Root/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 * Torbj√∂rn 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
44LIST_HEAD(super_blocks);
45DEFINE_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 */
54static 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 const 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_files);
66        INIT_LIST_HEAD(&s->s_instances);
67        INIT_HLIST_HEAD(&s->s_anon);
68        INIT_LIST_HEAD(&s->s_inodes);
69        INIT_LIST_HEAD(&s->s_dentry_lru);
70        init_rwsem(&s->s_umount);
71        mutex_init(&s->s_lock);
72        lockdep_set_class(&s->s_umount, &type->s_umount_key);
73        /*
74         * The locking rules for s_lock are up to the
75         * filesystem. For example ext3fs has different
76         * lock ordering than usbfs:
77         */
78        lockdep_set_class(&s->s_lock, &type->s_lock_key);
79        /*
80         * sget() can have s_umount recursion.
81         *
82         * When it cannot find a suitable sb, it allocates a new
83         * one (this one), and tries again to find a suitable old
84         * one.
85         *
86         * In case that succeeds, it will acquire the s_umount
87         * lock of the old one. Since these are clearly distrinct
88         * locks, and this object isn't exposed yet, there's no
89         * risk of deadlocks.
90         *
91         * Annotate this by putting this lock in a different
92         * subclass.
93         */
94        down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
95        s->s_count = S_BIAS;
96        atomic_set(&s->s_active, 1);
97        mutex_init(&s->s_vfs_rename_mutex);
98        mutex_init(&s->s_dquot.dqio_mutex);
99        mutex_init(&s->s_dquot.dqonoff_mutex);
100        init_rwsem(&s->s_dquot.dqptr_sem);
101        init_waitqueue_head(&s->s_wait_unfrozen);
102        s->s_maxbytes = MAX_NON_LFS;
103        s->dq_op = sb_dquot_ops;
104        s->s_qcop = sb_quotactl_ops;
105        s->s_op = &default_op;
106        s->s_time_gran = 1000000000;
107    }
108out:
109    return s;
110}
111
112/**
113 * destroy_super - frees a superblock
114 * @s: superblock to free
115 *
116 * Frees a superblock.
117 */
118static inline void destroy_super(struct super_block *s)
119{
120    security_sb_free(s);
121    kfree(s->s_subtype);
122    kfree(s->s_options);
123    kfree(s);
124}
125
126/* Superblock refcounting */
127
128/*
129 * Drop a superblock's refcount. Returns non-zero if the superblock was
130 * destroyed. The caller must hold sb_lock.
131 */
132static int __put_super(struct super_block *sb)
133{
134    int ret = 0;
135
136    if (!--sb->s_count) {
137        destroy_super(sb);
138        ret = 1;
139    }
140    return ret;
141}
142
143/*
144 * Drop a superblock's refcount.
145 * Returns non-zero if the superblock is about to be destroyed and
146 * at least is already removed from super_blocks list, so if we are
147 * making a loop through super blocks then we need to restart.
148 * The caller must hold sb_lock.
149 */
150int __put_super_and_need_restart(struct super_block *sb)
151{
152    /* check for race with generic_shutdown_super() */
153    if (list_empty(&sb->s_list)) {
154        /* super block is removed, need to restart... */
155        __put_super(sb);
156        return 1;
157    }
158    /* can't be the last, since s_list is still in use */
159    sb->s_count--;
160    BUG_ON(sb->s_count == 0);
161    return 0;
162}
163
164/**
165 * put_super - drop a temporary reference to superblock
166 * @sb: superblock in question
167 *
168 * Drops a temporary reference, frees superblock if there's no
169 * references left.
170 */
171void put_super(struct super_block *sb)
172{
173    spin_lock(&sb_lock);
174    __put_super(sb);
175    spin_unlock(&sb_lock);
176}
177
178
179/**
180 * deactivate_super - drop an active reference to superblock
181 * @s: superblock to deactivate
182 *
183 * Drops an active reference to superblock, acquiring a temprory one if
184 * there is no active references left. In that case we lock superblock,
185 * tell fs driver to shut it down and drop the temporary reference we
186 * had just acquired.
187 */
188void deactivate_super(struct super_block *s)
189{
190    struct file_system_type *fs = s->s_type;
191    if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
192        s->s_count -= S_BIAS-1;
193        spin_unlock(&sb_lock);
194        vfs_dq_off(s, 0);
195        down_write(&s->s_umount);
196        fs->kill_sb(s);
197        put_filesystem(fs);
198        put_super(s);
199    }
200}
201
202EXPORT_SYMBOL(deactivate_super);
203
204/**
205 * deactivate_locked_super - drop an active reference to superblock
206 * @s: superblock to deactivate
207 *
208 * Equivalent of up_write(&s->s_umount); deactivate_super(s);, except that
209 * it does not unlock it until it's all over. As the result, it's safe to
210 * use to dispose of new superblock on ->get_sb() failure exits - nobody
211 * will see the sucker until it's all over. Equivalent using up_write +
212 * deactivate_super is safe for that purpose only if superblock is either
213 * safe to use or has NULL ->s_root when we unlock.
214 */
215void deactivate_locked_super(struct super_block *s)
216{
217    struct file_system_type *fs = s->s_type;
218    if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
219        s->s_count -= S_BIAS-1;
220        spin_unlock(&sb_lock);
221        vfs_dq_off(s, 0);
222        fs->kill_sb(s);
223        put_filesystem(fs);
224        put_super(s);
225    } else {
226        up_write(&s->s_umount);
227    }
228}
229
230EXPORT_SYMBOL(deactivate_locked_super);
231
232/**
233 * grab_super - acquire an active reference
234 * @s: reference we are trying to make active
235 *
236 * Tries to acquire an active reference. grab_super() is used when we
237 * had just found a superblock in super_blocks or fs_type->fs_supers
238 * and want to turn it into a full-blown active reference. grab_super()
239 * is called with sb_lock held and drops it. Returns 1 in case of
240 * success, 0 if we had failed (superblock contents was already dead or
241 * dying when grab_super() had been called).
242 */
243static int grab_super(struct super_block *s) __releases(sb_lock)
244{
245    s->s_count++;
246    spin_unlock(&sb_lock);
247    down_write(&s->s_umount);
248    if (s->s_root) {
249        spin_lock(&sb_lock);
250        if (s->s_count > S_BIAS) {
251            atomic_inc(&s->s_active);
252            s->s_count--;
253            spin_unlock(&sb_lock);
254            return 1;
255        }
256        spin_unlock(&sb_lock);
257    }
258    up_write(&s->s_umount);
259    put_super(s);
260    yield();
261    return 0;
262}
263
264/*
265 * Superblock locking. We really ought to get rid of these two.
266 */
267void lock_super(struct super_block * sb)
268{
269    get_fs_excl();
270    mutex_lock(&sb->s_lock);
271}
272
273void unlock_super(struct super_block * sb)
274{
275    put_fs_excl();
276    mutex_unlock(&sb->s_lock);
277}
278
279EXPORT_SYMBOL(lock_super);
280EXPORT_SYMBOL(unlock_super);
281
282/**
283 * generic_shutdown_super - common helper for ->kill_sb()
284 * @sb: superblock to kill
285 *
286 * generic_shutdown_super() does all fs-independent work on superblock
287 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
288 * that need destruction out of superblock, call generic_shutdown_super()
289 * and release aforementioned objects. Note: dentries and inodes _are_
290 * taken care of and do not need specific handling.
291 *
292 * Upon calling this function, the filesystem may no longer alter or
293 * rearrange the set of dentries belonging to this super_block, nor may it
294 * change the attachments of dentries to inodes.
295 */
296void generic_shutdown_super(struct super_block *sb)
297{
298    const struct super_operations *sop = sb->s_op;
299
300
301    if (sb->s_root) {
302        shrink_dcache_for_umount(sb);
303        sync_filesystem(sb);
304        get_fs_excl();
305        sb->s_flags &= ~MS_ACTIVE;
306
307        /* bad name - it should be evict_inodes() */
308        invalidate_inodes(sb);
309
310        if (sop->put_super)
311            sop->put_super(sb);
312
313        /* Forget any remaining inodes */
314        if (invalidate_inodes(sb)) {
315            printk("VFS: Busy inodes after unmount of %s. "
316               "Self-destruct in 5 seconds. Have a nice day...\n",
317               sb->s_id);
318        }
319        put_fs_excl();
320    }
321    spin_lock(&sb_lock);
322    /* should be initialized for __put_super_and_need_restart() */
323    list_del_init(&sb->s_list);
324    list_del(&sb->s_instances);
325    spin_unlock(&sb_lock);
326    up_write(&sb->s_umount);
327}
328
329EXPORT_SYMBOL(generic_shutdown_super);
330
331/**
332 * sget - find or create a superblock
333 * @type: filesystem type superblock should belong to
334 * @test: comparison callback
335 * @set: setup callback
336 * @data: argument to each of them
337 */
338struct super_block *sget(struct file_system_type *type,
339            int (*test)(struct super_block *,void *),
340            int (*set)(struct super_block *,void *),
341            void *data)
342{
343    struct super_block *s = NULL;
344    struct super_block *old;
345    int err;
346
347retry:
348    spin_lock(&sb_lock);
349    if (test) {
350        list_for_each_entry(old, &type->fs_supers, s_instances) {
351            if (!test(old, data))
352                continue;
353            if (!grab_super(old))
354                goto retry;
355            if (s) {
356                up_write(&s->s_umount);
357                destroy_super(s);
358            }
359            return old;
360        }
361    }
362    if (!s) {
363        spin_unlock(&sb_lock);
364        s = alloc_super(type);
365        if (!s)
366            return ERR_PTR(-ENOMEM);
367        goto retry;
368    }
369        
370    err = set(s, data);
371    if (err) {
372        spin_unlock(&sb_lock);
373        up_write(&s->s_umount);
374        destroy_super(s);
375        return ERR_PTR(err);
376    }
377    s->s_type = type;
378    strlcpy(s->s_id, type->name, sizeof(s->s_id));
379    list_add_tail(&s->s_list, &super_blocks);
380    list_add(&s->s_instances, &type->fs_supers);
381    spin_unlock(&sb_lock);
382    get_filesystem(type);
383    return s;
384}
385
386EXPORT_SYMBOL(sget);
387
388void drop_super(struct super_block *sb)
389{
390    up_read(&sb->s_umount);
391    put_super(sb);
392}
393
394EXPORT_SYMBOL(drop_super);
395
396/**
397 * sync_supers - helper for periodic superblock writeback
398 *
399 * Call the write_super method if present on all dirty superblocks in
400 * the system. This is for the periodic writeback used by most older
401 * filesystems. For data integrity superblock writeback use
402 * sync_filesystems() instead.
403 *
404 * Note: check the dirty flag before waiting, so we don't
405 * hold up the sync while mounting a device. (The newly
406 * mounted device won't need syncing.)
407 */
408void sync_supers(void)
409{
410    struct super_block *sb;
411
412    spin_lock(&sb_lock);
413restart:
414    list_for_each_entry(sb, &super_blocks, s_list) {
415        if (sb->s_op->write_super && sb->s_dirt) {
416            sb->s_count++;
417            spin_unlock(&sb_lock);
418
419            down_read(&sb->s_umount);
420            if (sb->s_root && sb->s_dirt)
421                sb->s_op->write_super(sb);
422            up_read(&sb->s_umount);
423
424            spin_lock(&sb_lock);
425            if (__put_super_and_need_restart(sb))
426                goto restart;
427        }
428    }
429    spin_unlock(&sb_lock);
430}
431
432/**
433 * get_super - get the superblock of a device
434 * @bdev: device to get the superblock for
435 *
436 * Scans the superblock list and finds the superblock of the file system
437 * mounted on the device given. %NULL is returned if no match is found.
438 */
439
440struct super_block * get_super(struct block_device *bdev)
441{
442    struct super_block *sb;
443
444    if (!bdev)
445        return NULL;
446
447    spin_lock(&sb_lock);
448rescan:
449    list_for_each_entry(sb, &super_blocks, s_list) {
450        if (sb->s_bdev == bdev) {
451            sb->s_count++;
452            spin_unlock(&sb_lock);
453            down_read(&sb->s_umount);
454            if (sb->s_root)
455                return sb;
456            up_read(&sb->s_umount);
457            /* restart only when sb is no longer on the list */
458            spin_lock(&sb_lock);
459            if (__put_super_and_need_restart(sb))
460                goto rescan;
461        }
462    }
463    spin_unlock(&sb_lock);
464    return NULL;
465}
466
467EXPORT_SYMBOL(get_super);
468
469/**
470 * get_active_super - get an active reference to the superblock of a device
471 * @bdev: device to get the superblock for
472 *
473 * Scans the superblock list and finds the superblock of the file system
474 * mounted on the device given. Returns the superblock with an active
475 * reference and s_umount held exclusively or %NULL if none was found.
476 */
477struct super_block *get_active_super(struct block_device *bdev)
478{
479    struct super_block *sb;
480
481    if (!bdev)
482        return NULL;
483
484    spin_lock(&sb_lock);
485    list_for_each_entry(sb, &super_blocks, s_list) {
486        if (sb->s_bdev != bdev)
487            continue;
488
489        sb->s_count++;
490        spin_unlock(&sb_lock);
491        down_write(&sb->s_umount);
492        if (sb->s_root) {
493            spin_lock(&sb_lock);
494            if (sb->s_count > S_BIAS) {
495                atomic_inc(&sb->s_active);
496                sb->s_count--;
497                spin_unlock(&sb_lock);
498                return sb;
499            }
500            spin_unlock(&sb_lock);
501        }
502        up_write(&sb->s_umount);
503        put_super(sb);
504        yield();
505        spin_lock(&sb_lock);
506    }
507    spin_unlock(&sb_lock);
508    return NULL;
509}
510 
511struct super_block * user_get_super(dev_t dev)
512{
513    struct super_block *sb;
514
515    spin_lock(&sb_lock);
516rescan:
517    list_for_each_entry(sb, &super_blocks, s_list) {
518        if (sb->s_dev == dev) {
519            sb->s_count++;
520            spin_unlock(&sb_lock);
521            down_read(&sb->s_umount);
522            if (sb->s_root)
523                return sb;
524            up_read(&sb->s_umount);
525            /* restart only when sb is no longer on the list */
526            spin_lock(&sb_lock);
527            if (__put_super_and_need_restart(sb))
528                goto rescan;
529        }
530    }
531    spin_unlock(&sb_lock);
532    return NULL;
533}
534
535SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf)
536{
537        struct super_block *s;
538        struct ustat tmp;
539        struct kstatfs sbuf;
540    int err = -EINVAL;
541
542        s = user_get_super(new_decode_dev(dev));
543        if (s == NULL)
544                goto out;
545    err = vfs_statfs(s->s_root, &sbuf);
546    drop_super(s);
547    if (err)
548        goto out;
549
550        memset(&tmp,0,sizeof(struct ustat));
551        tmp.f_tfree = sbuf.f_bfree;
552        tmp.f_tinode = sbuf.f_ffree;
553
554        err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
555out:
556    return err;
557}
558
559/**
560 * do_remount_sb - asks filesystem to change mount options.
561 * @sb: superblock in question
562 * @flags: numeric part of options
563 * @data: the rest of options
564 * @force: whether or not to force the change
565 *
566 * Alters the mount options of a mounted file system.
567 */
568int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
569{
570    int retval;
571    int remount_rw, remount_ro;
572
573    if (sb->s_frozen != SB_UNFROZEN)
574        return -EBUSY;
575
576#ifdef CONFIG_BLOCK
577    if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
578        return -EACCES;
579#endif
580
581    if (flags & MS_RDONLY)
582        acct_auto_close(sb);
583    shrink_dcache_sb(sb);
584    sync_filesystem(sb);
585
586    remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY);
587    remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
588
589    /* If we are remounting RDONLY and current sb is read/write,
590       make sure there are no rw files opened */
591    if (remount_ro) {
592        if (force)
593            mark_files_ro(sb);
594        else if (!fs_may_remount_ro(sb))
595            return -EBUSY;
596        retval = vfs_dq_off(sb, 1);
597        if (retval < 0 && retval != -ENOSYS)
598            return -EBUSY;
599    }
600
601    if (sb->s_op->remount_fs) {
602        retval = sb->s_op->remount_fs(sb, &flags, data);
603        if (retval)
604            return retval;
605    }
606    sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
607    if (remount_rw)
608        vfs_dq_quota_on_remount(sb);
609    /*
610     * Some filesystems modify their metadata via some other path than the
611     * bdev buffer cache (eg. use a private mapping, or directories in
612     * pagecache, etc). Also file data modifications go via their own
613     * mappings. So If we try to mount readonly then copy the filesystem
614     * from bdev, we could get stale data, so invalidate it to give a best
615     * effort at coherency.
616     */
617    if (remount_ro && sb->s_bdev)
618        invalidate_bdev(sb->s_bdev);
619    return 0;
620}
621
622static void do_emergency_remount(struct work_struct *work)
623{
624    struct super_block *sb;
625
626    spin_lock(&sb_lock);
627    list_for_each_entry(sb, &super_blocks, s_list) {
628        sb->s_count++;
629        spin_unlock(&sb_lock);
630        down_write(&sb->s_umount);
631        if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
632            /*
633             * ->remount_fs needs lock_kernel().
634             *
635             * What lock protects sb->s_flags??
636             */
637            do_remount_sb(sb, MS_RDONLY, NULL, 1);
638        }
639        up_write(&sb->s_umount);
640        put_super(sb);
641        spin_lock(&sb_lock);
642    }
643    spin_unlock(&sb_lock);
644    kfree(work);
645    printk("Emergency Remount complete\n");
646}
647
648void emergency_remount(void)
649{
650    struct work_struct *work;
651
652    work = kmalloc(sizeof(*work), GFP_ATOMIC);
653    if (work) {
654        INIT_WORK(work, do_emergency_remount);
655        schedule_work(work);
656    }
657}
658
659/*
660 * Unnamed block devices are dummy devices used by virtual
661 * filesystems which don't use real block-devices. -- jrs
662 */
663
664static DEFINE_IDA(unnamed_dev_ida);
665static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
666static int unnamed_dev_start = 0; /* don't bother trying below it */
667
668int set_anon_super(struct super_block *s, void *data)
669{
670    int dev;
671    int error;
672
673 retry:
674    if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
675        return -ENOMEM;
676    spin_lock(&unnamed_dev_lock);
677    error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
678    if (!error)
679        unnamed_dev_start = dev + 1;
680    spin_unlock(&unnamed_dev_lock);
681    if (error == -EAGAIN)
682        /* We raced and lost with another CPU. */
683        goto retry;
684    else if (error)
685        return -EAGAIN;
686
687    if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
688        spin_lock(&unnamed_dev_lock);
689        ida_remove(&unnamed_dev_ida, dev);
690        if (unnamed_dev_start > dev)
691            unnamed_dev_start = dev;
692        spin_unlock(&unnamed_dev_lock);
693        return -EMFILE;
694    }
695    s->s_dev = MKDEV(0, dev & MINORMASK);
696    return 0;
697}
698
699EXPORT_SYMBOL(set_anon_super);
700
701void kill_anon_super(struct super_block *sb)
702{
703    int slot = MINOR(sb->s_dev);
704
705    generic_shutdown_super(sb);
706    spin_lock(&unnamed_dev_lock);
707    ida_remove(&unnamed_dev_ida, slot);
708    if (slot < unnamed_dev_start)
709        unnamed_dev_start = slot;
710    spin_unlock(&unnamed_dev_lock);
711}
712
713EXPORT_SYMBOL(kill_anon_super);
714
715void kill_litter_super(struct super_block *sb)
716{
717    if (sb->s_root)
718        d_genocide(sb->s_root);
719    kill_anon_super(sb);
720}
721
722EXPORT_SYMBOL(kill_litter_super);
723
724static int ns_test_super(struct super_block *sb, void *data)
725{
726    return sb->s_fs_info == data;
727}
728
729static int ns_set_super(struct super_block *sb, void *data)
730{
731    sb->s_fs_info = data;
732    return set_anon_super(sb, NULL);
733}
734
735int get_sb_ns(struct file_system_type *fs_type, int flags, void *data,
736    int (*fill_super)(struct super_block *, void *, int),
737    struct vfsmount *mnt)
738{
739    struct super_block *sb;
740
741    sb = sget(fs_type, ns_test_super, ns_set_super, data);
742    if (IS_ERR(sb))
743        return PTR_ERR(sb);
744
745    if (!sb->s_root) {
746        int err;
747        sb->s_flags = flags;
748        err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
749        if (err) {
750            deactivate_locked_super(sb);
751            return err;
752        }
753
754        sb->s_flags |= MS_ACTIVE;
755    }
756
757    simple_set_mnt(mnt, sb);
758    return 0;
759}
760
761EXPORT_SYMBOL(get_sb_ns);
762
763#ifdef CONFIG_BLOCK
764static int set_bdev_super(struct super_block *s, void *data)
765{
766    s->s_bdev = data;
767    s->s_dev = s->s_bdev->bd_dev;
768
769    /*
770     * We set the bdi here to the queue backing, file systems can
771     * overwrite this in ->fill_super()
772     */
773    s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
774    return 0;
775}
776
777static int test_bdev_super(struct super_block *s, void *data)
778{
779    return (void *)s->s_bdev == data;
780}
781
782int get_sb_bdev(struct file_system_type *fs_type,
783    int flags, const char *dev_name, void *data,
784    int (*fill_super)(struct super_block *, void *, int),
785    struct vfsmount *mnt)
786{
787    struct block_device *bdev;
788    struct super_block *s;
789    fmode_t mode = FMODE_READ;
790    int error = 0;
791
792    if (!(flags & MS_RDONLY))
793        mode |= FMODE_WRITE;
794
795    bdev = open_bdev_exclusive(dev_name, mode, fs_type);
796    if (IS_ERR(bdev))
797        return PTR_ERR(bdev);
798
799    /*
800     * once the super is inserted into the list by sget, s_umount
801     * will protect the lockfs code from trying to start a snapshot
802     * while we are mounting
803     */
804    mutex_lock(&bdev->bd_fsfreeze_mutex);
805    if (bdev->bd_fsfreeze_count > 0) {
806        mutex_unlock(&bdev->bd_fsfreeze_mutex);
807        error = -EBUSY;
808        goto error_bdev;
809    }
810    s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
811    mutex_unlock(&bdev->bd_fsfreeze_mutex);
812    if (IS_ERR(s))
813        goto error_s;
814
815    if (s->s_root) {
816        if ((flags ^ s->s_flags) & MS_RDONLY) {
817            deactivate_locked_super(s);
818            error = -EBUSY;
819            goto error_bdev;
820        }
821
822        close_bdev_exclusive(bdev, mode);
823    } else {
824        char b[BDEVNAME_SIZE];
825
826        s->s_flags = flags;
827        s->s_mode = mode;
828        strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
829        sb_set_blocksize(s, block_size(bdev));
830        error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
831        if (error) {
832            deactivate_locked_super(s);
833            goto error;
834        }
835
836        s->s_flags |= MS_ACTIVE;
837        bdev->bd_super = s;
838    }
839
840    simple_set_mnt(mnt, s);
841    return 0;
842
843error_s:
844    error = PTR_ERR(s);
845error_bdev:
846    close_bdev_exclusive(bdev, mode);
847error:
848    return error;
849}
850
851EXPORT_SYMBOL(get_sb_bdev);
852
853void kill_block_super(struct super_block *sb)
854{
855    struct block_device *bdev = sb->s_bdev;
856    fmode_t mode = sb->s_mode;
857
858    bdev->bd_super = NULL;
859    generic_shutdown_super(sb);
860    sync_blockdev(bdev);
861    close_bdev_exclusive(bdev, mode);
862}
863
864EXPORT_SYMBOL(kill_block_super);
865#endif
866
867int get_sb_nodev(struct file_system_type *fs_type,
868    int flags, void *data,
869    int (*fill_super)(struct super_block *, void *, int),
870    struct vfsmount *mnt)
871{
872    int error;
873    struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
874
875    if (IS_ERR(s))
876        return PTR_ERR(s);
877
878    s->s_flags = flags;
879
880    error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
881    if (error) {
882        deactivate_locked_super(s);
883        return error;
884    }
885    s->s_flags |= MS_ACTIVE;
886    simple_set_mnt(mnt, s);
887    return 0;
888}
889
890EXPORT_SYMBOL(get_sb_nodev);
891
892static int compare_single(struct super_block *s, void *p)
893{
894    return 1;
895}
896
897int get_sb_single(struct file_system_type *fs_type,
898    int flags, void *data,
899    int (*fill_super)(struct super_block *, void *, int),
900    struct vfsmount *mnt)
901{
902    struct super_block *s;
903    int error;
904
905    s = sget(fs_type, compare_single, set_anon_super, NULL);
906    if (IS_ERR(s))
907        return PTR_ERR(s);
908    if (!s->s_root) {
909        s->s_flags = flags;
910        error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
911        if (error) {
912            deactivate_locked_super(s);
913            return error;
914        }
915        s->s_flags |= MS_ACTIVE;
916    } else {
917        do_remount_sb(s, flags, data, 0);
918    }
919    simple_set_mnt(mnt, s);
920    return 0;
921}
922
923EXPORT_SYMBOL(get_sb_single);
924
925struct vfsmount *
926vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
927{
928    struct vfsmount *mnt;
929    char *secdata = NULL;
930    int error;
931
932    if (!type)
933        return ERR_PTR(-ENODEV);
934
935    error = -ENOMEM;
936    mnt = alloc_vfsmnt(name);
937    if (!mnt)
938        goto out;
939
940    if (flags & MS_KERNMOUNT)
941        mnt->mnt_flags = MNT_INTERNAL;
942
943    if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
944        secdata = alloc_secdata();
945        if (!secdata)
946            goto out_mnt;
947
948        error = security_sb_copy_data(data, secdata);
949        if (error)
950            goto out_free_secdata;
951    }
952
953    error = type->get_sb(type, flags, name, data, mnt);
954    if (error < 0)
955        goto out_free_secdata;
956    BUG_ON(!mnt->mnt_sb);
957
958     error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
959     if (error)
960         goto out_sb;
961
962    /*
963     * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
964     * but s_maxbytes was an unsigned long long for many releases. Throw
965     * this warning for a little while to try and catch filesystems that
966     * violate this rule. This warning should be either removed or
967     * converted to a BUG() in 2.6.34.
968     */
969    WARN((mnt->mnt_sb->s_maxbytes < 0), "%s set sb->s_maxbytes to "
970        "negative value (%lld)\n", type->name, mnt->mnt_sb->s_maxbytes);
971
972    mnt->mnt_mountpoint = mnt->mnt_root;
973    mnt->mnt_parent = mnt;
974    up_write(&mnt->mnt_sb->s_umount);
975    free_secdata(secdata);
976    return mnt;
977out_sb:
978    dput(mnt->mnt_root);
979    deactivate_locked_super(mnt->mnt_sb);
980out_free_secdata:
981    free_secdata(secdata);
982out_mnt:
983    free_vfsmnt(mnt);
984out:
985    return ERR_PTR(error);
986}
987
988EXPORT_SYMBOL_GPL(vfs_kern_mount);
989
990static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
991{
992    int err;
993    const char *subtype = strchr(fstype, '.');
994    if (subtype) {
995        subtype++;
996        err = -EINVAL;
997        if (!subtype[0])
998            goto err;
999    } else
1000        subtype = "";
1001
1002    mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
1003    err = -ENOMEM;
1004    if (!mnt->mnt_sb->s_subtype)
1005        goto err;
1006    return mnt;
1007
1008 err:
1009    mntput(mnt);
1010    return ERR_PTR(err);
1011}
1012
1013struct vfsmount *
1014do_kern_mount(const char *fstype, int flags, const char *name, void *data)
1015{
1016    struct file_system_type *type = get_fs_type(fstype);
1017    struct vfsmount *mnt;
1018    if (!type)
1019        return ERR_PTR(-ENODEV);
1020    mnt = vfs_kern_mount(type, flags, name, data);
1021    if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
1022        !mnt->mnt_sb->s_subtype)
1023        mnt = fs_set_subtype(mnt, fstype);
1024    put_filesystem(type);
1025    return mnt;
1026}
1027EXPORT_SYMBOL_GPL(do_kern_mount);
1028
1029struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
1030{
1031    return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
1032}
1033
1034EXPORT_SYMBOL_GPL(kern_mount_data);
1035

Archive Download this file



interactive