fs/nfsd/nfsctl.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * Syscall interface to knfsd.
   4  *
   5  * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
   6  */
   7
   8 #include <linux/slab.h>
   9 #include <linux/namei.h>
  10 #include <linux/ctype.h>
  11 #include <linux/fs_context.h>
  12
  13 #include <linux/sunrpc/svcsock.h>
  14 #include <linux/lockd/lockd.h>
  15 #include <linux/sunrpc/addr.h>
  16 #include <linux/sunrpc/gss_api.h>
  17 #include <linux/sunrpc/gss_krb5_enctypes.h>
  18 #include <linux/sunrpc/rpc_pipe_fs.h>
  19 #include <linux/module.h>
  20 #include <linux/fsnotify.h>
  21
  22 #include "idmap.h"
  23 #include "nfsd.h"
  24 #include "cache.h"
  25 #include "state.h"
  26 #include "netns.h"
  27 #include "pnfs.h"
  28 #include "filecache.h"
  29
  30 /*
  31  *      We have a single directory with several nodes in it.
  32  */
  33 enum {
  34         NFSD_Root = 1,
  35         NFSD_List,
  36         NFSD_Export_Stats,
  37         NFSD_Export_features,
  38         NFSD_Fh,
  39         NFSD_FO_UnlockIP,
  40         NFSD_FO_UnlockFS,
  41         NFSD_Threads,
  42         NFSD_Pool_Threads,
  43         NFSD_Pool_Stats,
  44         NFSD_Reply_Cache_Stats,
  45         NFSD_Versions,
  46         NFSD_Ports,
  47         NFSD_MaxBlkSize,
  48         NFSD_MaxConnections,
  49         NFSD_Filecache,
  50         NFSD_SupportedEnctypes,
  51         /*
  52          * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
  53          * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
  54          */
  55 #ifdef CONFIG_NFSD_V4
  56         NFSD_Leasetime,
  57         NFSD_Gracetime,
  58         NFSD_RecoveryDir,
  59         NFSD_V4EndGrace,
  60 #endif
  61         NFSD_MaxReserved
  62 };
  63
  64 /*
  65  * write() for these nodes.
  66  */
  67 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
  68 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
  69 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
  70 static ssize_t write_threads(struct file *file, char *buf, size_t size);
  71 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
  72 static ssize_t write_versions(struct file *file, char *buf, size_t size);
  73 static ssize_t write_ports(struct file *file, char *buf, size_t size);
  74 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
  75 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
  76 #ifdef CONFIG_NFSD_V4
  77 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
  78 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
  79 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
  80 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
  81 #endif
  82
  83 static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
  84         [NFSD_Fh] = write_filehandle,
  85         [NFSD_FO_UnlockIP] = write_unlock_ip,
  86         [NFSD_FO_UnlockFS] = write_unlock_fs,
  87         [NFSD_Threads] = write_threads,
  88         [NFSD_Pool_Threads] = write_pool_threads,
  89         [NFSD_Versions] = write_versions,
  90         [NFSD_Ports] = write_ports,
  91         [NFSD_MaxBlkSize] = write_maxblksize,
  92         [NFSD_MaxConnections] = write_maxconn,
  93 #ifdef CONFIG_NFSD_V4
  94         [NFSD_Leasetime] = write_leasetime,
  95         [NFSD_Gracetime] = write_gracetime,
  96         [NFSD_RecoveryDir] = write_recoverydir,
  97         [NFSD_V4EndGrace] = write_v4_end_grace,
  98 #endif
  99 };
 100
 101 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
 102 {
 103         ino_t ino =  file_inode(file)->i_ino;
 104         char *data;
 105         ssize_t rv;
 106
 107         if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
 108                 return -EINVAL;
 109
 110         data = simple_transaction_get(file, buf, size);
 111         if (IS_ERR(data))
 112                 return PTR_ERR(data);
 113
 114         rv =  write_op[ino](file, data, size);
 115         if (rv >= 0) {
 116                 simple_transaction_set(file, rv);
 117                 rv = size;
 118         }
 119         return rv;
 120 }
 121
 122 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
 123 {
 124         if (! file->private_data) {
 125                 /* An attempt to read a transaction file without writing
 126                  * causes a 0-byte write so that the file can return
 127                  * state information
 128                  */
 129                 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
 130                 if (rv < 0)
 131                         return rv;
 132         }
 133         return simple_transaction_read(file, buf, size, pos);
 134 }
 135
 136 static const struct file_operations transaction_ops = {
 137         .write          = nfsctl_transaction_write,
 138         .read           = nfsctl_transaction_read,
 139         .release        = simple_transaction_release,
 140         .llseek         = default_llseek,
 141 };
 142
 143 static int exports_net_open(struct net *net, struct file *file)
 144 {
 145         int err;
 146         struct seq_file *seq;
 147         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 148
 149         err = seq_open(file, &nfs_exports_op);
 150         if (err)
 151                 return err;
 152
 153         seq = file->private_data;
 154         seq->private = nn->svc_export_cache;
 155         return 0;
 156 }
 157
 158 static int exports_proc_open(struct inode *inode, struct file *file)
 159 {
 160         return exports_net_open(current->nsproxy->net_ns, file);
 161 }
 162
 163 static const struct proc_ops exports_proc_ops = {
 164         .proc_open      = exports_proc_open,
 165         .proc_read      = seq_read,
 166         .proc_lseek     = seq_lseek,
 167         .proc_release   = seq_release,
 168 };
 169
 170 static int exports_nfsd_open(struct inode *inode, struct file *file)
 171 {
 172         return exports_net_open(inode->i_sb->s_fs_info, file);
 173 }
 174
 175 static const struct file_operations exports_nfsd_operations = {
 176         .open           = exports_nfsd_open,
 177         .read           = seq_read,
 178         .llseek         = seq_lseek,
 179         .release        = seq_release,
 180 };
 181
 182 static int export_features_show(struct seq_file *m, void *v)
 183 {
 184         seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
 185         return 0;
 186 }
 187
 188 static int export_features_open(struct inode *inode, struct file *file)
 189 {
 190         return single_open(file, export_features_show, NULL);
 191 }
 192
 193 static const struct file_operations export_features_operations = {
 194         .open           = export_features_open,
 195         .read           = seq_read,
 196         .llseek         = seq_lseek,
 197         .release        = single_release,
 198 };
 199
 200 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
 201 static int supported_enctypes_show(struct seq_file *m, void *v)
 202 {
 203         seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
 204         return 0;
 205 }
 206
 207 static int supported_enctypes_open(struct inode *inode, struct file *file)
 208 {
 209         return single_open(file, supported_enctypes_show, NULL);
 210 }
 211
 212 static const struct file_operations supported_enctypes_ops = {
 213         .open           = supported_enctypes_open,
 214         .read           = seq_read,
 215         .llseek         = seq_lseek,
 216         .release        = single_release,
 217 };
 218 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
 219
 220 static const struct file_operations pool_stats_operations = {
 221         .open           = nfsd_pool_stats_open,
 222         .read           = seq_read,
 223         .llseek         = seq_lseek,
 224         .release        = nfsd_pool_stats_release,
 225 };
 226
 227 static const struct file_operations reply_cache_stats_operations = {
 228         .open           = nfsd_reply_cache_stats_open,
 229         .read           = seq_read,
 230         .llseek         = seq_lseek,
 231         .release        = single_release,
 232 };
 233
 234 static const struct file_operations filecache_ops = {
 235         .open           = nfsd_file_cache_stats_open,
 236         .read           = seq_read,
 237         .llseek         = seq_lseek,
 238         .release        = single_release,
 239 };
 240
 241 /*----------------------------------------------------------------------------*/
 242 /*
 243  * payload - write methods
 244  */
 245
 246 static inline struct net *netns(struct file *file)
 247 {
 248         return file_inode(file)->i_sb->s_fs_info;
 249 }
 250
 251 /*
 252  * write_unlock_ip - Release all locks used by a client
 253  *
 254  * Experimental.
 255  *
 256  * Input:
 257  *                      buf:    '\n'-terminated C string containing a
 258  *                              presentation format IP address
 259  *                      size:   length of C string in @buf
 260  * Output:
 261  *      On success:     returns zero if all specified locks were released;
 262  *                      returns one if one or more locks were not released
 263  *      On error:       return code is negative errno value
 264  */
 265 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
 266 {
 267         struct sockaddr_storage address;
 268         struct sockaddr *sap = (struct sockaddr *)&address;
 269         size_t salen = sizeof(address);
 270         char *fo_path;
 271         struct net *net = netns(file);
 272
 273         /* sanity check */
 274         if (size == 0)
 275                 return -EINVAL;
 276
 277         if (buf[size-1] != '\n')
 278                 return -EINVAL;
 279
 280         fo_path = buf;
 281         if (qword_get(&buf, fo_path, size) < 0)
 282                 return -EINVAL;
 283
 284         if (rpc_pton(net, fo_path, size, sap, salen) == 0)
 285                 return -EINVAL;
 286
 287         return nlmsvc_unlock_all_by_ip(sap);
 288 }
 289
 290 /*
 291  * write_unlock_fs - Release all locks on a local file system
 292  *
 293  * Experimental.
 294  *
 295  * Input:
 296  *                      buf:    '\n'-terminated C string containing the
 297  *                              absolute pathname of a local file system
 298  *                      size:   length of C string in @buf
 299  * Output:
 300  *      On success:     returns zero if all specified locks were released;
 301  *                      returns one if one or more locks were not released
 302  *      On error:       return code is negative errno value
 303  */
 304 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
 305 {
 306         struct path path;
 307         char *fo_path;
 308         int error;
 309
 310         /* sanity check */
 311         if (size == 0)
 312                 return -EINVAL;
 313
 314         if (buf[size-1] != '\n')
 315                 return -EINVAL;
 316
 317         fo_path = buf;
 318         if (qword_get(&buf, fo_path, size) < 0)
 319                 return -EINVAL;
 320
 321         error = kern_path(fo_path, 0, &path);
 322         if (error)
 323                 return error;
 324
 325         /*
 326          * XXX: Needs better sanity checking.  Otherwise we could end up
 327          * releasing locks on the wrong file system.
 328          *
 329          * For example:
 330          * 1.  Does the path refer to a directory?
 331          * 2.  Is that directory a mount point, or
 332          * 3.  Is that directory the root of an exported file system?
 333          */
 334         error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
 335
 336         path_put(&path);
 337         return error;
 338 }
 339
 340 /*
 341  * write_filehandle - Get a variable-length NFS file handle by path
 342  *
 343  * On input, the buffer contains a '\n'-terminated C string comprised of
 344  * three alphanumeric words separated by whitespace.  The string may
 345  * contain escape sequences.
 346  *
 347  * Input:
 348  *                      buf:
 349  *                              domain:         client domain name
 350  *                              path:           export pathname
 351  *                              maxsize:        numeric maximum size of
 352  *                                              @buf
 353  *                      size:   length of C string in @buf
 354  * Output:
 355  *      On success:     passed-in buffer filled with '\n'-terminated C
 356  *                      string containing a ASCII hex text version
 357  *                      of the NFS file handle;
 358  *                      return code is the size in bytes of the string
 359  *      On error:       return code is negative errno value
 360  */
 361 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
 362 {
 363         char *dname, *path;
 364         int maxsize;
 365         char *mesg = buf;
 366         int len;
 367         struct auth_domain *dom;
 368         struct knfsd_fh fh;
 369
 370         if (size == 0)
 371                 return -EINVAL;
 372
 373         if (buf[size-1] != '\n')
 374                 return -EINVAL;
 375         buf[size-1] = 0;
 376
 377         dname = mesg;
 378         len = qword_get(&mesg, dname, size);
 379         if (len <= 0)
 380                 return -EINVAL;
 381
 382         path = dname+len+1;
 383         len = qword_get(&mesg, path, size);
 384         if (len <= 0)
 385                 return -EINVAL;
 386
 387         len = get_int(&mesg, &maxsize);
 388         if (len)
 389                 return len;
 390
 391         if (maxsize < NFS_FHSIZE)
 392                 return -EINVAL;
 393         maxsize = min(maxsize, NFS3_FHSIZE);
 394
 395         if (qword_get(&mesg, mesg, size)>0)
 396                 return -EINVAL;
 397
 398         /* we have all the words, they are in buf.. */
 399         dom = unix_domain_find(dname);
 400         if (!dom)
 401                 return -ENOMEM;
 402
 403         len = exp_rootfh(netns(file), dom, path, &fh,  maxsize);
 404         auth_domain_put(dom);
 405         if (len)
 406                 return len;
 407
 408         mesg = buf;
 409         len = SIMPLE_TRANSACTION_LIMIT;
 410         qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
 411         mesg[-1] = '\n';
 412         return mesg - buf;
 413 }
 414
 415 /*
 416  * write_threads - Start NFSD, or report the current number of running threads
 417  *
 418  * Input:
 419  *                      buf:            ignored
 420  *                      size:           zero
 421  * Output:
 422  *      On success:     passed-in buffer filled with '\n'-terminated C
 423  *                      string numeric value representing the number of
 424  *                      running NFSD threads;
 425  *                      return code is the size in bytes of the string
 426  *      On error:       return code is zero
 427  *
 428  * OR
 429  *
 430  * Input:
 431  *                      buf:            C string containing an unsigned
 432  *                                      integer value representing the
 433  *                                      number of NFSD threads to start
 434  *                      size:           non-zero length of C string in @buf
 435  * Output:
 436  *      On success:     NFS service is started;
 437  *                      passed-in buffer filled with '\n'-terminated C
 438  *                      string numeric value representing the number of
 439  *                      running NFSD threads;
 440  *                      return code is the size in bytes of the string
 441  *      On error:       return code is zero or a negative errno value
 442  */
 443 static ssize_t write_threads(struct file *file, char *buf, size_t size)
 444 {
 445         char *mesg = buf;
 446         int rv;
 447         struct net *net = netns(file);
 448
 449         if (size > 0) {
 450                 int newthreads;
 451                 rv = get_int(&mesg, &newthreads);
 452                 if (rv)
 453                         return rv;
 454                 if (newthreads < 0)
 455                         return -EINVAL;
 456                 rv = nfsd_svc(newthreads, net, file->f_cred);
 457                 if (rv < 0)
 458                         return rv;
 459         } else
 460                 rv = nfsd_nrthreads(net);
 461
 462         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
 463 }
 464
 465 /*
 466  * write_pool_threads - Set or report the current number of threads per pool
 467  *
 468  * Input:
 469  *                      buf:            ignored
 470  *                      size:           zero
 471  *
 472  * OR
 473  *
 474  * Input:
 475  *                      buf:            C string containing whitespace-
 476  *                                      separated unsigned integer values
 477  *                                      representing the number of NFSD
 478  *                                      threads to start in each pool
 479  *                      size:           non-zero length of C string in @buf
 480  * Output:
 481  *      On success:     passed-in buffer filled with '\n'-terminated C
 482  *                      string containing integer values representing the
 483  *                      number of NFSD threads in each pool;
 484  *                      return code is the size in bytes of the string
 485  *      On error:       return code is zero or a negative errno value
 486  */
 487 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
 488 {
 489         /* if size > 0, look for an array of number of threads per node
 490          * and apply them  then write out number of threads per node as reply
 491          */
 492         char *mesg = buf;
 493         int i;
 494         int rv;
 495         int len;
 496         int npools;
 497         int *nthreads;
 498         struct net *net = netns(file);
 499
 500         mutex_lock(&nfsd_mutex);
 501         npools = nfsd_nrpools(net);
 502         if (npools == 0) {
 503                 /*
 504                  * NFS is shut down.  The admin can start it by
 505                  * writing to the threads file but NOT the pool_threads
 506                  * file, sorry.  Report zero threads.
 507                  */
 508                 mutex_unlock(&nfsd_mutex);
 509                 strcpy(buf, "0\n");
 510                 return strlen(buf);
 511         }
 512
 513         nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
 514         rv = -ENOMEM;
 515         if (nthreads == NULL)
 516                 goto out_free;
 517
 518         if (size > 0) {
 519                 for (i = 0; i < npools; i++) {
 520                         rv = get_int(&mesg, &nthreads[i]);
 521                         if (rv == -ENOENT)
 522                                 break;          /* fewer numbers than pools */
 523                         if (rv)
 524                                 goto out_free;  /* syntax error */
 525                         rv = -EINVAL;
 526                         if (nthreads[i] < 0)
 527                                 goto out_free;
 528                 }
 529                 rv = nfsd_set_nrthreads(i, nthreads, net);
 530                 if (rv)
 531                         goto out_free;
 532         }
 533
 534         rv = nfsd_get_nrthreads(npools, nthreads, net);
 535         if (rv)
 536                 goto out_free;
 537
 538         mesg = buf;
 539         size = SIMPLE_TRANSACTION_LIMIT;
 540         for (i = 0; i < npools && size > 0; i++) {
 541                 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
 542                 len = strlen(mesg);
 543                 size -= len;
 544                 mesg += len;
 545         }
 546         rv = mesg - buf;
 547 out_free:
 548         kfree(nthreads);
 549         mutex_unlock(&nfsd_mutex);
 550         return rv;
 551 }
 552
 553 static ssize_t
 554 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
 555                 const char *sep, unsigned vers, int minor)
 556 {
 557         const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
 558         bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
 559
 560         if (vers == 4 && minor >= 0 &&
 561             !nfsd_minorversion(nn, minor, NFSD_TEST))
 562                 supported = false;
 563         if (minor == 0 && supported)
 564                 /*
 565                  * special case for backward compatability.
 566                  * +4.0 is never reported, it is implied by
 567                  * +4, unless -4.0 is present.
 568                  */
 569                 return 0;
 570         return snprintf(buf, remaining, format, sep,
 571                         supported ? '+' : '-', vers, minor);
 572 }
 573
 574 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
 575 {
 576         char *mesg = buf;
 577         char *vers, *minorp, sign;
 578         int len, num, remaining;
 579         ssize_t tlen = 0;
 580         char *sep;
 581         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 582
 583         if (size>0) {
 584                 if (nn->nfsd_serv)
 585                         /* Cannot change versions without updating
 586                          * nn->nfsd_serv->sv_xdrsize, and reallocing
 587                          * rq_argp and rq_resp
 588                          */
 589                         return -EBUSY;
 590                 if (buf[size-1] != '\n')
 591                         return -EINVAL;
 592                 buf[size-1] = 0;
 593
 594                 vers = mesg;
 595                 len = qword_get(&mesg, vers, size);
 596                 if (len <= 0) return -EINVAL;
 597                 do {
 598                         enum vers_op cmd;
 599                         unsigned minor;
 600                         sign = *vers;
 601                         if (sign == '+' || sign == '-')
 602                                 num = simple_strtol((vers+1), &minorp, 0);
 603                         else
 604                                 num = simple_strtol(vers, &minorp, 0);
 605                         if (*minorp == '.') {
 606                                 if (num != 4)
 607                                         return -EINVAL;
 608                                 if (kstrtouint(minorp+1, 0, &minor) < 0)
 609                                         return -EINVAL;
 610                         }
 611
 612                         cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
 613                         switch(num) {
 614                         case 2:
 615                         case 3:
 616                                 nfsd_vers(nn, num, cmd);
 617                                 break;
 618                         case 4:
 619                                 if (*minorp == '.') {
 620                                         if (nfsd_minorversion(nn, minor, cmd) < 0)
 621                                                 return -EINVAL;
 622                                 } else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
 623                                         /*
 624                                          * Either we have +4 and no minors are enabled,
 625                                          * or we have -4 and at least one minor is enabled.
 626                                          * In either case, propagate 'cmd' to all minors.
 627                                          */
 628                                         minor = 0;
 629                                         while (nfsd_minorversion(nn, minor, cmd) >= 0)
 630                                                 minor++;
 631                                 }
 632                                 break;
 633                         default:
 634                                 return -EINVAL;
 635                         }
 636                         vers += len + 1;
 637                 } while ((len = qword_get(&mesg, vers, size)) > 0);
 638                 /* If all get turned off, turn them back on, as
 639                  * having no versions is BAD
 640                  */
 641                 nfsd_reset_versions(nn);
 642         }
 643
 644         /* Now write current state into reply buffer */
 645         sep = "";
 646         remaining = SIMPLE_TRANSACTION_LIMIT;
 647         for (num=2 ; num <= 4 ; num++) {
 648                 int minor;
 649                 if (!nfsd_vers(nn, num, NFSD_AVAIL))
 650                         continue;
 651
 652                 minor = -1;
 653                 do {
 654                         len = nfsd_print_version_support(nn, buf, remaining,
 655                                         sep, num, minor);
 656                         if (len >= remaining)
 657                                 goto out;
 658                         remaining -= len;
 659                         buf += len;
 660                         tlen += len;
 661                         minor++;
 662                         if (len)
 663                                 sep = " ";
 664                 } while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
 665         }
 666 out:
 667         len = snprintf(buf, remaining, "\n");
 668         if (len >= remaining)
 669                 return -EINVAL;
 670         return tlen + len;
 671 }
 672
 673 /*
 674  * write_versions - Set or report the available NFS protocol versions
 675  *
 676  * Input:
 677  *                      buf:            ignored
 678  *                      size:           zero
 679  * Output:
 680  *      On success:     passed-in buffer filled with '\n'-terminated C
 681  *                      string containing positive or negative integer
 682  *                      values representing the current status of each
 683  *                      protocol version;
 684  *                      return code is the size in bytes of the string
 685  *      On error:       return code is zero or a negative errno value
 686  *
 687  * OR
 688  *
 689  * Input:
 690  *                      buf:            C string containing whitespace-
 691  *                                      separated positive or negative
 692  *                                      integer values representing NFS
 693  *                                      protocol versions to enable ("+n")
 694  *                                      or disable ("-n")
 695  *                      size:           non-zero length of C string in @buf
 696  * Output:
 697  *      On success:     status of zero or more protocol versions has
 698  *                      been updated; passed-in buffer filled with
 699  *                      '\n'-terminated C string containing positive
 700  *                      or negative integer values representing the
 701  *                      current status of each protocol version;
 702  *                      return code is the size in bytes of the string
 703  *      On error:       return code is zero or a negative errno value
 704  */
 705 static ssize_t write_versions(struct file *file, char *buf, size_t size)
 706 {
 707         ssize_t rv;
 708
 709         mutex_lock(&nfsd_mutex);
 710         rv = __write_versions(file, buf, size);
 711         mutex_unlock(&nfsd_mutex);
 712         return rv;
 713 }
 714
 715 /*
 716  * Zero-length write.  Return a list of NFSD's current listener
 717  * transports.
 718  */
 719 static ssize_t __write_ports_names(char *buf, struct net *net)
 720 {
 721         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 722
 723         if (nn->nfsd_serv == NULL)
 724                 return 0;
 725         return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
 726 }
 727
 728 /*
 729  * A single 'fd' number was written, in which case it must be for
 730  * a socket of a supported family/protocol, and we use it as an
 731  * nfsd listener.
 732  */
 733 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
 734 {
 735         char *mesg = buf;
 736         int fd, err;
 737         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 738
 739         err = get_int(&mesg, &fd);
 740         if (err != 0 || fd < 0)
 741                 return -EINVAL;
 742
 743         if (svc_alien_sock(net, fd)) {
 744                 printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
 745                 return -EINVAL;
 746         }
 747
 748         err = nfsd_create_serv(net);
 749         if (err != 0)
 750                 return err;
 751
 752         err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
 753
 754         if (err >= 0 &&
 755             !nn->nfsd_serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
 756                 svc_get(nn->nfsd_serv);
 757
 758         nfsd_put(net);
 759         return err;
 760 }
 761
 762 /*
 763  * A transport listener is added by writing it's transport name and
 764  * a port number.
 765  */
 766 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
 767 {
 768         char transport[16];
 769         struct svc_xprt *xprt;
 770         int port, err;
 771         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 772
 773         if (sscanf(buf, "%15s %5u", transport, &port) != 2)
 774                 return -EINVAL;
 775
 776         if (port < 1 || port > USHRT_MAX)
 777                 return -EINVAL;
 778
 779         err = nfsd_create_serv(net);
 780         if (err != 0)
 781                 return err;
 782
 783         err = svc_xprt_create(nn->nfsd_serv, transport, net,
 784                               PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
 785         if (err < 0)
 786                 goto out_err;
 787
 788         err = svc_xprt_create(nn->nfsd_serv, transport, net,
 789                               PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
 790         if (err < 0 && err != -EAFNOSUPPORT)
 791                 goto out_close;
 792
 793         if (!nn->nfsd_serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
 794                 svc_get(nn->nfsd_serv);
 795
 796         nfsd_put(net);
 797         return 0;
 798 out_close:
 799         xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
 800         if (xprt != NULL) {
 801                 svc_xprt_close(xprt);
 802                 svc_xprt_put(xprt);
 803         }
 804 out_err:
 805         nfsd_put(net);
 806         return err;
 807 }
 808
 809 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
 810                              struct net *net)
 811 {
 812         if (size == 0)
 813                 return __write_ports_names(buf, net);
 814
 815         if (isdigit(buf[0]))
 816                 return __write_ports_addfd(buf, net, file->f_cred);
 817
 818         if (isalpha(buf[0]))
 819                 return __write_ports_addxprt(buf, net, file->f_cred);
 820
 821         return -EINVAL;
 822 }
 823
 824 /*
 825  * write_ports - Pass a socket file descriptor or transport name to listen on
 826  *
 827  * Input:
 828  *                      buf:            ignored
 829  *                      size:           zero
 830  * Output:
 831  *      On success:     passed-in buffer filled with a '\n'-terminated C
 832  *                      string containing a whitespace-separated list of
 833  *                      named NFSD listeners;
 834  *                      return code is the size in bytes of the string
 835  *      On error:       return code is zero or a negative errno value
 836  *
 837  * OR
 838  *
 839  * Input:
 840  *                      buf:            C string containing an unsigned
 841  *                                      integer value representing a bound
 842  *                                      but unconnected socket that is to be
 843  *                                      used as an NFSD listener; listen(3)
 844  *                                      must be called for a SOCK_STREAM
 845  *                                      socket, otherwise it is ignored
 846  *                      size:           non-zero length of C string in @buf
 847  * Output:
 848  *      On success:     NFS service is started;
 849  *                      passed-in buffer filled with a '\n'-terminated C
 850  *                      string containing a unique alphanumeric name of
 851  *                      the listener;
 852  *                      return code is the size in bytes of the string
 853  *      On error:       return code is a negative errno value
 854  *
 855  * OR
 856  *
 857  * Input:
 858  *                      buf:            C string containing a transport
 859  *                                      name and an unsigned integer value
 860  *                                      representing the port to listen on,
 861  *                                      separated by whitespace
 862  *                      size:           non-zero length of C string in @buf
 863  * Output:
 864  *      On success:     returns zero; NFS service is started
 865  *      On error:       return code is a negative errno value
 866  */
 867 static ssize_t write_ports(struct file *file, char *buf, size_t size)
 868 {
 869         ssize_t rv;
 870
 871         mutex_lock(&nfsd_mutex);
 872         rv = __write_ports(file, buf, size, netns(file));
 873         mutex_unlock(&nfsd_mutex);
 874         return rv;
 875 }
 876
 877
 878 int nfsd_max_blksize;
 879
 880 /*
 881  * write_maxblksize - Set or report the current NFS blksize
 882  *
 883  * Input:
 884  *                      buf:            ignored
 885  *                      size:           zero
 886  *
 887  * OR
 888  *
 889  * Input:
 890  *                      buf:            C string containing an unsigned
 891  *                                      integer value representing the new
 892  *                                      NFS blksize
 893  *                      size:           non-zero length of C string in @buf
 894  * Output:
 895  *      On success:     passed-in buffer filled with '\n'-terminated C string
 896  *                      containing numeric value of the current NFS blksize
 897  *                      setting;
 898  *                      return code is the size in bytes of the string
 899  *      On error:       return code is zero or a negative errno value
 900  */
 901 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
 902 {
 903         char *mesg = buf;
 904         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 905
 906         if (size > 0) {
 907                 int bsize;
 908                 int rv = get_int(&mesg, &bsize);
 909                 if (rv)
 910                         return rv;
 911                 /* force bsize into allowed range and
 912                  * required alignment.
 913                  */
 914                 bsize = max_t(int, bsize, 1024);
 915                 bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
 916                 bsize &= ~(1024-1);
 917                 mutex_lock(&nfsd_mutex);
 918                 if (nn->nfsd_serv) {
 919                         mutex_unlock(&nfsd_mutex);
 920                         return -EBUSY;
 921                 }
 922                 nfsd_max_blksize = bsize;
 923                 mutex_unlock(&nfsd_mutex);
 924         }
 925
 926         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
 927                                                         nfsd_max_blksize);
 928 }
 929
 930 /*
 931  * write_maxconn - Set or report the current max number of connections
 932  *
 933  * Input:
 934  *                      buf:            ignored
 935  *                      size:           zero
 936  * OR
 937  *
 938  * Input:
 939  *                      buf:            C string containing an unsigned
 940  *                                      integer value representing the new
 941  *                                      number of max connections
 942  *                      size:           non-zero length of C string in @buf
 943  * Output:
 944  *      On success:     passed-in buffer filled with '\n'-terminated C string
 945  *                      containing numeric value of max_connections setting
 946  *                      for this net namespace;
 947  *                      return code is the size in bytes of the string
 948  *      On error:       return code is zero or a negative errno value
 949  */
 950 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
 951 {
 952         char *mesg = buf;
 953         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 954         unsigned int maxconn = nn->max_connections;
 955
 956         if (size > 0) {
 957                 int rv = get_uint(&mesg, &maxconn);
 958
 959                 if (rv)
 960                         return rv;
 961                 nn->max_connections = maxconn;
 962         }
 963
 964         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
 965 }
 966
 967 #ifdef CONFIG_NFSD_V4
 968 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
 969                                   time64_t *time, struct nfsd_net *nn)
 970 {
 971         char *mesg = buf;
 972         int rv, i;
 973
 974         if (size > 0) {
 975                 if (nn->nfsd_serv)
 976                         return -EBUSY;
 977                 rv = get_int(&mesg, &i);
 978                 if (rv)
 979                         return rv;
 980                 /*
 981                  * Some sanity checking.  We don't have a reason for
 982                  * these particular numbers, but problems with the
 983                  * extremes are:
 984                  *      - Too short: the briefest network outage may
 985                  *        cause clients to lose all their locks.  Also,
 986                  *        the frequent polling may be wasteful.
 987                  *      - Too long: do you really want reboot recovery
 988                  *        to take more than an hour?  Or to make other
 989                  *        clients wait an hour before being able to
 990                  *        revoke a dead client's locks?
 991                  */
 992                 if (i < 10 || i > 3600)
 993                         return -EINVAL;
 994                 *time = i;
 995         }
 996
 997         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
 998 }
 999
1000 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
1001                                 time64_t *time, struct nfsd_net *nn)
1002 {
1003         ssize_t rv;
1004
1005         mutex_lock(&nfsd_mutex);
1006         rv = __nfsd4_write_time(file, buf, size, time, nn);
1007         mutex_unlock(&nfsd_mutex);
1008         return rv;
1009 }
1010
1011 /*
1012  * write_leasetime - Set or report the current NFSv4 lease time
1013  *
1014  * Input:
1015  *                      buf:            ignored
1016  *                      size:           zero
1017  *
1018  * OR
1019  *
1020  * Input:
1021  *                      buf:            C string containing an unsigned
1022  *                                      integer value representing the new
1023  *                                      NFSv4 lease expiry time
1024  *                      size:           non-zero length of C string in @buf
1025  * Output:
1026  *      On success:     passed-in buffer filled with '\n'-terminated C
1027  *                      string containing unsigned integer value of the
1028  *                      current lease expiry time;
1029  *                      return code is the size in bytes of the string
1030  *      On error:       return code is zero or a negative errno value
1031  */
1032 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1033 {
1034         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1035         return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1036 }
1037
1038 /*
1039  * write_gracetime - Set or report current NFSv4 grace period time
1040  *
1041  * As above, but sets the time of the NFSv4 grace period.
1042  *
1043  * Note this should never be set to less than the *previous*
1044  * lease-period time, but we don't try to enforce this.  (In the common
1045  * case (a new boot), we don't know what the previous lease time was
1046  * anyway.)
1047  */
1048 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1049 {
1050         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1051         return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1052 }
1053
1054 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1055                                    struct nfsd_net *nn)
1056 {
1057         char *mesg = buf;
1058         char *recdir;
1059         int len, status;
1060
1061         if (size > 0) {
1062                 if (nn->nfsd_serv)
1063                         return -EBUSY;
1064                 if (size > PATH_MAX || buf[size-1] != '\n')
1065                         return -EINVAL;
1066                 buf[size-1] = 0;
1067
1068                 recdir = mesg;
1069                 len = qword_get(&mesg, recdir, size);
1070                 if (len <= 0)
1071                         return -EINVAL;
1072
1073                 status = nfs4_reset_recoverydir(recdir);
1074                 if (status)
1075                         return status;
1076         }
1077
1078         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1079                                                         nfs4_recoverydir());
1080 }
1081
1082 /*
1083  * write_recoverydir - Set or report the pathname of the recovery directory
1084  *
1085  * Input:
1086  *                      buf:            ignored
1087  *                      size:           zero
1088  *
1089  * OR
1090  *
1091  * Input:
1092  *                      buf:            C string containing the pathname
1093  *                                      of the directory on a local file
1094  *                                      system containing permanent NFSv4
1095  *                                      recovery data
1096  *                      size:           non-zero length of C string in @buf
1097  * Output:
1098  *      On success:     passed-in buffer filled with '\n'-terminated C string
1099  *                      containing the current recovery pathname setting;
1100  *                      return code is the size in bytes of the string
1101  *      On error:       return code is zero or a negative errno value
1102  */
1103 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1104 {
1105         ssize_t rv;
1106         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1107
1108         mutex_lock(&nfsd_mutex);
1109         rv = __write_recoverydir(file, buf, size, nn);
1110         mutex_unlock(&nfsd_mutex);
1111         return rv;
1112 }
1113
1114 /*
1115  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1116  *
1117  * Input:
1118  *                      buf:            ignored
1119  *                      size:           zero
1120  * OR
1121  *
1122  * Input:
1123  *                      buf:            any value
1124  *                      size:           non-zero length of C string in @buf
1125  * Output:
1126  *                      passed-in buffer filled with "Y" or "N" with a newline
1127  *                      and NULL-terminated C string. This indicates whether
1128  *                      the grace period has ended in the current net
1129  *                      namespace. Return code is the size in bytes of the
1130  *                      string. Writing a string that starts with 'Y', 'y', or
1131  *                      '1' to the file will end the grace period for nfsd's v4
1132  *                      lock manager.
1133  */
1134 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1135 {
1136         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1137
1138         if (size > 0) {
1139                 switch(buf[0]) {
1140                 case 'Y':
1141                 case 'y':
1142                 case '1':
1143                         if (!nn->nfsd_serv)
1144                                 return -EBUSY;
1145                         nfsd4_end_grace(nn);
1146                         break;
1147                 default:
1148                         return -EINVAL;
1149                 }
1150         }
1151
1152         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1153                          nn->grace_ended ? 'Y' : 'N');
1154 }
1155
1156 #endif
1157
1158 /*----------------------------------------------------------------------------*/
1159 /*
1160  *      populating the filesystem.
1161  */
1162
1163 /* Basically copying rpc_get_inode. */
1164 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1165 {
1166         struct inode *inode = new_inode(sb);
1167         if (!inode)
1168                 return NULL;
1169         /* Following advice from simple_fill_super documentation: */
1170         inode->i_ino = iunique(sb, NFSD_MaxReserved);
1171         inode->i_mode = mode;
1172         inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1173         switch (mode & S_IFMT) {
1174         case S_IFDIR:
1175                 inode->i_fop = &simple_dir_operations;
1176                 inode->i_op = &simple_dir_inode_operations;
1177                 inc_nlink(inode);
1178                 break;
1179         default:
1180                 break;
1181         }
1182         return inode;
1183 }
1184
1185 static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1186 {
1187         struct inode *inode;
1188
1189         inode = nfsd_get_inode(dir->i_sb, mode);
1190         if (!inode)
1191                 return -ENOMEM;
1192         if (ncl) {
1193                 inode->i_private = ncl;
1194                 kref_get(&ncl->cl_ref);
1195         }
1196         d_add(dentry, inode);
1197         inc_nlink(dir);
1198         fsnotify_mkdir(dir, dentry);
1199         return 0;
1200 }
1201
1202 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1203 {
1204         struct inode *dir = parent->d_inode;
1205         struct dentry *dentry;
1206         int ret = -ENOMEM;
1207
1208         inode_lock(dir);
1209         dentry = d_alloc_name(parent, name);
1210         if (!dentry)
1211                 goto out_err;
1212         ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1213         if (ret)
1214                 goto out_err;
1215 out:
1216         inode_unlock(dir);
1217         return dentry;
1218 out_err:
1219         dput(dentry);
1220         dentry = ERR_PTR(ret);
1221         goto out;
1222 }
1223
1224 static void clear_ncl(struct inode *inode)
1225 {
1226         struct nfsdfs_client *ncl = inode->i_private;
1227
1228         inode->i_private = NULL;
1229         kref_put(&ncl->cl_ref, ncl->cl_release);
1230 }
1231
1232 static struct nfsdfs_client *__get_nfsdfs_client(struct inode *inode)
1233 {
1234         struct nfsdfs_client *nc = inode->i_private;
1235
1236         if (nc)
1237                 kref_get(&nc->cl_ref);
1238         return nc;
1239 }
1240
1241 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1242 {
1243         struct nfsdfs_client *nc;
1244
1245         inode_lock_shared(inode);
1246         nc = __get_nfsdfs_client(inode);
1247         inode_unlock_shared(inode);
1248         return nc;
1249 }
1250 /* from __rpc_unlink */
1251 static void nfsdfs_remove_file(struct inode *dir, struct dentry *dentry)
1252 {
1253         int ret;
1254
1255         clear_ncl(d_inode(dentry));
1256         dget(dentry);
1257         ret = simple_unlink(dir, dentry);
1258         d_drop(dentry);
1259         fsnotify_unlink(dir, dentry);
1260         dput(dentry);
1261         WARN_ON_ONCE(ret);
1262 }
1263
1264 static void nfsdfs_remove_files(struct dentry *root)
1265 {
1266         struct dentry *dentry, *tmp;
1267
1268         list_for_each_entry_safe(dentry, tmp, &root->d_subdirs, d_child) {
1269                 if (!simple_positive(dentry)) {
1270                         WARN_ON_ONCE(1); /* I think this can't happen? */
1271                         continue;
1272                 }
1273                 nfsdfs_remove_file(d_inode(root), dentry);
1274         }
1275 }
1276
1277 /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1278  * code instead. */
1279 static  int nfsdfs_create_files(struct dentry *root,
1280                                 const struct tree_descr *files,
1281                                 struct dentry **fdentries)
1282 {
1283         struct inode *dir = d_inode(root);
1284         struct inode *inode;
1285         struct dentry *dentry;
1286         int i;
1287
1288         inode_lock(dir);
1289         for (i = 0; files->name && files->name[0]; i++, files++) {
1290                 dentry = d_alloc_name(root, files->name);
1291                 if (!dentry)
1292                         goto out;
1293                 inode = nfsd_get_inode(d_inode(root)->i_sb,
1294                                         S_IFREG | files->mode);
1295                 if (!inode) {
1296                         dput(dentry);
1297                         goto out;
1298                 }
1299                 inode->i_fop = files->ops;
1300                 inode->i_private = __get_nfsdfs_client(dir);
1301                 d_add(dentry, inode);
1302                 fsnotify_create(dir, dentry);
1303                 if (fdentries)
1304                         fdentries[i] = dentry;
1305         }
1306         inode_unlock(dir);
1307         return 0;
1308 out:
1309         nfsdfs_remove_files(root);
1310         inode_unlock(dir);
1311         return -ENOMEM;
1312 }
1313
1314 /* on success, returns positive number unique to that client. */
1315 struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1316                                  struct nfsdfs_client *ncl, u32 id,
1317                                  const struct tree_descr *files,
1318                                  struct dentry **fdentries)
1319 {
1320         struct dentry *dentry;
1321         char name[11];
1322         int ret;
1323
1324         sprintf(name, "%u", id);
1325
1326         dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1327         if (IS_ERR(dentry)) /* XXX: tossing errors? */
1328                 return NULL;
1329         ret = nfsdfs_create_files(dentry, files, fdentries);
1330         if (ret) {
1331                 nfsd_client_rmdir(dentry);
1332                 return NULL;
1333         }
1334         return dentry;
1335 }
1336
1337 /* Taken from __rpc_rmdir: */
1338 void nfsd_client_rmdir(struct dentry *dentry)
1339 {
1340         struct inode *dir = d_inode(dentry->d_parent);
1341         struct inode *inode = d_inode(dentry);
1342         int ret;
1343
1344         inode_lock(dir);
1345         nfsdfs_remove_files(dentry);
1346         clear_ncl(inode);
1347         dget(dentry);
1348         ret = simple_rmdir(dir, dentry);
1349         WARN_ON_ONCE(ret);
1350         d_drop(dentry);
1351         fsnotify_rmdir(dir, dentry);
1352         dput(dentry);
1353         inode_unlock(dir);
1354 }
1355
1356 static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1357 {
1358         struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1359                                                         nfsd_net_id);
1360         struct dentry *dentry;
1361         int ret;
1362
1363         static const struct tree_descr nfsd_files[] = {
1364                 [NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1365                 /* Per-export io stats use same ops as exports file */
1366                 [NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1367                 [NFSD_Export_features] = {"export_features",
1368                                         &export_features_operations, S_IRUGO},
1369                 [NFSD_FO_UnlockIP] = {"unlock_ip",
1370                                         &transaction_ops, S_IWUSR|S_IRUSR},
1371                 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1372                                         &transaction_ops, S_IWUSR|S_IRUSR},
1373                 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1374                 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1375                 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1376                 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1377                 [NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
1378                 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1379                 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1380                 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1381                 [NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1382                 [NFSD_Filecache] = {"filecache", &filecache_ops, S_IRUGO},
1383 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1384                 [NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1385 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1386 #ifdef CONFIG_NFSD_V4
1387                 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1388                 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1389                 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1390                 [NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1391 #endif
1392                 /* last one */ {""}
1393         };
1394
1395         ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1396         if (ret)
1397                 return ret;
1398         dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1399         if (IS_ERR(dentry))
1400                 return PTR_ERR(dentry);
1401         nn->nfsd_client_dir = dentry;
1402         return 0;
1403 }
1404
1405 static int nfsd_fs_get_tree(struct fs_context *fc)
1406 {
1407         return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1408 }
1409
1410 static void nfsd_fs_free_fc(struct fs_context *fc)
1411 {
1412         if (fc->s_fs_info)
1413                 put_net(fc->s_fs_info);
1414 }
1415
1416 static const struct fs_context_operations nfsd_fs_context_ops = {
1417         .free           = nfsd_fs_free_fc,
1418         .get_tree       = nfsd_fs_get_tree,
1419 };
1420
1421 static int nfsd_init_fs_context(struct fs_context *fc)
1422 {
1423         put_user_ns(fc->user_ns);
1424         fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1425         fc->ops = &nfsd_fs_context_ops;
1426         return 0;
1427 }
1428
1429 static void nfsd_umount(struct super_block *sb)
1430 {
1431         struct net *net = sb->s_fs_info;
1432
1433         nfsd_shutdown_threads(net);
1434
1435         kill_litter_super(sb);
1436         put_net(net);
1437 }
1438
1439 static struct file_system_type nfsd_fs_type = {
1440         .owner          = THIS_MODULE,
1441         .name           = "nfsd",
1442         .init_fs_context = nfsd_init_fs_context,
1443         .kill_sb        = nfsd_umount,
1444 };
1445 MODULE_ALIAS_FS("nfsd");
1446
1447 #ifdef CONFIG_PROC_FS
1448 static int create_proc_exports_entry(void)
1449 {
1450         struct proc_dir_entry *entry;
1451
1452         entry = proc_mkdir("fs/nfs", NULL);
1453         if (!entry)
1454                 return -ENOMEM;
1455         entry = proc_create("exports", 0, entry, &exports_proc_ops);
1456         if (!entry) {
1457                 remove_proc_entry("fs/nfs", NULL);
1458                 return -ENOMEM;
1459         }
1460         return 0;
1461 }
1462 #else /* CONFIG_PROC_FS */
1463 static int create_proc_exports_entry(void)
1464 {
1465         return 0;
1466 }
1467 #endif
1468
1469 unsigned int nfsd_net_id;
1470
1471 static __net_init int nfsd_init_net(struct net *net)
1472 {
1473         int retval;
1474         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1475
1476         retval = nfsd_export_init(net);
1477         if (retval)
1478                 goto out_export_error;
1479         retval = nfsd_idmap_init(net);
1480         if (retval)
1481                 goto out_idmap_error;
1482         nn->nfsd_versions = NULL;
1483         nn->nfsd4_minorversions = NULL;
1484         retval = nfsd_reply_cache_init(nn);
1485         if (retval)
1486                 goto out_drc_error;
1487         nfsd4_init_leases_net(nn);
1488
1489         get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
1490         seqlock_init(&nn->writeverf_lock);
1491
1492         return 0;
1493
1494 out_drc_error:
1495         nfsd_idmap_shutdown(net);
1496 out_idmap_error:
1497         nfsd_export_shutdown(net);
1498 out_export_error:
1499         return retval;
1500 }
1501
1502 static __net_exit void nfsd_exit_net(struct net *net)
1503 {
1504         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1505
1506         nfsd_reply_cache_shutdown(nn);
1507         nfsd_idmap_shutdown(net);
1508         nfsd_export_shutdown(net);
1509         nfsd_netns_free_versions(net_generic(net, nfsd_net_id));
1510 }
1511
1512 static struct pernet_operations nfsd_net_ops = {
1513         .init = nfsd_init_net,
1514         .exit = nfsd_exit_net,
1515         .id   = &nfsd_net_id,
1516         .size = sizeof(struct nfsd_net),
1517 };
1518
1519 static int __init init_nfsd(void)
1520 {
1521         int retval;
1522
1523         retval = nfsd4_init_slabs();
1524         if (retval)
1525                 return retval;
1526         retval = nfsd4_init_pnfs();
1527         if (retval)
1528                 goto out_free_slabs;
1529         retval = nfsd_stat_init();      /* Statistics */
1530         if (retval)
1531                 goto out_free_pnfs;
1532         retval = nfsd_drc_slab_create();
1533         if (retval)
1534                 goto out_free_stat;
1535         nfsd_lockd_init();      /* lockd->nfsd callbacks */
1536         retval = create_proc_exports_entry();
1537         if (retval)
1538                 goto out_free_lockd;
1539         retval = register_pernet_subsys(&nfsd_net_ops);
1540         if (retval < 0)
1541                 goto out_free_exports;
1542         retval = register_cld_notifier();
1543         if (retval)
1544                 goto out_free_subsys;
1545         retval = nfsd4_create_laundry_wq();
1546         if (retval)
1547                 goto out_free_cld;
1548         retval = register_filesystem(&nfsd_fs_type);
1549         if (retval)
1550                 goto out_free_all;
1551         return 0;
1552 out_free_all:
1553         nfsd4_destroy_laundry_wq();
1554 out_free_cld:
1555         unregister_cld_notifier();
1556 out_free_subsys:
1557         unregister_pernet_subsys(&nfsd_net_ops);
1558 out_free_exports:
1559         remove_proc_entry("fs/nfs/exports", NULL);
1560         remove_proc_entry("fs/nfs", NULL);
1561 out_free_lockd:
1562         nfsd_lockd_shutdown();
1563         nfsd_drc_slab_free();
1564 out_free_stat:
1565         nfsd_stat_shutdown();
1566 out_free_pnfs:
1567         nfsd4_exit_pnfs();
1568 out_free_slabs:
1569         nfsd4_free_slabs();
1570         return retval;
1571 }
1572
1573 static void __exit exit_nfsd(void)
1574 {
1575         unregister_filesystem(&nfsd_fs_type);
1576         nfsd4_destroy_laundry_wq();
1577         unregister_cld_notifier();
1578         unregister_pernet_subsys(&nfsd_net_ops);
1579         nfsd_drc_slab_free();
1580         remove_proc_entry("fs/nfs/exports", NULL);
1581         remove_proc_entry("fs/nfs", NULL);
1582         nfsd_stat_shutdown();
1583         nfsd_lockd_shutdown();
1584         nfsd4_free_slabs();
1585         nfsd4_exit_pnfs();
1586 }
1587
1588 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1589 MODULE_LICENSE("GPL");
1590 module_init(init_nfsd)
1591 module_exit(exit_nfsd)