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/device_cgroup.h>
15 #include <linux/highmem.h>
16 #include <linux/blkdev.h>
17 #include <linux/module.h>
18 #include <linux/blkpg.h>
19 #include <linux/magic.h>
20 #include <linux/buffer_head.h>
21 #include <linux/swap.h>
22 #include <linux/pagevec.h>
23 #include <linux/writeback.h>
24 #include <linux/mpage.h>
25 #include <linux/mount.h>
26 #include <linux/uio.h>
27 #include <linux/namei.h>
28 #include <linux/log2.h>
29 #include <linux/cleancache.h>
30 #include <linux/aio.h>
31 #include <asm/uaccess.h>
35 struct block_device bdev
;
36 struct inode vfs_inode
;
39 static const struct address_space_operations def_blk_aops
;
41 static inline struct bdev_inode
*BDEV_I(struct inode
*inode
)
43 return container_of(inode
, struct bdev_inode
, vfs_inode
);
46 inline struct block_device
*I_BDEV(struct inode
*inode
)
48 return &BDEV_I(inode
)->bdev
;
50 EXPORT_SYMBOL(I_BDEV
);
52 static void bdev_write_inode(struct inode
*inode
)
54 spin_lock(&inode
->i_lock
);
55 while (inode
->i_state
& I_DIRTY
) {
56 spin_unlock(&inode
->i_lock
);
57 WARN_ON_ONCE(write_inode_now(inode
, true));
58 spin_lock(&inode
->i_lock
);
60 spin_unlock(&inode
->i_lock
);
63 /* Kill _all_ buffers and pagecache , dirty or not.. */
64 void kill_bdev(struct block_device
*bdev
)
66 struct address_space
*mapping
= bdev
->bd_inode
->i_mapping
;
68 if (mapping
->nrpages
== 0 && mapping
->nrshadows
== 0)
72 truncate_inode_pages(mapping
, 0);
74 EXPORT_SYMBOL(kill_bdev
);
76 /* Invalidate clean unused buffers and pagecache. */
77 void invalidate_bdev(struct block_device
*bdev
)
79 struct address_space
*mapping
= bdev
->bd_inode
->i_mapping
;
81 if (mapping
->nrpages
== 0)
85 lru_add_drain_all(); /* make sure all lru add caches are flushed */
86 invalidate_mapping_pages(mapping
, 0, -1);
87 /* 99% of the time, we don't need to flush the cleancache on the bdev.
88 * But, for the strange corners, lets be cautious
90 cleancache_invalidate_inode(mapping
);
92 EXPORT_SYMBOL(invalidate_bdev
);
94 int set_blocksize(struct block_device
*bdev
, int size
)
96 /* Size must be a power of two, and between 512 and PAGE_SIZE */
97 if (size
> PAGE_SIZE
|| size
< 512 || !is_power_of_2(size
))
100 /* Size cannot be smaller than the size supported by the device */
101 if (size
< bdev_logical_block_size(bdev
))
104 /* Don't change the size if it is same as current */
105 if (bdev
->bd_block_size
!= size
) {
107 bdev
->bd_block_size
= size
;
108 bdev
->bd_inode
->i_blkbits
= blksize_bits(size
);
114 EXPORT_SYMBOL(set_blocksize
);
116 int sb_set_blocksize(struct super_block
*sb
, int size
)
118 if (set_blocksize(sb
->s_bdev
, size
))
120 /* If we get here, we know size is power of two
121 * and it's value is between 512 and PAGE_SIZE */
122 sb
->s_blocksize
= size
;
123 sb
->s_blocksize_bits
= blksize_bits(size
);
124 return sb
->s_blocksize
;
127 EXPORT_SYMBOL(sb_set_blocksize
);
129 int sb_min_blocksize(struct super_block
*sb
, int size
)
131 int minsize
= bdev_logical_block_size(sb
->s_bdev
);
134 return sb_set_blocksize(sb
, size
);
137 EXPORT_SYMBOL(sb_min_blocksize
);
140 blkdev_get_block(struct inode
*inode
, sector_t iblock
,
141 struct buffer_head
*bh
, int create
)
143 bh
->b_bdev
= I_BDEV(inode
);
144 bh
->b_blocknr
= iblock
;
145 set_buffer_mapped(bh
);
150 blkdev_direct_IO(int rw
, struct kiocb
*iocb
, struct iov_iter
*iter
,
153 struct file
*file
= iocb
->ki_filp
;
154 struct inode
*inode
= file
->f_mapping
->host
;
156 return __blockdev_direct_IO(rw
, iocb
, inode
, I_BDEV(inode
), iter
,
157 offset
, blkdev_get_block
,
161 int __sync_blockdev(struct block_device
*bdev
, int wait
)
166 return filemap_flush(bdev
->bd_inode
->i_mapping
);
167 return filemap_write_and_wait(bdev
->bd_inode
->i_mapping
);
171 * Write out and wait upon all the dirty data associated with a block
172 * device via its mapping. Does not take the superblock lock.
174 int sync_blockdev(struct block_device
*bdev
)
176 return __sync_blockdev(bdev
, 1);
178 EXPORT_SYMBOL(sync_blockdev
);
181 * Write out and wait upon all dirty data associated with this
182 * device. Filesystem data as well as the underlying block
183 * device. Takes the superblock lock.
185 int fsync_bdev(struct block_device
*bdev
)
187 struct super_block
*sb
= get_super(bdev
);
189 int res
= sync_filesystem(sb
);
193 return sync_blockdev(bdev
);
195 EXPORT_SYMBOL(fsync_bdev
);
198 * freeze_bdev -- lock a filesystem and force it into a consistent state
199 * @bdev: blockdevice to lock
201 * If a superblock is found on this device, we take the s_umount semaphore
202 * on it to make sure nobody unmounts until the snapshot creation is done.
203 * The reference counter (bd_fsfreeze_count) guarantees that only the last
204 * unfreeze process can unfreeze the frozen filesystem actually when multiple
205 * freeze requests arrive simultaneously. It counts up in freeze_bdev() and
206 * count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze
209 struct super_block
*freeze_bdev(struct block_device
*bdev
)
211 struct super_block
*sb
;
214 mutex_lock(&bdev
->bd_fsfreeze_mutex
);
215 if (++bdev
->bd_fsfreeze_count
> 1) {
217 * We don't even need to grab a reference - the first call
218 * to freeze_bdev grab an active reference and only the last
219 * thaw_bdev drops it.
221 sb
= get_super(bdev
);
223 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
227 sb
= get_active_super(bdev
);
230 if (sb
->s_op
->freeze_super
)
231 error
= sb
->s_op
->freeze_super(sb
);
233 error
= freeze_super(sb
);
235 deactivate_super(sb
);
236 bdev
->bd_fsfreeze_count
--;
237 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
238 return ERR_PTR(error
);
240 deactivate_super(sb
);
243 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
244 return sb
; /* thaw_bdev releases s->s_umount */
246 EXPORT_SYMBOL(freeze_bdev
);
249 * thaw_bdev -- unlock filesystem
250 * @bdev: blockdevice to unlock
251 * @sb: associated superblock
253 * Unlocks the filesystem and marks it writeable again after freeze_bdev().
255 int thaw_bdev(struct block_device
*bdev
, struct super_block
*sb
)
259 mutex_lock(&bdev
->bd_fsfreeze_mutex
);
260 if (!bdev
->bd_fsfreeze_count
)
264 if (--bdev
->bd_fsfreeze_count
> 0)
270 if (sb
->s_op
->thaw_super
)
271 error
= sb
->s_op
->thaw_super(sb
);
273 error
= thaw_super(sb
);
275 bdev
->bd_fsfreeze_count
++;
276 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
280 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
283 EXPORT_SYMBOL(thaw_bdev
);
285 static int blkdev_writepage(struct page
*page
, struct writeback_control
*wbc
)
287 return block_write_full_page(page
, blkdev_get_block
, wbc
);
290 static int blkdev_readpage(struct file
* file
, struct page
* page
)
292 return block_read_full_page(page
, blkdev_get_block
);
295 static int blkdev_readpages(struct file
*file
, struct address_space
*mapping
,
296 struct list_head
*pages
, unsigned nr_pages
)
298 return mpage_readpages(mapping
, pages
, nr_pages
, blkdev_get_block
);
301 static int blkdev_write_begin(struct file
*file
, struct address_space
*mapping
,
302 loff_t pos
, unsigned len
, unsigned flags
,
303 struct page
**pagep
, void **fsdata
)
305 return block_write_begin(mapping
, pos
, len
, flags
, pagep
,
309 static int blkdev_write_end(struct file
*file
, struct address_space
*mapping
,
310 loff_t pos
, unsigned len
, unsigned copied
,
311 struct page
*page
, void *fsdata
)
314 ret
= block_write_end(file
, mapping
, pos
, len
, copied
, page
, fsdata
);
317 page_cache_release(page
);
324 * for a block special file file_inode(file)->i_size is zero
325 * so we compute the size by hand (just as in block_read/write above)
327 static loff_t
block_llseek(struct file
*file
, loff_t offset
, int whence
)
329 struct inode
*bd_inode
= file
->f_mapping
->host
;
332 mutex_lock(&bd_inode
->i_mutex
);
333 retval
= fixed_size_llseek(file
, offset
, whence
, i_size_read(bd_inode
));
334 mutex_unlock(&bd_inode
->i_mutex
);
338 int blkdev_fsync(struct file
*filp
, loff_t start
, loff_t end
, int datasync
)
340 struct inode
*bd_inode
= filp
->f_mapping
->host
;
341 struct block_device
*bdev
= I_BDEV(bd_inode
);
344 error
= filemap_write_and_wait_range(filp
->f_mapping
, start
, end
);
349 * There is no need to serialise calls to blkdev_issue_flush with
350 * i_mutex and doing so causes performance issues with concurrent
351 * O_SYNC writers to a block device.
353 error
= blkdev_issue_flush(bdev
, GFP_KERNEL
, NULL
);
354 if (error
== -EOPNOTSUPP
)
359 EXPORT_SYMBOL(blkdev_fsync
);
362 * bdev_read_page() - Start reading a page from a block device
363 * @bdev: The device to read the page from
364 * @sector: The offset on the device to read the page to (need not be aligned)
365 * @page: The page to read
367 * On entry, the page should be locked. It will be unlocked when the page
368 * has been read. If the block driver implements rw_page synchronously,
369 * that will be true on exit from this function, but it need not be.
371 * Errors returned by this function are usually "soft", eg out of memory, or
372 * queue full; callers should try a different route to read this page rather
373 * than propagate an error back up the stack.
375 * Return: negative errno if an error occurs, 0 if submission was successful.
377 int bdev_read_page(struct block_device
*bdev
, sector_t sector
,
380 const struct block_device_operations
*ops
= bdev
->bd_disk
->fops
;
383 return ops
->rw_page(bdev
, sector
+ get_start_sect(bdev
), page
, READ
);
385 EXPORT_SYMBOL_GPL(bdev_read_page
);
388 * bdev_write_page() - Start writing a page to a block device
389 * @bdev: The device to write the page to
390 * @sector: The offset on the device to write the page to (need not be aligned)
391 * @page: The page to write
392 * @wbc: The writeback_control for the write
394 * On entry, the page should be locked and not currently under writeback.
395 * On exit, if the write started successfully, the page will be unlocked and
396 * under writeback. If the write failed already (eg the driver failed to
397 * queue the page to the device), the page will still be locked. If the
398 * caller is a ->writepage implementation, it will need to unlock the page.
400 * Errors returned by this function are usually "soft", eg out of memory, or
401 * queue full; callers should try a different route to write this page rather
402 * than propagate an error back up the stack.
404 * Return: negative errno if an error occurs, 0 if submission was successful.
406 int bdev_write_page(struct block_device
*bdev
, sector_t sector
,
407 struct page
*page
, struct writeback_control
*wbc
)
410 int rw
= (wbc
->sync_mode
== WB_SYNC_ALL
) ? WRITE_SYNC
: WRITE
;
411 const struct block_device_operations
*ops
= bdev
->bd_disk
->fops
;
414 set_page_writeback(page
);
415 result
= ops
->rw_page(bdev
, sector
+ get_start_sect(bdev
), page
, rw
);
417 end_page_writeback(page
);
422 EXPORT_SYMBOL_GPL(bdev_write_page
);
428 static __cacheline_aligned_in_smp
DEFINE_SPINLOCK(bdev_lock
);
429 static struct kmem_cache
* bdev_cachep __read_mostly
;
431 static struct inode
*bdev_alloc_inode(struct super_block
*sb
)
433 struct bdev_inode
*ei
= kmem_cache_alloc(bdev_cachep
, GFP_KERNEL
);
436 return &ei
->vfs_inode
;
439 static void bdev_i_callback(struct rcu_head
*head
)
441 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
442 struct bdev_inode
*bdi
= BDEV_I(inode
);
444 kmem_cache_free(bdev_cachep
, bdi
);
447 static void bdev_destroy_inode(struct inode
*inode
)
449 call_rcu(&inode
->i_rcu
, bdev_i_callback
);
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_disks
);
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_evict_inode(struct inode
*inode
)
478 struct block_device
*bdev
= &BDEV_I(inode
)->bdev
;
480 truncate_inode_pages_final(&inode
->i_data
);
481 invalidate_inode_buffers(inode
); /* is it needed here? */
483 spin_lock(&bdev_lock
);
484 while ( (p
= bdev
->bd_inodes
.next
) != &bdev
->bd_inodes
) {
485 __bd_forget(list_entry(p
, struct inode
, i_devices
));
487 list_del_init(&bdev
->bd_list
);
488 spin_unlock(&bdev_lock
);
491 static const struct super_operations bdev_sops
= {
492 .statfs
= simple_statfs
,
493 .alloc_inode
= bdev_alloc_inode
,
494 .destroy_inode
= bdev_destroy_inode
,
495 .drop_inode
= generic_delete_inode
,
496 .evict_inode
= bdev_evict_inode
,
499 static struct dentry
*bd_mount(struct file_system_type
*fs_type
,
500 int flags
, const char *dev_name
, void *data
)
502 return mount_pseudo(fs_type
, "bdev:", &bdev_sops
, NULL
, BDEVFS_MAGIC
);
505 static struct file_system_type bd_type
= {
508 .kill_sb
= kill_anon_super
,
511 static struct super_block
*blockdev_superblock __read_mostly
;
513 void __init
bdev_cache_init(void)
516 static struct vfsmount
*bd_mnt
;
518 bdev_cachep
= kmem_cache_create("bdev_cache", sizeof(struct bdev_inode
),
519 0, (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
520 SLAB_MEM_SPREAD
|SLAB_PANIC
),
522 err
= register_filesystem(&bd_type
);
524 panic("Cannot register bdev pseudo-fs");
525 bd_mnt
= kern_mount(&bd_type
);
527 panic("Cannot create bdev pseudo-fs");
528 blockdev_superblock
= bd_mnt
->mnt_sb
; /* For writeback */
532 * Most likely _very_ bad one - but then it's hardly critical for small
533 * /dev and can be fixed when somebody will need really large one.
534 * Keep in mind that it will be fed through icache hash function too.
536 static inline unsigned long hash(dev_t dev
)
538 return MAJOR(dev
)+MINOR(dev
);
541 static int bdev_test(struct inode
*inode
, void *data
)
543 return BDEV_I(inode
)->bdev
.bd_dev
== *(dev_t
*)data
;
546 static int bdev_set(struct inode
*inode
, void *data
)
548 BDEV_I(inode
)->bdev
.bd_dev
= *(dev_t
*)data
;
552 static LIST_HEAD(all_bdevs
);
554 struct block_device
*bdget(dev_t dev
)
556 struct block_device
*bdev
;
559 inode
= iget5_locked(blockdev_superblock
, hash(dev
),
560 bdev_test
, bdev_set
, &dev
);
565 bdev
= &BDEV_I(inode
)->bdev
;
567 if (inode
->i_state
& I_NEW
) {
568 bdev
->bd_contains
= NULL
;
569 bdev
->bd_super
= NULL
;
570 bdev
->bd_inode
= inode
;
571 bdev
->bd_block_size
= (1 << inode
->i_blkbits
);
572 bdev
->bd_part_count
= 0;
573 bdev
->bd_invalidated
= 0;
574 inode
->i_mode
= S_IFBLK
;
576 inode
->i_bdev
= bdev
;
577 inode
->i_data
.a_ops
= &def_blk_aops
;
578 mapping_set_gfp_mask(&inode
->i_data
, GFP_USER
);
579 spin_lock(&bdev_lock
);
580 list_add(&bdev
->bd_list
, &all_bdevs
);
581 spin_unlock(&bdev_lock
);
582 unlock_new_inode(inode
);
587 EXPORT_SYMBOL(bdget
);
590 * bdgrab -- Grab a reference to an already referenced block device
591 * @bdev: Block device to grab a reference to.
593 struct block_device
*bdgrab(struct block_device
*bdev
)
595 ihold(bdev
->bd_inode
);
598 EXPORT_SYMBOL(bdgrab
);
600 long nr_blockdev_pages(void)
602 struct block_device
*bdev
;
604 spin_lock(&bdev_lock
);
605 list_for_each_entry(bdev
, &all_bdevs
, bd_list
) {
606 ret
+= bdev
->bd_inode
->i_mapping
->nrpages
;
608 spin_unlock(&bdev_lock
);
612 void bdput(struct block_device
*bdev
)
614 iput(bdev
->bd_inode
);
617 EXPORT_SYMBOL(bdput
);
619 static struct block_device
*bd_acquire(struct inode
*inode
)
621 struct block_device
*bdev
;
623 spin_lock(&bdev_lock
);
624 bdev
= inode
->i_bdev
;
626 ihold(bdev
->bd_inode
);
627 spin_unlock(&bdev_lock
);
630 spin_unlock(&bdev_lock
);
632 bdev
= bdget(inode
->i_rdev
);
634 spin_lock(&bdev_lock
);
635 if (!inode
->i_bdev
) {
637 * We take an additional reference to bd_inode,
638 * and it's released in clear_inode() of inode.
639 * So, we can access it via ->i_mapping always
642 ihold(bdev
->bd_inode
);
643 inode
->i_bdev
= bdev
;
644 inode
->i_mapping
= bdev
->bd_inode
->i_mapping
;
645 list_add(&inode
->i_devices
, &bdev
->bd_inodes
);
647 spin_unlock(&bdev_lock
);
652 int sb_is_blkdev_sb(struct super_block
*sb
)
654 return sb
== blockdev_superblock
;
657 /* Call when you free inode */
659 void bd_forget(struct inode
*inode
)
661 struct block_device
*bdev
= NULL
;
663 spin_lock(&bdev_lock
);
664 if (!sb_is_blkdev_sb(inode
->i_sb
))
665 bdev
= inode
->i_bdev
;
667 spin_unlock(&bdev_lock
);
670 iput(bdev
->bd_inode
);
674 * bd_may_claim - test whether a block device can be claimed
675 * @bdev: block device of interest
676 * @whole: whole block device containing @bdev, may equal @bdev
677 * @holder: holder trying to claim @bdev
679 * Test whether @bdev can be claimed by @holder.
682 * spin_lock(&bdev_lock).
685 * %true if @bdev can be claimed, %false otherwise.
687 static bool bd_may_claim(struct block_device
*bdev
, struct block_device
*whole
,
690 if (bdev
->bd_holder
== holder
)
691 return true; /* already a holder */
692 else if (bdev
->bd_holder
!= NULL
)
693 return false; /* held by someone else */
694 else if (bdev
->bd_contains
== bdev
)
695 return true; /* is a whole device which isn't held */
697 else if (whole
->bd_holder
== bd_may_claim
)
698 return true; /* is a partition of a device that is being partitioned */
699 else if (whole
->bd_holder
!= NULL
)
700 return false; /* is a partition of a held device */
702 return true; /* is a partition of an un-held device */
706 * bd_prepare_to_claim - prepare to claim a block device
707 * @bdev: block device of interest
708 * @whole: the whole device containing @bdev, may equal @bdev
709 * @holder: holder trying to claim @bdev
711 * Prepare to claim @bdev. This function fails if @bdev is already
712 * claimed by another holder and waits if another claiming is in
713 * progress. This function doesn't actually claim. On successful
714 * return, the caller has ownership of bd_claiming and bd_holder[s].
717 * spin_lock(&bdev_lock). Might release bdev_lock, sleep and regrab
721 * 0 if @bdev can be claimed, -EBUSY otherwise.
723 static int bd_prepare_to_claim(struct block_device
*bdev
,
724 struct block_device
*whole
, void *holder
)
727 /* if someone else claimed, fail */
728 if (!bd_may_claim(bdev
, whole
, holder
))
731 /* if claiming is already in progress, wait for it to finish */
732 if (whole
->bd_claiming
) {
733 wait_queue_head_t
*wq
= bit_waitqueue(&whole
->bd_claiming
, 0);
736 prepare_to_wait(wq
, &wait
, TASK_UNINTERRUPTIBLE
);
737 spin_unlock(&bdev_lock
);
739 finish_wait(wq
, &wait
);
740 spin_lock(&bdev_lock
);
749 * bd_start_claiming - start claiming a block device
750 * @bdev: block device of interest
751 * @holder: holder trying to claim @bdev
753 * @bdev is about to be opened exclusively. Check @bdev can be opened
754 * exclusively and mark that an exclusive open is in progress. Each
755 * successful call to this function must be matched with a call to
756 * either bd_finish_claiming() or bd_abort_claiming() (which do not
759 * This function is used to gain exclusive access to the block device
760 * without actually causing other exclusive open attempts to fail. It
761 * should be used when the open sequence itself requires exclusive
762 * access but may subsequently fail.
768 * Pointer to the block device containing @bdev on success, ERR_PTR()
771 static struct block_device
*bd_start_claiming(struct block_device
*bdev
,
774 struct gendisk
*disk
;
775 struct block_device
*whole
;
781 * @bdev might not have been initialized properly yet, look up
782 * and grab the outer block device the hard way.
784 disk
= get_gendisk(bdev
->bd_dev
, &partno
);
786 return ERR_PTR(-ENXIO
);
789 * Normally, @bdev should equal what's returned from bdget_disk()
790 * if partno is 0; however, some drivers (floppy) use multiple
791 * bdev's for the same physical device and @bdev may be one of the
792 * aliases. Keep @bdev if partno is 0. This means claimer
793 * tracking is broken for those devices but it has always been that
797 whole
= bdget_disk(disk
, 0);
799 whole
= bdgrab(bdev
);
801 module_put(disk
->fops
->owner
);
804 return ERR_PTR(-ENOMEM
);
806 /* prepare to claim, if successful, mark claiming in progress */
807 spin_lock(&bdev_lock
);
809 err
= bd_prepare_to_claim(bdev
, whole
, holder
);
811 whole
->bd_claiming
= holder
;
812 spin_unlock(&bdev_lock
);
815 spin_unlock(&bdev_lock
);
822 struct bd_holder_disk
{
823 struct list_head list
;
824 struct gendisk
*disk
;
828 static struct bd_holder_disk
*bd_find_holder_disk(struct block_device
*bdev
,
829 struct gendisk
*disk
)
831 struct bd_holder_disk
*holder
;
833 list_for_each_entry(holder
, &bdev
->bd_holder_disks
, list
)
834 if (holder
->disk
== disk
)
839 static int add_symlink(struct kobject
*from
, struct kobject
*to
)
841 return sysfs_create_link(from
, to
, kobject_name(to
));
844 static void del_symlink(struct kobject
*from
, struct kobject
*to
)
846 sysfs_remove_link(from
, kobject_name(to
));
850 * bd_link_disk_holder - create symlinks between holding disk and slave bdev
851 * @bdev: the claimed slave bdev
852 * @disk: the holding disk
854 * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT.
856 * This functions creates the following sysfs symlinks.
858 * - from "slaves" directory of the holder @disk to the claimed @bdev
859 * - from "holders" directory of the @bdev to the holder @disk
861 * For example, if /dev/dm-0 maps to /dev/sda and disk for dm-0 is
862 * passed to bd_link_disk_holder(), then:
864 * /sys/block/dm-0/slaves/sda --> /sys/block/sda
865 * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0
867 * The caller must have claimed @bdev before calling this function and
868 * ensure that both @bdev and @disk are valid during the creation and
869 * lifetime of these symlinks.
875 * 0 on success, -errno on failure.
877 int bd_link_disk_holder(struct block_device
*bdev
, struct gendisk
*disk
)
879 struct bd_holder_disk
*holder
;
882 mutex_lock(&bdev
->bd_mutex
);
884 WARN_ON_ONCE(!bdev
->bd_holder
);
886 /* FIXME: remove the following once add_disk() handles errors */
887 if (WARN_ON(!disk
->slave_dir
|| !bdev
->bd_part
->holder_dir
))
890 holder
= bd_find_holder_disk(bdev
, disk
);
896 holder
= kzalloc(sizeof(*holder
), GFP_KERNEL
);
902 INIT_LIST_HEAD(&holder
->list
);
906 ret
= add_symlink(disk
->slave_dir
, &part_to_dev(bdev
->bd_part
)->kobj
);
910 ret
= add_symlink(bdev
->bd_part
->holder_dir
, &disk_to_dev(disk
)->kobj
);
914 * bdev could be deleted beneath us which would implicitly destroy
915 * the holder directory. Hold on to it.
917 kobject_get(bdev
->bd_part
->holder_dir
);
919 list_add(&holder
->list
, &bdev
->bd_holder_disks
);
923 del_symlink(disk
->slave_dir
, &part_to_dev(bdev
->bd_part
)->kobj
);
927 mutex_unlock(&bdev
->bd_mutex
);
930 EXPORT_SYMBOL_GPL(bd_link_disk_holder
);
933 * bd_unlink_disk_holder - destroy symlinks created by bd_link_disk_holder()
934 * @bdev: the calimed slave bdev
935 * @disk: the holding disk
937 * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT.
942 void bd_unlink_disk_holder(struct block_device
*bdev
, struct gendisk
*disk
)
944 struct bd_holder_disk
*holder
;
946 mutex_lock(&bdev
->bd_mutex
);
948 holder
= bd_find_holder_disk(bdev
, disk
);
950 if (!WARN_ON_ONCE(holder
== NULL
) && !--holder
->refcnt
) {
951 del_symlink(disk
->slave_dir
, &part_to_dev(bdev
->bd_part
)->kobj
);
952 del_symlink(bdev
->bd_part
->holder_dir
,
953 &disk_to_dev(disk
)->kobj
);
954 kobject_put(bdev
->bd_part
->holder_dir
);
955 list_del_init(&holder
->list
);
959 mutex_unlock(&bdev
->bd_mutex
);
961 EXPORT_SYMBOL_GPL(bd_unlink_disk_holder
);
965 * flush_disk - invalidates all buffer-cache entries on a disk
967 * @bdev: struct block device to be flushed
968 * @kill_dirty: flag to guide handling of dirty inodes
970 * Invalidates all buffer-cache entries on a disk. It should be called
971 * when a disk has been changed -- either by a media change or online
974 static void flush_disk(struct block_device
*bdev
, bool kill_dirty
)
976 if (__invalidate_device(bdev
, kill_dirty
)) {
977 char name
[BDEVNAME_SIZE
] = "";
980 disk_name(bdev
->bd_disk
, 0, name
);
981 printk(KERN_WARNING
"VFS: busy inodes on changed media or "
982 "resized disk %s\n", name
);
987 if (disk_part_scan_enabled(bdev
->bd_disk
))
988 bdev
->bd_invalidated
= 1;
992 * check_disk_size_change - checks for disk size change and adjusts bdev size.
993 * @disk: struct gendisk to check
994 * @bdev: struct bdev to adjust.
996 * This routine checks to see if the bdev size does not match the disk size
997 * and adjusts it if it differs.
999 void check_disk_size_change(struct gendisk
*disk
, struct block_device
*bdev
)
1001 loff_t disk_size
, bdev_size
;
1003 disk_size
= (loff_t
)get_capacity(disk
) << 9;
1004 bdev_size
= i_size_read(bdev
->bd_inode
);
1005 if (disk_size
!= bdev_size
) {
1006 char name
[BDEVNAME_SIZE
];
1008 disk_name(disk
, 0, name
);
1010 "%s: detected capacity change from %lld to %lld\n",
1011 name
, bdev_size
, disk_size
);
1012 i_size_write(bdev
->bd_inode
, disk_size
);
1013 flush_disk(bdev
, false);
1016 EXPORT_SYMBOL(check_disk_size_change
);
1019 * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back
1020 * @disk: struct gendisk to be revalidated
1022 * This routine is a wrapper for lower-level driver's revalidate_disk
1023 * call-backs. It is used to do common pre and post operations needed
1024 * for all revalidate_disk operations.
1026 int revalidate_disk(struct gendisk
*disk
)
1028 struct block_device
*bdev
;
1031 if (disk
->fops
->revalidate_disk
)
1032 ret
= disk
->fops
->revalidate_disk(disk
);
1034 bdev
= bdget_disk(disk
, 0);
1038 mutex_lock(&bdev
->bd_mutex
);
1039 check_disk_size_change(disk
, bdev
);
1040 bdev
->bd_invalidated
= 0;
1041 mutex_unlock(&bdev
->bd_mutex
);
1045 EXPORT_SYMBOL(revalidate_disk
);
1048 * This routine checks whether a removable media has been changed,
1049 * and invalidates all buffer-cache-entries in that case. This
1050 * is a relatively slow routine, so we have to try to minimize using
1051 * it. Thus it is called only upon a 'mount' or 'open'. This
1052 * is the best way of combining speed and utility, I think.
1053 * People changing diskettes in the middle of an operation deserve
1056 int check_disk_change(struct block_device
*bdev
)
1058 struct gendisk
*disk
= bdev
->bd_disk
;
1059 const struct block_device_operations
*bdops
= disk
->fops
;
1060 unsigned int events
;
1062 events
= disk_clear_events(disk
, DISK_EVENT_MEDIA_CHANGE
|
1063 DISK_EVENT_EJECT_REQUEST
);
1064 if (!(events
& DISK_EVENT_MEDIA_CHANGE
))
1067 flush_disk(bdev
, true);
1068 if (bdops
->revalidate_disk
)
1069 bdops
->revalidate_disk(bdev
->bd_disk
);
1073 EXPORT_SYMBOL(check_disk_change
);
1075 void bd_set_size(struct block_device
*bdev
, loff_t size
)
1077 unsigned bsize
= bdev_logical_block_size(bdev
);
1079 mutex_lock(&bdev
->bd_inode
->i_mutex
);
1080 i_size_write(bdev
->bd_inode
, size
);
1081 mutex_unlock(&bdev
->bd_inode
->i_mutex
);
1082 while (bsize
< PAGE_CACHE_SIZE
) {
1087 bdev
->bd_block_size
= bsize
;
1088 bdev
->bd_inode
->i_blkbits
= blksize_bits(bsize
);
1090 EXPORT_SYMBOL(bd_set_size
);
1092 static void __blkdev_put(struct block_device
*bdev
, fmode_t mode
, int for_part
);
1097 * mutex_lock(part->bd_mutex)
1098 * mutex_lock_nested(whole->bd_mutex, 1)
1101 static int __blkdev_get(struct block_device
*bdev
, fmode_t mode
, int for_part
)
1103 struct gendisk
*disk
;
1104 struct module
*owner
;
1109 if (mode
& FMODE_READ
)
1111 if (mode
& FMODE_WRITE
)
1114 * hooks: /n/, see "layering violations".
1117 ret
= devcgroup_inode_permission(bdev
->bd_inode
, perm
);
1127 disk
= get_gendisk(bdev
->bd_dev
, &partno
);
1130 owner
= disk
->fops
->owner
;
1132 disk_block_events(disk
);
1133 mutex_lock_nested(&bdev
->bd_mutex
, for_part
);
1134 if (!bdev
->bd_openers
) {
1135 bdev
->bd_disk
= disk
;
1136 bdev
->bd_queue
= disk
->queue
;
1137 bdev
->bd_contains
= bdev
;
1140 bdev
->bd_part
= disk_get_part(disk
, partno
);
1145 if (disk
->fops
->open
) {
1146 ret
= disk
->fops
->open(bdev
, mode
);
1147 if (ret
== -ERESTARTSYS
) {
1148 /* Lost a race with 'disk' being
1149 * deleted, try again.
1152 disk_put_part(bdev
->bd_part
);
1153 bdev
->bd_part
= NULL
;
1154 bdev
->bd_disk
= NULL
;
1155 bdev
->bd_queue
= NULL
;
1156 mutex_unlock(&bdev
->bd_mutex
);
1157 disk_unblock_events(disk
);
1165 bd_set_size(bdev
,(loff_t
)get_capacity(disk
)<<9);
1168 * If the device is invalidated, rescan partition
1169 * if open succeeded or failed with -ENOMEDIUM.
1170 * The latter is necessary to prevent ghost
1171 * partitions on a removed medium.
1173 if (bdev
->bd_invalidated
) {
1175 rescan_partitions(disk
, bdev
);
1176 else if (ret
== -ENOMEDIUM
)
1177 invalidate_partitions(disk
, bdev
);
1182 struct block_device
*whole
;
1183 whole
= bdget_disk(disk
, 0);
1188 ret
= __blkdev_get(whole
, mode
, 1);
1191 bdev
->bd_contains
= whole
;
1192 bdev
->bd_part
= disk_get_part(disk
, partno
);
1193 if (!(disk
->flags
& GENHD_FL_UP
) ||
1194 !bdev
->bd_part
|| !bdev
->bd_part
->nr_sects
) {
1198 bd_set_size(bdev
, (loff_t
)bdev
->bd_part
->nr_sects
<< 9);
1201 if (bdev
->bd_contains
== bdev
) {
1203 if (bdev
->bd_disk
->fops
->open
)
1204 ret
= bdev
->bd_disk
->fops
->open(bdev
, mode
);
1205 /* the same as first opener case, read comment there */
1206 if (bdev
->bd_invalidated
) {
1208 rescan_partitions(bdev
->bd_disk
, bdev
);
1209 else if (ret
== -ENOMEDIUM
)
1210 invalidate_partitions(bdev
->bd_disk
, bdev
);
1213 goto out_unlock_bdev
;
1215 /* only one opener holds refs to the module and disk */
1221 bdev
->bd_part_count
++;
1222 mutex_unlock(&bdev
->bd_mutex
);
1223 disk_unblock_events(disk
);
1227 disk_put_part(bdev
->bd_part
);
1228 bdev
->bd_disk
= NULL
;
1229 bdev
->bd_part
= NULL
;
1230 bdev
->bd_queue
= NULL
;
1231 if (bdev
!= bdev
->bd_contains
)
1232 __blkdev_put(bdev
->bd_contains
, mode
, 1);
1233 bdev
->bd_contains
= NULL
;
1235 mutex_unlock(&bdev
->bd_mutex
);
1236 disk_unblock_events(disk
);
1246 * blkdev_get - open a block device
1247 * @bdev: block_device to open
1248 * @mode: FMODE_* mask
1249 * @holder: exclusive holder identifier
1251 * Open @bdev with @mode. If @mode includes %FMODE_EXCL, @bdev is
1252 * open with exclusive access. Specifying %FMODE_EXCL with %NULL
1253 * @holder is invalid. Exclusive opens may nest for the same @holder.
1255 * On success, the reference count of @bdev is unchanged. On failure,
1262 * 0 on success, -errno on failure.
1264 int blkdev_get(struct block_device
*bdev
, fmode_t mode
, void *holder
)
1266 struct block_device
*whole
= NULL
;
1269 WARN_ON_ONCE((mode
& FMODE_EXCL
) && !holder
);
1271 if ((mode
& FMODE_EXCL
) && holder
) {
1272 whole
= bd_start_claiming(bdev
, holder
);
1273 if (IS_ERR(whole
)) {
1275 return PTR_ERR(whole
);
1279 res
= __blkdev_get(bdev
, mode
, 0);
1282 struct gendisk
*disk
= whole
->bd_disk
;
1284 /* finish claiming */
1285 mutex_lock(&bdev
->bd_mutex
);
1286 spin_lock(&bdev_lock
);
1289 BUG_ON(!bd_may_claim(bdev
, whole
, holder
));
1291 * Note that for a whole device bd_holders
1292 * will be incremented twice, and bd_holder
1293 * will be set to bd_may_claim before being
1296 whole
->bd_holders
++;
1297 whole
->bd_holder
= bd_may_claim
;
1299 bdev
->bd_holder
= holder
;
1302 /* tell others that we're done */
1303 BUG_ON(whole
->bd_claiming
!= holder
);
1304 whole
->bd_claiming
= NULL
;
1305 wake_up_bit(&whole
->bd_claiming
, 0);
1307 spin_unlock(&bdev_lock
);
1310 * Block event polling for write claims if requested. Any
1311 * write holder makes the write_holder state stick until
1312 * all are released. This is good enough and tracking
1313 * individual writeable reference is too fragile given the
1314 * way @mode is used in blkdev_get/put().
1316 if (!res
&& (mode
& FMODE_WRITE
) && !bdev
->bd_write_holder
&&
1317 (disk
->flags
& GENHD_FL_BLOCK_EVENTS_ON_EXCL_WRITE
)) {
1318 bdev
->bd_write_holder
= true;
1319 disk_block_events(disk
);
1322 mutex_unlock(&bdev
->bd_mutex
);
1328 EXPORT_SYMBOL(blkdev_get
);
1331 * blkdev_get_by_path - open a block device by name
1332 * @path: path to the block device to open
1333 * @mode: FMODE_* mask
1334 * @holder: exclusive holder identifier
1336 * Open the blockdevice described by the device file at @path. @mode
1337 * and @holder are identical to blkdev_get().
1339 * On success, the returned block_device has reference count of one.
1345 * Pointer to block_device on success, ERR_PTR(-errno) on failure.
1347 struct block_device
*blkdev_get_by_path(const char *path
, fmode_t mode
,
1350 struct block_device
*bdev
;
1353 bdev
= lookup_bdev(path
);
1357 err
= blkdev_get(bdev
, mode
, holder
);
1359 return ERR_PTR(err
);
1361 if ((mode
& FMODE_WRITE
) && bdev_read_only(bdev
)) {
1362 blkdev_put(bdev
, mode
);
1363 return ERR_PTR(-EACCES
);
1368 EXPORT_SYMBOL(blkdev_get_by_path
);
1371 * blkdev_get_by_dev - open a block device by device number
1372 * @dev: device number of block device to open
1373 * @mode: FMODE_* mask
1374 * @holder: exclusive holder identifier
1376 * Open the blockdevice described by device number @dev. @mode and
1377 * @holder are identical to blkdev_get().
1379 * Use it ONLY if you really do not have anything better - i.e. when
1380 * you are behind a truly sucky interface and all you are given is a
1381 * device number. _Never_ to be used for internal purposes. If you
1382 * ever need it - reconsider your API.
1384 * On success, the returned block_device has reference count of one.
1390 * Pointer to block_device on success, ERR_PTR(-errno) on failure.
1392 struct block_device
*blkdev_get_by_dev(dev_t dev
, fmode_t mode
, void *holder
)
1394 struct block_device
*bdev
;
1399 return ERR_PTR(-ENOMEM
);
1401 err
= blkdev_get(bdev
, mode
, holder
);
1403 return ERR_PTR(err
);
1407 EXPORT_SYMBOL(blkdev_get_by_dev
);
1409 static int blkdev_open(struct inode
* inode
, struct file
* filp
)
1411 struct block_device
*bdev
;
1414 * Preserve backwards compatibility and allow large file access
1415 * even if userspace doesn't ask for it explicitly. Some mkfs
1416 * binary needs it. We might want to drop this workaround
1417 * during an unstable branch.
1419 filp
->f_flags
|= O_LARGEFILE
;
1421 if (filp
->f_flags
& O_NDELAY
)
1422 filp
->f_mode
|= FMODE_NDELAY
;
1423 if (filp
->f_flags
& O_EXCL
)
1424 filp
->f_mode
|= FMODE_EXCL
;
1425 if ((filp
->f_flags
& O_ACCMODE
) == 3)
1426 filp
->f_mode
|= FMODE_WRITE_IOCTL
;
1428 bdev
= bd_acquire(inode
);
1432 filp
->f_mapping
= bdev
->bd_inode
->i_mapping
;
1434 return blkdev_get(bdev
, filp
->f_mode
, filp
);
1437 static void __blkdev_put(struct block_device
*bdev
, fmode_t mode
, int for_part
)
1439 struct gendisk
*disk
= bdev
->bd_disk
;
1440 struct block_device
*victim
= NULL
;
1442 mutex_lock_nested(&bdev
->bd_mutex
, for_part
);
1444 bdev
->bd_part_count
--;
1446 if (!--bdev
->bd_openers
) {
1447 WARN_ON_ONCE(bdev
->bd_holders
);
1448 sync_blockdev(bdev
);
1451 * ->release can cause the queue to disappear, so flush all
1452 * dirty data before.
1454 bdev_write_inode(bdev
->bd_inode
);
1456 if (bdev
->bd_contains
== bdev
) {
1457 if (disk
->fops
->release
)
1458 disk
->fops
->release(disk
, mode
);
1460 if (!bdev
->bd_openers
) {
1461 struct module
*owner
= disk
->fops
->owner
;
1463 disk_put_part(bdev
->bd_part
);
1464 bdev
->bd_part
= NULL
;
1465 bdev
->bd_disk
= NULL
;
1466 if (bdev
!= bdev
->bd_contains
)
1467 victim
= bdev
->bd_contains
;
1468 bdev
->bd_contains
= NULL
;
1473 mutex_unlock(&bdev
->bd_mutex
);
1476 __blkdev_put(victim
, mode
, 1);
1479 void blkdev_put(struct block_device
*bdev
, fmode_t mode
)
1481 mutex_lock(&bdev
->bd_mutex
);
1483 if (mode
& FMODE_EXCL
) {
1487 * Release a claim on the device. The holder fields
1488 * are protected with bdev_lock. bd_mutex is to
1489 * synchronize disk_holder unlinking.
1491 spin_lock(&bdev_lock
);
1493 WARN_ON_ONCE(--bdev
->bd_holders
< 0);
1494 WARN_ON_ONCE(--bdev
->bd_contains
->bd_holders
< 0);
1496 /* bd_contains might point to self, check in a separate step */
1497 if ((bdev_free
= !bdev
->bd_holders
))
1498 bdev
->bd_holder
= NULL
;
1499 if (!bdev
->bd_contains
->bd_holders
)
1500 bdev
->bd_contains
->bd_holder
= NULL
;
1502 spin_unlock(&bdev_lock
);
1505 * If this was the last claim, remove holder link and
1506 * unblock evpoll if it was a write holder.
1508 if (bdev_free
&& bdev
->bd_write_holder
) {
1509 disk_unblock_events(bdev
->bd_disk
);
1510 bdev
->bd_write_holder
= false;
1515 * Trigger event checking and tell drivers to flush MEDIA_CHANGE
1516 * event. This is to ensure detection of media removal commanded
1517 * from userland - e.g. eject(1).
1519 disk_flush_events(bdev
->bd_disk
, DISK_EVENT_MEDIA_CHANGE
);
1521 mutex_unlock(&bdev
->bd_mutex
);
1523 __blkdev_put(bdev
, mode
, 0);
1525 EXPORT_SYMBOL(blkdev_put
);
1527 static int blkdev_close(struct inode
* inode
, struct file
* filp
)
1529 struct block_device
*bdev
= I_BDEV(filp
->f_mapping
->host
);
1530 blkdev_put(bdev
, filp
->f_mode
);
1534 static long block_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1536 struct block_device
*bdev
= I_BDEV(file
->f_mapping
->host
);
1537 fmode_t mode
= file
->f_mode
;
1540 * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have
1541 * to updated it before every ioctl.
1543 if (file
->f_flags
& O_NDELAY
)
1544 mode
|= FMODE_NDELAY
;
1546 mode
&= ~FMODE_NDELAY
;
1548 return blkdev_ioctl(bdev
, mode
, cmd
, arg
);
1552 * Write data to the block device. Only intended for the block device itself
1553 * and the raw driver which basically is a fake block device.
1555 * Does not take i_mutex for the write and thus is not for general purpose
1558 ssize_t
blkdev_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
1560 struct file
*file
= iocb
->ki_filp
;
1561 struct blk_plug plug
;
1564 blk_start_plug(&plug
);
1565 ret
= __generic_file_write_iter(iocb
, from
);
1568 err
= generic_write_sync(file
, iocb
->ki_pos
- ret
, ret
);
1572 blk_finish_plug(&plug
);
1575 EXPORT_SYMBOL_GPL(blkdev_write_iter
);
1577 ssize_t
blkdev_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
1579 struct file
*file
= iocb
->ki_filp
;
1580 struct inode
*bd_inode
= file
->f_mapping
->host
;
1581 loff_t size
= i_size_read(bd_inode
);
1582 loff_t pos
= iocb
->ki_pos
;
1588 iov_iter_truncate(to
, size
);
1589 return generic_file_read_iter(iocb
, to
);
1591 EXPORT_SYMBOL_GPL(blkdev_read_iter
);
1594 * Try to release a page associated with block device when the system
1595 * is under memory pressure.
1597 static int blkdev_releasepage(struct page
*page
, gfp_t wait
)
1599 struct super_block
*super
= BDEV_I(page
->mapping
->host
)->bdev
.bd_super
;
1601 if (super
&& super
->s_op
->bdev_try_to_free_page
)
1602 return super
->s_op
->bdev_try_to_free_page(super
, page
, wait
);
1604 return try_to_free_buffers(page
);
1607 static const struct address_space_operations def_blk_aops
= {
1608 .readpage
= blkdev_readpage
,
1609 .readpages
= blkdev_readpages
,
1610 .writepage
= blkdev_writepage
,
1611 .write_begin
= blkdev_write_begin
,
1612 .write_end
= blkdev_write_end
,
1613 .writepages
= generic_writepages
,
1614 .releasepage
= blkdev_releasepage
,
1615 .direct_IO
= blkdev_direct_IO
,
1616 .is_dirty_writeback
= buffer_check_dirty_writeback
,
1619 const struct file_operations def_blk_fops
= {
1620 .open
= blkdev_open
,
1621 .release
= blkdev_close
,
1622 .llseek
= block_llseek
,
1623 .read
= new_sync_read
,
1624 .write
= new_sync_write
,
1625 .read_iter
= blkdev_read_iter
,
1626 .write_iter
= blkdev_write_iter
,
1627 .mmap
= generic_file_mmap
,
1628 .fsync
= blkdev_fsync
,
1629 .unlocked_ioctl
= block_ioctl
,
1630 #ifdef CONFIG_COMPAT
1631 .compat_ioctl
= compat_blkdev_ioctl
,
1633 .splice_read
= generic_file_splice_read
,
1634 .splice_write
= iter_file_splice_write
,
1637 int ioctl_by_bdev(struct block_device
*bdev
, unsigned cmd
, unsigned long arg
)
1640 mm_segment_t old_fs
= get_fs();
1642 res
= blkdev_ioctl(bdev
, 0, cmd
, arg
);
1647 EXPORT_SYMBOL(ioctl_by_bdev
);
1650 * lookup_bdev - lookup a struct block_device by name
1651 * @pathname: special file representing the block device
1653 * Get a reference to the blockdevice at @pathname in the current
1654 * namespace if possible and return it. Return ERR_PTR(error)
1657 struct block_device
*lookup_bdev(const char *pathname
)
1659 struct block_device
*bdev
;
1660 struct inode
*inode
;
1664 if (!pathname
|| !*pathname
)
1665 return ERR_PTR(-EINVAL
);
1667 error
= kern_path(pathname
, LOOKUP_FOLLOW
, &path
);
1669 return ERR_PTR(error
);
1671 inode
= path
.dentry
->d_inode
;
1673 if (!S_ISBLK(inode
->i_mode
))
1676 if (path
.mnt
->mnt_flags
& MNT_NODEV
)
1679 bdev
= bd_acquire(inode
);
1686 bdev
= ERR_PTR(error
);
1689 EXPORT_SYMBOL(lookup_bdev
);
1691 int __invalidate_device(struct block_device
*bdev
, bool kill_dirty
)
1693 struct super_block
*sb
= get_super(bdev
);
1698 * no need to lock the super, get_super holds the
1699 * read mutex so the filesystem cannot go away
1700 * under us (->put_super runs with the write lock
1703 shrink_dcache_sb(sb
);
1704 res
= invalidate_inodes(sb
, kill_dirty
);
1707 invalidate_bdev(bdev
);
1710 EXPORT_SYMBOL(__invalidate_device
);
1712 void iterate_bdevs(void (*func
)(struct block_device
*, void *), void *arg
)
1714 struct inode
*inode
, *old_inode
= NULL
;
1716 spin_lock(&inode_sb_list_lock
);
1717 list_for_each_entry(inode
, &blockdev_superblock
->s_inodes
, i_sb_list
) {
1718 struct address_space
*mapping
= inode
->i_mapping
;
1720 spin_lock(&inode
->i_lock
);
1721 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
|I_NEW
) ||
1722 mapping
->nrpages
== 0) {
1723 spin_unlock(&inode
->i_lock
);
1727 spin_unlock(&inode
->i_lock
);
1728 spin_unlock(&inode_sb_list_lock
);
1730 * We hold a reference to 'inode' so it couldn't have been
1731 * removed from s_inodes list while we dropped the
1732 * inode_sb_list_lock. We cannot iput the inode now as we can
1733 * be holding the last reference and we cannot iput it under
1734 * inode_sb_list_lock. So we keep the reference and iput it
1740 func(I_BDEV(inode
), arg
);
1742 spin_lock(&inode_sb_list_lock
);
1744 spin_unlock(&inode_sb_list_lock
);