffer filled with a '\n'-terminated C
813  *			string containing a unique alphanumeric name of
814  *			the listener;
815  *			return code is the size in bytes of the string
816  *	On error:	return code is a negative errno value
817  *
818  * OR
819  *
820  * Input:
821  *			buf:		C string containing a transport
822  *					name and an unsigned integer value
823  *					representing the port to listen on,
824  *					separated by whitespace
825  *			size:		non-zero length of C string in @buf
826  * Output:
827  *	On success:	returns zero; NFS service is started
828  *	On error:	return code is a negative errno value
829  */
830 static ssize_t write_ports(struct file *file, char *buf, size_t size)
831 {
832 	ssize_t rv;
833 	struct net *net = file->f_dentry->d_sb->s_fs_info;
834 
835 	mutex_lock(&nfsd_mutex);
836 	rv = __write_ports(file, buf, size, net);
837 	mutex_unlock(&nfsd_mutex);
838 	return rv;
839 }
840 
841 
842 int nfsd_max_blksize;
843 
844 /**
845  * write_maxblksize - Set or report the current NFS blksize
846  *
847  * Input:
848  *			buf:		ignored
849  *			size:		zero
850  *
851  * OR
852  *
853  * Input:
854  * 			buf:		C string containing an unsigned
855  * 					integer value representing the new
856  * 					NFS blksize
857  *			size:		non-zero length of C string in @buf
858  * Output:
859  *	On success:	passed-in buffer filled with '\n'-terminated C string
860  *			containing numeric value of the current NFS blksize
861  *			setting;
862  *			return code is the size in bytes of the string
863  *	On error:	return code is zero or a negative errno value
864  */
865 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
866 {
867 	char *mesg = buf;
868 	struct net *net = file->f_dentry->d_sb->s_fs_info;
869 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
870 
871 	if (size > 0) {
872 		int bsize;
873 		int rv = get_int(&mesg, &bsize);
874 		if (rv)
875 			return rv;
876 		/* force bsize into allowed range and
877 		 * required alignment.
878 		 */
879 		bsize = max_t(int, bsize, 1024);
880 		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
881 		bsize &= ~(1024-1);
882 		mutex_lock(&nfsd_mutex);
883 		if (nn->nfsd_serv) {
884 			mutex_unlock(&nfsd_mutex);
885 			return -EBUSY;
886 		}
887 		nfsd_max_blksize = bsize;
888 		mutex_unlock(&nfsd_mutex);
889 	}
890 
891 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
892 							nfsd_max_blksize);
893 }
894 
895 /**
896  * write_maxconn - Set or report the current max number of connections
897  *
898  * Input:
899  *			buf:		ignored
900  *			size:		zero
901  * OR
902  *
903  * Input:
904  * 			buf:		C string containing an unsigned
905  * 					integer value representing the new
906  * 					number of max connections
907  *			size:		non-zero length of C string in @buf
908  * Output:
909  *	On success:	passed-in buffer filled with '\n'-terminated C string
910  *			containing numeric value of max_connections setting
911  *			for this net namespace;
912  *			return code is the size in bytes of the string
913  *	On error:	return code is zero or a negative errno value
914  */
915 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
916 {
917 	char *mesg = buf;
918 	struct net *net = file->f_dentry->d_sb->s_fs_info;
919 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
920 	unsigned int maxconn = nn->max_connections;
921 
922 	if (size > 0) {
923 		int rv = get_uint(&mesg, &maxconn);
924 
925 		if (rv)
926 			return rv;
927 		nn->max_connections = maxconn;
928 	}
929 
930 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
931 }
932 
933 #ifdef CONFIG_NFSD_V4
934 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
935 				  time_t *time, struct nfsd_net *nn)
936 {
937 	char *mesg = buf;
938 	int rv, i;
939 
940 	if (size > 0) {
941 		if (nn->nfsd_serv)
942 			return -EBUSY;
943 		rv = get_int(&mesg, &i);
944 		if (rv)
945 			return rv;
946 		/*
947 		 * Some sanity checking.  We don't have a reason for
948 		 * these particular numbers, but problems with the
949 		 * extremes are:
950 		 *	- Too short: the briefest network outage may
951 		 *	  cause clients to lose all their locks.  Also,
952 		 *	  the frequent polling may be wasteful.
953 		 *	- Too long: do you really want reboot recovery
954 		 *	  to take more than an hour?  Or to make other
955 		 *	  clients wait an hour before being able to
956 		 *	  revoke a dead client's locks?
957 		 */
958 		if (i < 10 || i > 3600)
959 			return -EINVAL;
960 		*time = i;
961 	}
962 
963 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%ld\n", *time);
964 }
965 
966 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
967 				time_t *time, struct nfsd_net *nn)
968 {
969 	ssize_t rv;
970 
971 	mutex_lock(&nfsd_mutex);
972 	rv = __nfsd4_write_time(file, buf, size, time, nn);
973 	mutex_unlock(&nfsd_mutex);
974 	return rv;
975 }
976 
977 /**
978  * write_leasetime - Set or report the current NFSv4 lease time
979  *
980  * Input:
981  *			buf:		ignored
982  *			size:		zero
983  *
984  * OR
985  *
986  * Input:
987  *			buf:		C string containing an unsigned
988  *					integer value representing the new
989  *					NFSv4 lease expiry time
990  *			size:		non-zero length of C string in @buf
991  * Output:
992  *	On success:	passed-in buffer filled with '\n'-terminated C
993  *			string containing unsigned integer value of the
994  *			current lease expiry time;
995  *			return code is the size in bytes of the string
996  *	On error:	return code is zero or a negative errno value
997  */
998 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
999 {
1000 	struct net *net = file->f_dentry->d_sb->s_fs_info;
1001 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1002 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1003 }
1004 
1005 /**
1006  * write_gracetime - Set or report current NFSv4 grace period time
1007  *
1008  * As above, but sets the time of the NFSv4 grace period.
1009  *
1010  * Note this should never be set to less than the *previous*
1011  * lease-period time, but we don't try to enforce this.  (In the common
1012  * case (a new boot), we don't know what the previous lease time was
1013  * anyway.)
1014  */
1015 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1016 {
1017 	struct net *net = file->f_dentry->d_sb->s_fs_info;
1018 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1019 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1020 }
1021 
1022 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1023 				   struct nfsd_net *nn)
1024 {
1025 	char *mesg = buf;
1026 	char *recdir;
1027 	int len, status;
1028 
1029 	if (size > 0) {
1030 		if (nn->nfsd_serv)
1031 			return -EBUSY;
1032 		if (size > PATH_MAX || buf[size-1] != '\n')
1033 			return -EINVAL;
1034 		buf[size-1] = 0;
1035 
1036 		recdir = mesg;
1037 		len = qword_get(&mesg, recdir, size);
1038 		if (len <= 0)
1039 			return -EINVAL;
1040 
1041 		status = nfs4_reset_recoverydir(recdir);
1042 		if (status)
1043 			return status;
1044 	}
1045 
1046 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1047 							nfs4_recoverydir());
1048 }
1049 
1050 /**
1051  * write_recoverydir - Set or report the pathname of the recovery directory
1052  *
1053  * Input:
1054  *			buf:		ignored
1055  *			size:		zero
1056  *
1057  * OR
1058  *
1059  * Input:
1060  *			buf:		C string containing the pathname
1061  *					of the directory on a local file
1062  *					system containing permanent NFSv4
1063  *					recovery data
1064  *			size:		non-zero length of C string in @buf
1065  * Output:
1066  *	On success:	passed-in buffer filled with '\n'-terminated C string
1067  *			containing the current recovery pathname setting;
1068  *			return code is the size in bytes of the string
1069  *	On error:	return code is zero or a negative errno value
1070  */
1071 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1072 {
1073 	ssize_t rv;
1074 	struct net *net = file->f_dentry->d_sb->s_fs_info;
1075 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1076 
1077 	mutex_lock(&nfsd_mutex);
1078 	rv = __write_recoverydir(file, buf, size, nn);
1079 	mutex_unlock(&nfsd_mutex);
1080 	return rv;
1081 }
1082 
1083 /**
1084  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1085  *
1086  * Input:
1087  *			buf:		ignored
1088  *			size:		zero
1089  * OR
1090  *
1091  * Input:
1092  * 			buf:		any value
1093  *			size:		non-zero length of C string in @buf
1094  * Output:
1095  *			passed-in buffer filled with "Y" or "N" with a newline
1096  *			and NULL-terminated C string. This indicates whether
1097  *			the grace period has ended in the current net
1098  *			namespace. Return code is the size in bytes of the
1099  *			string. Writing a string that starts with 'Y', 'y', or
1100  *			'1' to the file will end the grace period for nfsd's v4
1101  *			lock manager.
1102  */
1103 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1104 {
1105 	struct net *net = file->f_dentry->d_sb->s_fs_info;
1106 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1107 
1108 	if (size > 0) {
1109 		switch(buf[0]) {
1110 		case 'Y':
1111 		case 'y':
1112 		case '1':
1113 			nfsd4_end_grace(nn);
1114 			break;
1115 		default:
1116 			return -EINVAL;
1117 		}
1118 	}
1119 
1120 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1121 			 nn->grace_ended ? 'Y' : 'N');
1122 }
1123 
1124 #endif
1125 
1126 /*----------------------------------------------------------------------------*/
1127 /*
1128  *	populating the filesystem.
1129  */
1130 
1131 static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
1132 {
1133 	static struct tree_descr nfsd_files[] = {
1134 		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1135 		[NFSD_Export_features] = {"export_features",
1136 					&export_features_operations, S_IRUGO},
1137 		[NFSD_FO_UnlockIP] = {"unlock_ip",
1138 					&transaction_ops, S_IWUSR|S_IRUSR},
1139 		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1140 					&transaction_ops, S_IWUSR|S_IRUSR},
1141 		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1142 		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1143 		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1144 		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1145 		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
1146 		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1147 		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1148 		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1149 		[NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1150 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1151 		[NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1152 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1153 #ifdef CONFIG_NFSD_V4
1154 		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1155 		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1156 		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1157 		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1158 #endif
1159 		/* last one */ {""}
1160 	};
1161 	struct net *net = data;
1162 	int ret;
1163 
1164 	ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1165 	if (ret)
1166 		return ret;
1167 	sb->s_fs_info = get_net(net);
1168 	return 0;
1169 }
1170 
1171 static struct dentry *nfsd_mount(struct file_system_type *fs_type,
1172 	int flags, const char *dev_name, void *data)
1173 {
1174 	return mount_ns(fs_type, flags, current->nsproxy->net_ns, nfsd_fill_super);
1175 }
1176 
1177 static void nfsd_umount(struct super_block *sb)
1178 {
1179 	struct net *net = sb->s_fs_info;
1180 
1181 	kill_litter_super(sb);
1182 	put_net(net);
1183 }
1184 
1185 static struct file_system_type nfsd_fs_type = {
1186 	.owner		= THIS_MODULE,
1187 	.name		= "nfsd",
1188 	.mount		= nfsd_mount,
1189 	.kill_sb	= nfsd_umount,
1190 };
1191 MODULE_ALIAS_FS("nfsd");
1192 
1193 #ifdef CONFIG_PROC_FS
1194 static int create_proc_exports_entry(void)
1195 {
1196 	struct proc_dir_entry *entry;
1197 
1198 	entry = proc_mkdir("fs/nfs", NULL);
1199 	if (!entry)
1200 		return -ENOMEM;
1201 	entry = proc_create("exports", 0, entry,
1202 				 &exports_proc_operations);
1203 	if (!entry) {
1204 		remove_proc_entry("fs/nfs", NULL);
1205 		return -ENOMEM;
1206 	}
1207 	return 0;
1208 }
1209 #else /* CONFIG_PROC_FS */
1210 static int create_proc_exports_entry(void)
1211 {
1212 	return 0;
1213 }
1214 #endif
1215 
1216 int nfsd_net_id;
1217 
1218 static __net_init int nfsd_init_net(struct net *net)
1219 {
1220 	int retval;
1221 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1222 
1223 	retval = nfsd_export_init(net);
1224 	if (retval)
1225 		goto out_export_error;
1226 	retval = nfsd_idmap_init(net);
1227 	if (retval)
1228 		goto out_idmap_error;
1229 	nn->nfsd4_lease = 90;	/* default lease time */
1230 	nn->nfsd4_grace = 90;
1231 	return 0;
1232 
1233 out_idmap_error:
1234 	nfsd_export_shutdown(net);
1235 out_export_error:
1236 	return retval;
1237 }
1238 
1239 static __net_exit void nfsd_exit_net(struct net *net)
1240 {
1241 	nfsd_idmap_shutdown(net);
1242 	nfsd_export_shutdown(net);
1243 }
1244 
1245 static struct pernet_operations nfsd_net_ops = {
1246 	.init = nfsd_init_net,
1247 	.exit = nfsd_exit_net,
1248 	.id   = &nfsd_net_id,
1249 	.size = sizeof(struct nfsd_net),
1250 };
1251 
1252 static int __init init_nfsd(void)
1253 {
1254 	int retval;
1255 	printk(KERN_INFO "Installing knfsd (copyright (C) 1996 okir@monad.swb.de).\n");
1256 
1257 	retval = register_cld_notifier();
1258 	if (retval)
1259 		return retval;
1260 	retval = register_pernet_subsys(&nfsd_net_ops);
1261 	if (retval < 0)
1262 		goto out_unregister_notifier;
1263 	retval = nfsd4_init_slabs();
1264 	if (retval)
1265 		goto out_unregister_pernet;
1266 	retval = nfsd_fault_inject_init(); /* nfsd fault injection controls */
1267 	if (retval)
1268 		goto out_free_slabs;
1269 	nfsd_stat_init();	/* Statistics */
1270 	retval = nfsd_reply_cache_init();
1271 	if (retval)
1272 		goto out_free_stat;
1273 	nfsd_lockd_init();	/* lockd->nfsd callbacks */
1274 	retval = create_proc_exports_entry();
1275 	if (retval)
1276 		goto out_free_lockd;
1277 	retval = register_filesystem(&nfsd_fs_type);
1278 	if (retval)
1279 		goto out_free_all;
1280 	return 0;
1281 out_free_all:
1282 	remove_proc_entry("fs/nfs/exports", NULL);
1283 	remove_proc_entry("fs/nfs", NULL);
1284 out_free_lockd:
1285 	nfsd_lockd_shutdown();
1286 	nfsd_reply_cache_shutdown();
1287 out_free_stat:
1288 	nfsd_stat_shutdown();
1289 	nfsd_fault_inject_cleanup();
1290 out_free_slabs:
1291 	nfsd4_free_slabs();
1292 out_unregister_pernet:
1293 	unregister_pernet_subsys(&nfsd_net_ops);
1294 out_unregister_notifier:
1295 	unregister_cld_notifier();
1296 	return retval;
1297 }
1298 
1299 static void __exit exit_nfsd(void)
1300 {
1301 	nfsd_reply_cache_shutdown();
1302 	remove_proc_entry("fs/nfs/exports", NULL);
1303 	remove_proc_entry("fs/nfs", NULL);
1304 	nfsd_stat_shutdown();
1305 	nfsd_lockd_shutdown();
1306 	nfsd4_free_slabs();
1307 	nfsd_fault_inject_cleanup();
1308 	unregister_filesystem(&nfsd_fs_type);
1309 	unregister_pernet_subsys(&nfsd_net_ops);
1310 	unregister_cld_notifier();
1311 }
1312 
1313 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1314 MODULE_LICENSE("GPL");
1315 module_init(init_nfsd)
1316 module_exit(exit_nfsd)
1317