]> git.kernelconcepts.de Git - karo-tx-linux.git/blob - fs/nfsd/nfsctl.c
Merge tag 'staging-3.16-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
[karo-tx-linux.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
25 /*
26  *      We have a single directory with several nodes in it.
27  */
28 enum {
29         NFSD_Root = 1,
30         NFSD_List,
31         NFSD_Export_features,
32         NFSD_Fh,
33         NFSD_FO_UnlockIP,
34         NFSD_FO_UnlockFS,
35         NFSD_Threads,
36         NFSD_Pool_Threads,
37         NFSD_Pool_Stats,
38         NFSD_Reply_Cache_Stats,
39         NFSD_Versions,
40         NFSD_Ports,
41         NFSD_MaxBlkSize,
42         NFSD_SupportedEnctypes,
43         /*
44          * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
45          * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
46          */
47 #ifdef CONFIG_NFSD_V4
48         NFSD_Leasetime,
49         NFSD_Gracetime,
50         NFSD_RecoveryDir,
51 #endif
52 };
53
54 /*
55  * write() for these nodes.
56  */
57 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
58 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
59 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
60 static ssize_t write_threads(struct file *file, char *buf, size_t size);
61 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
62 static ssize_t write_versions(struct file *file, char *buf, size_t size);
63 static ssize_t write_ports(struct file *file, char *buf, size_t size);
64 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
65 #ifdef CONFIG_NFSD_V4
66 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
67 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
68 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
69 #endif
70
71 static ssize_t (*write_op[])(struct file *, char *, size_t) = {
72         [NFSD_Fh] = write_filehandle,
73         [NFSD_FO_UnlockIP] = write_unlock_ip,
74         [NFSD_FO_UnlockFS] = write_unlock_fs,
75         [NFSD_Threads] = write_threads,
76         [NFSD_Pool_Threads] = write_pool_threads,
77         [NFSD_Versions] = write_versions,
78         [NFSD_Ports] = write_ports,
79         [NFSD_MaxBlkSize] = write_maxblksize,
80 #ifdef CONFIG_NFSD_V4
81         [NFSD_Leasetime] = write_leasetime,
82         [NFSD_Gracetime] = write_gracetime,
83         [NFSD_RecoveryDir] = write_recoverydir,
84 #endif
85 };
86
87 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
88 {
89         ino_t ino =  file_inode(file)->i_ino;
90         char *data;
91         ssize_t rv;
92
93         if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
94                 return -EINVAL;
95
96         data = simple_transaction_get(file, buf, size);
97         if (IS_ERR(data))
98                 return PTR_ERR(data);
99
100         rv =  write_op[ino](file, data, size);
101         if (rv >= 0) {
102                 simple_transaction_set(file, rv);
103                 rv = size;
104         }
105         return rv;
106 }
107
108 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
109 {
110         if (! file->private_data) {
111                 /* An attempt to read a transaction file without writing
112                  * causes a 0-byte write so that the file can return
113                  * state information
114                  */
115                 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
116                 if (rv < 0)
117                         return rv;
118         }
119         return simple_transaction_read(file, buf, size, pos);
120 }
121
122 static const struct file_operations transaction_ops = {
123         .write          = nfsctl_transaction_write,
124         .read           = nfsctl_transaction_read,
125         .release        = simple_transaction_release,
126         .llseek         = default_llseek,
127 };
128
129 static int exports_net_open(struct net *net, struct file *file)
130 {
131         int err;
132         struct seq_file *seq;
133         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
134
135         err = seq_open(file, &nfs_exports_op);
136         if (err)
137                 return err;
138
139         seq = file->private_data;
140         seq->private = nn->svc_export_cache;
141         return 0;
142 }
143
144 static int exports_proc_open(struct inode *inode, struct file *file)
145 {
146         return exports_net_open(current->nsproxy->net_ns, file);
147 }
148
149 static const struct file_operations exports_proc_operations = {
150         .open           = exports_proc_open,
151         .read           = seq_read,
152         .llseek         = seq_lseek,
153         .release        = seq_release,
154         .owner          = THIS_MODULE,
155 };
156
157 static int exports_nfsd_open(struct inode *inode, struct file *file)
158 {
159         return exports_net_open(inode->i_sb->s_fs_info, file);
160 }
161
162 static const struct file_operations exports_nfsd_operations = {
163         .open           = exports_nfsd_open,
164         .read           = seq_read,
165         .llseek         = seq_lseek,
166         .release        = seq_release,
167         .owner          = THIS_MODULE,
168 };
169
170 static int export_features_show(struct seq_file *m, void *v)
171 {
172         seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
173         return 0;
174 }
175
176 static int export_features_open(struct inode *inode, struct file *file)
177 {
178         return single_open(file, export_features_show, NULL);
179 }
180
181 static const struct file_operations export_features_operations = {
182         .open           = export_features_open,
183         .read           = seq_read,
184         .llseek         = seq_lseek,
185         .release        = single_release,
186 };
187
188 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
189 static int supported_enctypes_show(struct seq_file *m, void *v)
190 {
191         seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
192         return 0;
193 }
194
195 static int supported_enctypes_open(struct inode *inode, struct file *file)
196 {
197         return single_open(file, supported_enctypes_show, NULL);
198 }
199
200 static const struct file_operations supported_enctypes_ops = {
201         .open           = supported_enctypes_open,
202         .read           = seq_read,
203         .llseek         = seq_lseek,
204         .release        = single_release,
205 };
206 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
207
208 static const struct file_operations pool_stats_operations = {
209         .open           = nfsd_pool_stats_open,
210         .read           = seq_read,
211         .llseek         = seq_lseek,
212         .release        = nfsd_pool_stats_release,
213         .owner          = THIS_MODULE,
214 };
215
216 static struct file_operations reply_cache_stats_operations = {
217         .open           = nfsd_reply_cache_stats_open,
218         .read           = seq_read,
219         .llseek         = seq_lseek,
220         .release        = single_release,
221 };
222
223 /*----------------------------------------------------------------------------*/
224 /*
225  * payload - write methods
226  */
227
228
229 /**
230  * write_unlock_ip - Release all locks used by a client
231  *
232  * Experimental.
233  *
234  * Input:
235  *                      buf:    '\n'-terminated C string containing a
236  *                              presentation format IP address
237  *                      size:   length of C string in @buf
238  * Output:
239  *      On success:     returns zero if all specified locks were released;
240  *                      returns one if one or more locks were not released
241  *      On error:       return code is negative errno value
242  */
243 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
244 {
245         struct sockaddr_storage address;
246         struct sockaddr *sap = (struct sockaddr *)&address;
247         size_t salen = sizeof(address);
248         char *fo_path;
249         struct net *net = file->f_dentry->d_sb->s_fs_info;
250
251         /* sanity check */
252         if (size == 0)
253                 return -EINVAL;
254
255         if (buf[size-1] != '\n')
256                 return -EINVAL;
257
258         fo_path = buf;
259         if (qword_get(&buf, fo_path, size) < 0)
260                 return -EINVAL;
261
262         if (rpc_pton(net, fo_path, size, sap, salen) == 0)
263                 return -EINVAL;
264
265         return nlmsvc_unlock_all_by_ip(sap);
266 }
267
268 /**
269  * write_unlock_fs - Release all locks on a local file system
270  *
271  * Experimental.
272  *
273  * Input:
274  *                      buf:    '\n'-terminated C string containing the
275  *                              absolute pathname of a local file system
276  *                      size:   length of C string in @buf
277  * Output:
278  *      On success:     returns zero if all specified locks were released;
279  *                      returns one if one or more locks were not released
280  *      On error:       return code is negative errno value
281  */
282 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
283 {
284         struct path path;
285         char *fo_path;
286         int error;
287
288         /* sanity check */
289         if (size == 0)
290                 return -EINVAL;
291
292         if (buf[size-1] != '\n')
293                 return -EINVAL;
294
295         fo_path = buf;
296         if (qword_get(&buf, fo_path, size) < 0)
297                 return -EINVAL;
298
299         error = kern_path(fo_path, 0, &path);
300         if (error)
301                 return error;
302
303         /*
304          * XXX: Needs better sanity checking.  Otherwise we could end up
305          * releasing locks on the wrong file system.
306          *
307          * For example:
308          * 1.  Does the path refer to a directory?
309          * 2.  Is that directory a mount point, or
310          * 3.  Is that directory the root of an exported file system?
311          */
312         error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
313
314         path_put(&path);
315         return error;
316 }
317
318 /**
319  * write_filehandle - Get a variable-length NFS file handle by path
320  *
321  * On input, the buffer contains a '\n'-terminated C string comprised of
322  * three alphanumeric words separated by whitespace.  The string may
323  * contain escape sequences.
324  *
325  * Input:
326  *                      buf:
327  *                              domain:         client domain name
328  *                              path:           export pathname
329  *                              maxsize:        numeric maximum size of
330  *                                              @buf
331  *                      size:   length of C string in @buf
332  * Output:
333  *      On success:     passed-in buffer filled with '\n'-terminated C
334  *                      string containing a ASCII hex text version
335  *                      of the NFS file handle;
336  *                      return code is the size in bytes of the string
337  *      On error:       return code is negative errno value
338  */
339 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
340 {
341         char *dname, *path;
342         int uninitialized_var(maxsize);
343         char *mesg = buf;
344         int len;
345         struct auth_domain *dom;
346         struct knfsd_fh fh;
347         struct net *net = file->f_dentry->d_sb->s_fs_info;
348
349         if (size == 0)
350                 return -EINVAL;
351
352         if (buf[size-1] != '\n')
353                 return -EINVAL;
354         buf[size-1] = 0;
355
356         dname = mesg;
357         len = qword_get(&mesg, dname, size);
358         if (len <= 0)
359                 return -EINVAL;
360         
361         path = dname+len+1;
362         len = qword_get(&mesg, path, size);
363         if (len <= 0)
364                 return -EINVAL;
365
366         len = get_int(&mesg, &maxsize);
367         if (len)
368                 return len;
369
370         if (maxsize < NFS_FHSIZE)
371                 return -EINVAL;
372         if (maxsize > NFS3_FHSIZE)
373                 maxsize = NFS3_FHSIZE;
374
375         if (qword_get(&mesg, mesg, size)>0)
376                 return -EINVAL;
377
378         /* we have all the words, they are in buf.. */
379         dom = unix_domain_find(dname);
380         if (!dom)
381                 return -ENOMEM;
382
383         len = exp_rootfh(net, dom, path, &fh,  maxsize);
384         auth_domain_put(dom);
385         if (len)
386                 return len;
387         
388         mesg = buf;
389         len = SIMPLE_TRANSACTION_LIMIT;
390         qword_addhex(&mesg, &len, (char*)&fh.fh_base, fh.fh_size);
391         mesg[-1] = '\n';
392         return mesg - buf;      
393 }
394
395 /**
396  * write_threads - Start NFSD, or report the current number of running threads
397  *
398  * Input:
399  *                      buf:            ignored
400  *                      size:           zero
401  * Output:
402  *      On success:     passed-in buffer filled with '\n'-terminated C
403  *                      string numeric value representing the number of
404  *                      running NFSD threads;
405  *                      return code is the size in bytes of the string
406  *      On error:       return code is zero
407  *
408  * OR
409  *
410  * Input:
411  *                      buf:            C string containing an unsigned
412  *                                      integer value representing the
413  *                                      number of NFSD threads to start
414  *                      size:           non-zero length of C string in @buf
415  * Output:
416  *      On success:     NFS service is started;
417  *                      passed-in buffer filled with '\n'-terminated C
418  *                      string numeric value representing the number of
419  *                      running NFSD threads;
420  *                      return code is the size in bytes of the string
421  *      On error:       return code is zero or a negative errno value
422  */
423 static ssize_t write_threads(struct file *file, char *buf, size_t size)
424 {
425         char *mesg = buf;
426         int rv;
427         struct net *net = file->f_dentry->d_sb->s_fs_info;
428
429         if (size > 0) {
430                 int newthreads;
431                 rv = get_int(&mesg, &newthreads);
432                 if (rv)
433                         return rv;
434                 if (newthreads < 0)
435                         return -EINVAL;
436                 rv = nfsd_svc(newthreads, net);
437                 if (rv < 0)
438                         return rv;
439         } else
440                 rv = nfsd_nrthreads(net);
441
442         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
443 }
444
445 /**
446  * write_pool_threads - Set or report the current number of threads per pool
447  *
448  * Input:
449  *                      buf:            ignored
450  *                      size:           zero
451  *
452  * OR
453  *
454  * Input:
455  *                      buf:            C string containing whitespace-
456  *                                      separated unsigned integer values
457  *                                      representing the number of NFSD
458  *                                      threads to start in each pool
459  *                      size:           non-zero length of C string in @buf
460  * Output:
461  *      On success:     passed-in buffer filled with '\n'-terminated C
462  *                      string containing integer values representing the
463  *                      number of NFSD threads in each pool;
464  *                      return code is the size in bytes of the string
465  *      On error:       return code is zero or a negative errno value
466  */
467 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
468 {
469         /* if size > 0, look for an array of number of threads per node
470          * and apply them  then write out number of threads per node as reply
471          */
472         char *mesg = buf;
473         int i;
474         int rv;
475         int len;
476         int npools;
477         int *nthreads;
478         struct net *net = file->f_dentry->d_sb->s_fs_info;
479
480         mutex_lock(&nfsd_mutex);
481         npools = nfsd_nrpools(net);
482         if (npools == 0) {
483                 /*
484                  * NFS is shut down.  The admin can start it by
485                  * writing to the threads file but NOT the pool_threads
486                  * file, sorry.  Report zero threads.
487                  */
488                 mutex_unlock(&nfsd_mutex);
489                 strcpy(buf, "0\n");
490                 return strlen(buf);
491         }
492
493         nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
494         rv = -ENOMEM;
495         if (nthreads == NULL)
496                 goto out_free;
497
498         if (size > 0) {
499                 for (i = 0; i < npools; i++) {
500                         rv = get_int(&mesg, &nthreads[i]);
501                         if (rv == -ENOENT)
502                                 break;          /* fewer numbers than pools */
503                         if (rv)
504                                 goto out_free;  /* syntax error */
505                         rv = -EINVAL;
506                         if (nthreads[i] < 0)
507                                 goto out_free;
508                 }
509                 rv = nfsd_set_nrthreads(i, nthreads, net);
510                 if (rv)
511                         goto out_free;
512         }
513
514         rv = nfsd_get_nrthreads(npools, nthreads, net);
515         if (rv)
516                 goto out_free;
517
518         mesg = buf;
519         size = SIMPLE_TRANSACTION_LIMIT;
520         for (i = 0; i < npools && size > 0; i++) {
521                 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
522                 len = strlen(mesg);
523                 size -= len;
524                 mesg += len;
525         }
526         rv = mesg - buf;
527 out_free:
528         kfree(nthreads);
529         mutex_unlock(&nfsd_mutex);
530         return rv;
531 }
532
533 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
534 {
535         char *mesg = buf;
536         char *vers, *minorp, sign;
537         int len, num, remaining;
538         unsigned minor;
539         ssize_t tlen = 0;
540         char *sep;
541         struct net *net = file->f_dentry->d_sb->s_fs_info;
542         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
543
544         if (size>0) {
545                 if (nn->nfsd_serv)
546                         /* Cannot change versions without updating
547                          * nn->nfsd_serv->sv_xdrsize, and reallocing
548                          * rq_argp and rq_resp
549                          */
550                         return -EBUSY;
551                 if (buf[size-1] != '\n')
552                         return -EINVAL;
553                 buf[size-1] = 0;
554
555                 vers = mesg;
556                 len = qword_get(&mesg, vers, size);
557                 if (len <= 0) return -EINVAL;
558                 do {
559                         sign = *vers;
560                         if (sign == '+' || sign == '-')
561                                 num = simple_strtol((vers+1), &minorp, 0);
562                         else
563                                 num = simple_strtol(vers, &minorp, 0);
564                         if (*minorp == '.') {
565                                 if (num != 4)
566                                         return -EINVAL;
567                                 minor = simple_strtoul(minorp+1, NULL, 0);
568                                 if (minor == 0)
569                                         return -EINVAL;
570                                 if (nfsd_minorversion(minor, sign == '-' ?
571                                                      NFSD_CLEAR : NFSD_SET) < 0)
572                                         return -EINVAL;
573                                 goto next;
574                         }
575                         switch(num) {
576                         case 2:
577                         case 3:
578                         case 4:
579                                 nfsd_vers(num, sign == '-' ? NFSD_CLEAR : NFSD_SET);
580                                 break;
581                         default:
582                                 return -EINVAL;
583                         }
584                 next:
585                         vers += len + 1;
586                 } while ((len = qword_get(&mesg, vers, size)) > 0);
587                 /* If all get turned off, turn them back on, as
588                  * having no versions is BAD
589                  */
590                 nfsd_reset_versions();
591         }
592
593         /* Now write current state into reply buffer */
594         len = 0;
595         sep = "";
596         remaining = SIMPLE_TRANSACTION_LIMIT;
597         for (num=2 ; num <= 4 ; num++)
598                 if (nfsd_vers(num, NFSD_AVAIL)) {
599                         len = snprintf(buf, remaining, "%s%c%d", sep,
600                                        nfsd_vers(num, NFSD_TEST)?'+':'-',
601                                        num);
602                         sep = " ";
603
604                         if (len > remaining)
605                                 break;
606                         remaining -= len;
607                         buf += len;
608                         tlen += len;
609                 }
610         if (nfsd_vers(4, NFSD_AVAIL))
611                 for (minor = 1; minor <= NFSD_SUPPORTED_MINOR_VERSION;
612                      minor++) {
613                         len = snprintf(buf, remaining, " %c4.%u",
614                                         (nfsd_vers(4, NFSD_TEST) &&
615                                          nfsd_minorversion(minor, NFSD_TEST)) ?
616                                                 '+' : '-',
617                                         minor);
618
619                         if (len > remaining)
620                                 break;
621                         remaining -= len;
622                         buf += len;
623                         tlen += len;
624                 }
625
626         len = snprintf(buf, remaining, "\n");
627         if (len > remaining)
628                 return -EINVAL;
629         return tlen + len;
630 }
631
632 /**
633  * write_versions - Set or report the available NFS protocol versions
634  *
635  * Input:
636  *                      buf:            ignored
637  *                      size:           zero
638  * Output:
639  *      On success:     passed-in buffer filled with '\n'-terminated C
640  *                      string containing positive or negative integer
641  *                      values representing the current status of each
642  *                      protocol version;
643  *                      return code is the size in bytes of the string
644  *      On error:       return code is zero or a negative errno value
645  *
646  * OR
647  *
648  * Input:
649  *                      buf:            C string containing whitespace-
650  *                                      separated positive or negative
651  *                                      integer values representing NFS
652  *                                      protocol versions to enable ("+n")
653  *                                      or disable ("-n")
654  *                      size:           non-zero length of C string in @buf
655  * Output:
656  *      On success:     status of zero or more protocol versions has
657  *                      been updated; passed-in buffer filled with
658  *                      '\n'-terminated C string containing positive
659  *                      or negative integer values representing the
660  *                      current status of each protocol version;
661  *                      return code is the size in bytes of the string
662  *      On error:       return code is zero or a negative errno value
663  */
664 static ssize_t write_versions(struct file *file, char *buf, size_t size)
665 {
666         ssize_t rv;
667
668         mutex_lock(&nfsd_mutex);
669         rv = __write_versions(file, buf, size);
670         mutex_unlock(&nfsd_mutex);
671         return rv;
672 }
673
674 /*
675  * Zero-length write.  Return a list of NFSD's current listener
676  * transports.
677  */
678 static ssize_t __write_ports_names(char *buf, struct net *net)
679 {
680         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
681
682         if (nn->nfsd_serv == NULL)
683                 return 0;
684         return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
685 }
686
687 /*
688  * A single 'fd' number was written, in which case it must be for
689  * a socket of a supported family/protocol, and we use it as an
690  * nfsd listener.
691  */
692 static ssize_t __write_ports_addfd(char *buf, struct net *net)
693 {
694         char *mesg = buf;
695         int fd, err;
696         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
697
698         err = get_int(&mesg, &fd);
699         if (err != 0 || fd < 0)
700                 return -EINVAL;
701
702         if (svc_alien_sock(net, fd)) {
703                 printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
704                 return -EINVAL;
705         }
706
707         err = nfsd_create_serv(net);
708         if (err != 0)
709                 return err;
710
711         err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT);
712         if (err < 0) {
713                 nfsd_destroy(net);
714                 return err;
715         }
716
717         /* Decrease the count, but don't shut down the service */
718         nn->nfsd_serv->sv_nrthreads--;
719         return err;
720 }
721
722 /*
723  * A transport listener is added by writing it's transport name and
724  * a port number.
725  */
726 static ssize_t __write_ports_addxprt(char *buf, struct net *net)
727 {
728         char transport[16];
729         struct svc_xprt *xprt;
730         int port, err;
731         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
732
733         if (sscanf(buf, "%15s %5u", transport, &port) != 2)
734                 return -EINVAL;
735
736         if (port < 1 || port > USHRT_MAX)
737                 return -EINVAL;
738
739         err = nfsd_create_serv(net);
740         if (err != 0)
741                 return err;
742
743         err = svc_create_xprt(nn->nfsd_serv, transport, net,
744                                 PF_INET, port, SVC_SOCK_ANONYMOUS);
745         if (err < 0)
746                 goto out_err;
747
748         err = svc_create_xprt(nn->nfsd_serv, transport, net,
749                                 PF_INET6, port, SVC_SOCK_ANONYMOUS);
750         if (err < 0 && err != -EAFNOSUPPORT)
751                 goto out_close;
752
753         /* Decrease the count, but don't shut down the service */
754         nn->nfsd_serv->sv_nrthreads--;
755         return 0;
756 out_close:
757         xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
758         if (xprt != NULL) {
759                 svc_close_xprt(xprt);
760                 svc_xprt_put(xprt);
761         }
762 out_err:
763         nfsd_destroy(net);
764         return err;
765 }
766
767 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
768                              struct net *net)
769 {
770         if (size == 0)
771                 return __write_ports_names(buf, net);
772
773         if (isdigit(buf[0]))
774                 return __write_ports_addfd(buf, net);
775
776         if (isalpha(buf[0]))
777                 return __write_ports_addxprt(buf, net);
778
779         return -EINVAL;
780 }
781
782 /**
783  * write_ports - Pass a socket file descriptor or transport name to listen on
784  *
785  * Input:
786  *                      buf:            ignored
787  *                      size:           zero
788  * Output:
789  *      On success:     passed-in buffer filled with a '\n'-terminated C
790  *                      string containing a whitespace-separated list of
791  *                      named NFSD listeners;
792  *                      return code is the size in bytes of the string
793  *      On error:       return code is zero or a negative errno value
794  *
795  * OR
796  *
797  * Input:
798  *                      buf:            C string containing an unsigned
799  *                                      integer value representing a bound
800  *                                      but unconnected socket that is to be
801  *                                      used as an NFSD listener; listen(3)
802  *                                      must be called for a SOCK_STREAM
803  *                                      socket, otherwise it is ignored
804  *                      size:           non-zero length of C string in @buf
805  * Output:
806  *      On success:     NFS service is started;
807  *                      passed-in buffer filled with a '\n'-terminated C
808  *                      string containing a unique alphanumeric name of
809  *                      the listener;
810  *                      return code is the size in bytes of the string
811  *      On error:       return code is a negative errno value
812  *
813  * OR
814  *
815  * Input:
816  *                      buf:            C string containing a transport
817  *                                      name and an unsigned integer value
818  *                                      representing the port to listen on,
819  *                                      separated by whitespace
820  *                      size:           non-zero length of C string in @buf
821  * Output:
822  *      On success:     returns zero; NFS service is started
823  *      On error:       return code is a negative errno value
824  */
825 static ssize_t write_ports(struct file *file, char *buf, size_t size)
826 {
827         ssize_t rv;
828         struct net *net = file->f_dentry->d_sb->s_fs_info;
829
830         mutex_lock(&nfsd_mutex);
831         rv = __write_ports(file, buf, size, net);
832         mutex_unlock(&nfsd_mutex);
833         return rv;
834 }
835
836
837 int nfsd_max_blksize;
838
839 /**
840  * write_maxblksize - Set or report the current NFS blksize
841  *
842  * Input:
843  *                      buf:            ignored
844  *                      size:           zero
845  *
846  * OR
847  *
848  * Input:
849  *                      buf:            C string containing an unsigned
850  *                                      integer value representing the new
851  *                                      NFS blksize
852  *                      size:           non-zero length of C string in @buf
853  * Output:
854  *      On success:     passed-in buffer filled with '\n'-terminated C string
855  *                      containing numeric value of the current NFS blksize
856  *                      setting;
857  *                      return code is the size in bytes of the string
858  *      On error:       return code is zero or a negative errno value
859  */
860 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
861 {
862         char *mesg = buf;
863         struct net *net = file->f_dentry->d_sb->s_fs_info;
864         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
865
866         if (size > 0) {
867                 int bsize;
868                 int rv = get_int(&mesg, &bsize);
869                 if (rv)
870                         return rv;
871                 /* force bsize into allowed range and
872                  * required alignment.
873                  */
874                 if (bsize < 1024)
875                         bsize = 1024;
876                 if (bsize > NFSSVC_MAXBLKSIZE)
877                         bsize = NFSSVC_MAXBLKSIZE;
878                 bsize &= ~(1024-1);
879                 mutex_lock(&nfsd_mutex);
880                 if (nn->nfsd_serv) {
881                         mutex_unlock(&nfsd_mutex);
882                         return -EBUSY;
883                 }
884                 nfsd_max_blksize = bsize;
885                 mutex_unlock(&nfsd_mutex);
886         }
887
888         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
889                                                         nfsd_max_blksize);
890 }
891
892 #ifdef CONFIG_NFSD_V4
893 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
894                                   time_t *time, struct nfsd_net *nn)
895 {
896         char *mesg = buf;
897         int rv, i;
898
899         if (size > 0) {
900                 if (nn->nfsd_serv)
901                         return -EBUSY;
902                 rv = get_int(&mesg, &i);
903                 if (rv)
904                         return rv;
905                 /*
906                  * Some sanity checking.  We don't have a reason for
907                  * these particular numbers, but problems with the
908                  * extremes are:
909                  *      - Too short: the briefest network outage may
910                  *        cause clients to lose all their locks.  Also,
911                  *        the frequent polling may be wasteful.
912                  *      - Too long: do you really want reboot recovery
913                  *        to take more than an hour?  Or to make other
914                  *        clients wait an hour before being able to
915                  *        revoke a dead client's locks?
916                  */
917                 if (i < 10 || i > 3600)
918                         return -EINVAL;
919                 *time = i;
920         }
921
922         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%ld\n", *time);
923 }
924
925 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
926                                 time_t *time, struct nfsd_net *nn)
927 {
928         ssize_t rv;
929
930         mutex_lock(&nfsd_mutex);
931         rv = __nfsd4_write_time(file, buf, size, time, nn);
932         mutex_unlock(&nfsd_mutex);
933         return rv;
934 }
935
936 /**
937  * write_leasetime - Set or report the current NFSv4 lease time
938  *
939  * Input:
940  *                      buf:            ignored
941  *                      size:           zero
942  *
943  * OR
944  *
945  * Input:
946  *                      buf:            C string containing an unsigned
947  *                                      integer value representing the new
948  *                                      NFSv4 lease expiry time
949  *                      size:           non-zero length of C string in @buf
950  * Output:
951  *      On success:     passed-in buffer filled with '\n'-terminated C
952  *                      string containing unsigned integer value of the
953  *                      current lease expiry time;
954  *                      return code is the size in bytes of the string
955  *      On error:       return code is zero or a negative errno value
956  */
957 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
958 {
959         struct net *net = file->f_dentry->d_sb->s_fs_info;
960         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
961         return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
962 }
963
964 /**
965  * write_gracetime - Set or report current NFSv4 grace period time
966  *
967  * As above, but sets the time of the NFSv4 grace period.
968  *
969  * Note this should never be set to less than the *previous*
970  * lease-period time, but we don't try to enforce this.  (In the common
971  * case (a new boot), we don't know what the previous lease time was
972  * anyway.)
973  */
974 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
975 {
976         struct net *net = file->f_dentry->d_sb->s_fs_info;
977         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
978         return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
979 }
980
981 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
982                                    struct nfsd_net *nn)
983 {
984         char *mesg = buf;
985         char *recdir;
986         int len, status;
987
988         if (size > 0) {
989                 if (nn->nfsd_serv)
990                         return -EBUSY;
991                 if (size > PATH_MAX || buf[size-1] != '\n')
992                         return -EINVAL;
993                 buf[size-1] = 0;
994
995                 recdir = mesg;
996                 len = qword_get(&mesg, recdir, size);
997                 if (len <= 0)
998                         return -EINVAL;
999
1000                 status = nfs4_reset_recoverydir(recdir);
1001                 if (status)
1002                         return status;
1003         }
1004
1005         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1006                                                         nfs4_recoverydir());
1007 }
1008
1009 /**
1010  * write_recoverydir - Set or report the pathname of the recovery directory
1011  *
1012  * Input:
1013  *                      buf:            ignored
1014  *                      size:           zero
1015  *
1016  * OR
1017  *
1018  * Input:
1019  *                      buf:            C string containing the pathname
1020  *                                      of the directory on a local file
1021  *                                      system containing permanent NFSv4
1022  *                                      recovery data
1023  *                      size:           non-zero length of C string in @buf
1024  * Output:
1025  *      On success:     passed-in buffer filled with '\n'-terminated C string
1026  *                      containing the current recovery pathname setting;
1027  *                      return code is the size in bytes of the string
1028  *      On error:       return code is zero or a negative errno value
1029  */
1030 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1031 {
1032         ssize_t rv;
1033         struct net *net = file->f_dentry->d_sb->s_fs_info;
1034         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1035
1036         mutex_lock(&nfsd_mutex);
1037         rv = __write_recoverydir(file, buf, size, nn);
1038         mutex_unlock(&nfsd_mutex);
1039         return rv;
1040 }
1041
1042 #endif
1043
1044 /*----------------------------------------------------------------------------*/
1045 /*
1046  *      populating the filesystem.
1047  */
1048
1049 static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
1050 {
1051         static struct tree_descr nfsd_files[] = {
1052                 [NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1053                 [NFSD_Export_features] = {"export_features",
1054                                         &export_features_operations, S_IRUGO},
1055                 [NFSD_FO_UnlockIP] = {"unlock_ip",
1056                                         &transaction_ops, S_IWUSR|S_IRUSR},
1057                 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1058                                         &transaction_ops, S_IWUSR|S_IRUSR},
1059                 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1060                 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1061                 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1062                 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1063                 [NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
1064                 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1065                 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1066                 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1067 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1068                 [NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1069 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1070 #ifdef CONFIG_NFSD_V4
1071                 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1072                 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1073                 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1074 #endif
1075                 /* last one */ {""}
1076         };
1077         struct net *net = data;
1078         int ret;
1079
1080         ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1081         if (ret)
1082                 return ret;
1083         sb->s_fs_info = get_net(net);
1084         return 0;
1085 }
1086
1087 static struct dentry *nfsd_mount(struct file_system_type *fs_type,
1088         int flags, const char *dev_name, void *data)
1089 {
1090         return mount_ns(fs_type, flags, current->nsproxy->net_ns, nfsd_fill_super);
1091 }
1092
1093 static void nfsd_umount(struct super_block *sb)
1094 {
1095         struct net *net = sb->s_fs_info;
1096
1097         kill_litter_super(sb);
1098         put_net(net);
1099 }
1100
1101 static struct file_system_type nfsd_fs_type = {
1102         .owner          = THIS_MODULE,
1103         .name           = "nfsd",
1104         .mount          = nfsd_mount,
1105         .kill_sb        = nfsd_umount,
1106 };
1107 MODULE_ALIAS_FS("nfsd");
1108
1109 #ifdef CONFIG_PROC_FS
1110 static int create_proc_exports_entry(void)
1111 {
1112         struct proc_dir_entry *entry;
1113
1114         entry = proc_mkdir("fs/nfs", NULL);
1115         if (!entry)
1116                 return -ENOMEM;
1117         entry = proc_create("exports", 0, entry,
1118                                  &exports_proc_operations);
1119         if (!entry) {
1120                 remove_proc_entry("fs/nfs", NULL);
1121                 return -ENOMEM;
1122         }
1123         return 0;
1124 }
1125 #else /* CONFIG_PROC_FS */
1126 static int create_proc_exports_entry(void)
1127 {
1128         return 0;
1129 }
1130 #endif
1131
1132 int nfsd_net_id;
1133
1134 static __net_init int nfsd_init_net(struct net *net)
1135 {
1136         int retval;
1137         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1138
1139         retval = nfsd_export_init(net);
1140         if (retval)
1141                 goto out_export_error;
1142         retval = nfsd_idmap_init(net);
1143         if (retval)
1144                 goto out_idmap_error;
1145         nn->nfsd4_lease = 90;   /* default lease time */
1146         nn->nfsd4_grace = 90;
1147         return 0;
1148
1149 out_idmap_error:
1150         nfsd_export_shutdown(net);
1151 out_export_error:
1152         return retval;
1153 }
1154
1155 static __net_exit void nfsd_exit_net(struct net *net)
1156 {
1157         nfsd_idmap_shutdown(net);
1158         nfsd_export_shutdown(net);
1159 }
1160
1161 static struct pernet_operations nfsd_net_ops = {
1162         .init = nfsd_init_net,
1163         .exit = nfsd_exit_net,
1164         .id   = &nfsd_net_id,
1165         .size = sizeof(struct nfsd_net),
1166 };
1167
1168 static int __init init_nfsd(void)
1169 {
1170         int retval;
1171         printk(KERN_INFO "Installing knfsd (copyright (C) 1996 okir@monad.swb.de).\n");
1172
1173         retval = register_cld_notifier();
1174         if (retval)
1175                 return retval;
1176         retval = register_pernet_subsys(&nfsd_net_ops);
1177         if (retval < 0)
1178                 goto out_unregister_notifier;
1179         retval = nfsd4_init_slabs();
1180         if (retval)
1181                 goto out_unregister_pernet;
1182         retval = nfsd_fault_inject_init(); /* nfsd fault injection controls */
1183         if (retval)
1184                 goto out_free_slabs;
1185         nfsd_stat_init();       /* Statistics */
1186         retval = nfsd_reply_cache_init();
1187         if (retval)
1188                 goto out_free_stat;
1189         nfsd_lockd_init();      /* lockd->nfsd callbacks */
1190         retval = create_proc_exports_entry();
1191         if (retval)
1192                 goto out_free_lockd;
1193         retval = register_filesystem(&nfsd_fs_type);
1194         if (retval)
1195                 goto out_free_all;
1196         return 0;
1197 out_free_all:
1198         remove_proc_entry("fs/nfs/exports", NULL);
1199         remove_proc_entry("fs/nfs", NULL);
1200 out_free_lockd:
1201         nfsd_lockd_shutdown();
1202         nfsd_reply_cache_shutdown();
1203 out_free_stat:
1204         nfsd_stat_shutdown();
1205         nfsd_fault_inject_cleanup();
1206 out_free_slabs:
1207         nfsd4_free_slabs();
1208 out_unregister_pernet:
1209         unregister_pernet_subsys(&nfsd_net_ops);
1210 out_unregister_notifier:
1211         unregister_cld_notifier();
1212         return retval;
1213 }
1214
1215 static void __exit exit_nfsd(void)
1216 {
1217         nfsd_reply_cache_shutdown();
1218         remove_proc_entry("fs/nfs/exports", NULL);
1219         remove_proc_entry("fs/nfs", NULL);
1220         nfsd_stat_shutdown();
1221         nfsd_lockd_shutdown();
1222         nfsd4_free_slabs();
1223         nfsd_fault_inject_cleanup();
1224         unregister_filesystem(&nfsd_fs_type);
1225         unregister_pernet_subsys(&nfsd_net_ops);
1226         unregister_cld_notifier();
1227 }
1228
1229 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1230 MODULE_LICENSE("GPL");
1231 module_init(init_nfsd)
1232 module_exit(exit_nfsd)