Merge tag 'dax-for-5.3' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm
[sfrench/cifs-2.6.git] / fs / nfsd / nfsctl.c
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
12 #include <linux/sunrpc/svcsock.h>
13 #include <linux/lockd/lockd.h>
14 #include <linux/sunrpc/addr.h>
15 #include <linux/sunrpc/gss_api.h>
16 #include <linux/sunrpc/gss_krb5_enctypes.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
28 /*
29  *      We have a single directory with several nodes in it.
30  */
31 enum {
32         NFSD_Root = 1,
33         NFSD_List,
34         NFSD_Export_features,
35         NFSD_Fh,
36         NFSD_FO_UnlockIP,
37         NFSD_FO_UnlockFS,
38         NFSD_Threads,
39         NFSD_Pool_Threads,
40         NFSD_Pool_Stats,
41         NFSD_Reply_Cache_Stats,
42         NFSD_Versions,
43         NFSD_Ports,
44         NFSD_MaxBlkSize,
45         NFSD_MaxConnections,
46         NFSD_SupportedEnctypes,
47         /*
48          * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
49          * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
50          */
51 #ifdef CONFIG_NFSD_V4
52         NFSD_Leasetime,
53         NFSD_Gracetime,
54         NFSD_RecoveryDir,
55         NFSD_V4EndGrace,
56 #endif
57         NFSD_MaxReserved
58 };
59
60 /*
61  * write() for these nodes.
62  */
63 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
64 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
65 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
66 static ssize_t write_threads(struct file *file, char *buf, size_t size);
67 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
68 static ssize_t write_versions(struct file *file, char *buf, size_t size);
69 static ssize_t write_ports(struct file *file, char *buf, size_t size);
70 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
71 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
72 #ifdef CONFIG_NFSD_V4
73 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
74 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
75 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
76 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
77 #endif
78
79 static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
80         [NFSD_Fh] = write_filehandle,
81         [NFSD_FO_UnlockIP] = write_unlock_ip,
82         [NFSD_FO_UnlockFS] = write_unlock_fs,
83         [NFSD_Threads] = write_threads,
84         [NFSD_Pool_Threads] = write_pool_threads,
85         [NFSD_Versions] = write_versions,
86         [NFSD_Ports] = write_ports,
87         [NFSD_MaxBlkSize] = write_maxblksize,
88         [NFSD_MaxConnections] = write_maxconn,
89 #ifdef CONFIG_NFSD_V4
90         [NFSD_Leasetime] = write_leasetime,
91         [NFSD_Gracetime] = write_gracetime,
92         [NFSD_RecoveryDir] = write_recoverydir,
93         [NFSD_V4EndGrace] = write_v4_end_grace,
94 #endif
95 };
96
97 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
98 {
99         ino_t ino =  file_inode(file)->i_ino;
100         char *data;
101         ssize_t rv;
102
103         if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
104                 return -EINVAL;
105
106         data = simple_transaction_get(file, buf, size);
107         if (IS_ERR(data))
108                 return PTR_ERR(data);
109
110         rv =  write_op[ino](file, data, size);
111         if (rv >= 0) {
112                 simple_transaction_set(file, rv);
113                 rv = size;
114         }
115         return rv;
116 }
117
118 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
119 {
120         if (! file->private_data) {
121                 /* An attempt to read a transaction file without writing
122                  * causes a 0-byte write so that the file can return
123                  * state information
124                  */
125                 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
126                 if (rv < 0)
127                         return rv;
128         }
129         return simple_transaction_read(file, buf, size, pos);
130 }
131
132 static const struct file_operations transaction_ops = {
133         .write          = nfsctl_transaction_write,
134         .read           = nfsctl_transaction_read,
135         .release        = simple_transaction_release,
136         .llseek         = default_llseek,
137 };
138
139 static int exports_net_open(struct net *net, struct file *file)
140 {
141         int err;
142         struct seq_file *seq;
143         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
144
145         err = seq_open(file, &nfs_exports_op);
146         if (err)
147                 return err;
148
149         seq = file->private_data;
150         seq->private = nn->svc_export_cache;
151         return 0;
152 }
153
154 static int exports_proc_open(struct inode *inode, struct file *file)
155 {
156         return exports_net_open(current->nsproxy->net_ns, file);
157 }
158
159 static const struct file_operations exports_proc_operations = {
160         .open           = exports_proc_open,
161         .read           = seq_read,
162         .llseek         = seq_lseek,
163         .release        = seq_release,
164 };
165
166 static int exports_nfsd_open(struct inode *inode, struct file *file)
167 {
168         return exports_net_open(inode->i_sb->s_fs_info, file);
169 }
170
171 static const struct file_operations exports_nfsd_operations = {
172         .open           = exports_nfsd_open,
173         .read           = seq_read,
174         .llseek         = seq_lseek,
175         .release        = seq_release,
176 };
177
178 static int export_features_show(struct seq_file *m, void *v)
179 {
180         seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
181         return 0;
182 }
183
184 static int export_features_open(struct inode *inode, struct file *file)
185 {
186         return single_open(file, export_features_show, NULL);
187 }
188
189 static const struct file_operations export_features_operations = {
190         .open           = export_features_open,
191         .read           = seq_read,
192         .llseek         = seq_lseek,
193         .release        = single_release,
194 };
195
196 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
197 static int supported_enctypes_show(struct seq_file *m, void *v)
198 {
199         seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
200         return 0;
201 }
202
203 static int supported_enctypes_open(struct inode *inode, struct file *file)
204 {
205         return single_open(file, supported_enctypes_show, NULL);
206 }
207
208 static const struct file_operations supported_enctypes_ops = {
209         .open           = supported_enctypes_open,
210         .read           = seq_read,
211         .llseek         = seq_lseek,
212         .release        = single_release,
213 };
214 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
215
216 static const struct file_operations pool_stats_operations = {
217         .open           = nfsd_pool_stats_open,
218         .read           = seq_read,
219         .llseek         = seq_lseek,
220         .release        = nfsd_pool_stats_release,
221 };
222
223 static const struct file_operations reply_cache_stats_operations = {
224         .open           = nfsd_reply_cache_stats_open,
225         .read           = seq_read,
226         .llseek         = seq_lseek,
227         .release        = single_release,
228 };
229
230 /*----------------------------------------------------------------------------*/
231 /*
232  * payload - write methods
233  */
234
235 static inline struct net *netns(struct file *file)
236 {
237         return file_inode(file)->i_sb->s_fs_info;
238 }
239
240 /**
241  * write_unlock_ip - Release all locks used by a client
242  *
243  * Experimental.
244  *
245  * Input:
246  *                      buf:    '\n'-terminated C string containing a
247  *                              presentation format IP address
248  *                      size:   length of C string in @buf
249  * Output:
250  *      On success:     returns zero if all specified locks were released;
251  *                      returns one if one or more locks were not released
252  *      On error:       return code is negative errno value
253  */
254 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
255 {
256         struct sockaddr_storage address;
257         struct sockaddr *sap = (struct sockaddr *)&address;
258         size_t salen = sizeof(address);
259         char *fo_path;
260         struct net *net = netns(file);
261
262         /* sanity check */
263         if (size == 0)
264                 return -EINVAL;
265
266         if (buf[size-1] != '\n')
267                 return -EINVAL;
268
269         fo_path = buf;
270         if (qword_get(&buf, fo_path, size) < 0)
271                 return -EINVAL;
272
273         if (rpc_pton(net, fo_path, size, sap, salen) == 0)
274                 return -EINVAL;
275
276         return nlmsvc_unlock_all_by_ip(sap);
277 }
278
279 /**
280  * write_unlock_fs - Release all locks on a local file system
281  *
282  * Experimental.
283  *
284  * Input:
285  *                      buf:    '\n'-terminated C string containing the
286  *                              absolute pathname of a local file system
287  *                      size:   length of C string in @buf
288  * Output:
289  *      On success:     returns zero if all specified locks were released;
290  *                      returns one if one or more locks were not released
291  *      On error:       return code is negative errno value
292  */
293 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
294 {
295         struct path path;
296         char *fo_path;
297         int error;
298
299         /* sanity check */
300         if (size == 0)
301                 return -EINVAL;
302
303         if (buf[size-1] != '\n')
304                 return -EINVAL;
305
306         fo_path = buf;
307         if (qword_get(&buf, fo_path, size) < 0)
308                 return -EINVAL;
309
310         error = kern_path(fo_path, 0, &path);
311         if (error)
312                 return error;
313
314         /*
315          * XXX: Needs better sanity checking.  Otherwise we could end up
316          * releasing locks on the wrong file system.
317          *
318          * For example:
319          * 1.  Does the path refer to a directory?
320          * 2.  Is that directory a mount point, or
321          * 3.  Is that directory the root of an exported file system?
322          */
323         error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
324
325         path_put(&path);
326         return error;
327 }
328
329 /**
330  * write_filehandle - Get a variable-length NFS file handle by path
331  *
332  * On input, the buffer contains a '\n'-terminated C string comprised of
333  * three alphanumeric words separated by whitespace.  The string may
334  * contain escape sequences.
335  *
336  * Input:
337  *                      buf:
338  *                              domain:         client domain name
339  *                              path:           export pathname
340  *                              maxsize:        numeric maximum size of
341  *                                              @buf
342  *                      size:   length of C string in @buf
343  * Output:
344  *      On success:     passed-in buffer filled with '\n'-terminated C
345  *                      string containing a ASCII hex text version
346  *                      of the NFS file handle;
347  *                      return code is the size in bytes of the string
348  *      On error:       return code is negative errno value
349  */
350 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
351 {
352         char *dname, *path;
353         int uninitialized_var(maxsize);
354         char *mesg = buf;
355         int len;
356         struct auth_domain *dom;
357         struct knfsd_fh fh;
358
359         if (size == 0)
360                 return -EINVAL;
361
362         if (buf[size-1] != '\n')
363                 return -EINVAL;
364         buf[size-1] = 0;
365
366         dname = mesg;
367         len = qword_get(&mesg, dname, size);
368         if (len <= 0)
369                 return -EINVAL;
370         
371         path = dname+len+1;
372         len = qword_get(&mesg, path, size);
373         if (len <= 0)
374                 return -EINVAL;
375
376         len = get_int(&mesg, &maxsize);
377         if (len)
378                 return len;
379
380         if (maxsize < NFS_FHSIZE)
381                 return -EINVAL;
382         maxsize = min(maxsize, NFS3_FHSIZE);
383
384         if (qword_get(&mesg, mesg, size)>0)
385                 return -EINVAL;
386
387         /* we have all the words, they are in buf.. */
388         dom = unix_domain_find(dname);
389         if (!dom)
390                 return -ENOMEM;
391
392         len = exp_rootfh(netns(file), dom, path, &fh,  maxsize);
393         auth_domain_put(dom);
394         if (len)
395                 return len;
396         
397         mesg = buf;
398         len = SIMPLE_TRANSACTION_LIMIT;
399         qword_addhex(&mesg, &len, (char*)&fh.fh_base, fh.fh_size);
400         mesg[-1] = '\n';
401         return mesg - buf;      
402 }
403
404 /**
405  * write_threads - Start NFSD, or report the current number of running threads
406  *
407  * Input:
408  *                      buf:            ignored
409  *                      size:           zero
410  * Output:
411  *      On success:     passed-in buffer filled with '\n'-terminated C
412  *                      string numeric value representing the number of
413  *                      running NFSD threads;
414  *                      return code is the size in bytes of the string
415  *      On error:       return code is zero
416  *
417  * OR
418  *
419  * Input:
420  *                      buf:            C string containing an unsigned
421  *                                      integer value representing the
422  *                                      number of NFSD threads to start
423  *                      size:           non-zero length of C string in @buf
424  * Output:
425  *      On success:     NFS service is started;
426  *                      passed-in buffer filled with '\n'-terminated C
427  *                      string numeric value representing the number of
428  *                      running NFSD threads;
429  *                      return code is the size in bytes of the string
430  *      On error:       return code is zero or a negative errno value
431  */
432 static ssize_t write_threads(struct file *file, char *buf, size_t size)
433 {
434         char *mesg = buf;
435         int rv;
436         struct net *net = netns(file);
437
438         if (size > 0) {
439                 int newthreads;
440                 rv = get_int(&mesg, &newthreads);
441                 if (rv)
442                         return rv;
443                 if (newthreads < 0)
444                         return -EINVAL;
445                 rv = nfsd_svc(newthreads, net, file->f_cred);
446                 if (rv < 0)
447                         return rv;
448         } else
449                 rv = nfsd_nrthreads(net);
450
451         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
452 }
453
454 /**
455  * write_pool_threads - Set or report the current number of threads per pool
456  *
457  * Input:
458  *                      buf:            ignored
459  *                      size:           zero
460  *
461  * OR
462  *
463  * Input:
464  *                      buf:            C string containing whitespace-
465  *                                      separated unsigned integer values
466  *                                      representing the number of NFSD
467  *                                      threads to start in each pool
468  *                      size:           non-zero length of C string in @buf
469  * Output:
470  *      On success:     passed-in buffer filled with '\n'-terminated C
471  *                      string containing integer values representing the
472  *                      number of NFSD threads in each pool;
473  *                      return code is the size in bytes of the string
474  *      On error:       return code is zero or a negative errno value
475  */
476 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
477 {
478         /* if size > 0, look for an array of number of threads per node
479          * and apply them  then write out number of threads per node as reply
480          */
481         char *mesg = buf;
482         int i;
483         int rv;
484         int len;
485         int npools;
486         int *nthreads;
487         struct net *net = netns(file);
488
489         mutex_lock(&nfsd_mutex);
490         npools = nfsd_nrpools(net);
491         if (npools == 0) {
492                 /*
493                  * NFS is shut down.  The admin can start it by
494                  * writing to the threads file but NOT the pool_threads
495                  * file, sorry.  Report zero threads.
496                  */
497                 mutex_unlock(&nfsd_mutex);
498                 strcpy(buf, "0\n");
499                 return strlen(buf);
500         }
501
502         nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
503         rv = -ENOMEM;
504         if (nthreads == NULL)
505                 goto out_free;
506
507         if (size > 0) {
508                 for (i = 0; i < npools; i++) {
509                         rv = get_int(&mesg, &nthreads[i]);
510                         if (rv == -ENOENT)
511                                 break;          /* fewer numbers than pools */
512                         if (rv)
513                                 goto out_free;  /* syntax error */
514                         rv = -EINVAL;
515                         if (nthreads[i] < 0)
516                                 goto out_free;
517                 }
518                 rv = nfsd_set_nrthreads(i, nthreads, net);
519                 if (rv)
520                         goto out_free;
521         }
522
523         rv = nfsd_get_nrthreads(npools, nthreads, net);
524         if (rv)
525                 goto out_free;
526
527         mesg = buf;
528         size = SIMPLE_TRANSACTION_LIMIT;
529         for (i = 0; i < npools && size > 0; i++) {
530                 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
531                 len = strlen(mesg);
532                 size -= len;
533                 mesg += len;
534         }
535         rv = mesg - buf;
536 out_free:
537         kfree(nthreads);
538         mutex_unlock(&nfsd_mutex);
539         return rv;
540 }
541
542 static ssize_t
543 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
544                 const char *sep, unsigned vers, int minor)
545 {
546         const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
547         bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
548
549         if (vers == 4 && minor >= 0 &&
550             !nfsd_minorversion(nn, minor, NFSD_TEST))
551                 supported = false;
552         if (minor == 0 && supported)
553                 /*
554                  * special case for backward compatability.
555                  * +4.0 is never reported, it is implied by
556                  * +4, unless -4.0 is present.
557                  */
558                 return 0;
559         return snprintf(buf, remaining, format, sep,
560                         supported ? '+' : '-', vers, minor);
561 }
562
563 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
564 {
565         char *mesg = buf;
566         char *vers, *minorp, sign;
567         int len, num, remaining;
568         ssize_t tlen = 0;
569         char *sep;
570         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
571
572         if (size>0) {
573                 if (nn->nfsd_serv)
574                         /* Cannot change versions without updating
575                          * nn->nfsd_serv->sv_xdrsize, and reallocing
576                          * rq_argp and rq_resp
577                          */
578                         return -EBUSY;
579                 if (buf[size-1] != '\n')
580                         return -EINVAL;
581                 buf[size-1] = 0;
582
583                 vers = mesg;
584                 len = qword_get(&mesg, vers, size);
585                 if (len <= 0) return -EINVAL;
586                 do {
587                         enum vers_op cmd;
588                         unsigned minor;
589                         sign = *vers;
590                         if (sign == '+' || sign == '-')
591                                 num = simple_strtol((vers+1), &minorp, 0);
592                         else
593                                 num = simple_strtol(vers, &minorp, 0);
594                         if (*minorp == '.') {
595                                 if (num != 4)
596                                         return -EINVAL;
597                                 if (kstrtouint(minorp+1, 0, &minor) < 0)
598                                         return -EINVAL;
599                         }
600
601                         cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
602                         switch(num) {
603                         case 2:
604                         case 3:
605                                 nfsd_vers(nn, num, cmd);
606                                 break;
607                         case 4:
608                                 if (*minorp == '.') {
609                                         if (nfsd_minorversion(nn, minor, cmd) < 0)
610                                                 return -EINVAL;
611                                 } else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
612                                         /*
613                                          * Either we have +4 and no minors are enabled,
614                                          * or we have -4 and at least one minor is enabled.
615                                          * In either case, propagate 'cmd' to all minors.
616                                          */
617                                         minor = 0;
618                                         while (nfsd_minorversion(nn, minor, cmd) >= 0)
619                                                 minor++;
620                                 }
621                                 break;
622                         default:
623                                 return -EINVAL;
624                         }
625                         vers += len + 1;
626                 } while ((len = qword_get(&mesg, vers, size)) > 0);
627                 /* If all get turned off, turn them back on, as
628                  * having no versions is BAD
629                  */
630                 nfsd_reset_versions(nn);
631         }
632
633         /* Now write current state into reply buffer */
634         len = 0;
635         sep = "";
636         remaining = SIMPLE_TRANSACTION_LIMIT;
637         for (num=2 ; num <= 4 ; num++) {
638                 int minor;
639                 if (!nfsd_vers(nn, num, NFSD_AVAIL))
640                         continue;
641
642                 minor = -1;
643                 do {
644                         len = nfsd_print_version_support(nn, buf, remaining,
645                                         sep, num, minor);
646                         if (len >= remaining)
647                                 goto out;
648                         remaining -= len;
649                         buf += len;
650                         tlen += len;
651                         minor++;
652                         if (len)
653                                 sep = " ";
654                 } while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
655         }
656 out:
657         len = snprintf(buf, remaining, "\n");
658         if (len >= remaining)
659                 return -EINVAL;
660         return tlen + len;
661 }
662
663 /**
664  * write_versions - Set or report the available NFS protocol versions
665  *
666  * Input:
667  *                      buf:            ignored
668  *                      size:           zero
669  * Output:
670  *      On success:     passed-in buffer filled with '\n'-terminated C
671  *                      string containing positive or negative integer
672  *                      values representing the current status of each
673  *                      protocol version;
674  *                      return code is the size in bytes of the string
675  *      On error:       return code is zero or a negative errno value
676  *
677  * OR
678  *
679  * Input:
680  *                      buf:            C string containing whitespace-
681  *                                      separated positive or negative
682  *                                      integer values representing NFS
683  *                                      protocol versions to enable ("+n")
684  *                                      or disable ("-n")
685  *                      size:           non-zero length of C string in @buf
686  * Output:
687  *      On success:     status of zero or more protocol versions has
688  *                      been updated; passed-in buffer filled with
689  *                      '\n'-terminated C string containing positive
690  *                      or negative integer values representing the
691  *                      current status of each protocol version;
692  *                      return code is the size in bytes of the string
693  *      On error:       return code is zero or a negative errno value
694  */
695 static ssize_t write_versions(struct file *file, char *buf, size_t size)
696 {
697         ssize_t rv;
698
699         mutex_lock(&nfsd_mutex);
700         rv = __write_versions(file, buf, size);
701         mutex_unlock(&nfsd_mutex);
702         return rv;
703 }
704
705 /*
706  * Zero-length write.  Return a list of NFSD's current listener
707  * transports.
708  */
709 static ssize_t __write_ports_names(char *buf, struct net *net)
710 {
711         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
712
713         if (nn->nfsd_serv == NULL)
714                 return 0;
715         return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
716 }
717
718 /*
719  * A single 'fd' number was written, in which case it must be for
720  * a socket of a supported family/protocol, and we use it as an
721  * nfsd listener.
722  */
723 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
724 {
725         char *mesg = buf;
726         int fd, err;
727         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
728
729         err = get_int(&mesg, &fd);
730         if (err != 0 || fd < 0)
731                 return -EINVAL;
732
733         if (svc_alien_sock(net, fd)) {
734                 printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
735                 return -EINVAL;
736         }
737
738         err = nfsd_create_serv(net);
739         if (err != 0)
740                 return err;
741
742         err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
743         if (err < 0) {
744                 nfsd_destroy(net);
745                 return err;
746         }
747
748         /* Decrease the count, but don't shut down the service */
749         nn->nfsd_serv->sv_nrthreads--;
750         return err;
751 }
752
753 /*
754  * A transport listener is added by writing it's transport name and
755  * a port number.
756  */
757 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
758 {
759         char transport[16];
760         struct svc_xprt *xprt;
761         int port, err;
762         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
763
764         if (sscanf(buf, "%15s %5u", transport, &port) != 2)
765                 return -EINVAL;
766
767         if (port < 1 || port > USHRT_MAX)
768                 return -EINVAL;
769
770         err = nfsd_create_serv(net);
771         if (err != 0)
772                 return err;
773
774         err = svc_create_xprt(nn->nfsd_serv, transport, net,
775                                 PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
776         if (err < 0)
777                 goto out_err;
778
779         err = svc_create_xprt(nn->nfsd_serv, transport, net,
780                                 PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
781         if (err < 0 && err != -EAFNOSUPPORT)
782                 goto out_close;
783
784         /* Decrease the count, but don't shut down the service */
785         nn->nfsd_serv->sv_nrthreads--;
786         return 0;
787 out_close:
788         xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
789         if (xprt != NULL) {
790                 svc_close_xprt(xprt);
791                 svc_xprt_put(xprt);
792         }
793 out_err:
794         nfsd_destroy(net);
795         return err;
796 }
797
798 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
799                              struct net *net)
800 {
801         if (size == 0)
802                 return __write_ports_names(buf, net);
803
804         if (isdigit(buf[0]))
805                 return __write_ports_addfd(buf, net, file->f_cred);
806
807         if (isalpha(buf[0]))
808                 return __write_ports_addxprt(buf, net, file->f_cred);
809
810         return -EINVAL;
811 }
812
813 /**
814  * write_ports - Pass a socket file descriptor or transport name to listen on
815  *
816  * Input:
817  *                      buf:            ignored
818  *                      size:           zero
819  * Output:
820  *      On success:     passed-in buffer filled with a '\n'-terminated C
821  *                      string containing a whitespace-separated list of
822  *                      named NFSD listeners;
823  *                      return code is the size in bytes of the string
824  *      On error:       return code is zero or a negative errno value
825  *
826  * OR
827  *
828  * Input:
829  *                      buf:            C string containing an unsigned
830  *                                      integer value representing a bound
831  *                                      but unconnected socket that is to be
832  *                                      used as an NFSD listener; listen(3)
833  *                                      must be called for a SOCK_STREAM
834  *                                      socket, otherwise it is ignored
835  *                      size:           non-zero length of C string in @buf
836  * Output:
837  *      On success:     NFS service is started;
838  *                      passed-in buffer filled with a '\n'-terminated C
839  *                      string containing a unique alphanumeric name of
840  *                      the listener;
841  *                      return code is the size in bytes of the string
842  *      On error:       return code is a negative errno value
843  *
844  * OR
845  *
846  * Input:
847  *                      buf:            C string containing a transport
848  *                                      name and an unsigned integer value
849  *                                      representing the port to listen on,
850  *                                      separated by whitespace
851  *                      size:           non-zero length of C string in @buf
852  * Output:
853  *      On success:     returns zero; NFS service is started
854  *      On error:       return code is a negative errno value
855  */
856 static ssize_t write_ports(struct file *file, char *buf, size_t size)
857 {
858         ssize_t rv;
859
860         mutex_lock(&nfsd_mutex);
861         rv = __write_ports(file, buf, size, netns(file));
862         mutex_unlock(&nfsd_mutex);
863         return rv;
864 }
865
866
867 int nfsd_max_blksize;
868
869 /**
870  * write_maxblksize - Set or report the current NFS blksize
871  *
872  * Input:
873  *                      buf:            ignored
874  *                      size:           zero
875  *
876  * OR
877  *
878  * Input:
879  *                      buf:            C string containing an unsigned
880  *                                      integer value representing the new
881  *                                      NFS blksize
882  *                      size:           non-zero length of C string in @buf
883  * Output:
884  *      On success:     passed-in buffer filled with '\n'-terminated C string
885  *                      containing numeric value of the current NFS blksize
886  *                      setting;
887  *                      return code is the size in bytes of the string
888  *      On error:       return code is zero or a negative errno value
889  */
890 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
891 {
892         char *mesg = buf;
893         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
894
895         if (size > 0) {
896                 int bsize;
897                 int rv = get_int(&mesg, &bsize);
898                 if (rv)
899                         return rv;
900                 /* force bsize into allowed range and
901                  * required alignment.
902                  */
903                 bsize = max_t(int, bsize, 1024);
904                 bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
905                 bsize &= ~(1024-1);
906                 mutex_lock(&nfsd_mutex);
907                 if (nn->nfsd_serv) {
908                         mutex_unlock(&nfsd_mutex);
909                         return -EBUSY;
910                 }
911                 nfsd_max_blksize = bsize;
912                 mutex_unlock(&nfsd_mutex);
913         }
914
915         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
916                                                         nfsd_max_blksize);
917 }
918
919 /**
920  * write_maxconn - Set or report the current max number of connections
921  *
922  * Input:
923  *                      buf:            ignored
924  *                      size:           zero
925  * OR
926  *
927  * Input:
928  *                      buf:            C string containing an unsigned
929  *                                      integer value representing the new
930  *                                      number of max connections
931  *                      size:           non-zero length of C string in @buf
932  * Output:
933  *      On success:     passed-in buffer filled with '\n'-terminated C string
934  *                      containing numeric value of max_connections setting
935  *                      for this net namespace;
936  *                      return code is the size in bytes of the string
937  *      On error:       return code is zero or a negative errno value
938  */
939 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
940 {
941         char *mesg = buf;
942         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
943         unsigned int maxconn = nn->max_connections;
944
945         if (size > 0) {
946                 int rv = get_uint(&mesg, &maxconn);
947
948                 if (rv)
949                         return rv;
950                 nn->max_connections = maxconn;
951         }
952
953         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
954 }
955
956 #ifdef CONFIG_NFSD_V4
957 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
958                                   time_t *time, struct nfsd_net *nn)
959 {
960         char *mesg = buf;
961         int rv, i;
962
963         if (size > 0) {
964                 if (nn->nfsd_serv)
965                         return -EBUSY;
966                 rv = get_int(&mesg, &i);
967                 if (rv)
968                         return rv;
969                 /*
970                  * Some sanity checking.  We don't have a reason for
971                  * these particular numbers, but problems with the
972                  * extremes are:
973                  *      - Too short: the briefest network outage may
974                  *        cause clients to lose all their locks.  Also,
975                  *        the frequent polling may be wasteful.
976                  *      - Too long: do you really want reboot recovery
977                  *        to take more than an hour?  Or to make other
978                  *        clients wait an hour before being able to
979                  *        revoke a dead client's locks?
980                  */
981                 if (i < 10 || i > 3600)
982                         return -EINVAL;
983                 *time = i;
984         }
985
986         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%ld\n", *time);
987 }
988
989 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
990                                 time_t *time, struct nfsd_net *nn)
991 {
992         ssize_t rv;
993
994         mutex_lock(&nfsd_mutex);
995         rv = __nfsd4_write_time(file, buf, size, time, nn);
996         mutex_unlock(&nfsd_mutex);
997         return rv;
998 }
999
1000 /**
1001  * write_leasetime - Set or report the current NFSv4 lease time
1002  *
1003  * Input:
1004  *                      buf:            ignored
1005  *                      size:           zero
1006  *
1007  * OR
1008  *
1009  * Input:
1010  *                      buf:            C string containing an unsigned
1011  *                                      integer value representing the new
1012  *                                      NFSv4 lease expiry time
1013  *                      size:           non-zero length of C string in @buf
1014  * Output:
1015  *      On success:     passed-in buffer filled with '\n'-terminated C
1016  *                      string containing unsigned integer value of the
1017  *                      current lease expiry time;
1018  *                      return code is the size in bytes of the string
1019  *      On error:       return code is zero or a negative errno value
1020  */
1021 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1022 {
1023         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1024         return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1025 }
1026
1027 /**
1028  * write_gracetime - Set or report current NFSv4 grace period time
1029  *
1030  * As above, but sets the time of the NFSv4 grace period.
1031  *
1032  * Note this should never be set to less than the *previous*
1033  * lease-period time, but we don't try to enforce this.  (In the common
1034  * case (a new boot), we don't know what the previous lease time was
1035  * anyway.)
1036  */
1037 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1038 {
1039         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1040         return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1041 }
1042
1043 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1044                                    struct nfsd_net *nn)
1045 {
1046         char *mesg = buf;
1047         char *recdir;
1048         int len, status;
1049
1050         if (size > 0) {
1051                 if (nn->nfsd_serv)
1052                         return -EBUSY;
1053                 if (size > PATH_MAX || buf[size-1] != '\n')
1054                         return -EINVAL;
1055                 buf[size-1] = 0;
1056
1057                 recdir = mesg;
1058                 len = qword_get(&mesg, recdir, size);
1059                 if (len <= 0)
1060                         return -EINVAL;
1061
1062                 status = nfs4_reset_recoverydir(recdir);
1063                 if (status)
1064                         return status;
1065         }
1066
1067         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1068                                                         nfs4_recoverydir());
1069 }
1070
1071 /**
1072  * write_recoverydir - Set or report the pathname of the recovery directory
1073  *
1074  * Input:
1075  *                      buf:            ignored
1076  *                      size:           zero
1077  *
1078  * OR
1079  *
1080  * Input:
1081  *                      buf:            C string containing the pathname
1082  *                                      of the directory on a local file
1083  *                                      system containing permanent NFSv4
1084  *                                      recovery data
1085  *                      size:           non-zero length of C string in @buf
1086  * Output:
1087  *      On success:     passed-in buffer filled with '\n'-terminated C string
1088  *                      containing the current recovery pathname setting;
1089  *                      return code is the size in bytes of the string
1090  *      On error:       return code is zero or a negative errno value
1091  */
1092 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1093 {
1094         ssize_t rv;
1095         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1096
1097         mutex_lock(&nfsd_mutex);
1098         rv = __write_recoverydir(file, buf, size, nn);
1099         mutex_unlock(&nfsd_mutex);
1100         return rv;
1101 }
1102
1103 /**
1104  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1105  *
1106  * Input:
1107  *                      buf:            ignored
1108  *                      size:           zero
1109  * OR
1110  *
1111  * Input:
1112  *                      buf:            any value
1113  *                      size:           non-zero length of C string in @buf
1114  * Output:
1115  *                      passed-in buffer filled with "Y" or "N" with a newline
1116  *                      and NULL-terminated C string. This indicates whether
1117  *                      the grace period has ended in the current net
1118  *                      namespace. Return code is the size in bytes of the
1119  *                      string. Writing a string that starts with 'Y', 'y', or
1120  *                      '1' to the file will end the grace period for nfsd's v4
1121  *                      lock manager.
1122  */
1123 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1124 {
1125         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1126
1127         if (size > 0) {
1128                 switch(buf[0]) {
1129                 case 'Y':
1130                 case 'y':
1131                 case '1':
1132                         if (!nn->nfsd_serv)
1133                                 return -EBUSY;
1134                         nfsd4_end_grace(nn);
1135                         break;
1136                 default:
1137                         return -EINVAL;
1138                 }
1139         }
1140
1141         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1142                          nn->grace_ended ? 'Y' : 'N');
1143 }
1144
1145 #endif
1146
1147 /*----------------------------------------------------------------------------*/
1148 /*
1149  *      populating the filesystem.
1150  */
1151
1152 /* Basically copying rpc_get_inode. */
1153 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1154 {
1155         struct inode *inode = new_inode(sb);
1156         if (!inode)
1157                 return NULL;
1158         /* Following advice from simple_fill_super documentation: */
1159         inode->i_ino = iunique(sb, NFSD_MaxReserved);
1160         inode->i_mode = mode;
1161         inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1162         switch (mode & S_IFMT) {
1163         case S_IFDIR:
1164                 inode->i_fop = &simple_dir_operations;
1165                 inode->i_op = &simple_dir_inode_operations;
1166                 inc_nlink(inode);
1167         default:
1168                 break;
1169         }
1170         return inode;
1171 }
1172
1173 static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
1174 {
1175         struct inode *inode;
1176
1177         inode = nfsd_get_inode(dir->i_sb, mode);
1178         if (!inode)
1179                 return -ENOMEM;
1180         d_add(dentry, inode);
1181         inc_nlink(dir);
1182         fsnotify_mkdir(dir, dentry);
1183         return 0;
1184 }
1185
1186 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1187 {
1188         struct inode *dir = parent->d_inode;
1189         struct dentry *dentry;
1190         int ret = -ENOMEM;
1191
1192         inode_lock(dir);
1193         dentry = d_alloc_name(parent, name);
1194         if (!dentry)
1195                 goto out_err;
1196         ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600);
1197         if (ret)
1198                 goto out_err;
1199         if (ncl) {
1200                 d_inode(dentry)->i_private = ncl;
1201                 kref_get(&ncl->cl_ref);
1202         }
1203 out:
1204         inode_unlock(dir);
1205         return dentry;
1206 out_err:
1207         dentry = ERR_PTR(ret);
1208         goto out;
1209 }
1210
1211 static void clear_ncl(struct inode *inode)
1212 {
1213         struct nfsdfs_client *ncl = inode->i_private;
1214
1215         inode->i_private = NULL;
1216         synchronize_rcu();
1217         kref_put(&ncl->cl_ref, ncl->cl_release);
1218 }
1219
1220
1221 static 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
1230 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1231 {
1232         struct nfsdfs_client *nc;
1233
1234         rcu_read_lock();
1235         nc = __get_nfsdfs_client(inode);
1236         rcu_read_unlock();
1237         return nc;
1238 }
1239 /* from __rpc_unlink */
1240 static 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
1252 static 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. */
1267 static  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;
1295 out:
1296         nfsdfs_remove_files(root);
1297         inode_unlock(dir);
1298         return -ENOMEM;
1299 }
1300
1301 /* on success, returns positive number unique to that client. */
1302 struct 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: */
1324 void 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
1340 static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
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         get_net(sb->s_fs_info);
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 }
1386
1387 static struct dentry *nfsd_mount(struct file_system_type *fs_type,
1388         int flags, const char *dev_name, void *data)
1389 {
1390         struct net *net = current->nsproxy->net_ns;
1391         return mount_ns(fs_type, flags, data, net, net->user_ns, nfsd_fill_super);
1392 }
1393
1394 static void nfsd_umount(struct super_block *sb)
1395 {
1396         struct net *net = sb->s_fs_info;
1397
1398         kill_litter_super(sb);
1399         put_net(net);
1400 }
1401
1402 static struct file_system_type nfsd_fs_type = {
1403         .owner          = THIS_MODULE,
1404         .name           = "nfsd",
1405         .mount          = nfsd_mount,
1406         .kill_sb        = nfsd_umount,
1407 };
1408 MODULE_ALIAS_FS("nfsd");
1409
1410 #ifdef CONFIG_PROC_FS
1411 static int create_proc_exports_entry(void)
1412 {
1413         struct proc_dir_entry *entry;
1414
1415         entry = proc_mkdir("fs/nfs", NULL);
1416         if (!entry)
1417                 return -ENOMEM;
1418         entry = proc_create("exports", 0, entry,
1419                                  &exports_proc_operations);
1420         if (!entry) {
1421                 remove_proc_entry("fs/nfs", NULL);
1422                 return -ENOMEM;
1423         }
1424         return 0;
1425 }
1426 #else /* CONFIG_PROC_FS */
1427 static int create_proc_exports_entry(void)
1428 {
1429         return 0;
1430 }
1431 #endif
1432
1433 unsigned int nfsd_net_id;
1434
1435 static __net_init int nfsd_init_net(struct net *net)
1436 {
1437         int retval;
1438         struct vfsmount *mnt;
1439         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1440
1441         retval = nfsd_export_init(net);
1442         if (retval)
1443                 goto out_export_error;
1444         retval = nfsd_idmap_init(net);
1445         if (retval)
1446                 goto out_idmap_error;
1447         nn->nfsd_versions = NULL;
1448         nn->nfsd4_minorversions = NULL;
1449         retval = nfsd_reply_cache_init(nn);
1450         if (retval)
1451                 goto out_drc_error;
1452         nn->nfsd4_lease = 90;   /* default lease time */
1453         nn->nfsd4_grace = 90;
1454         nn->somebody_reclaimed = false;
1455         nn->track_reclaim_completes = false;
1456         nn->clverifier_counter = prandom_u32();
1457         nn->clientid_base = prandom_u32();
1458         nn->clientid_counter = nn->clientid_base + 1;
1459         nn->s2s_cp_cl_id = nn->clientid_counter++;
1460
1461         atomic_set(&nn->ntf_refcnt, 0);
1462         init_waitqueue_head(&nn->ntf_wq);
1463
1464         mnt =  vfs_kern_mount(&nfsd_fs_type, SB_KERNMOUNT, "nfsd", NULL);
1465         if (IS_ERR(mnt)) {
1466                 retval = PTR_ERR(mnt);
1467                 goto out_mount_err;
1468         }
1469         nn->nfsd_mnt = mnt;
1470         return 0;
1471
1472 out_mount_err:
1473         nfsd_reply_cache_shutdown(nn);
1474 out_drc_error:
1475         nfsd_idmap_shutdown(net);
1476 out_idmap_error:
1477         nfsd_export_shutdown(net);
1478 out_export_error:
1479         return retval;
1480 }
1481
1482 static __net_exit void nfsd_exit_net(struct net *net)
1483 {
1484         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1485
1486         mntput(nn->nfsd_mnt);
1487         nfsd_reply_cache_shutdown(nn);
1488         nfsd_idmap_shutdown(net);
1489         nfsd_export_shutdown(net);
1490         nfsd_netns_free_versions(net_generic(net, nfsd_net_id));
1491 }
1492
1493 static struct pernet_operations nfsd_net_ops = {
1494         .init = nfsd_init_net,
1495         .exit = nfsd_exit_net,
1496         .id   = &nfsd_net_id,
1497         .size = sizeof(struct nfsd_net),
1498 };
1499
1500 static int __init init_nfsd(void)
1501 {
1502         int retval;
1503         printk(KERN_INFO "Installing knfsd (copyright (C) 1996 okir@monad.swb.de).\n");
1504
1505         retval = register_pernet_subsys(&nfsd_net_ops);
1506         if (retval < 0)
1507                 return retval;
1508         retval = register_cld_notifier();
1509         if (retval)
1510                 goto out_unregister_pernet;
1511         retval = nfsd4_init_slabs();
1512         if (retval)
1513                 goto out_unregister_notifier;
1514         retval = nfsd4_init_pnfs();
1515         if (retval)
1516                 goto out_free_slabs;
1517         nfsd_fault_inject_init(); /* nfsd fault injection controls */
1518         nfsd_stat_init();       /* Statistics */
1519         nfsd_lockd_init();      /* lockd->nfsd callbacks */
1520         retval = create_proc_exports_entry();
1521         if (retval)
1522                 goto out_free_lockd;
1523         retval = register_filesystem(&nfsd_fs_type);
1524         if (retval)
1525                 goto out_free_all;
1526         return 0;
1527 out_free_all:
1528         remove_proc_entry("fs/nfs/exports", NULL);
1529         remove_proc_entry("fs/nfs", NULL);
1530 out_free_lockd:
1531         nfsd_lockd_shutdown();
1532         nfsd_stat_shutdown();
1533         nfsd_fault_inject_cleanup();
1534         nfsd4_exit_pnfs();
1535 out_free_slabs:
1536         nfsd4_free_slabs();
1537 out_unregister_notifier:
1538         unregister_cld_notifier();
1539 out_unregister_pernet:
1540         unregister_pernet_subsys(&nfsd_net_ops);
1541         return retval;
1542 }
1543
1544 static void __exit exit_nfsd(void)
1545 {
1546         remove_proc_entry("fs/nfs/exports", NULL);
1547         remove_proc_entry("fs/nfs", NULL);
1548         nfsd_stat_shutdown();
1549         nfsd_lockd_shutdown();
1550         nfsd4_free_slabs();
1551         nfsd4_exit_pnfs();
1552         nfsd_fault_inject_cleanup();
1553         unregister_filesystem(&nfsd_fs_type);
1554         unregister_cld_notifier();
1555         unregister_pernet_subsys(&nfsd_net_ops);
1556 }
1557
1558 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1559 MODULE_LICENSE("GPL");
1560 module_init(init_nfsd)
1561 module_exit(exit_nfsd)