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