fs/nfsd/nfsctl.c

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