GNU Linux-libre 4.4.300-gnu1
[releases.git] / fs / nfsd / nfsctl.c
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/addr.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 "state.h"
23 #include "netns.h"
24 #include "pnfs.h"
25
26 /*
27  *      We have a single directory with several nodes in it.
28  */
29 enum {
30         NFSD_Root = 1,
31         NFSD_List,
32         NFSD_Export_features,
33         NFSD_Fh,
34         NFSD_FO_UnlockIP,
35         NFSD_FO_UnlockFS,
36         NFSD_Threads,
37         NFSD_Pool_Threads,
38         NFSD_Pool_Stats,
39         NFSD_Reply_Cache_Stats,
40         NFSD_Versions,
41         NFSD_Ports,
42         NFSD_MaxBlkSize,
43         NFSD_MaxConnections,
44         NFSD_SupportedEnctypes,
45         /*
46          * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
47          * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
48          */
49 #ifdef CONFIG_NFSD_V4
50         NFSD_Leasetime,
51         NFSD_Gracetime,
52         NFSD_RecoveryDir,
53         NFSD_V4EndGrace,
54 #endif
55 };
56
57 /*
58  * write() for these nodes.
59  */
60 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
61 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
62 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
63 static ssize_t write_threads(struct file *file, char *buf, size_t size);
64 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
65 static ssize_t write_versions(struct file *file, char *buf, size_t size);
66 static ssize_t write_ports(struct file *file, char *buf, size_t size);
67 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
68 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
69 #ifdef CONFIG_NFSD_V4
70 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
71 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
72 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
73 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
74 #endif
75
76 static ssize_t (*write_op[])(struct file *, char *, size_t) = {
77         [NFSD_Fh] = write_filehandle,
78         [NFSD_FO_UnlockIP] = write_unlock_ip,
79         [NFSD_FO_UnlockFS] = write_unlock_fs,
80         [NFSD_Threads] = write_threads,
81         [NFSD_Pool_Threads] = write_pool_threads,
82         [NFSD_Versions] = write_versions,
83         [NFSD_Ports] = write_ports,
84         [NFSD_MaxBlkSize] = write_maxblksize,
85         [NFSD_MaxConnections] = write_maxconn,
86 #ifdef CONFIG_NFSD_V4
87         [NFSD_Leasetime] = write_leasetime,
88         [NFSD_Gracetime] = write_gracetime,
89         [NFSD_RecoveryDir] = write_recoverydir,
90         [NFSD_V4EndGrace] = write_v4_end_grace,
91 #endif
92 };
93
94 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
95 {
96         ino_t ino =  file_inode(file)->i_ino;
97         char *data;
98         ssize_t rv;
99
100         if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
101                 return -EINVAL;
102
103         data = simple_transaction_get(file, buf, size);
104         if (IS_ERR(data))
105                 return PTR_ERR(data);
106
107         rv =  write_op[ino](file, data, size);
108         if (rv >= 0) {
109                 simple_transaction_set(file, rv);
110                 rv = size;
111         }
112         return rv;
113 }
114
115 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
116 {
117         if (! file->private_data) {
118                 /* An attempt to read a transaction file without writing
119                  * causes a 0-byte write so that the file can return
120                  * state information
121                  */
122                 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
123                 if (rv < 0)
124                         return rv;
125         }
126         return simple_transaction_read(file, buf, size, pos);
127 }
128
129 static const struct file_operations transaction_ops = {
130         .write          = nfsctl_transaction_write,
131         .read           = nfsctl_transaction_read,
132         .release        = simple_transaction_release,
133         .llseek         = default_llseek,
134 };
135
136 static int exports_net_open(struct net *net, struct file *file)
137 {
138         int err;
139         struct seq_file *seq;
140         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
141
142         err = seq_open(file, &nfs_exports_op);
143         if (err)
144                 return err;
145
146         seq = file->private_data;
147         seq->private = nn->svc_export_cache;
148         return 0;
149 }
150
151 static int exports_proc_open(struct inode *inode, struct file *file)
152 {
153         return exports_net_open(current->nsproxy->net_ns, file);
154 }
155
156 static const struct file_operations exports_proc_operations = {
157         .open           = exports_proc_open,
158         .read           = seq_read,
159         .llseek         = seq_lseek,
160         .release        = seq_release,
161         .owner          = THIS_MODULE,
162 };
163
164 static int exports_nfsd_open(struct inode *inode, struct file *file)
165 {
166         return exports_net_open(inode->i_sb->s_fs_info, file);
167 }
168
169 static const struct file_operations exports_nfsd_operations = {
170         .open           = exports_nfsd_open,
171         .read           = seq_read,
172         .llseek         = seq_lseek,
173         .release        = seq_release,
174         .owner          = THIS_MODULE,
175 };
176
177 static int export_features_show(struct seq_file *m, void *v)
178 {
179         seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
180         return 0;
181 }
182
183 static int export_features_open(struct inode *inode, struct file *file)
184 {
185         return single_open(file, export_features_show, NULL);
186 }
187
188 static const struct file_operations export_features_operations = {
189         .open           = export_features_open,
190         .read           = seq_read,
191         .llseek         = seq_lseek,
192         .release        = single_release,
193 };
194
195 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
196 static int supported_enctypes_show(struct seq_file *m, void *v)
197 {
198         seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
199         return 0;
200 }
201
202 static int supported_enctypes_open(struct inode *inode, struct file *file)
203 {
204         return single_open(file, supported_enctypes_show, NULL);
205 }
206
207 static const struct file_operations supported_enctypes_ops = {
208         .open           = supported_enctypes_open,
209         .read           = seq_read,
210         .llseek         = seq_lseek,
211         .release        = single_release,
212 };
213 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
214
215 static const struct file_operations pool_stats_operations = {
216         .open           = nfsd_pool_stats_open,
217         .read           = seq_read,
218         .llseek         = seq_lseek,
219         .release        = nfsd_pool_stats_release,
220         .owner          = THIS_MODULE,
221 };
222
223 static 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);
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 __write_versions(struct file *file, char *buf, size_t size)
543 {
544         char *mesg = buf;
545         char *vers, *minorp, sign;
546         int len, num, remaining;
547         unsigned minor;
548         ssize_t tlen = 0;
549         char *sep;
550         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
551
552         if (size>0) {
553                 if (nn->nfsd_serv)
554                         /* Cannot change versions without updating
555                          * nn->nfsd_serv->sv_xdrsize, and reallocing
556                          * rq_argp and rq_resp
557                          */
558                         return -EBUSY;
559                 if (buf[size-1] != '\n')
560                         return -EINVAL;
561                 buf[size-1] = 0;
562
563                 vers = mesg;
564                 len = qword_get(&mesg, vers, size);
565                 if (len <= 0) return -EINVAL;
566                 do {
567                         sign = *vers;
568                         if (sign == '+' || sign == '-')
569                                 num = simple_strtol((vers+1), &minorp, 0);
570                         else
571                                 num = simple_strtol(vers, &minorp, 0);
572                         if (*minorp == '.') {
573                                 if (num != 4)
574                                         return -EINVAL;
575                                 minor = simple_strtoul(minorp+1, NULL, 0);
576                                 if (minor == 0)
577                                         return -EINVAL;
578                                 if (nfsd_minorversion(minor, sign == '-' ?
579                                                      NFSD_CLEAR : NFSD_SET) < 0)
580                                         return -EINVAL;
581                                 goto next;
582                         }
583                         switch(num) {
584                         case 2:
585                         case 3:
586                         case 4:
587                                 nfsd_vers(num, sign == '-' ? NFSD_CLEAR : NFSD_SET);
588                                 break;
589                         default:
590                                 return -EINVAL;
591                         }
592                 next:
593                         vers += len + 1;
594                 } while ((len = qword_get(&mesg, vers, size)) > 0);
595                 /* If all get turned off, turn them back on, as
596                  * having no versions is BAD
597                  */
598                 nfsd_reset_versions();
599         }
600
601         /* Now write current state into reply buffer */
602         len = 0;
603         sep = "";
604         remaining = SIMPLE_TRANSACTION_LIMIT;
605         for (num=2 ; num <= 4 ; num++)
606                 if (nfsd_vers(num, NFSD_AVAIL)) {
607                         len = snprintf(buf, remaining, "%s%c%d", sep,
608                                        nfsd_vers(num, NFSD_TEST)?'+':'-',
609                                        num);
610                         sep = " ";
611
612                         if (len >= remaining)
613                                 break;
614                         remaining -= len;
615                         buf += len;
616                         tlen += len;
617                 }
618         if (nfsd_vers(4, NFSD_AVAIL))
619                 for (minor = 1; minor <= NFSD_SUPPORTED_MINOR_VERSION;
620                      minor++) {
621                         len = snprintf(buf, remaining, " %c4.%u",
622                                         (nfsd_vers(4, NFSD_TEST) &&
623                                          nfsd_minorversion(minor, NFSD_TEST)) ?
624                                                 '+' : '-',
625                                         minor);
626
627                         if (len >= remaining)
628                                 break;
629                         remaining -= len;
630                         buf += len;
631                         tlen += len;
632                 }
633
634         len = snprintf(buf, remaining, "\n");
635         if (len >= remaining)
636                 return -EINVAL;
637         return tlen + len;
638 }
639
640 /**
641  * write_versions - Set or report the available NFS protocol versions
642  *
643  * Input:
644  *                      buf:            ignored
645  *                      size:           zero
646  * Output:
647  *      On success:     passed-in buffer filled with '\n'-terminated C
648  *                      string containing positive or negative integer
649  *                      values representing the current status of each
650  *                      protocol version;
651  *                      return code is the size in bytes of the string
652  *      On error:       return code is zero or a negative errno value
653  *
654  * OR
655  *
656  * Input:
657  *                      buf:            C string containing whitespace-
658  *                                      separated positive or negative
659  *                                      integer values representing NFS
660  *                                      protocol versions to enable ("+n")
661  *                                      or disable ("-n")
662  *                      size:           non-zero length of C string in @buf
663  * Output:
664  *      On success:     status of zero or more protocol versions has
665  *                      been updated; passed-in buffer filled with
666  *                      '\n'-terminated C string containing positive
667  *                      or negative integer values representing the
668  *                      current status of each protocol version;
669  *                      return code is the size in bytes of the string
670  *      On error:       return code is zero or a negative errno value
671  */
672 static ssize_t write_versions(struct file *file, char *buf, size_t size)
673 {
674         ssize_t rv;
675
676         mutex_lock(&nfsd_mutex);
677         rv = __write_versions(file, buf, size);
678         mutex_unlock(&nfsd_mutex);
679         return rv;
680 }
681
682 /*
683  * Zero-length write.  Return a list of NFSD's current listener
684  * transports.
685  */
686 static ssize_t __write_ports_names(char *buf, struct net *net)
687 {
688         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
689
690         if (nn->nfsd_serv == NULL)
691                 return 0;
692         return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
693 }
694
695 /*
696  * A single 'fd' number was written, in which case it must be for
697  * a socket of a supported family/protocol, and we use it as an
698  * nfsd listener.
699  */
700 static ssize_t __write_ports_addfd(char *buf, struct net *net)
701 {
702         char *mesg = buf;
703         int fd, err;
704         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
705
706         err = get_int(&mesg, &fd);
707         if (err != 0 || fd < 0)
708                 return -EINVAL;
709
710         if (svc_alien_sock(net, fd)) {
711                 printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
712                 return -EINVAL;
713         }
714
715         err = nfsd_create_serv(net);
716         if (err != 0)
717                 return err;
718
719         err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT);
720         if (err < 0) {
721                 nfsd_destroy(net);
722                 return err;
723         }
724
725         /* Decrease the count, but don't shut down the service */
726         nn->nfsd_serv->sv_nrthreads--;
727         return err;
728 }
729
730 /*
731  * A transport listener is added by writing it's transport name and
732  * a port number.
733  */
734 static ssize_t __write_ports_addxprt(char *buf, struct net *net)
735 {
736         char transport[16];
737         struct svc_xprt *xprt;
738         int port, err;
739         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
740
741         if (sscanf(buf, "%15s %5u", transport, &port) != 2)
742                 return -EINVAL;
743
744         if (port < 1 || port > USHRT_MAX)
745                 return -EINVAL;
746
747         err = nfsd_create_serv(net);
748         if (err != 0)
749                 return err;
750
751         err = svc_create_xprt(nn->nfsd_serv, transport, net,
752                                 PF_INET, port, SVC_SOCK_ANONYMOUS);
753         if (err < 0)
754                 goto out_err;
755
756         err = svc_create_xprt(nn->nfsd_serv, transport, net,
757                                 PF_INET6, port, SVC_SOCK_ANONYMOUS);
758         if (err < 0 && err != -EAFNOSUPPORT)
759                 goto out_close;
760
761         /* Decrease the count, but don't shut down the service */
762         nn->nfsd_serv->sv_nrthreads--;
763         return 0;
764 out_close:
765         xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
766         if (xprt != NULL) {
767                 svc_close_xprt(xprt);
768                 svc_xprt_put(xprt);
769         }
770 out_err:
771         if (!list_empty(&nn->nfsd_serv->sv_permsocks))
772                 nn->nfsd_serv->sv_nrthreads--;
773          else
774                 nfsd_destroy(net);
775         return err;
776 }
777
778 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
779                              struct net *net)
780 {
781         if (size == 0)
782                 return __write_ports_names(buf, net);
783
784         if (isdigit(buf[0]))
785                 return __write_ports_addfd(buf, net);
786
787         if (isalpha(buf[0]))
788                 return __write_ports_addxprt(buf, net);
789
790         return -EINVAL;
791 }
792
793 /**
794  * write_ports - Pass a socket file descriptor or transport name to listen on
795  *
796  * Input:
797  *                      buf:            ignored
798  *                      size:           zero
799  * Output:
800  *      On success:     passed-in buffer filled with a '\n'-terminated C
801  *                      string containing a whitespace-separated list of
802  *                      named NFSD listeners;
803  *                      return code is the size in bytes of the string
804  *      On error:       return code is zero or a negative errno value
805  *
806  * OR
807  *
808  * Input:
809  *                      buf:            C string containing an unsigned
810  *                                      integer value representing a bound
811  *                                      but unconnected socket that is to be
812  *                                      used as an NFSD listener; listen(3)
813  *                                      must be called for a SOCK_STREAM
814  *                                      socket, otherwise it is ignored
815  *                      size:           non-zero length of C string in @buf
816  * Output:
817  *      On success:     NFS service is started;
818  *                      passed-in buffer filled with a '\n'-terminated C
819  *                      string containing a unique alphanumeric name of
820  *                      the listener;
821  *                      return code is the size in bytes of the string
822  *      On error:       return code is a negative errno value
823  *
824  * OR
825  *
826  * Input:
827  *                      buf:            C string containing a transport
828  *                                      name and an unsigned integer value
829  *                                      representing the port to listen on,
830  *                                      separated by whitespace
831  *                      size:           non-zero length of C string in @buf
832  * Output:
833  *      On success:     returns zero; NFS service is started
834  *      On error:       return code is a negative errno value
835  */
836 static ssize_t write_ports(struct file *file, char *buf, size_t size)
837 {
838         ssize_t rv;
839
840         mutex_lock(&nfsd_mutex);
841         rv = __write_ports(file, buf, size, netns(file));
842         mutex_unlock(&nfsd_mutex);
843         return rv;
844 }
845
846
847 int nfsd_max_blksize;
848
849 /**
850  * write_maxblksize - Set or report the current NFS blksize
851  *
852  * Input:
853  *                      buf:            ignored
854  *                      size:           zero
855  *
856  * OR
857  *
858  * Input:
859  *                      buf:            C string containing an unsigned
860  *                                      integer value representing the new
861  *                                      NFS blksize
862  *                      size:           non-zero length of C string in @buf
863  * Output:
864  *      On success:     passed-in buffer filled with '\n'-terminated C string
865  *                      containing numeric value of the current NFS blksize
866  *                      setting;
867  *                      return code is the size in bytes of the string
868  *      On error:       return code is zero or a negative errno value
869  */
870 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
871 {
872         char *mesg = buf;
873         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
874
875         if (size > 0) {
876                 int bsize;
877                 int rv = get_int(&mesg, &bsize);
878                 if (rv)
879                         return rv;
880                 /* force bsize into allowed range and
881                  * required alignment.
882                  */
883                 bsize = max_t(int, bsize, 1024);
884                 bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
885                 bsize &= ~(1024-1);
886                 mutex_lock(&nfsd_mutex);
887                 if (nn->nfsd_serv) {
888                         mutex_unlock(&nfsd_mutex);
889                         return -EBUSY;
890                 }
891                 nfsd_max_blksize = bsize;
892                 mutex_unlock(&nfsd_mutex);
893         }
894
895         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
896                                                         nfsd_max_blksize);
897 }
898
899 /**
900  * write_maxconn - Set or report the current max number of connections
901  *
902  * Input:
903  *                      buf:            ignored
904  *                      size:           zero
905  * OR
906  *
907  * Input:
908  *                      buf:            C string containing an unsigned
909  *                                      integer value representing the new
910  *                                      number of max connections
911  *                      size:           non-zero length of C string in @buf
912  * Output:
913  *      On success:     passed-in buffer filled with '\n'-terminated C string
914  *                      containing numeric value of max_connections setting
915  *                      for this net namespace;
916  *                      return code is the size in bytes of the string
917  *      On error:       return code is zero or a negative errno value
918  */
919 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
920 {
921         char *mesg = buf;
922         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
923         unsigned int maxconn = nn->max_connections;
924
925         if (size > 0) {
926                 int rv = get_uint(&mesg, &maxconn);
927
928                 if (rv)
929                         return rv;
930                 nn->max_connections = maxconn;
931         }
932
933         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
934 }
935
936 #ifdef CONFIG_NFSD_V4
937 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
938                                   time_t *time, struct nfsd_net *nn)
939 {
940         char *mesg = buf;
941         int rv, i;
942
943         if (size > 0) {
944                 if (nn->nfsd_serv)
945                         return -EBUSY;
946                 rv = get_int(&mesg, &i);
947                 if (rv)
948                         return rv;
949                 /*
950                  * Some sanity checking.  We don't have a reason for
951                  * these particular numbers, but problems with the
952                  * extremes are:
953                  *      - Too short: the briefest network outage may
954                  *        cause clients to lose all their locks.  Also,
955                  *        the frequent polling may be wasteful.
956                  *      - Too long: do you really want reboot recovery
957                  *        to take more than an hour?  Or to make other
958                  *        clients wait an hour before being able to
959                  *        revoke a dead client's locks?
960                  */
961                 if (i < 10 || i > 3600)
962                         return -EINVAL;
963                 *time = i;
964         }
965
966         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%ld\n", *time);
967 }
968
969 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
970                                 time_t *time, struct nfsd_net *nn)
971 {
972         ssize_t rv;
973
974         mutex_lock(&nfsd_mutex);
975         rv = __nfsd4_write_time(file, buf, size, time, nn);
976         mutex_unlock(&nfsd_mutex);
977         return rv;
978 }
979
980 /**
981  * write_leasetime - Set or report the current NFSv4 lease time
982  *
983  * Input:
984  *                      buf:            ignored
985  *                      size:           zero
986  *
987  * OR
988  *
989  * Input:
990  *                      buf:            C string containing an unsigned
991  *                                      integer value representing the new
992  *                                      NFSv4 lease expiry time
993  *                      size:           non-zero length of C string in @buf
994  * Output:
995  *      On success:     passed-in buffer filled with '\n'-terminated C
996  *                      string containing unsigned integer value of the
997  *                      current lease expiry time;
998  *                      return code is the size in bytes of the string
999  *      On error:       return code is zero or a negative errno value
1000  */
1001 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1002 {
1003         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1004         return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1005 }
1006
1007 /**
1008  * write_gracetime - Set or report current NFSv4 grace period time
1009  *
1010  * As above, but sets the time of the NFSv4 grace period.
1011  *
1012  * Note this should never be set to less than the *previous*
1013  * lease-period time, but we don't try to enforce this.  (In the common
1014  * case (a new boot), we don't know what the previous lease time was
1015  * anyway.)
1016  */
1017 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1018 {
1019         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1020         return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1021 }
1022
1023 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1024                                    struct nfsd_net *nn)
1025 {
1026         char *mesg = buf;
1027         char *recdir;
1028         int len, status;
1029
1030         if (size > 0) {
1031                 if (nn->nfsd_serv)
1032                         return -EBUSY;
1033                 if (size > PATH_MAX || buf[size-1] != '\n')
1034                         return -EINVAL;
1035                 buf[size-1] = 0;
1036
1037                 recdir = mesg;
1038                 len = qword_get(&mesg, recdir, size);
1039                 if (len <= 0)
1040                         return -EINVAL;
1041
1042                 status = nfs4_reset_recoverydir(recdir);
1043                 if (status)
1044                         return status;
1045         }
1046
1047         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1048                                                         nfs4_recoverydir());
1049 }
1050
1051 /**
1052  * write_recoverydir - Set or report the pathname of the recovery directory
1053  *
1054  * Input:
1055  *                      buf:            ignored
1056  *                      size:           zero
1057  *
1058  * OR
1059  *
1060  * Input:
1061  *                      buf:            C string containing the pathname
1062  *                                      of the directory on a local file
1063  *                                      system containing permanent NFSv4
1064  *                                      recovery data
1065  *                      size:           non-zero length of C string in @buf
1066  * Output:
1067  *      On success:     passed-in buffer filled with '\n'-terminated C string
1068  *                      containing the current recovery pathname setting;
1069  *                      return code is the size in bytes of the string
1070  *      On error:       return code is zero or a negative errno value
1071  */
1072 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1073 {
1074         ssize_t rv;
1075         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1076
1077         mutex_lock(&nfsd_mutex);
1078         rv = __write_recoverydir(file, buf, size, nn);
1079         mutex_unlock(&nfsd_mutex);
1080         return rv;
1081 }
1082
1083 /**
1084  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1085  *
1086  * Input:
1087  *                      buf:            ignored
1088  *                      size:           zero
1089  * OR
1090  *
1091  * Input:
1092  *                      buf:            any value
1093  *                      size:           non-zero length of C string in @buf
1094  * Output:
1095  *                      passed-in buffer filled with "Y" or "N" with a newline
1096  *                      and NULL-terminated C string. This indicates whether
1097  *                      the grace period has ended in the current net
1098  *                      namespace. Return code is the size in bytes of the
1099  *                      string. Writing a string that starts with 'Y', 'y', or
1100  *                      '1' to the file will end the grace period for nfsd's v4
1101  *                      lock manager.
1102  */
1103 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1104 {
1105         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1106
1107         if (size > 0) {
1108                 switch(buf[0]) {
1109                 case 'Y':
1110                 case 'y':
1111                 case '1':
1112                         if (!nn->nfsd_serv)
1113                                 return -EBUSY;
1114                         nfsd4_end_grace(nn);
1115                         break;
1116                 default:
1117                         return -EINVAL;
1118                 }
1119         }
1120
1121         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1122                          nn->grace_ended ? 'Y' : 'N');
1123 }
1124
1125 #endif
1126
1127 /*----------------------------------------------------------------------------*/
1128 /*
1129  *      populating the filesystem.
1130  */
1131
1132 static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
1133 {
1134         static struct tree_descr nfsd_files[] = {
1135                 [NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1136                 [NFSD_Export_features] = {"export_features",
1137                                         &export_features_operations, S_IRUGO},
1138                 [NFSD_FO_UnlockIP] = {"unlock_ip",
1139                                         &transaction_ops, S_IWUSR|S_IRUSR},
1140                 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1141                                         &transaction_ops, S_IWUSR|S_IRUSR},
1142                 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1143                 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1144                 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1145                 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1146                 [NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
1147                 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1148                 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1149                 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1150                 [NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1151 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1152                 [NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1153 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1154 #ifdef CONFIG_NFSD_V4
1155                 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1156                 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1157                 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1158                 [NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1159 #endif
1160                 /* last one */ {""}
1161         };
1162         struct net *net = data;
1163         int ret;
1164
1165         ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1166         if (ret)
1167                 return ret;
1168         sb->s_fs_info = get_net(net);
1169         return 0;
1170 }
1171
1172 static struct dentry *nfsd_mount(struct file_system_type *fs_type,
1173         int flags, const char *dev_name, void *data)
1174 {
1175         return mount_ns(fs_type, flags, current->nsproxy->net_ns, nfsd_fill_super);
1176 }
1177
1178 static void nfsd_umount(struct super_block *sb)
1179 {
1180         struct net *net = sb->s_fs_info;
1181
1182         kill_litter_super(sb);
1183         put_net(net);
1184 }
1185
1186 static struct file_system_type nfsd_fs_type = {
1187         .owner          = THIS_MODULE,
1188         .name           = "nfsd",
1189         .mount          = nfsd_mount,
1190         .kill_sb        = nfsd_umount,
1191 };
1192 MODULE_ALIAS_FS("nfsd");
1193
1194 #ifdef CONFIG_PROC_FS
1195 static int create_proc_exports_entry(void)
1196 {
1197         struct proc_dir_entry *entry;
1198
1199         entry = proc_mkdir("fs/nfs", NULL);
1200         if (!entry)
1201                 return -ENOMEM;
1202         entry = proc_create("exports", 0, entry,
1203                                  &exports_proc_operations);
1204         if (!entry) {
1205                 remove_proc_entry("fs/nfs", NULL);
1206                 return -ENOMEM;
1207         }
1208         return 0;
1209 }
1210 #else /* CONFIG_PROC_FS */
1211 static int create_proc_exports_entry(void)
1212 {
1213         return 0;
1214 }
1215 #endif
1216
1217 int nfsd_net_id;
1218
1219 static __net_init int nfsd_init_net(struct net *net)
1220 {
1221         int retval;
1222         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1223
1224         retval = nfsd_export_init(net);
1225         if (retval)
1226                 goto out_export_error;
1227         retval = nfsd_idmap_init(net);
1228         if (retval)
1229                 goto out_idmap_error;
1230         nn->nfsd4_lease = 90;   /* default lease time */
1231         nn->nfsd4_grace = 90;
1232         return 0;
1233
1234 out_idmap_error:
1235         nfsd_export_shutdown(net);
1236 out_export_error:
1237         return retval;
1238 }
1239
1240 static __net_exit void nfsd_exit_net(struct net *net)
1241 {
1242         nfsd_idmap_shutdown(net);
1243         nfsd_export_shutdown(net);
1244 }
1245
1246 static struct pernet_operations nfsd_net_ops = {
1247         .init = nfsd_init_net,
1248         .exit = nfsd_exit_net,
1249         .id   = &nfsd_net_id,
1250         .size = sizeof(struct nfsd_net),
1251 };
1252
1253 static int __init init_nfsd(void)
1254 {
1255         int retval;
1256         printk(KERN_INFO "Installing knfsd (copyright (C) 1996 okir@monad.swb.de).\n");
1257
1258         retval = register_pernet_subsys(&nfsd_net_ops);
1259         if (retval < 0)
1260                 return retval;
1261         retval = register_cld_notifier();
1262         if (retval)
1263                 goto out_unregister_pernet;
1264         retval = nfsd4_init_slabs();
1265         if (retval)
1266                 goto out_unregister_notifier;
1267         retval = nfsd4_init_pnfs();
1268         if (retval)
1269                 goto out_free_slabs;
1270         retval = nfsd_fault_inject_init(); /* nfsd fault injection controls */
1271         if (retval)
1272                 goto out_exit_pnfs;
1273         nfsd_stat_init();       /* Statistics */
1274         retval = nfsd_reply_cache_init();
1275         if (retval)
1276                 goto out_free_stat;
1277         nfsd_lockd_init();      /* lockd->nfsd callbacks */
1278         retval = create_proc_exports_entry();
1279         if (retval)
1280                 goto out_free_lockd;
1281         retval = register_filesystem(&nfsd_fs_type);
1282         if (retval)
1283                 goto out_free_all;
1284         return 0;
1285 out_free_all:
1286         remove_proc_entry("fs/nfs/exports", NULL);
1287         remove_proc_entry("fs/nfs", NULL);
1288 out_free_lockd:
1289         nfsd_lockd_shutdown();
1290         nfsd_reply_cache_shutdown();
1291 out_free_stat:
1292         nfsd_stat_shutdown();
1293         nfsd_fault_inject_cleanup();
1294 out_exit_pnfs:
1295         nfsd4_exit_pnfs();
1296 out_free_slabs:
1297         nfsd4_free_slabs();
1298 out_unregister_notifier:
1299         unregister_cld_notifier();
1300 out_unregister_pernet:
1301         unregister_pernet_subsys(&nfsd_net_ops);
1302         return retval;
1303 }
1304
1305 static void __exit exit_nfsd(void)
1306 {
1307         nfsd_reply_cache_shutdown();
1308         remove_proc_entry("fs/nfs/exports", NULL);
1309         remove_proc_entry("fs/nfs", NULL);
1310         nfsd_stat_shutdown();
1311         nfsd_lockd_shutdown();
1312         nfsd4_free_slabs();
1313         nfsd4_exit_pnfs();
1314         nfsd_fault_inject_cleanup();
1315         unregister_filesystem(&nfsd_fs_type);
1316         unregister_cld_notifier();
1317         unregister_pernet_subsys(&nfsd_net_ops);
1318 }
1319
1320 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1321 MODULE_LICENSE("GPL");
1322 module_init(init_nfsd)
1323 module_exit(exit_nfsd)