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 NFSD_Leasetime,
52 NFSD_Gracetime,
53 NFSD_RecoveryDir,
54 NFSD_V4EndGrace,
55 NFSD_MaxReserved
56};
57
58/*
59 * write() for these nodes.
60 */
61static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
62static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
63static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
64static ssize_t write_threads(struct file *file, char *buf, size_t size);
65static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
66static ssize_t write_versions(struct file *file, char *buf, size_t size);
67static ssize_t write_ports(struct file *file, char *buf, size_t size);
68static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
69static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
70#ifdef CONFIG_NFSD_V4
71static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
72static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
73#ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
74static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
75#endif
76static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
77#endif
78
79static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
80 [NFSD_Fh] = write_filehandle,
81 [NFSD_FO_UnlockIP] = write_unlock_ip,
82 [NFSD_FO_UnlockFS] = write_unlock_fs,
83 [NFSD_Threads] = write_threads,
84 [NFSD_Pool_Threads] = write_pool_threads,
85 [NFSD_Versions] = write_versions,
86 [NFSD_Ports] = write_ports,
87 [NFSD_MaxBlkSize] = write_maxblksize,
88 [NFSD_MaxConnections] = write_maxconn,
89#ifdef CONFIG_NFSD_V4
90 [NFSD_Leasetime] = write_leasetime,
91 [NFSD_Gracetime] = write_gracetime,
92#ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
93 [NFSD_RecoveryDir] = write_recoverydir,
94#endif
95 [NFSD_V4EndGrace] = write_v4_end_grace,
96#endif
97};
98
99static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
100{
101 ino_t ino = file_inode(file)->i_ino;
102 char *data;
103 ssize_t rv;
104
105 if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
106 return -EINVAL;
107
108 data = simple_transaction_get(file, buf, size);
109 if (IS_ERR(data))
110 return PTR_ERR(data);
111
112 rv = write_op[ino](file, data, size);
113 if (rv < 0)
114 return rv;
115
116 simple_transaction_set(file, rv);
117 return size;
118}
119
120static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
121{
122 if (! file->private_data) {
123 /* An attempt to read a transaction file without writing
124 * causes a 0-byte write so that the file can return
125 * state information
126 */
127 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
128 if (rv < 0)
129 return rv;
130 }
131 return simple_transaction_read(file, buf, size, pos);
132}
133
134static const struct file_operations transaction_ops = {
135 .write = nfsctl_transaction_write,
136 .read = nfsctl_transaction_read,
137 .release = simple_transaction_release,
138 .llseek = default_llseek,
139};
140
141static int exports_net_open(struct net *net, struct file *file)
142{
143 int err;
144 struct seq_file *seq;
145 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
146
147 err = seq_open(file, &nfs_exports_op);
148 if (err)
149 return err;
150
151 seq = file->private_data;
152 seq->private = nn->svc_export_cache;
153 return 0;
154}
155
156static int exports_nfsd_open(struct inode *inode, struct file *file)
157{
158 return exports_net_open(inode->i_sb->s_fs_info, file);
159}
160
161static const struct file_operations exports_nfsd_operations = {
162 .open = exports_nfsd_open,
163 .read = seq_read,
164 .llseek = seq_lseek,
165 .release = seq_release,
166};
167
168static int export_features_show(struct seq_file *m, void *v)
169{
170 seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
171 return 0;
172}
173
174DEFINE_SHOW_ATTRIBUTE(export_features);
175
176static const struct file_operations pool_stats_operations = {
177 .open = nfsd_pool_stats_open,
178 .read = seq_read,
179 .llseek = seq_lseek,
180 .release = seq_release,
181};
182
183DEFINE_SHOW_ATTRIBUTE(nfsd_reply_cache_stats);
184
185DEFINE_SHOW_ATTRIBUTE(nfsd_file_cache_stats);
186
187/*----------------------------------------------------------------------------*/
188/*
189 * payload - write methods
190 */
191
192static inline struct net *netns(struct file *file)
193{
194 return file_inode(file)->i_sb->s_fs_info;
195}
196
197/*
198 * write_unlock_ip - Release all locks used by a client
199 *
200 * Experimental.
201 *
202 * Input:
203 * buf: '\n'-terminated C string containing a
204 * presentation format IP address
205 * size: length of C string in @buf
206 * Output:
207 * On success: returns zero if all specified locks were released;
208 * returns one if one or more locks were not released
209 * On error: return code is negative errno value
210 */
211static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
212{
213 struct sockaddr_storage address;
214 struct sockaddr *sap = (struct sockaddr *)&address;
215 size_t salen = sizeof(address);
216 char *fo_path;
217 struct net *net = netns(file);
218
219 /* sanity check */
220 if (size == 0)
221 return -EINVAL;
222
223 if (buf[size-1] != '\n')
224 return -EINVAL;
225
226 fo_path = buf;
227 if (qword_get(&buf, fo_path, size) < 0)
228 return -EINVAL;
229
230 if (rpc_pton(net, fo_path, size, sap, salen) == 0)
231 return -EINVAL;
232
233 trace_nfsd_ctl_unlock_ip(net, buf);
234 return nlmsvc_unlock_all_by_ip(sap);
235}
236
237/*
238 * write_unlock_fs - Release all locks on a local file system
239 *
240 * Experimental.
241 *
242 * Input:
243 * buf: '\n'-terminated C string containing the
244 * absolute pathname of a local file system
245 * size: length of C string in @buf
246 * Output:
247 * On success: returns zero if all specified locks were released;
248 * returns one if one or more locks were not released
249 * On error: return code is negative errno value
250 */
251static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
252{
253 struct path path;
254 char *fo_path;
255 int error;
256
257 /* sanity check */
258 if (size == 0)
259 return -EINVAL;
260
261 if (buf[size-1] != '\n')
262 return -EINVAL;
263
264 fo_path = buf;
265 if (qword_get(&buf, fo_path, size) < 0)
266 return -EINVAL;
267 trace_nfsd_ctl_unlock_fs(netns(file), fo_path);
268 error = kern_path(fo_path, 0, &path);
269 if (error)
270 return error;
271
272 /*
273 * XXX: Needs better sanity checking. Otherwise we could end up
274 * releasing locks on the wrong file system.
275 *
276 * For example:
277 * 1. Does the path refer to a directory?
278 * 2. Is that directory a mount point, or
279 * 3. Is that directory the root of an exported file system?
280 */
281 error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
282 nfsd4_revoke_states(netns(file), path.dentry->d_sb);
283
284 path_put(&path);
285 return error;
286}
287
288/*
289 * write_filehandle - Get a variable-length NFS file handle by path
290 *
291 * On input, the buffer contains a '\n'-terminated C string comprised of
292 * three alphanumeric words separated by whitespace. The string may
293 * contain escape sequences.
294 *
295 * Input:
296 * buf:
297 * domain: client domain name
298 * path: export pathname
299 * maxsize: numeric maximum size of
300 * @buf
301 * size: length of C string in @buf
302 * Output:
303 * On success: passed-in buffer filled with '\n'-terminated C
304 * string containing a ASCII hex text version
305 * of the NFS file handle;
306 * return code is the size in bytes of the string
307 * On error: return code is negative errno value
308 */
309static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
310{
311 char *dname, *path;
312 int maxsize;
313 char *mesg = buf;
314 int len;
315 struct auth_domain *dom;
316 struct knfsd_fh fh;
317
318 if (size == 0)
319 return -EINVAL;
320
321 if (buf[size-1] != '\n')
322 return -EINVAL;
323 buf[size-1] = 0;
324
325 dname = mesg;
326 len = qword_get(&mesg, dname, size);
327 if (len <= 0)
328 return -EINVAL;
329
330 path = dname+len+1;
331 len = qword_get(&mesg, path, size);
332 if (len <= 0)
333 return -EINVAL;
334
335 len = get_int(&mesg, &maxsize);
336 if (len)
337 return len;
338
339 if (maxsize < NFS_FHSIZE)
340 return -EINVAL;
341 maxsize = min(maxsize, NFS3_FHSIZE);
342
343 if (qword_get(&mesg, mesg, size) > 0)
344 return -EINVAL;
345
346 trace_nfsd_ctl_filehandle(netns(file), dname, path, maxsize);
347
348 /* we have all the words, they are in buf.. */
349 dom = unix_domain_find(dname);
350 if (!dom)
351 return -ENOMEM;
352
353 len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
354 auth_domain_put(dom);
355 if (len)
356 return len;
357
358 mesg = buf;
359 len = SIMPLE_TRANSACTION_LIMIT;
360 qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
361 mesg[-1] = '\n';
362 return mesg - buf;
363}
364
365/*
366 * write_threads - Start NFSD, or report the current number of running threads
367 *
368 * Input:
369 * buf: ignored
370 * size: zero
371 * Output:
372 * On success: passed-in buffer filled with '\n'-terminated C
373 * string numeric value representing the number of
374 * running NFSD threads;
375 * return code is the size in bytes of the string
376 * On error: return code is zero
377 *
378 * OR
379 *
380 * Input:
381 * buf: C string containing an unsigned
382 * integer value representing the
383 * number of NFSD threads to start
384 * size: non-zero length of C string in @buf
385 * Output:
386 * On success: NFS service is started;
387 * passed-in buffer filled with '\n'-terminated C
388 * string numeric value representing the number of
389 * running NFSD threads;
390 * return code is the size in bytes of the string
391 * On error: return code is zero or a negative errno value
392 */
393static ssize_t write_threads(struct file *file, char *buf, size_t size)
394{
395 char *mesg = buf;
396 int rv;
397 struct net *net = netns(file);
398
399 if (size > 0) {
400 int newthreads;
401 rv = get_int(&mesg, &newthreads);
402 if (rv)
403 return rv;
404 if (newthreads < 0)
405 return -EINVAL;
406 trace_nfsd_ctl_threads(net, newthreads);
407 rv = nfsd_svc(newthreads, net, file->f_cred);
408 if (rv < 0)
409 return rv;
410 } else
411 rv = nfsd_nrthreads(net);
412
413 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
414}
415
416/*
417 * write_pool_threads - Set or report the current number of threads per pool
418 *
419 * Input:
420 * buf: ignored
421 * size: zero
422 *
423 * OR
424 *
425 * Input:
426 * buf: C string containing whitespace-
427 * separated unsigned integer values
428 * representing the number of NFSD
429 * threads to start in each pool
430 * size: non-zero length of C string in @buf
431 * Output:
432 * On success: passed-in buffer filled with '\n'-terminated C
433 * string containing integer values representing the
434 * number of NFSD threads in each pool;
435 * return code is the size in bytes of the string
436 * On error: return code is zero or a negative errno value
437 */
438static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
439{
440 /* if size > 0, look for an array of number of threads per node
441 * and apply them then write out number of threads per node as reply
442 */
443 char *mesg = buf;
444 int i;
445 int rv;
446 int len;
447 int npools;
448 int *nthreads;
449 struct net *net = netns(file);
450
451 mutex_lock(&nfsd_mutex);
452 npools = nfsd_nrpools(net);
453 if (npools == 0) {
454 /*
455 * NFS is shut down. The admin can start it by
456 * writing to the threads file but NOT the pool_threads
457 * file, sorry. Report zero threads.
458 */
459 mutex_unlock(&nfsd_mutex);
460 strcpy(buf, "0\n");
461 return strlen(buf);
462 }
463
464 nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
465 rv = -ENOMEM;
466 if (nthreads == NULL)
467 goto out_free;
468
469 if (size > 0) {
470 for (i = 0; i < npools; i++) {
471 rv = get_int(&mesg, &nthreads[i]);
472 if (rv == -ENOENT)
473 break; /* fewer numbers than pools */
474 if (rv)
475 goto out_free; /* syntax error */
476 rv = -EINVAL;
477 if (nthreads[i] < 0)
478 goto out_free;
479 trace_nfsd_ctl_pool_threads(net, i, nthreads[i]);
480 }
481 rv = nfsd_set_nrthreads(i, nthreads, net);
482 if (rv)
483 goto out_free;
484 }
485
486 rv = nfsd_get_nrthreads(npools, nthreads, net);
487 if (rv)
488 goto out_free;
489
490 mesg = buf;
491 size = SIMPLE_TRANSACTION_LIMIT;
492 for (i = 0; i < npools && size > 0; i++) {
493 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
494 len = strlen(mesg);
495 size -= len;
496 mesg += len;
497 }
498 rv = mesg - buf;
499out_free:
500 kfree(nthreads);
501 mutex_unlock(&nfsd_mutex);
502 return rv;
503}
504
505static ssize_t
506nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
507 const char *sep, unsigned vers, int minor)
508{
509 const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
510 bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
511
512 if (vers == 4 && minor >= 0 &&
513 !nfsd_minorversion(nn, minor, NFSD_TEST))
514 supported = false;
515 if (minor == 0 && supported)
516 /*
517 * special case for backward compatability.
518 * +4.0 is never reported, it is implied by
519 * +4, unless -4.0 is present.
520 */
521 return 0;
522 return snprintf(buf, remaining, format, sep,
523 supported ? '+' : '-', vers, minor);
524}
525
526static ssize_t __write_versions(struct file *file, char *buf, size_t size)
527{
528 char *mesg = buf;
529 char *vers, *minorp, sign;
530 int len, num, remaining;
531 ssize_t tlen = 0;
532 char *sep;
533 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
534
535 if (size > 0) {
536 if (nn->nfsd_serv)
537 /* Cannot change versions without updating
538 * nn->nfsd_serv->sv_xdrsize, and reallocing
539 * rq_argp and rq_resp
540 */
541 return -EBUSY;
542 if (buf[size-1] != '\n')
543 return -EINVAL;
544 buf[size-1] = 0;
545 trace_nfsd_ctl_version(netns(file), buf);
546
547 vers = mesg;
548 len = qword_get(&mesg, vers, size);
549 if (len <= 0) return -EINVAL;
550 do {
551 enum vers_op cmd;
552 unsigned minor;
553 sign = *vers;
554 if (sign == '+' || sign == '-')
555 num = simple_strtol((vers+1), &minorp, 0);
556 else
557 num = simple_strtol(vers, &minorp, 0);
558 if (*minorp == '.') {
559 if (num != 4)
560 return -EINVAL;
561 if (kstrtouint(minorp+1, 0, &minor) < 0)
562 return -EINVAL;
563 }
564
565 cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
566 switch(num) {
567#ifdef CONFIG_NFSD_V2
568 case 2:
569#endif
570 case 3:
571 nfsd_vers(nn, num, cmd);
572 break;
573 case 4:
574 if (*minorp == '.') {
575 if (nfsd_minorversion(nn, minor, cmd) < 0)
576 return -EINVAL;
577 } else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
578 /*
579 * Either we have +4 and no minors are enabled,
580 * or we have -4 and at least one minor is enabled.
581 * In either case, propagate 'cmd' to all minors.
582 */
583 minor = 0;
584 while (nfsd_minorversion(nn, minor, cmd) >= 0)
585 minor++;
586 }
587 break;
588 default:
589 /* Ignore requests to disable non-existent versions */
590 if (cmd == NFSD_SET)
591 return -EINVAL;
592 }
593 vers += len + 1;
594 } while ((len = qword_get(&mesg, vers, size)) > 0);
595 /* If all get turned off, turn them back on, as
596 * having no versions is BAD
597 */
598 nfsd_reset_versions(nn);
599 }
600
601 /* Now write current state into reply buffer */
602 sep = "";
603 remaining = SIMPLE_TRANSACTION_LIMIT;
604 for (num=2 ; num <= 4 ; num++) {
605 int minor;
606 if (!nfsd_vers(nn, num, NFSD_AVAIL))
607 continue;
608
609 minor = -1;
610 do {
611 len = nfsd_print_version_support(nn, buf, remaining,
612 sep, num, minor);
613 if (len >= remaining)
614 goto out;
615 remaining -= len;
616 buf += len;
617 tlen += len;
618 minor++;
619 if (len)
620 sep = " ";
621 } while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
622 }
623out:
624 len = snprintf(buf, remaining, "\n");
625 if (len >= remaining)
626 return -EINVAL;
627 return tlen + len;
628}
629
630/*
631 * write_versions - Set or report the available NFS protocol versions
632 *
633 * Input:
634 * buf: ignored
635 * size: zero
636 * Output:
637 * On success: passed-in buffer filled with '\n'-terminated C
638 * string containing positive or negative integer
639 * values representing the current status of each
640 * protocol version;
641 * return code is the size in bytes of the string
642 * On error: return code is zero or a negative errno value
643 *
644 * OR
645 *
646 * Input:
647 * buf: C string containing whitespace-
648 * separated positive or negative
649 * integer values representing NFS
650 * protocol versions to enable ("+n")
651 * or disable ("-n")
652 * size: non-zero length of C string in @buf
653 * Output:
654 * On success: status of zero or more protocol versions has
655 * been updated; passed-in buffer filled with
656 * '\n'-terminated C string containing positive
657 * or negative integer values representing the
658 * current status of each protocol version;
659 * return code is the size in bytes of the string
660 * On error: return code is zero or a negative errno value
661 */
662static ssize_t write_versions(struct file *file, char *buf, size_t size)
663{
664 ssize_t rv;
665
666 mutex_lock(&nfsd_mutex);
667 rv = __write_versions(file, buf, size);
668 mutex_unlock(&nfsd_mutex);
669 return rv;
670}
671
672/*
673 * Zero-length write. Return a list of NFSD's current listener
674 * transports.
675 */
676static ssize_t __write_ports_names(char *buf, struct net *net)
677{
678 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
679
680 if (nn->nfsd_serv == NULL)
681 return 0;
682 return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
683}
684
685/*
686 * A single 'fd' number was written, in which case it must be for
687 * a socket of a supported family/protocol, and we use it as an
688 * nfsd listener.
689 */
690static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
691{
692 char *mesg = buf;
693 int fd, err;
694 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
695 struct svc_serv *serv;
696
697 err = get_int(&mesg, &fd);
698 if (err != 0 || fd < 0)
699 return -EINVAL;
700 trace_nfsd_ctl_ports_addfd(net, fd);
701
702 err = nfsd_create_serv(net);
703 if (err != 0)
704 return err;
705
706 serv = nn->nfsd_serv;
707 err = svc_addsock(serv, net, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
708
709 if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
710 nfsd_destroy_serv(net);
711
712 return err;
713}
714
715/*
716 * A transport listener is added by writing its transport name and
717 * a port number.
718 */
719static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
720{
721 char transport[16];
722 struct svc_xprt *xprt;
723 int port, err;
724 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
725 struct svc_serv *serv;
726
727 if (sscanf(buf, "%15s %5u", transport, &port) != 2)
728 return -EINVAL;
729
730 if (port < 1 || port > USHRT_MAX)
731 return -EINVAL;
732 trace_nfsd_ctl_ports_addxprt(net, transport, port);
733
734 err = nfsd_create_serv(net);
735 if (err != 0)
736 return err;
737
738 serv = nn->nfsd_serv;
739 err = svc_xprt_create(serv, transport, net,
740 PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
741 if (err < 0)
742 goto out_err;
743
744 err = svc_xprt_create(serv, transport, net,
745 PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
746 if (err < 0 && err != -EAFNOSUPPORT)
747 goto out_close;
748
749 return 0;
750out_close:
751 xprt = svc_find_xprt(serv, transport, net, PF_INET, port);
752 if (xprt != NULL) {
753 svc_xprt_close(xprt);
754 svc_xprt_put(xprt);
755 }
756out_err:
757 if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
758 nfsd_destroy_serv(net);
759
760 return err;
761}
762
763static ssize_t __write_ports(struct file *file, char *buf, size_t size,
764 struct net *net)
765{
766 if (size == 0)
767 return __write_ports_names(buf, net);
768
769 if (isdigit(buf[0]))
770 return __write_ports_addfd(buf, net, file->f_cred);
771
772 if (isalpha(buf[0]))
773 return __write_ports_addxprt(buf, net, file->f_cred);
774
775 return -EINVAL;
776}
777
778/*
779 * write_ports - Pass a socket file descriptor or transport name to listen on
780 *
781 * Input:
782 * buf: ignored
783 * size: zero
784 * Output:
785 * On success: passed-in buffer filled with a '\n'-terminated C
786 * string containing a whitespace-separated list of
787 * named NFSD listeners;
788 * return code is the size in bytes of the string
789 * On error: return code is zero or a negative errno value
790 *
791 * OR
792 *
793 * Input:
794 * buf: C string containing an unsigned
795 * integer value representing a bound
796 * but unconnected socket that is to be
797 * used as an NFSD listener; listen(3)
798 * must be called for a SOCK_STREAM
799 * socket, otherwise it is ignored
800 * size: non-zero length of C string in @buf
801 * Output:
802 * On success: NFS service is started;
803 * passed-in buffer filled with a '\n'-terminated C
804 * string containing a unique alphanumeric name of
805 * the listener;
806 * return code is the size in bytes of the string
807 * On error: return code is a negative errno value
808 *
809 * OR
810 *
811 * Input:
812 * buf: C string containing a transport
813 * name and an unsigned integer value
814 * representing the port to listen on,
815 * separated by whitespace
816 * size: non-zero length of C string in @buf
817 * Output:
818 * On success: returns zero; NFS service is started
819 * On error: return code is a negative errno value
820 */
821static ssize_t write_ports(struct file *file, char *buf, size_t size)
822{
823 ssize_t rv;
824
825 mutex_lock(&nfsd_mutex);
826 rv = __write_ports(file, buf, size, netns(file));
827 mutex_unlock(&nfsd_mutex);
828 return rv;
829}
830
831
832int nfsd_max_blksize;
833
834/*
835 * write_maxblksize - Set or report the current NFS blksize
836 *
837 * Input:
838 * buf: ignored
839 * size: zero
840 *
841 * OR
842 *
843 * Input:
844 * buf: C string containing an unsigned
845 * integer value representing the new
846 * NFS blksize
847 * size: non-zero length of C string in @buf
848 * Output:
849 * On success: passed-in buffer filled with '\n'-terminated C string
850 * containing numeric value of the current NFS blksize
851 * setting;
852 * return code is the size in bytes of the string
853 * On error: return code is zero or a negative errno value
854 */
855static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
856{
857 char *mesg = buf;
858 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
859
860 if (size > 0) {
861 int bsize;
862 int rv = get_int(&mesg, &bsize);
863 if (rv)
864 return rv;
865 trace_nfsd_ctl_maxblksize(netns(file), bsize);
866
867 /* force bsize into allowed range and
868 * required alignment.
869 */
870 bsize = max_t(int, bsize, 1024);
871 bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
872 bsize &= ~(1024-1);
873 mutex_lock(&nfsd_mutex);
874 if (nn->nfsd_serv) {
875 mutex_unlock(&nfsd_mutex);
876 return -EBUSY;
877 }
878 nfsd_max_blksize = bsize;
879 mutex_unlock(&nfsd_mutex);
880 }
881
882 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
883 nfsd_max_blksize);
884}
885
886/*
887 * write_maxconn - Set or report the current max number of connections
888 *
889 * Input:
890 * buf: ignored
891 * size: zero
892 * OR
893 *
894 * Input:
895 * buf: C string containing an unsigned
896 * integer value representing the new
897 * number of max connections
898 * size: non-zero length of C string in @buf
899 * Output:
900 * On success: passed-in buffer filled with '\n'-terminated C string
901 * containing numeric value of max_connections setting
902 * for this net namespace;
903 * return code is the size in bytes of the string
904 * On error: return code is zero or a negative errno value
905 */
906static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
907{
908 char *mesg = buf;
909 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
910 unsigned int maxconn = nn->max_connections;
911
912 if (size > 0) {
913 int rv = get_uint(&mesg, &maxconn);
914
915 if (rv)
916 return rv;
917 trace_nfsd_ctl_maxconn(netns(file), maxconn);
918 nn->max_connections = maxconn;
919 }
920
921 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
922}
923
924#ifdef CONFIG_NFSD_V4
925static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
926 time64_t *time, struct nfsd_net *nn)
927{
928 struct dentry *dentry = file_dentry(file);
929 char *mesg = buf;
930 int rv, i;
931
932 if (size > 0) {
933 if (nn->nfsd_serv)
934 return -EBUSY;
935 rv = get_int(&mesg, &i);
936 if (rv)
937 return rv;
938 trace_nfsd_ctl_time(netns(file), dentry->d_name.name,
939 dentry->d_name.len, i);
940
941 /*
942 * Some sanity checking. We don't have a reason for
943 * these particular numbers, but problems with the
944 * extremes are:
945 * - Too short: the briefest network outage may
946 * cause clients to lose all their locks. Also,
947 * the frequent polling may be wasteful.
948 * - Too long: do you really want reboot recovery
949 * to take more than an hour? Or to make other
950 * clients wait an hour before being able to
951 * revoke a dead client's locks?
952 */
953 if (i < 10 || i > 3600)
954 return -EINVAL;
955 *time = i;
956 }
957
958 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
959}
960
961static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
962 time64_t *time, struct nfsd_net *nn)
963{
964 ssize_t rv;
965
966 mutex_lock(&nfsd_mutex);
967 rv = __nfsd4_write_time(file, buf, size, time, nn);
968 mutex_unlock(&nfsd_mutex);
969 return rv;
970}
971
972/*
973 * write_leasetime - Set or report the current NFSv4 lease time
974 *
975 * Input:
976 * buf: ignored
977 * size: zero
978 *
979 * OR
980 *
981 * Input:
982 * buf: C string containing an unsigned
983 * integer value representing the new
984 * NFSv4 lease expiry time
985 * size: non-zero length of C string in @buf
986 * Output:
987 * On success: passed-in buffer filled with '\n'-terminated C
988 * string containing unsigned integer value of the
989 * current lease expiry time;
990 * return code is the size in bytes of the string
991 * On error: return code is zero or a negative errno value
992 */
993static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
994{
995 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
996 return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
997}
998
999/*
1000 * write_gracetime - Set or report current NFSv4 grace period time
1001 *
1002 * As above, but sets the time of the NFSv4 grace period.
1003 *
1004 * Note this should never be set to less than the *previous*
1005 * lease-period time, but we don't try to enforce this. (In the common
1006 * case (a new boot), we don't know what the previous lease time was
1007 * anyway.)
1008 */
1009static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1010{
1011 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1012 return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1013}
1014
1015#ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1016static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1017 struct nfsd_net *nn)
1018{
1019 char *mesg = buf;
1020 char *recdir;
1021 int len, status;
1022
1023 if (size > 0) {
1024 if (nn->nfsd_serv)
1025 return -EBUSY;
1026 if (size > PATH_MAX || buf[size-1] != '\n')
1027 return -EINVAL;
1028 buf[size-1] = 0;
1029
1030 recdir = mesg;
1031 len = qword_get(&mesg, recdir, size);
1032 if (len <= 0)
1033 return -EINVAL;
1034 trace_nfsd_ctl_recoverydir(netns(file), recdir);
1035
1036 status = nfs4_reset_recoverydir(recdir);
1037 if (status)
1038 return status;
1039 }
1040
1041 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1042 nfs4_recoverydir());
1043}
1044
1045/*
1046 * write_recoverydir - Set or report the pathname of the recovery directory
1047 *
1048 * Input:
1049 * buf: ignored
1050 * size: zero
1051 *
1052 * OR
1053 *
1054 * Input:
1055 * buf: C string containing the pathname
1056 * of the directory on a local file
1057 * system containing permanent NFSv4
1058 * recovery data
1059 * size: non-zero length of C string in @buf
1060 * Output:
1061 * On success: passed-in buffer filled with '\n'-terminated C string
1062 * containing the current recovery pathname setting;
1063 * return code is the size in bytes of the string
1064 * On error: return code is zero or a negative errno value
1065 */
1066static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1067{
1068 ssize_t rv;
1069 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1070
1071 mutex_lock(&nfsd_mutex);
1072 rv = __write_recoverydir(file, buf, size, nn);
1073 mutex_unlock(&nfsd_mutex);
1074 return rv;
1075}
1076#endif
1077
1078/*
1079 * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1080 *
1081 * Input:
1082 * buf: ignored
1083 * size: zero
1084 * OR
1085 *
1086 * Input:
1087 * buf: any value
1088 * size: non-zero length of C string in @buf
1089 * Output:
1090 * passed-in buffer filled with "Y" or "N" with a newline
1091 * and NULL-terminated C string. This indicates whether
1092 * the grace period has ended in the current net
1093 * namespace. Return code is the size in bytes of the
1094 * string. Writing a string that starts with 'Y', 'y', or
1095 * '1' to the file will end the grace period for nfsd's v4
1096 * lock manager.
1097 */
1098static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1099{
1100 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1101
1102 if (size > 0) {
1103 switch(buf[0]) {
1104 case 'Y':
1105 case 'y':
1106 case '1':
1107 if (!nn->nfsd_serv)
1108 return -EBUSY;
1109 trace_nfsd_end_grace(netns(file));
1110 nfsd4_end_grace(nn);
1111 break;
1112 default:
1113 return -EINVAL;
1114 }
1115 }
1116
1117 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1118 nn->grace_ended ? 'Y' : 'N');
1119}
1120
1121#endif
1122
1123/*----------------------------------------------------------------------------*/
1124/*
1125 * populating the filesystem.
1126 */
1127
1128/* Basically copying rpc_get_inode. */
1129static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1130{
1131 struct inode *inode = new_inode(sb);
1132 if (!inode)
1133 return NULL;
1134 /* Following advice from simple_fill_super documentation: */
1135 inode->i_ino = iunique(sb, NFSD_MaxReserved);
1136 inode->i_mode = mode;
1137 simple_inode_init_ts(inode);
1138 switch (mode & S_IFMT) {
1139 case S_IFDIR:
1140 inode->i_fop = &simple_dir_operations;
1141 inode->i_op = &simple_dir_inode_operations;
1142 inc_nlink(inode);
1143 break;
1144 case S_IFLNK:
1145 inode->i_op = &simple_symlink_inode_operations;
1146 break;
1147 default:
1148 break;
1149 }
1150 return inode;
1151}
1152
1153static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1154{
1155 struct inode *inode;
1156
1157 inode = nfsd_get_inode(dir->i_sb, mode);
1158 if (!inode)
1159 return -ENOMEM;
1160 if (ncl) {
1161 inode->i_private = ncl;
1162 kref_get(&ncl->cl_ref);
1163 }
1164 d_add(dentry, inode);
1165 inc_nlink(dir);
1166 fsnotify_mkdir(dir, dentry);
1167 return 0;
1168}
1169
1170static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1171{
1172 struct inode *dir = parent->d_inode;
1173 struct dentry *dentry;
1174 int ret = -ENOMEM;
1175
1176 inode_lock(dir);
1177 dentry = d_alloc_name(parent, name);
1178 if (!dentry)
1179 goto out_err;
1180 ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1181 if (ret)
1182 goto out_err;
1183out:
1184 inode_unlock(dir);
1185 return dentry;
1186out_err:
1187 dput(dentry);
1188 dentry = ERR_PTR(ret);
1189 goto out;
1190}
1191
1192#if IS_ENABLED(CONFIG_SUNRPC_GSS)
1193static int __nfsd_symlink(struct inode *dir, struct dentry *dentry,
1194 umode_t mode, const char *content)
1195{
1196 struct inode *inode;
1197
1198 inode = nfsd_get_inode(dir->i_sb, mode);
1199 if (!inode)
1200 return -ENOMEM;
1201
1202 inode->i_link = (char *)content;
1203 inode->i_size = strlen(content);
1204
1205 d_add(dentry, inode);
1206 inc_nlink(dir);
1207 fsnotify_create(dir, dentry);
1208 return 0;
1209}
1210
1211/*
1212 * @content is assumed to be a NUL-terminated string that lives
1213 * longer than the symlink itself.
1214 */
1215static void _nfsd_symlink(struct dentry *parent, const char *name,
1216 const char *content)
1217{
1218 struct inode *dir = parent->d_inode;
1219 struct dentry *dentry;
1220 int ret;
1221
1222 inode_lock(dir);
1223 dentry = d_alloc_name(parent, name);
1224 if (!dentry)
1225 goto out;
1226 ret = __nfsd_symlink(d_inode(parent), dentry, S_IFLNK | 0777, content);
1227 if (ret)
1228 dput(dentry);
1229out:
1230 inode_unlock(dir);
1231}
1232#else
1233static inline void _nfsd_symlink(struct dentry *parent, const char *name,
1234 const char *content)
1235{
1236}
1237
1238#endif
1239
1240static void clear_ncl(struct dentry *dentry)
1241{
1242 struct inode *inode = d_inode(dentry);
1243 struct nfsdfs_client *ncl = inode->i_private;
1244
1245 spin_lock(&inode->i_lock);
1246 inode->i_private = NULL;
1247 spin_unlock(&inode->i_lock);
1248 kref_put(&ncl->cl_ref, ncl->cl_release);
1249}
1250
1251struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1252{
1253 struct nfsdfs_client *nc;
1254
1255 spin_lock(&inode->i_lock);
1256 nc = inode->i_private;
1257 if (nc)
1258 kref_get(&nc->cl_ref);
1259 spin_unlock(&inode->i_lock);
1260 return nc;
1261}
1262
1263/* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1264 * code instead. */
1265static int nfsdfs_create_files(struct dentry *root,
1266 const struct tree_descr *files,
1267 struct nfsdfs_client *ncl,
1268 struct dentry **fdentries)
1269{
1270 struct inode *dir = d_inode(root);
1271 struct inode *inode;
1272 struct dentry *dentry;
1273 int i;
1274
1275 inode_lock(dir);
1276 for (i = 0; files->name && files->name[0]; i++, files++) {
1277 dentry = d_alloc_name(root, files->name);
1278 if (!dentry)
1279 goto out;
1280 inode = nfsd_get_inode(d_inode(root)->i_sb,
1281 S_IFREG | files->mode);
1282 if (!inode) {
1283 dput(dentry);
1284 goto out;
1285 }
1286 kref_get(&ncl->cl_ref);
1287 inode->i_fop = files->ops;
1288 inode->i_private = ncl;
1289 d_add(dentry, inode);
1290 fsnotify_create(dir, dentry);
1291 if (fdentries)
1292 fdentries[i] = dentry;
1293 }
1294 inode_unlock(dir);
1295 return 0;
1296out:
1297 inode_unlock(dir);
1298 return -ENOMEM;
1299}
1300
1301/* on success, returns positive number unique to that client. */
1302struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1303 struct nfsdfs_client *ncl, u32 id,
1304 const struct tree_descr *files,
1305 struct dentry **fdentries)
1306{
1307 struct dentry *dentry;
1308 char name[11];
1309 int ret;
1310
1311 sprintf(name, "%u", id);
1312
1313 dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1314 if (IS_ERR(dentry)) /* XXX: tossing errors? */
1315 return NULL;
1316 ret = nfsdfs_create_files(dentry, files, ncl, fdentries);
1317 if (ret) {
1318 nfsd_client_rmdir(dentry);
1319 return NULL;
1320 }
1321 return dentry;
1322}
1323
1324/* Taken from __rpc_rmdir: */
1325void nfsd_client_rmdir(struct dentry *dentry)
1326{
1327 simple_recursive_removal(dentry, clear_ncl);
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#ifdef CONFIG_NFSD_V4
1359 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1360 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1361#ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1362 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1363#endif
1364 [NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1365#endif
1366 /* last one */ {""}
1367 };
1368
1369 ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1370 if (ret)
1371 return ret;
1372 _nfsd_symlink(sb->s_root, "supported_krb5_enctypes",
1373 "/proc/net/rpc/gss_krb5_enctypes");
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
1424
1425static int exports_proc_open(struct inode *inode, struct file *file)
1426{
1427 return exports_net_open(current->nsproxy->net_ns, file);
1428}
1429
1430static const struct proc_ops exports_proc_ops = {
1431 .proc_open = exports_proc_open,
1432 .proc_read = seq_read,
1433 .proc_lseek = seq_lseek,
1434 .proc_release = seq_release,
1435};
1436
1437static int create_proc_exports_entry(void)
1438{
1439 struct proc_dir_entry *entry;
1440
1441 entry = proc_mkdir("fs/nfs", NULL);
1442 if (!entry)
1443 return -ENOMEM;
1444 entry = proc_create("exports", 0, entry, &exports_proc_ops);
1445 if (!entry) {
1446 remove_proc_entry("fs/nfs", NULL);
1447 return -ENOMEM;
1448 }
1449 return 0;
1450}
1451#else /* CONFIG_PROC_FS */
1452static int create_proc_exports_entry(void)
1453{
1454 return 0;
1455}
1456#endif
1457
1458unsigned int nfsd_net_id;
1459
1460/**
1461 * nfsd_nl_rpc_status_get_start - Prepare rpc_status_get dumpit
1462 * @cb: netlink metadata and command arguments
1463 *
1464 * Return values:
1465 * %0: The rpc_status_get command may proceed
1466 * %-ENODEV: There is no NFSD running in this namespace
1467 */
1468int nfsd_nl_rpc_status_get_start(struct netlink_callback *cb)
1469{
1470 struct nfsd_net *nn = net_generic(sock_net(cb->skb->sk), nfsd_net_id);
1471 int ret = -ENODEV;
1472
1473 mutex_lock(&nfsd_mutex);
1474 if (nn->nfsd_serv)
1475 ret = 0;
1476 else
1477 mutex_unlock(&nfsd_mutex);
1478
1479 return ret;
1480}
1481
1482static int nfsd_genl_rpc_status_compose_msg(struct sk_buff *skb,
1483 struct netlink_callback *cb,
1484 struct nfsd_genl_rqstp *rqstp)
1485{
1486 void *hdr;
1487 u32 i;
1488
1489 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
1490 &nfsd_nl_family, 0, NFSD_CMD_RPC_STATUS_GET);
1491 if (!hdr)
1492 return -ENOBUFS;
1493
1494 if (nla_put_be32(skb, NFSD_A_RPC_STATUS_XID, rqstp->rq_xid) ||
1495 nla_put_u32(skb, NFSD_A_RPC_STATUS_FLAGS, rqstp->rq_flags) ||
1496 nla_put_u32(skb, NFSD_A_RPC_STATUS_PROG, rqstp->rq_prog) ||
1497 nla_put_u32(skb, NFSD_A_RPC_STATUS_PROC, rqstp->rq_proc) ||
1498 nla_put_u8(skb, NFSD_A_RPC_STATUS_VERSION, rqstp->rq_vers) ||
1499 nla_put_s64(skb, NFSD_A_RPC_STATUS_SERVICE_TIME,
1500 ktime_to_us(rqstp->rq_stime),
1501 NFSD_A_RPC_STATUS_PAD))
1502 return -ENOBUFS;
1503
1504 switch (rqstp->rq_saddr.sa_family) {
1505 case AF_INET: {
1506 const struct sockaddr_in *s_in, *d_in;
1507
1508 s_in = (const struct sockaddr_in *)&rqstp->rq_saddr;
1509 d_in = (const struct sockaddr_in *)&rqstp->rq_daddr;
1510 if (nla_put_in_addr(skb, NFSD_A_RPC_STATUS_SADDR4,
1511 s_in->sin_addr.s_addr) ||
1512 nla_put_in_addr(skb, NFSD_A_RPC_STATUS_DADDR4,
1513 d_in->sin_addr.s_addr) ||
1514 nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1515 s_in->sin_port) ||
1516 nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1517 d_in->sin_port))
1518 return -ENOBUFS;
1519 break;
1520 }
1521 case AF_INET6: {
1522 const struct sockaddr_in6 *s_in, *d_in;
1523
1524 s_in = (const struct sockaddr_in6 *)&rqstp->rq_saddr;
1525 d_in = (const struct sockaddr_in6 *)&rqstp->rq_daddr;
1526 if (nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_SADDR6,
1527 &s_in->sin6_addr) ||
1528 nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_DADDR6,
1529 &d_in->sin6_addr) ||
1530 nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1531 s_in->sin6_port) ||
1532 nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1533 d_in->sin6_port))
1534 return -ENOBUFS;
1535 break;
1536 }
1537 }
1538
1539 for (i = 0; i < rqstp->rq_opcnt; i++)
1540 if (nla_put_u32(skb, NFSD_A_RPC_STATUS_COMPOUND_OPS,
1541 rqstp->rq_opnum[i]))
1542 return -ENOBUFS;
1543
1544 genlmsg_end(skb, hdr);
1545 return 0;
1546}
1547
1548/**
1549 * nfsd_nl_rpc_status_get_dumpit - Handle rpc_status_get dumpit
1550 * @skb: reply buffer
1551 * @cb: netlink metadata and command arguments
1552 *
1553 * Returns the size of the reply or a negative errno.
1554 */
1555int nfsd_nl_rpc_status_get_dumpit(struct sk_buff *skb,
1556 struct netlink_callback *cb)
1557{
1558 struct nfsd_net *nn = net_generic(sock_net(skb->sk), nfsd_net_id);
1559 int i, ret, rqstp_index = 0;
1560
1561 rcu_read_lock();
1562
1563 for (i = 0; i < nn->nfsd_serv->sv_nrpools; i++) {
1564 struct svc_rqst *rqstp;
1565
1566 if (i < cb->args[0]) /* already consumed */
1567 continue;
1568
1569 rqstp_index = 0;
1570 list_for_each_entry_rcu(rqstp,
1571 &nn->nfsd_serv->sv_pools[i].sp_all_threads,
1572 rq_all) {
1573 struct nfsd_genl_rqstp genl_rqstp;
1574 unsigned int status_counter;
1575
1576 if (rqstp_index++ < cb->args[1]) /* already consumed */
1577 continue;
1578 /*
1579 * Acquire rq_status_counter before parsing the rqst
1580 * fields. rq_status_counter is set to an odd value in
1581 * order to notify the consumers the rqstp fields are
1582 * meaningful.
1583 */
1584 status_counter =
1585 smp_load_acquire(&rqstp->rq_status_counter);
1586 if (!(status_counter & 1))
1587 continue;
1588
1589 genl_rqstp.rq_xid = rqstp->rq_xid;
1590 genl_rqstp.rq_flags = rqstp->rq_flags;
1591 genl_rqstp.rq_vers = rqstp->rq_vers;
1592 genl_rqstp.rq_prog = rqstp->rq_prog;
1593 genl_rqstp.rq_proc = rqstp->rq_proc;
1594 genl_rqstp.rq_stime = rqstp->rq_stime;
1595 genl_rqstp.rq_opcnt = 0;
1596 memcpy(&genl_rqstp.rq_daddr, svc_daddr(rqstp),
1597 sizeof(struct sockaddr));
1598 memcpy(&genl_rqstp.rq_saddr, svc_addr(rqstp),
1599 sizeof(struct sockaddr));
1600
1601#ifdef CONFIG_NFSD_V4
1602 if (rqstp->rq_vers == NFS4_VERSION &&
1603 rqstp->rq_proc == NFSPROC4_COMPOUND) {
1604 /* NFSv4 compound */
1605 struct nfsd4_compoundargs *args;
1606 int j;
1607
1608 args = rqstp->rq_argp;
1609 genl_rqstp.rq_opcnt = args->opcnt;
1610 for (j = 0; j < genl_rqstp.rq_opcnt; j++)
1611 genl_rqstp.rq_opnum[j] =
1612 args->ops[j].opnum;
1613 }
1614#endif /* CONFIG_NFSD_V4 */
1615
1616 /*
1617 * Acquire rq_status_counter before reporting the rqst
1618 * fields to the user.
1619 */
1620 if (smp_load_acquire(&rqstp->rq_status_counter) !=
1621 status_counter)
1622 continue;
1623
1624 ret = nfsd_genl_rpc_status_compose_msg(skb, cb,
1625 &genl_rqstp);
1626 if (ret)
1627 goto out;
1628 }
1629 }
1630
1631 cb->args[0] = i;
1632 cb->args[1] = rqstp_index;
1633 ret = skb->len;
1634out:
1635 rcu_read_unlock();
1636
1637 return ret;
1638}
1639
1640/**
1641 * nfsd_nl_rpc_status_get_done - rpc_status_get dumpit post-processing
1642 * @cb: netlink metadata and command arguments
1643 *
1644 * Return values:
1645 * %0: Success
1646 */
1647int nfsd_nl_rpc_status_get_done(struct netlink_callback *cb)
1648{
1649 mutex_unlock(&nfsd_mutex);
1650
1651 return 0;
1652}
1653
1654/**
1655 * nfsd_net_init - Prepare the nfsd_net portion of a new net namespace
1656 * @net: a freshly-created network namespace
1657 *
1658 * This information stays around as long as the network namespace is
1659 * alive whether or not there is an NFSD instance running in the
1660 * namespace.
1661 *
1662 * Returns zero on success, or a negative errno otherwise.
1663 */
1664static __net_init int nfsd_net_init(struct net *net)
1665{
1666 int retval;
1667 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1668
1669 retval = nfsd_export_init(net);
1670 if (retval)
1671 goto out_export_error;
1672 retval = nfsd_idmap_init(net);
1673 if (retval)
1674 goto out_idmap_error;
1675 retval = nfsd_stat_counters_init(nn);
1676 if (retval)
1677 goto out_repcache_error;
1678 memset(&nn->nfsd_svcstats, 0, sizeof(nn->nfsd_svcstats));
1679 nn->nfsd_svcstats.program = &nfsd_program;
1680 nn->nfsd_versions = NULL;
1681 nn->nfsd4_minorversions = NULL;
1682 nfsd4_init_leases_net(nn);
1683 get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
1684 seqlock_init(&nn->writeverf_lock);
1685 nfsd_proc_stat_init(net);
1686
1687 return 0;
1688
1689out_repcache_error:
1690 nfsd_idmap_shutdown(net);
1691out_idmap_error:
1692 nfsd_export_shutdown(net);
1693out_export_error:
1694 return retval;
1695}
1696
1697/**
1698 * nfsd_net_exit - Release the nfsd_net portion of a net namespace
1699 * @net: a network namespace that is about to be destroyed
1700 *
1701 */
1702static __net_exit void nfsd_net_exit(struct net *net)
1703{
1704 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1705
1706 nfsd_proc_stat_shutdown(net);
1707 nfsd_stat_counters_destroy(nn);
1708 nfsd_idmap_shutdown(net);
1709 nfsd_export_shutdown(net);
1710 nfsd_netns_free_versions(nn);
1711}
1712
1713static struct pernet_operations nfsd_net_ops = {
1714 .init = nfsd_net_init,
1715 .exit = nfsd_net_exit,
1716 .id = &nfsd_net_id,
1717 .size = sizeof(struct nfsd_net),
1718};
1719
1720static int __init init_nfsd(void)
1721{
1722 int retval;
1723
1724 retval = nfsd4_init_slabs();
1725 if (retval)
1726 return retval;
1727 retval = nfsd4_init_pnfs();
1728 if (retval)
1729 goto out_free_slabs;
1730 retval = nfsd_drc_slab_create();
1731 if (retval)
1732 goto out_free_pnfs;
1733 nfsd_lockd_init(); /* lockd->nfsd callbacks */
1734 retval = create_proc_exports_entry();
1735 if (retval)
1736 goto out_free_lockd;
1737 retval = register_pernet_subsys(&nfsd_net_ops);
1738 if (retval < 0)
1739 goto out_free_exports;
1740 retval = register_cld_notifier();
1741 if (retval)
1742 goto out_free_subsys;
1743 retval = nfsd4_create_laundry_wq();
1744 if (retval)
1745 goto out_free_cld;
1746 retval = register_filesystem(&nfsd_fs_type);
1747 if (retval)
1748 goto out_free_all;
1749 retval = genl_register_family(&nfsd_nl_family);
1750 if (retval)
1751 goto out_free_all;
1752
1753 return 0;
1754out_free_all:
1755 nfsd4_destroy_laundry_wq();
1756out_free_cld:
1757 unregister_cld_notifier();
1758out_free_subsys:
1759 unregister_pernet_subsys(&nfsd_net_ops);
1760out_free_exports:
1761 remove_proc_entry("fs/nfs/exports", NULL);
1762 remove_proc_entry("fs/nfs", NULL);
1763out_free_lockd:
1764 nfsd_lockd_shutdown();
1765 nfsd_drc_slab_free();
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_lockd_shutdown();
1784 nfsd4_free_slabs();
1785 nfsd4_exit_pnfs();
1786}
1787
1788MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1789MODULE_DESCRIPTION("In-kernel NFS server");
1790MODULE_LICENSE("GPL");
1791module_init(init_nfsd)
1792module_exit(exit_nfsd)