4 * Copyright (C) 1991, 1992 Linus Torvalds
5 * Copyright (C) 2001 Andrea Arcangeli <andrea@suse.de> SuSE
8 #include <linux/init.h>
10 #include <linux/fcntl.h>
11 #include <linux/slab.h>
12 #include <linux/kmod.h>
13 #include <linux/major.h>
14 #include <linux/smp_lock.h>
15 #include <linux/device_cgroup.h>
16 #include <linux/highmem.h>
17 #include <linux/blkdev.h>
18 #include <linux/module.h>
19 #include <linux/blkpg.h>
20 #include <linux/buffer_head.h>
21 #include <linux/pagevec.h>
22 #include <linux/writeback.h>
23 #include <linux/mpage.h>
24 #include <linux/mount.h>
25 #include <linux/uio.h>
26 #include <linux/namei.h>
27 #include <linux/log2.h>
28 #include <linux/kmemleak.h>
29 #include <asm/uaccess.h>
33 struct block_device bdev
;
34 struct inode vfs_inode
;
37 static const struct address_space_operations def_blk_aops
;
39 static inline struct bdev_inode
*BDEV_I(struct inode
*inode
)
41 return container_of(inode
, struct bdev_inode
, vfs_inode
);
44 inline struct block_device
*I_BDEV(struct inode
*inode
)
46 return &BDEV_I(inode
)->bdev
;
49 EXPORT_SYMBOL(I_BDEV
);
51 static sector_t
max_block(struct block_device
*bdev
)
53 sector_t retval
= ~((sector_t
)0);
54 loff_t sz
= i_size_read(bdev
->bd_inode
);
57 unsigned int size
= block_size(bdev
);
58 unsigned int sizebits
= blksize_bits(size
);
59 retval
= (sz
>> sizebits
);
64 /* Kill _all_ buffers and pagecache , dirty or not.. */
65 static void kill_bdev(struct block_device
*bdev
)
67 if (bdev
->bd_inode
->i_mapping
->nrpages
== 0)
70 truncate_inode_pages(bdev
->bd_inode
->i_mapping
, 0);
73 int set_blocksize(struct block_device
*bdev
, int size
)
75 /* Size must be a power of two, and between 512 and PAGE_SIZE */
76 if (size
> PAGE_SIZE
|| size
< 512 || !is_power_of_2(size
))
79 /* Size cannot be smaller than the size supported by the device */
80 if (size
< bdev_logical_block_size(bdev
))
83 /* Don't change the size if it is same as current */
84 if (bdev
->bd_block_size
!= size
) {
86 bdev
->bd_block_size
= size
;
87 bdev
->bd_inode
->i_blkbits
= blksize_bits(size
);
93 EXPORT_SYMBOL(set_blocksize
);
95 int sb_set_blocksize(struct super_block
*sb
, int size
)
97 if (set_blocksize(sb
->s_bdev
, size
))
99 /* If we get here, we know size is power of two
100 * and it's value is between 512 and PAGE_SIZE */
101 sb
->s_blocksize
= size
;
102 sb
->s_blocksize_bits
= blksize_bits(size
);
103 return sb
->s_blocksize
;
106 EXPORT_SYMBOL(sb_set_blocksize
);
108 int sb_min_blocksize(struct super_block
*sb
, int size
)
110 int minsize
= bdev_logical_block_size(sb
->s_bdev
);
113 return sb_set_blocksize(sb
, size
);
116 EXPORT_SYMBOL(sb_min_blocksize
);
119 blkdev_get_block(struct inode
*inode
, sector_t iblock
,
120 struct buffer_head
*bh
, int create
)
122 if (iblock
>= max_block(I_BDEV(inode
))) {
127 * for reads, we're just trying to fill a partial page.
128 * return a hole, they will have to call get_block again
129 * before they can fill it, and they will get -EIO at that
134 bh
->b_bdev
= I_BDEV(inode
);
135 bh
->b_blocknr
= iblock
;
136 set_buffer_mapped(bh
);
141 blkdev_get_blocks(struct inode
*inode
, sector_t iblock
,
142 struct buffer_head
*bh
, int create
)
144 sector_t end_block
= max_block(I_BDEV(inode
));
145 unsigned long max_blocks
= bh
->b_size
>> inode
->i_blkbits
;
147 if ((iblock
+ max_blocks
) > end_block
) {
148 max_blocks
= end_block
- iblock
;
149 if ((long)max_blocks
<= 0) {
151 return -EIO
; /* write fully beyond EOF */
153 * It is a read which is fully beyond EOF. We return
154 * a !buffer_mapped buffer
160 bh
->b_bdev
= I_BDEV(inode
);
161 bh
->b_blocknr
= iblock
;
162 bh
->b_size
= max_blocks
<< inode
->i_blkbits
;
164 set_buffer_mapped(bh
);
169 blkdev_direct_IO(int rw
, struct kiocb
*iocb
, const struct iovec
*iov
,
170 loff_t offset
, unsigned long nr_segs
)
172 struct file
*file
= iocb
->ki_filp
;
173 struct inode
*inode
= file
->f_mapping
->host
;
175 return blockdev_direct_IO_no_locking(rw
, iocb
, inode
, I_BDEV(inode
),
176 iov
, offset
, nr_segs
, blkdev_get_blocks
, NULL
);
179 int __sync_blockdev(struct block_device
*bdev
, int wait
)
184 return filemap_flush(bdev
->bd_inode
->i_mapping
);
185 return filemap_write_and_wait(bdev
->bd_inode
->i_mapping
);
189 * Write out and wait upon all the dirty data associated with a block
190 * device via its mapping. Does not take the superblock lock.
192 int sync_blockdev(struct block_device
*bdev
)
194 return __sync_blockdev(bdev
, 1);
196 EXPORT_SYMBOL(sync_blockdev
);
199 * Write out and wait upon all dirty data associated with this
200 * device. Filesystem data as well as the underlying block
201 * device. Takes the superblock lock.
203 int fsync_bdev(struct block_device
*bdev
)
205 struct super_block
*sb
= get_super(bdev
);
207 int res
= sync_filesystem(sb
);
211 return sync_blockdev(bdev
);
213 EXPORT_SYMBOL(fsync_bdev
);
216 * freeze_bdev -- lock a filesystem and force it into a consistent state
217 * @bdev: blockdevice to lock
219 * If a superblock is found on this device, we take the s_umount semaphore
220 * on it to make sure nobody unmounts until the snapshot creation is done.
221 * The reference counter (bd_fsfreeze_count) guarantees that only the last
222 * unfreeze process can unfreeze the frozen filesystem actually when multiple
223 * freeze requests arrive simultaneously. It counts up in freeze_bdev() and
224 * count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze
227 struct super_block
*freeze_bdev(struct block_device
*bdev
)
229 struct super_block
*sb
;
232 mutex_lock(&bdev
->bd_fsfreeze_mutex
);
233 if (++bdev
->bd_fsfreeze_count
> 1) {
235 * We don't even need to grab a reference - the first call
236 * to freeze_bdev grab an active reference and only the last
237 * thaw_bdev drops it.
239 sb
= get_super(bdev
);
241 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
245 sb
= get_active_super(bdev
);
248 if (sb
->s_flags
& MS_RDONLY
) {
249 sb
->s_frozen
= SB_FREEZE_TRANS
;
250 up_write(&sb
->s_umount
);
251 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
255 sb
->s_frozen
= SB_FREEZE_WRITE
;
260 sb
->s_frozen
= SB_FREEZE_TRANS
;
263 sync_blockdev(sb
->s_bdev
);
265 if (sb
->s_op
->freeze_fs
) {
266 error
= sb
->s_op
->freeze_fs(sb
);
269 "VFS:Filesystem freeze failed\n");
270 sb
->s_frozen
= SB_UNFROZEN
;
271 deactivate_locked_super(sb
);
272 bdev
->bd_fsfreeze_count
--;
273 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
274 return ERR_PTR(error
);
277 up_write(&sb
->s_umount
);
281 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
282 return sb
; /* thaw_bdev releases s->s_umount */
284 EXPORT_SYMBOL(freeze_bdev
);
287 * thaw_bdev -- unlock filesystem
288 * @bdev: blockdevice to unlock
289 * @sb: associated superblock
291 * Unlocks the filesystem and marks it writeable again after freeze_bdev().
293 int thaw_bdev(struct block_device
*bdev
, struct super_block
*sb
)
297 mutex_lock(&bdev
->bd_fsfreeze_mutex
);
298 if (!bdev
->bd_fsfreeze_count
)
302 if (--bdev
->bd_fsfreeze_count
> 0)
308 BUG_ON(sb
->s_bdev
!= bdev
);
309 down_write(&sb
->s_umount
);
310 if (sb
->s_flags
& MS_RDONLY
)
313 if (sb
->s_op
->unfreeze_fs
) {
314 error
= sb
->s_op
->unfreeze_fs(sb
);
317 "VFS:Filesystem thaw failed\n");
318 sb
->s_frozen
= SB_FREEZE_TRANS
;
319 bdev
->bd_fsfreeze_count
++;
320 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
326 sb
->s_frozen
= SB_UNFROZEN
;
328 wake_up(&sb
->s_wait_unfrozen
);
331 deactivate_locked_super(sb
);
333 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
336 EXPORT_SYMBOL(thaw_bdev
);
338 static int blkdev_writepage(struct page
*page
, struct writeback_control
*wbc
)
340 return block_write_full_page(page
, blkdev_get_block
, wbc
);
343 static int blkdev_readpage(struct file
* file
, struct page
* page
)
345 return block_read_full_page(page
, blkdev_get_block
);
348 static int blkdev_write_begin(struct file
*file
, struct address_space
*mapping
,
349 loff_t pos
, unsigned len
, unsigned flags
,
350 struct page
**pagep
, void **fsdata
)
353 return block_write_begin(file
, mapping
, pos
, len
, flags
, pagep
, fsdata
,
357 static int blkdev_write_end(struct file
*file
, struct address_space
*mapping
,
358 loff_t pos
, unsigned len
, unsigned copied
,
359 struct page
*page
, void *fsdata
)
362 ret
= block_write_end(file
, mapping
, pos
, len
, copied
, page
, fsdata
);
365 page_cache_release(page
);
372 * for a block special file file->f_path.dentry->d_inode->i_size is zero
373 * so we compute the size by hand (just as in block_read/write above)
375 static loff_t
block_llseek(struct file
*file
, loff_t offset
, int origin
)
377 struct inode
*bd_inode
= file
->f_mapping
->host
;
381 mutex_lock(&bd_inode
->i_mutex
);
382 size
= i_size_read(bd_inode
);
389 offset
+= file
->f_pos
;
392 if (offset
>= 0 && offset
<= size
) {
393 if (offset
!= file
->f_pos
) {
394 file
->f_pos
= offset
;
398 mutex_unlock(&bd_inode
->i_mutex
);
403 * Filp is never NULL; the only case when ->fsync() is called with
404 * NULL first argument is nfsd_sync_dir() and that's not a directory.
407 int blkdev_fsync(struct file
*filp
, struct dentry
*dentry
, int datasync
)
409 struct inode
*bd_inode
= filp
->f_mapping
->host
;
410 struct block_device
*bdev
= I_BDEV(bd_inode
);
414 * There is no need to serialise calls to blkdev_issue_flush with
415 * i_mutex and doing so causes performance issues with concurrent
416 * O_SYNC writers to a block device.
418 mutex_unlock(&bd_inode
->i_mutex
);
420 error
= blkdev_issue_flush(bdev
, GFP_KERNEL
, NULL
, BLKDEV_IFL_WAIT
);
421 if (error
== -EOPNOTSUPP
)
424 mutex_lock(&bd_inode
->i_mutex
);
428 EXPORT_SYMBOL(blkdev_fsync
);
434 static __cacheline_aligned_in_smp
DEFINE_SPINLOCK(bdev_lock
);
435 static struct kmem_cache
* bdev_cachep __read_mostly
;
437 static struct inode
*bdev_alloc_inode(struct super_block
*sb
)
439 struct bdev_inode
*ei
= kmem_cache_alloc(bdev_cachep
, GFP_KERNEL
);
442 return &ei
->vfs_inode
;
445 static void bdev_destroy_inode(struct inode
*inode
)
447 struct bdev_inode
*bdi
= BDEV_I(inode
);
449 kmem_cache_free(bdev_cachep
, bdi
);
452 static void init_once(void *foo
)
454 struct bdev_inode
*ei
= (struct bdev_inode
*) foo
;
455 struct block_device
*bdev
= &ei
->bdev
;
457 memset(bdev
, 0, sizeof(*bdev
));
458 mutex_init(&bdev
->bd_mutex
);
459 INIT_LIST_HEAD(&bdev
->bd_inodes
);
460 INIT_LIST_HEAD(&bdev
->bd_list
);
462 INIT_LIST_HEAD(&bdev
->bd_holder_list
);
464 inode_init_once(&ei
->vfs_inode
);
465 /* Initialize mutex for freeze. */
466 mutex_init(&bdev
->bd_fsfreeze_mutex
);
469 static inline void __bd_forget(struct inode
*inode
)
471 list_del_init(&inode
->i_devices
);
472 inode
->i_bdev
= NULL
;
473 inode
->i_mapping
= &inode
->i_data
;
476 static void bdev_clear_inode(struct inode
*inode
)
478 struct block_device
*bdev
= &BDEV_I(inode
)->bdev
;
480 spin_lock(&bdev_lock
);
481 while ( (p
= bdev
->bd_inodes
.next
) != &bdev
->bd_inodes
) {
482 __bd_forget(list_entry(p
, struct inode
, i_devices
));
484 list_del_init(&bdev
->bd_list
);
485 spin_unlock(&bdev_lock
);
488 static const struct super_operations bdev_sops
= {
489 .statfs
= simple_statfs
,
490 .alloc_inode
= bdev_alloc_inode
,
491 .destroy_inode
= bdev_destroy_inode
,
492 .drop_inode
= generic_delete_inode
,
493 .clear_inode
= bdev_clear_inode
,
496 static int bd_get_sb(struct file_system_type
*fs_type
,
497 int flags
, const char *dev_name
, void *data
, struct vfsmount
*mnt
)
499 return get_sb_pseudo(fs_type
, "bdev:", &bdev_sops
, 0x62646576, mnt
);
502 static struct file_system_type bd_type
= {
505 .kill_sb
= kill_anon_super
,
508 struct super_block
*blockdev_superblock __read_mostly
;
510 void __init
bdev_cache_init(void)
513 struct vfsmount
*bd_mnt
;
515 bdev_cachep
= kmem_cache_create("bdev_cache", sizeof(struct bdev_inode
),
516 0, (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
517 SLAB_MEM_SPREAD
|SLAB_PANIC
),
519 err
= register_filesystem(&bd_type
);
521 panic("Cannot register bdev pseudo-fs");
522 bd_mnt
= kern_mount(&bd_type
);
524 panic("Cannot create bdev pseudo-fs");
526 * This vfsmount structure is only used to obtain the
527 * blockdev_superblock, so tell kmemleak not to report it.
529 kmemleak_not_leak(bd_mnt
);
530 blockdev_superblock
= bd_mnt
->mnt_sb
; /* For writeback */
534 * Most likely _very_ bad one - but then it's hardly critical for small
535 * /dev and can be fixed when somebody will need really large one.
536 * Keep in mind that it will be fed through icache hash function too.
538 static inline unsigned long hash(dev_t dev
)
540 return MAJOR(dev
)+MINOR(dev
);
543 static int bdev_test(struct inode
*inode
, void *data
)
545 return BDEV_I(inode
)->bdev
.bd_dev
== *(dev_t
*)data
;
548 static int bdev_set(struct inode
*inode
, void *data
)
550 BDEV_I(inode
)->bdev
.bd_dev
= *(dev_t
*)data
;
554 static LIST_HEAD(all_bdevs
);
556 struct block_device
*bdget(dev_t dev
)
558 struct block_device
*bdev
;
561 inode
= iget5_locked(blockdev_superblock
, hash(dev
),
562 bdev_test
, bdev_set
, &dev
);
567 bdev
= &BDEV_I(inode
)->bdev
;
569 if (inode
->i_state
& I_NEW
) {
570 bdev
->bd_contains
= NULL
;
571 bdev
->bd_inode
= inode
;
572 bdev
->bd_block_size
= (1 << inode
->i_blkbits
);
573 bdev
->bd_part_count
= 0;
574 bdev
->bd_invalidated
= 0;
575 inode
->i_mode
= S_IFBLK
;
577 inode
->i_bdev
= bdev
;
578 inode
->i_data
.a_ops
= &def_blk_aops
;
579 mapping_set_gfp_mask(&inode
->i_data
, GFP_USER
);
580 inode
->i_data
.backing_dev_info
= &default_backing_dev_info
;
581 spin_lock(&bdev_lock
);
582 list_add(&bdev
->bd_list
, &all_bdevs
);
583 spin_unlock(&bdev_lock
);
584 unlock_new_inode(inode
);
589 EXPORT_SYMBOL(bdget
);
592 * bdgrab -- Grab a reference to an already referenced block device
593 * @bdev: Block device to grab a reference to.
595 struct block_device
*bdgrab(struct block_device
*bdev
)
597 atomic_inc(&bdev
->bd_inode
->i_count
);
601 long nr_blockdev_pages(void)
603 struct block_device
*bdev
;
605 spin_lock(&bdev_lock
);
606 list_for_each_entry(bdev
, &all_bdevs
, bd_list
) {
607 ret
+= bdev
->bd_inode
->i_mapping
->nrpages
;
609 spin_unlock(&bdev_lock
);
613 void bdput(struct block_device
*bdev
)
615 iput(bdev
->bd_inode
);
618 EXPORT_SYMBOL(bdput
);
620 static struct block_device
*bd_acquire(struct inode
*inode
)
622 struct block_device
*bdev
;
624 spin_lock(&bdev_lock
);
625 bdev
= inode
->i_bdev
;
627 atomic_inc(&bdev
->bd_inode
->i_count
);
628 spin_unlock(&bdev_lock
);
631 spin_unlock(&bdev_lock
);
633 bdev
= bdget(inode
->i_rdev
);
635 spin_lock(&bdev_lock
);
636 if (!inode
->i_bdev
) {
638 * We take an additional bd_inode->i_count for inode,
639 * and it's released in clear_inode() of inode.
640 * So, we can access it via ->i_mapping always
643 atomic_inc(&bdev
->bd_inode
->i_count
);
644 inode
->i_bdev
= bdev
;
645 inode
->i_mapping
= bdev
->bd_inode
->i_mapping
;
646 list_add(&inode
->i_devices
, &bdev
->bd_inodes
);
648 spin_unlock(&bdev_lock
);
653 /* Call when you free inode */
655 void bd_forget(struct inode
*inode
)
657 struct block_device
*bdev
= NULL
;
659 spin_lock(&bdev_lock
);
661 if (!sb_is_blkdev_sb(inode
->i_sb
))
662 bdev
= inode
->i_bdev
;
665 spin_unlock(&bdev_lock
);
668 iput(bdev
->bd_inode
);
672 * bd_may_claim - test whether a block device can be claimed
673 * @bdev: block device of interest
674 * @whole: whole block device containing @bdev, may equal @bdev
675 * @holder: holder trying to claim @bdev
677 * Test whther @bdev can be claimed by @holder.
680 * spin_lock(&bdev_lock).
683 * %true if @bdev can be claimed, %false otherwise.
685 static bool bd_may_claim(struct block_device
*bdev
, struct block_device
*whole
,
688 if (bdev
->bd_holder
== holder
)
689 return true; /* already a holder */
690 else if (bdev
->bd_holder
!= NULL
)
691 return false; /* held by someone else */
692 else if (bdev
->bd_contains
== bdev
)
693 return true; /* is a whole device which isn't held */
695 else if (whole
->bd_holder
== bd_claim
)
696 return true; /* is a partition of a device that is being partitioned */
697 else if (whole
->bd_holder
!= NULL
)
698 return false; /* is a partition of a held device */
700 return true; /* is a partition of an un-held device */
704 * bd_prepare_to_claim - prepare to claim a block device
705 * @bdev: block device of interest
706 * @whole: the whole device containing @bdev, may equal @bdev
707 * @holder: holder trying to claim @bdev
709 * Prepare to claim @bdev. This function fails if @bdev is already
710 * claimed by another holder and waits if another claiming is in
711 * progress. This function doesn't actually claim. On successful
712 * return, the caller has ownership of bd_claiming and bd_holder[s].
715 * spin_lock(&bdev_lock). Might release bdev_lock, sleep and regrab
719 * 0 if @bdev can be claimed, -EBUSY otherwise.
721 static int bd_prepare_to_claim(struct block_device
*bdev
,
722 struct block_device
*whole
, void *holder
)
725 /* if someone else claimed, fail */
726 if (!bd_may_claim(bdev
, whole
, holder
))
729 /* if someone else is claiming, wait for it to finish */
730 if (whole
->bd_claiming
&& whole
->bd_claiming
!= holder
) {
731 wait_queue_head_t
*wq
= bit_waitqueue(&whole
->bd_claiming
, 0);
734 prepare_to_wait(wq
, &wait
, TASK_UNINTERRUPTIBLE
);
735 spin_unlock(&bdev_lock
);
737 finish_wait(wq
, &wait
);
738 spin_lock(&bdev_lock
);
747 * bd_start_claiming - start claiming a block device
748 * @bdev: block device of interest
749 * @holder: holder trying to claim @bdev
751 * @bdev is about to be opened exclusively. Check @bdev can be opened
752 * exclusively and mark that an exclusive open is in progress. Each
753 * successful call to this function must be matched with a call to
754 * either bd_claim() or bd_abort_claiming(). If this function
755 * succeeds, the matching bd_claim() is guaranteed to succeed.
761 * Pointer to the block device containing @bdev on success, ERR_PTR()
764 static struct block_device
*bd_start_claiming(struct block_device
*bdev
,
767 struct gendisk
*disk
;
768 struct block_device
*whole
;
774 * @bdev might not have been initialized properly yet, look up
775 * and grab the outer block device the hard way.
777 disk
= get_gendisk(bdev
->bd_dev
, &partno
);
779 return ERR_PTR(-ENXIO
);
781 whole
= bdget_disk(disk
, 0);
784 return ERR_PTR(-ENOMEM
);
786 /* prepare to claim, if successful, mark claiming in progress */
787 spin_lock(&bdev_lock
);
789 err
= bd_prepare_to_claim(bdev
, whole
, holder
);
791 whole
->bd_claiming
= holder
;
792 spin_unlock(&bdev_lock
);
795 spin_unlock(&bdev_lock
);
801 /* releases bdev_lock */
802 static void __bd_abort_claiming(struct block_device
*whole
, void *holder
)
804 BUG_ON(whole
->bd_claiming
!= holder
);
805 whole
->bd_claiming
= NULL
;
806 wake_up_bit(&whole
->bd_claiming
, 0);
808 spin_unlock(&bdev_lock
);
813 * bd_abort_claiming - abort claiming a block device
814 * @whole: whole block device returned by bd_start_claiming()
815 * @holder: holder trying to claim @bdev
817 * Abort a claiming block started by bd_start_claiming(). Note that
818 * @whole is not the block device to be claimed but the whole device
819 * returned by bd_start_claiming().
822 * Grabs and releases bdev_lock.
824 static void bd_abort_claiming(struct block_device
*whole
, void *holder
)
826 spin_lock(&bdev_lock
);
827 __bd_abort_claiming(whole
, holder
); /* releases bdev_lock */
831 * bd_claim - claim a block device
832 * @bdev: block device to claim
833 * @holder: holder trying to claim @bdev
835 * Try to claim @bdev which must have been opened successfully. This
836 * function may be called with or without preceding
837 * blk_start_claiming(). In the former case, this function is always
838 * successful and terminates the claiming block.
844 * 0 if successful, -EBUSY if @bdev is already claimed.
846 int bd_claim(struct block_device
*bdev
, void *holder
)
848 struct block_device
*whole
= bdev
->bd_contains
;
853 spin_lock(&bdev_lock
);
855 res
= bd_prepare_to_claim(bdev
, whole
, holder
);
857 /* note that for a whole device bd_holders
858 * will be incremented twice, and bd_holder will
859 * be set to bd_claim before being set to holder
862 whole
->bd_holder
= bd_claim
;
864 bdev
->bd_holder
= holder
;
867 if (whole
->bd_claiming
)
868 __bd_abort_claiming(whole
, holder
); /* releases bdev_lock */
870 spin_unlock(&bdev_lock
);
874 EXPORT_SYMBOL(bd_claim
);
876 void bd_release(struct block_device
*bdev
)
878 spin_lock(&bdev_lock
);
879 if (!--bdev
->bd_contains
->bd_holders
)
880 bdev
->bd_contains
->bd_holder
= NULL
;
881 if (!--bdev
->bd_holders
)
882 bdev
->bd_holder
= NULL
;
883 spin_unlock(&bdev_lock
);
886 EXPORT_SYMBOL(bd_release
);
890 * Functions for bd_claim_by_kobject / bd_release_from_kobject
892 * If a kobject is passed to bd_claim_by_kobject()
893 * and the kobject has a parent directory,
894 * following symlinks are created:
895 * o from the kobject to the claimed bdev
896 * o from "holders" directory of the bdev to the parent of the kobject
897 * bd_release_from_kobject() removes these symlinks.
900 * If /dev/dm-0 maps to /dev/sda, kobject corresponding to
901 * /sys/block/dm-0/slaves is passed to bd_claim_by_kobject(), then:
902 * /sys/block/dm-0/slaves/sda --> /sys/block/sda
903 * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0
906 static int add_symlink(struct kobject
*from
, struct kobject
*to
)
910 return sysfs_create_link(from
, to
, kobject_name(to
));
913 static void del_symlink(struct kobject
*from
, struct kobject
*to
)
917 sysfs_remove_link(from
, kobject_name(to
));
921 * 'struct bd_holder' contains pointers to kobjects symlinked by
922 * bd_claim_by_kobject.
923 * It's connected to bd_holder_list which is protected by bdev->bd_sem.
926 struct list_head list
; /* chain of holders of the bdev */
927 int count
; /* references from the holder */
928 struct kobject
*sdir
; /* holder object, e.g. "/block/dm-0/slaves" */
929 struct kobject
*hdev
; /* e.g. "/block/dm-0" */
930 struct kobject
*hdir
; /* e.g. "/block/sda/holders" */
931 struct kobject
*sdev
; /* e.g. "/block/sda" */
935 * Get references of related kobjects at once.
936 * Returns 1 on success. 0 on failure.
938 * Should call bd_holder_release_dirs() after successful use.
940 static int bd_holder_grab_dirs(struct block_device
*bdev
,
941 struct bd_holder
*bo
)
946 bo
->sdir
= kobject_get(bo
->sdir
);
950 bo
->hdev
= kobject_get(bo
->sdir
->parent
);
954 bo
->sdev
= kobject_get(&part_to_dev(bdev
->bd_part
)->kobj
);
958 bo
->hdir
= kobject_get(bdev
->bd_part
->holder_dir
);
965 kobject_put(bo
->sdev
);
967 kobject_put(bo
->hdev
);
969 kobject_put(bo
->sdir
);
974 /* Put references of related kobjects at once. */
975 static void bd_holder_release_dirs(struct bd_holder
*bo
)
977 kobject_put(bo
->hdir
);
978 kobject_put(bo
->sdev
);
979 kobject_put(bo
->hdev
);
980 kobject_put(bo
->sdir
);
983 static struct bd_holder
*alloc_bd_holder(struct kobject
*kobj
)
985 struct bd_holder
*bo
;
987 bo
= kzalloc(sizeof(*bo
), GFP_KERNEL
);
997 static void free_bd_holder(struct bd_holder
*bo
)
1003 * find_bd_holder - find matching struct bd_holder from the block device
1005 * @bdev: struct block device to be searched
1006 * @bo: target struct bd_holder
1008 * Returns matching entry with @bo in @bdev->bd_holder_list.
1009 * If found, increment the reference count and return the pointer.
1010 * If not found, returns NULL.
1012 static struct bd_holder
*find_bd_holder(struct block_device
*bdev
,
1013 struct bd_holder
*bo
)
1015 struct bd_holder
*tmp
;
1017 list_for_each_entry(tmp
, &bdev
->bd_holder_list
, list
)
1018 if (tmp
->sdir
== bo
->sdir
) {
1027 * add_bd_holder - create sysfs symlinks for bd_claim() relationship
1029 * @bdev: block device to be bd_claimed
1030 * @bo: preallocated and initialized by alloc_bd_holder()
1032 * Add @bo to @bdev->bd_holder_list, create symlinks.
1034 * Returns 0 if symlinks are created.
1035 * Returns -ve if something fails.
1037 static int add_bd_holder(struct block_device
*bdev
, struct bd_holder
*bo
)
1044 if (!bd_holder_grab_dirs(bdev
, bo
))
1047 err
= add_symlink(bo
->sdir
, bo
->sdev
);
1051 err
= add_symlink(bo
->hdir
, bo
->hdev
);
1053 del_symlink(bo
->sdir
, bo
->sdev
);
1057 list_add_tail(&bo
->list
, &bdev
->bd_holder_list
);
1062 * del_bd_holder - delete sysfs symlinks for bd_claim() relationship
1064 * @bdev: block device to be bd_claimed
1065 * @kobj: holder's kobject
1067 * If there is matching entry with @kobj in @bdev->bd_holder_list
1068 * and no other bd_claim() from the same kobject,
1069 * remove the struct bd_holder from the list, delete symlinks for it.
1071 * Returns a pointer to the struct bd_holder when it's removed from the list
1072 * and ready to be freed.
1073 * Returns NULL if matching claim isn't found or there is other bd_claim()
1074 * by the same kobject.
1076 static struct bd_holder
*del_bd_holder(struct block_device
*bdev
,
1077 struct kobject
*kobj
)
1079 struct bd_holder
*bo
;
1081 list_for_each_entry(bo
, &bdev
->bd_holder_list
, list
) {
1082 if (bo
->sdir
== kobj
) {
1084 BUG_ON(bo
->count
< 0);
1086 list_del(&bo
->list
);
1087 del_symlink(bo
->sdir
, bo
->sdev
);
1088 del_symlink(bo
->hdir
, bo
->hdev
);
1089 bd_holder_release_dirs(bo
);
1100 * bd_claim_by_kobject - bd_claim() with additional kobject signature
1102 * @bdev: block device to be claimed
1103 * @holder: holder's signature
1104 * @kobj: holder's kobject
1106 * Do bd_claim() and if it succeeds, create sysfs symlinks between
1107 * the bdev and the holder's kobject.
1108 * Use bd_release_from_kobject() when relesing the claimed bdev.
1110 * Returns 0 on success. (same as bd_claim())
1111 * Returns errno on failure.
1113 static int bd_claim_by_kobject(struct block_device
*bdev
, void *holder
,
1114 struct kobject
*kobj
)
1117 struct bd_holder
*bo
, *found
;
1122 bo
= alloc_bd_holder(kobj
);
1126 mutex_lock(&bdev
->bd_mutex
);
1128 err
= bd_claim(bdev
, holder
);
1132 found
= find_bd_holder(bdev
, bo
);
1136 err
= add_bd_holder(bdev
, bo
);
1142 mutex_unlock(&bdev
->bd_mutex
);
1148 * bd_release_from_kobject - bd_release() with additional kobject signature
1150 * @bdev: block device to be released
1151 * @kobj: holder's kobject
1153 * Do bd_release() and remove sysfs symlinks created by bd_claim_by_kobject().
1155 static void bd_release_from_kobject(struct block_device
*bdev
,
1156 struct kobject
*kobj
)
1161 mutex_lock(&bdev
->bd_mutex
);
1163 free_bd_holder(del_bd_holder(bdev
, kobj
));
1164 mutex_unlock(&bdev
->bd_mutex
);
1168 * bd_claim_by_disk - wrapper function for bd_claim_by_kobject()
1170 * @bdev: block device to be claimed
1171 * @holder: holder's signature
1172 * @disk: holder's gendisk
1174 * Call bd_claim_by_kobject() with getting @disk->slave_dir.
1176 int bd_claim_by_disk(struct block_device
*bdev
, void *holder
,
1177 struct gendisk
*disk
)
1179 return bd_claim_by_kobject(bdev
, holder
, kobject_get(disk
->slave_dir
));
1181 EXPORT_SYMBOL_GPL(bd_claim_by_disk
);
1184 * bd_release_from_disk - wrapper function for bd_release_from_kobject()
1186 * @bdev: block device to be claimed
1187 * @disk: holder's gendisk
1189 * Call bd_release_from_kobject() and put @disk->slave_dir.
1191 void bd_release_from_disk(struct block_device
*bdev
, struct gendisk
*disk
)
1193 bd_release_from_kobject(bdev
, disk
->slave_dir
);
1194 kobject_put(disk
->slave_dir
);
1196 EXPORT_SYMBOL_GPL(bd_release_from_disk
);
1200 * Tries to open block device by device number. Use it ONLY if you
1201 * really do not have anything better - i.e. when you are behind a
1202 * truly sucky interface and all you are given is a device number. _Never_
1203 * to be used for internal purposes. If you ever need it - reconsider
1206 struct block_device
*open_by_devnum(dev_t dev
, fmode_t mode
)
1208 struct block_device
*bdev
= bdget(dev
);
1211 err
= blkdev_get(bdev
, mode
);
1212 return err
? ERR_PTR(err
) : bdev
;
1215 EXPORT_SYMBOL(open_by_devnum
);
1218 * flush_disk - invalidates all buffer-cache entries on a disk
1220 * @bdev: struct block device to be flushed
1222 * Invalidates all buffer-cache entries on a disk. It should be called
1223 * when a disk has been changed -- either by a media change or online
1226 static void flush_disk(struct block_device
*bdev
)
1228 if (__invalidate_device(bdev
)) {
1229 char name
[BDEVNAME_SIZE
] = "";
1232 disk_name(bdev
->bd_disk
, 0, name
);
1233 printk(KERN_WARNING
"VFS: busy inodes on changed media or "
1234 "resized disk %s\n", name
);
1239 if (disk_partitionable(bdev
->bd_disk
))
1240 bdev
->bd_invalidated
= 1;
1244 * check_disk_size_change - checks for disk size change and adjusts bdev size.
1245 * @disk: struct gendisk to check
1246 * @bdev: struct bdev to adjust.
1248 * This routine checks to see if the bdev size does not match the disk size
1249 * and adjusts it if it differs.
1251 void check_disk_size_change(struct gendisk
*disk
, struct block_device
*bdev
)
1253 loff_t disk_size
, bdev_size
;
1255 disk_size
= (loff_t
)get_capacity(disk
) << 9;
1256 bdev_size
= i_size_read(bdev
->bd_inode
);
1257 if (disk_size
!= bdev_size
) {
1258 char name
[BDEVNAME_SIZE
];
1260 disk_name(disk
, 0, name
);
1262 "%s: detected capacity change from %lld to %lld\n",
1263 name
, bdev_size
, disk_size
);
1264 i_size_write(bdev
->bd_inode
, disk_size
);
1268 EXPORT_SYMBOL(check_disk_size_change
);
1271 * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back
1272 * @disk: struct gendisk to be revalidated
1274 * This routine is a wrapper for lower-level driver's revalidate_disk
1275 * call-backs. It is used to do common pre and post operations needed
1276 * for all revalidate_disk operations.
1278 int revalidate_disk(struct gendisk
*disk
)
1280 struct block_device
*bdev
;
1283 if (disk
->fops
->revalidate_disk
)
1284 ret
= disk
->fops
->revalidate_disk(disk
);
1286 bdev
= bdget_disk(disk
, 0);
1290 mutex_lock(&bdev
->bd_mutex
);
1291 check_disk_size_change(disk
, bdev
);
1292 mutex_unlock(&bdev
->bd_mutex
);
1296 EXPORT_SYMBOL(revalidate_disk
);
1299 * This routine checks whether a removable media has been changed,
1300 * and invalidates all buffer-cache-entries in that case. This
1301 * is a relatively slow routine, so we have to try to minimize using
1302 * it. Thus it is called only upon a 'mount' or 'open'. This
1303 * is the best way of combining speed and utility, I think.
1304 * People changing diskettes in the middle of an operation deserve
1307 int check_disk_change(struct block_device
*bdev
)
1309 struct gendisk
*disk
= bdev
->bd_disk
;
1310 const struct block_device_operations
*bdops
= disk
->fops
;
1312 if (!bdops
->media_changed
)
1314 if (!bdops
->media_changed(bdev
->bd_disk
))
1318 if (bdops
->revalidate_disk
)
1319 bdops
->revalidate_disk(bdev
->bd_disk
);
1323 EXPORT_SYMBOL(check_disk_change
);
1325 void bd_set_size(struct block_device
*bdev
, loff_t size
)
1327 unsigned bsize
= bdev_logical_block_size(bdev
);
1329 bdev
->bd_inode
->i_size
= size
;
1330 while (bsize
< PAGE_CACHE_SIZE
) {
1335 bdev
->bd_block_size
= bsize
;
1336 bdev
->bd_inode
->i_blkbits
= blksize_bits(bsize
);
1338 EXPORT_SYMBOL(bd_set_size
);
1340 static int __blkdev_put(struct block_device
*bdev
, fmode_t mode
, int for_part
);
1345 * mutex_lock(part->bd_mutex)
1346 * mutex_lock_nested(whole->bd_mutex, 1)
1349 static int __blkdev_get(struct block_device
*bdev
, fmode_t mode
, int for_part
)
1351 struct gendisk
*disk
;
1356 if (mode
& FMODE_READ
)
1358 if (mode
& FMODE_WRITE
)
1361 * hooks: /n/, see "layering violations".
1363 ret
= devcgroup_inode_permission(bdev
->bd_inode
, perm
);
1373 disk
= get_gendisk(bdev
->bd_dev
, &partno
);
1375 goto out_unlock_kernel
;
1377 mutex_lock_nested(&bdev
->bd_mutex
, for_part
);
1378 if (!bdev
->bd_openers
) {
1379 bdev
->bd_disk
= disk
;
1380 bdev
->bd_contains
= bdev
;
1382 struct backing_dev_info
*bdi
;
1385 bdev
->bd_part
= disk_get_part(disk
, partno
);
1389 if (disk
->fops
->open
) {
1390 ret
= disk
->fops
->open(bdev
, mode
);
1391 if (ret
== -ERESTARTSYS
) {
1392 /* Lost a race with 'disk' being
1393 * deleted, try again.
1396 disk_put_part(bdev
->bd_part
);
1397 bdev
->bd_part
= NULL
;
1398 module_put(disk
->fops
->owner
);
1400 bdev
->bd_disk
= NULL
;
1401 mutex_unlock(&bdev
->bd_mutex
);
1407 if (!bdev
->bd_openers
) {
1408 bd_set_size(bdev
,(loff_t
)get_capacity(disk
)<<9);
1409 bdi
= blk_get_backing_dev_info(bdev
);
1411 bdi
= &default_backing_dev_info
;
1412 bdev
->bd_inode
->i_data
.backing_dev_info
= bdi
;
1414 if (bdev
->bd_invalidated
)
1415 rescan_partitions(disk
, bdev
);
1417 struct block_device
*whole
;
1418 whole
= bdget_disk(disk
, 0);
1423 ret
= __blkdev_get(whole
, mode
, 1);
1426 bdev
->bd_contains
= whole
;
1427 bdev
->bd_inode
->i_data
.backing_dev_info
=
1428 whole
->bd_inode
->i_data
.backing_dev_info
;
1429 bdev
->bd_part
= disk_get_part(disk
, partno
);
1430 if (!(disk
->flags
& GENHD_FL_UP
) ||
1431 !bdev
->bd_part
|| !bdev
->bd_part
->nr_sects
) {
1435 bd_set_size(bdev
, (loff_t
)bdev
->bd_part
->nr_sects
<< 9);
1438 module_put(disk
->fops
->owner
);
1441 if (bdev
->bd_contains
== bdev
) {
1442 if (bdev
->bd_disk
->fops
->open
) {
1443 ret
= bdev
->bd_disk
->fops
->open(bdev
, mode
);
1445 goto out_unlock_bdev
;
1447 if (bdev
->bd_invalidated
)
1448 rescan_partitions(bdev
->bd_disk
, bdev
);
1453 bdev
->bd_part_count
++;
1454 mutex_unlock(&bdev
->bd_mutex
);
1459 disk_put_part(bdev
->bd_part
);
1460 bdev
->bd_disk
= NULL
;
1461 bdev
->bd_part
= NULL
;
1462 bdev
->bd_inode
->i_data
.backing_dev_info
= &default_backing_dev_info
;
1463 if (bdev
!= bdev
->bd_contains
)
1464 __blkdev_put(bdev
->bd_contains
, mode
, 1);
1465 bdev
->bd_contains
= NULL
;
1467 mutex_unlock(&bdev
->bd_mutex
);
1472 module_put(disk
->fops
->owner
);
1479 int blkdev_get(struct block_device
*bdev
, fmode_t mode
)
1481 return __blkdev_get(bdev
, mode
, 0);
1483 EXPORT_SYMBOL(blkdev_get
);
1485 static int blkdev_open(struct inode
* inode
, struct file
* filp
)
1487 struct block_device
*whole
= NULL
;
1488 struct block_device
*bdev
;
1492 * Preserve backwards compatibility and allow large file access
1493 * even if userspace doesn't ask for it explicitly. Some mkfs
1494 * binary needs it. We might want to drop this workaround
1495 * during an unstable branch.
1497 filp
->f_flags
|= O_LARGEFILE
;
1499 if (filp
->f_flags
& O_NDELAY
)
1500 filp
->f_mode
|= FMODE_NDELAY
;
1501 if (filp
->f_flags
& O_EXCL
)
1502 filp
->f_mode
|= FMODE_EXCL
;
1503 if ((filp
->f_flags
& O_ACCMODE
) == 3)
1504 filp
->f_mode
|= FMODE_WRITE_IOCTL
;
1506 bdev
= bd_acquire(inode
);
1510 if (filp
->f_mode
& FMODE_EXCL
) {
1511 whole
= bd_start_claiming(bdev
, filp
);
1512 if (IS_ERR(whole
)) {
1514 return PTR_ERR(whole
);
1518 filp
->f_mapping
= bdev
->bd_inode
->i_mapping
;
1520 res
= blkdev_get(bdev
, filp
->f_mode
);
1524 BUG_ON(bd_claim(bdev
, filp
) != 0);
1526 bd_abort_claiming(whole
, filp
);
1532 static int __blkdev_put(struct block_device
*bdev
, fmode_t mode
, int for_part
)
1535 struct gendisk
*disk
= bdev
->bd_disk
;
1536 struct block_device
*victim
= NULL
;
1538 mutex_lock_nested(&bdev
->bd_mutex
, for_part
);
1541 bdev
->bd_part_count
--;
1543 if (!--bdev
->bd_openers
) {
1544 sync_blockdev(bdev
);
1547 if (bdev
->bd_contains
== bdev
) {
1548 if (disk
->fops
->release
)
1549 ret
= disk
->fops
->release(disk
, mode
);
1551 if (!bdev
->bd_openers
) {
1552 struct module
*owner
= disk
->fops
->owner
;
1556 disk_put_part(bdev
->bd_part
);
1557 bdev
->bd_part
= NULL
;
1558 bdev
->bd_disk
= NULL
;
1559 bdev
->bd_inode
->i_data
.backing_dev_info
= &default_backing_dev_info
;
1560 if (bdev
!= bdev
->bd_contains
)
1561 victim
= bdev
->bd_contains
;
1562 bdev
->bd_contains
= NULL
;
1565 mutex_unlock(&bdev
->bd_mutex
);
1568 __blkdev_put(victim
, mode
, 1);
1572 int blkdev_put(struct block_device
*bdev
, fmode_t mode
)
1574 return __blkdev_put(bdev
, mode
, 0);
1576 EXPORT_SYMBOL(blkdev_put
);
1578 static int blkdev_close(struct inode
* inode
, struct file
* filp
)
1580 struct block_device
*bdev
= I_BDEV(filp
->f_mapping
->host
);
1581 if (bdev
->bd_holder
== filp
)
1583 return blkdev_put(bdev
, filp
->f_mode
);
1586 static long block_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1588 struct block_device
*bdev
= I_BDEV(file
->f_mapping
->host
);
1589 fmode_t mode
= file
->f_mode
;
1592 * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have
1593 * to updated it before every ioctl.
1595 if (file
->f_flags
& O_NDELAY
)
1596 mode
|= FMODE_NDELAY
;
1598 mode
&= ~FMODE_NDELAY
;
1600 return blkdev_ioctl(bdev
, mode
, cmd
, arg
);
1604 * Write data to the block device. Only intended for the block device itself
1605 * and the raw driver which basically is a fake block device.
1607 * Does not take i_mutex for the write and thus is not for general purpose
1610 ssize_t
blkdev_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
1611 unsigned long nr_segs
, loff_t pos
)
1613 struct file
*file
= iocb
->ki_filp
;
1616 BUG_ON(iocb
->ki_pos
!= pos
);
1618 ret
= __generic_file_aio_write(iocb
, iov
, nr_segs
, &iocb
->ki_pos
);
1619 if (ret
> 0 || ret
== -EIOCBQUEUED
) {
1622 err
= generic_write_sync(file
, pos
, ret
);
1623 if (err
< 0 && ret
> 0)
1628 EXPORT_SYMBOL_GPL(blkdev_aio_write
);
1631 * Try to release a page associated with block device when the system
1632 * is under memory pressure.
1634 static int blkdev_releasepage(struct page
*page
, gfp_t wait
)
1636 struct super_block
*super
= BDEV_I(page
->mapping
->host
)->bdev
.bd_super
;
1638 if (super
&& super
->s_op
->bdev_try_to_free_page
)
1639 return super
->s_op
->bdev_try_to_free_page(super
, page
, wait
);
1641 return try_to_free_buffers(page
);
1644 static const struct address_space_operations def_blk_aops
= {
1645 .readpage
= blkdev_readpage
,
1646 .writepage
= blkdev_writepage
,
1647 .sync_page
= block_sync_page
,
1648 .write_begin
= blkdev_write_begin
,
1649 .write_end
= blkdev_write_end
,
1650 .writepages
= generic_writepages
,
1651 .releasepage
= blkdev_releasepage
,
1652 .direct_IO
= blkdev_direct_IO
,
1655 const struct file_operations def_blk_fops
= {
1656 .open
= blkdev_open
,
1657 .release
= blkdev_close
,
1658 .llseek
= block_llseek
,
1659 .read
= do_sync_read
,
1660 .write
= do_sync_write
,
1661 .aio_read
= generic_file_aio_read
,
1662 .aio_write
= blkdev_aio_write
,
1663 .mmap
= generic_file_mmap
,
1664 .fsync
= blkdev_fsync
,
1665 .unlocked_ioctl
= block_ioctl
,
1666 #ifdef CONFIG_COMPAT
1667 .compat_ioctl
= compat_blkdev_ioctl
,
1669 .splice_read
= generic_file_splice_read
,
1670 .splice_write
= generic_file_splice_write
,
1673 int ioctl_by_bdev(struct block_device
*bdev
, unsigned cmd
, unsigned long arg
)
1676 mm_segment_t old_fs
= get_fs();
1678 res
= blkdev_ioctl(bdev
, 0, cmd
, arg
);
1683 EXPORT_SYMBOL(ioctl_by_bdev
);
1686 * lookup_bdev - lookup a struct block_device by name
1687 * @pathname: special file representing the block device
1689 * Get a reference to the blockdevice at @pathname in the current
1690 * namespace if possible and return it. Return ERR_PTR(error)
1693 struct block_device
*lookup_bdev(const char *pathname
)
1695 struct block_device
*bdev
;
1696 struct inode
*inode
;
1700 if (!pathname
|| !*pathname
)
1701 return ERR_PTR(-EINVAL
);
1703 error
= kern_path(pathname
, LOOKUP_FOLLOW
, &path
);
1705 return ERR_PTR(error
);
1707 inode
= path
.dentry
->d_inode
;
1709 if (!S_ISBLK(inode
->i_mode
))
1712 if (path
.mnt
->mnt_flags
& MNT_NODEV
)
1715 bdev
= bd_acquire(inode
);
1722 bdev
= ERR_PTR(error
);
1725 EXPORT_SYMBOL(lookup_bdev
);
1728 * open_bdev_exclusive - open a block device by name and set it up for use
1730 * @path: special file representing the block device
1731 * @mode: FMODE_... combination to pass be used
1732 * @holder: owner for exclusion
1734 * Open the blockdevice described by the special file at @path, claim it
1737 struct block_device
*open_bdev_exclusive(const char *path
, fmode_t mode
, void *holder
)
1739 struct block_device
*bdev
, *whole
;
1742 bdev
= lookup_bdev(path
);
1746 whole
= bd_start_claiming(bdev
, holder
);
1747 if (IS_ERR(whole
)) {
1752 error
= blkdev_get(bdev
, mode
);
1754 goto out_abort_claiming
;
1757 if ((mode
& FMODE_WRITE
) && bdev_read_only(bdev
))
1758 goto out_blkdev_put
;
1760 BUG_ON(bd_claim(bdev
, holder
) != 0);
1764 blkdev_put(bdev
, mode
);
1766 bd_abort_claiming(whole
, holder
);
1767 return ERR_PTR(error
);
1770 EXPORT_SYMBOL(open_bdev_exclusive
);
1773 * close_bdev_exclusive - close a blockdevice opened by open_bdev_exclusive()
1775 * @bdev: blockdevice to close
1776 * @mode: mode, must match that used to open.
1778 * This is the counterpart to open_bdev_exclusive().
1780 void close_bdev_exclusive(struct block_device
*bdev
, fmode_t mode
)
1783 blkdev_put(bdev
, mode
);
1786 EXPORT_SYMBOL(close_bdev_exclusive
);
1788 int __invalidate_device(struct block_device
*bdev
)
1790 struct super_block
*sb
= get_super(bdev
);
1795 * no need to lock the super, get_super holds the
1796 * read mutex so the filesystem cannot go away
1797 * under us (->put_super runs with the write lock
1800 shrink_dcache_sb(sb
);
1801 res
= invalidate_inodes(sb
);
1804 invalidate_bdev(bdev
);
1807 EXPORT_SYMBOL(__invalidate_device
);