fs/nfsd/nfsctl.c

   1 /*
   2  * Syscall interface to knfsd.
   3  *
   4  * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
   5  */
   6
   7 #include <linux/slab.h>
   8 #include <linux/namei.h>
   9 #include <linux/ctype.h>
  10
  11 #include <linux/sunrpc/svcsock.h>
  12 #include <linux/lockd/lockd.h>
  13 #include <linux/sunrpc/addr.h>
  14 #include <linux/sunrpc/gss_api.h>
  15 #include <linux/sunrpc/gss_krb5_enctypes.h>
  16 #include <linux/sunrpc/rpc_pipe_fs.h>
  17 #include <linux/module.h>
  18
  19 #include "idmap.h"
  20 #include "nfsd.h"
  21 #include "cache.h"
  22 #include "state.h"
  23 #include "netns.h"
  24 #include "pnfs.h"
  25
  26 /*
  27  *      We have a single directory with several nodes in it.
  28  */
  29 enum {
  30         NFSD_Root = 1,
  31         NFSD_List,
  32         NFSD_Export_features,
  33         NFSD_Fh,
  34         NFSD_FO_UnlockIP,
  35         NFSD_FO_UnlockFS,
  36         NFSD_Threads,
  37         NFSD_Pool_Threads,
  38         NFSD_Pool_Stats,
  39         NFSD_Reply_Cache_Stats,
  40         NFSD_Versions,
  41         NFSD_Ports,
  42         NFSD_MaxBlkSize,
  43         NFSD_MaxConnections,
  44         NFSD_SupportedEnctypes,
  45         /*
  46          * The below MUST come last.  Otherwise we leave a hole in nfsd_files[]
  47          * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
  48          */
  49 #ifdef CONFIG_NFSD_V4
  50         NFSD_Leasetime,
  51         NFSD_Gracetime,
  52         NFSD_RecoveryDir,
  53         NFSD_V4EndGrace,
  54 #endif
  55 };
  56
  57 /*
  58  * write() for these nodes.
  59  */
  60 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
  61 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
  62 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
  63 static ssize_t write_threads(struct file *file, char *buf, size_t size);
  64 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
  65 static ssize_t write_versions(struct file *file, char *buf, size_t size);
  66 static ssize_t write_ports(struct file *file, char *buf, size_t size);
  67 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
  68 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
  69 #ifdef CONFIG_NFSD_V4
  70 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
  71 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
  72 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
  73 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
  74 #endif
  75
  76 static ssize_t (*write_op[])(struct file *, char *, size_t) = {
  77         [NFSD_Fh] = write_filehandle,
  78         [NFSD_FO_UnlockIP] = write_unlock_ip,
  79         [NFSD_FO_UnlockFS] = write_unlock_fs,
  80         [NFSD_Threads] = write_threads,
  81         [NFSD_Pool_Threads] = write_pool_threads,
  82         [NFSD_Versions] = write_versions,
  83         [NFSD_Ports] = write_ports,
  84         [NFSD_MaxBlkSize] = write_maxblksize,
  85         [NFSD_MaxConnections] = write_maxconn,
  86 #ifdef CONFIG_NFSD_V4
  87         [NFSD_Leasetime] = write_leasetime,
  88         [NFSD_Gracetime] = write_gracetime,
  89         [NFSD_RecoveryDir] = write_recoverydir,
  90         [NFSD_V4EndGrace] = write_v4_end_grace,
  91 #endif
  92 };
  93
  94 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
  95 {
  96         ino_t ino =  file_inode(file)->i_ino;
  97         char *data;
  98         ssize_t rv;
  99
 100         if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
 101                 return -EINVAL;
 102
 103         data = simple_transaction_get(file, buf, size);
 104         if (IS_ERR(data))
 105                 return PTR_ERR(data);
 106
 107         rv =  write_op[ino](file, data, size);
 108         if (rv >= 0) {
 109                 simple_transaction_set(file, rv);
 110                 rv = size;
 111         }
 112         return rv;
 113 }
 114
 115 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
 116 {
 117         if (! file->private_data) {
 118                 /* An attempt to read a transaction file without writing
 119                  * causes a 0-byte write so that the file can return
 120                  * state information
 121                  */
 122                 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
 123                 if (rv < 0)
 124                         return rv;
 125         }
 126         return simple_transaction_read(file, buf, size, pos);
 127 }
 128
 129 static const struct file_operations transaction_ops = {
 130         .write          = nfsctl_transaction_write,
 131         .read           = nfsctl_transaction_read,
 132         .release        = simple_transaction_release,
 133         .llseek         = default_llseek,
 134 };
 135
 136 static int exports_net_open(struct net *net, struct file *file)
 137 {
 138         int err;
 139         struct seq_file *seq;
 140         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 141
 142         err = seq_open(file, &nfs_exports_op);
 143         if (err)
 144                 return err;
 145
 146         seq = file->private_data;
 147         seq->private = nn->svc_export_cache;
 148         return 0;
 149 }
 150
 151 static int exports_proc_open(struct inode *inode, struct file *file)
 152 {
 153         return exports_net_open(current->nsproxy->net_ns, file);
 154 }
 155
 156 static const struct file_operations exports_proc_operations = {
 157         .open           = exports_proc_open,
 158         .read           = seq_read,
 159         .llseek         = seq_lseek,
 160         .release        = seq_release,
 161         .owner          = THIS_MODULE,
 162 };
 163
 164 static int exports_nfsd_open(struct inode *inode, struct file *file)
 165 {
 166         return exports_net_open(inode->i_sb->s_fs_info, file);
 167 }
 168
 169 static const struct file_operations exports_nfsd_operations = {
 170         .open           = exports_nfsd_open,
 171         .read           = seq_read,
 172         .llseek         = seq_lseek,
 173         .release        = seq_release,
 174         .owner          = THIS_MODULE,
 175 };
 176
 177 static int export_features_show(struct seq_file *m, void *v)
 178 {
 179         seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
 180         return 0;
 181 }
 182
 183 static int export_features_open(struct inode *inode, struct file *file)
 184 {
 185         return single_open(file, export_features_show, NULL);
 186 }
 187
 188 static const struct file_operations export_features_operations = {
 189         .open           = export_features_open,
 190         .read           = seq_read,
 191         .llseek         = seq_lseek,
 192         .release        = single_release,
 193 };
 194
 195 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
 196 static int supported_enctypes_show(struct seq_file *m, void *v)
 197 {
 198         seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
 199         return 0;
 200 }
 201
 202 static int supported_enctypes_open(struct inode *inode, struct file *file)
 203 {
 204         return single_open(file, supported_enctypes_show, NULL);
 205 }
 206
 207 static const struct file_operations supported_enctypes_ops = {
 208         .open           = supported_enctypes_open,
 209         .read           = seq_read,
 210         .llseek         = seq_lseek,
 211         .release        = single_release,
 212 };
 213 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
 214
 215 static const struct file_operations pool_stats_operations = {
 216         .open           = nfsd_pool_stats_open,
 217         .read           = seq_read,
 218         .llseek         = seq_lseek,
 219         .release        = nfsd_pool_stats_release,
 220         .owner          = THIS_MODULE,
 221 };
 222
 223 static struct file_operations reply_cache_stats_operations = {
 224         .open           = nfsd_reply_cache_stats_open,
 225         .read           = seq_read,
 226         .llseek         = seq_lseek,
 227         .release        = single_release,
 228 };
 229
 230 /*----------------------------------------------------------------------------*/
 231 /*
 232  * payload - write methods
 233  */
 234
 235 static inline struct net *netns(struct file *file)
 236 {
 237         return file_inode(file)->i_sb->s_fs_info;
 238 }
 239
 240 /**
 241  * write_unlock_ip - Release all locks used by a client
 242  *
 243  * Experimental.
 244  *
 245  * Input:
 246  *                      buf:    '\n'-terminated C string containing a
 247  *                              presentation format IP address
 248  *                      size:   length of C string in @buf
 249  * Output:
 250  *      On success:     returns zero if all specified locks were released;
 251  *                      returns one if one or more locks were not released
 252  *      On error:       return code is negative errno value
 253  */
 254 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
 255 {
 256         struct sockaddr_storage address;
 257         struct sockaddr *sap = (struct sockaddr *)&address;
 258         size_t salen = sizeof(address);
 259         char *fo_path;
 260         struct net *net = netns(file);
 261
 262         /* sanity check */
 263         if (size == 0)
 264                 return -EINVAL;
 265
 266         if (buf[size-1] != '\n')
 267                 return -EINVAL;
 268
 269         fo_path = buf;
 270         if (qword_get(&buf, fo_path, size) < 0)
 271                 return -EINVAL;
 272
 273         if (rpc_pton(net, fo_path, size, sap, salen) == 0)
 274                 return -EINVAL;
 275
 276         return nlmsvc_unlock_all_by_ip(sap);
 277 }
 278
 279 /**
 280  * write_unlock_fs - Release all locks on a local file system
 281  *
 282  * Experimental.
 283  *
 284  * Input:
 285  *                      buf:    '\n'-terminated C string containing the
 286  *                              absolute pathname of a local file system
 287  *                      size:   length of C string in @buf
 288  * Output:
 289  *      On success:     returns zero if all specified locks were released;
 290  *                      returns one if one or more locks were not released
 291  *      On error:       return code is negative errno value
 292  */
 293 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
 294 {
 295         struct path path;
 296         char *fo_path;
 297         int error;
 298
 299         /* sanity check */
 300         if (size == 0)
 301                 return -EINVAL;
 302
 303         if (buf[size-1] != '\n')
 304                 return -EINVAL;
 305
 306         fo_path = buf;
 307         if (qword_get(&buf, fo_path, size) < 0)
 308                 return -EINVAL;
 309
 310         error = kern_path(fo_path, 0, &path);
 311         if (error)
 312                 return error;
 313
 314         /*
 315          * XXX: Needs better sanity checking.  Otherwise we could end up
 316          * releasing locks on the wrong file system.
 317          *
 318          * For example:
 319          * 1.  Does the path refer to a directory?
 320          * 2.  Is that directory a mount point, or
 321          * 3.  Is that directory the root of an exported file system?
 322          */
 323         error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
 324
 325         path_put(&path);
 326         return error;
 327 }
 328
 329 /**
 330  * write_filehandle - Get a variable-length NFS file handle by path
 331  *
 332  * On input, the buffer contains a '\n'-terminated C string comprised of
 333  * three alphanumeric words separated by whitespace.  The string may
 334  * contain escape sequences.
 335  *
 336  * Input:
 337  *                      buf:
 338  *                              domain:         client domain name
 339  *                              path:           export pathname
 340  *                              maxsize:        numeric maximum size of
 341  *                                              @buf
 342  *                      size:   length of C string in @buf
 343  * Output:
 344  *      On success:     passed-in buffer filled with '\n'-terminated C
 345  *                      string containing a ASCII hex text version
 346  *                      of the NFS file handle;
 347  *                      return code is the size in bytes of the string
 348  *      On error:       return code is negative errno value
 349  */
 350 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
 351 {
 352         char *dname, *path;
 353         int uninitialized_var(maxsize);
 354         char *mesg = buf;
 355         int len;
 356         struct auth_domain *dom;
 357         struct knfsd_fh fh;
 358
 359         if (size == 0)
 360                 return -EINVAL;
 361
 362         if (buf[size-1] != '\n')
 363                 return -EINVAL;
 364         buf[size-1] = 0;
 365
 366         dname = mesg;
 367         len = qword_get(&mesg, dname, size);
 368         if (len <= 0)
 369                 return -EINVAL;
 370
 371         path = dname+len+1;
 372         len = qword_get(&mesg, path, size);
 373         if (len <= 0)
 374                 return -EINVAL;
 375
 376         len = get_int(&mesg, &maxsize);
 377         if (len)
 378                 return len;
 379
 380         if (maxsize < NFS_FHSIZE)
 381                 return -EINVAL;
 382         maxsize = min(maxsize, NFS3_FHSIZE);
 383
 384         if (qword_get(&mesg, mesg, size)>0)
 385                 return -EINVAL;
 386
 387         /* we have all the words, they are in buf.. */
 388         dom = unix_domain_find(dname);
 389         if (!dom)
 390                 return -ENOMEM;
 391
 392         len = exp_rootfh(netns(file), dom, path, &fh,  maxsize);
 393         auth_domain_put(dom);
 394         if (len)
 395                 return len;
 396
 397         mesg = buf;
 398         len = SIMPLE_TRANSACTION_LIMIT;
 399         qword_addhex(&mesg, &len, (char*)&fh.fh_base, fh.fh_size);
 400         mesg[-1] = '\n';
 401         return mesg - buf;
 402 }
 403
 404 /**
 405  * write_threads - Start NFSD, or report the current number of running threads
 406  *
 407  * Input:
 408  *                      buf:            ignored
 409  *                      size:           zero
 410  * Output:
 411  *      On success:     passed-in buffer filled with '\n'-terminated C
 412  *                      string numeric value representing the number of
 413  *                      running NFSD threads;
 414  *                      return code is the size in bytes of the string
 415  *      On error:       return code is zero
 416  *
 417  * OR
 418  *
 419  * Input:
 420  *                      buf:            C string containing an unsigned
 421  *                                      integer value representing the
 422  *                                      number of NFSD threads to start
 423  *                      size:           non-zero length of C string in @buf
 424  * Output:
 425  *      On success:     NFS service is started;
 426  *                      passed-in buffer filled with '\n'-terminated C
 427  *                      string numeric value representing the number of
 428  *                      running NFSD threads;
 429  *                      return code is the size in bytes of the string
 430  *      On error:       return code is zero or a negative errno value
 431  */
 432 static ssize_t write_threads(struct file *file, char *buf, size_t size)
 433 {
 434         char *mesg = buf;
 435         int rv;
 436         struct net *net = netns(file);
 437
 438         if (size > 0) {
 439                 int newthreads;
 440                 rv = get_int(&mesg, &newthreads);
 441                 if (rv)
 442                         return rv;
 443                 if (newthreads < 0)
 444                         return -EINVAL;
 445                 rv = nfsd_svc(newthreads, net);
 446                 if (rv < 0)
 447                         return rv;
 448         } else
 449                 rv = nfsd_nrthreads(net);
 450
 451         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
 452 }
 453
 454 /**
 455  * write_pool_threads - Set or report the current number of threads per pool
 456  *
 457  * Input:
 458  *                      buf:            ignored
 459  *                      size:           zero
 460  *
 461  * OR
 462  *
 463  * Input:
 464  *                      buf:            C string containing whitespace-
 465  *                                      separated unsigned integer values
 466  *                                      representing the number of NFSD
 467  *                                      threads to start in each pool
 468  *                      size:           non-zero length of C string in @buf
 469  * Output:
 470  *      On success:     passed-in buffer filled with '\n'-terminated C
 471  *                      string containing integer values representing the
 472  *                      number of NFSD threads in each pool;
 473  *                      return code is the size in bytes of the string
 474  *      On error:       return code is zero or a negative errno value
 475  */
 476 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
 477 {
 478         /* if size > 0, look for an array of number of threads per node
 479          * and apply them  then write out number of threads per node as reply
 480          */
 481         char *mesg = buf;
 482         int i;
 483         int rv;
 484         int len;
 485         int npools;
 486         int *nthreads;
 487         struct net *net = netns(file);
 488
 489         mutex_lock(&nfsd_mutex);
 490         npools = nfsd_nrpools(net);
 491         if (npools == 0) {
 492                 /*
 493                  * NFS is shut down.  The admin can start it by
 494                  * writing to the threads file but NOT the pool_threads
 495                  * file, sorry.  Report zero threads.
 496                  */
 497                 mutex_unlock(&nfsd_mutex);
 498                 strcpy(buf, "0\n");
 499                 return strlen(buf);
 500         }
 501
 502         nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
 503         rv = -ENOMEM;
 504         if (nthreads == NULL)
 505                 goto out_free;
 506
 507         if (size > 0) {
 508                 for (i = 0; i < npools; i++) {
 509                         rv = get_int(&mesg, &nthreads[i]);
 510                         if (rv == -ENOENT)
 511                                 break;          /* fewer numbers than pools */
 512                         if (rv)
 513                                 goto out_free;  /* syntax error */
 514                         rv = -EINVAL;
 515                         if (nthreads[i] < 0)
 516                                 goto out_free;
 517                 }
 518                 rv = nfsd_set_nrthreads(i, nthreads, net);
 519                 if (rv)
 520                         goto out_free;
 521         }
 522
 523         rv = nfsd_get_nrthreads(npools, nthreads, net);
 524         if (rv)
 525                 goto out_free;
 526
 527         mesg = buf;
 528         size = SIMPLE_TRANSACTION_LIMIT;
 529         for (i = 0; i < npools && size > 0; i++) {
 530                 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
 531                 len = strlen(mesg);
 532                 size -= len;
 533                 mesg += len;
 534         }
 535         rv = mesg - buf;
 536 out_free:
 537         kfree(nthreads);
 538         mutex_unlock(&nfsd_mutex);
 539         return rv;
 540 }
 541
 542 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
 543 {
 544         char *mesg = buf;
 545         char *vers, *minorp, sign;
 546         int len, num, remaining;
 547         unsigned minor;
 548         ssize_t tlen = 0;
 549         char *sep;
 550         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 551
 552         if (size>0) {
 553                 if (nn->nfsd_serv)
 554                         /* Cannot change versions without updating
 555                          * nn->nfsd_serv->sv_xdrsize, and reallocing
 556                          * rq_argp and rq_resp
 557                          */
 558                         return -EBUSY;
 559                 if (buf[size-1] != '\n')
 560                         return -EINVAL;
 561                 buf[size-1] = 0;
 562
 563                 vers = mesg;
 564                 len = qword_get(&mesg, vers, size);
 565                 if (len <= 0) return -EINVAL;
 566                 do {
 567                         sign = *vers;
 568                         if (sign == '+' || sign == '-')
 569                                 num = simple_strtol((vers+1), &minorp, 0);
 570                         else
 571                                 num = simple_strtol(vers, &minorp, 0);
 572                         if (*minorp == '.') {
 573                                 if (num != 4)
 574                                         return -EINVAL;
 575                                 minor = simple_strtoul(minorp+1, NULL, 0);
 576                                 if (minor == 0)
 577                                         return -EINVAL;
 578                                 if (nfsd_minorversion(minor, sign == '-' ?
 579                                                      NFSD_CLEAR : NFSD_SET) < 0)
 580                                         return -EINVAL;
 581                                 goto next;
 582                         }
 583                         switch(num) {
 584                         case 2:
 585                         case 3:
 586                         case 4:
 587                                 nfsd_vers(num, sign == '-' ? NFSD_CLEAR : NFSD_SET);
 588                                 break;
 589                         default:
 590                                 return -EINVAL;
 591                         }
 592                 next:
 593                         vers += len + 1;
 594                 } while ((len = qword_get(&mesg, vers, size)) > 0);
 595                 /* If all get turned off, turn them back on, as
 596                  * having no versions is BAD
 597                  */
 598                 nfsd_reset_versions();
 599         }
 600
 601         /* Now write current state into reply buffer */
 602         len = 0;
 603         sep = "";
 604         remaining = SIMPLE_TRANSACTION_LIMIT;
 605         for (num=2 ; num <= 4 ; num++)
 606                 if (nfsd_vers(num, NFSD_AVAIL)) {
 607                         len = snprintf(buf, remaining, "%s%c%d", sep,
 608                                        nfsd_vers(num, NFSD_TEST)?'+':'-',
 609                                        num);
 610                         sep = " ";
 611
 612                         if (len >= remaining)
 613                                 break;
 614                         remaining -= len;
 615                         buf += len;
 616                         tlen += len;
 617                 }
 618         if (nfsd_vers(4, NFSD_AVAIL))
 619                 for (minor = 1; minor <= NFSD_SUPPORTED_MINOR_VERSION;
 620                      minor++) {
 621                         len = snprintf(buf, remaining, " %c4.%u",
 622                                         (nfsd_vers(4, NFSD_TEST) &&
 623                                          nfsd_minorversion(minor, NFSD_TEST)) ?
 624                                                 '+' : '-',
 625                                         minor);
 626
 627                         if (len >= remaining)
 628                                 break;
 629                         remaining -= len;
 630                         buf += len;
 631                         tlen += len;
 632                 }
 633
 634         len = snprintf(buf, remaining, "\n");
 635         if (len >= remaining)
 636                 return -EINVAL;
 637         return tlen + len;
 638 }
 639
 640 /**
 641  * write_versions - Set or report the available NFS protocol versions
 642  *
 643  * Input:
 644  *                      buf:            ignored
 645  *                      size:           zero
 646  * Output:
 647  *      On success:     passed-in buffer filled with '\n'-terminated C
 648  *                      string containing positive or negative integer
 649  *                      values representing the current status of each
 650  *                      protocol version;
 651  *                      return code is the size in bytes of the string
 652  *      On error:       return code is zero or a negative errno value
 653  *
 654  * OR
 655  *
 656  * Input:
 657  *                      buf:            C string containing whitespace-
 658  *                                      separated positive or negative
 659  *                                      integer values representing NFS
 660  *                                      protocol versions to enable ("+n")
 661  *                                      or disable ("-n")
 662  *                      size:           non-zero length of C string in @buf
 663  * Output:
 664  *      On success:     status of zero or more protocol versions has
 665  *                      been updated; passed-in buffer filled with
 666  *                      '\n'-terminated C string containing positive
 667  *                      or negative integer values representing the
 668  *                      current status of each protocol version;
 669  *                      return code is the size in bytes of the string
 670  *      On error:       return code is zero or a negative errno value
 671  */
 672 static ssize_t write_versions(struct file *file, char *buf, size_t size)
 673 {
 674         ssize_t rv;
 675
 676         mutex_lock(&nfsd_mutex);
 677         rv = __write_versions(file, buf, size);
 678         mutex_unlock(&nfsd_mutex);
 679         return rv;
 680 }
 681
 682 /*
 683  * Zero-length write.  Return a list of NFSD's current listener
 684  * transports.
 685  */
 686 static ssize_t __write_ports_names(char *buf, struct net *net)
 687 {
 688         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 689
 690         if (nn->nfsd_serv == NULL)
 691                 return 0;
 692         return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
 693 }
 694
 695 /*
 696  * A single 'fd' number was written, in which case it must be for
 697  * a socket of a supported family/protocol, and we use it as an
 698  * nfsd listener.
 699  */
 700 static ssize_t __write_ports_addfd(char *buf, struct net *net)
 701 {
 702         char *mesg = buf;
 703         int fd, err;
 704         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 705
 706         err = get_int(&mesg, &fd);
 707         if (err != 0 || fd < 0)
 708                 return -EINVAL;
 709
 710         if (svc_alien_sock(net, fd)) {
 711                 printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
 712                 return -EINVAL;
 713         }
 714
 715         err = nfsd_create_serv(net);
 716         if (err != 0)
 717                 return err;
 718
 719         err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT);
 720         if (err < 0) {
 721                 nfsd_destroy(net);
 722                 return err;
 723         }
 724
 725         /* Decrease the count, but don't shut down the service */
 726         nn->nfsd_serv->sv_nrthreads--;
 727         return err;
 728 }
 729
 730 /*
 731  * A transport listener is added by writing it's transport name and
 732  * a port number.
 733  */
 734 static ssize_t __write_ports_addxprt(char *buf, struct net *net)
 735 {
 736         char transport[16];
 737         struct svc_xprt *xprt;
 738         int port, err;
 739         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 740
 741         if (sscanf(buf, "%15s %5u", transport, &port) != 2)
 742                 return -EINVAL;
 743
 744         if (port < 1 || port > USHRT_MAX)
 745                 return -EINVAL;
 746
 747         err = nfsd_create_serv(net);
 748         if (err != 0)
 749                 return err;
 750
 751         err = svc_create_xprt(nn->nfsd_serv, transport, net,
 752                                 PF_INET, port, SVC_SOCK_ANONYMOUS);
 753         if (err < 0)
 754                 goto out_err;
 755
 756         err = svc_create_xprt(nn->nfsd_serv, transport, net,
 757                                 PF_INET6, port, SVC_SOCK_ANONYMOUS);
 758         if (err < 0 && err != -EAFNOSUPPORT)
 759                 goto out_close;
 760
 761         /* Decrease the count, but don't shut down the service */
 762         nn->nfsd_serv->sv_nrthreads--;
 763         return 0;
 764 out_close:
 765         xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
 766         if (xprt != NULL) {
 767                 svc_close_xprt(xprt);
 768                 svc_xprt_put(xprt);
 769         }
 770 out_err:
 771         nfsd_destroy(net);
 772         return err;
 773 }
 774
 775 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
 776                              struct net *net)
 777 {
 778         if (size == 0)
 779                 return __write_ports_names(buf, net);
 780
 781         if (isdigit(buf[0]))
 782                 return __write_ports_addfd(buf, net);
 783
 784         if (isalpha(buf[0]))
 785                 return __write_ports_addxprt(buf, net);
 786
 787         return -EINVAL;
 788 }
 789
 790 /**
 791  * write_ports - Pass a socket file descriptor or transport name to listen on
 792  *
 793  * Input:
 794  *                      buf:            ignored
 795  *                      size:           zero
 796  * Output:
 797  *      On success:     passed-in buffer filled with a '\n'-terminated C
 798  *                      string containing a whitespace-separated list of
 799  *                      named NFSD listeners;
 800  *                      return code is the size in bytes of the string
 801  *      On error:       return code is zero or a negative errno value
 802  *
 803  * OR
 804  *
 805  * Input:
 806  *                      buf:            C string containing an unsigned
 807  *                                      integer value representing a bound
 808  *                                      but unconnected socket that is to be
 809  *                                      used as an NFSD listener; listen(3)
 810  *                                      must be called for a SOCK_STREAM
 811  *                                      socket, otherwise it is ignored
 812  *                      size:           non-zero length of C string in @buf
 813  * Output:
 814  *      On success:     NFS service is started;
 815  *                      passed-in buffer filled with a '\n'-terminated C
 816  *                      string containing a unique alphanumeric name of
 817  *                      the listener;
 818  *                      return code is the size in bytes of the string
 819  *      On error:       return code is a negative errno value
 820  *
 821  * OR
 822  *
 823  * Input:
 824  *                      buf:            C string containing a transport
 825  *                                      name and an unsigned integer value
 826  *                                      representing the port to listen on,
 827  *                                      separated by whitespace
 828  *                      size:           non-zero length of C string in @buf
 829  * Output:
 830  *      On success:     returns zero; NFS service is started
 831  *      On error:       return code is a negative errno value
 832  */
 833 static ssize_t write_ports(struct file *file, char *buf, size_t size)
 834 {
 835         ssize_t rv;
 836
 837         mutex_lock(&nfsd_mutex);
 838         rv = __write_ports(file, buf, size, netns(file));
 839         mutex_unlock(&nfsd_mutex);
 840         return rv;
 841 }
 842
 843
 844 int nfsd_max_blksize;
 845
 846 /**
 847  * write_maxblksize - Set or report the current NFS blksize
 848  *
 849  * Input:
 850  *                      buf:            ignored
 851  *                      size:           zero
 852  *
 853  * OR
 854  *
 855  * Input:
 856  *                      buf:            C string containing an unsigned
 857  *                                      integer value representing the new
 858  *                                      NFS blksize
 859  *                      size:           non-zero length of C string in @buf
 860  * Output:
 861  *      On success:     passed-in buffer filled with '\n'-terminated C string
 862  *                      containing numeric value of the current NFS blksize
 863  *                      setting;
 864  *                      return code is the size in bytes of the string
 865  *      On error:       return code is zero or a negative errno value
 866  */
 867 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
 868 {
 869         char *mesg = buf;
 870         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 871
 872         if (size > 0) {
 873                 int bsize;
 874                 int rv = get_int(&mesg, &bsize);
 875                 if (rv)
 876                         return rv;
 877                 /* force bsize into allowed range and
 878                  * required alignment.
 879                  */
 880                 bsize = max_t(int, bsize, 1024);
 881                 bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
 882                 bsize &= ~(1024-1);
 883                 mutex_lock(&nfsd_mutex);
 884                 if (nn->nfsd_serv) {
 885                         mutex_unlock(&nfsd_mutex);
 886                         return -EBUSY;
 887                 }
 888                 nfsd_max_blksize = bsize;
 889                 mutex_unlock(&nfsd_mutex);
 890         }
 891
 892         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
 893                                                         nfsd_max_blksize);
 894 }
 895
 896 /**
 897  * write_maxconn - Set or report the current max number of connections
 898  *
 899  * Input:
 900  *                      buf:            ignored
 901  *                      size:           zero
 902  * OR
 903  *
 904  * Input:
 905  *                      buf:            C string containing an unsigned
 906  *                                      integer value representing the new
 907  *                                      number of max connections
 908  *                      size:           non-zero length of C string in @buf
 909  * Output:
 910  *      On success:     passed-in buffer filled with '\n'-terminated C string
 911  *                      containing numeric value of max_connections setting
 912  *                      for this net namespace;
 913  *                      return code is the size in bytes of the string
 914  *      On error:       return code is zero or a negative errno value
 915  */
 916 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
 917 {
 918         char *mesg = buf;
 919         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 920         unsigned int maxconn = nn->max_connections;
 921
 922         if (size > 0) {
 923                 int rv = get_uint(&mesg, &maxconn);
 924
 925                 if (rv)
 926                         return rv;
 927                 nn->max_connections = maxconn;
 928         }
 929
 930         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
 931 }
 932
 933 #ifdef CONFIG_NFSD_V4
 934 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
 935                                   time_t *time, struct nfsd_net *nn)
 936 {
 937         char *mesg = buf;
 938         int rv, i;
 939
 940         if (size > 0) {
 941                 if (nn->nfsd_serv)
 942                         return -EBUSY;
 943                 rv = get_int(&mesg, &i);
 944                 if (rv)
 945                         return rv;
 946                 /*
 947                  * Some sanity checking.  We don't have a reason for
 948                  * these particular numbers, but problems with the
 949                  * extremes are:
 950                  *      - Too short: the briefest network outage may
 951                  *        cause clients to lose all their locks.  Also,
 952                  *        the frequent polling may be wasteful.
 953                  *      - Too long: do you really want reboot recovery
 954                  *        to take more than an hour?  Or to make other
 955                  *        clients wait an hour before being able to
 956                  *        revoke a dead client's locks?
 957                  */
 958                 if (i < 10 || i > 3600)
 959                         return -EINVAL;
 960                 *time = i;
 961         }
 962
 963         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%ld\n", *time);
 964 }
 965
 966 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
 967                                 time_t *time, struct nfsd_net *nn)
 968 {
 969         ssize_t rv;
 970
 971         mutex_lock(&nfsd_mutex);
 972         rv = __nfsd4_write_time(file, buf, size, time, nn);
 973         mutex_unlock(&nfsd_mutex);
 974         return rv;
 975 }
 976
 977 /**
 978  * write_leasetime - Set or report the current NFSv4 lease time
 979  *
 980  * Input:
 981  *                      buf:            ignored
 982  *                      size:           zero
 983  *
 984  * OR
 985  *
 986  * Input:
 987  *                      buf:            C string containing an unsigned
 988  *                                      integer value representing the new
 989  *                                      NFSv4 lease expiry time
 990  *                      size:           non-zero length of C string in @buf
 991  * Output:
 992  *      On success:     passed-in buffer filled with '\n'-terminated C
 993  *                      string containing unsigned integer value of the
 994  *                      current lease expiry time;
 995  *                      return code is the size in bytes of the string
 996  *      On error:       return code is zero or a negative errno value
 997  */
 998 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
 999 {
1000         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1001         return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1002 }
1003
1004 /**
1005  * write_gracetime - Set or report current NFSv4 grace period time
1006  *
1007  * As above, but sets the time of the NFSv4 grace period.
1008  *
1009  * Note this should never be set to less than the *previous*
1010  * lease-period time, but we don't try to enforce this.  (In the common
1011  * case (a new boot), we don't know what the previous lease time was
1012  * anyway.)
1013  */
1014 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1015 {
1016         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1017         return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1018 }
1019
1020 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1021                                    struct nfsd_net *nn)
1022 {
1023         char *mesg = buf;
1024         char *recdir;
1025         int len, status;
1026
1027         if (size > 0) {
1028                 if (nn->nfsd_serv)
1029                         return -EBUSY;
1030                 if (size > PATH_MAX || buf[size-1] != '\n')
1031                         return -EINVAL;
1032                 buf[size-1] = 0;
1033
1034                 recdir = mesg;
1035                 len = qword_get(&mesg, recdir, size);
1036                 if (len <= 0)
1037                         return -EINVAL;
1038
1039                 status = nfs4_reset_recoverydir(recdir);
1040                 if (status)
1041                         return status;
1042         }
1043
1044         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1045                                                         nfs4_recoverydir());
1046 }
1047
1048 /**
1049  * write_recoverydir - Set or report the pathname of the recovery directory
1050  *
1051  * Input:
1052  *                      buf:            ignored
1053  *                      size:           zero
1054  *
1055  * OR
1056  *
1057  * Input:
1058  *                      buf:            C string containing the pathname
1059  *                                      of the directory on a local file
1060  *                                      system containing permanent NFSv4
1061  *                                      recovery data
1062  *                      size:           non-zero length of C string in @buf
1063  * Output:
1064  *      On success:     passed-in buffer filled with '\n'-terminated C string
1065  *                      containing the current recovery pathname setting;
1066  *                      return code is the size in bytes of the string
1067  *      On error:       return code is zero or a negative errno value
1068  */
1069 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1070 {
1071         ssize_t rv;
1072         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1073
1074         mutex_lock(&nfsd_mutex);
1075         rv = __write_recoverydir(file, buf, size, nn);
1076         mutex_unlock(&nfsd_mutex);
1077         return rv;
1078 }
1079
1080 /**
1081  * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1082  *
1083  * Input:
1084  *                      buf:            ignored
1085  *                      size:           zero
1086  * OR
1087  *
1088  * Input:
1089  *                      buf:            any value
1090  *                      size:           non-zero length of C string in @buf
1091  * Output:
1092  *                      passed-in buffer filled with "Y" or "N" with a newline
1093  *                      and NULL-terminated C string. This indicates whether
1094  *                      the grace period has ended in the current net
1095  *                      namespace. Return code is the size in bytes of the
1096  *                      string. Writing a string that starts with 'Y', 'y', or
1097  *                      '1' to the file will end the grace period for nfsd's v4
1098  *                      lock manager.
1099  */
1100 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1101 {
1102         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1103
1104         if (size > 0) {
1105                 switch(buf[0]) {
1106                 case 'Y':
1107                 case 'y':
1108                 case '1':
1109                         nfsd4_end_grace(nn);
1110                         break;
1111                 default:
1112                         return -EINVAL;
1113                 }
1114         }
1115
1116         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1117                          nn->grace_ended ? 'Y' : 'N');
1118 }
1119
1120 #endif
1121
1122 /*----------------------------------------------------------------------------*/
1123 /*
1124  *      populating the filesystem.
1125  */
1126
1127 static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
1128 {
1129         static struct tree_descr nfsd_files[] = {
1130                 [NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1131                 [NFSD_Export_features] = {"export_features",
1132                                         &export_features_operations, S_IRUGO},
1133                 [NFSD_FO_UnlockIP] = {"unlock_ip",
1134                                         &transaction_ops, S_IWUSR|S_IRUSR},
1135                 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1136                                         &transaction_ops, S_IWUSR|S_IRUSR},
1137                 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1138                 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1139                 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1140                 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1141                 [NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
1142                 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1143                 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1144                 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1145                 [NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1146 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1147                 [NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1148 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1149 #ifdef CONFIG_NFSD_V4
1150                 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1151                 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1152                 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1153                 [NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1154 #endif
1155                 /* last one */ {""}
1156         };
1157         struct net *net = data;
1158         int ret;
1159
1160         ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1161         if (ret)
1162                 return ret;
1163         sb->s_fs_info = get_net(net);
1164         return 0;
1165 }
1166
1167 static struct dentry *nfsd_mount(struct file_system_type *fs_type,
1168         int flags, const char *dev_name, void *data)
1169 {
1170         return mount_ns(fs_type, flags, current->nsproxy->net_ns, nfsd_fill_super);
1171 }
1172
1173 static void nfsd_umount(struct super_block *sb)
1174 {
1175         struct net *net = sb->s_fs_info;
1176
1177         kill_litter_super(sb);
1178         put_net(net);
1179 }
1180
1181 static struct file_system_type nfsd_fs_type = {
1182         .owner          = THIS_MODULE,
1183         .name           = "nfsd",
1184         .mount          = nfsd_mount,
1185         .kill_sb        = nfsd_umount,
1186 };
1187 MODULE_ALIAS_FS("nfsd");
1188
1189 #ifdef CONFIG_PROC_FS
1190 static int create_proc_exports_entry(void)
1191 {
1192         struct proc_dir_entry *entry;
1193
1194         entry = proc_mkdir("fs/nfs", NULL);
1195         if (!entry)
1196                 return -ENOMEM;
1197         entry = proc_create("exports", 0, entry,
1198                                  &exports_proc_operations);
1199         if (!entry) {
1200                 remove_proc_entry("fs/nfs", NULL);
1201                 return -ENOMEM;
1202         }
1203         return 0;
1204 }
1205 #else /* CONFIG_PROC_FS */
1206 static int create_proc_exports_entry(void)
1207 {
1208         return 0;
1209 }
1210 #endif
1211
1212 int nfsd_net_id;
1213
1214 static __net_init int nfsd_init_net(struct net *net)
1215 {
1216         int retval;
1217         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1218
1219         retval = nfsd_export_init(net);
1220         if (retval)
1221                 goto out_export_error;
1222         retval = nfsd_idmap_init(net);
1223         if (retval)
1224                 goto out_idmap_error;
1225         nn->nfsd4_lease = 90;   /* default lease time */
1226         nn->nfsd4_grace = 90;
1227         return 0;
1228
1229 out_idmap_error:
1230         nfsd_export_shutdown(net);
1231 out_export_error:
1232         return retval;
1233 }
1234
1235 static __net_exit void nfsd_exit_net(struct net *net)
1236 {
1237         nfsd_idmap_shutdown(net);
1238         nfsd_export_shutdown(net);
1239 }
1240
1241 static struct pernet_operations nfsd_net_ops = {
1242         .init = nfsd_init_net,
1243         .exit = nfsd_exit_net,
1244         .id   = &nfsd_net_id,
1245         .size = sizeof(struct nfsd_net),
1246 };
1247
1248 static int __init init_nfsd(void)
1249 {
1250         int retval;
1251         printk(KERN_INFO "Installing knfsd (copyright (C) 1996 okir@monad.swb.de).\n");
1252
1253         retval = register_pernet_subsys(&nfsd_net_ops);
1254         if (retval < 0)
1255                 return retval;
1256         retval = register_cld_notifier();
1257         if (retval)
1258                 goto out_unregister_pernet;
1259         retval = nfsd4_init_slabs();
1260         if (retval)
1261                 goto out_unregister_notifier;
1262         retval = nfsd4_init_pnfs();
1263         if (retval)
1264                 goto out_free_slabs;
1265         retval = nfsd_fault_inject_init(); /* nfsd fault injection controls */
1266         if (retval)
1267                 goto out_exit_pnfs;
1268         nfsd_stat_init();       /* Statistics */
1269         retval = nfsd_reply_cache_init();
1270         if (retval)
1271                 goto out_free_stat;
1272         nfsd_lockd_init();      /* lockd->nfsd callbacks */
1273         retval = create_proc_exports_entry();
1274         if (retval)
1275                 goto out_free_lockd;
1276         retval = register_filesystem(&nfsd_fs_type);
1277         if (retval)
1278                 goto out_free_all;
1279         return 0;
1280 out_free_all:
1281         remove_proc_entry("fs/nfs/exports", NULL);
1282         remove_proc_entry("fs/nfs", NULL);
1283 out_free_lockd:
1284         nfsd_lockd_shutdown();
1285         nfsd_reply_cache_shutdown();
1286 out_free_stat:
1287         nfsd_stat_shutdown();
1288         nfsd_fault_inject_cleanup();
1289 out_exit_pnfs:
1290         nfsd4_exit_pnfs();
1291 out_free_slabs:
1292         nfsd4_free_slabs();
1293 out_unregister_notifier:
1294         unregister_cld_notifier();
1295 out_unregister_pernet:
1296         unregister_pernet_subsys(&nfsd_net_ops);
1297         return retval;
1298 }
1299
1300 static void __exit exit_nfsd(void)
1301 {
1302         nfsd_reply_cache_shutdown();
1303         remove_proc_entry("fs/nfs/exports", NULL);
1304         remove_proc_entry("fs/nfs", NULL);
1305         nfsd_stat_shutdown();
1306         nfsd_lockd_shutdown();
1307         nfsd4_free_slabs();
1308         nfsd4_exit_pnfs();
1309         nfsd_fault_inject_cleanup();
1310         unregister_filesystem(&nfsd_fs_type);
1311         unregister_cld_notifier();
1312         unregister_pernet_subsys(&nfsd_net_ops);
1313 }
1314
1315 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1316 MODULE_LICENSE("GPL");
1317 module_init(init_nfsd)
1318 module_exit(exit_nfsd)