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