xref: /linux/fs/nfsd/nfsctl.c (revision 070a542f08acb7e8cf197287f5c44658c715d2d1)
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 
nfsctl_transaction_write(struct file * file,const char __user * buf,size_t size,loff_t * pos)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 
nfsctl_transaction_read(struct file * file,char __user * buf,size_t size,loff_t * pos)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 
exports_net_open(struct net * net,struct file * file)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 
exports_nfsd_open(struct inode * inode,struct file * file)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 
export_features_show(struct seq_file * m,void * v)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 
nfsd_pool_stats_open(struct inode * inode,struct file * file)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 
netns(struct file * file)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  */
write_unlock_ip(struct file * file,char * buf,size_t size)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  */
write_unlock_fs(struct file * file,char * buf,size_t size)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 
263 	/* sanity check */
264 	if (size == 0)
265 		return -EINVAL;
266 
267 	if (buf[size-1] != '\n')
268 		return -EINVAL;
269 
270 	fo_path = buf;
271 	if (qword_get(&buf, fo_path, size) < 0)
272 		return -EINVAL;
273 	trace_nfsd_ctl_unlock_fs(netns(file), fo_path);
274 	error = kern_path(fo_path, 0, &path);
275 	if (error)
276 		return error;
277 
278 	/*
279 	 * XXX: Needs better sanity checking.  Otherwise we could end up
280 	 * releasing locks on the wrong file system.
281 	 *
282 	 * For example:
283 	 * 1.  Does the path refer to a directory?
284 	 * 2.  Is that directory a mount point, or
285 	 * 3.  Is that directory the root of an exported file system?
286 	 */
287 	error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
288 	nfsd4_revoke_states(netns(file), path.dentry->d_sb);
289 
290 	path_put(&path);
291 	return error;
292 }
293 
294 /*
295  * write_filehandle - Get a variable-length NFS file handle by path
296  *
297  * On input, the buffer contains a '\n'-terminated C string comprised of
298  * three alphanumeric words separated by whitespace.  The string may
299  * contain escape sequences.
300  *
301  * Input:
302  *			buf:
303  *				domain:		client domain name
304  *				path:		export pathname
305  *				maxsize:	numeric maximum size of
306  *						@buf
307  *			size:	length of C string in @buf
308  * Output:
309  *	On success:	passed-in buffer filled with '\n'-terminated C
310  *			string containing a ASCII hex text version
311  *			of the NFS file handle;
312  *			return code is the size in bytes of the string
313  *	On error:	return code is negative errno value
314  */
write_filehandle(struct file * file,char * buf,size_t size)315 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
316 {
317 	char *dname, *path;
318 	int maxsize;
319 	char *mesg = buf;
320 	int len;
321 	struct auth_domain *dom;
322 	struct knfsd_fh fh;
323 
324 	if (size == 0)
325 		return -EINVAL;
326 
327 	if (buf[size-1] != '\n')
328 		return -EINVAL;
329 	buf[size-1] = 0;
330 
331 	dname = mesg;
332 	len = qword_get(&mesg, dname, size);
333 	if (len <= 0)
334 		return -EINVAL;
335 
336 	path = dname+len+1;
337 	len = qword_get(&mesg, path, size);
338 	if (len <= 0)
339 		return -EINVAL;
340 
341 	len = get_int(&mesg, &maxsize);
342 	if (len)
343 		return len;
344 
345 	if (maxsize < NFS_FHSIZE)
346 		return -EINVAL;
347 	maxsize = min(maxsize, NFS3_FHSIZE);
348 
349 	if (qword_get(&mesg, mesg, size) > 0)
350 		return -EINVAL;
351 
352 	trace_nfsd_ctl_filehandle(netns(file), dname, path, maxsize);
353 
354 	/* we have all the words, they are in buf.. */
355 	dom = unix_domain_find(dname);
356 	if (!dom)
357 		return -ENOMEM;
358 
359 	len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
360 	auth_domain_put(dom);
361 	if (len)
362 		return len;
363 
364 	mesg = buf;
365 	len = SIMPLE_TRANSACTION_LIMIT;
366 	qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
367 	mesg[-1] = '\n';
368 	return mesg - buf;
369 }
370 
371 /*
372  * write_threads - Start NFSD, or report the current number of running threads
373  *
374  * Input:
375  *			buf:		ignored
376  *			size:		zero
377  * Output:
378  *	On success:	passed-in buffer filled with '\n'-terminated C
379  *			string numeric value representing the number of
380  *			running NFSD threads;
381  *			return code is the size in bytes of the string
382  *	On error:	return code is zero
383  *
384  * OR
385  *
386  * Input:
387  *			buf:		C string containing an unsigned
388  *					integer value representing the
389  *					number of NFSD threads to start
390  *			size:		non-zero length of C string in @buf
391  * Output:
392  *	On success:	NFS service is started;
393  *			passed-in buffer filled with '\n'-terminated C
394  *			string numeric value representing the number of
395  *			running NFSD threads;
396  *			return code is the size in bytes of the string
397  *	On error:	return code is zero or a negative errno value
398  */
write_threads(struct file * file,char * buf,size_t size)399 static ssize_t write_threads(struct file *file, char *buf, size_t size)
400 {
401 	char *mesg = buf;
402 	int rv;
403 	struct net *net = netns(file);
404 
405 	if (size > 0) {
406 		int newthreads;
407 		rv = get_int(&mesg, &newthreads);
408 		if (rv)
409 			return rv;
410 		if (newthreads < 0)
411 			return -EINVAL;
412 		trace_nfsd_ctl_threads(net, newthreads);
413 		mutex_lock(&nfsd_mutex);
414 		rv = nfsd_svc(1, &newthreads, net, file->f_cred, NULL);
415 		mutex_unlock(&nfsd_mutex);
416 		if (rv < 0)
417 			return rv;
418 	} else
419 		rv = nfsd_nrthreads(net);
420 
421 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
422 }
423 
424 /*
425  * write_pool_threads - Set or report the current number of threads per pool
426  *
427  * Input:
428  *			buf:		ignored
429  *			size:		zero
430  *
431  * OR
432  *
433  * Input:
434  *			buf:		C string containing whitespace-
435  *					separated unsigned integer values
436  *					representing the number of NFSD
437  *					threads to start in each pool
438  *			size:		non-zero length of C string in @buf
439  * Output:
440  *	On success:	passed-in buffer filled with '\n'-terminated C
441  *			string containing integer values representing the
442  *			number of NFSD threads in each pool;
443  *			return code is the size in bytes of the string
444  *	On error:	return code is zero or a negative errno value
445  */
write_pool_threads(struct file * file,char * buf,size_t size)446 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
447 {
448 	/* if size > 0, look for an array of number of threads per node
449 	 * and apply them  then write out number of threads per node as reply
450 	 */
451 	char *mesg = buf;
452 	int i;
453 	int rv;
454 	int len;
455 	int npools;
456 	int *nthreads;
457 	struct net *net = netns(file);
458 
459 	mutex_lock(&nfsd_mutex);
460 	npools = nfsd_nrpools(net);
461 	if (npools == 0) {
462 		/*
463 		 * NFS is shut down.  The admin can start it by
464 		 * writing to the threads file but NOT the pool_threads
465 		 * file, sorry.  Report zero threads.
466 		 */
467 		mutex_unlock(&nfsd_mutex);
468 		strcpy(buf, "0\n");
469 		return strlen(buf);
470 	}
471 
472 	nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
473 	rv = -ENOMEM;
474 	if (nthreads == NULL)
475 		goto out_free;
476 
477 	if (size > 0) {
478 		for (i = 0; i < npools; i++) {
479 			rv = get_int(&mesg, &nthreads[i]);
480 			if (rv == -ENOENT)
481 				break;		/* fewer numbers than pools */
482 			if (rv)
483 				goto out_free;	/* syntax error */
484 			rv = -EINVAL;
485 			if (nthreads[i] < 0)
486 				goto out_free;
487 			trace_nfsd_ctl_pool_threads(net, i, nthreads[i]);
488 		}
489 
490 		/*
491 		 * There must always be a thread in pool 0; the admin
492 		 * can't shut down NFS completely using pool_threads.
493 		 */
494 		if (nthreads[0] == 0)
495 			nthreads[0] = 1;
496 
497 		rv = nfsd_set_nrthreads(i, nthreads, net);
498 		if (rv)
499 			goto out_free;
500 	}
501 
502 	rv = nfsd_get_nrthreads(npools, nthreads, net);
503 	if (rv)
504 		goto out_free;
505 
506 	mesg = buf;
507 	size = SIMPLE_TRANSACTION_LIMIT;
508 	for (i = 0; i < npools && size > 0; i++) {
509 		snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
510 		len = strlen(mesg);
511 		size -= len;
512 		mesg += len;
513 	}
514 	rv = mesg - buf;
515 out_free:
516 	kfree(nthreads);
517 	mutex_unlock(&nfsd_mutex);
518 	return rv;
519 }
520 
521 static ssize_t
nfsd_print_version_support(struct nfsd_net * nn,char * buf,int remaining,const char * sep,unsigned vers,int minor)522 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
523 		const char *sep, unsigned vers, int minor)
524 {
525 	const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
526 	bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
527 
528 	if (vers == 4 && minor >= 0 &&
529 	    !nfsd_minorversion(nn, minor, NFSD_TEST))
530 		supported = false;
531 	if (minor == 0 && supported)
532 		/*
533 		 * special case for backward compatability.
534 		 * +4.0 is never reported, it is implied by
535 		 * +4, unless -4.0 is present.
536 		 */
537 		return 0;
538 	return snprintf(buf, remaining, format, sep,
539 			supported ? '+' : '-', vers, minor);
540 }
541 
__write_versions(struct file * file,char * buf,size_t size)542 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
543 {
544 	char *mesg = buf;
545 	char *vers, *minorp, sign;
546 	int len, num, remaining;
547 	ssize_t tlen = 0;
548 	char *sep;
549 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
550 
551 	if (size > 0) {
552 		if (nn->nfsd_serv)
553 			/* Cannot change versions without updating
554 			 * nn->nfsd_serv->sv_xdrsize, and reallocing
555 			 * rq_argp and rq_resp
556 			 */
557 			return -EBUSY;
558 		if (buf[size-1] != '\n')
559 			return -EINVAL;
560 		buf[size-1] = 0;
561 		trace_nfsd_ctl_version(netns(file), buf);
562 
563 		vers = mesg;
564 		len = qword_get(&mesg, vers, size);
565 		if (len <= 0) return -EINVAL;
566 		do {
567 			enum vers_op cmd;
568 			unsigned minor;
569 			sign = *vers;
570 			if (sign == '+' || sign == '-')
571 				num = simple_strtol((vers+1), &minorp, 0);
572 			else
573 				num = simple_strtol(vers, &minorp, 0);
574 			if (*minorp == '.') {
575 				if (num != 4)
576 					return -EINVAL;
577 				if (kstrtouint(minorp+1, 0, &minor) < 0)
578 					return -EINVAL;
579 			}
580 
581 			cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
582 			switch(num) {
583 #ifdef CONFIG_NFSD_V2
584 			case 2:
585 #endif
586 			case 3:
587 				nfsd_vers(nn, num, cmd);
588 				break;
589 			case 4:
590 				if (*minorp == '.') {
591 					if (nfsd_minorversion(nn, minor, cmd) < 0)
592 						return -EINVAL;
593 				} else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
594 					/*
595 					 * Either we have +4 and no minors are enabled,
596 					 * or we have -4 and at least one minor is enabled.
597 					 * In either case, propagate 'cmd' to all minors.
598 					 */
599 					minor = 0;
600 					while (nfsd_minorversion(nn, minor, cmd) >= 0)
601 						minor++;
602 				}
603 				break;
604 			default:
605 				/* Ignore requests to disable non-existent versions */
606 				if (cmd == NFSD_SET)
607 					return -EINVAL;
608 			}
609 			vers += len + 1;
610 		} while ((len = qword_get(&mesg, vers, size)) > 0);
611 		/* If all get turned off, turn them back on, as
612 		 * having no versions is BAD
613 		 */
614 		nfsd_reset_versions(nn);
615 	}
616 
617 	/* Now write current state into reply buffer */
618 	sep = "";
619 	remaining = SIMPLE_TRANSACTION_LIMIT;
620 	for (num=2 ; num <= 4 ; num++) {
621 		int minor;
622 		if (!nfsd_vers(nn, num, NFSD_AVAIL))
623 			continue;
624 
625 		minor = -1;
626 		do {
627 			len = nfsd_print_version_support(nn, buf, remaining,
628 					sep, num, minor);
629 			if (len >= remaining)
630 				goto out;
631 			remaining -= len;
632 			buf += len;
633 			tlen += len;
634 			minor++;
635 			if (len)
636 				sep = " ";
637 		} while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
638 	}
639 out:
640 	len = snprintf(buf, remaining, "\n");
641 	if (len >= remaining)
642 		return -EINVAL;
643 	return tlen + len;
644 }
645 
646 /*
647  * write_versions - Set or report the available NFS protocol versions
648  *
649  * Input:
650  *			buf:		ignored
651  *			size:		zero
652  * Output:
653  *	On success:	passed-in buffer filled with '\n'-terminated C
654  *			string containing positive or negative integer
655  *			values representing the current status of each
656  *			protocol version;
657  *			return code is the size in bytes of the string
658  *	On error:	return code is zero or a negative errno value
659  *
660  * OR
661  *
662  * Input:
663  *			buf:		C string containing whitespace-
664  *					separated positive or negative
665  *					integer values representing NFS
666  *					protocol versions to enable ("+n")
667  *					or disable ("-n")
668  *			size:		non-zero length of C string in @buf
669  * Output:
670  *	On success:	status of zero or more protocol versions has
671  *			been updated; passed-in buffer filled with
672  *			'\n'-terminated C string containing positive
673  *			or negative integer values representing the
674  *			current status of each protocol version;
675  *			return code is the size in bytes of the string
676  *	On error:	return code is zero or a negative errno value
677  */
write_versions(struct file * file,char * buf,size_t size)678 static ssize_t write_versions(struct file *file, char *buf, size_t size)
679 {
680 	ssize_t rv;
681 
682 	mutex_lock(&nfsd_mutex);
683 	rv = __write_versions(file, buf, size);
684 	mutex_unlock(&nfsd_mutex);
685 	return rv;
686 }
687 
688 /*
689  * Zero-length write.  Return a list of NFSD's current listener
690  * transports.
691  */
__write_ports_names(char * buf,struct net * net)692 static ssize_t __write_ports_names(char *buf, struct net *net)
693 {
694 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
695 
696 	if (nn->nfsd_serv == NULL)
697 		return 0;
698 	return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
699 }
700 
701 /*
702  * A single 'fd' number was written, in which case it must be for
703  * a socket of a supported family/protocol, and we use it as an
704  * nfsd listener.
705  */
__write_ports_addfd(char * buf,struct net * net,const struct cred * cred)706 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
707 {
708 	char *mesg = buf;
709 	int fd, err;
710 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
711 	struct svc_serv *serv;
712 
713 	err = get_int(&mesg, &fd);
714 	if (err != 0 || fd < 0)
715 		return -EINVAL;
716 	trace_nfsd_ctl_ports_addfd(net, fd);
717 
718 	err = nfsd_create_serv(net);
719 	if (err != 0)
720 		return err;
721 
722 	serv = nn->nfsd_serv;
723 	err = svc_addsock(serv, net, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
724 
725 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
726 		nfsd_destroy_serv(net);
727 
728 	return err;
729 }
730 
731 /*
732  * A transport listener is added by writing its transport name and
733  * a port number.
734  */
__write_ports_addxprt(char * buf,struct net * net,const struct cred * cred)735 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
736 {
737 	char transport[16];
738 	struct svc_xprt *xprt;
739 	int port, err;
740 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
741 	struct svc_serv *serv;
742 
743 	if (sscanf(buf, "%15s %5u", transport, &port) != 2)
744 		return -EINVAL;
745 
746 	if (port < 1 || port > USHRT_MAX)
747 		return -EINVAL;
748 	trace_nfsd_ctl_ports_addxprt(net, transport, port);
749 
750 	err = nfsd_create_serv(net);
751 	if (err != 0)
752 		return err;
753 
754 	serv = nn->nfsd_serv;
755 	err = svc_xprt_create(serv, transport, net,
756 			      PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
757 	if (err < 0)
758 		goto out_err;
759 
760 	err = svc_xprt_create(serv, transport, net,
761 			      PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
762 	if (err < 0 && err != -EAFNOSUPPORT)
763 		goto out_close;
764 
765 	return 0;
766 out_close:
767 	xprt = svc_find_xprt(serv, transport, net, PF_INET, port);
768 	if (xprt != NULL) {
769 		svc_xprt_close(xprt);
770 		svc_xprt_put(xprt);
771 	}
772 out_err:
773 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
774 		nfsd_destroy_serv(net);
775 
776 	return err;
777 }
778 
__write_ports(struct file * file,char * buf,size_t size,struct net * net)779 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
780 			     struct net *net)
781 {
782 	if (size == 0)
783 		return __write_ports_names(buf, net);
784 
785 	if (isdigit(buf[0]))
786 		return __write_ports_addfd(buf, net, file->f_cred);
787 
788 	if (isalpha(buf[0]))
789 		return __write_ports_addxprt(buf, net, file->f_cred);
790 
791 	return -EINVAL;
792 }
793 
794 /*
795  * write_ports - Pass a socket file descriptor or transport name to listen on
796  *
797  * Input:
798  *			buf:		ignored
799  *			size:		zero
800  * Output:
801  *	On success:	passed-in buffer filled with a '\n'-terminated C
802  *			string containing a whitespace-separated list of
803  *			named NFSD listeners;
804  *			return code is the size in bytes of the string
805  *	On error:	return code is zero or a negative errno value
806  *
807  * OR
808  *
809  * Input:
810  *			buf:		C string containing an unsigned
811  *					integer value representing a bound
812  *					but unconnected socket that is to be
813  *					used as an NFSD listener; listen(3)
814  *					must be called for a SOCK_STREAM
815  *					socket, otherwise it is ignored
816  *			size:		non-zero length of C string in @buf
817  * Output:
818  *	On success:	NFS service is started;
819  *			passed-in buffer filled with a '\n'-terminated C
820  *			string containing a unique alphanumeric name of
821  *			the listener;
822  *			return code is the size in bytes of the string
823  *	On error:	return code is a negative errno value
824  *
825  * OR
826  *
827  * Input:
828  *			buf:		C string containing a transport
829  *					name and an unsigned integer value
830  *					representing the port to listen on,
831  *					separated by whitespace
832  *			size:		non-zero length of C string in @buf
833  * Output:
834  *	On success:	returns zero; NFS service is started
835  *	On error:	return code is a negative errno value
836  */
write_ports(struct file * file,char * buf,size_t size)837 static ssize_t write_ports(struct file *file, char *buf, size_t size)
838 {
839 	ssize_t rv;
840 
841 	mutex_lock(&nfsd_mutex);
842 	rv = __write_ports(file, buf, size, netns(file));
843 	mutex_unlock(&nfsd_mutex);
844 	return rv;
845 }
846 
847 
848 int nfsd_max_blksize;
849 
850 /*
851  * write_maxblksize - Set or report the current NFS blksize
852  *
853  * Input:
854  *			buf:		ignored
855  *			size:		zero
856  *
857  * OR
858  *
859  * Input:
860  *			buf:		C string containing an unsigned
861  *					integer value representing the new
862  *					NFS blksize
863  *			size:		non-zero length of C string in @buf
864  * Output:
865  *	On success:	passed-in buffer filled with '\n'-terminated C string
866  *			containing numeric value of the current NFS blksize
867  *			setting;
868  *			return code is the size in bytes of the string
869  *	On error:	return code is zero or a negative errno value
870  */
write_maxblksize(struct file * file,char * buf,size_t size)871 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
872 {
873 	char *mesg = buf;
874 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
875 
876 	if (size > 0) {
877 		int bsize;
878 		int rv = get_int(&mesg, &bsize);
879 		if (rv)
880 			return rv;
881 		trace_nfsd_ctl_maxblksize(netns(file), bsize);
882 
883 		/* force bsize into allowed range and
884 		 * required alignment.
885 		 */
886 		bsize = max_t(int, bsize, 1024);
887 		bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
888 		bsize &= ~(1024-1);
889 		mutex_lock(&nfsd_mutex);
890 		if (nn->nfsd_serv) {
891 			mutex_unlock(&nfsd_mutex);
892 			return -EBUSY;
893 		}
894 		nfsd_max_blksize = bsize;
895 		mutex_unlock(&nfsd_mutex);
896 	}
897 
898 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
899 							nfsd_max_blksize);
900 }
901 
902 #ifdef CONFIG_NFSD_V4
__nfsd4_write_time(struct file * file,char * buf,size_t size,time64_t * time,struct nfsd_net * nn)903 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
904 				  time64_t *time, struct nfsd_net *nn)
905 {
906 	struct dentry *dentry = file_dentry(file);
907 	char *mesg = buf;
908 	int rv, i;
909 
910 	if (size > 0) {
911 		if (nn->nfsd_serv)
912 			return -EBUSY;
913 		rv = get_int(&mesg, &i);
914 		if (rv)
915 			return rv;
916 		trace_nfsd_ctl_time(netns(file), dentry->d_name.name,
917 				    dentry->d_name.len, i);
918 
919 		/*
920 		 * Some sanity checking.  We don't have a reason for
921 		 * these particular numbers, but problems with the
922 		 * extremes are:
923 		 *	- Too short: the briefest network outage may
924 		 *	  cause clients to lose all their locks.  Also,
925 		 *	  the frequent polling may be wasteful.
926 		 *	- Too long: do you really want reboot recovery
927 		 *	  to take more than an hour?  Or to make other
928 		 *	  clients wait an hour before being able to
929 		 *	  revoke a dead client's locks?
930 		 */
931 		if (i < 10 || i > 3600)
932 			return -EINVAL;
933 		*time = i;
934 	}
935 
936 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
937 }
938 
nfsd4_write_time(struct file * file,char * buf,size_t size,time64_t * time,struct nfsd_net * nn)939 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
940 				time64_t *time, struct nfsd_net *nn)
941 {
942 	ssize_t rv;
943 
944 	mutex_lock(&nfsd_mutex);
945 	rv = __nfsd4_write_time(file, buf, size, time, nn);
946 	mutex_unlock(&nfsd_mutex);
947 	return rv;
948 }
949 
950 /*
951  * write_leasetime - Set or report the current NFSv4 lease time
952  *
953  * Input:
954  *			buf:		ignored
955  *			size:		zero
956  *
957  * OR
958  *
959  * Input:
960  *			buf:		C string containing an unsigned
961  *					integer value representing the new
962  *					NFSv4 lease expiry time
963  *			size:		non-zero length of C string in @buf
964  * Output:
965  *	On success:	passed-in buffer filled with '\n'-terminated C
966  *			string containing unsigned integer value of the
967  *			current lease expiry time;
968  *			return code is the size in bytes of the string
969  *	On error:	return code is zero or a negative errno value
970  */
write_leasetime(struct file * file,char * buf,size_t size)971 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
972 {
973 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
974 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
975 }
976 
977 /*
978  * write_gracetime - Set or report current NFSv4 grace period time
979  *
980  * As above, but sets the time of the NFSv4 grace period.
981  *
982  * Note this should never be set to less than the *previous*
983  * lease-period time, but we don't try to enforce this.  (In the common
984  * case (a new boot), we don't know what the previous lease time was
985  * anyway.)
986  */
write_gracetime(struct file * file,char * buf,size_t size)987 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
988 {
989 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
990 	return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
991 }
992 
993 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
__write_recoverydir(struct file * file,char * buf,size_t size,struct nfsd_net * nn)994 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
995 				   struct nfsd_net *nn)
996 {
997 	char *mesg = buf;
998 	char *recdir;
999 	int len, status;
1000 
1001 	if (size > 0) {
1002 		if (nn->nfsd_serv)
1003 			return -EBUSY;
1004 		if (size > PATH_MAX || buf[size-1] != '\n')
1005 			return -EINVAL;
1006 		buf[size-1] = 0;
1007 
1008 		recdir = mesg;
1009 		len = qword_get(&mesg, recdir, size);
1010 		if (len <= 0)
1011 			return -EINVAL;
1012 		trace_nfsd_ctl_recoverydir(netns(file), recdir);
1013 
1014 		status = nfs4_reset_recoverydir(recdir);
1015 		if (status)
1016 			return status;
1017 	}
1018 
1019 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1020 							nfs4_recoverydir());
1021 }
1022 
1023 /*
1024  * write_recoverydir - Set or report the pathname of the recovery directory
1025  *
1026  * Input:
1027  *			buf:		ignored
1028  *			size:		zero
1029  *
1030  * OR
1031  *
1032  * Input:
1033  *			buf:		C string containing the pathname
1034  *					of the directory on a local file
1035  *					system containing permanent NFSv4
1036  *					recovery data
1037  *			size:		non-zero length of C string in @buf
1038  * Output:
1039  *	On success:	passed-in buffer filled with '\n'-terminated C string
1040  *			containing the current recovery pathname setting;
1041  *			return code is the size in bytes of the string
1042  *	On error:	return code is zero or a negative errno value
1043  */
write_recoverydir(struct file * file,char * buf,size_t size)1044 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1045 {
1046 	ssize_t rv;
1047 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1048 
1049 	mutex_lock(&nfsd_mutex);
1050 	rv = __write_recoverydir(file, buf, size, nn);
1051 	mutex_unlock(&nfsd_mutex);
1052 	return rv;
1053 }
1054 #endif
1055 
1056 /*
1057  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1058  *
1059  * Input:
1060  *			buf:		ignored
1061  *			size:		zero
1062  * OR
1063  *
1064  * Input:
1065  *			buf:		any value
1066  *			size:		non-zero length of C string in @buf
1067  * Output:
1068  *			passed-in buffer filled with "Y" or "N" with a newline
1069  *			and NULL-terminated C string. This indicates whether
1070  *			the grace period has ended in the current net
1071  *			namespace. Return code is the size in bytes of the
1072  *			string. Writing a string that starts with 'Y', 'y', or
1073  *			'1' to the file will end the grace period for nfsd's v4
1074  *			lock manager.
1075  */
write_v4_end_grace(struct file * file,char * buf,size_t size)1076 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1077 {
1078 	struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1079 
1080 	if (size > 0) {
1081 		switch(buf[0]) {
1082 		case 'Y':
1083 		case 'y':
1084 		case '1':
1085 			if (!nn->nfsd_serv)
1086 				return -EBUSY;
1087 			trace_nfsd_end_grace(netns(file));
1088 			nfsd4_end_grace(nn);
1089 			break;
1090 		default:
1091 			return -EINVAL;
1092 		}
1093 	}
1094 
1095 	return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1096 			 nn->grace_ended ? 'Y' : 'N');
1097 }
1098 
1099 #endif
1100 
1101 /*----------------------------------------------------------------------------*/
1102 /*
1103  *	populating the filesystem.
1104  */
1105 
nfsd_get_inode(struct super_block * sb,umode_t mode)1106 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1107 {
1108 	struct inode *inode = new_inode(sb);
1109 	if (inode) {
1110 		/* Following advice from simple_fill_super documentation: */
1111 		inode->i_ino = iunique(sb, NFSD_MaxReserved);
1112 		inode->i_mode = mode;
1113 		simple_inode_init_ts(inode);
1114 	}
1115 	return inode;
1116 }
1117 
nfsd_mkdir(struct dentry * parent,struct nfsdfs_client * ncl,char * name)1118 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1119 {
1120 	struct inode *dir = parent->d_inode;
1121 	struct dentry *dentry;
1122 	struct inode *inode;
1123 
1124 	inode = nfsd_get_inode(parent->d_sb, S_IFDIR | 0600);
1125 	if (!inode)
1126 		return ERR_PTR(-ENOMEM);
1127 
1128 	dentry = simple_start_creating(parent, name);
1129 	if (IS_ERR(dentry)) {
1130 		iput(inode);
1131 		return dentry;
1132 	}
1133 	inode->i_fop = &simple_dir_operations;
1134 	inode->i_op = &simple_dir_inode_operations;
1135 	inc_nlink(inode);
1136 	if (ncl) {
1137 		inode->i_private = ncl;
1138 		kref_get(&ncl->cl_ref);
1139 	}
1140 	d_instantiate(dentry, inode);
1141 	inc_nlink(dir);
1142 	fsnotify_mkdir(dir, dentry);
1143 	inode_unlock(dir);
1144 	return dentry;
1145 }
1146 
1147 #if IS_ENABLED(CONFIG_SUNRPC_GSS)
1148 /*
1149  * @content is assumed to be a NUL-terminated string that lives
1150  * longer than the symlink itself.
1151  */
_nfsd_symlink(struct dentry * parent,const char * name,const char * content)1152 static void _nfsd_symlink(struct dentry *parent, const char *name,
1153 			  const char *content)
1154 {
1155 	struct inode *dir = parent->d_inode;
1156 	struct inode *inode;
1157 	struct dentry *dentry;
1158 
1159 	inode = nfsd_get_inode(dir->i_sb, S_IFLNK | 0777);
1160 	if (!inode)
1161 		return;
1162 
1163 	dentry = simple_start_creating(parent, name);
1164 	if (IS_ERR(dentry)) {
1165 		iput(inode);
1166 		return;
1167 	}
1168 
1169 	inode->i_op = &simple_symlink_inode_operations;
1170 	inode->i_link = (char *)content;
1171 	inode->i_size = strlen(content);
1172 
1173 	d_instantiate(dentry, inode);
1174 	fsnotify_create(dir, dentry);
1175 	inode_unlock(dir);
1176 }
1177 #else
_nfsd_symlink(struct dentry * parent,const char * name,const char * content)1178 static inline void _nfsd_symlink(struct dentry *parent, const char *name,
1179 				 const char *content)
1180 {
1181 }
1182 
1183 #endif
1184 
clear_ncl(struct dentry * dentry)1185 static void clear_ncl(struct dentry *dentry)
1186 {
1187 	struct inode *inode = d_inode(dentry);
1188 	struct nfsdfs_client *ncl = inode->i_private;
1189 
1190 	spin_lock(&inode->i_lock);
1191 	inode->i_private = NULL;
1192 	spin_unlock(&inode->i_lock);
1193 	kref_put(&ncl->cl_ref, ncl->cl_release);
1194 }
1195 
get_nfsdfs_client(struct inode * inode)1196 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1197 {
1198 	struct nfsdfs_client *nc;
1199 
1200 	spin_lock(&inode->i_lock);
1201 	nc = inode->i_private;
1202 	if (nc)
1203 		kref_get(&nc->cl_ref);
1204 	spin_unlock(&inode->i_lock);
1205 	return nc;
1206 }
1207 
1208 /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1209  * code instead. */
nfsdfs_create_files(struct dentry * root,const struct tree_descr * files,struct nfsdfs_client * ncl,struct dentry ** fdentries)1210 static int nfsdfs_create_files(struct dentry *root,
1211 				const struct tree_descr *files,
1212 				struct nfsdfs_client *ncl,
1213 				struct dentry **fdentries)
1214 {
1215 	struct inode *dir = d_inode(root);
1216 	struct dentry *dentry;
1217 
1218 	for (int i = 0; files->name && files->name[0]; i++, files++) {
1219 		struct inode *inode = nfsd_get_inode(root->d_sb,
1220 						     S_IFREG | files->mode);
1221 		if (!inode)
1222 			return -ENOMEM;
1223 		dentry = simple_start_creating(root, files->name);
1224 		if (IS_ERR(dentry)) {
1225 			iput(inode);
1226 			return PTR_ERR(dentry);
1227 		}
1228 		kref_get(&ncl->cl_ref);
1229 		inode->i_fop = files->ops;
1230 		inode->i_private = ncl;
1231 		d_instantiate(dentry, inode);
1232 		fsnotify_create(dir, dentry);
1233 		if (fdentries)
1234 			fdentries[i] = dentry;
1235 		inode_unlock(dir);
1236 	}
1237 	return 0;
1238 }
1239 
1240 /* on success, returns positive number unique to that client. */
nfsd_client_mkdir(struct nfsd_net * nn,struct nfsdfs_client * ncl,u32 id,const struct tree_descr * files,struct dentry ** fdentries)1241 struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1242 				 struct nfsdfs_client *ncl, u32 id,
1243 				 const struct tree_descr *files,
1244 				 struct dentry **fdentries)
1245 {
1246 	struct dentry *dentry;
1247 	char name[11];
1248 	int ret;
1249 
1250 	sprintf(name, "%u", id);
1251 
1252 	dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1253 	if (IS_ERR(dentry)) /* XXX: tossing errors? */
1254 		return NULL;
1255 	ret = nfsdfs_create_files(dentry, files, ncl, fdentries);
1256 	if (ret) {
1257 		nfsd_client_rmdir(dentry);
1258 		return NULL;
1259 	}
1260 	return dentry;
1261 }
1262 
1263 /* Taken from __rpc_rmdir: */
nfsd_client_rmdir(struct dentry * dentry)1264 void nfsd_client_rmdir(struct dentry *dentry)
1265 {
1266 	simple_recursive_removal(dentry, clear_ncl);
1267 }
1268 
nfsd_fill_super(struct super_block * sb,struct fs_context * fc)1269 static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1270 {
1271 	struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1272 							nfsd_net_id);
1273 	struct dentry *dentry;
1274 	int ret;
1275 
1276 	static const struct tree_descr nfsd_files[] = {
1277 		[NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1278 		/* Per-export io stats use same ops as exports file */
1279 		[NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1280 		[NFSD_Export_features] = {"export_features",
1281 					&export_features_fops, S_IRUGO},
1282 		[NFSD_FO_UnlockIP] = {"unlock_ip",
1283 					&transaction_ops, S_IWUSR|S_IRUSR},
1284 		[NFSD_FO_UnlockFS] = {"unlock_filesystem",
1285 					&transaction_ops, S_IWUSR|S_IRUSR},
1286 		[NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1287 		[NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1288 		[NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1289 		[NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1290 		[NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1291 					&nfsd_reply_cache_stats_fops, S_IRUGO},
1292 		[NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1293 		[NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1294 		[NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1295 		[NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
1296 #ifdef CONFIG_NFSD_V4
1297 		[NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1298 		[NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1299 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1300 		[NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1301 #endif
1302 		[NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1303 #endif
1304 		/* last one */ {""}
1305 	};
1306 
1307 	ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1308 	if (ret)
1309 		return ret;
1310 	_nfsd_symlink(sb->s_root, "supported_krb5_enctypes",
1311 		      "/proc/net/rpc/gss_krb5_enctypes");
1312 	dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1313 	if (IS_ERR(dentry))
1314 		return PTR_ERR(dentry);
1315 	nn->nfsd_client_dir = dentry;
1316 	return 0;
1317 }
1318 
nfsd_fs_get_tree(struct fs_context * fc)1319 static int nfsd_fs_get_tree(struct fs_context *fc)
1320 {
1321 	return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1322 }
1323 
nfsd_fs_free_fc(struct fs_context * fc)1324 static void nfsd_fs_free_fc(struct fs_context *fc)
1325 {
1326 	if (fc->s_fs_info)
1327 		put_net(fc->s_fs_info);
1328 }
1329 
1330 static const struct fs_context_operations nfsd_fs_context_ops = {
1331 	.free		= nfsd_fs_free_fc,
1332 	.get_tree	= nfsd_fs_get_tree,
1333 };
1334 
nfsd_init_fs_context(struct fs_context * fc)1335 static int nfsd_init_fs_context(struct fs_context *fc)
1336 {
1337 	put_user_ns(fc->user_ns);
1338 	fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1339 	fc->ops = &nfsd_fs_context_ops;
1340 	return 0;
1341 }
1342 
nfsd_umount(struct super_block * sb)1343 static void nfsd_umount(struct super_block *sb)
1344 {
1345 	struct net *net = sb->s_fs_info;
1346 
1347 	nfsd_shutdown_threads(net);
1348 
1349 	kill_litter_super(sb);
1350 	put_net(net);
1351 }
1352 
1353 static struct file_system_type nfsd_fs_type = {
1354 	.owner		= THIS_MODULE,
1355 	.name		= "nfsd",
1356 	.init_fs_context = nfsd_init_fs_context,
1357 	.kill_sb	= nfsd_umount,
1358 };
1359 MODULE_ALIAS_FS("nfsd");
1360 
1361 #ifdef CONFIG_PROC_FS
1362 
exports_proc_open(struct inode * inode,struct file * file)1363 static int exports_proc_open(struct inode *inode, struct file *file)
1364 {
1365 	return exports_net_open(current->nsproxy->net_ns, file);
1366 }
1367 
1368 static const struct proc_ops exports_proc_ops = {
1369 	.proc_open	= exports_proc_open,
1370 	.proc_read	= seq_read,
1371 	.proc_lseek	= seq_lseek,
1372 	.proc_release	= seq_release,
1373 };
1374 
create_proc_exports_entry(void)1375 static int create_proc_exports_entry(void)
1376 {
1377 	struct proc_dir_entry *entry;
1378 
1379 	entry = proc_mkdir("fs/nfs", NULL);
1380 	if (!entry)
1381 		return -ENOMEM;
1382 	entry = proc_create("exports", 0, entry, &exports_proc_ops);
1383 	if (!entry) {
1384 		remove_proc_entry("fs/nfs", NULL);
1385 		return -ENOMEM;
1386 	}
1387 	return 0;
1388 }
1389 #else /* CONFIG_PROC_FS */
create_proc_exports_entry(void)1390 static int create_proc_exports_entry(void)
1391 {
1392 	return 0;
1393 }
1394 #endif
1395 
1396 unsigned int nfsd_net_id;
1397 
nfsd_genl_rpc_status_compose_msg(struct sk_buff * skb,struct netlink_callback * cb,struct nfsd_genl_rqstp * genl_rqstp)1398 static int nfsd_genl_rpc_status_compose_msg(struct sk_buff *skb,
1399 					    struct netlink_callback *cb,
1400 					    struct nfsd_genl_rqstp *genl_rqstp)
1401 {
1402 	void *hdr;
1403 	u32 i;
1404 
1405 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
1406 			  &nfsd_nl_family, 0, NFSD_CMD_RPC_STATUS_GET);
1407 	if (!hdr)
1408 		return -ENOBUFS;
1409 
1410 	if (nla_put_be32(skb, NFSD_A_RPC_STATUS_XID, genl_rqstp->rq_xid) ||
1411 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_FLAGS, genl_rqstp->rq_flags) ||
1412 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROG, genl_rqstp->rq_prog) ||
1413 	    nla_put_u32(skb, NFSD_A_RPC_STATUS_PROC, genl_rqstp->rq_proc) ||
1414 	    nla_put_u8(skb, NFSD_A_RPC_STATUS_VERSION, genl_rqstp->rq_vers) ||
1415 	    nla_put_s64(skb, NFSD_A_RPC_STATUS_SERVICE_TIME,
1416 			ktime_to_us(genl_rqstp->rq_stime),
1417 			NFSD_A_RPC_STATUS_PAD))
1418 		return -ENOBUFS;
1419 
1420 	switch (genl_rqstp->rq_saddr.sa_family) {
1421 	case AF_INET: {
1422 		const struct sockaddr_in *s_in, *d_in;
1423 
1424 		s_in = (const struct sockaddr_in *)&genl_rqstp->rq_saddr;
1425 		d_in = (const struct sockaddr_in *)&genl_rqstp->rq_daddr;
1426 		if (nla_put_in_addr(skb, NFSD_A_RPC_STATUS_SADDR4,
1427 				    s_in->sin_addr.s_addr) ||
1428 		    nla_put_in_addr(skb, NFSD_A_RPC_STATUS_DADDR4,
1429 				    d_in->sin_addr.s_addr) ||
1430 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1431 				 s_in->sin_port) ||
1432 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1433 				 d_in->sin_port))
1434 			return -ENOBUFS;
1435 		break;
1436 	}
1437 	case AF_INET6: {
1438 		const struct sockaddr_in6 *s_in, *d_in;
1439 
1440 		s_in = (const struct sockaddr_in6 *)&genl_rqstp->rq_saddr;
1441 		d_in = (const struct sockaddr_in6 *)&genl_rqstp->rq_daddr;
1442 		if (nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_SADDR6,
1443 				     &s_in->sin6_addr) ||
1444 		    nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_DADDR6,
1445 				     &d_in->sin6_addr) ||
1446 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1447 				 s_in->sin6_port) ||
1448 		    nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1449 				 d_in->sin6_port))
1450 			return -ENOBUFS;
1451 		break;
1452 	}
1453 	}
1454 
1455 	for (i = 0; i < genl_rqstp->rq_opcnt; i++)
1456 		if (nla_put_u32(skb, NFSD_A_RPC_STATUS_COMPOUND_OPS,
1457 				genl_rqstp->rq_opnum[i]))
1458 			return -ENOBUFS;
1459 
1460 	genlmsg_end(skb, hdr);
1461 	return 0;
1462 }
1463 
1464 /**
1465  * nfsd_nl_rpc_status_get_dumpit - Handle rpc_status_get dumpit
1466  * @skb: reply buffer
1467  * @cb: netlink metadata and command arguments
1468  *
1469  * Returns the size of the reply or a negative errno.
1470  */
nfsd_nl_rpc_status_get_dumpit(struct sk_buff * skb,struct netlink_callback * cb)1471 int nfsd_nl_rpc_status_get_dumpit(struct sk_buff *skb,
1472 				  struct netlink_callback *cb)
1473 {
1474 	int i, ret, rqstp_index = 0;
1475 	struct nfsd_net *nn;
1476 
1477 	mutex_lock(&nfsd_mutex);
1478 
1479 	nn = net_generic(sock_net(skb->sk), nfsd_net_id);
1480 	if (!nn->nfsd_serv) {
1481 		ret = -ENODEV;
1482 		goto out_unlock;
1483 	}
1484 
1485 	rcu_read_lock();
1486 
1487 	for (i = 0; i < nn->nfsd_serv->sv_nrpools; i++) {
1488 		struct svc_rqst *rqstp;
1489 
1490 		if (i < cb->args[0]) /* already consumed */
1491 			continue;
1492 
1493 		rqstp_index = 0;
1494 		list_for_each_entry_rcu(rqstp,
1495 				&nn->nfsd_serv->sv_pools[i].sp_all_threads,
1496 				rq_all) {
1497 			struct nfsd_genl_rqstp genl_rqstp;
1498 			unsigned int status_counter;
1499 
1500 			if (rqstp_index++ < cb->args[1]) /* already consumed */
1501 				continue;
1502 			/*
1503 			 * Acquire rq_status_counter before parsing the rqst
1504 			 * fields. rq_status_counter is set to an odd value in
1505 			 * order to notify the consumers the rqstp fields are
1506 			 * meaningful.
1507 			 */
1508 			status_counter =
1509 				smp_load_acquire(&rqstp->rq_status_counter);
1510 			if (!(status_counter & 1))
1511 				continue;
1512 
1513 			genl_rqstp.rq_xid = rqstp->rq_xid;
1514 			genl_rqstp.rq_flags = rqstp->rq_flags;
1515 			genl_rqstp.rq_vers = rqstp->rq_vers;
1516 			genl_rqstp.rq_prog = rqstp->rq_prog;
1517 			genl_rqstp.rq_proc = rqstp->rq_proc;
1518 			genl_rqstp.rq_stime = rqstp->rq_stime;
1519 			genl_rqstp.rq_opcnt = 0;
1520 			memcpy(&genl_rqstp.rq_daddr, svc_daddr(rqstp),
1521 			       sizeof(struct sockaddr));
1522 			memcpy(&genl_rqstp.rq_saddr, svc_addr(rqstp),
1523 			       sizeof(struct sockaddr));
1524 
1525 #ifdef CONFIG_NFSD_V4
1526 			if (rqstp->rq_vers == NFS4_VERSION &&
1527 			    rqstp->rq_proc == NFSPROC4_COMPOUND) {
1528 				/* NFSv4 compound */
1529 				struct nfsd4_compoundargs *args;
1530 				int j;
1531 
1532 				args = rqstp->rq_argp;
1533 				genl_rqstp.rq_opcnt = min_t(u32, args->opcnt,
1534 							    ARRAY_SIZE(genl_rqstp.rq_opnum));
1535 				for (j = 0; j < genl_rqstp.rq_opcnt; j++)
1536 					genl_rqstp.rq_opnum[j] =
1537 						args->ops[j].opnum;
1538 			}
1539 #endif /* CONFIG_NFSD_V4 */
1540 
1541 			/*
1542 			 * Acquire rq_status_counter before reporting the rqst
1543 			 * fields to the user.
1544 			 */
1545 			if (smp_load_acquire(&rqstp->rq_status_counter) !=
1546 			    status_counter)
1547 				continue;
1548 
1549 			ret = nfsd_genl_rpc_status_compose_msg(skb, cb,
1550 							       &genl_rqstp);
1551 			if (ret)
1552 				goto out;
1553 		}
1554 	}
1555 
1556 	cb->args[0] = i;
1557 	cb->args[1] = rqstp_index;
1558 	ret = skb->len;
1559 out:
1560 	rcu_read_unlock();
1561 out_unlock:
1562 	mutex_unlock(&nfsd_mutex);
1563 
1564 	return ret;
1565 }
1566 
1567 /**
1568  * nfsd_nl_threads_set_doit - set the number of running threads
1569  * @skb: reply buffer
1570  * @info: netlink metadata and command arguments
1571  *
1572  * Return 0 on success or a negative errno.
1573  */
nfsd_nl_threads_set_doit(struct sk_buff * skb,struct genl_info * info)1574 int nfsd_nl_threads_set_doit(struct sk_buff *skb, struct genl_info *info)
1575 {
1576 	int *nthreads, nrpools = 0, i, ret = -EOPNOTSUPP, rem;
1577 	struct net *net = genl_info_net(info);
1578 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1579 	const struct nlattr *attr;
1580 	const char *scope = NULL;
1581 
1582 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_SERVER_THREADS))
1583 		return -EINVAL;
1584 
1585 	/* count number of SERVER_THREADS values */
1586 	nlmsg_for_each_attr_type(attr, NFSD_A_SERVER_THREADS, info->nlhdr,
1587 				 GENL_HDRLEN, rem)
1588 		nrpools++;
1589 
1590 	mutex_lock(&nfsd_mutex);
1591 
1592 	nthreads = kcalloc(nrpools, sizeof(int), GFP_KERNEL);
1593 	if (!nthreads) {
1594 		ret = -ENOMEM;
1595 		goto out_unlock;
1596 	}
1597 
1598 	i = 0;
1599 	nlmsg_for_each_attr_type(attr, NFSD_A_SERVER_THREADS, info->nlhdr,
1600 				 GENL_HDRLEN, rem) {
1601 		nthreads[i++] = nla_get_u32(attr);
1602 		if (i >= nrpools)
1603 			break;
1604 	}
1605 
1606 	if (info->attrs[NFSD_A_SERVER_GRACETIME] ||
1607 	    info->attrs[NFSD_A_SERVER_LEASETIME] ||
1608 	    info->attrs[NFSD_A_SERVER_SCOPE]) {
1609 		ret = -EBUSY;
1610 		if (nn->nfsd_serv && nn->nfsd_serv->sv_nrthreads)
1611 			goto out_unlock;
1612 
1613 		ret = -EINVAL;
1614 		attr = info->attrs[NFSD_A_SERVER_GRACETIME];
1615 		if (attr) {
1616 			u32 gracetime = nla_get_u32(attr);
1617 
1618 			if (gracetime < 10 || gracetime > 3600)
1619 				goto out_unlock;
1620 
1621 			nn->nfsd4_grace = gracetime;
1622 		}
1623 
1624 		attr = info->attrs[NFSD_A_SERVER_LEASETIME];
1625 		if (attr) {
1626 			u32 leasetime = nla_get_u32(attr);
1627 
1628 			if (leasetime < 10 || leasetime > 3600)
1629 				goto out_unlock;
1630 
1631 			nn->nfsd4_lease = leasetime;
1632 		}
1633 
1634 		attr = info->attrs[NFSD_A_SERVER_SCOPE];
1635 		if (attr)
1636 			scope = nla_data(attr);
1637 	}
1638 
1639 	ret = nfsd_svc(nrpools, nthreads, net, get_current_cred(), scope);
1640 	if (ret > 0)
1641 		ret = 0;
1642 out_unlock:
1643 	mutex_unlock(&nfsd_mutex);
1644 	kfree(nthreads);
1645 	return ret;
1646 }
1647 
1648 /**
1649  * nfsd_nl_threads_get_doit - get the number of running threads
1650  * @skb: reply buffer
1651  * @info: netlink metadata and command arguments
1652  *
1653  * Return 0 on success or a negative errno.
1654  */
nfsd_nl_threads_get_doit(struct sk_buff * skb,struct genl_info * info)1655 int nfsd_nl_threads_get_doit(struct sk_buff *skb, struct genl_info *info)
1656 {
1657 	struct net *net = genl_info_net(info);
1658 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1659 	void *hdr;
1660 	int err;
1661 
1662 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
1663 	if (!skb)
1664 		return -ENOMEM;
1665 
1666 	hdr = genlmsg_iput(skb, info);
1667 	if (!hdr) {
1668 		err = -EMSGSIZE;
1669 		goto err_free_msg;
1670 	}
1671 
1672 	mutex_lock(&nfsd_mutex);
1673 
1674 	err = nla_put_u32(skb, NFSD_A_SERVER_GRACETIME,
1675 			  nn->nfsd4_grace) ||
1676 	      nla_put_u32(skb, NFSD_A_SERVER_LEASETIME,
1677 			  nn->nfsd4_lease) ||
1678 	      nla_put_string(skb, NFSD_A_SERVER_SCOPE,
1679 			  nn->nfsd_name);
1680 	if (err)
1681 		goto err_unlock;
1682 
1683 	if (nn->nfsd_serv) {
1684 		int i;
1685 
1686 		for (i = 0; i < nfsd_nrpools(net); ++i) {
1687 			struct svc_pool *sp = &nn->nfsd_serv->sv_pools[i];
1688 
1689 			err = nla_put_u32(skb, NFSD_A_SERVER_THREADS,
1690 					  sp->sp_nrthreads);
1691 			if (err)
1692 				goto err_unlock;
1693 		}
1694 	} else {
1695 		err = nla_put_u32(skb, NFSD_A_SERVER_THREADS, 0);
1696 		if (err)
1697 			goto err_unlock;
1698 	}
1699 
1700 	mutex_unlock(&nfsd_mutex);
1701 
1702 	genlmsg_end(skb, hdr);
1703 
1704 	return genlmsg_reply(skb, info);
1705 
1706 err_unlock:
1707 	mutex_unlock(&nfsd_mutex);
1708 err_free_msg:
1709 	nlmsg_free(skb);
1710 
1711 	return err;
1712 }
1713 
1714 /**
1715  * nfsd_nl_version_set_doit - set the nfs enabled versions
1716  * @skb: reply buffer
1717  * @info: netlink metadata and command arguments
1718  *
1719  * Return 0 on success or a negative errno.
1720  */
nfsd_nl_version_set_doit(struct sk_buff * skb,struct genl_info * info)1721 int nfsd_nl_version_set_doit(struct sk_buff *skb, struct genl_info *info)
1722 {
1723 	const struct nlattr *attr;
1724 	struct nfsd_net *nn;
1725 	int i, rem;
1726 
1727 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_SERVER_PROTO_VERSION))
1728 		return -EINVAL;
1729 
1730 	mutex_lock(&nfsd_mutex);
1731 
1732 	nn = net_generic(genl_info_net(info), nfsd_net_id);
1733 	if (nn->nfsd_serv) {
1734 		mutex_unlock(&nfsd_mutex);
1735 		return -EBUSY;
1736 	}
1737 
1738 	/* clear current supported versions. */
1739 	nfsd_vers(nn, 2, NFSD_CLEAR);
1740 	nfsd_vers(nn, 3, NFSD_CLEAR);
1741 	for (i = 0; i <= NFSD_SUPPORTED_MINOR_VERSION; i++)
1742 		nfsd_minorversion(nn, i, NFSD_CLEAR);
1743 
1744 	nlmsg_for_each_attr_type(attr, NFSD_A_SERVER_PROTO_VERSION, info->nlhdr,
1745 				 GENL_HDRLEN, rem) {
1746 		struct nlattr *tb[NFSD_A_VERSION_MAX + 1];
1747 		u32 major, minor = 0;
1748 		bool enabled;
1749 
1750 		if (nla_parse_nested(tb, NFSD_A_VERSION_MAX, attr,
1751 				     nfsd_version_nl_policy, info->extack) < 0)
1752 			continue;
1753 
1754 		if (!tb[NFSD_A_VERSION_MAJOR])
1755 			continue;
1756 
1757 		major = nla_get_u32(tb[NFSD_A_VERSION_MAJOR]);
1758 		if (tb[NFSD_A_VERSION_MINOR])
1759 			minor = nla_get_u32(tb[NFSD_A_VERSION_MINOR]);
1760 
1761 		enabled = nla_get_flag(tb[NFSD_A_VERSION_ENABLED]);
1762 
1763 		switch (major) {
1764 		case 4:
1765 			nfsd_minorversion(nn, minor, enabled ? NFSD_SET : NFSD_CLEAR);
1766 			break;
1767 		case 3:
1768 		case 2:
1769 			if (!minor)
1770 				nfsd_vers(nn, major, enabled ? NFSD_SET : NFSD_CLEAR);
1771 			break;
1772 		default:
1773 			break;
1774 		}
1775 	}
1776 
1777 	mutex_unlock(&nfsd_mutex);
1778 
1779 	return 0;
1780 }
1781 
1782 /**
1783  * nfsd_nl_version_get_doit - get the enabled status for all supported nfs versions
1784  * @skb: reply buffer
1785  * @info: netlink metadata and command arguments
1786  *
1787  * Return 0 on success or a negative errno.
1788  */
nfsd_nl_version_get_doit(struct sk_buff * skb,struct genl_info * info)1789 int nfsd_nl_version_get_doit(struct sk_buff *skb, struct genl_info *info)
1790 {
1791 	struct nfsd_net *nn;
1792 	int i, err;
1793 	void *hdr;
1794 
1795 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
1796 	if (!skb)
1797 		return -ENOMEM;
1798 
1799 	hdr = genlmsg_iput(skb, info);
1800 	if (!hdr) {
1801 		err = -EMSGSIZE;
1802 		goto err_free_msg;
1803 	}
1804 
1805 	mutex_lock(&nfsd_mutex);
1806 	nn = net_generic(genl_info_net(info), nfsd_net_id);
1807 
1808 	for (i = 2; i <= 4; i++) {
1809 		int j;
1810 
1811 		for (j = 0; j <= NFSD_SUPPORTED_MINOR_VERSION; j++) {
1812 			struct nlattr *attr;
1813 
1814 			/* Don't record any versions the kernel doesn't have
1815 			 * compiled in
1816 			 */
1817 			if (!nfsd_support_version(i))
1818 				continue;
1819 
1820 			/* NFSv{2,3} does not support minor numbers */
1821 			if (i < 4 && j)
1822 				continue;
1823 
1824 			attr = nla_nest_start(skb,
1825 					      NFSD_A_SERVER_PROTO_VERSION);
1826 			if (!attr) {
1827 				err = -EINVAL;
1828 				goto err_nfsd_unlock;
1829 			}
1830 
1831 			if (nla_put_u32(skb, NFSD_A_VERSION_MAJOR, i) ||
1832 			    nla_put_u32(skb, NFSD_A_VERSION_MINOR, j)) {
1833 				err = -EINVAL;
1834 				goto err_nfsd_unlock;
1835 			}
1836 
1837 			/* Set the enabled flag if the version is enabled */
1838 			if (nfsd_vers(nn, i, NFSD_TEST) &&
1839 			    (i < 4 || nfsd_minorversion(nn, j, NFSD_TEST)) &&
1840 			    nla_put_flag(skb, NFSD_A_VERSION_ENABLED)) {
1841 				err = -EINVAL;
1842 				goto err_nfsd_unlock;
1843 			}
1844 
1845 			nla_nest_end(skb, attr);
1846 		}
1847 	}
1848 
1849 	mutex_unlock(&nfsd_mutex);
1850 	genlmsg_end(skb, hdr);
1851 
1852 	return genlmsg_reply(skb, info);
1853 
1854 err_nfsd_unlock:
1855 	mutex_unlock(&nfsd_mutex);
1856 err_free_msg:
1857 	nlmsg_free(skb);
1858 
1859 	return err;
1860 }
1861 
1862 /**
1863  * nfsd_nl_listener_set_doit - set the nfs running sockets
1864  * @skb: reply buffer
1865  * @info: netlink metadata and command arguments
1866  *
1867  * Return 0 on success or a negative errno.
1868  */
nfsd_nl_listener_set_doit(struct sk_buff * skb,struct genl_info * info)1869 int nfsd_nl_listener_set_doit(struct sk_buff *skb, struct genl_info *info)
1870 {
1871 	struct net *net = genl_info_net(info);
1872 	struct svc_xprt *xprt, *tmp;
1873 	const struct nlattr *attr;
1874 	struct svc_serv *serv;
1875 	LIST_HEAD(permsocks);
1876 	struct nfsd_net *nn;
1877 	bool delete = false;
1878 	int err, rem;
1879 
1880 	mutex_lock(&nfsd_mutex);
1881 
1882 	err = nfsd_create_serv(net);
1883 	if (err) {
1884 		mutex_unlock(&nfsd_mutex);
1885 		return err;
1886 	}
1887 
1888 	nn = net_generic(net, nfsd_net_id);
1889 	serv = nn->nfsd_serv;
1890 
1891 	spin_lock_bh(&serv->sv_lock);
1892 
1893 	/* Move all of the old listener sockets to a temp list */
1894 	list_splice_init(&serv->sv_permsocks, &permsocks);
1895 
1896 	/*
1897 	 * Walk the list of server_socks from userland and move any that match
1898 	 * back to sv_permsocks
1899 	 */
1900 	nlmsg_for_each_attr_type(attr, NFSD_A_SERVER_SOCK_ADDR, info->nlhdr,
1901 				 GENL_HDRLEN, rem) {
1902 		struct nlattr *tb[NFSD_A_SOCK_MAX + 1];
1903 		const char *xcl_name;
1904 		struct sockaddr *sa;
1905 
1906 		if (nla_parse_nested(tb, NFSD_A_SOCK_MAX, attr,
1907 				     nfsd_sock_nl_policy, info->extack) < 0)
1908 			continue;
1909 
1910 		if (!tb[NFSD_A_SOCK_ADDR] || !tb[NFSD_A_SOCK_TRANSPORT_NAME])
1911 			continue;
1912 
1913 		if (nla_len(tb[NFSD_A_SOCK_ADDR]) < sizeof(*sa))
1914 			continue;
1915 
1916 		xcl_name = nla_data(tb[NFSD_A_SOCK_TRANSPORT_NAME]);
1917 		sa = nla_data(tb[NFSD_A_SOCK_ADDR]);
1918 
1919 		/* Put back any matching sockets */
1920 		list_for_each_entry_safe(xprt, tmp, &permsocks, xpt_list) {
1921 			/* This shouldn't be possible */
1922 			if (WARN_ON_ONCE(xprt->xpt_net != net)) {
1923 				list_move(&xprt->xpt_list, &serv->sv_permsocks);
1924 				continue;
1925 			}
1926 
1927 			/* If everything matches, put it back */
1928 			if (!strcmp(xprt->xpt_class->xcl_name, xcl_name) &&
1929 			    rpc_cmp_addr_port(sa, (struct sockaddr *)&xprt->xpt_local)) {
1930 				list_move(&xprt->xpt_list, &serv->sv_permsocks);
1931 				break;
1932 			}
1933 		}
1934 	}
1935 
1936 	/*
1937 	 * If there are listener transports remaining on the permsocks list,
1938 	 * it means we were asked to remove a listener.
1939 	 */
1940 	if (!list_empty(&permsocks)) {
1941 		list_splice_init(&permsocks, &serv->sv_permsocks);
1942 		delete = true;
1943 	}
1944 	spin_unlock_bh(&serv->sv_lock);
1945 
1946 	/* Do not remove listeners while there are active threads. */
1947 	if (serv->sv_nrthreads) {
1948 		err = -EBUSY;
1949 		goto out_unlock_mtx;
1950 	}
1951 
1952 	/*
1953 	 * Since we can't delete an arbitrary llist entry, destroy the
1954 	 * remaining listeners and recreate the list.
1955 	 */
1956 	if (delete)
1957 		svc_xprt_destroy_all(serv, net, false);
1958 
1959 	/* walk list of addrs again, open any that still don't exist */
1960 	nlmsg_for_each_attr_type(attr, NFSD_A_SERVER_SOCK_ADDR, info->nlhdr,
1961 				 GENL_HDRLEN, rem) {
1962 		struct nlattr *tb[NFSD_A_SOCK_MAX + 1];
1963 		const char *xcl_name;
1964 		struct sockaddr *sa;
1965 		int ret;
1966 
1967 		if (nla_parse_nested(tb, NFSD_A_SOCK_MAX, attr,
1968 				     nfsd_sock_nl_policy, info->extack) < 0)
1969 			continue;
1970 
1971 		if (!tb[NFSD_A_SOCK_ADDR] || !tb[NFSD_A_SOCK_TRANSPORT_NAME])
1972 			continue;
1973 
1974 		if (nla_len(tb[NFSD_A_SOCK_ADDR]) < sizeof(*sa))
1975 			continue;
1976 
1977 		xcl_name = nla_data(tb[NFSD_A_SOCK_TRANSPORT_NAME]);
1978 		sa = nla_data(tb[NFSD_A_SOCK_ADDR]);
1979 
1980 		xprt = svc_find_listener(serv, xcl_name, net, sa);
1981 		if (xprt) {
1982 			if (delete)
1983 				WARN_ONCE(1, "Transport type=%s already exists\n",
1984 					  xcl_name);
1985 			svc_xprt_put(xprt);
1986 			continue;
1987 		}
1988 
1989 		ret = svc_xprt_create_from_sa(serv, xcl_name, net, sa, 0,
1990 					      get_current_cred());
1991 		/* always save the latest error */
1992 		if (ret < 0)
1993 			err = ret;
1994 	}
1995 
1996 	if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
1997 		nfsd_destroy_serv(net);
1998 
1999 out_unlock_mtx:
2000 	mutex_unlock(&nfsd_mutex);
2001 
2002 	return err;
2003 }
2004 
2005 /**
2006  * nfsd_nl_listener_get_doit - get the nfs running listeners
2007  * @skb: reply buffer
2008  * @info: netlink metadata and command arguments
2009  *
2010  * Return 0 on success or a negative errno.
2011  */
nfsd_nl_listener_get_doit(struct sk_buff * skb,struct genl_info * info)2012 int nfsd_nl_listener_get_doit(struct sk_buff *skb, struct genl_info *info)
2013 {
2014 	struct svc_xprt *xprt;
2015 	struct svc_serv *serv;
2016 	struct nfsd_net *nn;
2017 	void *hdr;
2018 	int err;
2019 
2020 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
2021 	if (!skb)
2022 		return -ENOMEM;
2023 
2024 	hdr = genlmsg_iput(skb, info);
2025 	if (!hdr) {
2026 		err = -EMSGSIZE;
2027 		goto err_free_msg;
2028 	}
2029 
2030 	mutex_lock(&nfsd_mutex);
2031 	nn = net_generic(genl_info_net(info), nfsd_net_id);
2032 
2033 	/* no nfs server? Just send empty socket list */
2034 	if (!nn->nfsd_serv)
2035 		goto out_unlock_mtx;
2036 
2037 	serv = nn->nfsd_serv;
2038 	spin_lock_bh(&serv->sv_lock);
2039 	list_for_each_entry(xprt, &serv->sv_permsocks, xpt_list) {
2040 		struct nlattr *attr;
2041 
2042 		attr = nla_nest_start(skb, NFSD_A_SERVER_SOCK_ADDR);
2043 		if (!attr) {
2044 			err = -EINVAL;
2045 			goto err_serv_unlock;
2046 		}
2047 
2048 		if (nla_put_string(skb, NFSD_A_SOCK_TRANSPORT_NAME,
2049 				   xprt->xpt_class->xcl_name) ||
2050 		    nla_put(skb, NFSD_A_SOCK_ADDR,
2051 			    sizeof(struct sockaddr_storage),
2052 			    &xprt->xpt_local)) {
2053 			err = -EINVAL;
2054 			goto err_serv_unlock;
2055 		}
2056 
2057 		nla_nest_end(skb, attr);
2058 	}
2059 	spin_unlock_bh(&serv->sv_lock);
2060 out_unlock_mtx:
2061 	mutex_unlock(&nfsd_mutex);
2062 	genlmsg_end(skb, hdr);
2063 
2064 	return genlmsg_reply(skb, info);
2065 
2066 err_serv_unlock:
2067 	spin_unlock_bh(&serv->sv_lock);
2068 	mutex_unlock(&nfsd_mutex);
2069 err_free_msg:
2070 	nlmsg_free(skb);
2071 
2072 	return err;
2073 }
2074 
2075 /**
2076  * nfsd_nl_pool_mode_set_doit - set the number of running threads
2077  * @skb: reply buffer
2078  * @info: netlink metadata and command arguments
2079  *
2080  * Return 0 on success or a negative errno.
2081  */
nfsd_nl_pool_mode_set_doit(struct sk_buff * skb,struct genl_info * info)2082 int nfsd_nl_pool_mode_set_doit(struct sk_buff *skb, struct genl_info *info)
2083 {
2084 	const struct nlattr *attr;
2085 
2086 	if (GENL_REQ_ATTR_CHECK(info, NFSD_A_POOL_MODE_MODE))
2087 		return -EINVAL;
2088 
2089 	attr = info->attrs[NFSD_A_POOL_MODE_MODE];
2090 	return sunrpc_set_pool_mode(nla_data(attr));
2091 }
2092 
2093 /**
2094  * nfsd_nl_pool_mode_get_doit - get info about pool_mode
2095  * @skb: reply buffer
2096  * @info: netlink metadata and command arguments
2097  *
2098  * Return 0 on success or a negative errno.
2099  */
nfsd_nl_pool_mode_get_doit(struct sk_buff * skb,struct genl_info * info)2100 int nfsd_nl_pool_mode_get_doit(struct sk_buff *skb, struct genl_info *info)
2101 {
2102 	struct net *net = genl_info_net(info);
2103 	char buf[16];
2104 	void *hdr;
2105 	int err;
2106 
2107 	if (sunrpc_get_pool_mode(buf, ARRAY_SIZE(buf)) >= ARRAY_SIZE(buf))
2108 		return -ERANGE;
2109 
2110 	skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
2111 	if (!skb)
2112 		return -ENOMEM;
2113 
2114 	err = -EMSGSIZE;
2115 	hdr = genlmsg_iput(skb, info);
2116 	if (!hdr)
2117 		goto err_free_msg;
2118 
2119 	err = nla_put_string(skb, NFSD_A_POOL_MODE_MODE, buf) |
2120 	      nla_put_u32(skb, NFSD_A_POOL_MODE_NPOOLS, nfsd_nrpools(net));
2121 	if (err)
2122 		goto err_free_msg;
2123 
2124 	genlmsg_end(skb, hdr);
2125 	return genlmsg_reply(skb, info);
2126 
2127 err_free_msg:
2128 	nlmsg_free(skb);
2129 	return err;
2130 }
2131 
2132 /**
2133  * nfsd_net_init - Prepare the nfsd_net portion of a new net namespace
2134  * @net: a freshly-created network namespace
2135  *
2136  * This information stays around as long as the network namespace is
2137  * alive whether or not there is an NFSD instance running in the
2138  * namespace.
2139  *
2140  * Returns zero on success, or a negative errno otherwise.
2141  */
nfsd_net_init(struct net * net)2142 static __net_init int nfsd_net_init(struct net *net)
2143 {
2144 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2145 	int retval;
2146 	int i;
2147 
2148 	retval = nfsd_export_init(net);
2149 	if (retval)
2150 		goto out_export_error;
2151 	retval = nfsd_idmap_init(net);
2152 	if (retval)
2153 		goto out_idmap_error;
2154 	retval = percpu_counter_init_many(nn->counter, 0, GFP_KERNEL,
2155 					  NFSD_STATS_COUNTERS_NUM);
2156 	if (retval)
2157 		goto out_repcache_error;
2158 
2159 	memset(&nn->nfsd_svcstats, 0, sizeof(nn->nfsd_svcstats));
2160 	nn->nfsd_svcstats.program = &nfsd_programs[0];
2161 	if (!nfsd_proc_stat_init(net)) {
2162 		retval = -ENOMEM;
2163 		goto out_proc_error;
2164 	}
2165 
2166 	for (i = 0; i < sizeof(nn->nfsd_versions); i++)
2167 		nn->nfsd_versions[i] = nfsd_support_version(i);
2168 	for (i = 0; i < sizeof(nn->nfsd4_minorversions); i++)
2169 		nn->nfsd4_minorversions[i] = nfsd_support_version(4);
2170 	nn->nfsd_info.mutex = &nfsd_mutex;
2171 	nn->nfsd_serv = NULL;
2172 	nfsd4_init_leases_net(nn);
2173 	get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
2174 	seqlock_init(&nn->writeverf_lock);
2175 #if IS_ENABLED(CONFIG_NFS_LOCALIO)
2176 	spin_lock_init(&nn->local_clients_lock);
2177 	INIT_LIST_HEAD(&nn->local_clients);
2178 #endif
2179 	return 0;
2180 
2181 out_proc_error:
2182 	percpu_counter_destroy_many(nn->counter, NFSD_STATS_COUNTERS_NUM);
2183 out_repcache_error:
2184 	nfsd_idmap_shutdown(net);
2185 out_idmap_error:
2186 	nfsd_export_shutdown(net);
2187 out_export_error:
2188 	return retval;
2189 }
2190 
2191 #if IS_ENABLED(CONFIG_NFS_LOCALIO)
2192 /**
2193  * nfsd_net_pre_exit - Disconnect localio clients from net namespace
2194  * @net: a network namespace that is about to be destroyed
2195  *
2196  * This invalidates ->net pointers held by localio clients
2197  * while they can still safely access nn->counter.
2198  */
nfsd_net_pre_exit(struct net * net)2199 static __net_exit void nfsd_net_pre_exit(struct net *net)
2200 {
2201 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2202 
2203 	nfs_localio_invalidate_clients(&nn->local_clients,
2204 				       &nn->local_clients_lock);
2205 }
2206 #endif
2207 
2208 /**
2209  * nfsd_net_exit - Release the nfsd_net portion of a net namespace
2210  * @net: a network namespace that is about to be destroyed
2211  *
2212  */
nfsd_net_exit(struct net * net)2213 static __net_exit void nfsd_net_exit(struct net *net)
2214 {
2215 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2216 
2217 	nfsd_proc_stat_shutdown(net);
2218 	percpu_counter_destroy_many(nn->counter, NFSD_STATS_COUNTERS_NUM);
2219 	nfsd_idmap_shutdown(net);
2220 	nfsd_export_shutdown(net);
2221 }
2222 
2223 static struct pernet_operations nfsd_net_ops = {
2224 	.init = nfsd_net_init,
2225 #if IS_ENABLED(CONFIG_NFS_LOCALIO)
2226 	.pre_exit = nfsd_net_pre_exit,
2227 #endif
2228 	.exit = nfsd_net_exit,
2229 	.id   = &nfsd_net_id,
2230 	.size = sizeof(struct nfsd_net),
2231 };
2232 
init_nfsd(void)2233 static int __init init_nfsd(void)
2234 {
2235 	int retval;
2236 
2237 	nfsd_debugfs_init();
2238 
2239 	retval = nfsd4_init_slabs();
2240 	if (retval)
2241 		return retval;
2242 	retval = nfsd4_init_pnfs();
2243 	if (retval)
2244 		goto out_free_slabs;
2245 	retval = nfsd_drc_slab_create();
2246 	if (retval)
2247 		goto out_free_pnfs;
2248 	nfsd_lockd_init();	/* lockd->nfsd callbacks */
2249 	retval = register_pernet_subsys(&nfsd_net_ops);
2250 	if (retval < 0)
2251 		goto out_free_lockd;
2252 	retval = register_cld_notifier();
2253 	if (retval)
2254 		goto out_free_subsys;
2255 	retval = nfsd4_create_laundry_wq();
2256 	if (retval)
2257 		goto out_free_cld;
2258 	retval = register_filesystem(&nfsd_fs_type);
2259 	if (retval)
2260 		goto out_free_nfsd4;
2261 	retval = genl_register_family(&nfsd_nl_family);
2262 	if (retval)
2263 		goto out_free_filesystem;
2264 	retval = create_proc_exports_entry();
2265 	if (retval)
2266 		goto out_free_all;
2267 	nfsd_localio_ops_init();
2268 
2269 	return 0;
2270 out_free_all:
2271 	genl_unregister_family(&nfsd_nl_family);
2272 out_free_filesystem:
2273 	unregister_filesystem(&nfsd_fs_type);
2274 out_free_nfsd4:
2275 	nfsd4_destroy_laundry_wq();
2276 out_free_cld:
2277 	unregister_cld_notifier();
2278 out_free_subsys:
2279 	unregister_pernet_subsys(&nfsd_net_ops);
2280 out_free_lockd:
2281 	nfsd_lockd_shutdown();
2282 	nfsd_drc_slab_free();
2283 out_free_pnfs:
2284 	nfsd4_exit_pnfs();
2285 out_free_slabs:
2286 	nfsd4_free_slabs();
2287 	nfsd_debugfs_exit();
2288 	return retval;
2289 }
2290 
exit_nfsd(void)2291 static void __exit exit_nfsd(void)
2292 {
2293 	remove_proc_entry("fs/nfs/exports", NULL);
2294 	remove_proc_entry("fs/nfs", NULL);
2295 	genl_unregister_family(&nfsd_nl_family);
2296 	unregister_filesystem(&nfsd_fs_type);
2297 	nfsd4_destroy_laundry_wq();
2298 	unregister_cld_notifier();
2299 	unregister_pernet_subsys(&nfsd_net_ops);
2300 	nfsd_drc_slab_free();
2301 	nfsd_lockd_shutdown();
2302 	nfsd4_free_slabs();
2303 	nfsd4_exit_pnfs();
2304 	nfsd_debugfs_exit();
2305 }
2306 
2307 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2308 MODULE_DESCRIPTION("In-kernel NFS server");
2309 MODULE_LICENSE("GPL");
2310 module_init(init_nfsd)
2311 module_exit(exit_nfsd)
2312