Loading...
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/gss_krb5_enctypes.h>
18#include <linux/sunrpc/rpc_pipe_fs.h>
19#include <linux/module.h>
20#include <linux/fsnotify.h>
21
22#include "idmap.h"
23#include "nfsd.h"
24#include "cache.h"
25#include "state.h"
26#include "netns.h"
27#include "pnfs.h"
28#include "filecache.h"
29
30/*
31 * We have a single directory with several nodes in it.
32 */
33enum {
34 NFSD_Root = 1,
35 NFSD_List,
36 NFSD_Export_Stats,
37 NFSD_Export_features,
38 NFSD_Fh,
39 NFSD_FO_UnlockIP,
40 NFSD_FO_UnlockFS,
41 NFSD_Threads,
42 NFSD_Pool_Threads,
43 NFSD_Pool_Stats,
44 NFSD_Reply_Cache_Stats,
45 NFSD_Versions,
46 NFSD_Ports,
47 NFSD_MaxBlkSize,
48 NFSD_MaxConnections,
49 NFSD_Filecache,
50 NFSD_SupportedEnctypes,
51 /*
52 * The below MUST come last. Otherwise we leave a hole in nfsd_files[]
53 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
54 */
55#ifdef CONFIG_NFSD_V4
56 NFSD_Leasetime,
57 NFSD_Gracetime,
58 NFSD_RecoveryDir,
59 NFSD_V4EndGrace,
60#endif
61 NFSD_MaxReserved
62};
63
64/*
65 * write() for these nodes.
66 */
67static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
68static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
69static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
70static ssize_t write_threads(struct file *file, char *buf, size_t size);
71static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
72static ssize_t write_versions(struct file *file, char *buf, size_t size);
73static ssize_t write_ports(struct file *file, char *buf, size_t size);
74static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
75static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
76#ifdef CONFIG_NFSD_V4
77static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
78static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
79static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
80static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
81#endif
82
83static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
84 [NFSD_Fh] = write_filehandle,
85 [NFSD_FO_UnlockIP] = write_unlock_ip,
86 [NFSD_FO_UnlockFS] = write_unlock_fs,
87 [NFSD_Threads] = write_threads,
88 [NFSD_Pool_Threads] = write_pool_threads,
89 [NFSD_Versions] = write_versions,
90 [NFSD_Ports] = write_ports,
91 [NFSD_MaxBlkSize] = write_maxblksize,
92 [NFSD_MaxConnections] = write_maxconn,
93#ifdef CONFIG_NFSD_V4
94 [NFSD_Leasetime] = write_leasetime,
95 [NFSD_Gracetime] = write_gracetime,
96 [NFSD_RecoveryDir] = write_recoverydir,
97 [NFSD_V4EndGrace] = write_v4_end_grace,
98#endif
99};
100
101static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
102{
103 ino_t ino = file_inode(file)->i_ino;
104 char *data;
105 ssize_t rv;
106
107 if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
108 return -EINVAL;
109
110 data = simple_transaction_get(file, buf, size);
111 if (IS_ERR(data))
112 return PTR_ERR(data);
113
114 rv = write_op[ino](file, data, size);
115 if (rv >= 0) {
116 simple_transaction_set(file, rv);
117 rv = size;
118 }
119 return rv;
120}
121
122static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
123{
124 if (! file->private_data) {
125 /* An attempt to read a transaction file without writing
126 * causes a 0-byte write so that the file can return
127 * state information
128 */
129 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
130 if (rv < 0)
131 return rv;
132 }
133 return simple_transaction_read(file, buf, size, pos);
134}
135
136static const struct file_operations transaction_ops = {
137 .write = nfsctl_transaction_write,
138 .read = nfsctl_transaction_read,
139 .release = simple_transaction_release,
140 .llseek = default_llseek,
141};
142
143static int exports_net_open(struct net *net, struct file *file)
144{
145 int err;
146 struct seq_file *seq;
147 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
148
149 err = seq_open(file, &nfs_exports_op);
150 if (err)
151 return err;
152
153 seq = file->private_data;
154 seq->private = nn->svc_export_cache;
155 return 0;
156}
157
158static int exports_proc_open(struct inode *inode, struct file *file)
159{
160 return exports_net_open(current->nsproxy->net_ns, file);
161}
162
163static const struct proc_ops exports_proc_ops = {
164 .proc_open = exports_proc_open,
165 .proc_read = seq_read,
166 .proc_lseek = seq_lseek,
167 .proc_release = seq_release,
168};
169
170static int exports_nfsd_open(struct inode *inode, struct file *file)
171{
172 return exports_net_open(inode->i_sb->s_fs_info, file);
173}
174
175static const struct file_operations exports_nfsd_operations = {
176 .open = exports_nfsd_open,
177 .read = seq_read,
178 .llseek = seq_lseek,
179 .release = seq_release,
180};
181
182static int export_features_show(struct seq_file *m, void *v)
183{
184 seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
185 return 0;
186}
187
188DEFINE_SHOW_ATTRIBUTE(export_features);
189
190#if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
191static int supported_enctypes_show(struct seq_file *m, void *v)
192{
193 seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
194 return 0;
195}
196
197DEFINE_SHOW_ATTRIBUTE(supported_enctypes);
198#endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
199
200static const struct file_operations pool_stats_operations = {
201 .open = nfsd_pool_stats_open,
202 .read = seq_read,
203 .llseek = seq_lseek,
204 .release = nfsd_pool_stats_release,
205};
206
207DEFINE_SHOW_ATTRIBUTE(nfsd_reply_cache_stats);
208
209DEFINE_SHOW_ATTRIBUTE(nfsd_file_cache_stats);
210
211/*----------------------------------------------------------------------------*/
212/*
213 * payload - write methods
214 */
215
216static inline struct net *netns(struct file *file)
217{
218 return file_inode(file)->i_sb->s_fs_info;
219}
220
221/*
222 * write_unlock_ip - Release all locks used by a client
223 *
224 * Experimental.
225 *
226 * Input:
227 * buf: '\n'-terminated C string containing a
228 * presentation format IP address
229 * size: length of C string in @buf
230 * Output:
231 * On success: returns zero if all specified locks were released;
232 * returns one if one or more locks were not released
233 * On error: return code is negative errno value
234 */
235static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
236{
237 struct sockaddr_storage address;
238 struct sockaddr *sap = (struct sockaddr *)&address;
239 size_t salen = sizeof(address);
240 char *fo_path;
241 struct net *net = netns(file);
242
243 /* sanity check */
244 if (size == 0)
245 return -EINVAL;
246
247 if (buf[size-1] != '\n')
248 return -EINVAL;
249
250 fo_path = buf;
251 if (qword_get(&buf, fo_path, size) < 0)
252 return -EINVAL;
253
254 if (rpc_pton(net, fo_path, size, sap, salen) == 0)
255 return -EINVAL;
256
257 return nlmsvc_unlock_all_by_ip(sap);
258}
259
260/*
261 * write_unlock_fs - Release all locks on a local file system
262 *
263 * Experimental.
264 *
265 * Input:
266 * buf: '\n'-terminated C string containing the
267 * absolute pathname of a local file system
268 * size: length of C string in @buf
269 * Output:
270 * On success: returns zero if all specified locks were released;
271 * returns one if one or more locks were not released
272 * On error: return code is negative errno value
273 */
274static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
275{
276 struct path path;
277 char *fo_path;
278 int error;
279
280 /* sanity check */
281 if (size == 0)
282 return -EINVAL;
283
284 if (buf[size-1] != '\n')
285 return -EINVAL;
286
287 fo_path = buf;
288 if (qword_get(&buf, fo_path, size) < 0)
289 return -EINVAL;
290
291 error = kern_path(fo_path, 0, &path);
292 if (error)
293 return error;
294
295 /*
296 * XXX: Needs better sanity checking. Otherwise we could end up
297 * releasing locks on the wrong file system.
298 *
299 * For example:
300 * 1. Does the path refer to a directory?
301 * 2. Is that directory a mount point, or
302 * 3. Is that directory the root of an exported file system?
303 */
304 error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
305
306 path_put(&path);
307 return error;
308}
309
310/*
311 * write_filehandle - Get a variable-length NFS file handle by path
312 *
313 * On input, the buffer contains a '\n'-terminated C string comprised of
314 * three alphanumeric words separated by whitespace. The string may
315 * contain escape sequences.
316 *
317 * Input:
318 * buf:
319 * domain: client domain name
320 * path: export pathname
321 * maxsize: numeric maximum size of
322 * @buf
323 * size: length of C string in @buf
324 * Output:
325 * On success: passed-in buffer filled with '\n'-terminated C
326 * string containing a ASCII hex text version
327 * of the NFS file handle;
328 * return code is the size in bytes of the string
329 * On error: return code is negative errno value
330 */
331static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
332{
333 char *dname, *path;
334 int maxsize;
335 char *mesg = buf;
336 int len;
337 struct auth_domain *dom;
338 struct knfsd_fh fh;
339
340 if (size == 0)
341 return -EINVAL;
342
343 if (buf[size-1] != '\n')
344 return -EINVAL;
345 buf[size-1] = 0;
346
347 dname = mesg;
348 len = qword_get(&mesg, dname, size);
349 if (len <= 0)
350 return -EINVAL;
351
352 path = dname+len+1;
353 len = qword_get(&mesg, path, size);
354 if (len <= 0)
355 return -EINVAL;
356
357 len = get_int(&mesg, &maxsize);
358 if (len)
359 return len;
360
361 if (maxsize < NFS_FHSIZE)
362 return -EINVAL;
363 maxsize = min(maxsize, NFS3_FHSIZE);
364
365 if (qword_get(&mesg, mesg, size)>0)
366 return -EINVAL;
367
368 /* we have all the words, they are in buf.. */
369 dom = unix_domain_find(dname);
370 if (!dom)
371 return -ENOMEM;
372
373 len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
374 auth_domain_put(dom);
375 if (len)
376 return len;
377
378 mesg = buf;
379 len = SIMPLE_TRANSACTION_LIMIT;
380 qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
381 mesg[-1] = '\n';
382 return mesg - buf;
383}
384
385/*
386 * write_threads - Start NFSD, or report the current number of running threads
387 *
388 * Input:
389 * buf: ignored
390 * size: zero
391 * Output:
392 * On success: passed-in buffer filled with '\n'-terminated C
393 * string numeric value representing the number of
394 * running NFSD threads;
395 * return code is the size in bytes of the string
396 * On error: return code is zero
397 *
398 * OR
399 *
400 * Input:
401 * buf: C string containing an unsigned
402 * integer value representing the
403 * number of NFSD threads to start
404 * size: non-zero length of C string in @buf
405 * Output:
406 * On success: NFS service is started;
407 * passed-in buffer filled with '\n'-terminated C
408 * string numeric value representing the number of
409 * running NFSD threads;
410 * return code is the size in bytes of the string
411 * On error: return code is zero or a negative errno value
412 */
413static ssize_t write_threads(struct file *file, char *buf, size_t size)
414{
415 char *mesg = buf;
416 int rv;
417 struct net *net = netns(file);
418
419 if (size > 0) {
420 int newthreads;
421 rv = get_int(&mesg, &newthreads);
422 if (rv)
423 return rv;
424 if (newthreads < 0)
425 return -EINVAL;
426 rv = nfsd_svc(newthreads, net, file->f_cred);
427 if (rv < 0)
428 return rv;
429 } else
430 rv = nfsd_nrthreads(net);
431
432 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
433}
434
435/*
436 * write_pool_threads - Set or report the current number of threads per pool
437 *
438 * Input:
439 * buf: ignored
440 * size: zero
441 *
442 * OR
443 *
444 * Input:
445 * buf: C string containing whitespace-
446 * separated unsigned integer values
447 * representing the number of NFSD
448 * threads to start in each pool
449 * size: non-zero length of C string in @buf
450 * Output:
451 * On success: passed-in buffer filled with '\n'-terminated C
452 * string containing integer values representing the
453 * number of NFSD threads in each pool;
454 * return code is the size in bytes of the string
455 * On error: return code is zero or a negative errno value
456 */
457static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
458{
459 /* if size > 0, look for an array of number of threads per node
460 * and apply them then write out number of threads per node as reply
461 */
462 char *mesg = buf;
463 int i;
464 int rv;
465 int len;
466 int npools;
467 int *nthreads;
468 struct net *net = netns(file);
469
470 mutex_lock(&nfsd_mutex);
471 npools = nfsd_nrpools(net);
472 if (npools == 0) {
473 /*
474 * NFS is shut down. The admin can start it by
475 * writing to the threads file but NOT the pool_threads
476 * file, sorry. Report zero threads.
477 */
478 mutex_unlock(&nfsd_mutex);
479 strcpy(buf, "0\n");
480 return strlen(buf);
481 }
482
483 nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
484 rv = -ENOMEM;
485 if (nthreads == NULL)
486 goto out_free;
487
488 if (size > 0) {
489 for (i = 0; i < npools; i++) {
490 rv = get_int(&mesg, &nthreads[i]);
491 if (rv == -ENOENT)
492 break; /* fewer numbers than pools */
493 if (rv)
494 goto out_free; /* syntax error */
495 rv = -EINVAL;
496 if (nthreads[i] < 0)
497 goto out_free;
498 }
499 rv = nfsd_set_nrthreads(i, nthreads, net);
500 if (rv)
501 goto out_free;
502 }
503
504 rv = nfsd_get_nrthreads(npools, nthreads, net);
505 if (rv)
506 goto out_free;
507
508 mesg = buf;
509 size = SIMPLE_TRANSACTION_LIMIT;
510 for (i = 0; i < npools && size > 0; i++) {
511 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
512 len = strlen(mesg);
513 size -= len;
514 mesg += len;
515 }
516 rv = mesg - buf;
517out_free:
518 kfree(nthreads);
519 mutex_unlock(&nfsd_mutex);
520 return rv;
521}
522
523static ssize_t
524nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
525 const char *sep, unsigned vers, int minor)
526{
527 const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
528 bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
529
530 if (vers == 4 && minor >= 0 &&
531 !nfsd_minorversion(nn, minor, NFSD_TEST))
532 supported = false;
533 if (minor == 0 && supported)
534 /*
535 * special case for backward compatability.
536 * +4.0 is never reported, it is implied by
537 * +4, unless -4.0 is present.
538 */
539 return 0;
540 return snprintf(buf, remaining, format, sep,
541 supported ? '+' : '-', vers, minor);
542}
543
544static ssize_t __write_versions(struct file *file, char *buf, size_t size)
545{
546 char *mesg = buf;
547 char *vers, *minorp, sign;
548 int len, num, remaining;
549 ssize_t tlen = 0;
550 char *sep;
551 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
552
553 if (size>0) {
554 if (nn->nfsd_serv)
555 /* Cannot change versions without updating
556 * nn->nfsd_serv->sv_xdrsize, and reallocing
557 * rq_argp and rq_resp
558 */
559 return -EBUSY;
560 if (buf[size-1] != '\n')
561 return -EINVAL;
562 buf[size-1] = 0;
563
564 vers = mesg;
565 len = qword_get(&mesg, vers, size);
566 if (len <= 0) return -EINVAL;
567 do {
568 enum vers_op cmd;
569 unsigned minor;
570 sign = *vers;
571 if (sign == '+' || sign == '-')
572 num = simple_strtol((vers+1), &minorp, 0);
573 else
574 num = simple_strtol(vers, &minorp, 0);
575 if (*minorp == '.') {
576 if (num != 4)
577 return -EINVAL;
578 if (kstrtouint(minorp+1, 0, &minor) < 0)
579 return -EINVAL;
580 }
581
582 cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
583 switch(num) {
584#ifdef CONFIG_NFSD_V2
585 case 2:
586#endif
587 case 3:
588 nfsd_vers(nn, num, cmd);
589 break;
590 case 4:
591 if (*minorp == '.') {
592 if (nfsd_minorversion(nn, minor, cmd) < 0)
593 return -EINVAL;
594 } else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
595 /*
596 * Either we have +4 and no minors are enabled,
597 * or we have -4 and at least one minor is enabled.
598 * In either case, propagate 'cmd' to all minors.
599 */
600 minor = 0;
601 while (nfsd_minorversion(nn, minor, cmd) >= 0)
602 minor++;
603 }
604 break;
605 default:
606 /* Ignore requests to disable non-existent versions */
607 if (cmd == NFSD_SET)
608 return -EINVAL;
609 }
610 vers += len + 1;
611 } while ((len = qword_get(&mesg, vers, size)) > 0);
612 /* If all get turned off, turn them back on, as
613 * having no versions is BAD
614 */
615 nfsd_reset_versions(nn);
616 }
617
618 /* Now write current state into reply buffer */
619 sep = "";
620 remaining = SIMPLE_TRANSACTION_LIMIT;
621 for (num=2 ; num <= 4 ; num++) {
622 int minor;
623 if (!nfsd_vers(nn, num, NFSD_AVAIL))
624 continue;
625
626 minor = -1;
627 do {
628 len = nfsd_print_version_support(nn, buf, remaining,
629 sep, num, minor);
630 if (len >= remaining)
631 goto out;
632 remaining -= len;
633 buf += len;
634 tlen += len;
635 minor++;
636 if (len)
637 sep = " ";
638 } while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
639 }
640out:
641 len = snprintf(buf, remaining, "\n");
642 if (len >= remaining)
643 return -EINVAL;
644 return tlen + len;
645}
646
647/*
648 * write_versions - Set or report the available NFS protocol versions
649 *
650 * Input:
651 * buf: ignored
652 * size: zero
653 * Output:
654 * On success: passed-in buffer filled with '\n'-terminated C
655 * string containing positive or negative integer
656 * values representing the current status of each
657 * protocol version;
658 * return code is the size in bytes of the string
659 * On error: return code is zero or a negative errno value
660 *
661 * OR
662 *
663 * Input:
664 * buf: C string containing whitespace-
665 * separated positive or negative
666 * integer values representing NFS
667 * protocol versions to enable ("+n")
668 * or disable ("-n")
669 * size: non-zero length of C string in @buf
670 * Output:
671 * On success: status of zero or more protocol versions has
672 * been updated; passed-in buffer filled with
673 * '\n'-terminated C string containing positive
674 * or negative integer values representing the
675 * current status of each protocol version;
676 * return code is the size in bytes of the string
677 * On error: return code is zero or a negative errno value
678 */
679static ssize_t write_versions(struct file *file, char *buf, size_t size)
680{
681 ssize_t rv;
682
683 mutex_lock(&nfsd_mutex);
684 rv = __write_versions(file, buf, size);
685 mutex_unlock(&nfsd_mutex);
686 return rv;
687}
688
689/*
690 * Zero-length write. Return a list of NFSD's current listener
691 * transports.
692 */
693static ssize_t __write_ports_names(char *buf, struct net *net)
694{
695 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
696
697 if (nn->nfsd_serv == NULL)
698 return 0;
699 return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
700}
701
702/*
703 * A single 'fd' number was written, in which case it must be for
704 * a socket of a supported family/protocol, and we use it as an
705 * nfsd listener.
706 */
707static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
708{
709 char *mesg = buf;
710 int fd, err;
711 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
712
713 err = get_int(&mesg, &fd);
714 if (err != 0 || fd < 0)
715 return -EINVAL;
716
717 if (svc_alien_sock(net, fd)) {
718 printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
719 return -EINVAL;
720 }
721
722 err = nfsd_create_serv(net);
723 if (err != 0)
724 return err;
725
726 err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
727
728 if (err >= 0 &&
729 !nn->nfsd_serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
730 svc_get(nn->nfsd_serv);
731
732 nfsd_put(net);
733 return err;
734}
735
736/*
737 * A transport listener is added by writing it's transport name and
738 * a port number.
739 */
740static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
741{
742 char transport[16];
743 struct svc_xprt *xprt;
744 int port, err;
745 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
746
747 if (sscanf(buf, "%15s %5u", transport, &port) != 2)
748 return -EINVAL;
749
750 if (port < 1 || port > USHRT_MAX)
751 return -EINVAL;
752
753 err = nfsd_create_serv(net);
754 if (err != 0)
755 return err;
756
757 err = svc_xprt_create(nn->nfsd_serv, transport, net,
758 PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
759 if (err < 0)
760 goto out_err;
761
762 err = svc_xprt_create(nn->nfsd_serv, transport, net,
763 PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
764 if (err < 0 && err != -EAFNOSUPPORT)
765 goto out_close;
766
767 if (!nn->nfsd_serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
768 svc_get(nn->nfsd_serv);
769
770 nfsd_put(net);
771 return 0;
772out_close:
773 xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
774 if (xprt != NULL) {
775 svc_xprt_close(xprt);
776 svc_xprt_put(xprt);
777 }
778out_err:
779 nfsd_put(net);
780 return err;
781}
782
783static ssize_t __write_ports(struct file *file, char *buf, size_t size,
784 struct net *net)
785{
786 if (size == 0)
787 return __write_ports_names(buf, net);
788
789 if (isdigit(buf[0]))
790 return __write_ports_addfd(buf, net, file->f_cred);
791
792 if (isalpha(buf[0]))
793 return __write_ports_addxprt(buf, net, file->f_cred);
794
795 return -EINVAL;
796}
797
798/*
799 * write_ports - Pass a socket file descriptor or transport name to listen on
800 *
801 * Input:
802 * buf: ignored
803 * size: zero
804 * Output:
805 * On success: passed-in buffer filled with a '\n'-terminated C
806 * string containing a whitespace-separated list of
807 * named NFSD listeners;
808 * return code is the size in bytes of the string
809 * On error: return code is zero or a negative errno value
810 *
811 * OR
812 *
813 * Input:
814 * buf: C string containing an unsigned
815 * integer value representing a bound
816 * but unconnected socket that is to be
817 * used as an NFSD listener; listen(3)
818 * must be called for a SOCK_STREAM
819 * socket, otherwise it is ignored
820 * size: non-zero length of C string in @buf
821 * Output:
822 * On success: NFS service is started;
823 * passed-in buffer filled with a '\n'-terminated C
824 * string containing a unique alphanumeric name of
825 * the listener;
826 * return code is the size in bytes of the string
827 * On error: return code is a negative errno value
828 *
829 * OR
830 *
831 * Input:
832 * buf: C string containing a transport
833 * name and an unsigned integer value
834 * representing the port to listen on,
835 * separated by whitespace
836 * size: non-zero length of C string in @buf
837 * Output:
838 * On success: returns zero; NFS service is started
839 * On error: return code is a negative errno value
840 */
841static ssize_t write_ports(struct file *file, char *buf, size_t size)
842{
843 ssize_t rv;
844
845 mutex_lock(&nfsd_mutex);
846 rv = __write_ports(file, buf, size, netns(file));
847 mutex_unlock(&nfsd_mutex);
848 return rv;
849}
850
851
852int nfsd_max_blksize;
853
854/*
855 * write_maxblksize - Set or report the current NFS blksize
856 *
857 * Input:
858 * buf: ignored
859 * size: zero
860 *
861 * OR
862 *
863 * Input:
864 * buf: C string containing an unsigned
865 * integer value representing the new
866 * NFS blksize
867 * size: non-zero length of C string in @buf
868 * Output:
869 * On success: passed-in buffer filled with '\n'-terminated C string
870 * containing numeric value of the current NFS blksize
871 * setting;
872 * return code is the size in bytes of the string
873 * On error: return code is zero or a negative errno value
874 */
875static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
876{
877 char *mesg = buf;
878 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
879
880 if (size > 0) {
881 int bsize;
882 int rv = get_int(&mesg, &bsize);
883 if (rv)
884 return rv;
885 /* force bsize into allowed range and
886 * required alignment.
887 */
888 bsize = max_t(int, bsize, 1024);
889 bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
890 bsize &= ~(1024-1);
891 mutex_lock(&nfsd_mutex);
892 if (nn->nfsd_serv) {
893 mutex_unlock(&nfsd_mutex);
894 return -EBUSY;
895 }
896 nfsd_max_blksize = bsize;
897 mutex_unlock(&nfsd_mutex);
898 }
899
900 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
901 nfsd_max_blksize);
902}
903
904/*
905 * write_maxconn - Set or report the current max number of connections
906 *
907 * Input:
908 * buf: ignored
909 * size: zero
910 * OR
911 *
912 * Input:
913 * buf: C string containing an unsigned
914 * integer value representing the new
915 * number of max connections
916 * size: non-zero length of C string in @buf
917 * Output:
918 * On success: passed-in buffer filled with '\n'-terminated C string
919 * containing numeric value of max_connections setting
920 * for this net namespace;
921 * return code is the size in bytes of the string
922 * On error: return code is zero or a negative errno value
923 */
924static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
925{
926 char *mesg = buf;
927 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
928 unsigned int maxconn = nn->max_connections;
929
930 if (size > 0) {
931 int rv = get_uint(&mesg, &maxconn);
932
933 if (rv)
934 return rv;
935 nn->max_connections = maxconn;
936 }
937
938 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
939}
940
941#ifdef CONFIG_NFSD_V4
942static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
943 time64_t *time, struct nfsd_net *nn)
944{
945 char *mesg = buf;
946 int rv, i;
947
948 if (size > 0) {
949 if (nn->nfsd_serv)
950 return -EBUSY;
951 rv = get_int(&mesg, &i);
952 if (rv)
953 return rv;
954 /*
955 * Some sanity checking. We don't have a reason for
956 * these particular numbers, but problems with the
957 * extremes are:
958 * - Too short: the briefest network outage may
959 * cause clients to lose all their locks. Also,
960 * the frequent polling may be wasteful.
961 * - Too long: do you really want reboot recovery
962 * to take more than an hour? Or to make other
963 * clients wait an hour before being able to
964 * revoke a dead client's locks?
965 */
966 if (i < 10 || i > 3600)
967 return -EINVAL;
968 *time = i;
969 }
970
971 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
972}
973
974static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
975 time64_t *time, struct nfsd_net *nn)
976{
977 ssize_t rv;
978
979 mutex_lock(&nfsd_mutex);
980 rv = __nfsd4_write_time(file, buf, size, time, nn);
981 mutex_unlock(&nfsd_mutex);
982 return rv;
983}
984
985/*
986 * write_leasetime - Set or report the current NFSv4 lease time
987 *
988 * Input:
989 * buf: ignored
990 * size: zero
991 *
992 * OR
993 *
994 * Input:
995 * buf: C string containing an unsigned
996 * integer value representing the new
997 * NFSv4 lease expiry time
998 * size: non-zero length of C string in @buf
999 * Output:
1000 * On success: passed-in buffer filled with '\n'-terminated C
1001 * string containing unsigned integer value of the
1002 * current lease expiry time;
1003 * return code is the size in bytes of the string
1004 * On error: return code is zero or a negative errno value
1005 */
1006static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1007{
1008 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1009 return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1010}
1011
1012/*
1013 * write_gracetime - Set or report current NFSv4 grace period time
1014 *
1015 * As above, but sets the time of the NFSv4 grace period.
1016 *
1017 * Note this should never be set to less than the *previous*
1018 * lease-period time, but we don't try to enforce this. (In the common
1019 * case (a new boot), we don't know what the previous lease time was
1020 * anyway.)
1021 */
1022static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1023{
1024 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1025 return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1026}
1027
1028static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1029 struct nfsd_net *nn)
1030{
1031 char *mesg = buf;
1032 char *recdir;
1033 int len, status;
1034
1035 if (size > 0) {
1036 if (nn->nfsd_serv)
1037 return -EBUSY;
1038 if (size > PATH_MAX || buf[size-1] != '\n')
1039 return -EINVAL;
1040 buf[size-1] = 0;
1041
1042 recdir = mesg;
1043 len = qword_get(&mesg, recdir, size);
1044 if (len <= 0)
1045 return -EINVAL;
1046
1047 status = nfs4_reset_recoverydir(recdir);
1048 if (status)
1049 return status;
1050 }
1051
1052 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1053 nfs4_recoverydir());
1054}
1055
1056/*
1057 * write_recoverydir - Set or report the pathname of the recovery directory
1058 *
1059 * Input:
1060 * buf: ignored
1061 * size: zero
1062 *
1063 * OR
1064 *
1065 * Input:
1066 * buf: C string containing the pathname
1067 * of the directory on a local file
1068 * system containing permanent NFSv4
1069 * recovery data
1070 * size: non-zero length of C string in @buf
1071 * Output:
1072 * On success: passed-in buffer filled with '\n'-terminated C string
1073 * containing the current recovery pathname setting;
1074 * return code is the size in bytes of the string
1075 * On error: return code is zero or a negative errno value
1076 */
1077static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1078{
1079 ssize_t rv;
1080 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1081
1082 mutex_lock(&nfsd_mutex);
1083 rv = __write_recoverydir(file, buf, size, nn);
1084 mutex_unlock(&nfsd_mutex);
1085 return rv;
1086}
1087
1088/*
1089 * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1090 *
1091 * Input:
1092 * buf: ignored
1093 * size: zero
1094 * OR
1095 *
1096 * Input:
1097 * buf: any value
1098 * size: non-zero length of C string in @buf
1099 * Output:
1100 * passed-in buffer filled with "Y" or "N" with a newline
1101 * and NULL-terminated C string. This indicates whether
1102 * the grace period has ended in the current net
1103 * namespace. Return code is the size in bytes of the
1104 * string. Writing a string that starts with 'Y', 'y', or
1105 * '1' to the file will end the grace period for nfsd's v4
1106 * lock manager.
1107 */
1108static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1109{
1110 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1111
1112 if (size > 0) {
1113 switch(buf[0]) {
1114 case 'Y':
1115 case 'y':
1116 case '1':
1117 if (!nn->nfsd_serv)
1118 return -EBUSY;
1119 nfsd4_end_grace(nn);
1120 break;
1121 default:
1122 return -EINVAL;
1123 }
1124 }
1125
1126 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1127 nn->grace_ended ? 'Y' : 'N');
1128}
1129
1130#endif
1131
1132/*----------------------------------------------------------------------------*/
1133/*
1134 * populating the filesystem.
1135 */
1136
1137/* Basically copying rpc_get_inode. */
1138static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1139{
1140 struct inode *inode = new_inode(sb);
1141 if (!inode)
1142 return NULL;
1143 /* Following advice from simple_fill_super documentation: */
1144 inode->i_ino = iunique(sb, NFSD_MaxReserved);
1145 inode->i_mode = mode;
1146 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1147 switch (mode & S_IFMT) {
1148 case S_IFDIR:
1149 inode->i_fop = &simple_dir_operations;
1150 inode->i_op = &simple_dir_inode_operations;
1151 inc_nlink(inode);
1152 break;
1153 default:
1154 break;
1155 }
1156 return inode;
1157}
1158
1159static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1160{
1161 struct inode *inode;
1162
1163 inode = nfsd_get_inode(dir->i_sb, mode);
1164 if (!inode)
1165 return -ENOMEM;
1166 if (ncl) {
1167 inode->i_private = ncl;
1168 kref_get(&ncl->cl_ref);
1169 }
1170 d_add(dentry, inode);
1171 inc_nlink(dir);
1172 fsnotify_mkdir(dir, dentry);
1173 return 0;
1174}
1175
1176static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1177{
1178 struct inode *dir = parent->d_inode;
1179 struct dentry *dentry;
1180 int ret = -ENOMEM;
1181
1182 inode_lock(dir);
1183 dentry = d_alloc_name(parent, name);
1184 if (!dentry)
1185 goto out_err;
1186 ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1187 if (ret)
1188 goto out_err;
1189out:
1190 inode_unlock(dir);
1191 return dentry;
1192out_err:
1193 dput(dentry);
1194 dentry = ERR_PTR(ret);
1195 goto out;
1196}
1197
1198static void clear_ncl(struct inode *inode)
1199{
1200 struct nfsdfs_client *ncl = inode->i_private;
1201
1202 inode->i_private = NULL;
1203 kref_put(&ncl->cl_ref, ncl->cl_release);
1204}
1205
1206static struct nfsdfs_client *__get_nfsdfs_client(struct inode *inode)
1207{
1208 struct nfsdfs_client *nc = inode->i_private;
1209
1210 if (nc)
1211 kref_get(&nc->cl_ref);
1212 return nc;
1213}
1214
1215struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1216{
1217 struct nfsdfs_client *nc;
1218
1219 inode_lock_shared(inode);
1220 nc = __get_nfsdfs_client(inode);
1221 inode_unlock_shared(inode);
1222 return nc;
1223}
1224/* from __rpc_unlink */
1225static void nfsdfs_remove_file(struct inode *dir, struct dentry *dentry)
1226{
1227 int ret;
1228
1229 clear_ncl(d_inode(dentry));
1230 dget(dentry);
1231 ret = simple_unlink(dir, dentry);
1232 d_drop(dentry);
1233 fsnotify_unlink(dir, dentry);
1234 dput(dentry);
1235 WARN_ON_ONCE(ret);
1236}
1237
1238static void nfsdfs_remove_files(struct dentry *root)
1239{
1240 struct dentry *dentry, *tmp;
1241
1242 list_for_each_entry_safe(dentry, tmp, &root->d_subdirs, d_child) {
1243 if (!simple_positive(dentry)) {
1244 WARN_ON_ONCE(1); /* I think this can't happen? */
1245 continue;
1246 }
1247 nfsdfs_remove_file(d_inode(root), dentry);
1248 }
1249}
1250
1251/* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1252 * code instead. */
1253static int nfsdfs_create_files(struct dentry *root,
1254 const struct tree_descr *files,
1255 struct dentry **fdentries)
1256{
1257 struct inode *dir = d_inode(root);
1258 struct inode *inode;
1259 struct dentry *dentry;
1260 int i;
1261
1262 inode_lock(dir);
1263 for (i = 0; files->name && files->name[0]; i++, files++) {
1264 dentry = d_alloc_name(root, files->name);
1265 if (!dentry)
1266 goto out;
1267 inode = nfsd_get_inode(d_inode(root)->i_sb,
1268 S_IFREG | files->mode);
1269 if (!inode) {
1270 dput(dentry);
1271 goto out;
1272 }
1273 inode->i_fop = files->ops;
1274 inode->i_private = __get_nfsdfs_client(dir);
1275 d_add(dentry, inode);
1276 fsnotify_create(dir, dentry);
1277 if (fdentries)
1278 fdentries[i] = dentry;
1279 }
1280 inode_unlock(dir);
1281 return 0;
1282out:
1283 nfsdfs_remove_files(root);
1284 inode_unlock(dir);
1285 return -ENOMEM;
1286}
1287
1288/* on success, returns positive number unique to that client. */
1289struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1290 struct nfsdfs_client *ncl, u32 id,
1291 const struct tree_descr *files,
1292 struct dentry **fdentries)
1293{
1294 struct dentry *dentry;
1295 char name[11];
1296 int ret;
1297
1298 sprintf(name, "%u", id);
1299
1300 dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1301 if (IS_ERR(dentry)) /* XXX: tossing errors? */
1302 return NULL;
1303 ret = nfsdfs_create_files(dentry, files, fdentries);
1304 if (ret) {
1305 nfsd_client_rmdir(dentry);
1306 return NULL;
1307 }
1308 return dentry;
1309}
1310
1311/* Taken from __rpc_rmdir: */
1312void nfsd_client_rmdir(struct dentry *dentry)
1313{
1314 struct inode *dir = d_inode(dentry->d_parent);
1315 struct inode *inode = d_inode(dentry);
1316 int ret;
1317
1318 inode_lock(dir);
1319 nfsdfs_remove_files(dentry);
1320 clear_ncl(inode);
1321 dget(dentry);
1322 ret = simple_rmdir(dir, dentry);
1323 WARN_ON_ONCE(ret);
1324 d_drop(dentry);
1325 fsnotify_rmdir(dir, dentry);
1326 dput(dentry);
1327 inode_unlock(dir);
1328}
1329
1330static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1331{
1332 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1333 nfsd_net_id);
1334 struct dentry *dentry;
1335 int ret;
1336
1337 static const struct tree_descr nfsd_files[] = {
1338 [NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1339 /* Per-export io stats use same ops as exports file */
1340 [NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1341 [NFSD_Export_features] = {"export_features",
1342 &export_features_fops, S_IRUGO},
1343 [NFSD_FO_UnlockIP] = {"unlock_ip",
1344 &transaction_ops, S_IWUSR|S_IRUSR},
1345 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1346 &transaction_ops, S_IWUSR|S_IRUSR},
1347 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1348 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1349 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1350 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1351 [NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1352 &nfsd_reply_cache_stats_fops, S_IRUGO},
1353 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1354 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1355 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1356 [NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1357 [NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
1358#if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1359 [NFSD_SupportedEnctypes] = {"supported_krb5_enctypes",
1360 &supported_enctypes_fops, S_IRUGO},
1361#endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1362#ifdef CONFIG_NFSD_V4
1363 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1364 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1365 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1366 [NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1367#endif
1368 /* last one */ {""}
1369 };
1370
1371 ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1372 if (ret)
1373 return ret;
1374 dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1375 if (IS_ERR(dentry))
1376 return PTR_ERR(dentry);
1377 nn->nfsd_client_dir = dentry;
1378 return 0;
1379}
1380
1381static int nfsd_fs_get_tree(struct fs_context *fc)
1382{
1383 return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1384}
1385
1386static void nfsd_fs_free_fc(struct fs_context *fc)
1387{
1388 if (fc->s_fs_info)
1389 put_net(fc->s_fs_info);
1390}
1391
1392static const struct fs_context_operations nfsd_fs_context_ops = {
1393 .free = nfsd_fs_free_fc,
1394 .get_tree = nfsd_fs_get_tree,
1395};
1396
1397static int nfsd_init_fs_context(struct fs_context *fc)
1398{
1399 put_user_ns(fc->user_ns);
1400 fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1401 fc->ops = &nfsd_fs_context_ops;
1402 return 0;
1403}
1404
1405static void nfsd_umount(struct super_block *sb)
1406{
1407 struct net *net = sb->s_fs_info;
1408
1409 nfsd_shutdown_threads(net);
1410
1411 kill_litter_super(sb);
1412 put_net(net);
1413}
1414
1415static struct file_system_type nfsd_fs_type = {
1416 .owner = THIS_MODULE,
1417 .name = "nfsd",
1418 .init_fs_context = nfsd_init_fs_context,
1419 .kill_sb = nfsd_umount,
1420};
1421MODULE_ALIAS_FS("nfsd");
1422
1423#ifdef CONFIG_PROC_FS
1424static int create_proc_exports_entry(void)
1425{
1426 struct proc_dir_entry *entry;
1427
1428 entry = proc_mkdir("fs/nfs", NULL);
1429 if (!entry)
1430 return -ENOMEM;
1431 entry = proc_create("exports", 0, entry, &exports_proc_ops);
1432 if (!entry) {
1433 remove_proc_entry("fs/nfs", NULL);
1434 return -ENOMEM;
1435 }
1436 return 0;
1437}
1438#else /* CONFIG_PROC_FS */
1439static int create_proc_exports_entry(void)
1440{
1441 return 0;
1442}
1443#endif
1444
1445unsigned int nfsd_net_id;
1446
1447static __net_init int nfsd_init_net(struct net *net)
1448{
1449 int retval;
1450 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1451
1452 retval = nfsd_export_init(net);
1453 if (retval)
1454 goto out_export_error;
1455 retval = nfsd_idmap_init(net);
1456 if (retval)
1457 goto out_idmap_error;
1458 nn->nfsd_versions = NULL;
1459 nn->nfsd4_minorversions = NULL;
1460 nfsd4_init_leases_net(nn);
1461 retval = nfsd_reply_cache_init(nn);
1462 if (retval)
1463 goto out_cache_error;
1464 get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
1465 seqlock_init(&nn->writeverf_lock);
1466
1467 return 0;
1468
1469out_cache_error:
1470 nfsd_idmap_shutdown(net);
1471out_idmap_error:
1472 nfsd_export_shutdown(net);
1473out_export_error:
1474 return retval;
1475}
1476
1477static __net_exit void nfsd_exit_net(struct net *net)
1478{
1479 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1480
1481 nfsd_reply_cache_shutdown(nn);
1482 nfsd_idmap_shutdown(net);
1483 nfsd_export_shutdown(net);
1484 nfsd_netns_free_versions(net_generic(net, nfsd_net_id));
1485}
1486
1487static struct pernet_operations nfsd_net_ops = {
1488 .init = nfsd_init_net,
1489 .exit = nfsd_exit_net,
1490 .id = &nfsd_net_id,
1491 .size = sizeof(struct nfsd_net),
1492};
1493
1494static int __init init_nfsd(void)
1495{
1496 int retval;
1497
1498 retval = nfsd4_init_slabs();
1499 if (retval)
1500 return retval;
1501 retval = nfsd4_init_pnfs();
1502 if (retval)
1503 goto out_free_slabs;
1504 retval = nfsd_stat_init(); /* Statistics */
1505 if (retval)
1506 goto out_free_pnfs;
1507 retval = nfsd_drc_slab_create();
1508 if (retval)
1509 goto out_free_stat;
1510 nfsd_lockd_init(); /* lockd->nfsd callbacks */
1511 retval = create_proc_exports_entry();
1512 if (retval)
1513 goto out_free_lockd;
1514 retval = register_pernet_subsys(&nfsd_net_ops);
1515 if (retval < 0)
1516 goto out_free_exports;
1517 retval = register_cld_notifier();
1518 if (retval)
1519 goto out_free_subsys;
1520 retval = nfsd4_create_laundry_wq();
1521 if (retval)
1522 goto out_free_cld;
1523 retval = register_filesystem(&nfsd_fs_type);
1524 if (retval)
1525 goto out_free_all;
1526 return 0;
1527out_free_all:
1528 nfsd4_destroy_laundry_wq();
1529out_free_cld:
1530 unregister_cld_notifier();
1531out_free_subsys:
1532 unregister_pernet_subsys(&nfsd_net_ops);
1533out_free_exports:
1534 remove_proc_entry("fs/nfs/exports", NULL);
1535 remove_proc_entry("fs/nfs", NULL);
1536out_free_lockd:
1537 nfsd_lockd_shutdown();
1538 nfsd_drc_slab_free();
1539out_free_stat:
1540 nfsd_stat_shutdown();
1541out_free_pnfs:
1542 nfsd4_exit_pnfs();
1543out_free_slabs:
1544 nfsd4_free_slabs();
1545 return retval;
1546}
1547
1548static void __exit exit_nfsd(void)
1549{
1550 unregister_filesystem(&nfsd_fs_type);
1551 nfsd4_destroy_laundry_wq();
1552 unregister_cld_notifier();
1553 unregister_pernet_subsys(&nfsd_net_ops);
1554 nfsd_drc_slab_free();
1555 remove_proc_entry("fs/nfs/exports", NULL);
1556 remove_proc_entry("fs/nfs", NULL);
1557 nfsd_stat_shutdown();
1558 nfsd_lockd_shutdown();
1559 nfsd4_free_slabs();
1560 nfsd4_exit_pnfs();
1561}
1562
1563MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1564MODULE_LICENSE("GPL");
1565module_init(init_nfsd)
1566module_exit(exit_nfsd)
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/module.h>
19#include <linux/fsnotify.h>
20
21#include "idmap.h"
22#include "nfsd.h"
23#include "cache.h"
24#include "state.h"
25#include "netns.h"
26#include "pnfs.h"
27#include "filecache.h"
28#include "trace.h"
29#include "netlink.h"
30
31/*
32 * We have a single directory with several nodes in it.
33 */
34enum {
35 NFSD_Root = 1,
36 NFSD_List,
37 NFSD_Export_Stats,
38 NFSD_Export_features,
39 NFSD_Fh,
40 NFSD_FO_UnlockIP,
41 NFSD_FO_UnlockFS,
42 NFSD_Threads,
43 NFSD_Pool_Threads,
44 NFSD_Pool_Stats,
45 NFSD_Reply_Cache_Stats,
46 NFSD_Versions,
47 NFSD_Ports,
48 NFSD_MaxBlkSize,
49 NFSD_MaxConnections,
50 NFSD_Filecache,
51#ifdef CONFIG_NFSD_V4
52 NFSD_Leasetime,
53 NFSD_Gracetime,
54 NFSD_RecoveryDir,
55 NFSD_V4EndGrace,
56#endif
57 NFSD_MaxReserved
58};
59
60/*
61 * write() for these nodes.
62 */
63static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
64static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
65static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
66static ssize_t write_threads(struct file *file, char *buf, size_t size);
67static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
68static ssize_t write_versions(struct file *file, char *buf, size_t size);
69static ssize_t write_ports(struct file *file, char *buf, size_t size);
70static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
71static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
72#ifdef CONFIG_NFSD_V4
73static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
74static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
75#ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
76static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
77#endif
78static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
79#endif
80
81static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
82 [NFSD_Fh] = write_filehandle,
83 [NFSD_FO_UnlockIP] = write_unlock_ip,
84 [NFSD_FO_UnlockFS] = write_unlock_fs,
85 [NFSD_Threads] = write_threads,
86 [NFSD_Pool_Threads] = write_pool_threads,
87 [NFSD_Versions] = write_versions,
88 [NFSD_Ports] = write_ports,
89 [NFSD_MaxBlkSize] = write_maxblksize,
90 [NFSD_MaxConnections] = write_maxconn,
91#ifdef CONFIG_NFSD_V4
92 [NFSD_Leasetime] = write_leasetime,
93 [NFSD_Gracetime] = write_gracetime,
94#ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
95 [NFSD_RecoveryDir] = write_recoverydir,
96#endif
97 [NFSD_V4EndGrace] = write_v4_end_grace,
98#endif
99};
100
101static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
102{
103 ino_t ino = file_inode(file)->i_ino;
104 char *data;
105 ssize_t rv;
106
107 if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
108 return -EINVAL;
109
110 data = simple_transaction_get(file, buf, size);
111 if (IS_ERR(data))
112 return PTR_ERR(data);
113
114 rv = write_op[ino](file, data, size);
115 if (rv < 0)
116 return rv;
117
118 simple_transaction_set(file, rv);
119 return size;
120}
121
122static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
123{
124 if (! file->private_data) {
125 /* An attempt to read a transaction file without writing
126 * causes a 0-byte write so that the file can return
127 * state information
128 */
129 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
130 if (rv < 0)
131 return rv;
132 }
133 return simple_transaction_read(file, buf, size, pos);
134}
135
136static const struct file_operations transaction_ops = {
137 .write = nfsctl_transaction_write,
138 .read = nfsctl_transaction_read,
139 .release = simple_transaction_release,
140 .llseek = default_llseek,
141};
142
143static int exports_net_open(struct net *net, struct file *file)
144{
145 int err;
146 struct seq_file *seq;
147 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
148
149 err = seq_open(file, &nfs_exports_op);
150 if (err)
151 return err;
152
153 seq = file->private_data;
154 seq->private = nn->svc_export_cache;
155 return 0;
156}
157
158static int exports_nfsd_open(struct inode *inode, struct file *file)
159{
160 return exports_net_open(inode->i_sb->s_fs_info, file);
161}
162
163static const struct file_operations exports_nfsd_operations = {
164 .open = exports_nfsd_open,
165 .read = seq_read,
166 .llseek = seq_lseek,
167 .release = seq_release,
168};
169
170static int export_features_show(struct seq_file *m, void *v)
171{
172 seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
173 return 0;
174}
175
176DEFINE_SHOW_ATTRIBUTE(export_features);
177
178static const struct file_operations pool_stats_operations = {
179 .open = nfsd_pool_stats_open,
180 .read = seq_read,
181 .llseek = seq_lseek,
182 .release = seq_release,
183};
184
185DEFINE_SHOW_ATTRIBUTE(nfsd_reply_cache_stats);
186
187DEFINE_SHOW_ATTRIBUTE(nfsd_file_cache_stats);
188
189/*----------------------------------------------------------------------------*/
190/*
191 * payload - write methods
192 */
193
194static inline struct net *netns(struct file *file)
195{
196 return file_inode(file)->i_sb->s_fs_info;
197}
198
199/*
200 * write_unlock_ip - Release all locks used by a client
201 *
202 * Experimental.
203 *
204 * Input:
205 * buf: '\n'-terminated C string containing a
206 * presentation format IP address
207 * size: length of C string in @buf
208 * Output:
209 * On success: returns zero if all specified locks were released;
210 * returns one if one or more locks were not released
211 * On error: return code is negative errno value
212 */
213static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
214{
215 struct sockaddr_storage address;
216 struct sockaddr *sap = (struct sockaddr *)&address;
217 size_t salen = sizeof(address);
218 char *fo_path;
219 struct net *net = netns(file);
220
221 /* sanity check */
222 if (size == 0)
223 return -EINVAL;
224
225 if (buf[size-1] != '\n')
226 return -EINVAL;
227
228 fo_path = buf;
229 if (qword_get(&buf, fo_path, size) < 0)
230 return -EINVAL;
231
232 if (rpc_pton(net, fo_path, size, sap, salen) == 0)
233 return -EINVAL;
234
235 trace_nfsd_ctl_unlock_ip(net, buf);
236 return nlmsvc_unlock_all_by_ip(sap);
237}
238
239/*
240 * write_unlock_fs - Release all locks on a local file system
241 *
242 * Experimental.
243 *
244 * Input:
245 * buf: '\n'-terminated C string containing the
246 * absolute pathname of a local file system
247 * size: length of C string in @buf
248 * Output:
249 * On success: returns zero if all specified locks were released;
250 * returns one if one or more locks were not released
251 * On error: return code is negative errno value
252 */
253static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
254{
255 struct path path;
256 char *fo_path;
257 int error;
258
259 /* sanity check */
260 if (size == 0)
261 return -EINVAL;
262
263 if (buf[size-1] != '\n')
264 return -EINVAL;
265
266 fo_path = buf;
267 if (qword_get(&buf, fo_path, size) < 0)
268 return -EINVAL;
269 trace_nfsd_ctl_unlock_fs(netns(file), fo_path);
270 error = kern_path(fo_path, 0, &path);
271 if (error)
272 return error;
273
274 /*
275 * XXX: Needs better sanity checking. Otherwise we could end up
276 * releasing locks on the wrong file system.
277 *
278 * For example:
279 * 1. Does the path refer to a directory?
280 * 2. Is that directory a mount point, or
281 * 3. Is that directory the root of an exported file system?
282 */
283 error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
284
285 path_put(&path);
286 return error;
287}
288
289/*
290 * write_filehandle - Get a variable-length NFS file handle by path
291 *
292 * On input, the buffer contains a '\n'-terminated C string comprised of
293 * three alphanumeric words separated by whitespace. The string may
294 * contain escape sequences.
295 *
296 * Input:
297 * buf:
298 * domain: client domain name
299 * path: export pathname
300 * maxsize: numeric maximum size of
301 * @buf
302 * size: length of C string in @buf
303 * Output:
304 * On success: passed-in buffer filled with '\n'-terminated C
305 * string containing a ASCII hex text version
306 * of the NFS file handle;
307 * return code is the size in bytes of the string
308 * On error: return code is negative errno value
309 */
310static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
311{
312 char *dname, *path;
313 int maxsize;
314 char *mesg = buf;
315 int len;
316 struct auth_domain *dom;
317 struct knfsd_fh fh;
318
319 if (size == 0)
320 return -EINVAL;
321
322 if (buf[size-1] != '\n')
323 return -EINVAL;
324 buf[size-1] = 0;
325
326 dname = mesg;
327 len = qword_get(&mesg, dname, size);
328 if (len <= 0)
329 return -EINVAL;
330
331 path = dname+len+1;
332 len = qword_get(&mesg, path, size);
333 if (len <= 0)
334 return -EINVAL;
335
336 len = get_int(&mesg, &maxsize);
337 if (len)
338 return len;
339
340 if (maxsize < NFS_FHSIZE)
341 return -EINVAL;
342 maxsize = min(maxsize, NFS3_FHSIZE);
343
344 if (qword_get(&mesg, mesg, size) > 0)
345 return -EINVAL;
346
347 trace_nfsd_ctl_filehandle(netns(file), dname, path, maxsize);
348
349 /* we have all the words, they are in buf.. */
350 dom = unix_domain_find(dname);
351 if (!dom)
352 return -ENOMEM;
353
354 len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
355 auth_domain_put(dom);
356 if (len)
357 return len;
358
359 mesg = buf;
360 len = SIMPLE_TRANSACTION_LIMIT;
361 qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
362 mesg[-1] = '\n';
363 return mesg - buf;
364}
365
366/*
367 * write_threads - Start NFSD, or report the current number of running threads
368 *
369 * Input:
370 * buf: ignored
371 * size: zero
372 * Output:
373 * On success: passed-in buffer filled with '\n'-terminated C
374 * string numeric value representing the number of
375 * running NFSD threads;
376 * return code is the size in bytes of the string
377 * On error: return code is zero
378 *
379 * OR
380 *
381 * Input:
382 * buf: C string containing an unsigned
383 * integer value representing the
384 * number of NFSD threads to start
385 * size: non-zero length of C string in @buf
386 * Output:
387 * On success: NFS service is started;
388 * passed-in buffer filled with '\n'-terminated C
389 * string numeric value representing the number of
390 * running NFSD threads;
391 * return code is the size in bytes of the string
392 * On error: return code is zero or a negative errno value
393 */
394static ssize_t write_threads(struct file *file, char *buf, size_t size)
395{
396 char *mesg = buf;
397 int rv;
398 struct net *net = netns(file);
399
400 if (size > 0) {
401 int newthreads;
402 rv = get_int(&mesg, &newthreads);
403 if (rv)
404 return rv;
405 if (newthreads < 0)
406 return -EINVAL;
407 trace_nfsd_ctl_threads(net, newthreads);
408 rv = nfsd_svc(newthreads, net, file->f_cred);
409 if (rv < 0)
410 return rv;
411 } else
412 rv = nfsd_nrthreads(net);
413
414 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
415}
416
417/*
418 * write_pool_threads - Set or report the current number of threads per pool
419 *
420 * Input:
421 * buf: ignored
422 * size: zero
423 *
424 * OR
425 *
426 * Input:
427 * buf: C string containing whitespace-
428 * separated unsigned integer values
429 * representing the number of NFSD
430 * threads to start in each pool
431 * size: non-zero length of C string in @buf
432 * Output:
433 * On success: passed-in buffer filled with '\n'-terminated C
434 * string containing integer values representing the
435 * number of NFSD threads in each pool;
436 * return code is the size in bytes of the string
437 * On error: return code is zero or a negative errno value
438 */
439static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
440{
441 /* if size > 0, look for an array of number of threads per node
442 * and apply them then write out number of threads per node as reply
443 */
444 char *mesg = buf;
445 int i;
446 int rv;
447 int len;
448 int npools;
449 int *nthreads;
450 struct net *net = netns(file);
451
452 mutex_lock(&nfsd_mutex);
453 npools = nfsd_nrpools(net);
454 if (npools == 0) {
455 /*
456 * NFS is shut down. The admin can start it by
457 * writing to the threads file but NOT the pool_threads
458 * file, sorry. Report zero threads.
459 */
460 mutex_unlock(&nfsd_mutex);
461 strcpy(buf, "0\n");
462 return strlen(buf);
463 }
464
465 nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
466 rv = -ENOMEM;
467 if (nthreads == NULL)
468 goto out_free;
469
470 if (size > 0) {
471 for (i = 0; i < npools; i++) {
472 rv = get_int(&mesg, &nthreads[i]);
473 if (rv == -ENOENT)
474 break; /* fewer numbers than pools */
475 if (rv)
476 goto out_free; /* syntax error */
477 rv = -EINVAL;
478 if (nthreads[i] < 0)
479 goto out_free;
480 trace_nfsd_ctl_pool_threads(net, i, nthreads[i]);
481 }
482 rv = nfsd_set_nrthreads(i, nthreads, net);
483 if (rv)
484 goto out_free;
485 }
486
487 rv = nfsd_get_nrthreads(npools, nthreads, net);
488 if (rv)
489 goto out_free;
490
491 mesg = buf;
492 size = SIMPLE_TRANSACTION_LIMIT;
493 for (i = 0; i < npools && size > 0; i++) {
494 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
495 len = strlen(mesg);
496 size -= len;
497 mesg += len;
498 }
499 rv = mesg - buf;
500out_free:
501 kfree(nthreads);
502 mutex_unlock(&nfsd_mutex);
503 return rv;
504}
505
506static ssize_t
507nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
508 const char *sep, unsigned vers, int minor)
509{
510 const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
511 bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
512
513 if (vers == 4 && minor >= 0 &&
514 !nfsd_minorversion(nn, minor, NFSD_TEST))
515 supported = false;
516 if (minor == 0 && supported)
517 /*
518 * special case for backward compatability.
519 * +4.0 is never reported, it is implied by
520 * +4, unless -4.0 is present.
521 */
522 return 0;
523 return snprintf(buf, remaining, format, sep,
524 supported ? '+' : '-', vers, minor);
525}
526
527static ssize_t __write_versions(struct file *file, char *buf, size_t size)
528{
529 char *mesg = buf;
530 char *vers, *minorp, sign;
531 int len, num, remaining;
532 ssize_t tlen = 0;
533 char *sep;
534 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
535
536 if (size > 0) {
537 if (nn->nfsd_serv)
538 /* Cannot change versions without updating
539 * nn->nfsd_serv->sv_xdrsize, and reallocing
540 * rq_argp and rq_resp
541 */
542 return -EBUSY;
543 if (buf[size-1] != '\n')
544 return -EINVAL;
545 buf[size-1] = 0;
546 trace_nfsd_ctl_version(netns(file), buf);
547
548 vers = mesg;
549 len = qword_get(&mesg, vers, size);
550 if (len <= 0) return -EINVAL;
551 do {
552 enum vers_op cmd;
553 unsigned minor;
554 sign = *vers;
555 if (sign == '+' || sign == '-')
556 num = simple_strtol((vers+1), &minorp, 0);
557 else
558 num = simple_strtol(vers, &minorp, 0);
559 if (*minorp == '.') {
560 if (num != 4)
561 return -EINVAL;
562 if (kstrtouint(minorp+1, 0, &minor) < 0)
563 return -EINVAL;
564 }
565
566 cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
567 switch(num) {
568#ifdef CONFIG_NFSD_V2
569 case 2:
570#endif
571 case 3:
572 nfsd_vers(nn, num, cmd);
573 break;
574 case 4:
575 if (*minorp == '.') {
576 if (nfsd_minorversion(nn, minor, cmd) < 0)
577 return -EINVAL;
578 } else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
579 /*
580 * Either we have +4 and no minors are enabled,
581 * or we have -4 and at least one minor is enabled.
582 * In either case, propagate 'cmd' to all minors.
583 */
584 minor = 0;
585 while (nfsd_minorversion(nn, minor, cmd) >= 0)
586 minor++;
587 }
588 break;
589 default:
590 /* Ignore requests to disable non-existent versions */
591 if (cmd == NFSD_SET)
592 return -EINVAL;
593 }
594 vers += len + 1;
595 } while ((len = qword_get(&mesg, vers, size)) > 0);
596 /* If all get turned off, turn them back on, as
597 * having no versions is BAD
598 */
599 nfsd_reset_versions(nn);
600 }
601
602 /* Now write current state into reply buffer */
603 sep = "";
604 remaining = SIMPLE_TRANSACTION_LIMIT;
605 for (num=2 ; num <= 4 ; num++) {
606 int minor;
607 if (!nfsd_vers(nn, num, NFSD_AVAIL))
608 continue;
609
610 minor = -1;
611 do {
612 len = nfsd_print_version_support(nn, buf, remaining,
613 sep, num, minor);
614 if (len >= remaining)
615 goto out;
616 remaining -= len;
617 buf += len;
618 tlen += len;
619 minor++;
620 if (len)
621 sep = " ";
622 } while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
623 }
624out:
625 len = snprintf(buf, remaining, "\n");
626 if (len >= remaining)
627 return -EINVAL;
628 return tlen + len;
629}
630
631/*
632 * write_versions - Set or report the available NFS protocol versions
633 *
634 * Input:
635 * buf: ignored
636 * size: zero
637 * Output:
638 * On success: passed-in buffer filled with '\n'-terminated C
639 * string containing positive or negative integer
640 * values representing the current status of each
641 * protocol version;
642 * return code is the size in bytes of the string
643 * On error: return code is zero or a negative errno value
644 *
645 * OR
646 *
647 * Input:
648 * buf: C string containing whitespace-
649 * separated positive or negative
650 * integer values representing NFS
651 * protocol versions to enable ("+n")
652 * or disable ("-n")
653 * size: non-zero length of C string in @buf
654 * Output:
655 * On success: status of zero or more protocol versions has
656 * been updated; passed-in buffer filled with
657 * '\n'-terminated C string containing positive
658 * or negative integer values representing the
659 * current status of each protocol version;
660 * return code is the size in bytes of the string
661 * On error: return code is zero or a negative errno value
662 */
663static ssize_t write_versions(struct file *file, char *buf, size_t size)
664{
665 ssize_t rv;
666
667 mutex_lock(&nfsd_mutex);
668 rv = __write_versions(file, buf, size);
669 mutex_unlock(&nfsd_mutex);
670 return rv;
671}
672
673/*
674 * Zero-length write. Return a list of NFSD's current listener
675 * transports.
676 */
677static ssize_t __write_ports_names(char *buf, struct net *net)
678{
679 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
680
681 if (nn->nfsd_serv == NULL)
682 return 0;
683 return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
684}
685
686/*
687 * A single 'fd' number was written, in which case it must be for
688 * a socket of a supported family/protocol, and we use it as an
689 * nfsd listener.
690 */
691static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
692{
693 char *mesg = buf;
694 int fd, err;
695 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
696 struct svc_serv *serv;
697
698 err = get_int(&mesg, &fd);
699 if (err != 0 || fd < 0)
700 return -EINVAL;
701 trace_nfsd_ctl_ports_addfd(net, fd);
702
703 err = nfsd_create_serv(net);
704 if (err != 0)
705 return err;
706
707 serv = nn->nfsd_serv;
708 err = svc_addsock(serv, net, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
709
710 if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
711 nfsd_destroy_serv(net);
712
713 return err;
714}
715
716/*
717 * A transport listener is added by writing its transport name and
718 * a port number.
719 */
720static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
721{
722 char transport[16];
723 struct svc_xprt *xprt;
724 int port, err;
725 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
726 struct svc_serv *serv;
727
728 if (sscanf(buf, "%15s %5u", transport, &port) != 2)
729 return -EINVAL;
730
731 if (port < 1 || port > USHRT_MAX)
732 return -EINVAL;
733 trace_nfsd_ctl_ports_addxprt(net, transport, port);
734
735 err = nfsd_create_serv(net);
736 if (err != 0)
737 return err;
738
739 serv = nn->nfsd_serv;
740 err = svc_xprt_create(serv, transport, net,
741 PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
742 if (err < 0)
743 goto out_err;
744
745 err = svc_xprt_create(serv, transport, net,
746 PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
747 if (err < 0 && err != -EAFNOSUPPORT)
748 goto out_close;
749
750 return 0;
751out_close:
752 xprt = svc_find_xprt(serv, transport, net, PF_INET, port);
753 if (xprt != NULL) {
754 svc_xprt_close(xprt);
755 svc_xprt_put(xprt);
756 }
757out_err:
758 if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
759 nfsd_destroy_serv(net);
760
761 return err;
762}
763
764static ssize_t __write_ports(struct file *file, char *buf, size_t size,
765 struct net *net)
766{
767 if (size == 0)
768 return __write_ports_names(buf, net);
769
770 if (isdigit(buf[0]))
771 return __write_ports_addfd(buf, net, file->f_cred);
772
773 if (isalpha(buf[0]))
774 return __write_ports_addxprt(buf, net, file->f_cred);
775
776 return -EINVAL;
777}
778
779/*
780 * write_ports - Pass a socket file descriptor or transport name to listen on
781 *
782 * Input:
783 * buf: ignored
784 * size: zero
785 * Output:
786 * On success: passed-in buffer filled with a '\n'-terminated C
787 * string containing a whitespace-separated list of
788 * named NFSD listeners;
789 * return code is the size in bytes of the string
790 * On error: return code is zero or a negative errno value
791 *
792 * OR
793 *
794 * Input:
795 * buf: C string containing an unsigned
796 * integer value representing a bound
797 * but unconnected socket that is to be
798 * used as an NFSD listener; listen(3)
799 * must be called for a SOCK_STREAM
800 * socket, otherwise it is ignored
801 * size: non-zero length of C string in @buf
802 * Output:
803 * On success: NFS service is started;
804 * passed-in buffer filled with a '\n'-terminated C
805 * string containing a unique alphanumeric name of
806 * the listener;
807 * return code is the size in bytes of the string
808 * On error: return code is a negative errno value
809 *
810 * OR
811 *
812 * Input:
813 * buf: C string containing a transport
814 * name and an unsigned integer value
815 * representing the port to listen on,
816 * separated by whitespace
817 * size: non-zero length of C string in @buf
818 * Output:
819 * On success: returns zero; NFS service is started
820 * On error: return code is a negative errno value
821 */
822static ssize_t write_ports(struct file *file, char *buf, size_t size)
823{
824 ssize_t rv;
825
826 mutex_lock(&nfsd_mutex);
827 rv = __write_ports(file, buf, size, netns(file));
828 mutex_unlock(&nfsd_mutex);
829 return rv;
830}
831
832
833int nfsd_max_blksize;
834
835/*
836 * write_maxblksize - Set or report the current NFS blksize
837 *
838 * Input:
839 * buf: ignored
840 * size: zero
841 *
842 * OR
843 *
844 * Input:
845 * buf: C string containing an unsigned
846 * integer value representing the new
847 * NFS blksize
848 * size: non-zero length of C string in @buf
849 * Output:
850 * On success: passed-in buffer filled with '\n'-terminated C string
851 * containing numeric value of the current NFS blksize
852 * setting;
853 * return code is the size in bytes of the string
854 * On error: return code is zero or a negative errno value
855 */
856static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
857{
858 char *mesg = buf;
859 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
860
861 if (size > 0) {
862 int bsize;
863 int rv = get_int(&mesg, &bsize);
864 if (rv)
865 return rv;
866 trace_nfsd_ctl_maxblksize(netns(file), bsize);
867
868 /* force bsize into allowed range and
869 * required alignment.
870 */
871 bsize = max_t(int, bsize, 1024);
872 bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
873 bsize &= ~(1024-1);
874 mutex_lock(&nfsd_mutex);
875 if (nn->nfsd_serv) {
876 mutex_unlock(&nfsd_mutex);
877 return -EBUSY;
878 }
879 nfsd_max_blksize = bsize;
880 mutex_unlock(&nfsd_mutex);
881 }
882
883 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
884 nfsd_max_blksize);
885}
886
887/*
888 * write_maxconn - Set or report the current max number of connections
889 *
890 * Input:
891 * buf: ignored
892 * size: zero
893 * OR
894 *
895 * Input:
896 * buf: C string containing an unsigned
897 * integer value representing the new
898 * number of max connections
899 * size: non-zero length of C string in @buf
900 * Output:
901 * On success: passed-in buffer filled with '\n'-terminated C string
902 * containing numeric value of max_connections setting
903 * for this net namespace;
904 * return code is the size in bytes of the string
905 * On error: return code is zero or a negative errno value
906 */
907static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
908{
909 char *mesg = buf;
910 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
911 unsigned int maxconn = nn->max_connections;
912
913 if (size > 0) {
914 int rv = get_uint(&mesg, &maxconn);
915
916 if (rv)
917 return rv;
918 trace_nfsd_ctl_maxconn(netns(file), maxconn);
919 nn->max_connections = maxconn;
920 }
921
922 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
923}
924
925#ifdef CONFIG_NFSD_V4
926static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
927 time64_t *time, struct nfsd_net *nn)
928{
929 struct dentry *dentry = file_dentry(file);
930 char *mesg = buf;
931 int rv, i;
932
933 if (size > 0) {
934 if (nn->nfsd_serv)
935 return -EBUSY;
936 rv = get_int(&mesg, &i);
937 if (rv)
938 return rv;
939 trace_nfsd_ctl_time(netns(file), dentry->d_name.name,
940 dentry->d_name.len, i);
941
942 /*
943 * Some sanity checking. We don't have a reason for
944 * these particular numbers, but problems with the
945 * extremes are:
946 * - Too short: the briefest network outage may
947 * cause clients to lose all their locks. Also,
948 * the frequent polling may be wasteful.
949 * - Too long: do you really want reboot recovery
950 * to take more than an hour? Or to make other
951 * clients wait an hour before being able to
952 * revoke a dead client's locks?
953 */
954 if (i < 10 || i > 3600)
955 return -EINVAL;
956 *time = i;
957 }
958
959 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
960}
961
962static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
963 time64_t *time, struct nfsd_net *nn)
964{
965 ssize_t rv;
966
967 mutex_lock(&nfsd_mutex);
968 rv = __nfsd4_write_time(file, buf, size, time, nn);
969 mutex_unlock(&nfsd_mutex);
970 return rv;
971}
972
973/*
974 * write_leasetime - Set or report the current NFSv4 lease time
975 *
976 * Input:
977 * buf: ignored
978 * size: zero
979 *
980 * OR
981 *
982 * Input:
983 * buf: C string containing an unsigned
984 * integer value representing the new
985 * NFSv4 lease expiry time
986 * size: non-zero length of C string in @buf
987 * Output:
988 * On success: passed-in buffer filled with '\n'-terminated C
989 * string containing unsigned integer value of the
990 * current lease expiry time;
991 * return code is the size in bytes of the string
992 * On error: return code is zero or a negative errno value
993 */
994static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
995{
996 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
997 return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
998}
999
1000/*
1001 * write_gracetime - Set or report current NFSv4 grace period time
1002 *
1003 * As above, but sets the time of the NFSv4 grace period.
1004 *
1005 * Note this should never be set to less than the *previous*
1006 * lease-period time, but we don't try to enforce this. (In the common
1007 * case (a new boot), we don't know what the previous lease time was
1008 * anyway.)
1009 */
1010static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1011{
1012 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1013 return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1014}
1015
1016#ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1017static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1018 struct nfsd_net *nn)
1019{
1020 char *mesg = buf;
1021 char *recdir;
1022 int len, status;
1023
1024 if (size > 0) {
1025 if (nn->nfsd_serv)
1026 return -EBUSY;
1027 if (size > PATH_MAX || buf[size-1] != '\n')
1028 return -EINVAL;
1029 buf[size-1] = 0;
1030
1031 recdir = mesg;
1032 len = qword_get(&mesg, recdir, size);
1033 if (len <= 0)
1034 return -EINVAL;
1035 trace_nfsd_ctl_recoverydir(netns(file), recdir);
1036
1037 status = nfs4_reset_recoverydir(recdir);
1038 if (status)
1039 return status;
1040 }
1041
1042 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1043 nfs4_recoverydir());
1044}
1045
1046/*
1047 * write_recoverydir - Set or report the pathname of the recovery directory
1048 *
1049 * Input:
1050 * buf: ignored
1051 * size: zero
1052 *
1053 * OR
1054 *
1055 * Input:
1056 * buf: C string containing the pathname
1057 * of the directory on a local file
1058 * system containing permanent NFSv4
1059 * recovery data
1060 * size: non-zero length of C string in @buf
1061 * Output:
1062 * On success: passed-in buffer filled with '\n'-terminated C string
1063 * containing the current recovery pathname setting;
1064 * return code is the size in bytes of the string
1065 * On error: return code is zero or a negative errno value
1066 */
1067static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1068{
1069 ssize_t rv;
1070 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1071
1072 mutex_lock(&nfsd_mutex);
1073 rv = __write_recoverydir(file, buf, size, nn);
1074 mutex_unlock(&nfsd_mutex);
1075 return rv;
1076}
1077#endif
1078
1079/*
1080 * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1081 *
1082 * Input:
1083 * buf: ignored
1084 * size: zero
1085 * OR
1086 *
1087 * Input:
1088 * buf: any value
1089 * size: non-zero length of C string in @buf
1090 * Output:
1091 * passed-in buffer filled with "Y" or "N" with a newline
1092 * and NULL-terminated C string. This indicates whether
1093 * the grace period has ended in the current net
1094 * namespace. Return code is the size in bytes of the
1095 * string. Writing a string that starts with 'Y', 'y', or
1096 * '1' to the file will end the grace period for nfsd's v4
1097 * lock manager.
1098 */
1099static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1100{
1101 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1102
1103 if (size > 0) {
1104 switch(buf[0]) {
1105 case 'Y':
1106 case 'y':
1107 case '1':
1108 if (!nn->nfsd_serv)
1109 return -EBUSY;
1110 trace_nfsd_end_grace(netns(file));
1111 nfsd4_end_grace(nn);
1112 break;
1113 default:
1114 return -EINVAL;
1115 }
1116 }
1117
1118 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1119 nn->grace_ended ? 'Y' : 'N');
1120}
1121
1122#endif
1123
1124/*----------------------------------------------------------------------------*/
1125/*
1126 * populating the filesystem.
1127 */
1128
1129/* Basically copying rpc_get_inode. */
1130static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1131{
1132 struct inode *inode = new_inode(sb);
1133 if (!inode)
1134 return NULL;
1135 /* Following advice from simple_fill_super documentation: */
1136 inode->i_ino = iunique(sb, NFSD_MaxReserved);
1137 inode->i_mode = mode;
1138 simple_inode_init_ts(inode);
1139 switch (mode & S_IFMT) {
1140 case S_IFDIR:
1141 inode->i_fop = &simple_dir_operations;
1142 inode->i_op = &simple_dir_inode_operations;
1143 inc_nlink(inode);
1144 break;
1145 case S_IFLNK:
1146 inode->i_op = &simple_symlink_inode_operations;
1147 break;
1148 default:
1149 break;
1150 }
1151 return inode;
1152}
1153
1154static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1155{
1156 struct inode *inode;
1157
1158 inode = nfsd_get_inode(dir->i_sb, mode);
1159 if (!inode)
1160 return -ENOMEM;
1161 if (ncl) {
1162 inode->i_private = ncl;
1163 kref_get(&ncl->cl_ref);
1164 }
1165 d_add(dentry, inode);
1166 inc_nlink(dir);
1167 fsnotify_mkdir(dir, dentry);
1168 return 0;
1169}
1170
1171static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1172{
1173 struct inode *dir = parent->d_inode;
1174 struct dentry *dentry;
1175 int ret = -ENOMEM;
1176
1177 inode_lock(dir);
1178 dentry = d_alloc_name(parent, name);
1179 if (!dentry)
1180 goto out_err;
1181 ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1182 if (ret)
1183 goto out_err;
1184out:
1185 inode_unlock(dir);
1186 return dentry;
1187out_err:
1188 dput(dentry);
1189 dentry = ERR_PTR(ret);
1190 goto out;
1191}
1192
1193#if IS_ENABLED(CONFIG_SUNRPC_GSS)
1194static int __nfsd_symlink(struct inode *dir, struct dentry *dentry,
1195 umode_t mode, const char *content)
1196{
1197 struct inode *inode;
1198
1199 inode = nfsd_get_inode(dir->i_sb, mode);
1200 if (!inode)
1201 return -ENOMEM;
1202
1203 inode->i_link = (char *)content;
1204 inode->i_size = strlen(content);
1205
1206 d_add(dentry, inode);
1207 inc_nlink(dir);
1208 fsnotify_create(dir, dentry);
1209 return 0;
1210}
1211
1212/*
1213 * @content is assumed to be a NUL-terminated string that lives
1214 * longer than the symlink itself.
1215 */
1216static void _nfsd_symlink(struct dentry *parent, const char *name,
1217 const char *content)
1218{
1219 struct inode *dir = parent->d_inode;
1220 struct dentry *dentry;
1221 int ret;
1222
1223 inode_lock(dir);
1224 dentry = d_alloc_name(parent, name);
1225 if (!dentry)
1226 goto out;
1227 ret = __nfsd_symlink(d_inode(parent), dentry, S_IFLNK | 0777, content);
1228 if (ret)
1229 dput(dentry);
1230out:
1231 inode_unlock(dir);
1232}
1233#else
1234static inline void _nfsd_symlink(struct dentry *parent, const char *name,
1235 const char *content)
1236{
1237}
1238
1239#endif
1240
1241static void clear_ncl(struct dentry *dentry)
1242{
1243 struct inode *inode = d_inode(dentry);
1244 struct nfsdfs_client *ncl = inode->i_private;
1245
1246 spin_lock(&inode->i_lock);
1247 inode->i_private = NULL;
1248 spin_unlock(&inode->i_lock);
1249 kref_put(&ncl->cl_ref, ncl->cl_release);
1250}
1251
1252struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1253{
1254 struct nfsdfs_client *nc;
1255
1256 spin_lock(&inode->i_lock);
1257 nc = inode->i_private;
1258 if (nc)
1259 kref_get(&nc->cl_ref);
1260 spin_unlock(&inode->i_lock);
1261 return nc;
1262}
1263
1264/* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1265 * code instead. */
1266static int nfsdfs_create_files(struct dentry *root,
1267 const struct tree_descr *files,
1268 struct nfsdfs_client *ncl,
1269 struct dentry **fdentries)
1270{
1271 struct inode *dir = d_inode(root);
1272 struct inode *inode;
1273 struct dentry *dentry;
1274 int i;
1275
1276 inode_lock(dir);
1277 for (i = 0; files->name && files->name[0]; i++, files++) {
1278 dentry = d_alloc_name(root, files->name);
1279 if (!dentry)
1280 goto out;
1281 inode = nfsd_get_inode(d_inode(root)->i_sb,
1282 S_IFREG | files->mode);
1283 if (!inode) {
1284 dput(dentry);
1285 goto out;
1286 }
1287 kref_get(&ncl->cl_ref);
1288 inode->i_fop = files->ops;
1289 inode->i_private = ncl;
1290 d_add(dentry, inode);
1291 fsnotify_create(dir, dentry);
1292 if (fdentries)
1293 fdentries[i] = dentry;
1294 }
1295 inode_unlock(dir);
1296 return 0;
1297out:
1298 inode_unlock(dir);
1299 return -ENOMEM;
1300}
1301
1302/* on success, returns positive number unique to that client. */
1303struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1304 struct nfsdfs_client *ncl, u32 id,
1305 const struct tree_descr *files,
1306 struct dentry **fdentries)
1307{
1308 struct dentry *dentry;
1309 char name[11];
1310 int ret;
1311
1312 sprintf(name, "%u", id);
1313
1314 dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1315 if (IS_ERR(dentry)) /* XXX: tossing errors? */
1316 return NULL;
1317 ret = nfsdfs_create_files(dentry, files, ncl, fdentries);
1318 if (ret) {
1319 nfsd_client_rmdir(dentry);
1320 return NULL;
1321 }
1322 return dentry;
1323}
1324
1325/* Taken from __rpc_rmdir: */
1326void nfsd_client_rmdir(struct dentry *dentry)
1327{
1328 simple_recursive_removal(dentry, clear_ncl);
1329}
1330
1331static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1332{
1333 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1334 nfsd_net_id);
1335 struct dentry *dentry;
1336 int ret;
1337
1338 static const struct tree_descr nfsd_files[] = {
1339 [NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1340 /* Per-export io stats use same ops as exports file */
1341 [NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1342 [NFSD_Export_features] = {"export_features",
1343 &export_features_fops, S_IRUGO},
1344 [NFSD_FO_UnlockIP] = {"unlock_ip",
1345 &transaction_ops, S_IWUSR|S_IRUSR},
1346 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1347 &transaction_ops, S_IWUSR|S_IRUSR},
1348 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1349 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1350 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1351 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1352 [NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1353 &nfsd_reply_cache_stats_fops, S_IRUGO},
1354 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1355 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1356 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1357 [NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1358 [NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
1359#ifdef CONFIG_NFSD_V4
1360 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1361 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1362 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1363 [NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1364#endif
1365 /* last one */ {""}
1366 };
1367
1368 ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1369 if (ret)
1370 return ret;
1371 _nfsd_symlink(sb->s_root, "supported_krb5_enctypes",
1372 "/proc/net/rpc/gss_krb5_enctypes");
1373 dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1374 if (IS_ERR(dentry))
1375 return PTR_ERR(dentry);
1376 nn->nfsd_client_dir = dentry;
1377 return 0;
1378}
1379
1380static int nfsd_fs_get_tree(struct fs_context *fc)
1381{
1382 return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1383}
1384
1385static void nfsd_fs_free_fc(struct fs_context *fc)
1386{
1387 if (fc->s_fs_info)
1388 put_net(fc->s_fs_info);
1389}
1390
1391static const struct fs_context_operations nfsd_fs_context_ops = {
1392 .free = nfsd_fs_free_fc,
1393 .get_tree = nfsd_fs_get_tree,
1394};
1395
1396static int nfsd_init_fs_context(struct fs_context *fc)
1397{
1398 put_user_ns(fc->user_ns);
1399 fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1400 fc->ops = &nfsd_fs_context_ops;
1401 return 0;
1402}
1403
1404static void nfsd_umount(struct super_block *sb)
1405{
1406 struct net *net = sb->s_fs_info;
1407
1408 nfsd_shutdown_threads(net);
1409
1410 kill_litter_super(sb);
1411 put_net(net);
1412}
1413
1414static struct file_system_type nfsd_fs_type = {
1415 .owner = THIS_MODULE,
1416 .name = "nfsd",
1417 .init_fs_context = nfsd_init_fs_context,
1418 .kill_sb = nfsd_umount,
1419};
1420MODULE_ALIAS_FS("nfsd");
1421
1422#ifdef CONFIG_PROC_FS
1423
1424static int exports_proc_open(struct inode *inode, struct file *file)
1425{
1426 return exports_net_open(current->nsproxy->net_ns, file);
1427}
1428
1429static const struct proc_ops exports_proc_ops = {
1430 .proc_open = exports_proc_open,
1431 .proc_read = seq_read,
1432 .proc_lseek = seq_lseek,
1433 .proc_release = seq_release,
1434};
1435
1436static int create_proc_exports_entry(void)
1437{
1438 struct proc_dir_entry *entry;
1439
1440 entry = proc_mkdir("fs/nfs", NULL);
1441 if (!entry)
1442 return -ENOMEM;
1443 entry = proc_create("exports", 0, entry, &exports_proc_ops);
1444 if (!entry) {
1445 remove_proc_entry("fs/nfs", NULL);
1446 return -ENOMEM;
1447 }
1448 return 0;
1449}
1450#else /* CONFIG_PROC_FS */
1451static int create_proc_exports_entry(void)
1452{
1453 return 0;
1454}
1455#endif
1456
1457unsigned int nfsd_net_id;
1458
1459/**
1460 * nfsd_nl_rpc_status_get_start - Prepare rpc_status_get dumpit
1461 * @cb: netlink metadata and command arguments
1462 *
1463 * Return values:
1464 * %0: The rpc_status_get command may proceed
1465 * %-ENODEV: There is no NFSD running in this namespace
1466 */
1467int nfsd_nl_rpc_status_get_start(struct netlink_callback *cb)
1468{
1469 struct nfsd_net *nn = net_generic(sock_net(cb->skb->sk), nfsd_net_id);
1470 int ret = -ENODEV;
1471
1472 mutex_lock(&nfsd_mutex);
1473 if (nn->nfsd_serv)
1474 ret = 0;
1475 else
1476 mutex_unlock(&nfsd_mutex);
1477
1478 return ret;
1479}
1480
1481static int nfsd_genl_rpc_status_compose_msg(struct sk_buff *skb,
1482 struct netlink_callback *cb,
1483 struct nfsd_genl_rqstp *rqstp)
1484{
1485 void *hdr;
1486 u32 i;
1487
1488 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
1489 &nfsd_nl_family, 0, NFSD_CMD_RPC_STATUS_GET);
1490 if (!hdr)
1491 return -ENOBUFS;
1492
1493 if (nla_put_be32(skb, NFSD_A_RPC_STATUS_XID, rqstp->rq_xid) ||
1494 nla_put_u32(skb, NFSD_A_RPC_STATUS_FLAGS, rqstp->rq_flags) ||
1495 nla_put_u32(skb, NFSD_A_RPC_STATUS_PROG, rqstp->rq_prog) ||
1496 nla_put_u32(skb, NFSD_A_RPC_STATUS_PROC, rqstp->rq_proc) ||
1497 nla_put_u8(skb, NFSD_A_RPC_STATUS_VERSION, rqstp->rq_vers) ||
1498 nla_put_s64(skb, NFSD_A_RPC_STATUS_SERVICE_TIME,
1499 ktime_to_us(rqstp->rq_stime),
1500 NFSD_A_RPC_STATUS_PAD))
1501 return -ENOBUFS;
1502
1503 switch (rqstp->rq_saddr.sa_family) {
1504 case AF_INET: {
1505 const struct sockaddr_in *s_in, *d_in;
1506
1507 s_in = (const struct sockaddr_in *)&rqstp->rq_saddr;
1508 d_in = (const struct sockaddr_in *)&rqstp->rq_daddr;
1509 if (nla_put_in_addr(skb, NFSD_A_RPC_STATUS_SADDR4,
1510 s_in->sin_addr.s_addr) ||
1511 nla_put_in_addr(skb, NFSD_A_RPC_STATUS_DADDR4,
1512 d_in->sin_addr.s_addr) ||
1513 nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1514 s_in->sin_port) ||
1515 nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1516 d_in->sin_port))
1517 return -ENOBUFS;
1518 break;
1519 }
1520 case AF_INET6: {
1521 const struct sockaddr_in6 *s_in, *d_in;
1522
1523 s_in = (const struct sockaddr_in6 *)&rqstp->rq_saddr;
1524 d_in = (const struct sockaddr_in6 *)&rqstp->rq_daddr;
1525 if (nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_SADDR6,
1526 &s_in->sin6_addr) ||
1527 nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_DADDR6,
1528 &d_in->sin6_addr) ||
1529 nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1530 s_in->sin6_port) ||
1531 nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1532 d_in->sin6_port))
1533 return -ENOBUFS;
1534 break;
1535 }
1536 }
1537
1538 for (i = 0; i < rqstp->rq_opcnt; i++)
1539 if (nla_put_u32(skb, NFSD_A_RPC_STATUS_COMPOUND_OPS,
1540 rqstp->rq_opnum[i]))
1541 return -ENOBUFS;
1542
1543 genlmsg_end(skb, hdr);
1544 return 0;
1545}
1546
1547/**
1548 * nfsd_nl_rpc_status_get_dumpit - Handle rpc_status_get dumpit
1549 * @skb: reply buffer
1550 * @cb: netlink metadata and command arguments
1551 *
1552 * Returns the size of the reply or a negative errno.
1553 */
1554int nfsd_nl_rpc_status_get_dumpit(struct sk_buff *skb,
1555 struct netlink_callback *cb)
1556{
1557 struct nfsd_net *nn = net_generic(sock_net(skb->sk), nfsd_net_id);
1558 int i, ret, rqstp_index = 0;
1559
1560 rcu_read_lock();
1561
1562 for (i = 0; i < nn->nfsd_serv->sv_nrpools; i++) {
1563 struct svc_rqst *rqstp;
1564
1565 if (i < cb->args[0]) /* already consumed */
1566 continue;
1567
1568 rqstp_index = 0;
1569 list_for_each_entry_rcu(rqstp,
1570 &nn->nfsd_serv->sv_pools[i].sp_all_threads,
1571 rq_all) {
1572 struct nfsd_genl_rqstp genl_rqstp;
1573 unsigned int status_counter;
1574
1575 if (rqstp_index++ < cb->args[1]) /* already consumed */
1576 continue;
1577 /*
1578 * Acquire rq_status_counter before parsing the rqst
1579 * fields. rq_status_counter is set to an odd value in
1580 * order to notify the consumers the rqstp fields are
1581 * meaningful.
1582 */
1583 status_counter =
1584 smp_load_acquire(&rqstp->rq_status_counter);
1585 if (!(status_counter & 1))
1586 continue;
1587
1588 genl_rqstp.rq_xid = rqstp->rq_xid;
1589 genl_rqstp.rq_flags = rqstp->rq_flags;
1590 genl_rqstp.rq_vers = rqstp->rq_vers;
1591 genl_rqstp.rq_prog = rqstp->rq_prog;
1592 genl_rqstp.rq_proc = rqstp->rq_proc;
1593 genl_rqstp.rq_stime = rqstp->rq_stime;
1594 genl_rqstp.rq_opcnt = 0;
1595 memcpy(&genl_rqstp.rq_daddr, svc_daddr(rqstp),
1596 sizeof(struct sockaddr));
1597 memcpy(&genl_rqstp.rq_saddr, svc_addr(rqstp),
1598 sizeof(struct sockaddr));
1599
1600#ifdef CONFIG_NFSD_V4
1601 if (rqstp->rq_vers == NFS4_VERSION &&
1602 rqstp->rq_proc == NFSPROC4_COMPOUND) {
1603 /* NFSv4 compound */
1604 struct nfsd4_compoundargs *args;
1605 int j;
1606
1607 args = rqstp->rq_argp;
1608 genl_rqstp.rq_opcnt = args->opcnt;
1609 for (j = 0; j < genl_rqstp.rq_opcnt; j++)
1610 genl_rqstp.rq_opnum[j] =
1611 args->ops[j].opnum;
1612 }
1613#endif /* CONFIG_NFSD_V4 */
1614
1615 /*
1616 * Acquire rq_status_counter before reporting the rqst
1617 * fields to the user.
1618 */
1619 if (smp_load_acquire(&rqstp->rq_status_counter) !=
1620 status_counter)
1621 continue;
1622
1623 ret = nfsd_genl_rpc_status_compose_msg(skb, cb,
1624 &genl_rqstp);
1625 if (ret)
1626 goto out;
1627 }
1628 }
1629
1630 cb->args[0] = i;
1631 cb->args[1] = rqstp_index;
1632 ret = skb->len;
1633out:
1634 rcu_read_unlock();
1635
1636 return ret;
1637}
1638
1639/**
1640 * nfsd_nl_rpc_status_get_done - rpc_status_get dumpit post-processing
1641 * @cb: netlink metadata and command arguments
1642 *
1643 * Return values:
1644 * %0: Success
1645 */
1646int nfsd_nl_rpc_status_get_done(struct netlink_callback *cb)
1647{
1648 mutex_unlock(&nfsd_mutex);
1649
1650 return 0;
1651}
1652
1653/**
1654 * nfsd_net_init - Prepare the nfsd_net portion of a new net namespace
1655 * @net: a freshly-created network namespace
1656 *
1657 * This information stays around as long as the network namespace is
1658 * alive whether or not there is an NFSD instance running in the
1659 * namespace.
1660 *
1661 * Returns zero on success, or a negative errno otherwise.
1662 */
1663static __net_init int nfsd_net_init(struct net *net)
1664{
1665 int retval;
1666 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1667
1668 retval = nfsd_export_init(net);
1669 if (retval)
1670 goto out_export_error;
1671 retval = nfsd_idmap_init(net);
1672 if (retval)
1673 goto out_idmap_error;
1674 retval = nfsd_net_reply_cache_init(nn);
1675 if (retval)
1676 goto out_repcache_error;
1677 nn->nfsd_versions = NULL;
1678 nn->nfsd4_minorversions = NULL;
1679 nfsd4_init_leases_net(nn);
1680 get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
1681 seqlock_init(&nn->writeverf_lock);
1682
1683 return 0;
1684
1685out_repcache_error:
1686 nfsd_idmap_shutdown(net);
1687out_idmap_error:
1688 nfsd_export_shutdown(net);
1689out_export_error:
1690 return retval;
1691}
1692
1693/**
1694 * nfsd_net_exit - Release the nfsd_net portion of a net namespace
1695 * @net: a network namespace that is about to be destroyed
1696 *
1697 */
1698static __net_exit void nfsd_net_exit(struct net *net)
1699{
1700 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1701
1702 nfsd_net_reply_cache_destroy(nn);
1703 nfsd_idmap_shutdown(net);
1704 nfsd_export_shutdown(net);
1705 nfsd_netns_free_versions(nn);
1706}
1707
1708static struct pernet_operations nfsd_net_ops = {
1709 .init = nfsd_net_init,
1710 .exit = nfsd_net_exit,
1711 .id = &nfsd_net_id,
1712 .size = sizeof(struct nfsd_net),
1713};
1714
1715static int __init init_nfsd(void)
1716{
1717 int retval;
1718
1719 retval = nfsd4_init_slabs();
1720 if (retval)
1721 return retval;
1722 retval = nfsd4_init_pnfs();
1723 if (retval)
1724 goto out_free_slabs;
1725 retval = nfsd_stat_init(); /* Statistics */
1726 if (retval)
1727 goto out_free_pnfs;
1728 retval = nfsd_drc_slab_create();
1729 if (retval)
1730 goto out_free_stat;
1731 nfsd_lockd_init(); /* lockd->nfsd callbacks */
1732 retval = create_proc_exports_entry();
1733 if (retval)
1734 goto out_free_lockd;
1735 retval = register_pernet_subsys(&nfsd_net_ops);
1736 if (retval < 0)
1737 goto out_free_exports;
1738 retval = register_cld_notifier();
1739 if (retval)
1740 goto out_free_subsys;
1741 retval = nfsd4_create_laundry_wq();
1742 if (retval)
1743 goto out_free_cld;
1744 retval = register_filesystem(&nfsd_fs_type);
1745 if (retval)
1746 goto out_free_all;
1747 retval = genl_register_family(&nfsd_nl_family);
1748 if (retval)
1749 goto out_free_all;
1750
1751 return 0;
1752out_free_all:
1753 nfsd4_destroy_laundry_wq();
1754out_free_cld:
1755 unregister_cld_notifier();
1756out_free_subsys:
1757 unregister_pernet_subsys(&nfsd_net_ops);
1758out_free_exports:
1759 remove_proc_entry("fs/nfs/exports", NULL);
1760 remove_proc_entry("fs/nfs", NULL);
1761out_free_lockd:
1762 nfsd_lockd_shutdown();
1763 nfsd_drc_slab_free();
1764out_free_stat:
1765 nfsd_stat_shutdown();
1766out_free_pnfs:
1767 nfsd4_exit_pnfs();
1768out_free_slabs:
1769 nfsd4_free_slabs();
1770 return retval;
1771}
1772
1773static void __exit exit_nfsd(void)
1774{
1775 genl_unregister_family(&nfsd_nl_family);
1776 unregister_filesystem(&nfsd_fs_type);
1777 nfsd4_destroy_laundry_wq();
1778 unregister_cld_notifier();
1779 unregister_pernet_subsys(&nfsd_net_ops);
1780 nfsd_drc_slab_free();
1781 remove_proc_entry("fs/nfs/exports", NULL);
1782 remove_proc_entry("fs/nfs", NULL);
1783 nfsd_stat_shutdown();
1784 nfsd_lockd_shutdown();
1785 nfsd4_free_slabs();
1786 nfsd4_exit_pnfs();
1787}
1788
1789MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1790MODULE_DESCRIPTION("In-kernel NFS server");
1791MODULE_LICENSE("GPL");
1792module_init(init_nfsd)
1793module_exit(exit_nfsd)