fs/afs/super.c

   1 /* AFS superblock handling
   2  *
   3  * Copyright (c) 2002, 2007 Red Hat, Inc. All rights reserved.
   4  *
   5  * This software may be freely redistributed under the terms of the
   6  * GNU General Public License.
   7  *
   8  * You should have received a copy of the GNU General Public License
   9  * along with this program; if not, write to the Free Software
  10  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  11  *
  12  * Authors: David Howells <dhowells@redhat.com>
  13  *          David Woodhouse <dwmw2@infradead.org>
  14  *
  15  */
  16
  17 #include <linux/kernel.h>
  18 #include <linux/module.h>
  19 #include <linux/mount.h>
  20 #include <linux/init.h>
  21 #include <linux/slab.h>
  22 #include <linux/fs.h>
  23 #include <linux/pagemap.h>
  24 #include <linux/parser.h>
  25 #include <linux/statfs.h>
  26 #include <linux/sched.h>
  27 #include <linux/nsproxy.h>
  28 #include <linux/magic.h>
  29 #include <net/net_namespace.h>
  30 #include "internal.h"
  31
  32 static void afs_i_init_once(void *foo);
  33 static struct dentry *afs_mount(struct file_system_type *fs_type,
  34                       int flags, const char *dev_name, void *data);
  35 static void afs_kill_super(struct super_block *sb);
  36 static struct inode *afs_alloc_inode(struct super_block *sb);
  37 static void afs_destroy_inode(struct inode *inode);
  38 static int afs_statfs(struct dentry *dentry, struct kstatfs *buf);
  39 static int afs_show_devname(struct seq_file *m, struct dentry *root);
  40 static int afs_show_options(struct seq_file *m, struct dentry *root);
  41
  42 struct file_system_type afs_fs_type = {
  43         .owner          = THIS_MODULE,
  44         .name           = "afs",
  45         .mount          = afs_mount,
  46         .kill_sb        = afs_kill_super,
  47         .fs_flags       = 0,
  48 };
  49 MODULE_ALIAS_FS("afs");
  50
  51 static const struct super_operations afs_super_ops = {
  52         .statfs         = afs_statfs,
  53         .alloc_inode    = afs_alloc_inode,
  54         .drop_inode     = afs_drop_inode,
  55         .destroy_inode  = afs_destroy_inode,
  56         .evict_inode    = afs_evict_inode,
  57         .show_devname   = afs_show_devname,
  58         .show_options   = afs_show_options,
  59 };
  60
  61 static struct kmem_cache *afs_inode_cachep;
  62 static atomic_t afs_count_active_inodes;
  63
  64 enum {
  65         afs_no_opt,
  66         afs_opt_cell,
  67         afs_opt_dyn,
  68         afs_opt_rwpath,
  69         afs_opt_vol,
  70         afs_opt_autocell,
  71 };
  72
  73 static const match_table_t afs_options_list = {
  74         { afs_opt_cell,         "cell=%s"       },
  75         { afs_opt_dyn,          "dyn"           },
  76         { afs_opt_rwpath,       "rwpath"        },
  77         { afs_opt_vol,          "vol=%s"        },
  78         { afs_opt_autocell,     "autocell"      },
  79         { afs_no_opt,           NULL            },
  80 };
  81
  82 /*
  83  * initialise the filesystem
  84  */
  85 int __init afs_fs_init(void)
  86 {
  87         int ret;
  88
  89         _enter("");
  90
  91         /* create ourselves an inode cache */
  92         atomic_set(&afs_count_active_inodes, 0);
  93
  94         ret = -ENOMEM;
  95         afs_inode_cachep = kmem_cache_create("afs_inode_cache",
  96                                              sizeof(struct afs_vnode),
  97                                              0,
  98                                              SLAB_HWCACHE_ALIGN|SLAB_ACCOUNT,
  99                                              afs_i_init_once);
 100         if (!afs_inode_cachep) {
 101                 printk(KERN_NOTICE "kAFS: Failed to allocate inode cache\n");
 102                 return ret;
 103         }
 104
 105         /* now export our filesystem to lesser mortals */
 106         ret = register_filesystem(&afs_fs_type);
 107         if (ret < 0) {
 108                 kmem_cache_destroy(afs_inode_cachep);
 109                 _leave(" = %d", ret);
 110                 return ret;
 111         }
 112
 113         _leave(" = 0");
 114         return 0;
 115 }
 116
 117 /*
 118  * clean up the filesystem
 119  */
 120 void __exit afs_fs_exit(void)
 121 {
 122         _enter("");
 123
 124         afs_mntpt_kill_timer();
 125         unregister_filesystem(&afs_fs_type);
 126
 127         if (atomic_read(&afs_count_active_inodes) != 0) {
 128                 printk("kAFS: %d active inode objects still present\n",
 129                        atomic_read(&afs_count_active_inodes));
 130                 BUG();
 131         }
 132
 133         /*
 134          * Make sure all delayed rcu free inodes are flushed before we
 135          * destroy cache.
 136          */
 137         rcu_barrier();
 138         kmem_cache_destroy(afs_inode_cachep);
 139         _leave("");
 140 }
 141
 142 /*
 143  * Display the mount device name in /proc/mounts.
 144  */
 145 static int afs_show_devname(struct seq_file *m, struct dentry *root)
 146 {
 147         struct afs_super_info *as = AFS_FS_S(root->d_sb);
 148         struct afs_volume *volume = as->volume;
 149         struct afs_cell *cell = as->cell;
 150         const char *suf = "";
 151         char pref = '%';
 152
 153         if (as->dyn_root) {
 154                 seq_puts(m, "none");
 155                 return 0;
 156         }
 157
 158         switch (volume->type) {
 159         case AFSVL_RWVOL:
 160                 break;
 161         case AFSVL_ROVOL:
 162                 pref = '#';
 163                 if (volume->type_force)
 164                         suf = ".readonly";
 165                 break;
 166         case AFSVL_BACKVOL:
 167                 pref = '#';
 168                 suf = ".backup";
 169                 break;
 170         }
 171
 172         seq_printf(m, "%c%s:%s%s", pref, cell->name, volume->name, suf);
 173         return 0;
 174 }
 175
 176 /*
 177  * Display the mount options in /proc/mounts.
 178  */
 179 static int afs_show_options(struct seq_file *m, struct dentry *root)
 180 {
 181         struct afs_super_info *as = AFS_FS_S(root->d_sb);
 182
 183         if (as->dyn_root)
 184                 seq_puts(m, ",dyn");
 185         if (test_bit(AFS_VNODE_AUTOCELL, &AFS_FS_I(d_inode(root))->flags))
 186                 seq_puts(m, ",autocell");
 187         return 0;
 188 }
 189
 190 /*
 191  * parse the mount options
 192  * - this function has been shamelessly adapted from the ext3 fs which
 193  *   shamelessly adapted it from the msdos fs
 194  */
 195 static int afs_parse_options(struct afs_mount_params *params,
 196                              char *options, const char **devname)
 197 {
 198         struct afs_cell *cell;
 199         substring_t args[MAX_OPT_ARGS];
 200         char *p;
 201         int token;
 202
 203         _enter("%s", options);
 204
 205         options[PAGE_SIZE - 1] = 0;
 206
 207         while ((p = strsep(&options, ","))) {
 208                 if (!*p)
 209                         continue;
 210
 211                 token = match_token(p, afs_options_list, args);
 212                 switch (token) {
 213                 case afs_opt_cell:
 214                         rcu_read_lock();
 215                         cell = afs_lookup_cell_rcu(params->net,
 216                                                    args[0].from,
 217                                                    args[0].to - args[0].from);
 218                         rcu_read_unlock();
 219                         if (IS_ERR(cell))
 220                                 return PTR_ERR(cell);
 221                         afs_put_cell(params->net, params->cell);
 222                         params->cell = cell;
 223                         break;
 224
 225                 case afs_opt_rwpath:
 226                         params->rwpath = true;
 227                         break;
 228
 229                 case afs_opt_vol:
 230                         *devname = args[0].from;
 231                         break;
 232
 233                 case afs_opt_autocell:
 234                         params->autocell = true;
 235                         break;
 236
 237                 case afs_opt_dyn:
 238                         params->dyn_root = true;
 239                         break;
 240
 241                 default:
 242                         printk(KERN_ERR "kAFS:"
 243                                " Unknown or invalid mount option: '%s'\n", p);
 244                         return -EINVAL;
 245                 }
 246         }
 247
 248         _leave(" = 0");
 249         return 0;
 250 }
 251
 252 /*
 253  * parse a device name to get cell name, volume name, volume type and R/W
 254  * selector
 255  * - this can be one of the following:
 256  *      "%[cell:]volume[.]"             R/W volume
 257  *      "#[cell:]volume[.]"             R/O or R/W volume (rwpath=0),
 258  *                                       or R/W (rwpath=1) volume
 259  *      "%[cell:]volume.readonly"       R/O volume
 260  *      "#[cell:]volume.readonly"       R/O volume
 261  *      "%[cell:]volume.backup"         Backup volume
 262  *      "#[cell:]volume.backup"         Backup volume
 263  */
 264 static int afs_parse_device_name(struct afs_mount_params *params,
 265                                  const char *name)
 266 {
 267         struct afs_cell *cell;
 268         const char *cellname, *suffix;
 269         int cellnamesz;
 270
 271         _enter(",%s", name);
 272
 273         if (!name) {
 274                 printk(KERN_ERR "kAFS: no volume name specified\n");
 275                 return -EINVAL;
 276         }
 277
 278         if ((name[0] != '%' && name[0] != '#') || !name[1]) {
 279                 printk(KERN_ERR "kAFS: unparsable volume name\n");
 280                 return -EINVAL;
 281         }
 282
 283         /* determine the type of volume we're looking for */
 284         params->type = AFSVL_ROVOL;
 285         params->force = false;
 286         if (params->rwpath || name[0] == '%') {
 287                 params->type = AFSVL_RWVOL;
 288                 params->force = true;
 289         }
 290         name++;
 291
 292         /* split the cell name out if there is one */
 293         params->volname = strchr(name, ':');
 294         if (params->volname) {
 295                 cellname = name;
 296                 cellnamesz = params->volname - name;
 297                 params->volname++;
 298         } else {
 299                 params->volname = name;
 300                 cellname = NULL;
 301                 cellnamesz = 0;
 302         }
 303
 304         /* the volume type is further affected by a possible suffix */
 305         suffix = strrchr(params->volname, '.');
 306         if (suffix) {
 307                 if (strcmp(suffix, ".readonly") == 0) {
 308                         params->type = AFSVL_ROVOL;
 309                         params->force = true;
 310                 } else if (strcmp(suffix, ".backup") == 0) {
 311                         params->type = AFSVL_BACKVOL;
 312                         params->force = true;
 313                 } else if (suffix[1] == 0) {
 314                 } else {
 315                         suffix = NULL;
 316                 }
 317         }
 318
 319         params->volnamesz = suffix ?
 320                 suffix - params->volname : strlen(params->volname);
 321
 322         _debug("cell %*.*s [%p]",
 323                cellnamesz, cellnamesz, cellname ?: "", params->cell);
 324
 325         /* lookup the cell record */
 326         if (cellname || !params->cell) {
 327                 cell = afs_lookup_cell(params->net, cellname, cellnamesz,
 328                                        NULL, false);
 329                 if (IS_ERR(cell)) {
 330                         printk(KERN_ERR "kAFS: unable to lookup cell '%*.*s'\n",
 331                                cellnamesz, cellnamesz, cellname ?: "");
 332                         return PTR_ERR(cell);
 333                 }
 334                 afs_put_cell(params->net, params->cell);
 335                 params->cell = cell;
 336         }
 337
 338         _debug("CELL:%s [%p] VOLUME:%*.*s SUFFIX:%s TYPE:%d%s",
 339                params->cell->name, params->cell,
 340                params->volnamesz, params->volnamesz, params->volname,
 341                suffix ?: "-", params->type, params->force ? " FORCE" : "");
 342
 343         return 0;
 344 }
 345
 346 /*
 347  * check a superblock to see if it's the one we're looking for
 348  */
 349 static int afs_test_super(struct super_block *sb, void *data)
 350 {
 351         struct afs_super_info *as1 = data;
 352         struct afs_super_info *as = AFS_FS_S(sb);
 353
 354         return (as->net == as1->net &&
 355                 as->volume &&
 356                 as->volume->vid == as1->volume->vid);
 357 }
 358
 359 static int afs_dynroot_test_super(struct super_block *sb, void *data)
 360 {
 361         return false;
 362 }
 363
 364 static int afs_set_super(struct super_block *sb, void *data)
 365 {
 366         struct afs_super_info *as = data;
 367
 368         sb->s_fs_info = as;
 369         return set_anon_super(sb, NULL);
 370 }
 371
 372 /*
 373  * fill in the superblock
 374  */
 375 static int afs_fill_super(struct super_block *sb,
 376                           struct afs_mount_params *params)
 377 {
 378         struct afs_super_info *as = AFS_FS_S(sb);
 379         struct afs_fid fid;
 380         struct inode *inode = NULL;
 381         int ret;
 382
 383         _enter("");
 384
 385         /* fill in the superblock */
 386         sb->s_blocksize         = PAGE_SIZE;
 387         sb->s_blocksize_bits    = PAGE_SHIFT;
 388         sb->s_magic             = AFS_FS_MAGIC;
 389         sb->s_op                = &afs_super_ops;
 390         if (!as->dyn_root)
 391                 sb->s_xattr     = afs_xattr_handlers;
 392         ret = super_setup_bdi(sb);
 393         if (ret)
 394                 return ret;
 395         sb->s_bdi->ra_pages     = VM_MAX_READAHEAD * 1024 / PAGE_SIZE;
 396
 397         /* allocate the root inode and dentry */
 398         if (as->dyn_root) {
 399                 inode = afs_iget_pseudo_dir(sb, true);
 400                 sb->s_flags     |= SB_RDONLY;
 401         } else {
 402                 sprintf(sb->s_id, "%u", as->volume->vid);
 403                 afs_activate_volume(as->volume);
 404                 fid.vid         = as->volume->vid;
 405                 fid.vnode       = 1;
 406                 fid.unique      = 1;
 407                 inode = afs_iget(sb, params->key, &fid, NULL, NULL, NULL);
 408         }
 409
 410         if (IS_ERR(inode))
 411                 return PTR_ERR(inode);
 412
 413         if (params->autocell || params->dyn_root)
 414                 set_bit(AFS_VNODE_AUTOCELL, &AFS_FS_I(inode)->flags);
 415
 416         ret = -ENOMEM;
 417         sb->s_root = d_make_root(inode);
 418         if (!sb->s_root)
 419                 goto error;
 420
 421         sb->s_d_op = &afs_fs_dentry_operations;
 422
 423         _leave(" = 0");
 424         return 0;
 425
 426 error:
 427         _leave(" = %d", ret);
 428         return ret;
 429 }
 430
 431 static struct afs_super_info *afs_alloc_sbi(struct afs_mount_params *params)
 432 {
 433         struct afs_super_info *as;
 434
 435         as = kzalloc(sizeof(struct afs_super_info), GFP_KERNEL);
 436         if (as) {
 437                 as->net = afs_get_net(params->net);
 438                 if (params->dyn_root)
 439                         as->dyn_root = true;
 440                 else
 441                         as->cell = afs_get_cell(params->cell);
 442         }
 443         return as;
 444 }
 445
 446 static void afs_destroy_sbi(struct afs_super_info *as)
 447 {
 448         if (as) {
 449                 afs_put_volume(as->cell, as->volume);
 450                 afs_put_cell(as->net, as->cell);
 451                 afs_put_net(as->net);
 452                 kfree(as);
 453         }
 454 }
 455
 456 /*
 457  * get an AFS superblock
 458  */
 459 static struct dentry *afs_mount(struct file_system_type *fs_type,
 460                                 int flags, const char *dev_name, void *options)
 461 {
 462         struct afs_mount_params params;
 463         struct super_block *sb;
 464         struct afs_volume *candidate;
 465         struct key *key;
 466         struct afs_super_info *as;
 467         int ret;
 468
 469         _enter(",,%s,%p", dev_name, options);
 470
 471         memset(&params, 0, sizeof(params));
 472         params.net = &__afs_net;
 473
 474         ret = -EINVAL;
 475         if (current->nsproxy->net_ns != &init_net)
 476                 goto error;
 477
 478         /* parse the options and device name */
 479         if (options) {
 480                 ret = afs_parse_options(&params, options, &dev_name);
 481                 if (ret < 0)
 482                         goto error;
 483         }
 484
 485         if (!params.dyn_root) {
 486                 ret = afs_parse_device_name(&params, dev_name);
 487                 if (ret < 0)
 488                         goto error;
 489
 490                 /* try and do the mount securely */
 491                 key = afs_request_key(params.cell);
 492                 if (IS_ERR(key)) {
 493                         _leave(" = %ld [key]", PTR_ERR(key));
 494                         ret = PTR_ERR(key);
 495                         goto error;
 496                 }
 497                 params.key = key;
 498         }
 499
 500         /* allocate a superblock info record */
 501         ret = -ENOMEM;
 502         as = afs_alloc_sbi(&params);
 503         if (!as)
 504                 goto error_key;
 505
 506         if (!params.dyn_root) {
 507                 /* Assume we're going to need a volume record; at the very
 508                  * least we can use it to update the volume record if we have
 509                  * one already.  This checks that the volume exists within the
 510                  * cell.
 511                  */
 512                 candidate = afs_create_volume(&params);
 513                 if (IS_ERR(candidate)) {
 514                         ret = PTR_ERR(candidate);
 515                         goto error_as;
 516                 }
 517
 518                 as->volume = candidate;
 519         }
 520
 521         /* allocate a deviceless superblock */
 522         sb = sget(fs_type,
 523                   as->dyn_root ? afs_dynroot_test_super : afs_test_super,
 524                   afs_set_super, flags, as);
 525         if (IS_ERR(sb)) {
 526                 ret = PTR_ERR(sb);
 527                 goto error_as;
 528         }
 529
 530         if (!sb->s_root) {
 531                 /* initial superblock/root creation */
 532                 _debug("create");
 533                 ret = afs_fill_super(sb, &params);
 534                 if (ret < 0)
 535                         goto error_sb;
 536                 as = NULL;
 537                 sb->s_flags |= SB_ACTIVE;
 538         } else {
 539                 _debug("reuse");
 540                 ASSERTCMP(sb->s_flags, &, SB_ACTIVE);
 541                 afs_destroy_sbi(as);
 542                 as = NULL;
 543         }
 544
 545         afs_put_cell(params.net, params.cell);
 546         key_put(params.key);
 547         _leave(" = 0 [%p]", sb);
 548         return dget(sb->s_root);
 549
 550 error_sb:
 551         deactivate_locked_super(sb);
 552         goto error_key;
 553 error_as:
 554         afs_destroy_sbi(as);
 555 error_key:
 556         key_put(params.key);
 557 error:
 558         afs_put_cell(params.net, params.cell);
 559         _leave(" = %d", ret);
 560         return ERR_PTR(ret);
 561 }
 562
 563 static void afs_kill_super(struct super_block *sb)
 564 {
 565         struct afs_super_info *as = AFS_FS_S(sb);
 566
 567         /* Clear the callback interests (which will do ilookup5) before
 568          * deactivating the superblock.
 569          */
 570         if (as->volume)
 571                 afs_clear_callback_interests(as->net, as->volume->servers);
 572         kill_anon_super(sb);
 573         if (as->volume)
 574                 afs_deactivate_volume(as->volume);
 575         afs_destroy_sbi(as);
 576 }
 577
 578 /*
 579  * Initialise an inode cache slab element prior to any use.  Note that
 580  * afs_alloc_inode() *must* reset anything that could incorrectly leak from one
 581  * inode to another.
 582  */
 583 static void afs_i_init_once(void *_vnode)
 584 {
 585         struct afs_vnode *vnode = _vnode;
 586
 587         memset(vnode, 0, sizeof(*vnode));
 588         inode_init_once(&vnode->vfs_inode);
 589         mutex_init(&vnode->io_lock);
 590         mutex_init(&vnode->validate_lock);
 591         spin_lock_init(&vnode->wb_lock);
 592         spin_lock_init(&vnode->lock);
 593         INIT_LIST_HEAD(&vnode->wb_keys);
 594         INIT_LIST_HEAD(&vnode->pending_locks);
 595         INIT_LIST_HEAD(&vnode->granted_locks);
 596         INIT_DELAYED_WORK(&vnode->lock_work, afs_lock_work);
 597         seqlock_init(&vnode->cb_lock);
 598 }
 599
 600 /*
 601  * allocate an AFS inode struct from our slab cache
 602  */
 603 static struct inode *afs_alloc_inode(struct super_block *sb)
 604 {
 605         struct afs_vnode *vnode;
 606
 607         vnode = kmem_cache_alloc(afs_inode_cachep, GFP_KERNEL);
 608         if (!vnode)
 609                 return NULL;
 610
 611         atomic_inc(&afs_count_active_inodes);
 612
 613         /* Reset anything that shouldn't leak from one inode to the next. */
 614         memset(&vnode->fid, 0, sizeof(vnode->fid));
 615         memset(&vnode->status, 0, sizeof(vnode->status));
 616
 617         vnode->volume           = NULL;
 618         vnode->lock_key         = NULL;
 619         vnode->permit_cache     = NULL;
 620         vnode->cb_interest      = NULL;
 621 #ifdef CONFIG_AFS_FSCACHE
 622         vnode->cache            = NULL;
 623 #endif
 624
 625         vnode->flags            = 1 << AFS_VNODE_UNSET;
 626         vnode->cb_type          = 0;
 627         vnode->lock_state       = AFS_VNODE_LOCK_NONE;
 628
 629         _leave(" = %p", &vnode->vfs_inode);
 630         return &vnode->vfs_inode;
 631 }
 632
 633 static void afs_i_callback(struct rcu_head *head)
 634 {
 635         struct inode *inode = container_of(head, struct inode, i_rcu);
 636         struct afs_vnode *vnode = AFS_FS_I(inode);
 637         kmem_cache_free(afs_inode_cachep, vnode);
 638 }
 639
 640 /*
 641  * destroy an AFS inode struct
 642  */
 643 static void afs_destroy_inode(struct inode *inode)
 644 {
 645         struct afs_vnode *vnode = AFS_FS_I(inode);
 646
 647         _enter("%p{%x:%u}", inode, vnode->fid.vid, vnode->fid.vnode);
 648
 649         _debug("DESTROY INODE %p", inode);
 650
 651         ASSERTCMP(vnode->cb_interest, ==, NULL);
 652
 653         call_rcu(&inode->i_rcu, afs_i_callback);
 654         atomic_dec(&afs_count_active_inodes);
 655 }
 656
 657 /*
 658  * return information about an AFS volume
 659  */
 660 static int afs_statfs(struct dentry *dentry, struct kstatfs *buf)
 661 {
 662         struct afs_super_info *as = AFS_FS_S(dentry->d_sb);
 663         struct afs_fs_cursor fc;
 664         struct afs_volume_status vs;
 665         struct afs_vnode *vnode = AFS_FS_I(d_inode(dentry));
 666         struct key *key;
 667         int ret;
 668
 669         buf->f_type     = dentry->d_sb->s_magic;
 670         buf->f_bsize    = AFS_BLOCK_SIZE;
 671         buf->f_namelen  = AFSNAMEMAX - 1;
 672
 673         if (as->dyn_root) {
 674                 buf->f_blocks   = 1;
 675                 buf->f_bavail   = 0;
 676                 buf->f_bfree    = 0;
 677                 return 0;
 678         }
 679
 680         key = afs_request_key(vnode->volume->cell);
 681         if (IS_ERR(key))
 682                 return PTR_ERR(key);
 683
 684         ret = -ERESTARTSYS;
 685         if (afs_begin_vnode_operation(&fc, vnode, key)) {
 686                 fc.flags |= AFS_FS_CURSOR_NO_VSLEEP;
 687                 while (afs_select_fileserver(&fc)) {
 688                         fc.cb_break = vnode->cb_break + vnode->cb_s_break;
 689                         afs_fs_get_volume_status(&fc, &vs);
 690                 }
 691
 692                 afs_check_for_remote_deletion(&fc, fc.vnode);
 693                 afs_vnode_commit_status(&fc, vnode, fc.cb_break);
 694                 ret = afs_end_vnode_operation(&fc);
 695         }
 696
 697         key_put(key);
 698
 699         if (ret == 0) {
 700                 if (vs.max_quota == 0)
 701                         buf->f_blocks = vs.part_max_blocks;
 702                 else
 703                         buf->f_blocks = vs.max_quota;
 704                 buf->f_bavail = buf->f_bfree = buf->f_blocks - vs.blocks_in_use;
 705         }
 706
 707         return ret;
 708 }