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/mount.h>
  32 #include <linux/security.h>
  33 #include <linux/syscalls.h>
  34 #include <linux/vfs.h>
  35 #include <linux/writeback.h>            /* for the emergency remount stuff */
  36 #include <linux/idr.h>
  37 #include <linux/kobject.h>
  38 #include <linux/mutex.h>
  39 #include <linux/file.h>
  40 #include <asm/uaccess.h>
  41 #include "internal.h"
  42
  43
  44 LIST_HEAD(super_blocks);
  45 DEFINE_SPINLOCK(sb_lock);
  46
  47 /**
  48  *      alloc_super     -       create new superblock
  49  *      @type:  filesystem type superblock should belong to
  50  *
  51  *      Allocates and initializes a new &struct super_block.  alloc_super()
  52  *      returns a pointer new superblock or %NULL if allocation had failed.
  53  */
  54 static struct super_block *alloc_super(struct file_system_type *type)
  55 {
  56         struct super_block *s = kzalloc(sizeof(struct super_block),  GFP_USER);
  57         static struct super_operations default_op;
  58
  59         if (s) {
  60                 if (security_sb_alloc(s)) {
  61                         kfree(s);
  62                         s = NULL;
  63                         goto out;
  64                 }
  65                 INIT_LIST_HEAD(&s->s_dirty);
  66                 INIT_LIST_HEAD(&s->s_io);
  67                 INIT_LIST_HEAD(&s->s_more_io);
  68                 INIT_LIST_HEAD(&s->s_files);
  69                 INIT_LIST_HEAD(&s->s_instances);
  70                 INIT_HLIST_HEAD(&s->s_anon);
  71                 INIT_LIST_HEAD(&s->s_inodes);
  72                 INIT_LIST_HEAD(&s->s_dentry_lru);
  73                 init_rwsem(&s->s_umount);
  74                 mutex_init(&s->s_lock);
  75                 lockdep_set_class(&s->s_umount, &type->s_umount_key);
  76                 /*
  77                  * The locking rules for s_lock are up to the
  78                  * filesystem. For example ext3fs has different
  79                  * lock ordering than usbfs:
  80                  */
  81                 lockdep_set_class(&s->s_lock, &type->s_lock_key);
  82                 /*
  83                  * sget() can have s_umount recursion.
  84                  *
  85                  * When it cannot find a suitable sb, it allocates a new
  86                  * one (this one), and tries again to find a suitable old
  87                  * one.
  88                  *
  89                  * In case that succeeds, it will acquire the s_umount
  90                  * lock of the old one. Since these are clearly distrinct
  91                  * locks, and this object isn't exposed yet, there's no
  92                  * risk of deadlocks.
  93                  *
  94                  * Annotate this by putting this lock in a different
  95                  * subclass.
  96                  */
  97                 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
  98                 s->s_count = S_BIAS;
  99                 atomic_set(&s->s_active, 1);
 100                 mutex_init(&s->s_vfs_rename_mutex);
 101                 mutex_init(&s->s_dquot.dqio_mutex);
 102                 mutex_init(&s->s_dquot.dqonoff_mutex);
 103                 init_rwsem(&s->s_dquot.dqptr_sem);
 104                 init_waitqueue_head(&s->s_wait_unfrozen);
 105                 s->s_maxbytes = MAX_NON_LFS;
 106                 s->dq_op = sb_dquot_ops;
 107                 s->s_qcop = sb_quotactl_ops;
 108                 s->s_op = &default_op;
 109                 s->s_time_gran = 1000000000;
 110         }
 111 out:
 112         return s;
 113 }
 114
 115 /**
 116  *      destroy_super   -       frees a superblock
 117  *      @s: superblock to free
 118  *
 119  *      Frees a superblock.
 120  */
 121 static inline void destroy_super(struct super_block *s)
 122 {
 123         security_sb_free(s);
 124         kfree(s->s_subtype);
 125         kfree(s->s_options);
 126         kfree(s);
 127 }
 128
 129 /* Superblock refcounting  */
 130
 131 /*
 132  * Drop a superblock's refcount.  Returns non-zero if the superblock was
 133  * destroyed.  The caller must hold sb_lock.
 134  */
 135 static int __put_super(struct super_block *sb)
 136 {
 137         int ret = 0;
 138
 139         if (!--sb->s_count) {
 140                 destroy_super(sb);
 141                 ret = 1;
 142         }
 143         return ret;
 144 }
 145
 146 /*
 147  * Drop a superblock's refcount.
 148  * Returns non-zero if the superblock is about to be destroyed and
 149  * at least is already removed from super_blocks list, so if we are
 150  * making a loop through super blocks then we need to restart.
 151  * The caller must hold sb_lock.
 152  */
 153 int __put_super_and_need_restart(struct super_block *sb)
 154 {
 155         /* check for race with generic_shutdown_super() */
 156         if (list_empty(&sb->s_list)) {
 157                 /* super block is removed, need to restart... */
 158                 __put_super(sb);
 159                 return 1;
 160         }
 161         /* can't be the last, since s_list is still in use */
 162         sb->s_count--;
 163         BUG_ON(sb->s_count == 0);
 164         return 0;
 165 }
 166
 167 /**
 168  *      put_super       -       drop a temporary reference to superblock
 169  *      @sb: superblock in question
 170  *
 171  *      Drops a temporary reference, frees superblock if there's no
 172  *      references left.
 173  */
 174 static void put_super(struct super_block *sb)
 175 {
 176         spin_lock(&sb_lock);
 177         __put_super(sb);
 178         spin_unlock(&sb_lock);
 179 }
 180
 181
 182 /**
 183  *      deactivate_super        -       drop an active reference to superblock
 184  *      @s: superblock to deactivate
 185  *
 186  *      Drops an active reference to superblock, acquiring a temprory one if
 187  *      there is no active references left.  In that case we lock superblock,
 188  *      tell fs driver to shut it down and drop the temporary reference we
 189  *      had just acquired.
 190  */
 191 void deactivate_super(struct super_block *s)
 192 {
 193         struct file_system_type *fs = s->s_type;
 194         if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
 195                 s->s_count -= S_BIAS-1;
 196                 spin_unlock(&sb_lock);
 197                 vfs_dq_off(s, 0);
 198                 down_write(&s->s_umount);
 199                 fs->kill_sb(s);
 200                 put_filesystem(fs);
 201                 put_super(s);
 202         }
 203 }
 204
 205 EXPORT_SYMBOL(deactivate_super);
 206
 207 /**
 208  *      deactivate_locked_super -       drop an active reference to superblock
 209  *      @s: superblock to deactivate
 210  *
 211  *      Equivalent of up_write(&s->s_umount); deactivate_super(s);, except that
 212  *      it does not unlock it until it's all over.  As the result, it's safe to
 213  *      use to dispose of new superblock on ->get_sb() failure exits - nobody
 214  *      will see the sucker until it's all over.  Equivalent using up_write +
 215  *      deactivate_super is safe for that purpose only if superblock is either
 216  *      safe to use or has NULL ->s_root when we unlock.
 217  */
 218 void deactivate_locked_super(struct super_block *s)
 219 {
 220         struct file_system_type *fs = s->s_type;
 221         if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
 222                 s->s_count -= S_BIAS-1;
 223                 spin_unlock(&sb_lock);
 224                 vfs_dq_off(s, 0);
 225                 fs->kill_sb(s);
 226                 put_filesystem(fs);
 227                 put_super(s);
 228         } else {
 229                 up_write(&s->s_umount);
 230         }
 231 }
 232
 233 EXPORT_SYMBOL(deactivate_locked_super);
 234
 235 /**
 236  *      grab_super - acquire an active reference
 237  *      @s: reference we are trying to make active
 238  *
 239  *      Tries to acquire an active reference.  grab_super() is used when we
 240  *      had just found a superblock in super_blocks or fs_type->fs_supers
 241  *      and want to turn it into a full-blown active reference.  grab_super()
 242  *      is called with sb_lock held and drops it.  Returns 1 in case of
 243  *      success, 0 if we had failed (superblock contents was already dead or
 244  *      dying when grab_super() had been called).
 245  */
 246 static int grab_super(struct super_block *s) __releases(sb_lock)
 247 {
 248         s->s_count++;
 249         spin_unlock(&sb_lock);
 250         down_write(&s->s_umount);
 251         if (s->s_root) {
 252                 spin_lock(&sb_lock);
 253                 if (s->s_count > S_BIAS) {
 254                         atomic_inc(&s->s_active);
 255                         s->s_count--;
 256                         spin_unlock(&sb_lock);
 257                         return 1;
 258                 }
 259                 spin_unlock(&sb_lock);
 260         }
 261         up_write(&s->s_umount);
 262         put_super(s);
 263         yield();
 264         return 0;
 265 }
 266
 267 /*
 268  * Superblock locking.  We really ought to get rid of these two.
 269  */
 270 void lock_super(struct super_block * sb)
 271 {
 272         get_fs_excl();
 273         mutex_lock(&sb->s_lock);
 274 }
 275
 276 void unlock_super(struct super_block * sb)
 277 {
 278         put_fs_excl();
 279         mutex_unlock(&sb->s_lock);
 280 }
 281
 282 EXPORT_SYMBOL(lock_super);
 283 EXPORT_SYMBOL(unlock_super);
 284
 285 /**
 286  *      generic_shutdown_super  -       common helper for ->kill_sb()
 287  *      @sb: superblock to kill
 288  *
 289  *      generic_shutdown_super() does all fs-independent work on superblock
 290  *      shutdown.  Typical ->kill_sb() should pick all fs-specific objects
 291  *      that need destruction out of superblock, call generic_shutdown_super()
 292  *      and release aforementioned objects.  Note: dentries and inodes _are_
 293  *      taken care of and do not need specific handling.
 294  *
 295  *      Upon calling this function, the filesystem may no longer alter or
 296  *      rearrange the set of dentries belonging to this super_block, nor may it
 297  *      change the attachments of dentries to inodes.
 298  */
 299 void generic_shutdown_super(struct super_block *sb)
 300 {
 301         const struct super_operations *sop = sb->s_op;
 302
 303
 304         if (sb->s_root) {
 305                 shrink_dcache_for_umount(sb);
 306                 sync_filesystem(sb);
 307                 get_fs_excl();
 308                 sb->s_flags &= ~MS_ACTIVE;
 309
 310                 /* bad name - it should be evict_inodes() */
 311                 invalidate_inodes(sb);
 312
 313                 if (sop->put_super)
 314                         sop->put_super(sb);
 315
 316                 /* Forget any remaining inodes */
 317                 if (invalidate_inodes(sb)) {
 318                         printk("VFS: Busy inodes after unmount of %s. "
 319                            "Self-destruct in 5 seconds.  Have a nice day...\n",
 320                            sb->s_id);
 321                 }
 322                 put_fs_excl();
 323         }
 324         spin_lock(&sb_lock);
 325         /* should be initialized for __put_super_and_need_restart() */
 326         list_del_init(&sb->s_list);
 327         list_del(&sb->s_instances);
 328         spin_unlock(&sb_lock);
 329         up_write(&sb->s_umount);
 330 }
 331
 332 EXPORT_SYMBOL(generic_shutdown_super);
 333
 334 /**
 335  *      sget    -       find or create a superblock
 336  *      @type:  filesystem type superblock should belong to
 337  *      @test:  comparison callback
 338  *      @set:   setup callback
 339  *      @data:  argument to each of them
 340  */
 341 struct super_block *sget(struct file_system_type *type,
 342                         int (*test)(struct super_block *,void *),
 343                         int (*set)(struct super_block *,void *),
 344                         void *data)
 345 {
 346         struct super_block *s = NULL;
 347         struct super_block *old;
 348         int err;
 349
 350 retry:
 351         spin_lock(&sb_lock);
 352         if (test) {
 353                 list_for_each_entry(old, &type->fs_supers, s_instances) {
 354                         if (!test(old, data))
 355                                 continue;
 356                         if (!grab_super(old))
 357                                 goto retry;
 358                         if (s) {
 359                                 up_write(&s->s_umount);
 360                                 destroy_super(s);
 361                         }
 362                         return old;
 363                 }
 364         }
 365         if (!s) {
 366                 spin_unlock(&sb_lock);
 367                 s = alloc_super(type);
 368                 if (!s)
 369                         return ERR_PTR(-ENOMEM);
 370                 goto retry;
 371         }
 372
 373         err = set(s, data);
 374         if (err) {
 375                 spin_unlock(&sb_lock);
 376                 up_write(&s->s_umount);
 377                 destroy_super(s);
 378                 return ERR_PTR(err);
 379         }
 380         s->s_type = type;
 381         strlcpy(s->s_id, type->name, sizeof(s->s_id));
 382         list_add_tail(&s->s_list, &super_blocks);
 383         list_add(&s->s_instances, &type->fs_supers);
 384         spin_unlock(&sb_lock);
 385         get_filesystem(type);
 386         return s;
 387 }
 388
 389 EXPORT_SYMBOL(sget);
 390
 391 void drop_super(struct super_block *sb)
 392 {
 393         up_read(&sb->s_umount);
 394         put_super(sb);
 395 }
 396
 397 EXPORT_SYMBOL(drop_super);
 398
 399 /**
 400  * sync_supers - helper for periodic superblock writeback
 401  *
 402  * Call the write_super method if present on all dirty superblocks in
 403  * the system.  This is for the periodic writeback used by most older
 404  * filesystems.  For data integrity superblock writeback use
 405  * sync_filesystems() instead.
 406  *
 407  * Note: check the dirty flag before waiting, so we don't
 408  * hold up the sync while mounting a device. (The newly
 409  * mounted device won't need syncing.)
 410  */
 411 void sync_supers(void)
 412 {
 413         struct super_block *sb;
 414
 415         spin_lock(&sb_lock);
 416 restart:
 417         list_for_each_entry(sb, &super_blocks, s_list) {
 418                 if (sb->s_op->write_super && sb->s_dirt) {
 419                         sb->s_count++;
 420                         spin_unlock(&sb_lock);
 421
 422                         down_read(&sb->s_umount);
 423                         if (sb->s_root && sb->s_dirt)
 424                                 sb->s_op->write_super(sb);
 425                         up_read(&sb->s_umount);
 426
 427                         spin_lock(&sb_lock);
 428                         if (__put_super_and_need_restart(sb))
 429                                 goto restart;
 430                 }
 431         }
 432         spin_unlock(&sb_lock);
 433 }
 434
 435 /**
 436  *      get_super - get the superblock of a device
 437  *      @bdev: device to get the superblock for
 438  *
 439  *      Scans the superblock list and finds the superblock of the file system
 440  *      mounted on the device given. %NULL is returned if no match is found.
 441  */
 442
 443 struct super_block * get_super(struct block_device *bdev)
 444 {
 445         struct super_block *sb;
 446
 447         if (!bdev)
 448                 return NULL;
 449
 450         spin_lock(&sb_lock);
 451 rescan:
 452         list_for_each_entry(sb, &super_blocks, s_list) {
 453                 if (sb->s_bdev == bdev) {
 454                         sb->s_count++;
 455                         spin_unlock(&sb_lock);
 456                         down_read(&sb->s_umount);
 457                         if (sb->s_root)
 458                                 return sb;
 459                         up_read(&sb->s_umount);
 460                         /* restart only when sb is no longer on the list */
 461                         spin_lock(&sb_lock);
 462                         if (__put_super_and_need_restart(sb))
 463                                 goto rescan;
 464                 }
 465         }
 466         spin_unlock(&sb_lock);
 467         return NULL;
 468 }
 469
 470 EXPORT_SYMBOL(get_super);
 471
 472 struct super_block * user_get_super(dev_t dev)
 473 {
 474         struct super_block *sb;
 475
 476         spin_lock(&sb_lock);
 477 rescan:
 478         list_for_each_entry(sb, &super_blocks, s_list) {
 479                 if (sb->s_dev ==  dev) {
 480                         sb->s_count++;
 481                         spin_unlock(&sb_lock);
 482                         down_read(&sb->s_umount);
 483                         if (sb->s_root)
 484                                 return sb;
 485                         up_read(&sb->s_umount);
 486                         /* restart only when sb is no longer on the list */
 487                         spin_lock(&sb_lock);
 488                         if (__put_super_and_need_restart(sb))
 489                                 goto rescan;
 490                 }
 491         }
 492         spin_unlock(&sb_lock);
 493         return NULL;
 494 }
 495
 496 SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf)
 497 {
 498         struct super_block *s;
 499         struct ustat tmp;
 500         struct kstatfs sbuf;
 501         int err = -EINVAL;
 502
 503         s = user_get_super(new_decode_dev(dev));
 504         if (s == NULL)
 505                 goto out;
 506         err = vfs_statfs(s->s_root, &sbuf);
 507         drop_super(s);
 508         if (err)
 509                 goto out;
 510
 511         memset(&tmp,0,sizeof(struct ustat));
 512         tmp.f_tfree = sbuf.f_bfree;
 513         tmp.f_tinode = sbuf.f_ffree;
 514
 515         err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
 516 out:
 517         return err;
 518 }
 519
 520 /**
 521  *      do_remount_sb - asks filesystem to change mount options.
 522  *      @sb:    superblock in question
 523  *      @flags: numeric part of options
 524  *      @data:  the rest of options
 525  *      @force: whether or not to force the change
 526  *
 527  *      Alters the mount options of a mounted file system.
 528  */
 529 int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
 530 {
 531         int retval;
 532         int remount_rw;
 533
 534 #ifdef CONFIG_BLOCK
 535         if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
 536                 return -EACCES;
 537 #endif
 538         if (flags & MS_RDONLY)
 539                 acct_auto_close(sb);
 540         shrink_dcache_sb(sb);
 541         sync_filesystem(sb);
 542
 543         /* If we are remounting RDONLY and current sb is read/write,
 544            make sure there are no rw files opened */
 545         if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
 546                 if (force)
 547                         mark_files_ro(sb);
 548                 else if (!fs_may_remount_ro(sb))
 549                         return -EBUSY;
 550                 retval = vfs_dq_off(sb, 1);
 551                 if (retval < 0 && retval != -ENOSYS)
 552                         return -EBUSY;
 553         }
 554         remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
 555
 556         if (sb->s_op->remount_fs) {
 557                 retval = sb->s_op->remount_fs(sb, &flags, data);
 558                 if (retval)
 559                         return retval;
 560         }
 561         sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
 562         if (remount_rw)
 563                 vfs_dq_quota_on_remount(sb);
 564         return 0;
 565 }
 566
 567 static void do_emergency_remount(struct work_struct *work)
 568 {
 569         struct super_block *sb;
 570
 571         spin_lock(&sb_lock);
 572         list_for_each_entry(sb, &super_blocks, s_list) {
 573                 sb->s_count++;
 574                 spin_unlock(&sb_lock);
 575                 down_write(&sb->s_umount);
 576                 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
 577                         /*
 578                          * ->remount_fs needs lock_kernel().
 579                          *
 580                          * What lock protects sb->s_flags??
 581                          */
 582                         do_remount_sb(sb, MS_RDONLY, NULL, 1);
 583                 }
 584                 up_write(&sb->s_umount);
 585                 put_super(sb);
 586                 spin_lock(&sb_lock);
 587         }
 588         spin_unlock(&sb_lock);
 589         kfree(work);
 590         printk("Emergency Remount complete\n");
 591 }
 592
 593 void emergency_remount(void)
 594 {
 595         struct work_struct *work;
 596
 597         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 598         if (work) {
 599                 INIT_WORK(work, do_emergency_remount);
 600                 schedule_work(work);
 601         }
 602 }
 603
 604 /*
 605  * Unnamed block devices are dummy devices used by virtual
 606  * filesystems which don't use real block-devices.  -- jrs
 607  */
 608
 609 static DEFINE_IDA(unnamed_dev_ida);
 610 static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
 611 static int unnamed_dev_start = 0; /* don't bother trying below it */
 612
 613 int set_anon_super(struct super_block *s, void *data)
 614 {
 615         int dev;
 616         int error;
 617
 618  retry:
 619         if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
 620                 return -ENOMEM;
 621         spin_lock(&unnamed_dev_lock);
 622         error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
 623         if (!error)
 624                 unnamed_dev_start = dev + 1;
 625         spin_unlock(&unnamed_dev_lock);
 626         if (error == -EAGAIN)
 627                 /* We raced and lost with another CPU. */
 628                 goto retry;
 629         else if (error)
 630                 return -EAGAIN;
 631
 632         if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
 633                 spin_lock(&unnamed_dev_lock);
 634                 ida_remove(&unnamed_dev_ida, dev);
 635                 if (unnamed_dev_start > dev)
 636                         unnamed_dev_start = dev;
 637                 spin_unlock(&unnamed_dev_lock);
 638                 return -EMFILE;
 639         }
 640         s->s_dev = MKDEV(0, dev & MINORMASK);
 641         return 0;
 642 }
 643
 644 EXPORT_SYMBOL(set_anon_super);
 645
 646 void kill_anon_super(struct super_block *sb)
 647 {
 648         int slot = MINOR(sb->s_dev);
 649
 650         generic_shutdown_super(sb);
 651         spin_lock(&unnamed_dev_lock);
 652         ida_remove(&unnamed_dev_ida, slot);
 653         if (slot < unnamed_dev_start)
 654                 unnamed_dev_start = slot;
 655         spin_unlock(&unnamed_dev_lock);
 656 }
 657
 658 EXPORT_SYMBOL(kill_anon_super);
 659
 660 void kill_litter_super(struct super_block *sb)
 661 {
 662         if (sb->s_root)
 663                 d_genocide(sb->s_root);
 664         kill_anon_super(sb);
 665 }
 666
 667 EXPORT_SYMBOL(kill_litter_super);
 668
 669 static int ns_test_super(struct super_block *sb, void *data)
 670 {
 671         return sb->s_fs_info == data;
 672 }
 673
 674 static int ns_set_super(struct super_block *sb, void *data)
 675 {
 676         sb->s_fs_info = data;
 677         return set_anon_super(sb, NULL);
 678 }
 679
 680 int get_sb_ns(struct file_system_type *fs_type, int flags, void *data,
 681         int (*fill_super)(struct super_block *, void *, int),
 682         struct vfsmount *mnt)
 683 {
 684         struct super_block *sb;
 685
 686         sb = sget(fs_type, ns_test_super, ns_set_super, data);
 687         if (IS_ERR(sb))
 688                 return PTR_ERR(sb);
 689
 690         if (!sb->s_root) {
 691                 int err;
 692                 sb->s_flags = flags;
 693                 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
 694                 if (err) {
 695                         deactivate_locked_super(sb);
 696                         return err;
 697                 }
 698
 699                 sb->s_flags |= MS_ACTIVE;
 700         }
 701
 702         simple_set_mnt(mnt, sb);
 703         return 0;
 704 }
 705
 706 EXPORT_SYMBOL(get_sb_ns);
 707
 708 #ifdef CONFIG_BLOCK
 709 static int set_bdev_super(struct super_block *s, void *data)
 710 {
 711         s->s_bdev = data;
 712         s->s_dev = s->s_bdev->bd_dev;
 713         return 0;
 714 }
 715
 716 static int test_bdev_super(struct super_block *s, void *data)
 717 {
 718         return (void *)s->s_bdev == data;
 719 }
 720
 721 int get_sb_bdev(struct file_system_type *fs_type,
 722         int flags, const char *dev_name, void *data,
 723         int (*fill_super)(struct super_block *, void *, int),
 724         struct vfsmount *mnt)
 725 {
 726         struct block_device *bdev;
 727         struct super_block *s;
 728         fmode_t mode = FMODE_READ;
 729         int error = 0;
 730
 731         if (!(flags & MS_RDONLY))
 732                 mode |= FMODE_WRITE;
 733
 734         bdev = open_bdev_exclusive(dev_name, mode, fs_type);
 735         if (IS_ERR(bdev))
 736                 return PTR_ERR(bdev);
 737
 738         /*
 739          * once the super is inserted into the list by sget, s_umount
 740          * will protect the lockfs code from trying to start a snapshot
 741          * while we are mounting
 742          */
 743         down(&bdev->bd_mount_sem);
 744         s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
 745         up(&bdev->bd_mount_sem);
 746         if (IS_ERR(s))
 747                 goto error_s;
 748
 749         if (s->s_root) {
 750                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 751                         deactivate_locked_super(s);
 752                         error = -EBUSY;
 753                         goto error_bdev;
 754                 }
 755
 756                 close_bdev_exclusive(bdev, mode);
 757         } else {
 758                 char b[BDEVNAME_SIZE];
 759
 760                 s->s_flags = flags;
 761                 s->s_mode = mode;
 762                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 763                 sb_set_blocksize(s, block_size(bdev));
 764                 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
 765                 if (error) {
 766                         deactivate_locked_super(s);
 767                         goto error;
 768                 }
 769
 770                 s->s_flags |= MS_ACTIVE;
 771                 bdev->bd_super = s;
 772         }
 773
 774         simple_set_mnt(mnt, s);
 775         return 0;
 776
 777 error_s:
 778         error = PTR_ERR(s);
 779 error_bdev:
 780         close_bdev_exclusive(bdev, mode);
 781 error:
 782         return error;
 783 }
 784
 785 EXPORT_SYMBOL(get_sb_bdev);
 786
 787 void kill_block_super(struct super_block *sb)
 788 {
 789         struct block_device *bdev = sb->s_bdev;
 790         fmode_t mode = sb->s_mode;
 791
 792         bdev->bd_super = NULL;
 793         generic_shutdown_super(sb);
 794         sync_blockdev(bdev);
 795         close_bdev_exclusive(bdev, mode);
 796 }
 797
 798 EXPORT_SYMBOL(kill_block_super);
 799 #endif
 800
 801 int get_sb_nodev(struct file_system_type *fs_type,
 802         int flags, void *data,
 803         int (*fill_super)(struct super_block *, void *, int),
 804         struct vfsmount *mnt)
 805 {
 806         int error;
 807         struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
 808
 809         if (IS_ERR(s))
 810                 return PTR_ERR(s);
 811
 812         s->s_flags = flags;
 813
 814         error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
 815         if (error) {
 816                 deactivate_locked_super(s);
 817                 return error;
 818         }
 819         s->s_flags |= MS_ACTIVE;
 820         simple_set_mnt(mnt, s);
 821         return 0;
 822 }
 823
 824 EXPORT_SYMBOL(get_sb_nodev);
 825
 826 static int compare_single(struct super_block *s, void *p)
 827 {
 828         return 1;
 829 }
 830
 831 int get_sb_single(struct file_system_type *fs_type,
 832         int flags, void *data,
 833         int (*fill_super)(struct super_block *, void *, int),
 834         struct vfsmount *mnt)
 835 {
 836         struct super_block *s;
 837         int error;
 838
 839         s = sget(fs_type, compare_single, set_anon_super, NULL);
 840         if (IS_ERR(s))
 841                 return PTR_ERR(s);
 842         if (!s->s_root) {
 843                 s->s_flags = flags;
 844                 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
 845                 if (error) {
 846                         deactivate_locked_super(s);
 847                         return error;
 848                 }
 849                 s->s_flags |= MS_ACTIVE;
 850         }
 851         do_remount_sb(s, flags, data, 0);
 852         simple_set_mnt(mnt, s);
 853         return 0;
 854 }
 855
 856 EXPORT_SYMBOL(get_sb_single);
 857
 858 struct vfsmount *
 859 vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
 860 {
 861         struct vfsmount *mnt;
 862         char *secdata = NULL;
 863         int error;
 864
 865         if (!type)
 866                 return ERR_PTR(-ENODEV);
 867
 868         error = -ENOMEM;
 869         mnt = alloc_vfsmnt(name);
 870         if (!mnt)
 871                 goto out;
 872
 873         if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
 874                 secdata = alloc_secdata();
 875                 if (!secdata)
 876                         goto out_mnt;
 877
 878                 error = security_sb_copy_data(data, secdata);
 879                 if (error)
 880                         goto out_free_secdata;
 881         }
 882
 883         error = type->get_sb(type, flags, name, data, mnt);
 884         if (error < 0)
 885                 goto out_free_secdata;
 886         BUG_ON(!mnt->mnt_sb);
 887
 888         error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
 889         if (error)
 890                 goto out_sb;
 891
 892         mnt->mnt_mountpoint = mnt->mnt_root;
 893         mnt->mnt_parent = mnt;
 894         up_write(&mnt->mnt_sb->s_umount);
 895         free_secdata(secdata);
 896         return mnt;
 897 out_sb:
 898         dput(mnt->mnt_root);
 899         deactivate_locked_super(mnt->mnt_sb);
 900 out_free_secdata:
 901         free_secdata(secdata);
 902 out_mnt:
 903         free_vfsmnt(mnt);
 904 out:
 905         return ERR_PTR(error);
 906 }
 907
 908 EXPORT_SYMBOL_GPL(vfs_kern_mount);
 909
 910 static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
 911 {
 912         int err;
 913         const char *subtype = strchr(fstype, '.');
 914         if (subtype) {
 915                 subtype++;
 916                 err = -EINVAL;
 917                 if (!subtype[0])
 918                         goto err;
 919         } else
 920                 subtype = "";
 921
 922         mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
 923         err = -ENOMEM;
 924         if (!mnt->mnt_sb->s_subtype)
 925                 goto err;
 926         return mnt;
 927
 928  err:
 929         mntput(mnt);
 930         return ERR_PTR(err);
 931 }
 932
 933 struct vfsmount *
 934 do_kern_mount(const char *fstype, int flags, const char *name, void *data)
 935 {
 936         struct file_system_type *type = get_fs_type(fstype);
 937         struct vfsmount *mnt;
 938         if (!type)
 939                 return ERR_PTR(-ENODEV);
 940         mnt = vfs_kern_mount(type, flags, name, data);
 941         if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
 942             !mnt->mnt_sb->s_subtype)
 943                 mnt = fs_set_subtype(mnt, fstype);
 944         put_filesystem(type);
 945         return mnt;
 946 }
 947 EXPORT_SYMBOL_GPL(do_kern_mount);
 948
 949 struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
 950 {
 951         return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
 952 }
 953
 954 EXPORT_SYMBOL_GPL(kern_mount_data);