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 };
 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 };
 174
 175 static int export_features_show(struct seq_file *m, void *v)
 176 {
 177         seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
 178         return 0;
 179 }
 180
 181 static int export_features_open(struct inode *inode, struct file *file)
 182 {
 183         return single_open(file, export_features_show, NULL);
 184 }
 185
 186 static const struct file_operations export_features_operations = {
 187         .open           = export_features_open,
 188         .read           = seq_read,
 189         .llseek         = seq_lseek,
 190         .release        = single_release,
 191 };
 192
 193 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
 194 static int supported_enctypes_show(struct seq_file *m, void *v)
 195 {
 196         seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
 197         return 0;
 198 }
 199
 200 static int supported_enctypes_open(struct inode *inode, struct file *file)
 201 {
 202         return single_open(file, supported_enctypes_show, NULL);
 203 }
 204
 205 static const struct file_operations supported_enctypes_ops = {
 206         .open           = supported_enctypes_open,
 207         .read           = seq_read,
 208         .llseek         = seq_lseek,
 209         .release        = single_release,
 210 };
 211 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
 212
 213 static const struct file_operations pool_stats_operations = {
 214         .open           = nfsd_pool_stats_open,
 215         .read           = seq_read,
 216         .llseek         = seq_lseek,
 217         .release        = nfsd_pool_stats_release,
 218 };
 219
 220 static struct file_operations reply_cache_stats_operations = {
 221         .open           = nfsd_reply_cache_stats_open,
 222         .read           = seq_read,
 223         .llseek         = seq_lseek,
 224         .release        = single_release,
 225 };
 226
 227 /*----------------------------------------------------------------------------*/
 228 /*
 229  * payload - write methods
 230  */
 231
 232 static inline struct net *netns(struct file *file)
 233 {
 234         return file_inode(file)->i_sb->s_fs_info;
 235 }
 236
 237 /**
 238  * write_unlock_ip - Release all locks used by a client
 239  *
 240  * Experimental.
 241  *
 242  * Input:
 243  *                      buf:    '\n'-terminated C string containing a
 244  *                              presentation format IP address
 245  *                      size:   length of C string in @buf
 246  * Output:
 247  *      On success:     returns zero if all specified locks were released;
 248  *                      returns one if one or more locks were not released
 249  *      On error:       return code is negative errno value
 250  */
 251 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
 252 {
 253         struct sockaddr_storage address;
 254         struct sockaddr *sap = (struct sockaddr *)&address;
 255         size_t salen = sizeof(address);
 256         char *fo_path;
 257         struct net *net = netns(file);
 258
 259         /* sanity check */
 260         if (size == 0)
 261                 return -EINVAL;
 262
 263         if (buf[size-1] != '\n')
 264                 return -EINVAL;
 265
 266         fo_path = buf;
 267         if (qword_get(&buf, fo_path, size) < 0)
 268                 return -EINVAL;
 269
 270         if (rpc_pton(net, fo_path, size, sap, salen) == 0)
 271                 return -EINVAL;
 272
 273         return nlmsvc_unlock_all_by_ip(sap);
 274 }
 275
 276 /**
 277  * write_unlock_fs - Release all locks on a local file system
 278  *
 279  * Experimental.
 280  *
 281  * Input:
 282  *                      buf:    '\n'-terminated C string containing the
 283  *                              absolute pathname of a local file system
 284  *                      size:   length of C string in @buf
 285  * Output:
 286  *      On success:     returns zero if all specified locks were released;
 287  *                      returns one if one or more locks were not released
 288  *      On error:       return code is negative errno value
 289  */
 290 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
 291 {
 292         struct path path;
 293         char *fo_path;
 294         int error;
 295
 296         /* sanity check */
 297         if (size == 0)
 298                 return -EINVAL;
 299
 300         if (buf[size-1] != '\n')
 301                 return -EINVAL;
 302
 303         fo_path = buf;
 304         if (qword_get(&buf, fo_path, size) < 0)
 305                 return -EINVAL;
 306
 307         error = kern_path(fo_path, 0, &path);
 308         if (error)
 309                 return error;
 310
 311         /*
 312          * XXX: Needs better sanity checking.  Otherwise we could end up
 313          * releasing locks on the wrong file system.
 314          *
 315          * For example:
 316          * 1.  Does the path refer to a directory?
 317          * 2.  Is that directory a mount point, or
 318          * 3.  Is that directory the root of an exported file system?
 319          */
 320         error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
 321
 322         path_put(&path);
 323         return error;
 324 }
 325
 326 /**
 327  * write_filehandle - Get a variable-length NFS file handle by path
 328  *
 329  * On input, the buffer contains a '\n'-terminated C string comprised of
 330  * three alphanumeric words separated by whitespace.  The string may
 331  * contain escape sequences.
 332  *
 333  * Input:
 334  *                      buf:
 335  *                              domain:         client domain name
 336  *                              path:           export pathname
 337  *                              maxsize:        numeric maximum size of
 338  *                                              @buf
 339  *                      size:   length of C string in @buf
 340  * Output:
 341  *      On success:     passed-in buffer filled with '\n'-terminated C
 342  *                      string containing a ASCII hex text version
 343  *                      of the NFS file handle;
 344  *                      return code is the size in bytes of the string
 345  *      On error:       return code is negative errno value
 346  */
 347 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
 348 {
 349         char *dname, *path;
 350         int uninitialized_var(maxsize);
 351         char *mesg = buf;
 352         int len;
 353         struct auth_domain *dom;
 354         struct knfsd_fh fh;
 355
 356         if (size == 0)
 357                 return -EINVAL;
 358
 359         if (buf[size-1] != '\n')
 360                 return -EINVAL;
 361         buf[size-1] = 0;
 362
 363         dname = mesg;
 364         len = qword_get(&mesg, dname, size);
 365         if (len <= 0)
 366                 return -EINVAL;
 367
 368         path = dname+len+1;
 369         len = qword_get(&mesg, path, size);
 370         if (len <= 0)
 371                 return -EINVAL;
 372
 373         len = get_int(&mesg, &maxsize);
 374         if (len)
 375                 return len;
 376
 377         if (maxsize < NFS_FHSIZE)
 378                 return -EINVAL;
 379         maxsize = min(maxsize, NFS3_FHSIZE);
 380
 381         if (qword_get(&mesg, mesg, size)>0)
 382                 return -EINVAL;
 383
 384         /* we have all the words, they are in buf.. */
 385         dom = unix_domain_find(dname);
 386         if (!dom)
 387                 return -ENOMEM;
 388
 389         len = exp_rootfh(netns(file), dom, path, &fh,  maxsize);
 390         auth_domain_put(dom);
 391         if (len)
 392                 return len;
 393
 394         mesg = buf;
 395         len = SIMPLE_TRANSACTION_LIMIT;
 396         qword_addhex(&mesg, &len, (char*)&fh.fh_base, fh.fh_size);
 397         mesg[-1] = '\n';
 398         return mesg - buf;
 399 }
 400
 401 /**
 402  * write_threads - Start NFSD, or report the current number of running threads
 403  *
 404  * Input:
 405  *                      buf:            ignored
 406  *                      size:           zero
 407  * Output:
 408  *      On success:     passed-in buffer filled with '\n'-terminated C
 409  *                      string numeric value representing the number of
 410  *                      running NFSD threads;
 411  *                      return code is the size in bytes of the string
 412  *      On error:       return code is zero
 413  *
 414  * OR
 415  *
 416  * Input:
 417  *                      buf:            C string containing an unsigned
 418  *                                      integer value representing the
 419  *                                      number of NFSD threads to start
 420  *                      size:           non-zero length of C string in @buf
 421  * Output:
 422  *      On success:     NFS service is started;
 423  *                      passed-in buffer filled with '\n'-terminated C
 424  *                      string numeric value representing the number of
 425  *                      running NFSD threads;
 426  *                      return code is the size in bytes of the string
 427  *      On error:       return code is zero or a negative errno value
 428  */
 429 static ssize_t write_threads(struct file *file, char *buf, size_t size)
 430 {
 431         char *mesg = buf;
 432         int rv;
 433         struct net *net = netns(file);
 434
 435         if (size > 0) {
 436                 int newthreads;
 437                 rv = get_int(&mesg, &newthreads);
 438                 if (rv)
 439                         return rv;
 440                 if (newthreads < 0)
 441                         return -EINVAL;
 442                 rv = nfsd_svc(newthreads, net);
 443                 if (rv < 0)
 444                         return rv;
 445         } else
 446                 rv = nfsd_nrthreads(net);
 447
 448         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
 449 }
 450
 451 /**
 452  * write_pool_threads - Set or report the current number of threads per pool
 453  *
 454  * Input:
 455  *                      buf:            ignored
 456  *                      size:           zero
 457  *
 458  * OR
 459  *
 460  * Input:
 461  *                      buf:            C string containing whitespace-
 462  *                                      separated unsigned integer values
 463  *                                      representing the number of NFSD
 464  *                                      threads to start in each pool
 465  *                      size:           non-zero length of C string in @buf
 466  * Output:
 467  *      On success:     passed-in buffer filled with '\n'-terminated C
 468  *                      string containing integer values representing the
 469  *                      number of NFSD threads in each pool;
 470  *                      return code is the size in bytes of the string
 471  *      On error:       return code is zero or a negative errno value
 472  */
 473 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
 474 {
 475         /* if size > 0, look for an array of number of threads per node
 476          * and apply them  then write out number of threads per node as reply
 477          */
 478         char *mesg = buf;
 479         int i;
 480         int rv;
 481         int len;
 482         int npools;
 483         int *nthreads;
 484         struct net *net = netns(file);
 485
 486         mutex_lock(&nfsd_mutex);
 487         npools = nfsd_nrpools(net);
 488         if (npools == 0) {
 489                 /*
 490                  * NFS is shut down.  The admin can start it by
 491                  * writing to the threads file but NOT the pool_threads
 492                  * file, sorry.  Report zero threads.
 493                  */
 494                 mutex_unlock(&nfsd_mutex);
 495                 strcpy(buf, "0\n");
 496                 return strlen(buf);
 497         }
 498
 499         nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
 500         rv = -ENOMEM;
 501         if (nthreads == NULL)
 502                 goto out_free;
 503
 504         if (size > 0) {
 505                 for (i = 0; i < npools; i++) {
 506                         rv = get_int(&mesg, &nthreads[i]);
 507                         if (rv == -ENOENT)
 508                                 break;          /* fewer numbers than pools */
 509                         if (rv)
 510                                 goto out_free;  /* syntax error */
 511                         rv = -EINVAL;
 512                         if (nthreads[i] < 0)
 513                                 goto out_free;
 514                 }
 515                 rv = nfsd_set_nrthreads(i, nthreads, net);
 516                 if (rv)
 517                         goto out_free;
 518         }
 519
 520         rv = nfsd_get_nrthreads(npools, nthreads, net);
 521         if (rv)
 522                 goto out_free;
 523
 524         mesg = buf;
 525         size = SIMPLE_TRANSACTION_LIMIT;
 526         for (i = 0; i < npools && size > 0; i++) {
 527                 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
 528                 len = strlen(mesg);
 529                 size -= len;
 530                 mesg += len;
 531         }
 532         rv = mesg - buf;
 533 out_free:
 534         kfree(nthreads);
 535         mutex_unlock(&nfsd_mutex);
 536         return rv;
 537 }
 538
 539 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
 540 {
 541         char *mesg = buf;
 542         char *vers, *minorp, sign;
 543         int len, num, remaining;
 544         unsigned minor;
 545         ssize_t tlen = 0;
 546         char *sep;
 547         struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
 548
 549         if (size>0) {
 550                 if (nn->nfsd_serv)
 551                         /* Cannot change versions without updating
 552                          * nn->nfsd_serv->sv_xdrsize, and reallocing
 553                          * rq_argp and rq_resp
 554                          */
 555                         return -EBUSY;
 556                 if (buf[size-1] != '\n')
 557                         return -EINVAL;
 558                 buf[size-1] = 0;
 559
 560                 vers = mesg;
 561                 len = qword_get(&mesg, vers, size);
 562                 if (len <= 0) return -EINVAL;
 563                 do {
 564                         sign = *vers;
 565                         if (sign == '+' || sign == '-')
 566                                 num = simple_strtol((vers+1), &minorp, 0);
 567                         else
 568                                 num = simple_strtol(vers, &minorp, 0);
 569                         if (*minorp == '.') {
 570                                 if (num != 4)
 571                                         return -EINVAL;
 572                                 minor = simple_strtoul(minorp+1, NULL, 0);
 573                                 if (minor == 0)
 574                                         return -EINVAL;
 575                                 if (nfsd_minorversion(minor, sign == '-' ?
 576                                                      NFSD_CLEAR : NFSD_SET) < 0)
 577                                         return -EINVAL;
 578                                 goto next;
 579                         }
 580                         switch(num) {
 581                         case 2:
 582                         case 3:
 583                         case 4:
 584                                 nfsd_vers(num, sign == '-' ? NFSD_CLEAR : NFSD_SET);
 585                                 break;
 586                         default:
 587                                 return -EINVAL;
 588                         }
 589                 next:
 590                         vers += len + 1;
 591                 } while ((len = qword_get(&mesg, vers, size)) > 0);
 592                 /* If all get turned off, turn them back on, as
 593                  * having no versions is BAD
 594                  */
 595                 nfsd_reset_versions();
 596         }
 597
 598         /* Now write current state into reply buffer */
 599         len = 0;
 600         sep = "";
 601         remaining = SIMPLE_TRANSACTION_LIMIT;
 602         for (num=2 ; num <= 4 ; num++)
 603                 if (nfsd_vers(num, NFSD_AVAIL)) {
 604                         len = snprintf(buf, remaining, "%s%c%d", sep,
 605                                        nfsd_vers(num, NFSD_TEST)?'+':'-',
 606                                        num);
 607                         sep = " ";
 608
 609                         if (len >= remaining)
 610                                 break;
 611                         remaining -= len;
 612                         buf += len;
 613                         tlen += len;
 614                 }
 615         if (nfsd_vers(4, NFSD_AVAIL))
 616                 for (minor = 1; minor <= NFSD_SUPPORTED_MINOR_VERSION;
 617                      minor++) {
 618                         len = snprintf(buf, remaining, " %c4.%u",
 619                                         (nfsd_vers(4, NFSD_TEST) &&
 620                                          nfsd_minorversion(minor, NFSD_TEST)) ?
 621                                                 '+' : '-',
 622                                         minor);
 623
 624                         if (len >= remaining)
 625                                 break;
 626                         remaining -= len;
 627                         buf += len;
 628                         tlen += len;
 629                 }
 630
 631         len = snprintf(buf, remaining, "\n");
 632         if (len >= remaining)
 633                 return -EINVAL;
 634         return tlen + len;
 635 }
 636
 637 /**
 638  * write_versions - Set or report the available NFS protocol versions
 639  *
 640  * Input:
 641  *                      buf:            ignored
 642  *                      size:           zero
 643  * Output:
 644  *      On success:     passed-in buffer filled with '\n'-terminated C
 645  *                      string containing positive or negative integer
 646  *                      values representing the current status of each
 647  *                      protocol version;
 648  *                      return code is the size in bytes of the string
 649  *      On error:       return code is zero or a negative errno value
 650  *
 651  * OR
 652  *
 653  * Input:
 654  *                      buf:            C string containing whitespace-
 655  *                                      separated positive or negative
 656  *                                      integer values representing NFS
 657  *                                      protocol versions to enable ("+n")
 658  *                                      or disable ("-n")
 659  *                      size:           non-zero length of C string in @buf
 660  * Output:
 661  *      On success:     status of zero or more protocol versions has
 662  *                      been updated; passed-in buffer filled with
 663  *                      '\n'-terminated C string containing positive
 664  *                      or negative integer values representing the
 665  *                      current status of each protocol version;
 666  *                      return code is the size in bytes of the string
 667  *      On error:       return code is zero or a negative errno value
 668  */
 669 static ssize_t write_versions(struct file *file, char *buf, size_t size)
 670 {
 671         ssize_t rv;
 672
 673         mutex_lock(&nfsd_mutex);
 674         rv = __write_versions(file, buf, size);
 675         mutex_unlock(&nfsd_mutex);
 676         return rv;
 677 }
 678
 679 /*
 680  * Zero-length write.  Return a list of NFSD's current listener
 681  * transports.
 682  */
 683 static ssize_t __write_ports_names(char *buf, struct net *net)
 684 {
 685         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 686
 687         if (nn->nfsd_serv == NULL)
 688                 return 0;
 689         return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
 690 }
 691
 692 /*
 693  * A single 'fd' number was written, in which case it must be for
 694  * a socket of a supported family/protocol, and we use it as an
 695  * nfsd listener.
 696  */
 697 static ssize_t __write_ports_addfd(char *buf, struct net *net)
 698 {
 699         char *mesg = buf;
 700         int fd, err;
 701         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 702
 703         err = get_int(&mesg, &fd);
 704         if (err != 0 || fd < 0)
 705                 return -EINVAL;
 706
 707         if (svc_alien_sock(net, fd)) {
 708                 printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
 709                 return -EINVAL;
 710         }
 711
 712         err = nfsd_create_serv(net);
 713         if (err != 0)
 714                 return err;
 715
 716         err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT);
 717         if (err < 0) {
 718                 nfsd_destroy(net);
 719                 return err;
 720         }
 721
 722         /* Decrease the count, but don't shut down the service */
 723         nn->nfsd_serv->sv_nrthreads--;
 724         return err;
 725 }
 726
 727 /*
 728  * A transport listener is added by writing it's transport name and
 729  * a port number.
 730  */
 731 static ssize_t __write_ports_addxprt(char *buf, struct net *net)
 732 {
 733         char transport[16];
 734         struct svc_xprt *xprt;
 735         int port, err;
 736         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 737
 738         if (sscanf(buf, "%15s %5u", transport, &port) != 2)
 739                 return -EINVAL;
 740
 741         if (port < 1 || port > USHRT_MAX)
 742                 return -EINVAL;
 743
 744         err = nfsd_create_serv(net);
 745         if (err != 0)
 746                 return err;
 747
 748         err = svc_create_xprt(nn->nfsd_serv, transport, net,
 749                                 PF_INET, port, SVC_SOCK_ANONYMOUS);
 750         if (err < 0)
 751                 goto out_err;
 752
 753         err = svc_create_xprt(nn->nfsd_serv, transport, net,
 754                                 PF_INET6, port, SVC_SOCK_ANONYMOUS);
 755         if (err < 0 && err != -EAFNOSUPPORT)
 756                 goto out_close;
 757
 758         /* Decrease the count, but don't shut down the service */
 759         nn->nfsd_serv->sv_nrthreads--;
 760         return 0;
 761 out_close:
 762         xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
 763         if (xprt != NULL) {
 764                 svc_close_xprt(xprt);
 765                 svc_xprt_put(xprt);
 766         }
 767 out_err:
 768         if (!list_empty(&nn->nfsd_serv->sv_permsocks))
 769                 nn->nfsd_serv->sv_nrthreads--;
 770          else
 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                         if (!nn->nfsd_serv)
1110                                 return -EBUSY;
1111                         nfsd4_end_grace(nn);
1112                         break;
1113                 default:
1114                         return -EINVAL;
1115                 }
1116         }
1117
1118         return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1119                          nn->grace_ended ? 'Y' : 'N');
1120 }
1121
1122 #endif
1123
1124 /*----------------------------------------------------------------------------*/
1125 /*
1126  *      populating the filesystem.
1127  */
1128
1129 static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
1130 {
1131         static struct tree_descr nfsd_files[] = {
1132                 [NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1133                 [NFSD_Export_features] = {"export_features",
1134                                         &export_features_operations, S_IRUGO},
1135                 [NFSD_FO_UnlockIP] = {"unlock_ip",
1136                                         &transaction_ops, S_IWUSR|S_IRUSR},
1137                 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1138                                         &transaction_ops, S_IWUSR|S_IRUSR},
1139                 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1140                 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1141                 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1142                 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1143                 [NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
1144                 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1145                 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1146                 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1147                 [NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1148 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1149                 [NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1150 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1151 #ifdef CONFIG_NFSD_V4
1152                 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1153                 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1154                 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1155                 [NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1156 #endif
1157                 /* last one */ {""}
1158         };
1159         get_net(sb->s_fs_info);
1160         return simple_fill_super(sb, 0x6e667364, nfsd_files);
1161 }
1162
1163 static struct dentry *nfsd_mount(struct file_system_type *fs_type,
1164         int flags, const char *dev_name, void *data)
1165 {
1166         struct net *net = current->nsproxy->net_ns;
1167         return mount_ns(fs_type, flags, data, net, net->user_ns, nfsd_fill_super);
1168 }
1169
1170 static void nfsd_umount(struct super_block *sb)
1171 {
1172         struct net *net = sb->s_fs_info;
1173
1174         kill_litter_super(sb);
1175         put_net(net);
1176 }
1177
1178 static struct file_system_type nfsd_fs_type = {
1179         .owner          = THIS_MODULE,
1180         .name           = "nfsd",
1181         .mount          = nfsd_mount,
1182         .kill_sb        = nfsd_umount,
1183 };
1184 MODULE_ALIAS_FS("nfsd");
1185
1186 #ifdef CONFIG_PROC_FS
1187 static int create_proc_exports_entry(void)
1188 {
1189         struct proc_dir_entry *entry;
1190
1191         entry = proc_mkdir("fs/nfs", NULL);
1192         if (!entry)
1193                 return -ENOMEM;
1194         entry = proc_create("exports", 0, entry,
1195                                  &exports_proc_operations);
1196         if (!entry) {
1197                 remove_proc_entry("fs/nfs", NULL);
1198                 return -ENOMEM;
1199         }
1200         return 0;
1201 }
1202 #else /* CONFIG_PROC_FS */
1203 static int create_proc_exports_entry(void)
1204 {
1205         return 0;
1206 }
1207 #endif
1208
1209 int nfsd_net_id;
1210
1211 static __net_init int nfsd_init_net(struct net *net)
1212 {
1213         int retval;
1214         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1215
1216         retval = nfsd_export_init(net);
1217         if (retval)
1218                 goto out_export_error;
1219         retval = nfsd_idmap_init(net);
1220         if (retval)
1221                 goto out_idmap_error;
1222         nn->nfsd4_lease = 90;   /* default lease time */
1223         nn->nfsd4_grace = 90;
1224         nn->clverifier_counter = prandom_u32();
1225         nn->clientid_counter = prandom_u32();
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_pernet_subsys(&nfsd_net_ops);
1253         if (retval < 0)
1254                 return retval;
1255         retval = register_cld_notifier();
1256         if (retval)
1257                 goto out_unregister_pernet;
1258         retval = nfsd4_init_slabs();
1259         if (retval)
1260                 goto out_unregister_notifier;
1261         retval = nfsd4_init_pnfs();
1262         if (retval)
1263                 goto out_free_slabs;
1264         retval = nfsd_fault_inject_init(); /* nfsd fault injection controls */
1265         if (retval)
1266                 goto out_exit_pnfs;
1267         nfsd_stat_init();       /* Statistics */
1268         retval = nfsd_reply_cache_init();
1269         if (retval)
1270                 goto out_free_stat;
1271         nfsd_lockd_init();      /* lockd->nfsd callbacks */
1272         retval = create_proc_exports_entry();
1273         if (retval)
1274                 goto out_free_lockd;
1275         retval = register_filesystem(&nfsd_fs_type);
1276         if (retval)
1277                 goto out_free_all;
1278         return 0;
1279 out_free_all:
1280         remove_proc_entry("fs/nfs/exports", NULL);
1281         remove_proc_entry("fs/nfs", NULL);
1282 out_free_lockd:
1283         nfsd_lockd_shutdown();
1284         nfsd_reply_cache_shutdown();
1285 out_free_stat:
1286         nfsd_stat_shutdown();
1287         nfsd_fault_inject_cleanup();
1288 out_exit_pnfs:
1289         nfsd4_exit_pnfs();
1290 out_free_slabs:
1291         nfsd4_free_slabs();
1292 out_unregister_notifier:
1293         unregister_cld_notifier();
1294 out_unregister_pernet:
1295         unregister_pernet_subsys(&nfsd_net_ops);
1296         return retval;
1297 }
1298
1299 static void __exit exit_nfsd(void)
1300 {
1301         nfsd_reply_cache_shutdown();
1302         remove_proc_entry("fs/nfs/exports", NULL);
1303         remove_proc_entry("fs/nfs", NULL);
1304         nfsd_stat_shutdown();
1305         nfsd_lockd_shutdown();
1306         nfsd4_free_slabs();
1307         nfsd4_exit_pnfs();
1308         nfsd_fault_inject_cleanup();
1309         unregister_filesystem(&nfsd_fs_type);
1310         unregister_cld_notifier();
1311         unregister_pernet_subsys(&nfsd_net_ops);
1312 }
1313
1314 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1315 MODULE_LICENSE("GPL");
1316 module_init(init_nfsd)
1317 module_exit(exit_nfsd)