xref: /linux/fs/nfsd/nfsctl.c (revision 3daab3112f039cf849f96764019b096bb0a39d04)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Syscall interface to knfsd.
4  *
5  * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
6  */
7 
8 #include <linux/slab.h>
9 #include <linux/namei.h>
10 #include <linux/ctype.h>
11 #include <linux/fs_context.h>
12 
13 #include <linux/sunrpc/svcsock.h>
14 #include <linux/lockd/lockd.h>
15 #include <linux/sunrpc/addr.h>
16 #include <linux/sunrpc/gss_api.h>
17 #include <linux/sunrpc/rpc_pipe_fs.h>
18 #include <linux/sunrpc/svc.h>
19 #include <linux/module.h>
20 #include <linux/fsnotify.h>
21 #include <linux/nfslocalio.h>
22 
23 #include "idmap.h"
24 #include "nfsd.h"
25 #include "cache.h"
26 #include "state.h"
27 #include "netns.h"
28 #include "pnfs.h"
29 #include "filecache.h"
30 #include "trace.h"
31 #include "netlink.h"
32 
33 /*
34  *	We have a single directory with several nodes in it.
35  */
36 enum {
37 	NFSD_Root = 1,
38 	NFSD_List,
39 	NFSD_Export_Stats,
40 	NFSD_Export_features,
41 	NFSD_Fh,
42 	NFSD_FO_UnlockIP,
43 	NFSD_FO_UnlockFS,
44 	NFSD_Threads,
45 	NFSD_Pool_Threads,
46 	NFSD_Pool_Stats,
47 	NFSD_Reply_Cache_Stats,
48 	NFSD_Versions,
49 	NFSD_Ports,
50 	NFSD_MaxBlkSize,
51 	NFSD_Filecache,
52 	NFSD_Leasetime,
53 	NFSD_Gracetime,
54 	NFSD_RecoveryDir,
55 	NFSD_V4EndGrace,
56 	NFSD_MaxReserved
57 };
58 
59 /*
60  * write() for these nodes.
61  */
62 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
63 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
64 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
65 static ssize_t write_threads(struct file *file, char *buf, size_t size);
66 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
67 static ssize_t write_versions(struct file *file, char *buf, size_t size);
68 static ssize_t write_ports(struct file *file, char *buf, size_t size);
69 static ssize_t write_maxblksize(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 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
74 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
75 #endif
76 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
77 #endif
78 
79 static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
80 	[NFSD_Fh] = write_filehandle,
81 	[NFSD_FO_UnlockIP] = write_unlock_ip,
82 	[NFSD_FO_UnlockFS] = write_unlock_fs,
83 	[NFSD_Threads] = write_threads,
84 	[NFSD_Pool_Threads] = write_pool_threads,
85 	[NFSD_Versions] = write_versions,
86 	[NFSD_Ports] = write_ports,
87 	[NFSD_MaxBlkSize] = write_maxblksize,
88 #ifdef CONFIG_NFSD_V4
89 	[NFSD_Leasetime] = write_leasetime,
90 	[NFSD_Gracetime] = write_gracetime,
91 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
92 	[NFSD_RecoveryDir] = write_recoverydir,
93 #endif
94 	[NFSD_V4EndGrace] = write_v4_end_grace,
95 #endif
96 };
97 
98 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
99 {
100 	ino_t ino =  file_inode(file)->i_ino;
101 	char *data;
102 	ssize_t rv;
103 
104 	if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
105 		return -EINVAL;
106 
107 	data = simple_transaction_get(file, buf, size);
108 	if (IS_ERR(data))
109 		return PTR_ERR(data);
110 
111 	rv = write_op[ino](file, data, size);
112 	if (rv < 0)
113 		return rv;
114 
115 	simple_transaction_set(file, rv);
116 	return size;
117 }
118 
119 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
120 {
121 	if (! file->private_data) {
122 		/* An attempt to read a transaction file without writing
123 		 * causes a 0-byte write so that the file can return
124 		 * state information
125 		 */
126 		ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
127 		if (rv < 0)
128 			return rv;
129 	}
130 	return simple_transaction_read(file, buf, size, pos);
131 }
132 
133 static const struct file_operations transaction_ops = {
134 	.write		= nfsctl_transaction_write,
135 	.read		= nfsctl_transaction_read,
136 	.release	= simple_transaction_release,
137 	.llseek		= default_llseek,
138 };
139 
140 static int exports_net_open(struct net *net, struct file *file)
141 {
142 	int err;
143 	struct seq_file *seq;
144 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
145 
146 	err = seq_open(file, &nfs_exports_op);
147 	if (err)
148 		return err;
149 
150 	seq = file->private_data;
151 	seq->private = nn->svc_export_cache;
152 	return 0;
153 }
154 
155 static int exports_nfsd_open(struct inode *inode, struct file *file)
156 {
157 	return exports_net_open(inode->i_sb->s_fs_info, file);
158 }
159 
160 static const struct file_operations exports_nfsd_operations = {
161 	.open		= exports_nfsd_open,
162 	.read		= seq_read,
163 	.llseek		= seq_lseek,
164 	.release	= seq_release,
165 };
166 
167 static int export_features_show(struct seq_file *m, void *v)
168 {
169 	seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
170 	return 0;
171 }
172 
173 DEFINE_SHOW_ATTRIBUTE(export_features);
174 
175 static int nfsd_pool_stats_open(struct inode *inode, struct file *file)
176 {
177 	struct nfsd_net *nn = net_generic(inode->i_sb->s_fs_info, nfsd_net_id);
178 
179 	return svc_pool_stats_open(&nn->nfsd_info, file);
180 }
181 
182 static const struct file_operations pool_stats_operations = {
183 	.open		= nfsd_pool_stats_open,
184 	.read		= seq_read,
185 	.llseek		= seq_lseek,
186 	.release	= seq_release,
187 };
188 
189 DEFINE_SHOW_ATTRIBUTE(nfsd_reply_cache_stats);
190 
191 DEFINE_SHOW_ATTRIBUTE(nfsd_file_cache_stats);
192 
193 /*----------------------------------------------------------------------------*/
194 /*
195  * payload - write methods
196  */
197 
198 static inline struct net *netns(struct file *file)
199 {
200 	return file_inode(file)->i_sb->s_fs_info;
201 }
202 
203 /*
204  * write_unlock_ip - Release all locks used by a client
205  *
206  * Experimental.
207  *
208  * Input:
209  *			buf:	'\n'-terminated C string containing a
210  *				presentation format IP address
211  *			size:	length of C string in @buf
212  * Output:
213  *	On success:	returns zero if all specified locks were released;
214  *			returns one if one or more locks were not released
215  *	On error:	return code is negative errno value
216  */
217 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
218 {
219 	struct sockaddr_storage address;
220 	struct sockaddr *sap = (struct sockaddr *)&address;
221 	size_t salen = sizeof(address);
222 	char *fo_path;
223 	struct net *net = netns(file);
224 
225 	/* sanity check */
226 	if (size == 0)
227 		return -EINVAL;
228 
229 	if (buf[size-1] != '\n')
230 		return -EINVAL;
231 
232 	fo_path = buf;
233 	if (qword_get(&buf, fo_path, size) < 0)
234 		return -EINVAL;
235 
236 	if (rpc_pton(net, fo_path, size, sap, salen) == 0)
237 		return -EINVAL;
238 
239 	trace_nfsd_ctl_unlock_ip(net, buf);
240 	return nlmsvc_unlock_all_by_ip(sap);
241 }
242 
243 /*
244  * write_unlock_fs - Release all locks on a local file system
245  *
246  * Experimental.
247  *
248  * Input:
249  *			buf:	'\n'-terminated C string containing the
250  *				absolute pathname of a local file system
251  *			size:	length of C string in @buf
252  * Output:
253  *	On success:	returns zero if all specified locks were released;
254  *			returns one if one or more locks were not released
255  *	On error:	return code is negative errno value
256  */
257 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
258 {
259 	struct path path;
260 	char *fo_path;
261 	int error;
262 	struct nfsd_net *nn;
263 
264 	/* sanity check */
265 	if (size == 0)
266 		return -EINVAL;
267 
268 	if (buf[size-1] != '\n')
269 		return -EINVAL;
270 
271 	fo_path = buf;
272 	if (qword_get(&buf, fo_path, size) < 0)
273 		return -EINVAL;
274 	trace_nfsd_ctl_unlock_fs(netns(file), fo_path);
275 	error = kern_path(fo_path, 0, &path);
276 	if (error)
277 		return error;
278 
279 	/*
280 	 * XXX: Needs better sanity checking.  Otherwise we could end up
281 	 * releasing locks on the wrong file system.
282 	 *
283 	 * For example:
284 	 * 1.  Does the path refer to a directory?
285 	 * 2.  Is that directory a mount point, or
286 	 * 3.  Is that directory the root of an exported file system?
287 	 */
288 	nfsd4_cancel_copy_by_sb(netns(file), path.dentry->d_sb);
289 	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
290 	mutex_lock(&nfsd_mutex);
291 	nn = net_generic(netns(file), nfsd_net_id);
292 	if (nn->nfsd_serv)
293 		nfsd4_revoke_states(nn, path.dentry->d_sb);
294 	else
295 		error = -EINVAL;
296 	mutex_unlock(&nfsd_mutex);
297 
298 	path_put(&path);
299 	return error;
300 }
301 
302 /*
303  * write_filehandle - Get a variable-length NFS file handle by path
304  *
305  * On input, the buffer contains a '\n'-terminated C string comprised of
306  * three alphanumeric words separated by whitespace.  The string may
307  * contain escape sequences.
308  *
309  * Input:
310  *			buf:
311  *				domain:		client domain name
312  *				path:		export pathname
313  *				maxsize:	numeric maximum size of
314  *						@buf
315  *			size:	length of C string in @buf
316  * Output:
317  *	On success:	passed-in buffer filled with '\n'-terminated C
318  *			string containing a ASCII hex text version
319  *			of the NFS file handle;
320  *			return code is the size in bytes of the string
321  *	On error:	return code is negative errno value
322  */
323 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
324 {
325 	char *dname, *path;
326 	int maxsize;
327 	char *mesg = buf;
328 	int len;
329 	struct auth_domain *dom;
330 	struct knfsd_fh fh;
331 
332 	if (size == 0)
333 		return -EINVAL;
334 
335 	if (buf[size-1] != '\n')
336 		return -EINVAL;
337 	buf[size-1] = 0;
338 
339 	dname = mesg;
340 	len = qword_get(&mesg, dname, size);
341 	if (len <= 0)
342 		return -EINVAL;
343 
344 	path = dname+len+1;
345 	len = qword_get(&mesg, path, size);
346 	if (len <= 0)
347 		return -EINVAL;
348 
349 	len = get_int(&mesg, &maxsize);
350 	if (len)
351 		return len;
352 
353 	if (maxsize < NFS_FHSIZE)
354 		return -EINVAL;
355 	maxsize = min(maxsize, NFS3_FHSIZE);
356 
357 	if (qword_get(&mesg, mesg, size) > 0)
358 		return -EINVAL;
359 
360 	trace_nfsd_ctl_filehandle(netns(file), dname, path, maxsize);
361 
362 	/* we have all the words, they are in buf.. */
363 	dom = unix_domain_find(dname);
364 	if (!dom)
365 		return -ENOMEM;
366 
367 	len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
368 	auth_domain_put(dom);
369 	if (len)
370 		return len;
371 
372 	mesg = buf;
373 	len = SIMPLE_TRANSACTION_LIMIT;
374 	qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
375 	mesg[-1] = '\n';
376 	return mesg - buf;
377 }
378 
379 /*
380  * write_threads - Start NFSD, or report the current number of running threads
381  *
382  * Input:
383  *			buf:		ignored
384  *			size:		zero
385  * Output:
386  *	On success:	passed-in buffer filled with '\n'-terminated C
387  *			string numeric value representing the number of
388  *			running NFSD threads;
389  *			return code is the size in bytes of the string
390  *	On error:	return code is zero
391  *
392  * OR
393  *
394  * Input:
395  *			buf:		C string containing an unsigned
396  *					integer value representing the
397  *					number of NFSD threads to start
398  *			size:		non-zero length of C string in @buf
399  * Output:
400  *	On success:	NFS service is started;
401  *			passed-in buffer filled with '\n'-terminated C
402  *			string numeric value representing the number of
403  *			running NFSD threads;
404  *			return code is the size in bytes of the string
405  *	On error:	return code is zero or a negative errno value
406  */
407 static ssize_t write_threads(struct file *file, char *buf, size_t size)
408 {
409 	char *mesg = buf;
410 	int rv;
411 	struct net *net = netns(file);
412 
413 	if (size > 0) {
414 		int newthreads;
415 		rv = get_int(&mesg, &newthreads);
416 		if (rv)
417 			return rv;
418 		if (newthreads < 0)
419 			return -EINVAL;
420 		trace_nfsd_ctl_threads(net, newthreads);
421 		mutex_lock(&nfsd_mutex);
422 		rv = nfsd_svc(1, &newthreads, net, file->f_cred, NULL);
423 		mutex_unlock(&nfsd_mutex);
424 		if (rv < 0)
425 			return rv;
426 	} else
427 		rv = nfsd_nrthreads(net);
428 
429 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
430 }
431 
432 /*
433  * write_pool_threads - Set or report the current number of threads per pool
434  *
435  * Input:
436  *			buf:		ignored
437  *			size:		zero
438  *
439  * OR
440  *
441  * Input:
442  *			buf:		C string containing whitespace-
443  *					separated unsigned integer values
444  *					representing the number of NFSD
445  *					threads to start in each pool
446  *			size:		non-zero length of C string in @buf
447  * Output:
448  *	On success:	passed-in buffer filled with '\n'-terminated C
449  *			string containing integer values representing the
450  *			number of NFSD threads in each pool;
451  *			return code is the size in bytes of the string
452  *	On error:	return code is zero or a negative errno value
453  */
454 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
455 {
456 	/* if size > 0, look for an array of number of threads per node
457 	 * and apply them  then write out number of threads per node as reply
458 	 */
459 	char *mesg = buf;
460 	int i;
461 	int rv;
462 	int len;
463 	int npools;
464 	int *nthreads;
465 	struct net *net = netns(file);
466 
467 	mutex_lock(&nfsd_mutex);
468 	npools = nfsd_nrpools(net);
469 	if (npools == 0) {
470 		/*
471 		 * NFS is shut down.  The admin can start it by
472 		 * writing to the threads file but NOT the pool_threads
473 		 * file, sorry.  Report zero threads.
474 		 */
475 		mutex_unlock(&nfsd_mutex);
476 		strcpy(buf, "0\n");
477 		return strlen(buf);
478 	}
479 
480 	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
481 	rv = -ENOMEM;
482 	if (nthreads == NULL)
483 		goto out_free;
484 
485 	if (size > 0) {
486 		for (i = 0; i < npools; i++) {
487 			rv = get_int(&mesg, &nthreads[i]);
488 			if (rv == -ENOENT)
489 				break;		/* fewer numbers than pools */
490 			if (rv)
491 				goto out_free;	/* syntax error */
492 			rv = -EINVAL;
493 			if (nthreads[i] < 0)
494 				goto out_free;
495 			trace_nfsd_ctl_pool_threads(net, i, nthreads[i]);
496 		}
497 
498 		/*
499 		 * There must always be a thread in pool 0; the admin
500 		 * can't shut down NFS completely using pool_threads.
501 		 */
502 		if (nthreads[0] == 0)
503 			nthreads[0] = 1;
504 
505 		rv = nfsd_set_nrthreads(i, nthreads, net);
506 		if (rv)
507 			goto out_free;
508 	}
509 
510 	rv = nfsd_get_nrthreads(npools, nthreads, net);
511 	if (rv)
512 		goto out_free;
513 
514 	mesg = buf;
515 	size = SIMPLE_TRANSACTION_LIMIT;
516 	for (i = 0; i < npools && size > 0; i++) {
517 		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
518 		len = strlen(mesg);
519 		size -= len;
520 		mesg += len;
521 	}
522 	rv = mesg - buf;
523 out_free:
524 	kfree(nthreads);
525 	mutex_unlock(&nfsd_mutex);
526 	return rv;
527 }
528 
529 static ssize_t
530 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
531 		const char *sep, unsigned vers, int minor)
532 {
533 	const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
534 	bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
535 
536 	if (vers == 4 && minor >= 0 &&
537 	    !nfsd_minorversion(nn, minor, NFSD_TEST))
538 		supported = false;
539 	if (minor == 0 && supported)
540 		/*
541 		 * special case for backward compatability.
542 		 * +4.0 is never reported, it is implied by
543 		 * +4, unless -4.0 is present.
544 		 */
545 		return 0;
546 	return snprintf(buf, remaining, format, sep,
547 			supported ? '+' : '-', vers, minor);
548 }
549 
550 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
551 {
552 	char *mesg = buf;
553 	char *vers, *minorp, sign;
554 	int len, num, remaining;
555 	ssize_t tlen = 0;
556 	char *sep;
557 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
558 
559 	if (size > 0) {
560 		if (nn->nfsd_serv)
561 			/* Cannot change versions without updating
562 			 * nn->nfsd_serv->sv_xdrsize, and reallocing
563 			 * rq_argp and rq_resp
564 			 */
565 			return -EBUSY;
566 		if (buf[size-1] != '\n')
567 			return -EINVAL;
568 		buf[size-1] = 0;
569 		trace_nfsd_ctl_version(netns(file), buf);
570 
571 		vers = mesg;
572 		len = qword_get(&mesg, vers, size);
573 		if (len <= 0) return -EINVAL;
574 		do {
575 			enum vers_op cmd;
576 			unsigned minor;
577 			sign = *vers;
578 			if (sign == '+' || sign == '-')
579 				num = simple_strtol((vers+1), &minorp, 0);
580 			else
581 				num = simple_strtol(vers, &minorp, 0);
582 			if (*minorp == '.') {
583 				if (num != 4)
584 					return -EINVAL;
585 				if (kstrtouint(minorp+1, 0, &minor) < 0)
586 					return -EINVAL;
587 			}
588 
589 			cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
590 			switch(num) {
591 #ifdef CONFIG_NFSD_V2
592 			case 2:
593 #endif
594 			case 3:
595 				nfsd_vers(nn, num, cmd);
596 				break;
597 			case 4:
598 				if (*minorp == '.') {
599 					if (nfsd_minorversion(nn, minor, cmd) < 0)
600 						return -EINVAL;
601 				} else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
602 					/*
603 					 * Either we have +4 and no minors are enabled,
604 					 * or we have -4 and at least one minor is enabled.
605 					 * In either case, propagate 'cmd' to all minors.
606 					 */
607 					minor = 0;
608 					while (nfsd_minorversion(nn, minor, cmd) >= 0)
609 						minor++;
610 				}
611 				break;
612 			default:
613 				/* Ignore requests to disable non-existent versions */
614 				if (cmd == NFSD_SET)
615 					return -EINVAL;
616 			}
617 			vers += len + 1;
618 		} while ((len = qword_get(&mesg, vers, size)) > 0);
619 		/* If all get turned off, turn them back on, as
620 		 * having no versions is BAD
621 		 */
622 		nfsd_reset_versions(nn);
623 	}
624 
625 	/* Now write current state into reply buffer */
626 	sep = "";
627 	remaining = SIMPLE_TRANSACTION_LIMIT;
628 	for (num=2 ; num <= 4 ; num++) {
629 		int minor;
630 		if (!nfsd_vers(nn, num, NFSD_AVAIL))
631 			continue;
632 
633 		minor = -1;
634 		do {
635 			len = nfsd_print_version_support(nn, buf, remaining,
636 					sep, num, minor);
637 			if (len >= remaining)
638 				goto out;
639 			remaining -= len;
640 			buf += len;
641 			tlen += len;
642 			minor++;
643 			if (len)
644 				sep = " ";
645 		} while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
646 	}
647 out:
648 	len = snprintf(buf, remaining, "\n");
649 	if (len >= remaining)
650 		return -EINVAL;
651 	return tlen + len;
652 }
653 
654 /*
655  * write_versions - Set or report the available NFS protocol versions
656  *
657  * Input:
658  *			buf:		ignored
659  *			size:		zero
660  * Output:
661  *	On success:	passed-in buffer filled with '\n'-terminated C
662  *			string containing positive or negative integer
663  *			values representing the current status of each
664  *			protocol version;
665  *			return code is the size in bytes of the string
666  *	On error:	return code is zero or a negative errno value
667  *
668  * OR
669  *
670  * Input:
671  *			buf:		C string containing whitespace-
672  *					separated positive or negative
673  *					integer values representing NFS
674  *					protocol versions to enable ("+n")
675  *					or disable ("-n")
676  *			size:		non-zero length of C string in @buf
677  * Output:
678  *	On success:	status of zero or more protocol versions has
679  *			been updated; passed-in buffer filled with
680  *			'\n'-terminated C string containing positive
681  *			or negative integer values representing the
682  *			current status of each protocol version;
683  *			return code is the size in bytes of the string
684  *	On error:	return code is zero or a negative errno value
685  */
686 static ssize_t write_versions(struct file *file, char *buf, size_t size)
687 {
688 	ssize_t rv;
689 
690 	mutex_lock(&nfsd_mutex);
691 	rv = __write_versions(file, buf, size);
692 	mutex_unlock(&nfsd_mutex);
693 	return rv;
694 }
695 
696 /*
697  * Zero-length write.  Return a list of NFSD's current listener
698  * transports.
699  */
700 static ssize_t __write_ports_names(char *buf, struct net *net)
701 {
702 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
703 
704 	if (nn->nfsd_serv == NULL)
705 		return 0;
706 	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
707 }
708 
709 /*
710  * A single 'fd' number was written, in which case it must be for
711  * a socket of a supported family/protocol, and we use it as an
712  * nfsd listener.
713  */
714 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
715 {
716 	char *mesg = buf;
717 	int fd, err;
718 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
719 	struct svc_serv *serv;
720 
721 	err = get_int(&mesg, &fd);
722 	if (err != 0 || fd < 0)
723 		return -EINVAL;
724 	trace_nfsd_ctl_ports_addfd(net, fd);
725 
726 	err = nfsd_create_serv(net);
727 	if (err != 0)
728 		return err;
729 
730 	serv = nn->nfsd_serv;
731 	err = svc_addsock(serv, net, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
732 
733 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
734 		nfsd_destroy_serv(net);
735 
736 	return err;
737 }
738 
739 /*
740  * A transport listener is added by writing its transport name and
741  * a port number.
742  */
743 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
744 {
745 	char transport[16];
746 	struct svc_xprt *xprt;
747 	int port, err;
748 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
749 	struct svc_serv *serv;
750 
751 	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
752 		return -EINVAL;
753 
754 	if (port < 1 || port > USHRT_MAX)
755 		return -EINVAL;
756 	trace_nfsd_ctl_ports_addxprt(net, transport, port);
757 
758 	err = nfsd_create_serv(net);
759 	if (err != 0)
760 		return err;
761 
762 	serv = nn->nfsd_serv;
763 	err = svc_xprt_create(serv, transport, net,
764 			      PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
765 	if (err < 0)
766 		goto out_err;
767 
768 	err = svc_xprt_create(serv, transport, net,
769 			      PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
770 	if (err < 0 && err != -EAFNOSUPPORT)
771 		goto out_close;
772 
773 	return 0;
774 out_close:
775 	xprt = svc_find_xprt(serv, transport, net, PF_INET, port);
776 	if (xprt != NULL) {
777 		svc_xprt_close(xprt);
778 		svc_xprt_put(xprt);
779 	}
780 out_err:
781 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
782 		nfsd_destroy_serv(net);
783 
784 	return err;
785 }
786 
787 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
788 			     struct net *net)
789 {
790 	if (size == 0)
791 		return __write_ports_names(buf, net);
792 
793 	if (isdigit(buf[0]))
794 		return __write_ports_addfd(buf, net, file->f_cred);
795 
796 	if (isalpha(buf[0]))
797 		return __write_ports_addxprt(buf, net, file->f_cred);
798 
799 	return -EINVAL;
800 }
801 
802 /*
803  * write_ports - Pass a socket file descriptor or transport name to listen on
804  *
805  * Input:
806  *			buf:		ignored
807  *			size:		zero
808  * Output:
809  *	On success:	passed-in buffer filled with a '\n'-terminated C
810  *			string containing a whitespace-separated list of
811  *			named NFSD listeners;
812  *			return code is the size in bytes of the string
813  *	On error:	return code is zero or a negative errno value
814  *
815  * OR
816  *
817  * Input:
818  *			buf:		C string containing an unsigned
819  *					integer value representing a bound
820  *					but unconnected socket that is to be
821  *					used as an NFSD listener; listen(3)
822  *					must be called for a SOCK_STREAM
823  *					socket, otherwise it is ignored
824  *			size:		non-zero length of C string in @buf
825  * Output:
826  *	On success:	NFS service is started;
827  *			passed-in buffer filled with a '\n'-terminated C
828  *			string containing a unique alphanumeric name of
829  *			the listener;
830  *			return code is the size in bytes of the string
831  *	On error:	return code is a negative errno value
832  *
833  * OR
834  *
835  * Input:
836  *			buf:		C string containing a transport
837  *					name and an unsigned integer value
838  *					representing the port to listen on,
839  *					separated by whitespace
840  *			size:		non-zero length of C string in @buf
841  * Output:
842  *	On success:	returns zero; NFS service is started
843  *	On error:	return code is a negative errno value
844  */
845 static ssize_t write_ports(struct file *file, char *buf, size_t size)
846 {
847 	ssize_t rv;
848 
849 	mutex_lock(&nfsd_mutex);
850 	rv = __write_ports(file, buf, size, netns(file));
851 	mutex_unlock(&nfsd_mutex);
852 	return rv;
853 }
854 
855 
856 int nfsd_max_blksize;
857 
858 /*
859  * write_maxblksize - Set or report the current NFS blksize
860  *
861  * Input:
862  *			buf:		ignored
863  *			size:		zero
864  *
865  * OR
866  *
867  * Input:
868  *			buf:		C string containing an unsigned
869  *					integer value representing the new
870  *					NFS blksize
871  *			size:		non-zero length of C string in @buf
872  * Output:
873  *	On success:	passed-in buffer filled with '\n'-terminated C string
874  *			containing numeric value of the current NFS blksize
875  *			setting;
876  *			return code is the size in bytes of the string
877  *	On error:	return code is zero or a negative errno value
878  */
879 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
880 {
881 	char *mesg = buf;
882 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
883 
884 	if (size > 0) {
885 		int bsize;
886 		int rv = get_int(&mesg, &bsize);
887 		if (rv)
888 			return rv;
889 		trace_nfsd_ctl_maxblksize(netns(file), bsize);
890 
891 		/* force bsize into allowed range and
892 		 * required alignment.
893 		 */
894 		bsize = max_t(int, bsize, 1024);
895 		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
896 		bsize &= ~(1024-1);
897 		mutex_lock(&nfsd_mutex);
898 		if (nn->nfsd_serv) {
899 			mutex_unlock(&nfsd_mutex);
900 			return -EBUSY;
901 		}
902 		nfsd_max_blksize = bsize;
903 		mutex_unlock(&nfsd_mutex);
904 	}
905 
906 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
907 							nfsd_max_blksize);
908 }
909 
910 #ifdef CONFIG_NFSD_V4
911 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
912 				  time64_t *time, struct nfsd_net *nn)
913 {
914 	struct dentry *dentry = file_dentry(file);
915 	char *mesg = buf;
916 	int rv, i;
917 
918 	if (size > 0) {
919 		if (nn->nfsd_serv)
920 			return -EBUSY;
921 		rv = get_int(&mesg, &i);
922 		if (rv)
923 			return rv;
924 		trace_nfsd_ctl_time(netns(file), dentry->d_name.name,
925 				    dentry->d_name.len, i);
926 
927 		/*
928 		 * Some sanity checking.  We don't have a reason for
929 		 * these particular numbers, but problems with the
930 		 * extremes are:
931 		 *	- Too short: the briefest network outage may
932 		 *	  cause clients to lose all their locks.  Also,
933 		 *	  the frequent polling may be wasteful.
934 		 *	- Too long: do you really want reboot recovery
935 		 *	  to take more than an hour?  Or to make other
936 		 *	  clients wait an hour before being able to
937 		 *	  revoke a dead client's locks?
938 		 */
939 		if (i < 10 || i > 3600)
940 			return -EINVAL;
941 		*time = i;
942 	}
943 
944 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
945 }
946 
947 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
948 				time64_t *time, struct nfsd_net *nn)
949 {
950 	ssize_t rv;
951 
952 	mutex_lock(&nfsd_mutex);
953 	rv = __nfsd4_write_time(file, buf, size, time, nn);
954 	mutex_unlock(&nfsd_mutex);
955 	return rv;
956 }
957 
958 /*
959  * write_leasetime - Set or report the current NFSv4 lease time
960  *
961  * Input:
962  *			buf:		ignored
963  *			size:		zero
964  *
965  * OR
966  *
967  * Input:
968  *			buf:		C string containing an unsigned
969  *					integer value representing the new
970  *					NFSv4 lease expiry time
971  *			size:		non-zero length of C string in @buf
972  * Output:
973  *	On success:	passed-in buffer filled with '\n'-terminated C
974  *			string containing unsigned integer value of the
975  *			current lease expiry time;
976  *			return code is the size in bytes of the string
977  *	On error:	return code is zero or a negative errno value
978  */
979 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
980 {
981 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
982 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
983 }
984 
985 /*
986  * write_gracetime - Set or report current NFSv4 grace period time
987  *
988  * As above, but sets the time of the NFSv4 grace period.
989  *
990  * Note this should never be set to less than the *previous*
991  * lease-period time, but we don't try to enforce this.  (In the common
992  * case (a new boot), we don't know what the previous lease time was
993  * anyway.)
994  */
995 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
996 {
997 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
998 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
999 }
1000 
1001 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1002 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1003 				   struct nfsd_net *nn)
1004 {
1005 	char *mesg = buf;
1006 	char *recdir;
1007 	int len, status;
1008 
1009 	if (size > 0) {
1010 		if (nn->nfsd_serv)
1011 			return -EBUSY;
1012 		if (size > PATH_MAX || buf[size-1] != '\n')
1013 			return -EINVAL;
1014 		buf[size-1] = 0;
1015 
1016 		recdir = mesg;
1017 		len = qword_get(&mesg, recdir, size);
1018 		if (len <= 0)
1019 			return -EINVAL;
1020 		trace_nfsd_ctl_recoverydir(netns(file), recdir);
1021 
1022 		status = nfs4_reset_recoverydir(recdir);
1023 		if (status)
1024 			return status;
1025 	}
1026 
1027 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1028 							nfs4_recoverydir());
1029 }
1030 
1031 /*
1032  * write_recoverydir - Set or report the pathname of the recovery directory
1033  *
1034  * Input:
1035  *			buf:		ignored
1036  *			size:		zero
1037  *
1038  * OR
1039  *
1040  * Input:
1041  *			buf:		C string containing the pathname
1042  *					of the directory on a local file
1043  *					system containing permanent NFSv4
1044  *					recovery data
1045  *			size:		non-zero length of C string in @buf
1046  * Output:
1047  *	On success:	passed-in buffer filled with '\n'-terminated C string
1048  *			containing the current recovery pathname setting;
1049  *			return code is the size in bytes of the string
1050  *	On error:	return code is zero or a negative errno value
1051  */
1052 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1053 {
1054 	ssize_t rv;
1055 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1056 
1057 	mutex_lock(&nfsd_mutex);
1058 	rv = __write_recoverydir(file, buf, size, nn);
1059 	mutex_unlock(&nfsd_mutex);
1060 	return rv;
1061 }
1062 #endif
1063 
1064 /*
1065  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1066  *
1067  * Input:
1068  *			buf:		ignored
1069  *			size:		zero
1070  * OR
1071  *
1072  * Input:
1073  *			buf:		any value
1074  *			size:		non-zero length of C string in @buf
1075  * Output:
1076  *			passed-in buffer filled with "Y" or "N" with a newline
1077  *			and NULL-terminated C string. This indicates whether
1078  *			the grace period has ended in the current net
1079  *			namespace. Return code is the size in bytes of the
1080  *			string. Writing a string that starts with 'Y', 'y', or
1081  *			'1' to the file will end the grace period for nfsd's v4
1082  *			lock manager.
1083  */
1084 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1085 {
1086 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1087 
1088 	if (size > 0) {
1089 		switch(buf[0]) {
1090 		case 'Y':
1091 		case 'y':
1092 		case '1':
1093 			if (!nfsd4_force_end_grace(nn))
1094 				return -EBUSY;
1095 			trace_nfsd_end_grace(netns(file));
1096 			break;
1097 		default:
1098 			return -EINVAL;
1099 		}
1100 	}
1101 
1102 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1103 			 nn->grace_ended ? 'Y' : 'N');
1104 }
1105 
1106 #endif
1107 
1108 /*----------------------------------------------------------------------------*/
1109 /*
1110  *	populating the filesystem.
1111  */
1112 
1113 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1114 {
1115 	struct inode *inode = new_inode(sb);
1116 	if (inode) {
1117 		/* Following advice from simple_fill_super documentation: */
1118 		inode->i_ino = iunique(sb, NFSD_MaxReserved);
1119 		inode->i_mode = mode;
1120 		simple_inode_init_ts(inode);
1121 	}
1122 	return inode;
1123 }
1124 
1125 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1126 {
1127 	struct inode *dir = parent->d_inode;
1128 	struct dentry *dentry;
1129 	struct inode *inode;
1130 
1131 	inode = nfsd_get_inode(parent->d_sb, S_IFDIR | 0600);
1132 	if (!inode)
1133 		return ERR_PTR(-ENOMEM);
1134 
1135 	dentry = simple_start_creating(parent, name);
1136 	if (IS_ERR(dentry)) {
1137 		iput(inode);
1138 		return dentry;
1139 	}
1140 	inode->i_fop = &simple_dir_operations;
1141 	inode->i_op = &simple_dir_inode_operations;
1142 	inc_nlink(inode);
1143 	if (ncl) {
1144 		inode->i_private = ncl;
1145 		kref_get(&ncl->cl_ref);
1146 	}
1147 	d_make_persistent(dentry, inode);
1148 	inc_nlink(dir);
1149 	fsnotify_mkdir(dir, dentry);
1150 	simple_done_creating(dentry);
1151 	return dentry;	// borrowed
1152 }
1153 
1154 #if IS_ENABLED(CONFIG_SUNRPC_GSS)
1155 /*
1156  * @content is assumed to be a NUL-terminated string that lives
1157  * longer than the symlink itself.
1158  */
1159 static void _nfsd_symlink(struct dentry *parent, const char *name,
1160 			  const char *content)
1161 {
1162 	struct inode *dir = parent->d_inode;
1163 	struct inode *inode;
1164 	struct dentry *dentry;
1165 
1166 	inode = nfsd_get_inode(dir->i_sb, S_IFLNK | 0777);
1167 	if (!inode)
1168 		return;
1169 
1170 	dentry = simple_start_creating(parent, name);
1171 	if (IS_ERR(dentry)) {
1172 		iput(inode);
1173 		return;
1174 	}
1175 
1176 	inode->i_op = &simple_symlink_inode_operations;
1177 	inode->i_link = (char *)content;
1178 	inode->i_size = strlen(content);
1179 
1180 	d_make_persistent(dentry, inode);
1181 	fsnotify_create(dir, dentry);
1182 	simple_done_creating(dentry);
1183 }
1184 #else
1185 static inline void _nfsd_symlink(struct dentry *parent, const char *name,
1186 				 const char *content)
1187 {
1188 }
1189 
1190 #endif
1191 
1192 static void clear_ncl(struct dentry *dentry)
1193 {
1194 	struct inode *inode = d_inode(dentry);
1195 	struct nfsdfs_client *ncl = inode->i_private;
1196 
1197 	spin_lock(&inode->i_lock);
1198 	inode->i_private = NULL;
1199 	spin_unlock(&inode->i_lock);
1200 	kref_put(&ncl->cl_ref, ncl->cl_release);
1201 }
1202 
1203 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1204 {
1205 	struct nfsdfs_client *nc;
1206 
1207 	spin_lock(&inode->i_lock);
1208 	nc = inode->i_private;
1209 	if (nc)
1210 		kref_get(&nc->cl_ref);
1211 	spin_unlock(&inode->i_lock);
1212 	return nc;
1213 }
1214 
1215 /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1216  * code instead. */
1217 static int nfsdfs_create_files(struct dentry *root,
1218 				const struct tree_descr *files,
1219 				struct nfsdfs_client *ncl,
1220 				struct dentry **fdentries)
1221 {
1222 	struct inode *dir = d_inode(root);
1223 	struct dentry *dentry;
1224 
1225 	for (int i = 0; files->name && files->name[0]; i++, files++) {
1226 		struct inode *inode = nfsd_get_inode(root->d_sb,
1227 						     S_IFREG | files->mode);
1228 		if (!inode)
1229 			return -ENOMEM;
1230 		dentry = simple_start_creating(root, files->name);
1231 		if (IS_ERR(dentry)) {
1232 			iput(inode);
1233 			return PTR_ERR(dentry);
1234 		}
1235 		kref_get(&ncl->cl_ref);
1236 		inode->i_fop = files->ops;
1237 		inode->i_private = ncl;
1238 		d_make_persistent(dentry, inode);
1239 		fsnotify_create(dir, dentry);
1240 		if (fdentries)
1241 			fdentries[i] = dentry; // borrowed
1242 		simple_done_creating(dentry);
1243 	}
1244 	return 0;
1245 }
1246 
1247 /* on success, returns positive number unique to that client. */
1248 struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1249 				 struct nfsdfs_client *ncl, u32 id,
1250 				 const struct tree_descr *files,
1251 				 struct dentry **fdentries)
1252 {
1253 	struct dentry *dentry;
1254 	char name[11];
1255 	int ret;
1256 
1257 	sprintf(name, "%u", id);
1258 
1259 	dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1260 	if (IS_ERR(dentry)) /* XXX: tossing errors? */
1261 		return NULL;
1262 	ret = nfsdfs_create_files(dentry, files, ncl, fdentries);
1263 	if (ret) {
1264 		nfsd_client_rmdir(dentry);
1265 		return NULL;
1266 	}
1267 	return dentry;
1268 }
1269 
1270 /* Taken from __rpc_rmdir: */
1271 void nfsd_client_rmdir(struct dentry *dentry)
1272 {
1273 	simple_recursive_removal(dentry, clear_ncl);
1274 }
1275 
1276 static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1277 {
1278 	struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1279 							nfsd_net_id);
1280 	struct dentry *dentry;
1281 	int ret;
1282 
1283 	static const struct tree_descr nfsd_files[] = {
1284 		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1285 		/* Per-export io stats use same ops as exports file */
1286 		[NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1287 		[NFSD_Export_features] = {"export_features",
1288 					&export_features_fops, S_IRUGO},
1289 		[NFSD_FO_UnlockIP] = {"unlock_ip",
1290 					&transaction_ops, S_IWUSR|S_IRUSR},
1291 		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1292 					&transaction_ops, S_IWUSR|S_IRUSR},
1293 		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1294 		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1295 		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1296 		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1297 		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1298 					&nfsd_reply_cache_stats_fops, S_IRUGO},
1299 		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1300 		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1301 		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1302 		[NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
1303 #ifdef CONFIG_NFSD_V4
1304 		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1305 		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1306 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1307 		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1308 #endif
1309 		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1310 #endif
1311 		/* last one */ {""}
1312 	};
1313 
1314 	ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1315 	if (ret)
1316 		return ret;
1317 	_nfsd_symlink(sb->s_root, "supported_krb5_enctypes",
1318 		      "/proc/net/rpc/gss_krb5_enctypes");
1319 	dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1320 	if (IS_ERR(dentry))
1321 		return PTR_ERR(dentry);
1322 	nn->nfsd_client_dir = dentry;
1323 	return 0;
1324 }
1325 
1326 static int nfsd_fs_get_tree(struct fs_context *fc)
1327 {
1328 	return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1329 }
1330 
1331 static void nfsd_fs_free_fc(struct fs_context *fc)
1332 {
1333 	if (fc->s_fs_info)
1334 		put_net(fc->s_fs_info);
1335 }
1336 
1337 static const struct fs_context_operations nfsd_fs_context_ops = {
1338 	.free		= nfsd_fs_free_fc,
1339 	.get_tree	= nfsd_fs_get_tree,
1340 };
1341 
1342 static int nfsd_init_fs_context(struct fs_context *fc)
1343 {
1344 	put_user_ns(fc->user_ns);
1345 	fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1346 	fc->ops = &nfsd_fs_context_ops;
1347 	return 0;
1348 }
1349 
1350 static void nfsd_umount(struct super_block *sb)
1351 {
1352 	struct net *net = sb->s_fs_info;
1353 
1354 	nfsd_shutdown_threads(net);
1355 
1356 	kill_anon_super(sb);
1357 	put_net(net);
1358 }
1359 
1360 static struct file_system_type nfsd_fs_type = {
1361 	.owner		= THIS_MODULE,
1362 	.name		= "nfsd",
1363 	.init_fs_context = nfsd_init_fs_context,
1364 	.kill_sb	= nfsd_umount,
1365 };
1366 MODULE_ALIAS_FS("nfsd");
1367 
1368 #ifdef CONFIG_PROC_FS
1369 
1370 static int exports_proc_open(struct inode *inode, struct file *file)
1371 {
1372 	return exports_net_open(current->nsproxy->net_ns, file);
1373 }
1374 
1375 static const struct proc_ops exports_proc_ops = {
1376 	.proc_open	= exports_proc_open,
1377 	.proc_read	= seq_read,
1378 	.proc_lseek	= seq_lseek,
1379 	.proc_release	= seq_release,
1380 };
1381 
1382 static int create_proc_exports_entry(void)
1383 {
1384 	struct proc_dir_entry *entry;
1385 
1386 	entry = proc_mkdir("fs/nfs", NULL);
1387 	if (!entry)
1388 		return -ENOMEM;
1389 	entry = proc_create("exports", 0, entry, &exports_proc_ops);
1390 	if (!entry) {
1391 		remove_proc_entry("fs/nfs", NULL);
1392 		return -ENOMEM;
1393 	}
1394 	return 0;
1395 }
1396 #else /* CONFIG_PROC_FS */
1397 static int create_proc_exports_entry(void)
1398 {
1399 	return 0;
1400 }
1401 #endif
1402 
1403 unsigned int nfsd_net_id;
1404 
1405 static int nfsd_genl_rpc_status_compose_msg(struct sk_buff *skb,
1406 					    struct netlink_callback *cb,
1407 					    struct nfsd_genl_rqstp *genl_rqstp)
1408 {
1409 	void *hdr;
1410 	u32 i;
1411 
1412 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
1413 			  &nfsd_nl_family, 0, NFSD_CMD_RPC_STATUS_GET);
1414 	if (!hdr)
1415 		return -ENOBUFS;
1416 
1417 	if (nla_put_be32(skb, NFSD_A_RPC_STATUS_XID, genl_rqstp->rq_xid) ||
1418 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_FLAGS, genl_rqstp->rq_flags) ||
1419 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROG, genl_rqstp->rq_prog) ||
1420 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROC, genl_rqstp->rq_proc) ||
1421 	    nla_put_u8(skb, NFSD_A_RPC_STATUS_VERSION, genl_rqstp->rq_vers) ||
1422 	    nla_put_s64(skb, NFSD_A_RPC_STATUS_SERVICE_TIME,
1423 			ktime_to_us(genl_rqstp->rq_stime),
1424 			NFSD_A_RPC_STATUS_PAD))
1425 		return -ENOBUFS;
1426 
1427 	switch (genl_rqstp->rq_saddr.sa_family) {
1428 	case AF_INET: {
1429 		const struct sockaddr_in *s_in, *d_in;
1430 
1431 		s_in = (const struct sockaddr_in *)&genl_rqstp->rq_saddr;
1432 		d_in = (const struct sockaddr_in *)&genl_rqstp->rq_daddr;
1433 		if (nla_put_in_addr(skb, NFSD_A_RPC_STATUS_SADDR4,
1434 				    s_in->sin_addr.s_addr) ||
1435 		    nla_put_in_addr(skb, NFSD_A_RPC_STATUS_DADDR4,
1436 				    d_in->sin_addr.s_addr) ||
1437 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1438 				 s_in->sin_port) ||
1439 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1440 				 d_in->sin_port))
1441 			return -ENOBUFS;
1442 		break;
1443 	}
1444 	case AF_INET6: {
1445 		const struct sockaddr_in6 *s_in, *d_in;
1446 
1447 		s_in = (const struct sockaddr_in6 *)&genl_rqstp->rq_saddr;
1448 		d_in = (const struct sockaddr_in6 *)&genl_rqstp->rq_daddr;
1449 		if (nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_SADDR6,
1450 				     &s_in->sin6_addr) ||
1451 		    nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_DADDR6,
1452 				     &d_in->sin6_addr) ||
1453 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1454 				 s_in->sin6_port) ||
1455 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1456 				 d_in->sin6_port))
1457 			return -ENOBUFS;
1458 		break;
1459 	}
1460 	}
1461 
1462 	for (i = 0; i < genl_rqstp->rq_opcnt; i++)
1463 		if (nla_put_u32(skb, NFSD_A_RPC_STATUS_COMPOUND_OPS,
1464 				genl_rqstp->rq_opnum[i]))
1465 			return -ENOBUFS;
1466 
1467 	genlmsg_end(skb, hdr);
1468 	return 0;
1469 }
1470 
1471 /**
1472  * nfsd_nl_rpc_status_get_dumpit - Handle rpc_status_get dumpit
1473  * @skb: reply buffer
1474  * @cb: netlink metadata and command arguments
1475  *
1476  * Returns the size of the reply or a negative errno.
1477  */
1478 int nfsd_nl_rpc_status_get_dumpit(struct sk_buff *skb,
1479 				  struct netlink_callback *cb)
1480 {
1481 	int i, ret, rqstp_index = 0;
1482 	struct nfsd_net *nn;
1483 
1484 	mutex_lock(&nfsd_mutex);
1485 
1486 	nn = net_generic(sock_net(skb->sk), nfsd_net_id);
1487 	if (!nn->nfsd_serv) {
1488 		ret = -ENODEV;
1489 		goto out_unlock;
1490 	}
1491 
1492 	rcu_read_lock();
1493 
1494 	for (i = 0; i < nn->nfsd_serv->sv_nrpools; i++) {
1495 		struct svc_rqst *rqstp;
1496 
1497 		if (i < cb->args[0]) /* already consumed */
1498 			continue;
1499 
1500 		rqstp_index = 0;
1501 		list_for_each_entry_rcu(rqstp,
1502 				&nn->nfsd_serv->sv_pools[i].sp_all_threads,
1503 				rq_all) {
1504 			struct nfsd_genl_rqstp genl_rqstp;
1505 			unsigned int status_counter;
1506 
1507 			if (rqstp_index++ < cb->args[1]) /* already consumed */
1508 				continue;
1509 			/*
1510 			 * Acquire rq_status_counter before parsing the rqst
1511 			 * fields. rq_status_counter is set to an odd value in
1512 			 * order to notify the consumers the rqstp fields are
1513 			 * meaningful.
1514 			 */
1515 			status_counter =
1516 				smp_load_acquire(&rqstp->rq_status_counter);
1517 			if (!(status_counter & 1))
1518 				continue;
1519 
1520 			genl_rqstp.rq_xid = rqstp->rq_xid;
1521 			genl_rqstp.rq_flags = rqstp->rq_flags;
1522 			genl_rqstp.rq_vers = rqstp->rq_vers;
1523 			genl_rqstp.rq_prog = rqstp->rq_prog;
1524 			genl_rqstp.rq_proc = rqstp->rq_proc;
1525 			genl_rqstp.rq_stime = rqstp->rq_stime;
1526 			genl_rqstp.rq_opcnt = 0;
1527 			memcpy(&genl_rqstp.rq_daddr, svc_daddr(rqstp),
1528 			       sizeof(struct sockaddr));
1529 			memcpy(&genl_rqstp.rq_saddr, svc_addr(rqstp),
1530 			       sizeof(struct sockaddr));
1531 
1532 #ifdef CONFIG_NFSD_V4
1533 			if (rqstp->rq_vers == NFS4_VERSION &&
1534 			    rqstp->rq_proc == NFSPROC4_COMPOUND) {
1535 				/* NFSv4 compound */
1536 				struct nfsd4_compoundargs *args;
1537 				int j;
1538 
1539 				args = rqstp->rq_argp;
1540 				genl_rqstp.rq_opcnt = min_t(u32, args->opcnt,
1541 							    ARRAY_SIZE(genl_rqstp.rq_opnum));
1542 				for (j = 0; j < genl_rqstp.rq_opcnt; j++)
1543 					genl_rqstp.rq_opnum[j] =
1544 						args->ops[j].opnum;
1545 			}
1546 #endif /* CONFIG_NFSD_V4 */
1547 
1548 			/*
1549 			 * Acquire rq_status_counter before reporting the rqst
1550 			 * fields to the user.
1551 			 */
1552 			if (smp_load_acquire(&rqstp->rq_status_counter) !=
1553 			    status_counter)
1554 				continue;
1555 
1556 			ret = nfsd_genl_rpc_status_compose_msg(skb, cb,
1557 							       &genl_rqstp);
1558 			if (ret)
1559 				goto out;
1560 		}
1561 	}
1562 
1563 	cb->args[0] = i;
1564 	cb->args[1] = rqstp_index;
1565 	ret = skb->len;
1566 out:
1567 	rcu_read_unlock();
1568 out_unlock:
1569 	mutex_unlock(&nfsd_mutex);
1570 
1571 	return ret;
1572 }
1573 
1574 /**
1575  * nfsd_nl_threads_set_doit - set the number of running threads
1576  * @skb: reply buffer
1577  * @info: netlink metadata and command arguments
1578  *
1579  * Return 0 on success or a negative errno.
1580  */
1581 int nfsd_nl_threads_set_doit(struct sk_buff *skb, struct genl_info *info)
1582 {
1583 	int *nthreads, nrpools = 0, i, ret = -EOPNOTSUPP, rem;
1584 	struct net *net = genl_info_net(info);
1585 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1586 	const struct nlattr *attr;
1587 	const char *scope = NULL;
1588 
1589 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_SERVER_THREADS))
1590 		return -EINVAL;
1591 
1592 	/* count number of SERVER_THREADS values */
1593 	nlmsg_for_each_attr_type(attr, NFSD_A_SERVER_THREADS, info->nlhdr,
1594 				 GENL_HDRLEN, rem)
1595 		nrpools++;
1596 
1597 	mutex_lock(&nfsd_mutex);
1598 
1599 	nthreads = kcalloc(nrpools, sizeof(int), GFP_KERNEL);
1600 	if (!nthreads) {
1601 		ret = -ENOMEM;
1602 		goto out_unlock;
1603 	}
1604 
1605 	i = 0;
1606 	nlmsg_for_each_attr_type(attr, NFSD_A_SERVER_THREADS, info->nlhdr,
1607 				 GENL_HDRLEN, rem) {
1608 		nthreads[i++] = nla_get_u32(attr);
1609 		if (i >= nrpools)
1610 			break;
1611 	}
1612 
1613 	if (info->attrs[NFSD_A_SERVER_GRACETIME] ||
1614 	    info->attrs[NFSD_A_SERVER_LEASETIME] ||
1615 	    info->attrs[NFSD_A_SERVER_SCOPE]) {
1616 		ret = -EBUSY;
1617 		if (nn->nfsd_serv && nn->nfsd_serv->sv_nrthreads)
1618 			goto out_unlock;
1619 
1620 		ret = -EINVAL;
1621 		attr = info->attrs[NFSD_A_SERVER_GRACETIME];
1622 		if (attr) {
1623 			u32 gracetime = nla_get_u32(attr);
1624 
1625 			if (gracetime < 10 || gracetime > 3600)
1626 				goto out_unlock;
1627 
1628 			nn->nfsd4_grace = gracetime;
1629 		}
1630 
1631 		attr = info->attrs[NFSD_A_SERVER_LEASETIME];
1632 		if (attr) {
1633 			u32 leasetime = nla_get_u32(attr);
1634 
1635 			if (leasetime < 10 || leasetime > 3600)
1636 				goto out_unlock;
1637 
1638 			nn->nfsd4_lease = leasetime;
1639 		}
1640 
1641 		attr = info->attrs[NFSD_A_SERVER_SCOPE];
1642 		if (attr)
1643 			scope = nla_data(attr);
1644 	}
1645 
1646 	attr = info->attrs[NFSD_A_SERVER_MIN_THREADS];
1647 	if (attr)
1648 		nn->min_threads = nla_get_u32(attr);
1649 
1650 	ret = nfsd_svc(nrpools, nthreads, net, get_current_cred(), scope);
1651 	if (ret > 0)
1652 		ret = 0;
1653 out_unlock:
1654 	mutex_unlock(&nfsd_mutex);
1655 	kfree(nthreads);
1656 	return ret;
1657 }
1658 
1659 /**
1660  * nfsd_nl_threads_get_doit - get the number of running threads
1661  * @skb: reply buffer
1662  * @info: netlink metadata and command arguments
1663  *
1664  * Return 0 on success or a negative errno.
1665  */
1666 int nfsd_nl_threads_get_doit(struct sk_buff *skb, struct genl_info *info)
1667 {
1668 	struct net *net = genl_info_net(info);
1669 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1670 	void *hdr;
1671 	int err;
1672 
1673 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
1674 	if (!skb)
1675 		return -ENOMEM;
1676 
1677 	hdr = genlmsg_iput(skb, info);
1678 	if (!hdr) {
1679 		err = -EMSGSIZE;
1680 		goto err_free_msg;
1681 	}
1682 
1683 	mutex_lock(&nfsd_mutex);
1684 
1685 	err = nla_put_u32(skb, NFSD_A_SERVER_GRACETIME,
1686 			  nn->nfsd4_grace) ||
1687 	      nla_put_u32(skb, NFSD_A_SERVER_LEASETIME,
1688 			  nn->nfsd4_lease) ||
1689 	      nla_put_u32(skb, NFSD_A_SERVER_MIN_THREADS,
1690 			  nn->min_threads) ||
1691 	      nla_put_string(skb, NFSD_A_SERVER_SCOPE,
1692 			  nn->nfsd_name);
1693 	if (err)
1694 		goto err_unlock;
1695 
1696 	if (nn->nfsd_serv) {
1697 		int i;
1698 
1699 		for (i = 0; i < nfsd_nrpools(net); ++i) {
1700 			struct svc_pool *sp = &nn->nfsd_serv->sv_pools[i];
1701 
1702 			err = nla_put_u32(skb, NFSD_A_SERVER_THREADS,
1703 					  sp->sp_nrthreads);
1704 			if (err)
1705 				goto err_unlock;
1706 		}
1707 	} else {
1708 		err = nla_put_u32(skb, NFSD_A_SERVER_THREADS, 0);
1709 		if (err)
1710 			goto err_unlock;
1711 	}
1712 
1713 	mutex_unlock(&nfsd_mutex);
1714 
1715 	genlmsg_end(skb, hdr);
1716 
1717 	return genlmsg_reply(skb, info);
1718 
1719 err_unlock:
1720 	mutex_unlock(&nfsd_mutex);
1721 err_free_msg:
1722 	nlmsg_free(skb);
1723 
1724 	return err;
1725 }
1726 
1727 /**
1728  * nfsd_nl_version_set_doit - set the nfs enabled versions
1729  * @skb: reply buffer
1730  * @info: netlink metadata and command arguments
1731  *
1732  * Return 0 on success or a negative errno.
1733  */
1734 int nfsd_nl_version_set_doit(struct sk_buff *skb, struct genl_info *info)
1735 {
1736 	const struct nlattr *attr;
1737 	struct nfsd_net *nn;
1738 	int i, rem;
1739 
1740 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_SERVER_PROTO_VERSION))
1741 		return -EINVAL;
1742 
1743 	mutex_lock(&nfsd_mutex);
1744 
1745 	nn = net_generic(genl_info_net(info), nfsd_net_id);
1746 	if (nn->nfsd_serv) {
1747 		mutex_unlock(&nfsd_mutex);
1748 		return -EBUSY;
1749 	}
1750 
1751 	/* clear current supported versions. */
1752 	nfsd_vers(nn, 2, NFSD_CLEAR);
1753 	nfsd_vers(nn, 3, NFSD_CLEAR);
1754 	for (i = 0; i <= NFSD_SUPPORTED_MINOR_VERSION; i++)
1755 		nfsd_minorversion(nn, i, NFSD_CLEAR);
1756 
1757 	nlmsg_for_each_attr_type(attr, NFSD_A_SERVER_PROTO_VERSION, info->nlhdr,
1758 				 GENL_HDRLEN, rem) {
1759 		struct nlattr *tb[NFSD_A_VERSION_MAX + 1];
1760 		u32 major, minor = 0;
1761 		bool enabled;
1762 
1763 		if (nla_parse_nested(tb, NFSD_A_VERSION_MAX, attr,
1764 				     nfsd_version_nl_policy, info->extack) < 0)
1765 			continue;
1766 
1767 		if (!tb[NFSD_A_VERSION_MAJOR])
1768 			continue;
1769 
1770 		major = nla_get_u32(tb[NFSD_A_VERSION_MAJOR]);
1771 		if (tb[NFSD_A_VERSION_MINOR])
1772 			minor = nla_get_u32(tb[NFSD_A_VERSION_MINOR]);
1773 
1774 		enabled = nla_get_flag(tb[NFSD_A_VERSION_ENABLED]);
1775 
1776 		switch (major) {
1777 		case 4:
1778 			nfsd_minorversion(nn, minor, enabled ? NFSD_SET : NFSD_CLEAR);
1779 			break;
1780 		case 3:
1781 		case 2:
1782 			if (!minor)
1783 				nfsd_vers(nn, major, enabled ? NFSD_SET : NFSD_CLEAR);
1784 			break;
1785 		default:
1786 			break;
1787 		}
1788 	}
1789 
1790 	mutex_unlock(&nfsd_mutex);
1791 
1792 	return 0;
1793 }
1794 
1795 /**
1796  * nfsd_nl_version_get_doit - get the enabled status for all supported nfs versions
1797  * @skb: reply buffer
1798  * @info: netlink metadata and command arguments
1799  *
1800  * Return 0 on success or a negative errno.
1801  */
1802 int nfsd_nl_version_get_doit(struct sk_buff *skb, struct genl_info *info)
1803 {
1804 	struct nfsd_net *nn;
1805 	int i, err;
1806 	void *hdr;
1807 
1808 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
1809 	if (!skb)
1810 		return -ENOMEM;
1811 
1812 	hdr = genlmsg_iput(skb, info);
1813 	if (!hdr) {
1814 		err = -EMSGSIZE;
1815 		goto err_free_msg;
1816 	}
1817 
1818 	mutex_lock(&nfsd_mutex);
1819 	nn = net_generic(genl_info_net(info), nfsd_net_id);
1820 
1821 	for (i = 2; i <= 4; i++) {
1822 		int j;
1823 
1824 		for (j = 0; j <= NFSD_SUPPORTED_MINOR_VERSION; j++) {
1825 			struct nlattr *attr;
1826 
1827 			/* Don't record any versions the kernel doesn't have
1828 			 * compiled in
1829 			 */
1830 			if (!nfsd_support_version(i))
1831 				continue;
1832 
1833 			/* NFSv{2,3} does not support minor numbers */
1834 			if (i < 4 && j)
1835 				continue;
1836 
1837 			attr = nla_nest_start(skb,
1838 					      NFSD_A_SERVER_PROTO_VERSION);
1839 			if (!attr) {
1840 				err = -EINVAL;
1841 				goto err_nfsd_unlock;
1842 			}
1843 
1844 			if (nla_put_u32(skb, NFSD_A_VERSION_MAJOR, i) ||
1845 			    nla_put_u32(skb, NFSD_A_VERSION_MINOR, j)) {
1846 				err = -EINVAL;
1847 				goto err_nfsd_unlock;
1848 			}
1849 
1850 			/* Set the enabled flag if the version is enabled */
1851 			if (nfsd_vers(nn, i, NFSD_TEST) &&
1852 			    (i < 4 || nfsd_minorversion(nn, j, NFSD_TEST)) &&
1853 			    nla_put_flag(skb, NFSD_A_VERSION_ENABLED)) {
1854 				err = -EINVAL;
1855 				goto err_nfsd_unlock;
1856 			}
1857 
1858 			nla_nest_end(skb, attr);
1859 		}
1860 	}
1861 
1862 	mutex_unlock(&nfsd_mutex);
1863 	genlmsg_end(skb, hdr);
1864 
1865 	return genlmsg_reply(skb, info);
1866 
1867 err_nfsd_unlock:
1868 	mutex_unlock(&nfsd_mutex);
1869 err_free_msg:
1870 	nlmsg_free(skb);
1871 
1872 	return err;
1873 }
1874 
1875 /**
1876  * nfsd_nl_listener_set_doit - set the nfs running sockets
1877  * @skb: reply buffer
1878  * @info: netlink metadata and command arguments
1879  *
1880  * Return 0 on success or a negative errno.
1881  */
1882 int nfsd_nl_listener_set_doit(struct sk_buff *skb, struct genl_info *info)
1883 {
1884 	struct net *net = genl_info_net(info);
1885 	struct svc_xprt *xprt, *tmp;
1886 	const struct nlattr *attr;
1887 	struct svc_serv *serv;
1888 	LIST_HEAD(permsocks);
1889 	struct nfsd_net *nn;
1890 	bool delete = false;
1891 	int err, rem;
1892 
1893 	mutex_lock(&nfsd_mutex);
1894 
1895 	err = nfsd_create_serv(net);
1896 	if (err) {
1897 		mutex_unlock(&nfsd_mutex);
1898 		return err;
1899 	}
1900 
1901 	nn = net_generic(net, nfsd_net_id);
1902 	serv = nn->nfsd_serv;
1903 
1904 	spin_lock_bh(&serv->sv_lock);
1905 
1906 	/* Move all of the old listener sockets to a temp list */
1907 	list_splice_init(&serv->sv_permsocks, &permsocks);
1908 
1909 	/*
1910 	 * Walk the list of server_socks from userland and move any that match
1911 	 * back to sv_permsocks
1912 	 */
1913 	nlmsg_for_each_attr_type(attr, NFSD_A_SERVER_SOCK_ADDR, info->nlhdr,
1914 				 GENL_HDRLEN, rem) {
1915 		struct nlattr *tb[NFSD_A_SOCK_MAX + 1];
1916 		const char *xcl_name;
1917 		struct sockaddr *sa;
1918 
1919 		if (nla_parse_nested(tb, NFSD_A_SOCK_MAX, attr,
1920 				     nfsd_sock_nl_policy, info->extack) < 0)
1921 			continue;
1922 
1923 		if (!tb[NFSD_A_SOCK_ADDR] || !tb[NFSD_A_SOCK_TRANSPORT_NAME])
1924 			continue;
1925 
1926 		if (nla_len(tb[NFSD_A_SOCK_ADDR]) < sizeof(*sa))
1927 			continue;
1928 
1929 		xcl_name = nla_data(tb[NFSD_A_SOCK_TRANSPORT_NAME]);
1930 		sa = nla_data(tb[NFSD_A_SOCK_ADDR]);
1931 
1932 		/* Put back any matching sockets */
1933 		list_for_each_entry_safe(xprt, tmp, &permsocks, xpt_list) {
1934 			/* This shouldn't be possible */
1935 			if (WARN_ON_ONCE(xprt->xpt_net != net)) {
1936 				list_move(&xprt->xpt_list, &serv->sv_permsocks);
1937 				continue;
1938 			}
1939 
1940 			/* If everything matches, put it back */
1941 			if (!strcmp(xprt->xpt_class->xcl_name, xcl_name) &&
1942 			    rpc_cmp_addr_port(sa, (struct sockaddr *)&xprt->xpt_local)) {
1943 				list_move(&xprt->xpt_list, &serv->sv_permsocks);
1944 				break;
1945 			}
1946 		}
1947 	}
1948 
1949 	/*
1950 	 * If there are listener transports remaining on the permsocks list,
1951 	 * it means we were asked to remove a listener.
1952 	 */
1953 	if (!list_empty(&permsocks)) {
1954 		list_splice_init(&permsocks, &serv->sv_permsocks);
1955 		delete = true;
1956 	}
1957 	spin_unlock_bh(&serv->sv_lock);
1958 
1959 	/* Do not remove listeners while there are active threads. */
1960 	if (serv->sv_nrthreads) {
1961 		err = -EBUSY;
1962 		goto out_unlock_mtx;
1963 	}
1964 
1965 	/*
1966 	 * Since we can't delete an arbitrary llist entry, destroy the
1967 	 * remaining listeners and recreate the list.
1968 	 */
1969 	if (delete)
1970 		svc_xprt_destroy_all(serv, net, false);
1971 
1972 	/* walk list of addrs again, open any that still don't exist */
1973 	nlmsg_for_each_attr_type(attr, NFSD_A_SERVER_SOCK_ADDR, info->nlhdr,
1974 				 GENL_HDRLEN, rem) {
1975 		struct nlattr *tb[NFSD_A_SOCK_MAX + 1];
1976 		const char *xcl_name;
1977 		struct sockaddr *sa;
1978 		int ret;
1979 
1980 		if (nla_parse_nested(tb, NFSD_A_SOCK_MAX, attr,
1981 				     nfsd_sock_nl_policy, info->extack) < 0)
1982 			continue;
1983 
1984 		if (!tb[NFSD_A_SOCK_ADDR] || !tb[NFSD_A_SOCK_TRANSPORT_NAME])
1985 			continue;
1986 
1987 		if (nla_len(tb[NFSD_A_SOCK_ADDR]) < sizeof(*sa))
1988 			continue;
1989 
1990 		xcl_name = nla_data(tb[NFSD_A_SOCK_TRANSPORT_NAME]);
1991 		sa = nla_data(tb[NFSD_A_SOCK_ADDR]);
1992 
1993 		xprt = svc_find_listener(serv, xcl_name, net, sa);
1994 		if (xprt) {
1995 			if (delete)
1996 				WARN_ONCE(1, "Transport type=%s already exists\n",
1997 					  xcl_name);
1998 			svc_xprt_put(xprt);
1999 			continue;
2000 		}
2001 
2002 		ret = svc_xprt_create_from_sa(serv, xcl_name, net, sa, 0,
2003 					      get_current_cred());
2004 		/* always save the latest error */
2005 		if (ret < 0)
2006 			err = ret;
2007 	}
2008 
2009 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
2010 		nfsd_destroy_serv(net);
2011 
2012 out_unlock_mtx:
2013 	mutex_unlock(&nfsd_mutex);
2014 
2015 	return err;
2016 }
2017 
2018 /**
2019  * nfsd_nl_listener_get_doit - get the nfs running listeners
2020  * @skb: reply buffer
2021  * @info: netlink metadata and command arguments
2022  *
2023  * Return 0 on success or a negative errno.
2024  */
2025 int nfsd_nl_listener_get_doit(struct sk_buff *skb, struct genl_info *info)
2026 {
2027 	struct svc_xprt *xprt;
2028 	struct svc_serv *serv;
2029 	struct nfsd_net *nn;
2030 	void *hdr;
2031 	int err;
2032 
2033 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
2034 	if (!skb)
2035 		return -ENOMEM;
2036 
2037 	hdr = genlmsg_iput(skb, info);
2038 	if (!hdr) {
2039 		err = -EMSGSIZE;
2040 		goto err_free_msg;
2041 	}
2042 
2043 	mutex_lock(&nfsd_mutex);
2044 	nn = net_generic(genl_info_net(info), nfsd_net_id);
2045 
2046 	/* no nfs server? Just send empty socket list */
2047 	if (!nn->nfsd_serv)
2048 		goto out_unlock_mtx;
2049 
2050 	serv = nn->nfsd_serv;
2051 	spin_lock_bh(&serv->sv_lock);
2052 	list_for_each_entry(xprt, &serv->sv_permsocks, xpt_list) {
2053 		struct nlattr *attr;
2054 
2055 		attr = nla_nest_start(skb, NFSD_A_SERVER_SOCK_ADDR);
2056 		if (!attr) {
2057 			err = -EINVAL;
2058 			goto err_serv_unlock;
2059 		}
2060 
2061 		if (nla_put_string(skb, NFSD_A_SOCK_TRANSPORT_NAME,
2062 				   xprt->xpt_class->xcl_name) ||
2063 		    nla_put(skb, NFSD_A_SOCK_ADDR,
2064 			    sizeof(struct sockaddr_storage),
2065 			    &xprt->xpt_local)) {
2066 			err = -EINVAL;
2067 			goto err_serv_unlock;
2068 		}
2069 
2070 		nla_nest_end(skb, attr);
2071 	}
2072 	spin_unlock_bh(&serv->sv_lock);
2073 out_unlock_mtx:
2074 	mutex_unlock(&nfsd_mutex);
2075 	genlmsg_end(skb, hdr);
2076 
2077 	return genlmsg_reply(skb, info);
2078 
2079 err_serv_unlock:
2080 	spin_unlock_bh(&serv->sv_lock);
2081 	mutex_unlock(&nfsd_mutex);
2082 err_free_msg:
2083 	nlmsg_free(skb);
2084 
2085 	return err;
2086 }
2087 
2088 /**
2089  * nfsd_nl_pool_mode_set_doit - set the number of running threads
2090  * @skb: reply buffer
2091  * @info: netlink metadata and command arguments
2092  *
2093  * Return 0 on success or a negative errno.
2094  */
2095 int nfsd_nl_pool_mode_set_doit(struct sk_buff *skb, struct genl_info *info)
2096 {
2097 	const struct nlattr *attr;
2098 
2099 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_POOL_MODE_MODE))
2100 		return -EINVAL;
2101 
2102 	attr = info->attrs[NFSD_A_POOL_MODE_MODE];
2103 	return sunrpc_set_pool_mode(nla_data(attr));
2104 }
2105 
2106 /**
2107  * nfsd_nl_pool_mode_get_doit - get info about pool_mode
2108  * @skb: reply buffer
2109  * @info: netlink metadata and command arguments
2110  *
2111  * Return 0 on success or a negative errno.
2112  */
2113 int nfsd_nl_pool_mode_get_doit(struct sk_buff *skb, struct genl_info *info)
2114 {
2115 	struct net *net = genl_info_net(info);
2116 	char buf[16];
2117 	void *hdr;
2118 	int err;
2119 
2120 	if (sunrpc_get_pool_mode(buf, ARRAY_SIZE(buf)) >= ARRAY_SIZE(buf))
2121 		return -ERANGE;
2122 
2123 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
2124 	if (!skb)
2125 		return -ENOMEM;
2126 
2127 	err = -EMSGSIZE;
2128 	hdr = genlmsg_iput(skb, info);
2129 	if (!hdr)
2130 		goto err_free_msg;
2131 
2132 	err = nla_put_string(skb, NFSD_A_POOL_MODE_MODE, buf) |
2133 	      nla_put_u32(skb, NFSD_A_POOL_MODE_NPOOLS, nfsd_nrpools(net));
2134 	if (err)
2135 		goto err_free_msg;
2136 
2137 	genlmsg_end(skb, hdr);
2138 	return genlmsg_reply(skb, info);
2139 
2140 err_free_msg:
2141 	nlmsg_free(skb);
2142 	return err;
2143 }
2144 
2145 /**
2146  * nfsd_net_init - Prepare the nfsd_net portion of a new net namespace
2147  * @net: a freshly-created network namespace
2148  *
2149  * This information stays around as long as the network namespace is
2150  * alive whether or not there is an NFSD instance running in the
2151  * namespace.
2152  *
2153  * Returns zero on success, or a negative errno otherwise.
2154  */
2155 static __net_init int nfsd_net_init(struct net *net)
2156 {
2157 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2158 	int retval;
2159 	int i;
2160 
2161 	retval = nfsd_export_init(net);
2162 	if (retval)
2163 		goto out_export_error;
2164 	retval = nfsd_idmap_init(net);
2165 	if (retval)
2166 		goto out_idmap_error;
2167 	retval = percpu_counter_init_many(nn->counter, 0, GFP_KERNEL,
2168 					  NFSD_STATS_COUNTERS_NUM);
2169 	if (retval)
2170 		goto out_repcache_error;
2171 
2172 	memset(&nn->nfsd_svcstats, 0, sizeof(nn->nfsd_svcstats));
2173 	nn->nfsd_svcstats.program = &nfsd_programs[0];
2174 	if (!nfsd_proc_stat_init(net)) {
2175 		retval = -ENOMEM;
2176 		goto out_proc_error;
2177 	}
2178 
2179 	for (i = 0; i < sizeof(nn->nfsd_versions); i++)
2180 		nn->nfsd_versions[i] = nfsd_support_version(i);
2181 	for (i = 0; i < sizeof(nn->nfsd4_minorversions); i++)
2182 		nn->nfsd4_minorversions[i] = nfsd_support_version(4);
2183 	nn->nfsd_info.mutex = &nfsd_mutex;
2184 	nn->nfsd_serv = NULL;
2185 	nfsd4_init_leases_net(nn);
2186 	get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
2187 	seqlock_init(&nn->writeverf_lock);
2188 #if IS_ENABLED(CONFIG_NFS_LOCALIO)
2189 	spin_lock_init(&nn->local_clients_lock);
2190 	INIT_LIST_HEAD(&nn->local_clients);
2191 #endif
2192 	return 0;
2193 
2194 out_proc_error:
2195 	percpu_counter_destroy_many(nn->counter, NFSD_STATS_COUNTERS_NUM);
2196 out_repcache_error:
2197 	nfsd_idmap_shutdown(net);
2198 out_idmap_error:
2199 	nfsd_export_shutdown(net);
2200 out_export_error:
2201 	return retval;
2202 }
2203 
2204 #if IS_ENABLED(CONFIG_NFS_LOCALIO)
2205 /**
2206  * nfsd_net_pre_exit - Disconnect localio clients from net namespace
2207  * @net: a network namespace that is about to be destroyed
2208  *
2209  * This invalidates ->net pointers held by localio clients
2210  * while they can still safely access nn->counter.
2211  */
2212 static __net_exit void nfsd_net_pre_exit(struct net *net)
2213 {
2214 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2215 
2216 	nfs_localio_invalidate_clients(&nn->local_clients,
2217 				       &nn->local_clients_lock);
2218 }
2219 #endif
2220 
2221 /**
2222  * nfsd_net_exit - Release the nfsd_net portion of a net namespace
2223  * @net: a network namespace that is about to be destroyed
2224  *
2225  */
2226 static __net_exit void nfsd_net_exit(struct net *net)
2227 {
2228 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2229 
2230 	nfsd_proc_stat_shutdown(net);
2231 	percpu_counter_destroy_many(nn->counter, NFSD_STATS_COUNTERS_NUM);
2232 	nfsd_idmap_shutdown(net);
2233 	nfsd_export_shutdown(net);
2234 }
2235 
2236 static struct pernet_operations nfsd_net_ops = {
2237 	.init = nfsd_net_init,
2238 #if IS_ENABLED(CONFIG_NFS_LOCALIO)
2239 	.pre_exit = nfsd_net_pre_exit,
2240 #endif
2241 	.exit = nfsd_net_exit,
2242 	.id   = &nfsd_net_id,
2243 	.size = sizeof(struct nfsd_net),
2244 };
2245 
2246 static int __init init_nfsd(void)
2247 {
2248 	int retval;
2249 
2250 	nfsd_debugfs_init();
2251 
2252 	retval = nfsd4_init_slabs();
2253 	if (retval)
2254 		return retval;
2255 	retval = nfsd4_init_pnfs();
2256 	if (retval)
2257 		goto out_free_slabs;
2258 	retval = nfsd_drc_slab_create();
2259 	if (retval)
2260 		goto out_free_pnfs;
2261 	nfsd_lockd_init();	/* lockd->nfsd callbacks */
2262 	retval = register_pernet_subsys(&nfsd_net_ops);
2263 	if (retval < 0)
2264 		goto out_free_lockd;
2265 	retval = register_cld_notifier();
2266 	if (retval)
2267 		goto out_free_subsys;
2268 	retval = nfsd4_create_laundry_wq();
2269 	if (retval)
2270 		goto out_free_cld;
2271 	retval = register_filesystem(&nfsd_fs_type);
2272 	if (retval)
2273 		goto out_free_nfsd4;
2274 	retval = genl_register_family(&nfsd_nl_family);
2275 	if (retval)
2276 		goto out_free_filesystem;
2277 	retval = create_proc_exports_entry();
2278 	if (retval)
2279 		goto out_free_all;
2280 	nfsd_localio_ops_init();
2281 
2282 	return 0;
2283 out_free_all:
2284 	genl_unregister_family(&nfsd_nl_family);
2285 out_free_filesystem:
2286 	unregister_filesystem(&nfsd_fs_type);
2287 out_free_nfsd4:
2288 	nfsd4_destroy_laundry_wq();
2289 out_free_cld:
2290 	unregister_cld_notifier();
2291 out_free_subsys:
2292 	unregister_pernet_subsys(&nfsd_net_ops);
2293 out_free_lockd:
2294 	nfsd_lockd_shutdown();
2295 	nfsd_drc_slab_free();
2296 out_free_pnfs:
2297 	nfsd4_exit_pnfs();
2298 out_free_slabs:
2299 	nfsd4_free_slabs();
2300 	nfsd_debugfs_exit();
2301 	return retval;
2302 }
2303 
2304 static void __exit exit_nfsd(void)
2305 {
2306 	remove_proc_entry("fs/nfs/exports", NULL);
2307 	remove_proc_entry("fs/nfs", NULL);
2308 	genl_unregister_family(&nfsd_nl_family);
2309 	unregister_filesystem(&nfsd_fs_type);
2310 	nfsd4_destroy_laundry_wq();
2311 	unregister_cld_notifier();
2312 	unregister_pernet_subsys(&nfsd_net_ops);
2313 	nfsd_drc_slab_free();
2314 	nfsd_lockd_shutdown();
2315 	nfsd4_free_slabs();
2316 	nfsd4_exit_pnfs();
2317 	nfsd_debugfs_exit();
2318 }
2319 
2320 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2321 MODULE_DESCRIPTION("In-kernel NFS server");
2322 MODULE_LICENSE("GPL");
2323 module_init(init_nfsd)
2324 module_exit(exit_nfsd)
2325