xref: /linux/fs/nfsd/nfsctl.c (revision d2912cb15bdda8ba4a5dd73396ad62641af2f520)
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 
12 #include <linux/sunrpc/svcsock.h>
13 #include <linux/lockd/lockd.h>
14 #include <linux/sunrpc/addr.h>
15 #include <linux/sunrpc/gss_api.h>
16 #include <linux/sunrpc/gss_krb5_enctypes.h>
17 #include <linux/sunrpc/rpc_pipe_fs.h>
18 #include <linux/module.h>
19 
20 #include "idmap.h"
21 #include "nfsd.h"
22 #include "cache.h"
23 #include "state.h"
24 #include "netns.h"
25 #include "pnfs.h"
26 
27 /*
28  *	We have a single directory with several nodes in it.
29  */
30 enum {
31 	NFSD_Root = 1,
32 	NFSD_List,
33 	NFSD_Export_features,
34 	NFSD_Fh,
35 	NFSD_FO_UnlockIP,
36 	NFSD_FO_UnlockFS,
37 	NFSD_Threads,
38 	NFSD_Pool_Threads,
39 	NFSD_Pool_Stats,
40 	NFSD_Reply_Cache_Stats,
41 	NFSD_Versions,
42 	NFSD_Ports,
43 	NFSD_MaxBlkSize,
44 	NFSD_MaxConnections,
45 	NFSD_SupportedEnctypes,
46 	/*
47 	 * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
48 	 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
49 	 */
50 #ifdef CONFIG_NFSD_V4
51 	NFSD_Leasetime,
52 	NFSD_Gracetime,
53 	NFSD_RecoveryDir,
54 	NFSD_V4EndGrace,
55 #endif
56 };
57 
58 /*
59  * write() for these nodes.
60  */
61 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
62 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
63 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
64 static ssize_t write_threads(struct file *file, char *buf, size_t size);
65 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
66 static ssize_t write_versions(struct file *file, char *buf, size_t size);
67 static ssize_t write_ports(struct file *file, char *buf, size_t size);
68 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
69 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
70 #ifdef CONFIG_NFSD_V4
71 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
72 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
73 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
74 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
75 #endif
76 
77 static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
78 	[NFSD_Fh] = write_filehandle,
79 	[NFSD_FO_UnlockIP] = write_unlock_ip,
80 	[NFSD_FO_UnlockFS] = write_unlock_fs,
81 	[NFSD_Threads] = write_threads,
82 	[NFSD_Pool_Threads] = write_pool_threads,
83 	[NFSD_Versions] = write_versions,
84 	[NFSD_Ports] = write_ports,
85 	[NFSD_MaxBlkSize] = write_maxblksize,
86 	[NFSD_MaxConnections] = write_maxconn,
87 #ifdef CONFIG_NFSD_V4
88 	[NFSD_Leasetime] = write_leasetime,
89 	[NFSD_Gracetime] = write_gracetime,
90 	[NFSD_RecoveryDir] = write_recoverydir,
91 	[NFSD_V4EndGrace] = write_v4_end_grace,
92 #endif
93 };
94 
95 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
96 {
97 	ino_t ino =  file_inode(file)->i_ino;
98 	char *data;
99 	ssize_t rv;
100 
101 	if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
102 		return -EINVAL;
103 
104 	data = simple_transaction_get(file, buf, size);
105 	if (IS_ERR(data))
106 		return PTR_ERR(data);
107 
108 	rv =  write_op[ino](file, data, size);
109 	if (rv >= 0) {
110 		simple_transaction_set(file, rv);
111 		rv = size;
112 	}
113 	return rv;
114 }
115 
116 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
117 {
118 	if (! file->private_data) {
119 		/* An attempt to read a transaction file without writing
120 		 * causes a 0-byte write so that the file can return
121 		 * state information
122 		 */
123 		ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
124 		if (rv < 0)
125 			return rv;
126 	}
127 	return simple_transaction_read(file, buf, size, pos);
128 }
129 
130 static const struct file_operations transaction_ops = {
131 	.write		= nfsctl_transaction_write,
132 	.read		= nfsctl_transaction_read,
133 	.release	= simple_transaction_release,
134 	.llseek		= default_llseek,
135 };
136 
137 static int exports_net_open(struct net *net, struct file *file)
138 {
139 	int err;
140 	struct seq_file *seq;
141 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
142 
143 	err = seq_open(file, &nfs_exports_op);
144 	if (err)
145 		return err;
146 
147 	seq = file->private_data;
148 	seq->private = nn->svc_export_cache;
149 	return 0;
150 }
151 
152 static int exports_proc_open(struct inode *inode, struct file *file)
153 {
154 	return exports_net_open(current->nsproxy->net_ns, file);
155 }
156 
157 static const struct file_operations exports_proc_operations = {
158 	.open		= exports_proc_open,
159 	.read		= seq_read,
160 	.llseek		= seq_lseek,
161 	.release	= seq_release,
162 };
163 
164 static int exports_nfsd_open(struct inode *inode, struct file *file)
165 {
166 	return exports_net_open(inode->i_sb->s_fs_info, file);
167 }
168 
169 static const struct file_operations exports_nfsd_operations = {
170 	.open		= exports_nfsd_open,
171 	.read		= seq_read,
172 	.llseek		= seq_lseek,
173 	.release	= seq_release,
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 };
220 
221 static const struct file_operations reply_cache_stats_operations = {
222 	.open		= nfsd_reply_cache_stats_open,
223 	.read		= seq_read,
224 	.llseek		= seq_lseek,
225 	.release	= single_release,
226 };
227 
228 /*----------------------------------------------------------------------------*/
229 /*
230  * payload - write methods
231  */
232 
233 static inline struct net *netns(struct file *file)
234 {
235 	return file_inode(file)->i_sb->s_fs_info;
236 }
237 
238 /**
239  * write_unlock_ip - Release all locks used by a client
240  *
241  * Experimental.
242  *
243  * Input:
244  *			buf:	'\n'-terminated C string containing a
245  *				presentation format IP address
246  *			size:	length of C string in @buf
247  * Output:
248  *	On success:	returns zero if all specified locks were released;
249  *			returns one if one or more locks were not released
250  *	On error:	return code is negative errno value
251  */
252 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
253 {
254 	struct sockaddr_storage address;
255 	struct sockaddr *sap = (struct sockaddr *)&address;
256 	size_t salen = sizeof(address);
257 	char *fo_path;
258 	struct net *net = netns(file);
259 
260 	/* sanity check */
261 	if (size == 0)
262 		return -EINVAL;
263 
264 	if (buf[size-1] != '\n')
265 		return -EINVAL;
266 
267 	fo_path = buf;
268 	if (qword_get(&buf, fo_path, size) < 0)
269 		return -EINVAL;
270 
271 	if (rpc_pton(net, fo_path, size, sap, salen) == 0)
272 		return -EINVAL;
273 
274 	return nlmsvc_unlock_all_by_ip(sap);
275 }
276 
277 /**
278  * write_unlock_fs - Release all locks on a local file system
279  *
280  * Experimental.
281  *
282  * Input:
283  *			buf:	'\n'-terminated C string containing the
284  *				absolute pathname of a local file system
285  *			size:	length of C string in @buf
286  * Output:
287  *	On success:	returns zero if all specified locks were released;
288  *			returns one if one or more locks were not released
289  *	On error:	return code is negative errno value
290  */
291 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
292 {
293 	struct path path;
294 	char *fo_path;
295 	int error;
296 
297 	/* sanity check */
298 	if (size == 0)
299 		return -EINVAL;
300 
301 	if (buf[size-1] != '\n')
302 		return -EINVAL;
303 
304 	fo_path = buf;
305 	if (qword_get(&buf, fo_path, size) < 0)
306 		return -EINVAL;
307 
308 	error = kern_path(fo_path, 0, &path);
309 	if (error)
310 		return error;
311 
312 	/*
313 	 * XXX: Needs better sanity checking.  Otherwise we could end up
314 	 * releasing locks on the wrong file system.
315 	 *
316 	 * For example:
317 	 * 1.  Does the path refer to a directory?
318 	 * 2.  Is that directory a mount point, or
319 	 * 3.  Is that directory the root of an exported file system?
320 	 */
321 	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
322 
323 	path_put(&path);
324 	return error;
325 }
326 
327 /**
328  * write_filehandle - Get a variable-length NFS file handle by path
329  *
330  * On input, the buffer contains a '\n'-terminated C string comprised of
331  * three alphanumeric words separated by whitespace.  The string may
332  * contain escape sequences.
333  *
334  * Input:
335  *			buf:
336  *				domain:		client domain name
337  *				path:		export pathname
338  *				maxsize:	numeric maximum size of
339  *						@buf
340  *			size:	length of C string in @buf
341  * Output:
342  *	On success:	passed-in buffer filled with '\n'-terminated C
343  *			string containing a ASCII hex text version
344  *			of the NFS file handle;
345  *			return code is the size in bytes of the string
346  *	On error:	return code is negative errno value
347  */
348 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
349 {
350 	char *dname, *path;
351 	int uninitialized_var(maxsize);
352 	char *mesg = buf;
353 	int len;
354 	struct auth_domain *dom;
355 	struct knfsd_fh fh;
356 
357 	if (size == 0)
358 		return -EINVAL;
359 
360 	if (buf[size-1] != '\n')
361 		return -EINVAL;
362 	buf[size-1] = 0;
363 
364 	dname = mesg;
365 	len = qword_get(&mesg, dname, size);
366 	if (len <= 0)
367 		return -EINVAL;
368 
369 	path = dname+len+1;
370 	len = qword_get(&mesg, path, size);
371 	if (len <= 0)
372 		return -EINVAL;
373 
374 	len = get_int(&mesg, &maxsize);
375 	if (len)
376 		return len;
377 
378 	if (maxsize < NFS_FHSIZE)
379 		return -EINVAL;
380 	maxsize = min(maxsize, NFS3_FHSIZE);
381 
382 	if (qword_get(&mesg, mesg, size)>0)
383 		return -EINVAL;
384 
385 	/* we have all the words, they are in buf.. */
386 	dom = unix_domain_find(dname);
387 	if (!dom)
388 		return -ENOMEM;
389 
390 	len = exp_rootfh(netns(file), dom, path, &fh,  maxsize);
391 	auth_domain_put(dom);
392 	if (len)
393 		return len;
394 
395 	mesg = buf;
396 	len = SIMPLE_TRANSACTION_LIMIT;
397 	qword_addhex(&mesg, &len, (char*)&fh.fh_base, fh.fh_size);
398 	mesg[-1] = '\n';
399 	return mesg - buf;
400 }
401 
402 /**
403  * write_threads - Start NFSD, or report the current number of running threads
404  *
405  * Input:
406  *			buf:		ignored
407  *			size:		zero
408  * Output:
409  *	On success:	passed-in buffer filled with '\n'-terminated C
410  *			string numeric value representing the number of
411  *			running NFSD threads;
412  *			return code is the size in bytes of the string
413  *	On error:	return code is zero
414  *
415  * OR
416  *
417  * Input:
418  *			buf:		C string containing an unsigned
419  *					integer value representing the
420  *					number of NFSD threads to start
421  *			size:		non-zero length of C string in @buf
422  * Output:
423  *	On success:	NFS service is started;
424  *			passed-in buffer filled with '\n'-terminated C
425  *			string numeric value representing the number of
426  *			running NFSD threads;
427  *			return code is the size in bytes of the string
428  *	On error:	return code is zero or a negative errno value
429  */
430 static ssize_t write_threads(struct file *file, char *buf, size_t size)
431 {
432 	char *mesg = buf;
433 	int rv;
434 	struct net *net = netns(file);
435 
436 	if (size > 0) {
437 		int newthreads;
438 		rv = get_int(&mesg, &newthreads);
439 		if (rv)
440 			return rv;
441 		if (newthreads < 0)
442 			return -EINVAL;
443 		rv = nfsd_svc(newthreads, net, file->f_cred);
444 		if (rv < 0)
445 			return rv;
446 	} else
447 		rv = nfsd_nrthreads(net);
448 
449 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
450 }
451 
452 /**
453  * write_pool_threads - Set or report the current number of threads per pool
454  *
455  * Input:
456  *			buf:		ignored
457  *			size:		zero
458  *
459  * OR
460  *
461  * Input:
462  * 			buf:		C string containing whitespace-
463  * 					separated unsigned integer values
464  *					representing the number of NFSD
465  *					threads to start in each pool
466  *			size:		non-zero length of C string in @buf
467  * Output:
468  *	On success:	passed-in buffer filled with '\n'-terminated C
469  *			string containing integer values representing the
470  *			number of NFSD threads in each pool;
471  *			return code is the size in bytes of the string
472  *	On error:	return code is zero or a negative errno value
473  */
474 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
475 {
476 	/* if size > 0, look for an array of number of threads per node
477 	 * and apply them  then write out number of threads per node as reply
478 	 */
479 	char *mesg = buf;
480 	int i;
481 	int rv;
482 	int len;
483 	int npools;
484 	int *nthreads;
485 	struct net *net = netns(file);
486 
487 	mutex_lock(&nfsd_mutex);
488 	npools = nfsd_nrpools(net);
489 	if (npools == 0) {
490 		/*
491 		 * NFS is shut down.  The admin can start it by
492 		 * writing to the threads file but NOT the pool_threads
493 		 * file, sorry.  Report zero threads.
494 		 */
495 		mutex_unlock(&nfsd_mutex);
496 		strcpy(buf, "0\n");
497 		return strlen(buf);
498 	}
499 
500 	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
501 	rv = -ENOMEM;
502 	if (nthreads == NULL)
503 		goto out_free;
504 
505 	if (size > 0) {
506 		for (i = 0; i < npools; i++) {
507 			rv = get_int(&mesg, &nthreads[i]);
508 			if (rv == -ENOENT)
509 				break;		/* fewer numbers than pools */
510 			if (rv)
511 				goto out_free;	/* syntax error */
512 			rv = -EINVAL;
513 			if (nthreads[i] < 0)
514 				goto out_free;
515 		}
516 		rv = nfsd_set_nrthreads(i, nthreads, net);
517 		if (rv)
518 			goto out_free;
519 	}
520 
521 	rv = nfsd_get_nrthreads(npools, nthreads, net);
522 	if (rv)
523 		goto out_free;
524 
525 	mesg = buf;
526 	size = SIMPLE_TRANSACTION_LIMIT;
527 	for (i = 0; i < npools && size > 0; i++) {
528 		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
529 		len = strlen(mesg);
530 		size -= len;
531 		mesg += len;
532 	}
533 	rv = mesg - buf;
534 out_free:
535 	kfree(nthreads);
536 	mutex_unlock(&nfsd_mutex);
537 	return rv;
538 }
539 
540 static ssize_t
541 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
542 		const char *sep, unsigned vers, int minor)
543 {
544 	const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
545 	bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
546 
547 	if (vers == 4 && minor >= 0 &&
548 	    !nfsd_minorversion(nn, minor, NFSD_TEST))
549 		supported = false;
550 	if (minor == 0 && supported)
551 		/*
552 		 * special case for backward compatability.
553 		 * +4.0 is never reported, it is implied by
554 		 * +4, unless -4.0 is present.
555 		 */
556 		return 0;
557 	return snprintf(buf, remaining, format, sep,
558 			supported ? '+' : '-', vers, minor);
559 }
560 
561 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
562 {
563 	char *mesg = buf;
564 	char *vers, *minorp, sign;
565 	int len, num, remaining;
566 	ssize_t tlen = 0;
567 	char *sep;
568 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
569 
570 	if (size>0) {
571 		if (nn->nfsd_serv)
572 			/* Cannot change versions without updating
573 			 * nn->nfsd_serv->sv_xdrsize, and reallocing
574 			 * rq_argp and rq_resp
575 			 */
576 			return -EBUSY;
577 		if (buf[size-1] != '\n')
578 			return -EINVAL;
579 		buf[size-1] = 0;
580 
581 		vers = mesg;
582 		len = qword_get(&mesg, vers, size);
583 		if (len <= 0) return -EINVAL;
584 		do {
585 			enum vers_op cmd;
586 			unsigned minor;
587 			sign = *vers;
588 			if (sign == '+' || sign == '-')
589 				num = simple_strtol((vers+1), &minorp, 0);
590 			else
591 				num = simple_strtol(vers, &minorp, 0);
592 			if (*minorp == '.') {
593 				if (num != 4)
594 					return -EINVAL;
595 				if (kstrtouint(minorp+1, 0, &minor) < 0)
596 					return -EINVAL;
597 			}
598 
599 			cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
600 			switch(num) {
601 			case 2:
602 			case 3:
603 				nfsd_vers(nn, num, cmd);
604 				break;
605 			case 4:
606 				if (*minorp == '.') {
607 					if (nfsd_minorversion(nn, minor, cmd) < 0)
608 						return -EINVAL;
609 				} else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
610 					/*
611 					 * Either we have +4 and no minors are enabled,
612 					 * or we have -4 and at least one minor is enabled.
613 					 * In either case, propagate 'cmd' to all minors.
614 					 */
615 					minor = 0;
616 					while (nfsd_minorversion(nn, minor, cmd) >= 0)
617 						minor++;
618 				}
619 				break;
620 			default:
621 				return -EINVAL;
622 			}
623 			vers += len + 1;
624 		} while ((len = qword_get(&mesg, vers, size)) > 0);
625 		/* If all get turned off, turn them back on, as
626 		 * having no versions is BAD
627 		 */
628 		nfsd_reset_versions(nn);
629 	}
630 
631 	/* Now write current state into reply buffer */
632 	len = 0;
633 	sep = "";
634 	remaining = SIMPLE_TRANSACTION_LIMIT;
635 	for (num=2 ; num <= 4 ; num++) {
636 		int minor;
637 		if (!nfsd_vers(nn, num, NFSD_AVAIL))
638 			continue;
639 
640 		minor = -1;
641 		do {
642 			len = nfsd_print_version_support(nn, buf, remaining,
643 					sep, num, minor);
644 			if (len >= remaining)
645 				goto out;
646 			remaining -= len;
647 			buf += len;
648 			tlen += len;
649 			minor++;
650 			if (len)
651 				sep = " ";
652 		} while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
653 	}
654 out:
655 	len = snprintf(buf, remaining, "\n");
656 	if (len >= remaining)
657 		return -EINVAL;
658 	return tlen + len;
659 }
660 
661 /**
662  * write_versions - Set or report the available NFS protocol versions
663  *
664  * Input:
665  *			buf:		ignored
666  *			size:		zero
667  * Output:
668  *	On success:	passed-in buffer filled with '\n'-terminated C
669  *			string containing positive or negative integer
670  *			values representing the current status of each
671  *			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  * OR
676  *
677  * Input:
678  * 			buf:		C string containing whitespace-
679  * 					separated positive or negative
680  * 					integer values representing NFS
681  * 					protocol versions to enable ("+n")
682  * 					or disable ("-n")
683  *			size:		non-zero length of C string in @buf
684  * Output:
685  *	On success:	status of zero or more protocol versions has
686  *			been updated; passed-in buffer filled with
687  *			'\n'-terminated C string containing positive
688  *			or negative integer values representing the
689  *			current status of each protocol version;
690  *			return code is the size in bytes of the string
691  *	On error:	return code is zero or a negative errno value
692  */
693 static ssize_t write_versions(struct file *file, char *buf, size_t size)
694 {
695 	ssize_t rv;
696 
697 	mutex_lock(&nfsd_mutex);
698 	rv = __write_versions(file, buf, size);
699 	mutex_unlock(&nfsd_mutex);
700 	return rv;
701 }
702 
703 /*
704  * Zero-length write.  Return a list of NFSD's current listener
705  * transports.
706  */
707 static ssize_t __write_ports_names(char *buf, struct net *net)
708 {
709 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
710 
711 	if (nn->nfsd_serv == NULL)
712 		return 0;
713 	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
714 }
715 
716 /*
717  * A single 'fd' number was written, in which case it must be for
718  * a socket of a supported family/protocol, and we use it as an
719  * nfsd listener.
720  */
721 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
722 {
723 	char *mesg = buf;
724 	int fd, err;
725 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
726 
727 	err = get_int(&mesg, &fd);
728 	if (err != 0 || fd < 0)
729 		return -EINVAL;
730 
731 	if (svc_alien_sock(net, fd)) {
732 		printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
733 		return -EINVAL;
734 	}
735 
736 	err = nfsd_create_serv(net);
737 	if (err != 0)
738 		return err;
739 
740 	err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
741 	if (err < 0) {
742 		nfsd_destroy(net);
743 		return err;
744 	}
745 
746 	/* Decrease the count, but don't shut down the service */
747 	nn->nfsd_serv->sv_nrthreads--;
748 	return err;
749 }
750 
751 /*
752  * A transport listener is added by writing it's transport name and
753  * a port number.
754  */
755 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
756 {
757 	char transport[16];
758 	struct svc_xprt *xprt;
759 	int port, err;
760 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
761 
762 	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
763 		return -EINVAL;
764 
765 	if (port < 1 || port > USHRT_MAX)
766 		return -EINVAL;
767 
768 	err = nfsd_create_serv(net);
769 	if (err != 0)
770 		return err;
771 
772 	err = svc_create_xprt(nn->nfsd_serv, transport, net,
773 				PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
774 	if (err < 0)
775 		goto out_err;
776 
777 	err = svc_create_xprt(nn->nfsd_serv, transport, net,
778 				PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
779 	if (err < 0 && err != -EAFNOSUPPORT)
780 		goto out_close;
781 
782 	/* Decrease the count, but don't shut down the service */
783 	nn->nfsd_serv->sv_nrthreads--;
784 	return 0;
785 out_close:
786 	xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
787 	if (xprt != NULL) {
788 		svc_close_xprt(xprt);
789 		svc_xprt_put(xprt);
790 	}
791 out_err:
792 	nfsd_destroy(net);
793 	return err;
794 }
795 
796 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
797 			     struct net *net)
798 {
799 	if (size == 0)
800 		return __write_ports_names(buf, net);
801 
802 	if (isdigit(buf[0]))
803 		return __write_ports_addfd(buf, net, file->f_cred);
804 
805 	if (isalpha(buf[0]))
806 		return __write_ports_addxprt(buf, net, file->f_cred);
807 
808 	return -EINVAL;
809 }
810 
811 /**
812  * write_ports - Pass a socket file descriptor or transport name to listen on
813  *
814  * Input:
815  *			buf:		ignored
816  *			size:		zero
817  * Output:
818  *	On success:	passed-in buffer filled with a '\n'-terminated C
819  *			string containing a whitespace-separated list of
820  *			named NFSD listeners;
821  *			return code is the size in bytes of the string
822  *	On error:	return code is zero or a negative errno value
823  *
824  * OR
825  *
826  * Input:
827  *			buf:		C string containing an unsigned
828  *					integer value representing a bound
829  *					but unconnected socket that is to be
830  *					used as an NFSD listener; listen(3)
831  *					must be called for a SOCK_STREAM
832  *					socket, otherwise it is ignored
833  *			size:		non-zero length of C string in @buf
834  * Output:
835  *	On success:	NFS service is started;
836  *			passed-in buffer filled with a '\n'-terminated C
837  *			string containing a unique alphanumeric name of
838  *			the listener;
839  *			return code is the size in bytes of the string
840  *	On error:	return code is a negative errno value
841  *
842  * OR
843  *
844  * Input:
845  *			buf:		C string containing a transport
846  *					name and an unsigned integer value
847  *					representing the port to listen on,
848  *					separated by whitespace
849  *			size:		non-zero length of C string in @buf
850  * Output:
851  *	On success:	returns zero; NFS service is started
852  *	On error:	return code is a negative errno value
853  */
854 static ssize_t write_ports(struct file *file, char *buf, size_t size)
855 {
856 	ssize_t rv;
857 
858 	mutex_lock(&nfsd_mutex);
859 	rv = __write_ports(file, buf, size, netns(file));
860 	mutex_unlock(&nfsd_mutex);
861 	return rv;
862 }
863 
864 
865 int nfsd_max_blksize;
866 
867 /**
868  * write_maxblksize - Set or report the current NFS blksize
869  *
870  * Input:
871  *			buf:		ignored
872  *			size:		zero
873  *
874  * OR
875  *
876  * Input:
877  * 			buf:		C string containing an unsigned
878  * 					integer value representing the new
879  * 					NFS blksize
880  *			size:		non-zero length of C string in @buf
881  * Output:
882  *	On success:	passed-in buffer filled with '\n'-terminated C string
883  *			containing numeric value of the current NFS blksize
884  *			setting;
885  *			return code is the size in bytes of the string
886  *	On error:	return code is zero or a negative errno value
887  */
888 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
889 {
890 	char *mesg = buf;
891 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
892 
893 	if (size > 0) {
894 		int bsize;
895 		int rv = get_int(&mesg, &bsize);
896 		if (rv)
897 			return rv;
898 		/* force bsize into allowed range and
899 		 * required alignment.
900 		 */
901 		bsize = max_t(int, bsize, 1024);
902 		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
903 		bsize &= ~(1024-1);
904 		mutex_lock(&nfsd_mutex);
905 		if (nn->nfsd_serv) {
906 			mutex_unlock(&nfsd_mutex);
907 			return -EBUSY;
908 		}
909 		nfsd_max_blksize = bsize;
910 		mutex_unlock(&nfsd_mutex);
911 	}
912 
913 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
914 							nfsd_max_blksize);
915 }
916 
917 /**
918  * write_maxconn - Set or report the current max number of connections
919  *
920  * Input:
921  *			buf:		ignored
922  *			size:		zero
923  * OR
924  *
925  * Input:
926  * 			buf:		C string containing an unsigned
927  * 					integer value representing the new
928  * 					number of max connections
929  *			size:		non-zero length of C string in @buf
930  * Output:
931  *	On success:	passed-in buffer filled with '\n'-terminated C string
932  *			containing numeric value of max_connections setting
933  *			for this net namespace;
934  *			return code is the size in bytes of the string
935  *	On error:	return code is zero or a negative errno value
936  */
937 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
938 {
939 	char *mesg = buf;
940 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
941 	unsigned int maxconn = nn->max_connections;
942 
943 	if (size > 0) {
944 		int rv = get_uint(&mesg, &maxconn);
945 
946 		if (rv)
947 			return rv;
948 		nn->max_connections = maxconn;
949 	}
950 
951 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
952 }
953 
954 #ifdef CONFIG_NFSD_V4
955 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
956 				  time_t *time, struct nfsd_net *nn)
957 {
958 	char *mesg = buf;
959 	int rv, i;
960 
961 	if (size > 0) {
962 		if (nn->nfsd_serv)
963 			return -EBUSY;
964 		rv = get_int(&mesg, &i);
965 		if (rv)
966 			return rv;
967 		/*
968 		 * Some sanity checking.  We don't have a reason for
969 		 * these particular numbers, but problems with the
970 		 * extremes are:
971 		 *	- Too short: the briefest network outage may
972 		 *	  cause clients to lose all their locks.  Also,
973 		 *	  the frequent polling may be wasteful.
974 		 *	- Too long: do you really want reboot recovery
975 		 *	  to take more than an hour?  Or to make other
976 		 *	  clients wait an hour before being able to
977 		 *	  revoke a dead client's locks?
978 		 */
979 		if (i < 10 || i > 3600)
980 			return -EINVAL;
981 		*time = i;
982 	}
983 
984 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%ld\n", *time);
985 }
986 
987 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
988 				time_t *time, struct nfsd_net *nn)
989 {
990 	ssize_t rv;
991 
992 	mutex_lock(&nfsd_mutex);
993 	rv = __nfsd4_write_time(file, buf, size, time, nn);
994 	mutex_unlock(&nfsd_mutex);
995 	return rv;
996 }
997 
998 /**
999  * write_leasetime - Set or report the current NFSv4 lease time
1000  *
1001  * Input:
1002  *			buf:		ignored
1003  *			size:		zero
1004  *
1005  * OR
1006  *
1007  * Input:
1008  *			buf:		C string containing an unsigned
1009  *					integer value representing the new
1010  *					NFSv4 lease expiry time
1011  *			size:		non-zero length of C string in @buf
1012  * Output:
1013  *	On success:	passed-in buffer filled with '\n'-terminated C
1014  *			string containing unsigned integer value of the
1015  *			current lease expiry time;
1016  *			return code is the size in bytes of the string
1017  *	On error:	return code is zero or a negative errno value
1018  */
1019 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1020 {
1021 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1022 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1023 }
1024 
1025 /**
1026  * write_gracetime - Set or report current NFSv4 grace period time
1027  *
1028  * As above, but sets the time of the NFSv4 grace period.
1029  *
1030  * Note this should never be set to less than the *previous*
1031  * lease-period time, but we don't try to enforce this.  (In the common
1032  * case (a new boot), we don't know what the previous lease time was
1033  * anyway.)
1034  */
1035 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1036 {
1037 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1038 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1039 }
1040 
1041 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1042 				   struct nfsd_net *nn)
1043 {
1044 	char *mesg = buf;
1045 	char *recdir;
1046 	int len, status;
1047 
1048 	if (size > 0) {
1049 		if (nn->nfsd_serv)
1050 			return -EBUSY;
1051 		if (size > PATH_MAX || buf[size-1] != '\n')
1052 			return -EINVAL;
1053 		buf[size-1] = 0;
1054 
1055 		recdir = mesg;
1056 		len = qword_get(&mesg, recdir, size);
1057 		if (len <= 0)
1058 			return -EINVAL;
1059 
1060 		status = nfs4_reset_recoverydir(recdir);
1061 		if (status)
1062 			return status;
1063 	}
1064 
1065 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1066 							nfs4_recoverydir());
1067 }
1068 
1069 /**
1070  * write_recoverydir - Set or report the pathname of the recovery directory
1071  *
1072  * Input:
1073  *			buf:		ignored
1074  *			size:		zero
1075  *
1076  * OR
1077  *
1078  * Input:
1079  *			buf:		C string containing the pathname
1080  *					of the directory on a local file
1081  *					system containing permanent NFSv4
1082  *					recovery data
1083  *			size:		non-zero length of C string in @buf
1084  * Output:
1085  *	On success:	passed-in buffer filled with '\n'-terminated C string
1086  *			containing the current recovery pathname setting;
1087  *			return code is the size in bytes of the string
1088  *	On error:	return code is zero or a negative errno value
1089  */
1090 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1091 {
1092 	ssize_t rv;
1093 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1094 
1095 	mutex_lock(&nfsd_mutex);
1096 	rv = __write_recoverydir(file, buf, size, nn);
1097 	mutex_unlock(&nfsd_mutex);
1098 	return rv;
1099 }
1100 
1101 /**
1102  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1103  *
1104  * Input:
1105  *			buf:		ignored
1106  *			size:		zero
1107  * OR
1108  *
1109  * Input:
1110  * 			buf:		any value
1111  *			size:		non-zero length of C string in @buf
1112  * Output:
1113  *			passed-in buffer filled with "Y" or "N" with a newline
1114  *			and NULL-terminated C string. This indicates whether
1115  *			the grace period has ended in the current net
1116  *			namespace. Return code is the size in bytes of the
1117  *			string. Writing a string that starts with 'Y', 'y', or
1118  *			'1' to the file will end the grace period for nfsd's v4
1119  *			lock manager.
1120  */
1121 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1122 {
1123 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1124 
1125 	if (size > 0) {
1126 		switch(buf[0]) {
1127 		case 'Y':
1128 		case 'y':
1129 		case '1':
1130 			if (!nn->nfsd_serv)
1131 				return -EBUSY;
1132 			nfsd4_end_grace(nn);
1133 			break;
1134 		default:
1135 			return -EINVAL;
1136 		}
1137 	}
1138 
1139 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1140 			 nn->grace_ended ? 'Y' : 'N');
1141 }
1142 
1143 #endif
1144 
1145 /*----------------------------------------------------------------------------*/
1146 /*
1147  *	populating the filesystem.
1148  */
1149 
1150 static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
1151 {
1152 	static const struct tree_descr nfsd_files[] = {
1153 		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1154 		[NFSD_Export_features] = {"export_features",
1155 					&export_features_operations, S_IRUGO},
1156 		[NFSD_FO_UnlockIP] = {"unlock_ip",
1157 					&transaction_ops, S_IWUSR|S_IRUSR},
1158 		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1159 					&transaction_ops, S_IWUSR|S_IRUSR},
1160 		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1161 		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1162 		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1163 		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1164 		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
1165 		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1166 		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1167 		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1168 		[NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1169 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1170 		[NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1171 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1172 #ifdef CONFIG_NFSD_V4
1173 		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1174 		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1175 		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1176 		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1177 #endif
1178 		/* last one */ {""}
1179 	};
1180 	get_net(sb->s_fs_info);
1181 	return simple_fill_super(sb, 0x6e667364, nfsd_files);
1182 }
1183 
1184 static struct dentry *nfsd_mount(struct file_system_type *fs_type,
1185 	int flags, const char *dev_name, void *data)
1186 {
1187 	struct net *net = current->nsproxy->net_ns;
1188 	return mount_ns(fs_type, flags, data, net, net->user_ns, nfsd_fill_super);
1189 }
1190 
1191 static void nfsd_umount(struct super_block *sb)
1192 {
1193 	struct net *net = sb->s_fs_info;
1194 
1195 	kill_litter_super(sb);
1196 	put_net(net);
1197 }
1198 
1199 static struct file_system_type nfsd_fs_type = {
1200 	.owner		= THIS_MODULE,
1201 	.name		= "nfsd",
1202 	.mount		= nfsd_mount,
1203 	.kill_sb	= nfsd_umount,
1204 };
1205 MODULE_ALIAS_FS("nfsd");
1206 
1207 #ifdef CONFIG_PROC_FS
1208 static int create_proc_exports_entry(void)
1209 {
1210 	struct proc_dir_entry *entry;
1211 
1212 	entry = proc_mkdir("fs/nfs", NULL);
1213 	if (!entry)
1214 		return -ENOMEM;
1215 	entry = proc_create("exports", 0, entry,
1216 				 &exports_proc_operations);
1217 	if (!entry) {
1218 		remove_proc_entry("fs/nfs", NULL);
1219 		return -ENOMEM;
1220 	}
1221 	return 0;
1222 }
1223 #else /* CONFIG_PROC_FS */
1224 static int create_proc_exports_entry(void)
1225 {
1226 	return 0;
1227 }
1228 #endif
1229 
1230 unsigned int nfsd_net_id;
1231 
1232 static __net_init int nfsd_init_net(struct net *net)
1233 {
1234 	int retval;
1235 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1236 
1237 	retval = nfsd_export_init(net);
1238 	if (retval)
1239 		goto out_export_error;
1240 	retval = nfsd_idmap_init(net);
1241 	if (retval)
1242 		goto out_idmap_error;
1243 	nn->nfsd_versions = NULL;
1244 	nn->nfsd4_minorversions = NULL;
1245 	nn->nfsd4_lease = 90;	/* default lease time */
1246 	nn->nfsd4_grace = 90;
1247 	nn->somebody_reclaimed = false;
1248 	nn->track_reclaim_completes = false;
1249 	nn->clverifier_counter = prandom_u32();
1250 	nn->clientid_counter = prandom_u32();
1251 	nn->s2s_cp_cl_id = nn->clientid_counter++;
1252 
1253 	atomic_set(&nn->ntf_refcnt, 0);
1254 	init_waitqueue_head(&nn->ntf_wq);
1255 	return 0;
1256 
1257 out_idmap_error:
1258 	nfsd_export_shutdown(net);
1259 out_export_error:
1260 	return retval;
1261 }
1262 
1263 static __net_exit void nfsd_exit_net(struct net *net)
1264 {
1265 	nfsd_idmap_shutdown(net);
1266 	nfsd_export_shutdown(net);
1267 	nfsd_netns_free_versions(net_generic(net, nfsd_net_id));
1268 }
1269 
1270 static struct pernet_operations nfsd_net_ops = {
1271 	.init = nfsd_init_net,
1272 	.exit = nfsd_exit_net,
1273 	.id   = &nfsd_net_id,
1274 	.size = sizeof(struct nfsd_net),
1275 };
1276 
1277 static int __init init_nfsd(void)
1278 {
1279 	int retval;
1280 	printk(KERN_INFO "Installing knfsd (copyright (C) 1996 okir@monad.swb.de).\n");
1281 
1282 	retval = register_pernet_subsys(&nfsd_net_ops);
1283 	if (retval < 0)
1284 		return retval;
1285 	retval = register_cld_notifier();
1286 	if (retval)
1287 		goto out_unregister_pernet;
1288 	retval = nfsd4_init_slabs();
1289 	if (retval)
1290 		goto out_unregister_notifier;
1291 	retval = nfsd4_init_pnfs();
1292 	if (retval)
1293 		goto out_free_slabs;
1294 	retval = nfsd_fault_inject_init(); /* nfsd fault injection controls */
1295 	if (retval)
1296 		goto out_exit_pnfs;
1297 	nfsd_stat_init();	/* Statistics */
1298 	retval = nfsd_reply_cache_init();
1299 	if (retval)
1300 		goto out_free_stat;
1301 	nfsd_lockd_init();	/* lockd->nfsd callbacks */
1302 	retval = create_proc_exports_entry();
1303 	if (retval)
1304 		goto out_free_lockd;
1305 	retval = register_filesystem(&nfsd_fs_type);
1306 	if (retval)
1307 		goto out_free_all;
1308 	return 0;
1309 out_free_all:
1310 	remove_proc_entry("fs/nfs/exports", NULL);
1311 	remove_proc_entry("fs/nfs", NULL);
1312 out_free_lockd:
1313 	nfsd_lockd_shutdown();
1314 	nfsd_reply_cache_shutdown();
1315 out_free_stat:
1316 	nfsd_stat_shutdown();
1317 	nfsd_fault_inject_cleanup();
1318 out_exit_pnfs:
1319 	nfsd4_exit_pnfs();
1320 out_free_slabs:
1321 	nfsd4_free_slabs();
1322 out_unregister_notifier:
1323 	unregister_cld_notifier();
1324 out_unregister_pernet:
1325 	unregister_pernet_subsys(&nfsd_net_ops);
1326 	return retval;
1327 }
1328 
1329 static void __exit exit_nfsd(void)
1330 {
1331 	nfsd_reply_cache_shutdown();
1332 	remove_proc_entry("fs/nfs/exports", NULL);
1333 	remove_proc_entry("fs/nfs", NULL);
1334 	nfsd_stat_shutdown();
1335 	nfsd_lockd_shutdown();
1336 	nfsd4_free_slabs();
1337 	nfsd4_exit_pnfs();
1338 	nfsd_fault_inject_cleanup();
1339 	unregister_filesystem(&nfsd_fs_type);
1340 	unregister_cld_notifier();
1341 	unregister_pernet_subsys(&nfsd_net_ops);
1342 }
1343 
1344 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1345 MODULE_LICENSE("GPL");
1346 module_init(init_nfsd)
1347 module_exit(exit_nfsd)
1348