fs/super.c

   1 /*
   2  *  linux/fs/super.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  *
   6  *  super.c contains code to handle: - mount structures
   7  *                                   - super-block tables
   8  *                                   - filesystem drivers list
   9  *                                   - mount system call
  10  *                                   - umount system call
  11  *                                   - ustat system call
  12  *
  13  * GK 2/5/95  -  Changed to support mounting the root fs via NFS
  14  *
  15  *  Added kerneld support: Jacques Gelinas and Bjorn Ekwall
  16  *  Added change_root: Werner Almesberger & Hans Lermen, Feb '96
  17  *  Added options to /proc/mounts:
  18  *    Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996.
  19  *  Added devfs support: Richard Gooch <rgooch@atnf.csiro.au>, 13-JAN-1998
  20  *  Heavily rewritten for 'one fs - one tree' dcache architecture. AV, Mar 2000
  21  */
  22
  23 #include <linux/module.h>
  24 #include <linux/slab.h>
  25 #include <linux/init.h>
  26 #include <linux/smp_lock.h>
  27 #include <linux/acct.h>
  28 #include <linux/blkdev.h>
  29 #include <linux/quotaops.h>
  30 #include <linux/namei.h>
  31 #include <linux/buffer_head.h>          /* for fsync_super() */
  32 #include <linux/mount.h>
  33 #include <linux/security.h>
  34 #include <linux/syscalls.h>
  35 #include <linux/vfs.h>
  36 #include <linux/writeback.h>            /* for the emergency remount stuff */
  37 #include <linux/idr.h>
  38 #include <linux/kobject.h>
  39 #include <linux/mutex.h>
  40 #include <asm/uaccess.h>
  41
  42
  43 void get_filesystem(struct file_system_type *fs);
  44 void put_filesystem(struct file_system_type *fs);
  45 struct file_system_type *get_fs_type(const char *name);
  46
  47 LIST_HEAD(super_blocks);
  48 DEFINE_SPINLOCK(sb_lock);
  49
  50 /**
  51  *      alloc_super     -       create new superblock
  52  *      @type:  filesystem type superblock should belong to
  53  *
  54  *      Allocates and initializes a new &struct super_block.  alloc_super()
  55  *      returns a pointer new superblock or %NULL if allocation had failed.
  56  */
  57 static struct super_block *alloc_super(struct file_system_type *type)
  58 {
  59         struct super_block *s = kzalloc(sizeof(struct super_block),  GFP_USER);
  60         static struct super_operations default_op;
  61
  62         if (s) {
  63                 if (security_sb_alloc(s)) {
  64                         kfree(s);
  65                         s = NULL;
  66                         goto out;
  67                 }
  68                 INIT_LIST_HEAD(&s->s_dirty);
  69                 INIT_LIST_HEAD(&s->s_io);
  70                 INIT_LIST_HEAD(&s->s_files);
  71                 INIT_LIST_HEAD(&s->s_instances);
  72                 INIT_HLIST_HEAD(&s->s_anon);
  73                 INIT_LIST_HEAD(&s->s_inodes);
  74                 init_rwsem(&s->s_umount);
  75                 mutex_init(&s->s_lock);
  76                 lockdep_set_class(&s->s_umount, &type->s_umount_key);
  77                 /*
  78                  * The locking rules for s_lock are up to the
  79                  * filesystem. For example ext3fs has different
  80                  * lock ordering than usbfs:
  81                  */
  82                 lockdep_set_class(&s->s_lock, &type->s_lock_key);
  83                 down_write(&s->s_umount);
  84                 s->s_count = S_BIAS;
  85                 atomic_set(&s->s_active, 1);
  86                 mutex_init(&s->s_vfs_rename_mutex);
  87                 mutex_init(&s->s_dquot.dqio_mutex);
  88                 mutex_init(&s->s_dquot.dqonoff_mutex);
  89                 init_rwsem(&s->s_dquot.dqptr_sem);
  90                 init_waitqueue_head(&s->s_wait_unfrozen);
  91                 s->s_maxbytes = MAX_NON_LFS;
  92                 s->dq_op = sb_dquot_ops;
  93                 s->s_qcop = sb_quotactl_ops;
  94                 s->s_op = &default_op;
  95                 s->s_time_gran = 1000000000;
  96         }
  97 out:
  98         return s;
  99 }
 100
 101 /**
 102  *      destroy_super   -       frees a superblock
 103  *      @s: superblock to free
 104  *
 105  *      Frees a superblock.
 106  */
 107 static inline void destroy_super(struct super_block *s)
 108 {
 109         security_sb_free(s);
 110         kfree(s);
 111 }
 112
 113 /* Superblock refcounting  */
 114
 115 /*
 116  * Drop a superblock's refcount.  Returns non-zero if the superblock was
 117  * destroyed.  The caller must hold sb_lock.
 118  */
 119 int __put_super(struct super_block *sb)
 120 {
 121         int ret = 0;
 122
 123         if (!--sb->s_count) {
 124                 destroy_super(sb);
 125                 ret = 1;
 126         }
 127         return ret;
 128 }
 129
 130 /*
 131  * Drop a superblock's refcount.
 132  * Returns non-zero if the superblock is about to be destroyed and
 133  * at least is already removed from super_blocks list, so if we are
 134  * making a loop through super blocks then we need to restart.
 135  * The caller must hold sb_lock.
 136  */
 137 int __put_super_and_need_restart(struct super_block *sb)
 138 {
 139         /* check for race with generic_shutdown_super() */
 140         if (list_empty(&sb->s_list)) {
 141                 /* super block is removed, need to restart... */
 142                 __put_super(sb);
 143                 return 1;
 144         }
 145         /* can't be the last, since s_list is still in use */
 146         sb->s_count--;
 147         BUG_ON(sb->s_count == 0);
 148         return 0;
 149 }
 150
 151 /**
 152  *      put_super       -       drop a temporary reference to superblock
 153  *      @sb: superblock in question
 154  *
 155  *      Drops a temporary reference, frees superblock if there's no
 156  *      references left.
 157  */
 158 static void put_super(struct super_block *sb)
 159 {
 160         spin_lock(&sb_lock);
 161         __put_super(sb);
 162         spin_unlock(&sb_lock);
 163 }
 164
 165
 166 /**
 167  *      deactivate_super        -       drop an active reference to superblock
 168  *      @s: superblock to deactivate
 169  *
 170  *      Drops an active reference to superblock, acquiring a temprory one if
 171  *      there is no active references left.  In that case we lock superblock,
 172  *      tell fs driver to shut it down and drop the temporary reference we
 173  *      had just acquired.
 174  */
 175 void deactivate_super(struct super_block *s)
 176 {
 177         struct file_system_type *fs = s->s_type;
 178         if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
 179                 s->s_count -= S_BIAS-1;
 180                 spin_unlock(&sb_lock);
 181                 DQUOT_OFF(s);
 182                 down_write(&s->s_umount);
 183                 fs->kill_sb(s);
 184                 put_filesystem(fs);
 185                 put_super(s);
 186         }
 187 }
 188
 189 EXPORT_SYMBOL(deactivate_super);
 190
 191 /**
 192  *      grab_super - acquire an active reference
 193  *      @s: reference we are trying to make active
 194  *
 195  *      Tries to acquire an active reference.  grab_super() is used when we
 196  *      had just found a superblock in super_blocks or fs_type->fs_supers
 197  *      and want to turn it into a full-blown active reference.  grab_super()
 198  *      is called with sb_lock held and drops it.  Returns 1 in case of
 199  *      success, 0 if we had failed (superblock contents was already dead or
 200  *      dying when grab_super() had been called).
 201  */
 202 static int grab_super(struct super_block *s) __releases(sb_lock)
 203 {
 204         s->s_count++;
 205         spin_unlock(&sb_lock);
 206         down_write(&s->s_umount);
 207         if (s->s_root) {
 208                 spin_lock(&sb_lock);
 209                 if (s->s_count > S_BIAS) {
 210                         atomic_inc(&s->s_active);
 211                         s->s_count--;
 212                         spin_unlock(&sb_lock);
 213                         return 1;
 214                 }
 215                 spin_unlock(&sb_lock);
 216         }
 217         up_write(&s->s_umount);
 218         put_super(s);
 219         yield();
 220         return 0;
 221 }
 222
 223 /*
 224  * Write out and wait upon all dirty data associated with this
 225  * superblock.  Filesystem data as well as the underlying block
 226  * device.  Takes the superblock lock.  Requires a second blkdev
 227  * flush by the caller to complete the operation.
 228  */
 229 void __fsync_super(struct super_block *sb)
 230 {
 231         sync_inodes_sb(sb, 0);
 232         DQUOT_SYNC(sb);
 233         lock_super(sb);
 234         if (sb->s_dirt && sb->s_op->write_super)
 235                 sb->s_op->write_super(sb);
 236         unlock_super(sb);
 237         if (sb->s_op->sync_fs)
 238                 sb->s_op->sync_fs(sb, 1);
 239         sync_blockdev(sb->s_bdev);
 240         sync_inodes_sb(sb, 1);
 241 }
 242
 243 /*
 244  * Write out and wait upon all dirty data associated with this
 245  * superblock.  Filesystem data as well as the underlying block
 246  * device.  Takes the superblock lock.
 247  */
 248 int fsync_super(struct super_block *sb)
 249 {
 250         __fsync_super(sb);
 251         return sync_blockdev(sb->s_bdev);
 252 }
 253
 254 /**
 255  *      generic_shutdown_super  -       common helper for ->kill_sb()
 256  *      @sb: superblock to kill
 257  *
 258  *      generic_shutdown_super() does all fs-independent work on superblock
 259  *      shutdown.  Typical ->kill_sb() should pick all fs-specific objects
 260  *      that need destruction out of superblock, call generic_shutdown_super()
 261  *      and release aforementioned objects.  Note: dentries and inodes _are_
 262  *      taken care of and do not need specific handling.
 263  */
 264 void generic_shutdown_super(struct super_block *sb)
 265 {
 266         struct dentry *root = sb->s_root;
 267         struct super_operations *sop = sb->s_op;
 268
 269         if (root) {
 270                 sb->s_root = NULL;
 271                 shrink_dcache_parent(root);
 272                 shrink_dcache_sb(sb);
 273                 dput(root);
 274                 fsync_super(sb);
 275                 lock_super(sb);
 276                 sb->s_flags &= ~MS_ACTIVE;
 277                 /* bad name - it should be evict_inodes() */
 278                 invalidate_inodes(sb);
 279                 lock_kernel();
 280
 281                 if (sop->write_super && sb->s_dirt)
 282                         sop->write_super(sb);
 283                 if (sop->put_super)
 284                         sop->put_super(sb);
 285
 286                 /* Forget any remaining inodes */
 287                 if (invalidate_inodes(sb)) {
 288                         printk("VFS: Busy inodes after unmount of %s. "
 289                            "Self-destruct in 5 seconds.  Have a nice day...\n",
 290                            sb->s_id);
 291                 }
 292
 293                 unlock_kernel();
 294                 unlock_super(sb);
 295         }
 296         spin_lock(&sb_lock);
 297         /* should be initialized for __put_super_and_need_restart() */
 298         list_del_init(&sb->s_list);
 299         list_del(&sb->s_instances);
 300         spin_unlock(&sb_lock);
 301         up_write(&sb->s_umount);
 302 }
 303
 304 EXPORT_SYMBOL(generic_shutdown_super);
 305
 306 /**
 307  *      sget    -       find or create a superblock
 308  *      @type:  filesystem type superblock should belong to
 309  *      @test:  comparison callback
 310  *      @set:   setup callback
 311  *      @data:  argument to each of them
 312  */
 313 struct super_block *sget(struct file_system_type *type,
 314                         int (*test)(struct super_block *,void *),
 315                         int (*set)(struct super_block *,void *),
 316                         void *data)
 317 {
 318         struct super_block *s = NULL;
 319         struct list_head *p;
 320         int err;
 321
 322 retry:
 323         spin_lock(&sb_lock);
 324         if (test) list_for_each(p, &type->fs_supers) {
 325                 struct super_block *old;
 326                 old = list_entry(p, struct super_block, s_instances);
 327                 if (!test(old, data))
 328                         continue;
 329                 if (!grab_super(old))
 330                         goto retry;
 331                 if (s)
 332                         destroy_super(s);
 333                 return old;
 334         }
 335         if (!s) {
 336                 spin_unlock(&sb_lock);
 337                 s = alloc_super(type);
 338                 if (!s)
 339                         return ERR_PTR(-ENOMEM);
 340                 goto retry;
 341         }
 342
 343         err = set(s, data);
 344         if (err) {
 345                 spin_unlock(&sb_lock);
 346                 destroy_super(s);
 347                 return ERR_PTR(err);
 348         }
 349         s->s_type = type;
 350         strlcpy(s->s_id, type->name, sizeof(s->s_id));
 351         list_add_tail(&s->s_list, &super_blocks);
 352         list_add(&s->s_instances, &type->fs_supers);
 353         spin_unlock(&sb_lock);
 354         get_filesystem(type);
 355         return s;
 356 }
 357
 358 EXPORT_SYMBOL(sget);
 359
 360 void drop_super(struct super_block *sb)
 361 {
 362         up_read(&sb->s_umount);
 363         put_super(sb);
 364 }
 365
 366 EXPORT_SYMBOL(drop_super);
 367
 368 static inline void write_super(struct super_block *sb)
 369 {
 370         lock_super(sb);
 371         if (sb->s_root && sb->s_dirt)
 372                 if (sb->s_op->write_super)
 373                         sb->s_op->write_super(sb);
 374         unlock_super(sb);
 375 }
 376
 377 /*
 378  * Note: check the dirty flag before waiting, so we don't
 379  * hold up the sync while mounting a device. (The newly
 380  * mounted device won't need syncing.)
 381  */
 382 void sync_supers(void)
 383 {
 384         struct super_block *sb;
 385
 386         spin_lock(&sb_lock);
 387 restart:
 388         list_for_each_entry(sb, &super_blocks, s_list) {
 389                 if (sb->s_dirt) {
 390                         sb->s_count++;
 391                         spin_unlock(&sb_lock);
 392                         down_read(&sb->s_umount);
 393                         write_super(sb);
 394                         up_read(&sb->s_umount);
 395                         spin_lock(&sb_lock);
 396                         if (__put_super_and_need_restart(sb))
 397                                 goto restart;
 398                 }
 399         }
 400         spin_unlock(&sb_lock);
 401 }
 402
 403 /*
 404  * Call the ->sync_fs super_op against all filesytems which are r/w and
 405  * which implement it.
 406  *
 407  * This operation is careful to avoid the livelock which could easily happen
 408  * if two or more filesystems are being continuously dirtied.  s_need_sync_fs
 409  * is used only here.  We set it against all filesystems and then clear it as
 410  * we sync them.  So redirtied filesystems are skipped.
 411  *
 412  * But if process A is currently running sync_filesytems and then process B
 413  * calls sync_filesystems as well, process B will set all the s_need_sync_fs
 414  * flags again, which will cause process A to resync everything.  Fix that with
 415  * a local mutex.
 416  *
 417  * (Fabian) Avoid sync_fs with clean fs & wait mode 0
 418  */
 419 void sync_filesystems(int wait)
 420 {
 421         struct super_block *sb;
 422         static DEFINE_MUTEX(mutex);
 423
 424         mutex_lock(&mutex);             /* Could be down_interruptible */
 425         spin_lock(&sb_lock);
 426         list_for_each_entry(sb, &super_blocks, s_list) {
 427                 if (!sb->s_op->sync_fs)
 428                         continue;
 429                 if (sb->s_flags & MS_RDONLY)
 430                         continue;
 431                 sb->s_need_sync_fs = 1;
 432         }
 433
 434 restart:
 435         list_for_each_entry(sb, &super_blocks, s_list) {
 436                 if (!sb->s_need_sync_fs)
 437                         continue;
 438                 sb->s_need_sync_fs = 0;
 439                 if (sb->s_flags & MS_RDONLY)
 440                         continue;       /* hm.  Was remounted r/o meanwhile */
 441                 sb->s_count++;
 442                 spin_unlock(&sb_lock);
 443                 down_read(&sb->s_umount);
 444                 if (sb->s_root && (wait || sb->s_dirt))
 445                         sb->s_op->sync_fs(sb, wait);
 446                 up_read(&sb->s_umount);
 447                 /* restart only when sb is no longer on the list */
 448                 spin_lock(&sb_lock);
 449                 if (__put_super_and_need_restart(sb))
 450                         goto restart;
 451         }
 452         spin_unlock(&sb_lock);
 453         mutex_unlock(&mutex);
 454 }
 455
 456 /**
 457  *      get_super - get the superblock of a device
 458  *      @bdev: device to get the superblock for
 459  *
 460  *      Scans the superblock list and finds the superblock of the file system
 461  *      mounted on the device given. %NULL is returned if no match is found.
 462  */
 463
 464 struct super_block * get_super(struct block_device *bdev)
 465 {
 466         struct super_block *sb;
 467
 468         if (!bdev)
 469                 return NULL;
 470
 471         spin_lock(&sb_lock);
 472 rescan:
 473         list_for_each_entry(sb, &super_blocks, s_list) {
 474                 if (sb->s_bdev == bdev) {
 475                         sb->s_count++;
 476                         spin_unlock(&sb_lock);
 477                         down_read(&sb->s_umount);
 478                         if (sb->s_root)
 479                                 return sb;
 480                         up_read(&sb->s_umount);
 481                         /* restart only when sb is no longer on the list */
 482                         spin_lock(&sb_lock);
 483                         if (__put_super_and_need_restart(sb))
 484                                 goto rescan;
 485                 }
 486         }
 487         spin_unlock(&sb_lock);
 488         return NULL;
 489 }
 490
 491 EXPORT_SYMBOL(get_super);
 492
 493 struct super_block * user_get_super(dev_t dev)
 494 {
 495         struct super_block *sb;
 496
 497         spin_lock(&sb_lock);
 498 rescan:
 499         list_for_each_entry(sb, &super_blocks, s_list) {
 500                 if (sb->s_dev ==  dev) {
 501                         sb->s_count++;
 502                         spin_unlock(&sb_lock);
 503                         down_read(&sb->s_umount);
 504                         if (sb->s_root)
 505                                 return sb;
 506                         up_read(&sb->s_umount);
 507                         /* restart only when sb is no longer on the list */
 508                         spin_lock(&sb_lock);
 509                         if (__put_super_and_need_restart(sb))
 510                                 goto rescan;
 511                 }
 512         }
 513         spin_unlock(&sb_lock);
 514         return NULL;
 515 }
 516
 517 asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf)
 518 {
 519         struct super_block *s;
 520         struct ustat tmp;
 521         struct kstatfs sbuf;
 522         int err = -EINVAL;
 523
 524         s = user_get_super(new_decode_dev(dev));
 525         if (s == NULL)
 526                 goto out;
 527         err = vfs_statfs(s->s_root, &sbuf);
 528         drop_super(s);
 529         if (err)
 530                 goto out;
 531
 532         memset(&tmp,0,sizeof(struct ustat));
 533         tmp.f_tfree = sbuf.f_bfree;
 534         tmp.f_tinode = sbuf.f_ffree;
 535
 536         err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
 537 out:
 538         return err;
 539 }
 540
 541 /**
 542  *      mark_files_ro
 543  *      @sb: superblock in question
 544  *
 545  *      All files are marked read/only.  We don't care about pending
 546  *      delete files so this should be used in 'force' mode only
 547  */
 548
 549 static void mark_files_ro(struct super_block *sb)
 550 {
 551         struct file *f;
 552
 553         file_list_lock();
 554         list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
 555                 if (S_ISREG(f->f_dentry->d_inode->i_mode) && file_count(f))
 556                         f->f_mode &= ~FMODE_WRITE;
 557         }
 558         file_list_unlock();
 559 }
 560
 561 /**
 562  *      do_remount_sb - asks filesystem to change mount options.
 563  *      @sb:    superblock in question
 564  *      @flags: numeric part of options
 565  *      @data:  the rest of options
 566  *      @force: whether or not to force the change
 567  *
 568  *      Alters the mount options of a mounted file system.
 569  */
 570 int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
 571 {
 572         int retval;
 573
 574         if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
 575                 return -EACCES;
 576         if (flags & MS_RDONLY)
 577                 acct_auto_close(sb);
 578         shrink_dcache_sb(sb);
 579         fsync_super(sb);
 580
 581         /* If we are remounting RDONLY and current sb is read/write,
 582            make sure there are no rw files opened */
 583         if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
 584                 if (force)
 585                         mark_files_ro(sb);
 586                 else if (!fs_may_remount_ro(sb))
 587                         return -EBUSY;
 588         }
 589
 590         if (sb->s_op->remount_fs) {
 591                 lock_super(sb);
 592                 retval = sb->s_op->remount_fs(sb, &flags, data);
 593                 unlock_super(sb);
 594                 if (retval)
 595                         return retval;
 596         }
 597         sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
 598         return 0;
 599 }
 600
 601 static void do_emergency_remount(unsigned long foo)
 602 {
 603         struct super_block *sb;
 604
 605         spin_lock(&sb_lock);
 606         list_for_each_entry(sb, &super_blocks, s_list) {
 607                 sb->s_count++;
 608                 spin_unlock(&sb_lock);
 609                 down_read(&sb->s_umount);
 610                 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
 611                         /*
 612                          * ->remount_fs needs lock_kernel().
 613                          *
 614                          * What lock protects sb->s_flags??
 615                          */
 616                         lock_kernel();
 617                         do_remount_sb(sb, MS_RDONLY, NULL, 1);
 618                         unlock_kernel();
 619                 }
 620                 drop_super(sb);
 621                 spin_lock(&sb_lock);
 622         }
 623         spin_unlock(&sb_lock);
 624         printk("Emergency Remount complete\n");
 625 }
 626
 627 void emergency_remount(void)
 628 {
 629         pdflush_operation(do_emergency_remount, 0);
 630 }
 631
 632 /*
 633  * Unnamed block devices are dummy devices used by virtual
 634  * filesystems which don't use real block-devices.  -- jrs
 635  */
 636
 637 static struct idr unnamed_dev_idr;
 638 static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
 639
 640 int set_anon_super(struct super_block *s, void *data)
 641 {
 642         int dev;
 643         int error;
 644
 645  retry:
 646         if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0)
 647                 return -ENOMEM;
 648         spin_lock(&unnamed_dev_lock);
 649         error = idr_get_new(&unnamed_dev_idr, NULL, &dev);
 650         spin_unlock(&unnamed_dev_lock);
 651         if (error == -EAGAIN)
 652                 /* We raced and lost with another CPU. */
 653                 goto retry;
 654         else if (error)
 655                 return -EAGAIN;
 656
 657         if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
 658                 spin_lock(&unnamed_dev_lock);
 659                 idr_remove(&unnamed_dev_idr, dev);
 660                 spin_unlock(&unnamed_dev_lock);
 661                 return -EMFILE;
 662         }
 663         s->s_dev = MKDEV(0, dev & MINORMASK);
 664         return 0;
 665 }
 666
 667 EXPORT_SYMBOL(set_anon_super);
 668
 669 void kill_anon_super(struct super_block *sb)
 670 {
 671         int slot = MINOR(sb->s_dev);
 672
 673         generic_shutdown_super(sb);
 674         spin_lock(&unnamed_dev_lock);
 675         idr_remove(&unnamed_dev_idr, slot);
 676         spin_unlock(&unnamed_dev_lock);
 677 }
 678
 679 EXPORT_SYMBOL(kill_anon_super);
 680
 681 void __init unnamed_dev_init(void)
 682 {
 683         idr_init(&unnamed_dev_idr);
 684 }
 685
 686 void kill_litter_super(struct super_block *sb)
 687 {
 688         if (sb->s_root)
 689                 d_genocide(sb->s_root);
 690         kill_anon_super(sb);
 691 }
 692
 693 EXPORT_SYMBOL(kill_litter_super);
 694
 695 static int set_bdev_super(struct super_block *s, void *data)
 696 {
 697         s->s_bdev = data;
 698         s->s_dev = s->s_bdev->bd_dev;
 699         return 0;
 700 }
 701
 702 static int test_bdev_super(struct super_block *s, void *data)
 703 {
 704         return (void *)s->s_bdev == data;
 705 }
 706
 707 static void bdev_uevent(struct block_device *bdev, enum kobject_action action)
 708 {
 709         if (bdev->bd_disk) {
 710                 if (bdev->bd_part)
 711                         kobject_uevent(&bdev->bd_part->kobj, action);
 712                 else
 713                         kobject_uevent(&bdev->bd_disk->kobj, action);
 714         }
 715 }
 716
 717 int get_sb_bdev(struct file_system_type *fs_type,
 718         int flags, const char *dev_name, void *data,
 719         int (*fill_super)(struct super_block *, void *, int),
 720         struct vfsmount *mnt)
 721 {
 722         struct block_device *bdev;
 723         struct super_block *s;
 724         int error = 0;
 725
 726         bdev = open_bdev_excl(dev_name, flags, fs_type);
 727         if (IS_ERR(bdev))
 728                 return PTR_ERR(bdev);
 729
 730         /*
 731          * once the super is inserted into the list by sget, s_umount
 732          * will protect the lockfs code from trying to start a snapshot
 733          * while we are mounting
 734          */
 735         mutex_lock(&bdev->bd_mount_mutex);
 736         s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
 737         mutex_unlock(&bdev->bd_mount_mutex);
 738         if (IS_ERR(s))
 739                 goto error_s;
 740
 741         if (s->s_root) {
 742                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 743                         up_write(&s->s_umount);
 744                         deactivate_super(s);
 745                         error = -EBUSY;
 746                         goto error_bdev;
 747                 }
 748
 749                 close_bdev_excl(bdev);
 750         } else {
 751                 char b[BDEVNAME_SIZE];
 752
 753                 s->s_flags = flags;
 754                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 755                 sb_set_blocksize(s, block_size(bdev));
 756                 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
 757                 if (error) {
 758                         up_write(&s->s_umount);
 759                         deactivate_super(s);
 760                         goto error;
 761                 }
 762
 763                 s->s_flags |= MS_ACTIVE;
 764                 bdev_uevent(bdev, KOBJ_MOUNT);
 765         }
 766
 767         return simple_set_mnt(mnt, s);
 768
 769 error_s:
 770         error = PTR_ERR(s);
 771 error_bdev:
 772         close_bdev_excl(bdev);
 773 error:
 774         return error;
 775 }
 776
 777 EXPORT_SYMBOL(get_sb_bdev);
 778
 779 void kill_block_super(struct super_block *sb)
 780 {
 781         struct block_device *bdev = sb->s_bdev;
 782
 783         bdev_uevent(bdev, KOBJ_UMOUNT);
 784         generic_shutdown_super(sb);
 785         sync_blockdev(bdev);
 786         close_bdev_excl(bdev);
 787 }
 788
 789 EXPORT_SYMBOL(kill_block_super);
 790
 791 int get_sb_nodev(struct file_system_type *fs_type,
 792         int flags, void *data,
 793         int (*fill_super)(struct super_block *, void *, int),
 794         struct vfsmount *mnt)
 795 {
 796         int error;
 797         struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
 798
 799         if (IS_ERR(s))
 800                 return PTR_ERR(s);
 801
 802         s->s_flags = flags;
 803
 804         error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
 805         if (error) {
 806                 up_write(&s->s_umount);
 807                 deactivate_super(s);
 808                 return error;
 809         }
 810         s->s_flags |= MS_ACTIVE;
 811         return simple_set_mnt(mnt, s);
 812 }
 813
 814 EXPORT_SYMBOL(get_sb_nodev);
 815
 816 static int compare_single(struct super_block *s, void *p)
 817 {
 818         return 1;
 819 }
 820
 821 int get_sb_single(struct file_system_type *fs_type,
 822         int flags, void *data,
 823         int (*fill_super)(struct super_block *, void *, int),
 824         struct vfsmount *mnt)
 825 {
 826         struct super_block *s;
 827         int error;
 828
 829         s = sget(fs_type, compare_single, set_anon_super, NULL);
 830         if (IS_ERR(s))
 831                 return PTR_ERR(s);
 832         if (!s->s_root) {
 833                 s->s_flags = flags;
 834                 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
 835                 if (error) {
 836                         up_write(&s->s_umount);
 837                         deactivate_super(s);
 838                         return error;
 839                 }
 840                 s->s_flags |= MS_ACTIVE;
 841         }
 842         do_remount_sb(s, flags, data, 0);
 843         return simple_set_mnt(mnt, s);
 844 }
 845
 846 EXPORT_SYMBOL(get_sb_single);
 847
 848 struct vfsmount *
 849 vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
 850 {
 851         struct vfsmount *mnt;
 852         char *secdata = NULL;
 853         int error;
 854
 855         if (!type)
 856                 return ERR_PTR(-ENODEV);
 857
 858         error = -ENOMEM;
 859         mnt = alloc_vfsmnt(name);
 860         if (!mnt)
 861                 goto out;
 862
 863         if (data) {
 864                 secdata = alloc_secdata();
 865                 if (!secdata)
 866                         goto out_mnt;
 867
 868                 error = security_sb_copy_data(type, data, secdata);
 869                 if (error)
 870                         goto out_free_secdata;
 871         }
 872
 873         error = type->get_sb(type, flags, name, data, mnt);
 874         if (error < 0)
 875                 goto out_free_secdata;
 876
 877         error = security_sb_kern_mount(mnt->mnt_sb, secdata);
 878         if (error)
 879                 goto out_sb;
 880
 881         mnt->mnt_mountpoint = mnt->mnt_root;
 882         mnt->mnt_parent = mnt;
 883         up_write(&mnt->mnt_sb->s_umount);
 884         free_secdata(secdata);
 885         return mnt;
 886 out_sb:
 887         dput(mnt->mnt_root);
 888         up_write(&mnt->mnt_sb->s_umount);
 889         deactivate_super(mnt->mnt_sb);
 890 out_free_secdata:
 891         free_secdata(secdata);
 892 out_mnt:
 893         free_vfsmnt(mnt);
 894 out:
 895         return ERR_PTR(error);
 896 }
 897
 898 EXPORT_SYMBOL_GPL(vfs_kern_mount);
 899
 900 struct vfsmount *
 901 do_kern_mount(const char *fstype, int flags, const char *name, void *data)
 902 {
 903         struct file_system_type *type = get_fs_type(fstype);
 904         struct vfsmount *mnt;
 905         if (!type)
 906                 return ERR_PTR(-ENODEV);
 907         mnt = vfs_kern_mount(type, flags, name, data);
 908         put_filesystem(type);
 909         return mnt;
 910 }
 911
 912 struct vfsmount *kern_mount(struct file_system_type *type)
 913 {
 914         return vfs_kern_mount(type, 0, type->name, NULL);
 915 }
 916
 917 EXPORT_SYMBOL(kern_mount);