Loading...
1/*
2 * Syscall interface to knfsd.
3 *
4 * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
5 */
6
7#include <linux/slab.h>
8#include <linux/namei.h>
9#include <linux/ctype.h>
10
11#include <linux/sunrpc/svcsock.h>
12#include <linux/lockd/lockd.h>
13#include <linux/sunrpc/clnt.h>
14#include <linux/sunrpc/gss_api.h>
15#include <linux/sunrpc/gss_krb5_enctypes.h>
16#include <linux/sunrpc/rpc_pipe_fs.h>
17#include <linux/module.h>
18
19#include "idmap.h"
20#include "nfsd.h"
21#include "cache.h"
22#include "fault_inject.h"
23#include "netns.h"
24
25/*
26 * We have a single directory with several nodes in it.
27 */
28enum {
29 NFSD_Root = 1,
30 NFSD_List,
31 NFSD_Export_features,
32 NFSD_Fh,
33 NFSD_FO_UnlockIP,
34 NFSD_FO_UnlockFS,
35 NFSD_Threads,
36 NFSD_Pool_Threads,
37 NFSD_Pool_Stats,
38 NFSD_Versions,
39 NFSD_Ports,
40 NFSD_MaxBlkSize,
41 NFSD_SupportedEnctypes,
42 /*
43 * The below MUST come last. Otherwise we leave a hole in nfsd_files[]
44 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
45 */
46#ifdef CONFIG_NFSD_V4
47 NFSD_Leasetime,
48 NFSD_Gracetime,
49 NFSD_RecoveryDir,
50#endif
51};
52
53/*
54 * write() for these nodes.
55 */
56static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
57static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
58static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
59static ssize_t write_threads(struct file *file, char *buf, size_t size);
60static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
61static ssize_t write_versions(struct file *file, char *buf, size_t size);
62static ssize_t write_ports(struct file *file, char *buf, size_t size);
63static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
64#ifdef CONFIG_NFSD_V4
65static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
66static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
67static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
68#endif
69
70static ssize_t (*write_op[])(struct file *, char *, size_t) = {
71 [NFSD_Fh] = write_filehandle,
72 [NFSD_FO_UnlockIP] = write_unlock_ip,
73 [NFSD_FO_UnlockFS] = write_unlock_fs,
74 [NFSD_Threads] = write_threads,
75 [NFSD_Pool_Threads] = write_pool_threads,
76 [NFSD_Versions] = write_versions,
77 [NFSD_Ports] = write_ports,
78 [NFSD_MaxBlkSize] = write_maxblksize,
79#ifdef CONFIG_NFSD_V4
80 [NFSD_Leasetime] = write_leasetime,
81 [NFSD_Gracetime] = write_gracetime,
82 [NFSD_RecoveryDir] = write_recoverydir,
83#endif
84};
85
86static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
87{
88 ino_t ino = file->f_path.dentry->d_inode->i_ino;
89 char *data;
90 ssize_t rv;
91
92 if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
93 return -EINVAL;
94
95 data = simple_transaction_get(file, buf, size);
96 if (IS_ERR(data))
97 return PTR_ERR(data);
98
99 rv = write_op[ino](file, data, size);
100 if (rv >= 0) {
101 simple_transaction_set(file, rv);
102 rv = size;
103 }
104 return rv;
105}
106
107static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
108{
109 if (! file->private_data) {
110 /* An attempt to read a transaction file without writing
111 * causes a 0-byte write so that the file can return
112 * state information
113 */
114 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
115 if (rv < 0)
116 return rv;
117 }
118 return simple_transaction_read(file, buf, size, pos);
119}
120
121static const struct file_operations transaction_ops = {
122 .write = nfsctl_transaction_write,
123 .read = nfsctl_transaction_read,
124 .release = simple_transaction_release,
125 .llseek = default_llseek,
126};
127
128static int exports_open(struct inode *inode, struct file *file)
129{
130 int err;
131 struct seq_file *seq;
132 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
133
134 err = seq_open(file, &nfs_exports_op);
135 if (err)
136 return err;
137
138 seq = file->private_data;
139 seq->private = nn->svc_export_cache;
140 return 0;
141}
142
143static const struct file_operations exports_operations = {
144 .open = exports_open,
145 .read = seq_read,
146 .llseek = seq_lseek,
147 .release = seq_release,
148 .owner = THIS_MODULE,
149};
150
151static int export_features_show(struct seq_file *m, void *v)
152{
153 seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
154 return 0;
155}
156
157static int export_features_open(struct inode *inode, struct file *file)
158{
159 return single_open(file, export_features_show, NULL);
160}
161
162static struct file_operations export_features_operations = {
163 .open = export_features_open,
164 .read = seq_read,
165 .llseek = seq_lseek,
166 .release = single_release,
167};
168
169#if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
170static int supported_enctypes_show(struct seq_file *m, void *v)
171{
172 seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
173 return 0;
174}
175
176static int supported_enctypes_open(struct inode *inode, struct file *file)
177{
178 return single_open(file, supported_enctypes_show, NULL);
179}
180
181static struct file_operations supported_enctypes_ops = {
182 .open = supported_enctypes_open,
183 .read = seq_read,
184 .llseek = seq_lseek,
185 .release = single_release,
186};
187#endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
188
189extern int nfsd_pool_stats_open(struct inode *inode, struct file *file);
190extern int nfsd_pool_stats_release(struct inode *inode, struct file *file);
191
192static const struct file_operations pool_stats_operations = {
193 .open = nfsd_pool_stats_open,
194 .read = seq_read,
195 .llseek = seq_lseek,
196 .release = nfsd_pool_stats_release,
197 .owner = THIS_MODULE,
198};
199
200/*----------------------------------------------------------------------------*/
201/*
202 * payload - write methods
203 */
204
205
206/**
207 * write_unlock_ip - Release all locks used by a client
208 *
209 * Experimental.
210 *
211 * Input:
212 * buf: '\n'-terminated C string containing a
213 * presentation format IP address
214 * size: length of C string in @buf
215 * Output:
216 * On success: returns zero if all specified locks were released;
217 * returns one if one or more locks were not released
218 * On error: return code is negative errno value
219 */
220static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
221{
222 struct sockaddr_storage address;
223 struct sockaddr *sap = (struct sockaddr *)&address;
224 size_t salen = sizeof(address);
225 char *fo_path;
226
227 /* sanity check */
228 if (size == 0)
229 return -EINVAL;
230
231 if (buf[size-1] != '\n')
232 return -EINVAL;
233
234 fo_path = buf;
235 if (qword_get(&buf, fo_path, size) < 0)
236 return -EINVAL;
237
238 if (rpc_pton(&init_net, fo_path, size, sap, salen) == 0)
239 return -EINVAL;
240
241 return nlmsvc_unlock_all_by_ip(sap);
242}
243
244/**
245 * write_unlock_fs - Release all locks on a local file system
246 *
247 * Experimental.
248 *
249 * Input:
250 * buf: '\n'-terminated C string containing the
251 * absolute pathname of a local file system
252 * size: length of C string in @buf
253 * Output:
254 * On success: returns zero if all specified locks were released;
255 * returns one if one or more locks were not released
256 * On error: return code is negative errno value
257 */
258static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
259{
260 struct path path;
261 char *fo_path;
262 int error;
263
264 /* sanity check */
265 if (size == 0)
266 return -EINVAL;
267
268 if (buf[size-1] != '\n')
269 return -EINVAL;
270
271 fo_path = buf;
272 if (qword_get(&buf, fo_path, size) < 0)
273 return -EINVAL;
274
275 error = kern_path(fo_path, 0, &path);
276 if (error)
277 return error;
278
279 /*
280 * XXX: Needs better sanity checking. Otherwise we could end up
281 * releasing locks on the wrong file system.
282 *
283 * For example:
284 * 1. Does the path refer to a directory?
285 * 2. Is that directory a mount point, or
286 * 3. Is that directory the root of an exported file system?
287 */
288 error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
289
290 path_put(&path);
291 return error;
292}
293
294/**
295 * write_filehandle - Get a variable-length NFS file handle by path
296 *
297 * On input, the buffer contains a '\n'-terminated C string comprised of
298 * three alphanumeric words separated by whitespace. The string may
299 * contain escape sequences.
300 *
301 * Input:
302 * buf:
303 * domain: client domain name
304 * path: export pathname
305 * maxsize: numeric maximum size of
306 * @buf
307 * size: length of C string in @buf
308 * Output:
309 * On success: passed-in buffer filled with '\n'-terminated C
310 * string containing a ASCII hex text version
311 * of the NFS file handle;
312 * return code is the size in bytes of the string
313 * On error: return code is negative errno value
314 */
315static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
316{
317 char *dname, *path;
318 int uninitialized_var(maxsize);
319 char *mesg = buf;
320 int len;
321 struct auth_domain *dom;
322 struct knfsd_fh fh;
323
324 if (size == 0)
325 return -EINVAL;
326
327 if (buf[size-1] != '\n')
328 return -EINVAL;
329 buf[size-1] = 0;
330
331 dname = mesg;
332 len = qword_get(&mesg, dname, size);
333 if (len <= 0)
334 return -EINVAL;
335
336 path = dname+len+1;
337 len = qword_get(&mesg, path, size);
338 if (len <= 0)
339 return -EINVAL;
340
341 len = get_int(&mesg, &maxsize);
342 if (len)
343 return len;
344
345 if (maxsize < NFS_FHSIZE)
346 return -EINVAL;
347 if (maxsize > NFS3_FHSIZE)
348 maxsize = NFS3_FHSIZE;
349
350 if (qword_get(&mesg, mesg, size)>0)
351 return -EINVAL;
352
353 /* we have all the words, they are in buf.. */
354 dom = unix_domain_find(dname);
355 if (!dom)
356 return -ENOMEM;
357
358 len = exp_rootfh(&init_net, dom, path, &fh, maxsize);
359 auth_domain_put(dom);
360 if (len)
361 return len;
362
363 mesg = buf;
364 len = SIMPLE_TRANSACTION_LIMIT;
365 qword_addhex(&mesg, &len, (char*)&fh.fh_base, fh.fh_size);
366 mesg[-1] = '\n';
367 return mesg - buf;
368}
369
370/**
371 * write_threads - Start NFSD, or report the current number of running threads
372 *
373 * Input:
374 * buf: ignored
375 * size: zero
376 * Output:
377 * On success: passed-in buffer filled with '\n'-terminated C
378 * string numeric value representing the number of
379 * running NFSD threads;
380 * return code is the size in bytes of the string
381 * On error: return code is zero
382 *
383 * OR
384 *
385 * Input:
386 * buf: C string containing an unsigned
387 * integer value representing the
388 * number of NFSD threads to start
389 * size: non-zero length of C string in @buf
390 * Output:
391 * On success: NFS service is started;
392 * 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 or a negative errno value
397 */
398static ssize_t write_threads(struct file *file, char *buf, size_t size)
399{
400 char *mesg = buf;
401 int rv;
402 if (size > 0) {
403 int newthreads;
404 rv = get_int(&mesg, &newthreads);
405 if (rv)
406 return rv;
407 if (newthreads < 0)
408 return -EINVAL;
409 rv = nfsd_svc(NFS_PORT, newthreads);
410 if (rv < 0)
411 return rv;
412 } else
413 rv = nfsd_nrthreads();
414
415 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
416}
417
418/**
419 * write_pool_threads - Set or report the current number of threads per pool
420 *
421 * Input:
422 * buf: ignored
423 * size: zero
424 *
425 * OR
426 *
427 * Input:
428 * buf: C string containing whitespace-
429 * separated unsigned integer values
430 * representing the number of NFSD
431 * threads to start in each pool
432 * size: non-zero length of C string in @buf
433 * Output:
434 * On success: passed-in buffer filled with '\n'-terminated C
435 * string containing integer values representing the
436 * number of NFSD threads in each pool;
437 * return code is the size in bytes of the string
438 * On error: return code is zero or a negative errno value
439 */
440static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
441{
442 /* if size > 0, look for an array of number of threads per node
443 * and apply them then write out number of threads per node as reply
444 */
445 char *mesg = buf;
446 int i;
447 int rv;
448 int len;
449 int npools;
450 int *nthreads;
451
452 mutex_lock(&nfsd_mutex);
453 npools = nfsd_nrpools();
454 if (npools == 0) {
455 /*
456 * NFS is shut down. The admin can start it by
457 * writing to the threads file but NOT the pool_threads
458 * file, sorry. Report zero threads.
459 */
460 mutex_unlock(&nfsd_mutex);
461 strcpy(buf, "0\n");
462 return strlen(buf);
463 }
464
465 nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
466 rv = -ENOMEM;
467 if (nthreads == NULL)
468 goto out_free;
469
470 if (size > 0) {
471 for (i = 0; i < npools; i++) {
472 rv = get_int(&mesg, &nthreads[i]);
473 if (rv == -ENOENT)
474 break; /* fewer numbers than pools */
475 if (rv)
476 goto out_free; /* syntax error */
477 rv = -EINVAL;
478 if (nthreads[i] < 0)
479 goto out_free;
480 }
481 rv = nfsd_set_nrthreads(i, nthreads);
482 if (rv)
483 goto out_free;
484 }
485
486 rv = nfsd_get_nrthreads(npools, nthreads);
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 __write_versions(struct file *file, char *buf, size_t size)
506{
507 char *mesg = buf;
508 char *vers, *minorp, sign;
509 int len, num, remaining;
510 unsigned minor;
511 ssize_t tlen = 0;
512 char *sep;
513
514 if (size>0) {
515 if (nfsd_serv)
516 /* Cannot change versions without updating
517 * nfsd_serv->sv_xdrsize, and reallocing
518 * rq_argp and rq_resp
519 */
520 return -EBUSY;
521 if (buf[size-1] != '\n')
522 return -EINVAL;
523 buf[size-1] = 0;
524
525 vers = mesg;
526 len = qword_get(&mesg, vers, size);
527 if (len <= 0) return -EINVAL;
528 do {
529 sign = *vers;
530 if (sign == '+' || sign == '-')
531 num = simple_strtol((vers+1), &minorp, 0);
532 else
533 num = simple_strtol(vers, &minorp, 0);
534 if (*minorp == '.') {
535 if (num < 4)
536 return -EINVAL;
537 minor = simple_strtoul(minorp+1, NULL, 0);
538 if (minor == 0)
539 return -EINVAL;
540 if (nfsd_minorversion(minor, sign == '-' ?
541 NFSD_CLEAR : NFSD_SET) < 0)
542 return -EINVAL;
543 goto next;
544 }
545 switch(num) {
546 case 2:
547 case 3:
548 case 4:
549 nfsd_vers(num, sign == '-' ? NFSD_CLEAR : NFSD_SET);
550 break;
551 default:
552 return -EINVAL;
553 }
554 next:
555 vers += len + 1;
556 } while ((len = qword_get(&mesg, vers, size)) > 0);
557 /* If all get turned off, turn them back on, as
558 * having no versions is BAD
559 */
560 nfsd_reset_versions();
561 }
562
563 /* Now write current state into reply buffer */
564 len = 0;
565 sep = "";
566 remaining = SIMPLE_TRANSACTION_LIMIT;
567 for (num=2 ; num <= 4 ; num++)
568 if (nfsd_vers(num, NFSD_AVAIL)) {
569 len = snprintf(buf, remaining, "%s%c%d", sep,
570 nfsd_vers(num, NFSD_TEST)?'+':'-',
571 num);
572 sep = " ";
573
574 if (len > remaining)
575 break;
576 remaining -= len;
577 buf += len;
578 tlen += len;
579 }
580 if (nfsd_vers(4, NFSD_AVAIL))
581 for (minor = 1; minor <= NFSD_SUPPORTED_MINOR_VERSION;
582 minor++) {
583 len = snprintf(buf, remaining, " %c4.%u",
584 (nfsd_vers(4, NFSD_TEST) &&
585 nfsd_minorversion(minor, NFSD_TEST)) ?
586 '+' : '-',
587 minor);
588
589 if (len > remaining)
590 break;
591 remaining -= len;
592 buf += len;
593 tlen += len;
594 }
595
596 len = snprintf(buf, remaining, "\n");
597 if (len > remaining)
598 return -EINVAL;
599 return tlen + len;
600}
601
602/**
603 * write_versions - Set or report the available NFS protocol versions
604 *
605 * Input:
606 * buf: ignored
607 * size: zero
608 * Output:
609 * On success: passed-in buffer filled with '\n'-terminated C
610 * string containing positive or negative integer
611 * values representing the current status of each
612 * protocol version;
613 * return code is the size in bytes of the string
614 * On error: return code is zero or a negative errno value
615 *
616 * OR
617 *
618 * Input:
619 * buf: C string containing whitespace-
620 * separated positive or negative
621 * integer values representing NFS
622 * protocol versions to enable ("+n")
623 * or disable ("-n")
624 * size: non-zero length of C string in @buf
625 * Output:
626 * On success: status of zero or more protocol versions has
627 * been updated; passed-in buffer filled with
628 * '\n'-terminated C string containing positive
629 * or negative integer values representing the
630 * current status of each protocol version;
631 * return code is the size in bytes of the string
632 * On error: return code is zero or a negative errno value
633 */
634static ssize_t write_versions(struct file *file, char *buf, size_t size)
635{
636 ssize_t rv;
637
638 mutex_lock(&nfsd_mutex);
639 rv = __write_versions(file, buf, size);
640 mutex_unlock(&nfsd_mutex);
641 return rv;
642}
643
644/*
645 * Zero-length write. Return a list of NFSD's current listener
646 * transports.
647 */
648static ssize_t __write_ports_names(char *buf)
649{
650 if (nfsd_serv == NULL)
651 return 0;
652 return svc_xprt_names(nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
653}
654
655/*
656 * A single 'fd' number was written, in which case it must be for
657 * a socket of a supported family/protocol, and we use it as an
658 * nfsd listener.
659 */
660static ssize_t __write_ports_addfd(char *buf)
661{
662 char *mesg = buf;
663 int fd, err;
664 struct net *net = &init_net;
665
666 err = get_int(&mesg, &fd);
667 if (err != 0 || fd < 0)
668 return -EINVAL;
669
670 err = nfsd_create_serv();
671 if (err != 0)
672 return err;
673
674 err = svc_addsock(nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT);
675 if (err < 0) {
676 nfsd_destroy(net);
677 return err;
678 }
679
680 /* Decrease the count, but don't shut down the service */
681 nfsd_serv->sv_nrthreads--;
682 return err;
683}
684
685/*
686 * A '-' followed by the 'name' of a socket means we close the socket.
687 */
688static ssize_t __write_ports_delfd(char *buf)
689{
690 char *toclose;
691 int len = 0;
692
693 toclose = kstrdup(buf + 1, GFP_KERNEL);
694 if (toclose == NULL)
695 return -ENOMEM;
696
697 if (nfsd_serv != NULL)
698 len = svc_sock_names(nfsd_serv, buf,
699 SIMPLE_TRANSACTION_LIMIT, toclose);
700 kfree(toclose);
701 return len;
702}
703
704/*
705 * A transport listener is added by writing it's transport name and
706 * a port number.
707 */
708static ssize_t __write_ports_addxprt(char *buf)
709{
710 char transport[16];
711 struct svc_xprt *xprt;
712 int port, err;
713 struct net *net = &init_net;
714
715 if (sscanf(buf, "%15s %4u", transport, &port) != 2)
716 return -EINVAL;
717
718 if (port < 1 || port > USHRT_MAX)
719 return -EINVAL;
720
721 err = nfsd_create_serv();
722 if (err != 0)
723 return err;
724
725 err = svc_create_xprt(nfsd_serv, transport, net,
726 PF_INET, port, SVC_SOCK_ANONYMOUS);
727 if (err < 0)
728 goto out_err;
729
730 err = svc_create_xprt(nfsd_serv, transport, net,
731 PF_INET6, port, SVC_SOCK_ANONYMOUS);
732 if (err < 0 && err != -EAFNOSUPPORT)
733 goto out_close;
734
735 /* Decrease the count, but don't shut down the service */
736 nfsd_serv->sv_nrthreads--;
737 return 0;
738out_close:
739 xprt = svc_find_xprt(nfsd_serv, transport, net, PF_INET, port);
740 if (xprt != NULL) {
741 svc_close_xprt(xprt);
742 svc_xprt_put(xprt);
743 }
744out_err:
745 nfsd_destroy(net);
746 return err;
747}
748
749/*
750 * A transport listener is removed by writing a "-", it's transport
751 * name, and it's port number.
752 */
753static ssize_t __write_ports_delxprt(char *buf)
754{
755 struct svc_xprt *xprt;
756 char transport[16];
757 int port;
758
759 if (sscanf(&buf[1], "%15s %4u", transport, &port) != 2)
760 return -EINVAL;
761
762 if (port < 1 || port > USHRT_MAX || nfsd_serv == NULL)
763 return -EINVAL;
764
765 xprt = svc_find_xprt(nfsd_serv, transport, &init_net, AF_UNSPEC, port);
766 if (xprt == NULL)
767 return -ENOTCONN;
768
769 svc_close_xprt(xprt);
770 svc_xprt_put(xprt);
771 return 0;
772}
773
774static ssize_t __write_ports(struct file *file, char *buf, size_t size)
775{
776 if (size == 0)
777 return __write_ports_names(buf);
778
779 if (isdigit(buf[0]))
780 return __write_ports_addfd(buf);
781
782 if (buf[0] == '-' && isdigit(buf[1]))
783 return __write_ports_delfd(buf);
784
785 if (isalpha(buf[0]))
786 return __write_ports_addxprt(buf);
787
788 if (buf[0] == '-' && isalpha(buf[1]))
789 return __write_ports_delxprt(buf);
790
791 return -EINVAL;
792}
793
794/**
795 * write_ports - Pass a socket file descriptor or transport name to listen on
796 *
797 * Input:
798 * buf: ignored
799 * size: zero
800 * Output:
801 * On success: passed-in buffer filled with a '\n'-terminated C
802 * string containing a whitespace-separated list of
803 * named NFSD listeners;
804 * return code is the size in bytes of the string
805 * On error: return code is zero or a negative errno value
806 *
807 * OR
808 *
809 * Input:
810 * buf: C string containing an unsigned
811 * integer value representing a bound
812 * but unconnected socket that is to be
813 * used as an NFSD listener; listen(3)
814 * must be called for a SOCK_STREAM
815 * socket, otherwise it is ignored
816 * size: non-zero length of C string in @buf
817 * Output:
818 * On success: NFS service is started;
819 * passed-in buffer filled with a '\n'-terminated C
820 * string containing a unique alphanumeric name of
821 * the listener;
822 * return code is the size in bytes of the string
823 * On error: return code is a negative errno value
824 *
825 * OR
826 *
827 * Input:
828 * buf: C string containing a "-" followed
829 * by an integer value representing a
830 * previously passed in socket file
831 * descriptor
832 * size: non-zero length of C string in @buf
833 * Output:
834 * On success: NFS service no longer listens on that socket;
835 * passed-in buffer filled with a '\n'-terminated C
836 * string containing a unique name of the listener;
837 * return code is the size in bytes of the string
838 * On error: return code is a negative errno value
839 *
840 * OR
841 *
842 * Input:
843 * buf: C string containing a transport
844 * name and an unsigned integer value
845 * representing the port to listen on,
846 * separated by whitespace
847 * size: non-zero length of C string in @buf
848 * Output:
849 * On success: returns zero; NFS service is started
850 * On error: return code is a negative errno value
851 *
852 * OR
853 *
854 * Input:
855 * buf: C string containing a "-" followed
856 * by a transport name and an unsigned
857 * integer value representing the port
858 * to listen on, separated by whitespace
859 * size: non-zero length of C string in @buf
860 * Output:
861 * On success: returns zero; NFS service no longer listens
862 * on that transport
863 * On error: return code is a negative errno value
864 */
865static ssize_t write_ports(struct file *file, char *buf, size_t size)
866{
867 ssize_t rv;
868
869 mutex_lock(&nfsd_mutex);
870 rv = __write_ports(file, buf, size);
871 mutex_unlock(&nfsd_mutex);
872 return rv;
873}
874
875
876int nfsd_max_blksize;
877
878/**
879 * write_maxblksize - Set or report the current NFS blksize
880 *
881 * Input:
882 * buf: ignored
883 * size: zero
884 *
885 * OR
886 *
887 * Input:
888 * buf: C string containing an unsigned
889 * integer value representing the new
890 * NFS blksize
891 * size: non-zero length of C string in @buf
892 * Output:
893 * On success: passed-in buffer filled with '\n'-terminated C string
894 * containing numeric value of the current NFS blksize
895 * setting;
896 * return code is the size in bytes of the string
897 * On error: return code is zero or a negative errno value
898 */
899static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
900{
901 char *mesg = buf;
902 if (size > 0) {
903 int bsize;
904 int rv = get_int(&mesg, &bsize);
905 if (rv)
906 return rv;
907 /* force bsize into allowed range and
908 * required alignment.
909 */
910 if (bsize < 1024)
911 bsize = 1024;
912 if (bsize > NFSSVC_MAXBLKSIZE)
913 bsize = NFSSVC_MAXBLKSIZE;
914 bsize &= ~(1024-1);
915 mutex_lock(&nfsd_mutex);
916 if (nfsd_serv) {
917 mutex_unlock(&nfsd_mutex);
918 return -EBUSY;
919 }
920 nfsd_max_blksize = bsize;
921 mutex_unlock(&nfsd_mutex);
922 }
923
924 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
925 nfsd_max_blksize);
926}
927
928#ifdef CONFIG_NFSD_V4
929static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size, time_t *time)
930{
931 char *mesg = buf;
932 int rv, i;
933
934 if (size > 0) {
935 if (nfsd_serv)
936 return -EBUSY;
937 rv = get_int(&mesg, &i);
938 if (rv)
939 return rv;
940 /*
941 * Some sanity checking. We don't have a reason for
942 * these particular numbers, but problems with the
943 * extremes are:
944 * - Too short: the briefest network outage may
945 * cause clients to lose all their locks. Also,
946 * the frequent polling may be wasteful.
947 * - Too long: do you really want reboot recovery
948 * to take more than an hour? Or to make other
949 * clients wait an hour before being able to
950 * revoke a dead client's locks?
951 */
952 if (i < 10 || i > 3600)
953 return -EINVAL;
954 *time = i;
955 }
956
957 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%ld\n", *time);
958}
959
960static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size, time_t *time)
961{
962 ssize_t rv;
963
964 mutex_lock(&nfsd_mutex);
965 rv = __nfsd4_write_time(file, buf, size, time);
966 mutex_unlock(&nfsd_mutex);
967 return rv;
968}
969
970/**
971 * write_leasetime - Set or report the current NFSv4 lease time
972 *
973 * Input:
974 * buf: ignored
975 * size: zero
976 *
977 * OR
978 *
979 * Input:
980 * buf: C string containing an unsigned
981 * integer value representing the new
982 * NFSv4 lease expiry time
983 * size: non-zero length of C string in @buf
984 * Output:
985 * On success: passed-in buffer filled with '\n'-terminated C
986 * string containing unsigned integer value of the
987 * current lease expiry time;
988 * return code is the size in bytes of the string
989 * On error: return code is zero or a negative errno value
990 */
991static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
992{
993 return nfsd4_write_time(file, buf, size, &nfsd4_lease);
994}
995
996/**
997 * write_gracetime - Set or report current NFSv4 grace period time
998 *
999 * As above, but sets the time of the NFSv4 grace period.
1000 *
1001 * Note this should never be set to less than the *previous*
1002 * lease-period time, but we don't try to enforce this. (In the common
1003 * case (a new boot), we don't know what the previous lease time was
1004 * anyway.)
1005 */
1006static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1007{
1008 return nfsd4_write_time(file, buf, size, &nfsd4_grace);
1009}
1010
1011extern char *nfs4_recoverydir(void);
1012
1013static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size)
1014{
1015 char *mesg = buf;
1016 char *recdir;
1017 int len, status;
1018
1019 if (size > 0) {
1020 if (nfsd_serv)
1021 return -EBUSY;
1022 if (size > PATH_MAX || buf[size-1] != '\n')
1023 return -EINVAL;
1024 buf[size-1] = 0;
1025
1026 recdir = mesg;
1027 len = qword_get(&mesg, recdir, size);
1028 if (len <= 0)
1029 return -EINVAL;
1030
1031 status = nfs4_reset_recoverydir(recdir);
1032 if (status)
1033 return status;
1034 }
1035
1036 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1037 nfs4_recoverydir());
1038}
1039
1040/**
1041 * write_recoverydir - Set or report the pathname of the recovery directory
1042 *
1043 * Input:
1044 * buf: ignored
1045 * size: zero
1046 *
1047 * OR
1048 *
1049 * Input:
1050 * buf: C string containing the pathname
1051 * of the directory on a local file
1052 * system containing permanent NFSv4
1053 * recovery data
1054 * size: non-zero length of C string in @buf
1055 * Output:
1056 * On success: passed-in buffer filled with '\n'-terminated C string
1057 * containing the current recovery pathname setting;
1058 * return code is the size in bytes of the string
1059 * On error: return code is zero or a negative errno value
1060 */
1061static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1062{
1063 ssize_t rv;
1064
1065 mutex_lock(&nfsd_mutex);
1066 rv = __write_recoverydir(file, buf, size);
1067 mutex_unlock(&nfsd_mutex);
1068 return rv;
1069}
1070
1071#endif
1072
1073/*----------------------------------------------------------------------------*/
1074/*
1075 * populating the filesystem.
1076 */
1077
1078static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
1079{
1080 static struct tree_descr nfsd_files[] = {
1081 [NFSD_List] = {"exports", &exports_operations, S_IRUGO},
1082 [NFSD_Export_features] = {"export_features",
1083 &export_features_operations, S_IRUGO},
1084 [NFSD_FO_UnlockIP] = {"unlock_ip",
1085 &transaction_ops, S_IWUSR|S_IRUSR},
1086 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1087 &transaction_ops, S_IWUSR|S_IRUSR},
1088 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1089 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1090 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1091 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1092 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1093 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1094 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1095#if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1096 [NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1097#endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1098#ifdef CONFIG_NFSD_V4
1099 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1100 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1101 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1102#endif
1103 /* last one */ {""}
1104 };
1105 return simple_fill_super(sb, 0x6e667364, nfsd_files);
1106}
1107
1108static struct dentry *nfsd_mount(struct file_system_type *fs_type,
1109 int flags, const char *dev_name, void *data)
1110{
1111 return mount_single(fs_type, flags, data, nfsd_fill_super);
1112}
1113
1114static struct file_system_type nfsd_fs_type = {
1115 .owner = THIS_MODULE,
1116 .name = "nfsd",
1117 .mount = nfsd_mount,
1118 .kill_sb = kill_litter_super,
1119};
1120
1121#ifdef CONFIG_PROC_FS
1122static int create_proc_exports_entry(void)
1123{
1124 struct proc_dir_entry *entry;
1125
1126 entry = proc_mkdir("fs/nfs", NULL);
1127 if (!entry)
1128 return -ENOMEM;
1129 entry = proc_create("exports", 0, entry, &exports_operations);
1130 if (!entry)
1131 return -ENOMEM;
1132 return 0;
1133}
1134#else /* CONFIG_PROC_FS */
1135static int create_proc_exports_entry(void)
1136{
1137 return 0;
1138}
1139#endif
1140
1141int nfsd_net_id;
1142
1143static __net_init int nfsd_init_net(struct net *net)
1144{
1145 int retval;
1146
1147 retval = nfsd_export_init(net);
1148 if (retval)
1149 goto out_export_error;
1150 retval = nfsd_idmap_init(net);
1151 if (retval)
1152 goto out_idmap_error;
1153 return 0;
1154
1155out_idmap_error:
1156 nfsd_export_shutdown(net);
1157out_export_error:
1158 return retval;
1159}
1160
1161static __net_exit void nfsd_exit_net(struct net *net)
1162{
1163 nfsd_idmap_shutdown(net);
1164 nfsd_export_shutdown(net);
1165}
1166
1167static struct pernet_operations nfsd_net_ops = {
1168 .init = nfsd_init_net,
1169 .exit = nfsd_exit_net,
1170 .id = &nfsd_net_id,
1171 .size = sizeof(struct nfsd_net),
1172};
1173
1174static int __init init_nfsd(void)
1175{
1176 int retval;
1177 printk(KERN_INFO "Installing knfsd (copyright (C) 1996 okir@monad.swb.de).\n");
1178
1179 retval = register_cld_notifier();
1180 if (retval)
1181 return retval;
1182 retval = register_pernet_subsys(&nfsd_net_ops);
1183 if (retval < 0)
1184 goto out_unregister_notifier;
1185 retval = nfsd4_init_slabs();
1186 if (retval)
1187 goto out_unregister_pernet;
1188 nfs4_state_init();
1189 retval = nfsd_fault_inject_init(); /* nfsd fault injection controls */
1190 if (retval)
1191 goto out_free_slabs;
1192 nfsd_stat_init(); /* Statistics */
1193 retval = nfsd_reply_cache_init();
1194 if (retval)
1195 goto out_free_stat;
1196 nfsd_lockd_init(); /* lockd->nfsd callbacks */
1197 retval = create_proc_exports_entry();
1198 if (retval)
1199 goto out_free_lockd;
1200 retval = register_filesystem(&nfsd_fs_type);
1201 if (retval)
1202 goto out_free_all;
1203 return 0;
1204out_free_all:
1205 remove_proc_entry("fs/nfs/exports", NULL);
1206 remove_proc_entry("fs/nfs", NULL);
1207out_free_lockd:
1208 nfsd_lockd_shutdown();
1209 nfsd_reply_cache_shutdown();
1210out_free_stat:
1211 nfsd_stat_shutdown();
1212 nfsd_fault_inject_cleanup();
1213out_free_slabs:
1214 nfsd4_free_slabs();
1215out_unregister_pernet:
1216 unregister_pernet_subsys(&nfsd_net_ops);
1217out_unregister_notifier:
1218 unregister_cld_notifier();
1219 return retval;
1220}
1221
1222static void __exit exit_nfsd(void)
1223{
1224 nfsd_reply_cache_shutdown();
1225 remove_proc_entry("fs/nfs/exports", NULL);
1226 remove_proc_entry("fs/nfs", NULL);
1227 nfsd_stat_shutdown();
1228 nfsd_lockd_shutdown();
1229 nfsd4_free_slabs();
1230 nfsd_fault_inject_cleanup();
1231 unregister_filesystem(&nfsd_fs_type);
1232 unregister_pernet_subsys(&nfsd_net_ops);
1233 unregister_cld_notifier();
1234}
1235
1236MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1237MODULE_LICENSE("GPL");
1238module_init(init_nfsd)
1239module_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/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
29/*
30 * We have a single directory with several nodes in it.
31 */
32enum {
33 NFSD_Root = 1,
34 NFSD_List,
35 NFSD_Export_features,
36 NFSD_Fh,
37 NFSD_FO_UnlockIP,
38 NFSD_FO_UnlockFS,
39 NFSD_Threads,
40 NFSD_Pool_Threads,
41 NFSD_Pool_Stats,
42 NFSD_Reply_Cache_Stats,
43 NFSD_Versions,
44 NFSD_Ports,
45 NFSD_MaxBlkSize,
46 NFSD_MaxConnections,
47 NFSD_SupportedEnctypes,
48 /*
49 * The below MUST come last. Otherwise we leave a hole in nfsd_files[]
50 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
51 */
52#ifdef CONFIG_NFSD_V4
53 NFSD_Leasetime,
54 NFSD_Gracetime,
55 NFSD_RecoveryDir,
56 NFSD_V4EndGrace,
57#endif
58 NFSD_MaxReserved
59};
60
61/*
62 * write() for these nodes.
63 */
64static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
65static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
66static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
67static ssize_t write_threads(struct file *file, char *buf, size_t size);
68static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
69static ssize_t write_versions(struct file *file, char *buf, size_t size);
70static ssize_t write_ports(struct file *file, char *buf, size_t size);
71static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
72static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
73#ifdef CONFIG_NFSD_V4
74static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
75static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
76static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
77static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
78#endif
79
80static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
81 [NFSD_Fh] = write_filehandle,
82 [NFSD_FO_UnlockIP] = write_unlock_ip,
83 [NFSD_FO_UnlockFS] = write_unlock_fs,
84 [NFSD_Threads] = write_threads,
85 [NFSD_Pool_Threads] = write_pool_threads,
86 [NFSD_Versions] = write_versions,
87 [NFSD_Ports] = write_ports,
88 [NFSD_MaxBlkSize] = write_maxblksize,
89 [NFSD_MaxConnections] = write_maxconn,
90#ifdef CONFIG_NFSD_V4
91 [NFSD_Leasetime] = write_leasetime,
92 [NFSD_Gracetime] = write_gracetime,
93 [NFSD_RecoveryDir] = write_recoverydir,
94 [NFSD_V4EndGrace] = write_v4_end_grace,
95#endif
96};
97
98static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
99{
100 ino_t ino = file_inode(file)->i_ino;
101 char *data;
102 ssize_t rv;
103
104 if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
105 return -EINVAL;
106
107 data = simple_transaction_get(file, buf, size);
108 if (IS_ERR(data))
109 return PTR_ERR(data);
110
111 rv = write_op[ino](file, data, size);
112 if (rv >= 0) {
113 simple_transaction_set(file, rv);
114 rv = size;
115 }
116 return rv;
117}
118
119static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
120{
121 if (! file->private_data) {
122 /* An attempt to read a transaction file without writing
123 * causes a 0-byte write so that the file can return
124 * state information
125 */
126 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
127 if (rv < 0)
128 return rv;
129 }
130 return simple_transaction_read(file, buf, size, pos);
131}
132
133static const struct file_operations transaction_ops = {
134 .write = nfsctl_transaction_write,
135 .read = nfsctl_transaction_read,
136 .release = simple_transaction_release,
137 .llseek = default_llseek,
138};
139
140static int exports_net_open(struct net *net, struct file *file)
141{
142 int err;
143 struct seq_file *seq;
144 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
145
146 err = seq_open(file, &nfs_exports_op);
147 if (err)
148 return err;
149
150 seq = file->private_data;
151 seq->private = nn->svc_export_cache;
152 return 0;
153}
154
155static int exports_proc_open(struct inode *inode, struct file *file)
156{
157 return exports_net_open(current->nsproxy->net_ns, file);
158}
159
160static const struct file_operations exports_proc_operations = {
161 .open = exports_proc_open,
162 .read = seq_read,
163 .llseek = seq_lseek,
164 .release = seq_release,
165};
166
167static int exports_nfsd_open(struct inode *inode, struct file *file)
168{
169 return exports_net_open(inode->i_sb->s_fs_info, file);
170}
171
172static const struct file_operations exports_nfsd_operations = {
173 .open = exports_nfsd_open,
174 .read = seq_read,
175 .llseek = seq_lseek,
176 .release = seq_release,
177};
178
179static int export_features_show(struct seq_file *m, void *v)
180{
181 seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
182 return 0;
183}
184
185static int export_features_open(struct inode *inode, struct file *file)
186{
187 return single_open(file, export_features_show, NULL);
188}
189
190static const struct file_operations export_features_operations = {
191 .open = export_features_open,
192 .read = seq_read,
193 .llseek = seq_lseek,
194 .release = single_release,
195};
196
197#if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
198static int supported_enctypes_show(struct seq_file *m, void *v)
199{
200 seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
201 return 0;
202}
203
204static int supported_enctypes_open(struct inode *inode, struct file *file)
205{
206 return single_open(file, supported_enctypes_show, NULL);
207}
208
209static const struct file_operations supported_enctypes_ops = {
210 .open = supported_enctypes_open,
211 .read = seq_read,
212 .llseek = seq_lseek,
213 .release = single_release,
214};
215#endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
216
217static const struct file_operations pool_stats_operations = {
218 .open = nfsd_pool_stats_open,
219 .read = seq_read,
220 .llseek = seq_lseek,
221 .release = nfsd_pool_stats_release,
222};
223
224static const struct file_operations reply_cache_stats_operations = {
225 .open = nfsd_reply_cache_stats_open,
226 .read = seq_read,
227 .llseek = seq_lseek,
228 .release = single_release,
229};
230
231/*----------------------------------------------------------------------------*/
232/*
233 * payload - write methods
234 */
235
236static inline struct net *netns(struct file *file)
237{
238 return file_inode(file)->i_sb->s_fs_info;
239}
240
241/**
242 * write_unlock_ip - Release all locks used by a client
243 *
244 * Experimental.
245 *
246 * Input:
247 * buf: '\n'-terminated C string containing a
248 * presentation format IP address
249 * size: length of C string in @buf
250 * Output:
251 * On success: returns zero if all specified locks were released;
252 * returns one if one or more locks were not released
253 * On error: return code is negative errno value
254 */
255static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
256{
257 struct sockaddr_storage address;
258 struct sockaddr *sap = (struct sockaddr *)&address;
259 size_t salen = sizeof(address);
260 char *fo_path;
261 struct net *net = netns(file);
262
263 /* sanity check */
264 if (size == 0)
265 return -EINVAL;
266
267 if (buf[size-1] != '\n')
268 return -EINVAL;
269
270 fo_path = buf;
271 if (qword_get(&buf, fo_path, size) < 0)
272 return -EINVAL;
273
274 if (rpc_pton(net, fo_path, size, sap, salen) == 0)
275 return -EINVAL;
276
277 return nlmsvc_unlock_all_by_ip(sap);
278}
279
280/**
281 * write_unlock_fs - Release all locks on a local file system
282 *
283 * Experimental.
284 *
285 * Input:
286 * buf: '\n'-terminated C string containing the
287 * absolute pathname of a local file system
288 * size: length of C string in @buf
289 * Output:
290 * On success: returns zero if all specified locks were released;
291 * returns one if one or more locks were not released
292 * On error: return code is negative errno value
293 */
294static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
295{
296 struct path path;
297 char *fo_path;
298 int error;
299
300 /* sanity check */
301 if (size == 0)
302 return -EINVAL;
303
304 if (buf[size-1] != '\n')
305 return -EINVAL;
306
307 fo_path = buf;
308 if (qword_get(&buf, fo_path, size) < 0)
309 return -EINVAL;
310
311 error = kern_path(fo_path, 0, &path);
312 if (error)
313 return error;
314
315 /*
316 * XXX: Needs better sanity checking. Otherwise we could end up
317 * releasing locks on the wrong file system.
318 *
319 * For example:
320 * 1. Does the path refer to a directory?
321 * 2. Is that directory a mount point, or
322 * 3. Is that directory the root of an exported file system?
323 */
324 error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
325
326 path_put(&path);
327 return error;
328}
329
330/**
331 * write_filehandle - Get a variable-length NFS file handle by path
332 *
333 * On input, the buffer contains a '\n'-terminated C string comprised of
334 * three alphanumeric words separated by whitespace. The string may
335 * contain escape sequences.
336 *
337 * Input:
338 * buf:
339 * domain: client domain name
340 * path: export pathname
341 * maxsize: numeric maximum size of
342 * @buf
343 * size: length of C string in @buf
344 * Output:
345 * On success: passed-in buffer filled with '\n'-terminated C
346 * string containing a ASCII hex text version
347 * of the NFS file handle;
348 * return code is the size in bytes of the string
349 * On error: return code is negative errno value
350 */
351static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
352{
353 char *dname, *path;
354 int uninitialized_var(maxsize);
355 char *mesg = buf;
356 int len;
357 struct auth_domain *dom;
358 struct knfsd_fh fh;
359
360 if (size == 0)
361 return -EINVAL;
362
363 if (buf[size-1] != '\n')
364 return -EINVAL;
365 buf[size-1] = 0;
366
367 dname = mesg;
368 len = qword_get(&mesg, dname, size);
369 if (len <= 0)
370 return -EINVAL;
371
372 path = dname+len+1;
373 len = qword_get(&mesg, path, size);
374 if (len <= 0)
375 return -EINVAL;
376
377 len = get_int(&mesg, &maxsize);
378 if (len)
379 return len;
380
381 if (maxsize < NFS_FHSIZE)
382 return -EINVAL;
383 maxsize = min(maxsize, NFS3_FHSIZE);
384
385 if (qword_get(&mesg, mesg, size)>0)
386 return -EINVAL;
387
388 /* we have all the words, they are in buf.. */
389 dom = unix_domain_find(dname);
390 if (!dom)
391 return -ENOMEM;
392
393 len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
394 auth_domain_put(dom);
395 if (len)
396 return len;
397
398 mesg = buf;
399 len = SIMPLE_TRANSACTION_LIMIT;
400 qword_addhex(&mesg, &len, (char*)&fh.fh_base, fh.fh_size);
401 mesg[-1] = '\n';
402 return mesg - buf;
403}
404
405/**
406 * write_threads - Start NFSD, or report the current number of running threads
407 *
408 * Input:
409 * buf: ignored
410 * size: zero
411 * Output:
412 * On success: passed-in buffer filled with '\n'-terminated C
413 * string numeric value representing the number of
414 * running NFSD threads;
415 * return code is the size in bytes of the string
416 * On error: return code is zero
417 *
418 * OR
419 *
420 * Input:
421 * buf: C string containing an unsigned
422 * integer value representing the
423 * number of NFSD threads to start
424 * size: non-zero length of C string in @buf
425 * Output:
426 * On success: NFS service is started;
427 * passed-in buffer filled with '\n'-terminated C
428 * string numeric value representing the number of
429 * running NFSD threads;
430 * return code is the size in bytes of the string
431 * On error: return code is zero or a negative errno value
432 */
433static ssize_t write_threads(struct file *file, char *buf, size_t size)
434{
435 char *mesg = buf;
436 int rv;
437 struct net *net = netns(file);
438
439 if (size > 0) {
440 int newthreads;
441 rv = get_int(&mesg, &newthreads);
442 if (rv)
443 return rv;
444 if (newthreads < 0)
445 return -EINVAL;
446 rv = nfsd_svc(newthreads, net, file->f_cred);
447 if (rv < 0)
448 return rv;
449 } else
450 rv = nfsd_nrthreads(net);
451
452 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
453}
454
455/**
456 * write_pool_threads - Set or report the current number of threads per pool
457 *
458 * Input:
459 * buf: ignored
460 * size: zero
461 *
462 * OR
463 *
464 * Input:
465 * buf: C string containing whitespace-
466 * separated unsigned integer values
467 * representing the number of NFSD
468 * threads to start in each pool
469 * size: non-zero length of C string in @buf
470 * Output:
471 * On success: passed-in buffer filled with '\n'-terminated C
472 * string containing integer values representing the
473 * number of NFSD threads in each pool;
474 * return code is the size in bytes of the string
475 * On error: return code is zero or a negative errno value
476 */
477static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
478{
479 /* if size > 0, look for an array of number of threads per node
480 * and apply them then write out number of threads per node as reply
481 */
482 char *mesg = buf;
483 int i;
484 int rv;
485 int len;
486 int npools;
487 int *nthreads;
488 struct net *net = netns(file);
489
490 mutex_lock(&nfsd_mutex);
491 npools = nfsd_nrpools(net);
492 if (npools == 0) {
493 /*
494 * NFS is shut down. The admin can start it by
495 * writing to the threads file but NOT the pool_threads
496 * file, sorry. Report zero threads.
497 */
498 mutex_unlock(&nfsd_mutex);
499 strcpy(buf, "0\n");
500 return strlen(buf);
501 }
502
503 nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
504 rv = -ENOMEM;
505 if (nthreads == NULL)
506 goto out_free;
507
508 if (size > 0) {
509 for (i = 0; i < npools; i++) {
510 rv = get_int(&mesg, &nthreads[i]);
511 if (rv == -ENOENT)
512 break; /* fewer numbers than pools */
513 if (rv)
514 goto out_free; /* syntax error */
515 rv = -EINVAL;
516 if (nthreads[i] < 0)
517 goto out_free;
518 }
519 rv = nfsd_set_nrthreads(i, nthreads, net);
520 if (rv)
521 goto out_free;
522 }
523
524 rv = nfsd_get_nrthreads(npools, nthreads, net);
525 if (rv)
526 goto out_free;
527
528 mesg = buf;
529 size = SIMPLE_TRANSACTION_LIMIT;
530 for (i = 0; i < npools && size > 0; i++) {
531 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
532 len = strlen(mesg);
533 size -= len;
534 mesg += len;
535 }
536 rv = mesg - buf;
537out_free:
538 kfree(nthreads);
539 mutex_unlock(&nfsd_mutex);
540 return rv;
541}
542
543static ssize_t
544nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
545 const char *sep, unsigned vers, int minor)
546{
547 const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
548 bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
549
550 if (vers == 4 && minor >= 0 &&
551 !nfsd_minorversion(nn, minor, NFSD_TEST))
552 supported = false;
553 if (minor == 0 && supported)
554 /*
555 * special case for backward compatability.
556 * +4.0 is never reported, it is implied by
557 * +4, unless -4.0 is present.
558 */
559 return 0;
560 return snprintf(buf, remaining, format, sep,
561 supported ? '+' : '-', vers, minor);
562}
563
564static ssize_t __write_versions(struct file *file, char *buf, size_t size)
565{
566 char *mesg = buf;
567 char *vers, *minorp, sign;
568 int len, num, remaining;
569 ssize_t tlen = 0;
570 char *sep;
571 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
572
573 if (size>0) {
574 if (nn->nfsd_serv)
575 /* Cannot change versions without updating
576 * nn->nfsd_serv->sv_xdrsize, and reallocing
577 * rq_argp and rq_resp
578 */
579 return -EBUSY;
580 if (buf[size-1] != '\n')
581 return -EINVAL;
582 buf[size-1] = 0;
583
584 vers = mesg;
585 len = qword_get(&mesg, vers, size);
586 if (len <= 0) return -EINVAL;
587 do {
588 enum vers_op cmd;
589 unsigned minor;
590 sign = *vers;
591 if (sign == '+' || sign == '-')
592 num = simple_strtol((vers+1), &minorp, 0);
593 else
594 num = simple_strtol(vers, &minorp, 0);
595 if (*minorp == '.') {
596 if (num != 4)
597 return -EINVAL;
598 if (kstrtouint(minorp+1, 0, &minor) < 0)
599 return -EINVAL;
600 }
601
602 cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
603 switch(num) {
604 case 2:
605 case 3:
606 nfsd_vers(nn, num, cmd);
607 break;
608 case 4:
609 if (*minorp == '.') {
610 if (nfsd_minorversion(nn, minor, cmd) < 0)
611 return -EINVAL;
612 } else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
613 /*
614 * Either we have +4 and no minors are enabled,
615 * or we have -4 and at least one minor is enabled.
616 * In either case, propagate 'cmd' to all minors.
617 */
618 minor = 0;
619 while (nfsd_minorversion(nn, minor, cmd) >= 0)
620 minor++;
621 }
622 break;
623 default:
624 return -EINVAL;
625 }
626 vers += len + 1;
627 } while ((len = qword_get(&mesg, vers, size)) > 0);
628 /* If all get turned off, turn them back on, as
629 * having no versions is BAD
630 */
631 nfsd_reset_versions(nn);
632 }
633
634 /* Now write current state into reply buffer */
635 len = 0;
636 sep = "";
637 remaining = SIMPLE_TRANSACTION_LIMIT;
638 for (num=2 ; num <= 4 ; num++) {
639 int minor;
640 if (!nfsd_vers(nn, num, NFSD_AVAIL))
641 continue;
642
643 minor = -1;
644 do {
645 len = nfsd_print_version_support(nn, buf, remaining,
646 sep, num, minor);
647 if (len >= remaining)
648 goto out;
649 remaining -= len;
650 buf += len;
651 tlen += len;
652 minor++;
653 if (len)
654 sep = " ";
655 } while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
656 }
657out:
658 len = snprintf(buf, remaining, "\n");
659 if (len >= remaining)
660 return -EINVAL;
661 return tlen + len;
662}
663
664/**
665 * write_versions - Set or report the available NFS protocol versions
666 *
667 * Input:
668 * buf: ignored
669 * size: zero
670 * Output:
671 * On success: passed-in buffer filled with '\n'-terminated C
672 * string containing positive or negative integer
673 * values representing the current status of each
674 * protocol version;
675 * return code is the size in bytes of the string
676 * On error: return code is zero or a negative errno value
677 *
678 * OR
679 *
680 * Input:
681 * buf: C string containing whitespace-
682 * separated positive or negative
683 * integer values representing NFS
684 * protocol versions to enable ("+n")
685 * or disable ("-n")
686 * size: non-zero length of C string in @buf
687 * Output:
688 * On success: status of zero or more protocol versions has
689 * been updated; passed-in buffer filled with
690 * '\n'-terminated C string containing positive
691 * or negative integer values representing the
692 * current status of each protocol version;
693 * return code is the size in bytes of the string
694 * On error: return code is zero or a negative errno value
695 */
696static ssize_t write_versions(struct file *file, char *buf, size_t size)
697{
698 ssize_t rv;
699
700 mutex_lock(&nfsd_mutex);
701 rv = __write_versions(file, buf, size);
702 mutex_unlock(&nfsd_mutex);
703 return rv;
704}
705
706/*
707 * Zero-length write. Return a list of NFSD's current listener
708 * transports.
709 */
710static ssize_t __write_ports_names(char *buf, struct net *net)
711{
712 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
713
714 if (nn->nfsd_serv == NULL)
715 return 0;
716 return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
717}
718
719/*
720 * A single 'fd' number was written, in which case it must be for
721 * a socket of a supported family/protocol, and we use it as an
722 * nfsd listener.
723 */
724static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
725{
726 char *mesg = buf;
727 int fd, err;
728 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
729
730 err = get_int(&mesg, &fd);
731 if (err != 0 || fd < 0)
732 return -EINVAL;
733
734 if (svc_alien_sock(net, fd)) {
735 printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
736 return -EINVAL;
737 }
738
739 err = nfsd_create_serv(net);
740 if (err != 0)
741 return err;
742
743 err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
744 if (err < 0) {
745 nfsd_destroy(net);
746 return err;
747 }
748
749 /* Decrease the count, but don't shut down the service */
750 nn->nfsd_serv->sv_nrthreads--;
751 return err;
752}
753
754/*
755 * A transport listener is added by writing it's transport name and
756 * a port number.
757 */
758static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
759{
760 char transport[16];
761 struct svc_xprt *xprt;
762 int port, err;
763 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
764
765 if (sscanf(buf, "%15s %5u", transport, &port) != 2)
766 return -EINVAL;
767
768 if (port < 1 || port > USHRT_MAX)
769 return -EINVAL;
770
771 err = nfsd_create_serv(net);
772 if (err != 0)
773 return err;
774
775 err = svc_create_xprt(nn->nfsd_serv, transport, net,
776 PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
777 if (err < 0)
778 goto out_err;
779
780 err = svc_create_xprt(nn->nfsd_serv, transport, net,
781 PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
782 if (err < 0 && err != -EAFNOSUPPORT)
783 goto out_close;
784
785 /* Decrease the count, but don't shut down the service */
786 nn->nfsd_serv->sv_nrthreads--;
787 return 0;
788out_close:
789 xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
790 if (xprt != NULL) {
791 svc_close_xprt(xprt);
792 svc_xprt_put(xprt);
793 }
794out_err:
795 nfsd_destroy(net);
796 return err;
797}
798
799static ssize_t __write_ports(struct file *file, char *buf, size_t size,
800 struct net *net)
801{
802 if (size == 0)
803 return __write_ports_names(buf, net);
804
805 if (isdigit(buf[0]))
806 return __write_ports_addfd(buf, net, file->f_cred);
807
808 if (isalpha(buf[0]))
809 return __write_ports_addxprt(buf, net, file->f_cred);
810
811 return -EINVAL;
812}
813
814/**
815 * write_ports - Pass a socket file descriptor or transport name to listen on
816 *
817 * Input:
818 * buf: ignored
819 * size: zero
820 * Output:
821 * On success: passed-in buffer filled with a '\n'-terminated C
822 * string containing a whitespace-separated list of
823 * named NFSD listeners;
824 * return code is the size in bytes of the string
825 * On error: return code is zero or a negative errno value
826 *
827 * OR
828 *
829 * Input:
830 * buf: C string containing an unsigned
831 * integer value representing a bound
832 * but unconnected socket that is to be
833 * used as an NFSD listener; listen(3)
834 * must be called for a SOCK_STREAM
835 * socket, otherwise it is ignored
836 * size: non-zero length of C string in @buf
837 * Output:
838 * On success: NFS service is started;
839 * passed-in buffer filled with a '\n'-terminated C
840 * string containing a unique alphanumeric name of
841 * the listener;
842 * return code is the size in bytes of the string
843 * On error: return code is a negative errno value
844 *
845 * OR
846 *
847 * Input:
848 * buf: C string containing a transport
849 * name and an unsigned integer value
850 * representing the port to listen on,
851 * separated by whitespace
852 * size: non-zero length of C string in @buf
853 * Output:
854 * On success: returns zero; NFS service is started
855 * On error: return code is a negative errno value
856 */
857static ssize_t write_ports(struct file *file, char *buf, size_t size)
858{
859 ssize_t rv;
860
861 mutex_lock(&nfsd_mutex);
862 rv = __write_ports(file, buf, size, netns(file));
863 mutex_unlock(&nfsd_mutex);
864 return rv;
865}
866
867
868int nfsd_max_blksize;
869
870/**
871 * write_maxblksize - Set or report the current NFS blksize
872 *
873 * Input:
874 * buf: ignored
875 * size: zero
876 *
877 * OR
878 *
879 * Input:
880 * buf: C string containing an unsigned
881 * integer value representing the new
882 * NFS blksize
883 * size: non-zero length of C string in @buf
884 * Output:
885 * On success: passed-in buffer filled with '\n'-terminated C string
886 * containing numeric value of the current NFS blksize
887 * setting;
888 * return code is the size in bytes of the string
889 * On error: return code is zero or a negative errno value
890 */
891static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
892{
893 char *mesg = buf;
894 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
895
896 if (size > 0) {
897 int bsize;
898 int rv = get_int(&mesg, &bsize);
899 if (rv)
900 return rv;
901 /* force bsize into allowed range and
902 * required alignment.
903 */
904 bsize = max_t(int, bsize, 1024);
905 bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
906 bsize &= ~(1024-1);
907 mutex_lock(&nfsd_mutex);
908 if (nn->nfsd_serv) {
909 mutex_unlock(&nfsd_mutex);
910 return -EBUSY;
911 }
912 nfsd_max_blksize = bsize;
913 mutex_unlock(&nfsd_mutex);
914 }
915
916 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
917 nfsd_max_blksize);
918}
919
920/**
921 * write_maxconn - Set or report the current max number of connections
922 *
923 * Input:
924 * buf: ignored
925 * size: zero
926 * OR
927 *
928 * Input:
929 * buf: C string containing an unsigned
930 * integer value representing the new
931 * number of max connections
932 * size: non-zero length of C string in @buf
933 * Output:
934 * On success: passed-in buffer filled with '\n'-terminated C string
935 * containing numeric value of max_connections setting
936 * for this net namespace;
937 * return code is the size in bytes of the string
938 * On error: return code is zero or a negative errno value
939 */
940static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
941{
942 char *mesg = buf;
943 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
944 unsigned int maxconn = nn->max_connections;
945
946 if (size > 0) {
947 int rv = get_uint(&mesg, &maxconn);
948
949 if (rv)
950 return rv;
951 nn->max_connections = maxconn;
952 }
953
954 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
955}
956
957#ifdef CONFIG_NFSD_V4
958static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
959 time_t *time, struct nfsd_net *nn)
960{
961 char *mesg = buf;
962 int rv, i;
963
964 if (size > 0) {
965 if (nn->nfsd_serv)
966 return -EBUSY;
967 rv = get_int(&mesg, &i);
968 if (rv)
969 return rv;
970 /*
971 * Some sanity checking. We don't have a reason for
972 * these particular numbers, but problems with the
973 * extremes are:
974 * - Too short: the briefest network outage may
975 * cause clients to lose all their locks. Also,
976 * the frequent polling may be wasteful.
977 * - Too long: do you really want reboot recovery
978 * to take more than an hour? Or to make other
979 * clients wait an hour before being able to
980 * revoke a dead client's locks?
981 */
982 if (i < 10 || i > 3600)
983 return -EINVAL;
984 *time = i;
985 }
986
987 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%ld\n", *time);
988}
989
990static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
991 time_t *time, struct nfsd_net *nn)
992{
993 ssize_t rv;
994
995 mutex_lock(&nfsd_mutex);
996 rv = __nfsd4_write_time(file, buf, size, time, nn);
997 mutex_unlock(&nfsd_mutex);
998 return rv;
999}
1000
1001/**
1002 * write_leasetime - Set or report the current NFSv4 lease time
1003 *
1004 * Input:
1005 * buf: ignored
1006 * size: zero
1007 *
1008 * OR
1009 *
1010 * Input:
1011 * buf: C string containing an unsigned
1012 * integer value representing the new
1013 * NFSv4 lease expiry time
1014 * size: non-zero length of C string in @buf
1015 * Output:
1016 * On success: passed-in buffer filled with '\n'-terminated C
1017 * string containing unsigned integer value of the
1018 * current lease expiry time;
1019 * return code is the size in bytes of the string
1020 * On error: return code is zero or a negative errno value
1021 */
1022static ssize_t write_leasetime(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_lease, nn);
1026}
1027
1028/**
1029 * write_gracetime - Set or report current NFSv4 grace period time
1030 *
1031 * As above, but sets the time of the NFSv4 grace period.
1032 *
1033 * Note this should never be set to less than the *previous*
1034 * lease-period time, but we don't try to enforce this. (In the common
1035 * case (a new boot), we don't know what the previous lease time was
1036 * anyway.)
1037 */
1038static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1039{
1040 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1041 return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1042}
1043
1044static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1045 struct nfsd_net *nn)
1046{
1047 char *mesg = buf;
1048 char *recdir;
1049 int len, status;
1050
1051 if (size > 0) {
1052 if (nn->nfsd_serv)
1053 return -EBUSY;
1054 if (size > PATH_MAX || buf[size-1] != '\n')
1055 return -EINVAL;
1056 buf[size-1] = 0;
1057
1058 recdir = mesg;
1059 len = qword_get(&mesg, recdir, size);
1060 if (len <= 0)
1061 return -EINVAL;
1062
1063 status = nfs4_reset_recoverydir(recdir);
1064 if (status)
1065 return status;
1066 }
1067
1068 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1069 nfs4_recoverydir());
1070}
1071
1072/**
1073 * write_recoverydir - Set or report the pathname of the recovery directory
1074 *
1075 * Input:
1076 * buf: ignored
1077 * size: zero
1078 *
1079 * OR
1080 *
1081 * Input:
1082 * buf: C string containing the pathname
1083 * of the directory on a local file
1084 * system containing permanent NFSv4
1085 * recovery data
1086 * size: non-zero length of C string in @buf
1087 * Output:
1088 * On success: passed-in buffer filled with '\n'-terminated C string
1089 * containing the current recovery pathname setting;
1090 * return code is the size in bytes of the string
1091 * On error: return code is zero or a negative errno value
1092 */
1093static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1094{
1095 ssize_t rv;
1096 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1097
1098 mutex_lock(&nfsd_mutex);
1099 rv = __write_recoverydir(file, buf, size, nn);
1100 mutex_unlock(&nfsd_mutex);
1101 return rv;
1102}
1103
1104/**
1105 * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1106 *
1107 * Input:
1108 * buf: ignored
1109 * size: zero
1110 * OR
1111 *
1112 * Input:
1113 * buf: any value
1114 * size: non-zero length of C string in @buf
1115 * Output:
1116 * passed-in buffer filled with "Y" or "N" with a newline
1117 * and NULL-terminated C string. This indicates whether
1118 * the grace period has ended in the current net
1119 * namespace. Return code is the size in bytes of the
1120 * string. Writing a string that starts with 'Y', 'y', or
1121 * '1' to the file will end the grace period for nfsd's v4
1122 * lock manager.
1123 */
1124static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1125{
1126 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1127
1128 if (size > 0) {
1129 switch(buf[0]) {
1130 case 'Y':
1131 case 'y':
1132 case '1':
1133 if (!nn->nfsd_serv)
1134 return -EBUSY;
1135 nfsd4_end_grace(nn);
1136 break;
1137 default:
1138 return -EINVAL;
1139 }
1140 }
1141
1142 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1143 nn->grace_ended ? 'Y' : 'N');
1144}
1145
1146#endif
1147
1148/*----------------------------------------------------------------------------*/
1149/*
1150 * populating the filesystem.
1151 */
1152
1153/* Basically copying rpc_get_inode. */
1154static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1155{
1156 struct inode *inode = new_inode(sb);
1157 if (!inode)
1158 return NULL;
1159 /* Following advice from simple_fill_super documentation: */
1160 inode->i_ino = iunique(sb, NFSD_MaxReserved);
1161 inode->i_mode = mode;
1162 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1163 switch (mode & S_IFMT) {
1164 case S_IFDIR:
1165 inode->i_fop = &simple_dir_operations;
1166 inode->i_op = &simple_dir_inode_operations;
1167 inc_nlink(inode);
1168 default:
1169 break;
1170 }
1171 return inode;
1172}
1173
1174static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1175{
1176 struct inode *inode;
1177
1178 inode = nfsd_get_inode(dir->i_sb, mode);
1179 if (!inode)
1180 return -ENOMEM;
1181 if (ncl) {
1182 inode->i_private = ncl;
1183 kref_get(&ncl->cl_ref);
1184 }
1185 d_add(dentry, inode);
1186 inc_nlink(dir);
1187 fsnotify_mkdir(dir, dentry);
1188 return 0;
1189}
1190
1191static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1192{
1193 struct inode *dir = parent->d_inode;
1194 struct dentry *dentry;
1195 int ret = -ENOMEM;
1196
1197 inode_lock(dir);
1198 dentry = d_alloc_name(parent, name);
1199 if (!dentry)
1200 goto out_err;
1201 ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1202 if (ret)
1203 goto out_err;
1204out:
1205 inode_unlock(dir);
1206 return dentry;
1207out_err:
1208 dput(dentry);
1209 dentry = ERR_PTR(ret);
1210 goto out;
1211}
1212
1213static void clear_ncl(struct inode *inode)
1214{
1215 struct nfsdfs_client *ncl = inode->i_private;
1216
1217 inode->i_private = NULL;
1218 kref_put(&ncl->cl_ref, ncl->cl_release);
1219}
1220
1221static struct nfsdfs_client *__get_nfsdfs_client(struct inode *inode)
1222{
1223 struct nfsdfs_client *nc = inode->i_private;
1224
1225 if (nc)
1226 kref_get(&nc->cl_ref);
1227 return nc;
1228}
1229
1230struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1231{
1232 struct nfsdfs_client *nc;
1233
1234 inode_lock_shared(inode);
1235 nc = __get_nfsdfs_client(inode);
1236 inode_unlock_shared(inode);
1237 return nc;
1238}
1239/* from __rpc_unlink */
1240static void nfsdfs_remove_file(struct inode *dir, struct dentry *dentry)
1241{
1242 int ret;
1243
1244 clear_ncl(d_inode(dentry));
1245 dget(dentry);
1246 ret = simple_unlink(dir, dentry);
1247 d_delete(dentry);
1248 dput(dentry);
1249 WARN_ON_ONCE(ret);
1250}
1251
1252static void nfsdfs_remove_files(struct dentry *root)
1253{
1254 struct dentry *dentry, *tmp;
1255
1256 list_for_each_entry_safe(dentry, tmp, &root->d_subdirs, d_child) {
1257 if (!simple_positive(dentry)) {
1258 WARN_ON_ONCE(1); /* I think this can't happen? */
1259 continue;
1260 }
1261 nfsdfs_remove_file(d_inode(root), dentry);
1262 }
1263}
1264
1265/* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1266 * code instead. */
1267static int nfsdfs_create_files(struct dentry *root,
1268 const struct tree_descr *files)
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 if (!files->name)
1278 continue;
1279 dentry = d_alloc_name(root, files->name);
1280 if (!dentry)
1281 goto out;
1282 inode = nfsd_get_inode(d_inode(root)->i_sb,
1283 S_IFREG | files->mode);
1284 if (!inode) {
1285 dput(dentry);
1286 goto out;
1287 }
1288 inode->i_fop = files->ops;
1289 inode->i_private = __get_nfsdfs_client(dir);
1290 d_add(dentry, inode);
1291 fsnotify_create(dir, dentry);
1292 }
1293 inode_unlock(dir);
1294 return 0;
1295out:
1296 nfsdfs_remove_files(root);
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{
1306 struct dentry *dentry;
1307 char name[11];
1308 int ret;
1309
1310 sprintf(name, "%u", id);
1311
1312 dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1313 if (IS_ERR(dentry)) /* XXX: tossing errors? */
1314 return NULL;
1315 ret = nfsdfs_create_files(dentry, files);
1316 if (ret) {
1317 nfsd_client_rmdir(dentry);
1318 return NULL;
1319 }
1320 return dentry;
1321}
1322
1323/* Taken from __rpc_rmdir: */
1324void nfsd_client_rmdir(struct dentry *dentry)
1325{
1326 struct inode *dir = d_inode(dentry->d_parent);
1327 struct inode *inode = d_inode(dentry);
1328 int ret;
1329
1330 inode_lock(dir);
1331 nfsdfs_remove_files(dentry);
1332 clear_ncl(inode);
1333 dget(dentry);
1334 ret = simple_rmdir(dir, dentry);
1335 WARN_ON_ONCE(ret);
1336 d_delete(dentry);
1337 inode_unlock(dir);
1338}
1339
1340static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1341{
1342 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1343 nfsd_net_id);
1344 struct dentry *dentry;
1345 int ret;
1346
1347 static const struct tree_descr nfsd_files[] = {
1348 [NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1349 [NFSD_Export_features] = {"export_features",
1350 &export_features_operations, S_IRUGO},
1351 [NFSD_FO_UnlockIP] = {"unlock_ip",
1352 &transaction_ops, S_IWUSR|S_IRUSR},
1353 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1354 &transaction_ops, S_IWUSR|S_IRUSR},
1355 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1356 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1357 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1358 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1359 [NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
1360 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1361 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1362 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1363 [NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1364#if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1365 [NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1366#endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1367#ifdef CONFIG_NFSD_V4
1368 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1369 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1370 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1371 [NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1372#endif
1373 /* last one */ {""}
1374 };
1375
1376 ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1377 if (ret)
1378 return ret;
1379 dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1380 if (IS_ERR(dentry))
1381 return PTR_ERR(dentry);
1382 nn->nfsd_client_dir = dentry;
1383 return 0;
1384}
1385
1386static int nfsd_fs_get_tree(struct fs_context *fc)
1387{
1388 return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1389}
1390
1391static void nfsd_fs_free_fc(struct fs_context *fc)
1392{
1393 if (fc->s_fs_info)
1394 put_net(fc->s_fs_info);
1395}
1396
1397static const struct fs_context_operations nfsd_fs_context_ops = {
1398 .free = nfsd_fs_free_fc,
1399 .get_tree = nfsd_fs_get_tree,
1400};
1401
1402static int nfsd_init_fs_context(struct fs_context *fc)
1403{
1404 put_user_ns(fc->user_ns);
1405 fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1406 fc->ops = &nfsd_fs_context_ops;
1407 return 0;
1408}
1409
1410static void nfsd_umount(struct super_block *sb)
1411{
1412 struct net *net = sb->s_fs_info;
1413
1414 kill_litter_super(sb);
1415 put_net(net);
1416}
1417
1418static struct file_system_type nfsd_fs_type = {
1419 .owner = THIS_MODULE,
1420 .name = "nfsd",
1421 .init_fs_context = nfsd_init_fs_context,
1422 .kill_sb = nfsd_umount,
1423};
1424MODULE_ALIAS_FS("nfsd");
1425
1426#ifdef CONFIG_PROC_FS
1427static int create_proc_exports_entry(void)
1428{
1429 struct proc_dir_entry *entry;
1430
1431 entry = proc_mkdir("fs/nfs", NULL);
1432 if (!entry)
1433 return -ENOMEM;
1434 entry = proc_create("exports", 0, entry,
1435 &exports_proc_operations);
1436 if (!entry) {
1437 remove_proc_entry("fs/nfs", NULL);
1438 return -ENOMEM;
1439 }
1440 return 0;
1441}
1442#else /* CONFIG_PROC_FS */
1443static int create_proc_exports_entry(void)
1444{
1445 return 0;
1446}
1447#endif
1448
1449unsigned int nfsd_net_id;
1450
1451static __net_init int nfsd_init_net(struct net *net)
1452{
1453 int retval;
1454 struct vfsmount *mnt;
1455 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1456
1457 retval = nfsd_export_init(net);
1458 if (retval)
1459 goto out_export_error;
1460 retval = nfsd_idmap_init(net);
1461 if (retval)
1462 goto out_idmap_error;
1463 nn->nfsd_versions = NULL;
1464 nn->nfsd4_minorversions = NULL;
1465 retval = nfsd_reply_cache_init(nn);
1466 if (retval)
1467 goto out_drc_error;
1468 nn->nfsd4_lease = 90; /* default lease time */
1469 nn->nfsd4_grace = 90;
1470 nn->somebody_reclaimed = false;
1471 nn->track_reclaim_completes = false;
1472 nn->clverifier_counter = prandom_u32();
1473 nn->clientid_base = prandom_u32();
1474 nn->clientid_counter = nn->clientid_base + 1;
1475 nn->s2s_cp_cl_id = nn->clientid_counter++;
1476
1477 atomic_set(&nn->ntf_refcnt, 0);
1478 init_waitqueue_head(&nn->ntf_wq);
1479 seqlock_init(&nn->boot_lock);
1480
1481 mnt = vfs_kern_mount(&nfsd_fs_type, SB_KERNMOUNT, "nfsd", NULL);
1482 if (IS_ERR(mnt)) {
1483 retval = PTR_ERR(mnt);
1484 goto out_mount_err;
1485 }
1486 nn->nfsd_mnt = mnt;
1487 return 0;
1488
1489out_mount_err:
1490 nfsd_reply_cache_shutdown(nn);
1491out_drc_error:
1492 nfsd_idmap_shutdown(net);
1493out_idmap_error:
1494 nfsd_export_shutdown(net);
1495out_export_error:
1496 return retval;
1497}
1498
1499static __net_exit void nfsd_exit_net(struct net *net)
1500{
1501 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1502
1503 mntput(nn->nfsd_mnt);
1504 nfsd_reply_cache_shutdown(nn);
1505 nfsd_idmap_shutdown(net);
1506 nfsd_export_shutdown(net);
1507 nfsd_netns_free_versions(net_generic(net, nfsd_net_id));
1508}
1509
1510static struct pernet_operations nfsd_net_ops = {
1511 .init = nfsd_init_net,
1512 .exit = nfsd_exit_net,
1513 .id = &nfsd_net_id,
1514 .size = sizeof(struct nfsd_net),
1515};
1516
1517static int __init init_nfsd(void)
1518{
1519 int retval;
1520 printk(KERN_INFO "Installing knfsd (copyright (C) 1996 okir@monad.swb.de).\n");
1521
1522 retval = register_pernet_subsys(&nfsd_net_ops);
1523 if (retval < 0)
1524 return retval;
1525 retval = register_cld_notifier();
1526 if (retval)
1527 goto out_unregister_pernet;
1528 retval = nfsd4_init_slabs();
1529 if (retval)
1530 goto out_unregister_notifier;
1531 retval = nfsd4_init_pnfs();
1532 if (retval)
1533 goto out_free_slabs;
1534 nfsd_fault_inject_init(); /* nfsd fault injection controls */
1535 nfsd_stat_init(); /* Statistics */
1536 nfsd_lockd_init(); /* lockd->nfsd callbacks */
1537 retval = create_proc_exports_entry();
1538 if (retval)
1539 goto out_free_lockd;
1540 retval = register_filesystem(&nfsd_fs_type);
1541 if (retval)
1542 goto out_free_all;
1543 return 0;
1544out_free_all:
1545 remove_proc_entry("fs/nfs/exports", NULL);
1546 remove_proc_entry("fs/nfs", NULL);
1547out_free_lockd:
1548 nfsd_lockd_shutdown();
1549 nfsd_stat_shutdown();
1550 nfsd_fault_inject_cleanup();
1551 nfsd4_exit_pnfs();
1552out_free_slabs:
1553 nfsd4_free_slabs();
1554out_unregister_notifier:
1555 unregister_cld_notifier();
1556out_unregister_pernet:
1557 unregister_pernet_subsys(&nfsd_net_ops);
1558 return retval;
1559}
1560
1561static void __exit exit_nfsd(void)
1562{
1563 remove_proc_entry("fs/nfs/exports", NULL);
1564 remove_proc_entry("fs/nfs", NULL);
1565 nfsd_stat_shutdown();
1566 nfsd_lockd_shutdown();
1567 nfsd4_free_slabs();
1568 nfsd4_exit_pnfs();
1569 nfsd_fault_inject_cleanup();
1570 unregister_filesystem(&nfsd_fs_type);
1571 unregister_cld_notifier();
1572 unregister_pernet_subsys(&nfsd_net_ops);
1573}
1574
1575MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1576MODULE_LICENSE("GPL");
1577module_init(init_nfsd)
1578module_exit(exit_nfsd)