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
);
53 * Move the inode from its current bdi to a new bdi. If the inode is dirty we
54 * need to move it onto the dirty list of @dst so that the inode is always on
57 static void bdev_inode_switch_bdi(struct inode
*inode
,
58 struct backing_dev_info
*dst
)
60 struct backing_dev_info
*old
= inode
->i_data
.backing_dev_info
;
61 bool wakeup_bdi
= false;
63 if (unlikely(dst
== old
)) /* deadlock avoidance */
65 bdi_lock_two(&old
->wb
, &dst
->wb
);
66 spin_lock(&inode
->i_lock
);
67 inode
->i_data
.backing_dev_info
= dst
;
68 if (inode
->i_state
& I_DIRTY
) {
69 if (bdi_cap_writeback_dirty(dst
) && !wb_has_dirty_io(&dst
->wb
))
71 list_move(&inode
->i_wb_list
, &dst
->wb
.b_dirty
);
73 spin_unlock(&inode
->i_lock
);
74 spin_unlock(&old
->wb
.list_lock
);
75 spin_unlock(&dst
->wb
.list_lock
);
78 bdi_wakeup_thread_delayed(dst
);
81 /* Kill _all_ buffers and pagecache , dirty or not.. */
82 void kill_bdev(struct block_device
*bdev
)
84 struct address_space
*mapping
= bdev
->bd_inode
->i_mapping
;
86 if (mapping
->nrpages
== 0 && mapping
->nrshadows
== 0)
90 truncate_inode_pages(mapping
, 0);
92 EXPORT_SYMBOL(kill_bdev
);
94 /* Invalidate clean unused buffers and pagecache. */
95 void invalidate_bdev(struct block_device
*bdev
)
97 struct address_space
*mapping
= bdev
->bd_inode
->i_mapping
;
99 if (mapping
->nrpages
== 0)
102 invalidate_bh_lrus();
103 lru_add_drain_all(); /* make sure all lru add caches are flushed */
104 invalidate_mapping_pages(mapping
, 0, -1);
105 /* 99% of the time, we don't need to flush the cleancache on the bdev.
106 * But, for the strange corners, lets be cautious
108 cleancache_invalidate_inode(mapping
);
110 EXPORT_SYMBOL(invalidate_bdev
);
112 int set_blocksize(struct block_device
*bdev
, int size
)
114 /* Size must be a power of two, and between 512 and PAGE_SIZE */
115 if (size
> PAGE_SIZE
|| size
< 512 || !is_power_of_2(size
))
118 /* Size cannot be smaller than the size supported by the device */
119 if (size
< bdev_logical_block_size(bdev
))
122 /* Don't change the size if it is same as current */
123 if (bdev
->bd_block_size
!= size
) {
125 bdev
->bd_block_size
= size
;
126 bdev
->bd_inode
->i_blkbits
= blksize_bits(size
);
132 EXPORT_SYMBOL(set_blocksize
);
134 int sb_set_blocksize(struct super_block
*sb
, int size
)
136 if (set_blocksize(sb
->s_bdev
, size
))
138 /* If we get here, we know size is power of two
139 * and it's value is between 512 and PAGE_SIZE */
140 sb
->s_blocksize
= size
;
141 sb
->s_blocksize_bits
= blksize_bits(size
);
142 return sb
->s_blocksize
;
145 EXPORT_SYMBOL(sb_set_blocksize
);
147 int sb_min_blocksize(struct super_block
*sb
, int size
)
149 int minsize
= bdev_logical_block_size(sb
->s_bdev
);
152 return sb_set_blocksize(sb
, size
);
155 EXPORT_SYMBOL(sb_min_blocksize
);
158 blkdev_get_block(struct inode
*inode
, sector_t iblock
,
159 struct buffer_head
*bh
, int create
)
161 bh
->b_bdev
= I_BDEV(inode
);
162 bh
->b_blocknr
= iblock
;
163 set_buffer_mapped(bh
);
168 blkdev_direct_IO(int rw
, struct kiocb
*iocb
, struct iov_iter
*iter
,
171 struct file
*file
= iocb
->ki_filp
;
172 struct inode
*inode
= file
->f_mapping
->host
;
174 return __blockdev_direct_IO(rw
, iocb
, inode
, I_BDEV(inode
), iter
,
175 offset
, blkdev_get_block
,
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 error
= freeze_super(sb
);
250 deactivate_super(sb
);
251 bdev
->bd_fsfreeze_count
--;
252 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
253 return ERR_PTR(error
);
255 deactivate_super(sb
);
258 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
259 return sb
; /* thaw_bdev releases s->s_umount */
261 EXPORT_SYMBOL(freeze_bdev
);
264 * thaw_bdev -- unlock filesystem
265 * @bdev: blockdevice to unlock
266 * @sb: associated superblock
268 * Unlocks the filesystem and marks it writeable again after freeze_bdev().
270 int thaw_bdev(struct block_device
*bdev
, struct super_block
*sb
)
274 mutex_lock(&bdev
->bd_fsfreeze_mutex
);
275 if (!bdev
->bd_fsfreeze_count
)
279 if (--bdev
->bd_fsfreeze_count
> 0)
285 error
= thaw_super(sb
);
287 bdev
->bd_fsfreeze_count
++;
288 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
292 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
295 EXPORT_SYMBOL(thaw_bdev
);
297 static int blkdev_writepage(struct page
*page
, struct writeback_control
*wbc
)
299 return block_write_full_page(page
, blkdev_get_block
, wbc
);
302 static int blkdev_readpage(struct file
* file
, struct page
* page
)
304 return block_read_full_page(page
, blkdev_get_block
);
307 static int blkdev_readpages(struct file
*file
, struct address_space
*mapping
,
308 struct list_head
*pages
, unsigned nr_pages
)
310 return mpage_readpages(mapping
, pages
, nr_pages
, blkdev_get_block
);
313 static int blkdev_write_begin(struct file
*file
, struct address_space
*mapping
,
314 loff_t pos
, unsigned len
, unsigned flags
,
315 struct page
**pagep
, void **fsdata
)
317 return block_write_begin(mapping
, pos
, len
, flags
, pagep
,
321 static int blkdev_write_end(struct file
*file
, struct address_space
*mapping
,
322 loff_t pos
, unsigned len
, unsigned copied
,
323 struct page
*page
, void *fsdata
)
326 ret
= block_write_end(file
, mapping
, pos
, len
, copied
, page
, fsdata
);
329 page_cache_release(page
);
336 * for a block special file file_inode(file)->i_size is zero
337 * so we compute the size by hand (just as in block_read/write above)
339 static loff_t
block_llseek(struct file
*file
, loff_t offset
, int whence
)
341 struct inode
*bd_inode
= file
->f_mapping
->host
;
344 mutex_lock(&bd_inode
->i_mutex
);
345 retval
= fixed_size_llseek(file
, offset
, whence
, i_size_read(bd_inode
));
346 mutex_unlock(&bd_inode
->i_mutex
);
350 int blkdev_fsync(struct file
*filp
, loff_t start
, loff_t end
, int datasync
)
352 struct inode
*bd_inode
= filp
->f_mapping
->host
;
353 struct block_device
*bdev
= I_BDEV(bd_inode
);
356 error
= filemap_write_and_wait_range(filp
->f_mapping
, start
, end
);
361 * There is no need to serialise calls to blkdev_issue_flush with
362 * i_mutex and doing so causes performance issues with concurrent
363 * O_SYNC writers to a block device.
365 error
= blkdev_issue_flush(bdev
, GFP_KERNEL
, NULL
);
366 if (error
== -EOPNOTSUPP
)
371 EXPORT_SYMBOL(blkdev_fsync
);
374 * bdev_read_page() - Start reading a page from a block device
375 * @bdev: The device to read the page from
376 * @sector: The offset on the device to read the page to (need not be aligned)
377 * @page: The page to read
379 * On entry, the page should be locked. It will be unlocked when the page
380 * has been read. If the block driver implements rw_page synchronously,
381 * that will be true on exit from this function, but it need not be.
383 * Errors returned by this function are usually "soft", eg out of memory, or
384 * queue full; callers should try a different route to read this page rather
385 * than propagate an error back up the stack.
387 * Return: negative errno if an error occurs, 0 if submission was successful.
389 int bdev_read_page(struct block_device
*bdev
, sector_t sector
,
392 const struct block_device_operations
*ops
= bdev
->bd_disk
->fops
;
395 return ops
->rw_page(bdev
, sector
+ get_start_sect(bdev
), page
, READ
);
397 EXPORT_SYMBOL_GPL(bdev_read_page
);
400 * bdev_write_page() - Start writing a page to a block device
401 * @bdev: The device to write the page to
402 * @sector: The offset on the device to write the page to (need not be aligned)
403 * @page: The page to write
404 * @wbc: The writeback_control for the write
406 * On entry, the page should be locked and not currently under writeback.
407 * On exit, if the write started successfully, the page will be unlocked and
408 * under writeback. If the write failed already (eg the driver failed to
409 * queue the page to the device), the page will still be locked. If the
410 * caller is a ->writepage implementation, it will need to unlock the page.
412 * Errors returned by this function are usually "soft", eg out of memory, or
413 * queue full; callers should try a different route to write this page rather
414 * than propagate an error back up the stack.
416 * Return: negative errno if an error occurs, 0 if submission was successful.
418 int bdev_write_page(struct block_device
*bdev
, sector_t sector
,
419 struct page
*page
, struct writeback_control
*wbc
)
422 int rw
= (wbc
->sync_mode
== WB_SYNC_ALL
) ? WRITE_SYNC
: WRITE
;
423 const struct block_device_operations
*ops
= bdev
->bd_disk
->fops
;
426 set_page_writeback(page
);
427 result
= ops
->rw_page(bdev
, sector
+ get_start_sect(bdev
), page
, rw
);
429 end_page_writeback(page
);
434 EXPORT_SYMBOL_GPL(bdev_write_page
);
440 static __cacheline_aligned_in_smp
DEFINE_SPINLOCK(bdev_lock
);
441 static struct kmem_cache
* bdev_cachep __read_mostly
;
443 static struct inode
*bdev_alloc_inode(struct super_block
*sb
)
445 struct bdev_inode
*ei
= kmem_cache_alloc(bdev_cachep
, GFP_KERNEL
);
448 return &ei
->vfs_inode
;
451 static void bdev_i_callback(struct rcu_head
*head
)
453 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
454 struct bdev_inode
*bdi
= BDEV_I(inode
);
456 kmem_cache_free(bdev_cachep
, bdi
);
459 static void bdev_destroy_inode(struct inode
*inode
)
461 call_rcu(&inode
->i_rcu
, bdev_i_callback
);
464 static void init_once(void *foo
)
466 struct bdev_inode
*ei
= (struct bdev_inode
*) foo
;
467 struct block_device
*bdev
= &ei
->bdev
;
469 memset(bdev
, 0, sizeof(*bdev
));
470 mutex_init(&bdev
->bd_mutex
);
471 INIT_LIST_HEAD(&bdev
->bd_inodes
);
472 INIT_LIST_HEAD(&bdev
->bd_list
);
474 INIT_LIST_HEAD(&bdev
->bd_holder_disks
);
476 inode_init_once(&ei
->vfs_inode
);
477 /* Initialize mutex for freeze. */
478 mutex_init(&bdev
->bd_fsfreeze_mutex
);
481 static inline void __bd_forget(struct inode
*inode
)
483 list_del_init(&inode
->i_devices
);
484 inode
->i_bdev
= NULL
;
485 inode
->i_mapping
= &inode
->i_data
;
488 static void bdev_evict_inode(struct inode
*inode
)
490 struct block_device
*bdev
= &BDEV_I(inode
)->bdev
;
492 truncate_inode_pages_final(&inode
->i_data
);
493 invalidate_inode_buffers(inode
); /* is it needed here? */
495 spin_lock(&bdev_lock
);
496 while ( (p
= bdev
->bd_inodes
.next
) != &bdev
->bd_inodes
) {
497 __bd_forget(list_entry(p
, struct inode
, i_devices
));
499 list_del_init(&bdev
->bd_list
);
500 spin_unlock(&bdev_lock
);
503 static const struct super_operations bdev_sops
= {
504 .statfs
= simple_statfs
,
505 .alloc_inode
= bdev_alloc_inode
,
506 .destroy_inode
= bdev_destroy_inode
,
507 .drop_inode
= generic_delete_inode
,
508 .evict_inode
= bdev_evict_inode
,
511 static struct dentry
*bd_mount(struct file_system_type
*fs_type
,
512 int flags
, const char *dev_name
, void *data
)
514 return mount_pseudo(fs_type
, "bdev:", &bdev_sops
, NULL
, BDEVFS_MAGIC
);
517 static struct file_system_type bd_type
= {
520 .kill_sb
= kill_anon_super
,
523 static struct super_block
*blockdev_superblock __read_mostly
;
525 void __init
bdev_cache_init(void)
528 static struct vfsmount
*bd_mnt
;
530 bdev_cachep
= kmem_cache_create("bdev_cache", sizeof(struct bdev_inode
),
531 0, (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
532 SLAB_MEM_SPREAD
|SLAB_PANIC
),
534 err
= register_filesystem(&bd_type
);
536 panic("Cannot register bdev pseudo-fs");
537 bd_mnt
= kern_mount(&bd_type
);
539 panic("Cannot create bdev pseudo-fs");
540 blockdev_superblock
= bd_mnt
->mnt_sb
; /* For writeback */
544 * Most likely _very_ bad one - but then it's hardly critical for small
545 * /dev and can be fixed when somebody will need really large one.
546 * Keep in mind that it will be fed through icache hash function too.
548 static inline unsigned long hash(dev_t dev
)
550 return MAJOR(dev
)+MINOR(dev
);
553 static int bdev_test(struct inode
*inode
, void *data
)
555 return BDEV_I(inode
)->bdev
.bd_dev
== *(dev_t
*)data
;
558 static int bdev_set(struct inode
*inode
, void *data
)
560 BDEV_I(inode
)->bdev
.bd_dev
= *(dev_t
*)data
;
564 static LIST_HEAD(all_bdevs
);
566 struct block_device
*bdget(dev_t dev
)
568 struct block_device
*bdev
;
571 inode
= iget5_locked(blockdev_superblock
, hash(dev
),
572 bdev_test
, bdev_set
, &dev
);
577 bdev
= &BDEV_I(inode
)->bdev
;
579 if (inode
->i_state
& I_NEW
) {
580 bdev
->bd_contains
= NULL
;
581 bdev
->bd_super
= NULL
;
582 bdev
->bd_inode
= inode
;
583 bdev
->bd_block_size
= (1 << inode
->i_blkbits
);
584 bdev
->bd_part_count
= 0;
585 bdev
->bd_invalidated
= 0;
586 inode
->i_mode
= S_IFBLK
;
588 inode
->i_bdev
= bdev
;
589 inode
->i_data
.a_ops
= &def_blk_aops
;
590 mapping_set_gfp_mask(&inode
->i_data
, GFP_USER
);
591 inode
->i_data
.backing_dev_info
= &default_backing_dev_info
;
592 spin_lock(&bdev_lock
);
593 list_add(&bdev
->bd_list
, &all_bdevs
);
594 spin_unlock(&bdev_lock
);
595 unlock_new_inode(inode
);
600 EXPORT_SYMBOL(bdget
);
603 * bdgrab -- Grab a reference to an already referenced block device
604 * @bdev: Block device to grab a reference to.
606 struct block_device
*bdgrab(struct block_device
*bdev
)
608 ihold(bdev
->bd_inode
);
611 EXPORT_SYMBOL(bdgrab
);
613 long nr_blockdev_pages(void)
615 struct block_device
*bdev
;
617 spin_lock(&bdev_lock
);
618 list_for_each_entry(bdev
, &all_bdevs
, bd_list
) {
619 ret
+= bdev
->bd_inode
->i_mapping
->nrpages
;
621 spin_unlock(&bdev_lock
);
625 void bdput(struct block_device
*bdev
)
627 iput(bdev
->bd_inode
);
630 EXPORT_SYMBOL(bdput
);
632 static struct block_device
*bd_acquire(struct inode
*inode
)
634 struct block_device
*bdev
;
636 spin_lock(&bdev_lock
);
637 bdev
= inode
->i_bdev
;
639 ihold(bdev
->bd_inode
);
640 spin_unlock(&bdev_lock
);
643 spin_unlock(&bdev_lock
);
645 bdev
= bdget(inode
->i_rdev
);
647 spin_lock(&bdev_lock
);
648 if (!inode
->i_bdev
) {
650 * We take an additional reference to bd_inode,
651 * and it's released in clear_inode() of inode.
652 * So, we can access it via ->i_mapping always
655 ihold(bdev
->bd_inode
);
656 inode
->i_bdev
= bdev
;
657 inode
->i_mapping
= bdev
->bd_inode
->i_mapping
;
658 list_add(&inode
->i_devices
, &bdev
->bd_inodes
);
660 spin_unlock(&bdev_lock
);
665 int sb_is_blkdev_sb(struct super_block
*sb
)
667 return sb
== blockdev_superblock
;
670 /* Call when you free inode */
672 void bd_forget(struct inode
*inode
)
674 struct block_device
*bdev
= NULL
;
676 spin_lock(&bdev_lock
);
677 if (!sb_is_blkdev_sb(inode
->i_sb
))
678 bdev
= inode
->i_bdev
;
680 spin_unlock(&bdev_lock
);
683 iput(bdev
->bd_inode
);
687 * bd_may_claim - test whether a block device can be claimed
688 * @bdev: block device of interest
689 * @whole: whole block device containing @bdev, may equal @bdev
690 * @holder: holder trying to claim @bdev
692 * Test whether @bdev can be claimed by @holder.
695 * spin_lock(&bdev_lock).
698 * %true if @bdev can be claimed, %false otherwise.
700 static bool bd_may_claim(struct block_device
*bdev
, struct block_device
*whole
,
703 if (bdev
->bd_holder
== holder
)
704 return true; /* already a holder */
705 else if (bdev
->bd_holder
!= NULL
)
706 return false; /* held by someone else */
707 else if (bdev
->bd_contains
== bdev
)
708 return true; /* is a whole device which isn't held */
710 else if (whole
->bd_holder
== bd_may_claim
)
711 return true; /* is a partition of a device that is being partitioned */
712 else if (whole
->bd_holder
!= NULL
)
713 return false; /* is a partition of a held device */
715 return true; /* is a partition of an un-held device */
719 * bd_prepare_to_claim - prepare to claim a block device
720 * @bdev: block device of interest
721 * @whole: the whole device containing @bdev, may equal @bdev
722 * @holder: holder trying to claim @bdev
724 * Prepare to claim @bdev. This function fails if @bdev is already
725 * claimed by another holder and waits if another claiming is in
726 * progress. This function doesn't actually claim. On successful
727 * return, the caller has ownership of bd_claiming and bd_holder[s].
730 * spin_lock(&bdev_lock). Might release bdev_lock, sleep and regrab
734 * 0 if @bdev can be claimed, -EBUSY otherwise.
736 static int bd_prepare_to_claim(struct block_device
*bdev
,
737 struct block_device
*whole
, void *holder
)
740 /* if someone else claimed, fail */
741 if (!bd_may_claim(bdev
, whole
, holder
))
744 /* if claiming is already in progress, wait for it to finish */
745 if (whole
->bd_claiming
) {
746 wait_queue_head_t
*wq
= bit_waitqueue(&whole
->bd_claiming
, 0);
749 prepare_to_wait(wq
, &wait
, TASK_UNINTERRUPTIBLE
);
750 spin_unlock(&bdev_lock
);
752 finish_wait(wq
, &wait
);
753 spin_lock(&bdev_lock
);
762 * bd_start_claiming - start claiming a block device
763 * @bdev: block device of interest
764 * @holder: holder trying to claim @bdev
766 * @bdev is about to be opened exclusively. Check @bdev can be opened
767 * exclusively and mark that an exclusive open is in progress. Each
768 * successful call to this function must be matched with a call to
769 * either bd_finish_claiming() or bd_abort_claiming() (which do not
772 * This function is used to gain exclusive access to the block device
773 * without actually causing other exclusive open attempts to fail. It
774 * should be used when the open sequence itself requires exclusive
775 * access but may subsequently fail.
781 * Pointer to the block device containing @bdev on success, ERR_PTR()
784 static struct block_device
*bd_start_claiming(struct block_device
*bdev
,
787 struct gendisk
*disk
;
788 struct block_device
*whole
;
794 * @bdev might not have been initialized properly yet, look up
795 * and grab the outer block device the hard way.
797 disk
= get_gendisk(bdev
->bd_dev
, &partno
);
799 return ERR_PTR(-ENXIO
);
802 * Normally, @bdev should equal what's returned from bdget_disk()
803 * if partno is 0; however, some drivers (floppy) use multiple
804 * bdev's for the same physical device and @bdev may be one of the
805 * aliases. Keep @bdev if partno is 0. This means claimer
806 * tracking is broken for those devices but it has always been that
810 whole
= bdget_disk(disk
, 0);
812 whole
= bdgrab(bdev
);
814 module_put(disk
->fops
->owner
);
817 return ERR_PTR(-ENOMEM
);
819 /* prepare to claim, if successful, mark claiming in progress */
820 spin_lock(&bdev_lock
);
822 err
= bd_prepare_to_claim(bdev
, whole
, holder
);
824 whole
->bd_claiming
= holder
;
825 spin_unlock(&bdev_lock
);
828 spin_unlock(&bdev_lock
);
835 struct bd_holder_disk
{
836 struct list_head list
;
837 struct gendisk
*disk
;
841 static struct bd_holder_disk
*bd_find_holder_disk(struct block_device
*bdev
,
842 struct gendisk
*disk
)
844 struct bd_holder_disk
*holder
;
846 list_for_each_entry(holder
, &bdev
->bd_holder_disks
, list
)
847 if (holder
->disk
== disk
)
852 static int add_symlink(struct kobject
*from
, struct kobject
*to
)
854 return sysfs_create_link(from
, to
, kobject_name(to
));
857 static void del_symlink(struct kobject
*from
, struct kobject
*to
)
859 sysfs_remove_link(from
, kobject_name(to
));
863 * bd_link_disk_holder - create symlinks between holding disk and slave bdev
864 * @bdev: the claimed slave bdev
865 * @disk: the holding disk
867 * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT.
869 * This functions creates the following sysfs symlinks.
871 * - from "slaves" directory of the holder @disk to the claimed @bdev
872 * - from "holders" directory of the @bdev to the holder @disk
874 * For example, if /dev/dm-0 maps to /dev/sda and disk for dm-0 is
875 * passed to bd_link_disk_holder(), then:
877 * /sys/block/dm-0/slaves/sda --> /sys/block/sda
878 * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0
880 * The caller must have claimed @bdev before calling this function and
881 * ensure that both @bdev and @disk are valid during the creation and
882 * lifetime of these symlinks.
888 * 0 on success, -errno on failure.
890 int bd_link_disk_holder(struct block_device
*bdev
, struct gendisk
*disk
)
892 struct bd_holder_disk
*holder
;
895 mutex_lock(&bdev
->bd_mutex
);
897 WARN_ON_ONCE(!bdev
->bd_holder
);
899 /* FIXME: remove the following once add_disk() handles errors */
900 if (WARN_ON(!disk
->slave_dir
|| !bdev
->bd_part
->holder_dir
))
903 holder
= bd_find_holder_disk(bdev
, disk
);
909 holder
= kzalloc(sizeof(*holder
), GFP_KERNEL
);
915 INIT_LIST_HEAD(&holder
->list
);
919 ret
= add_symlink(disk
->slave_dir
, &part_to_dev(bdev
->bd_part
)->kobj
);
923 ret
= add_symlink(bdev
->bd_part
->holder_dir
, &disk_to_dev(disk
)->kobj
);
927 * bdev could be deleted beneath us which would implicitly destroy
928 * the holder directory. Hold on to it.
930 kobject_get(bdev
->bd_part
->holder_dir
);
932 list_add(&holder
->list
, &bdev
->bd_holder_disks
);
936 del_symlink(disk
->slave_dir
, &part_to_dev(bdev
->bd_part
)->kobj
);
940 mutex_unlock(&bdev
->bd_mutex
);
943 EXPORT_SYMBOL_GPL(bd_link_disk_holder
);
946 * bd_unlink_disk_holder - destroy symlinks created by bd_link_disk_holder()
947 * @bdev: the calimed slave bdev
948 * @disk: the holding disk
950 * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT.
955 void bd_unlink_disk_holder(struct block_device
*bdev
, struct gendisk
*disk
)
957 struct bd_holder_disk
*holder
;
959 mutex_lock(&bdev
->bd_mutex
);
961 holder
= bd_find_holder_disk(bdev
, disk
);
963 if (!WARN_ON_ONCE(holder
== NULL
) && !--holder
->refcnt
) {
964 del_symlink(disk
->slave_dir
, &part_to_dev(bdev
->bd_part
)->kobj
);
965 del_symlink(bdev
->bd_part
->holder_dir
,
966 &disk_to_dev(disk
)->kobj
);
967 kobject_put(bdev
->bd_part
->holder_dir
);
968 list_del_init(&holder
->list
);
972 mutex_unlock(&bdev
->bd_mutex
);
974 EXPORT_SYMBOL_GPL(bd_unlink_disk_holder
);
978 * flush_disk - invalidates all buffer-cache entries on a disk
980 * @bdev: struct block device to be flushed
981 * @kill_dirty: flag to guide handling of dirty inodes
983 * Invalidates all buffer-cache entries on a disk. It should be called
984 * when a disk has been changed -- either by a media change or online
987 static void flush_disk(struct block_device
*bdev
, bool kill_dirty
)
989 if (__invalidate_device(bdev
, kill_dirty
)) {
990 char name
[BDEVNAME_SIZE
] = "";
993 disk_name(bdev
->bd_disk
, 0, name
);
994 printk(KERN_WARNING
"VFS: busy inodes on changed media or "
995 "resized disk %s\n", name
);
1000 if (disk_part_scan_enabled(bdev
->bd_disk
))
1001 bdev
->bd_invalidated
= 1;
1005 * check_disk_size_change - checks for disk size change and adjusts bdev size.
1006 * @disk: struct gendisk to check
1007 * @bdev: struct bdev to adjust.
1009 * This routine checks to see if the bdev size does not match the disk size
1010 * and adjusts it if it differs.
1012 void check_disk_size_change(struct gendisk
*disk
, struct block_device
*bdev
)
1014 loff_t disk_size
, bdev_size
;
1016 disk_size
= (loff_t
)get_capacity(disk
) << 9;
1017 bdev_size
= i_size_read(bdev
->bd_inode
);
1018 if (disk_size
!= bdev_size
) {
1019 char name
[BDEVNAME_SIZE
];
1021 disk_name(disk
, 0, name
);
1023 "%s: detected capacity change from %lld to %lld\n",
1024 name
, bdev_size
, disk_size
);
1025 i_size_write(bdev
->bd_inode
, disk_size
);
1026 flush_disk(bdev
, false);
1029 EXPORT_SYMBOL(check_disk_size_change
);
1032 * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back
1033 * @disk: struct gendisk to be revalidated
1035 * This routine is a wrapper for lower-level driver's revalidate_disk
1036 * call-backs. It is used to do common pre and post operations needed
1037 * for all revalidate_disk operations.
1039 int revalidate_disk(struct gendisk
*disk
)
1041 struct block_device
*bdev
;
1044 if (disk
->fops
->revalidate_disk
)
1045 ret
= disk
->fops
->revalidate_disk(disk
);
1047 bdev
= bdget_disk(disk
, 0);
1051 mutex_lock(&bdev
->bd_mutex
);
1052 check_disk_size_change(disk
, bdev
);
1053 bdev
->bd_invalidated
= 0;
1054 mutex_unlock(&bdev
->bd_mutex
);
1058 EXPORT_SYMBOL(revalidate_disk
);
1061 * This routine checks whether a removable media has been changed,
1062 * and invalidates all buffer-cache-entries in that case. This
1063 * is a relatively slow routine, so we have to try to minimize using
1064 * it. Thus it is called only upon a 'mount' or 'open'. This
1065 * is the best way of combining speed and utility, I think.
1066 * People changing diskettes in the middle of an operation deserve
1069 int check_disk_change(struct block_device
*bdev
)
1071 struct gendisk
*disk
= bdev
->bd_disk
;
1072 const struct block_device_operations
*bdops
= disk
->fops
;
1073 unsigned int events
;
1075 events
= disk_clear_events(disk
, DISK_EVENT_MEDIA_CHANGE
|
1076 DISK_EVENT_EJECT_REQUEST
);
1077 if (!(events
& DISK_EVENT_MEDIA_CHANGE
))
1080 flush_disk(bdev
, true);
1081 if (bdops
->revalidate_disk
)
1082 bdops
->revalidate_disk(bdev
->bd_disk
);
1086 EXPORT_SYMBOL(check_disk_change
);
1088 void bd_set_size(struct block_device
*bdev
, loff_t size
)
1090 unsigned bsize
= bdev_logical_block_size(bdev
);
1092 mutex_lock(&bdev
->bd_inode
->i_mutex
);
1093 i_size_write(bdev
->bd_inode
, size
);
1094 mutex_unlock(&bdev
->bd_inode
->i_mutex
);
1095 while (bsize
< PAGE_CACHE_SIZE
) {
1100 bdev
->bd_block_size
= bsize
;
1101 bdev
->bd_inode
->i_blkbits
= blksize_bits(bsize
);
1103 EXPORT_SYMBOL(bd_set_size
);
1105 static void __blkdev_put(struct block_device
*bdev
, fmode_t mode
, int for_part
);
1110 * mutex_lock(part->bd_mutex)
1111 * mutex_lock_nested(whole->bd_mutex, 1)
1114 static int __blkdev_get(struct block_device
*bdev
, fmode_t mode
, int for_part
)
1116 struct gendisk
*disk
;
1117 struct module
*owner
;
1122 if (mode
& FMODE_READ
)
1124 if (mode
& FMODE_WRITE
)
1127 * hooks: /n/, see "layering violations".
1130 ret
= devcgroup_inode_permission(bdev
->bd_inode
, perm
);
1140 disk
= get_gendisk(bdev
->bd_dev
, &partno
);
1143 owner
= disk
->fops
->owner
;
1145 disk_block_events(disk
);
1146 mutex_lock_nested(&bdev
->bd_mutex
, for_part
);
1147 if (!bdev
->bd_openers
) {
1148 bdev
->bd_disk
= disk
;
1149 bdev
->bd_queue
= disk
->queue
;
1150 bdev
->bd_contains
= bdev
;
1152 struct backing_dev_info
*bdi
;
1155 bdev
->bd_part
= disk_get_part(disk
, partno
);
1160 if (disk
->fops
->open
) {
1161 ret
= disk
->fops
->open(bdev
, mode
);
1162 if (ret
== -ERESTARTSYS
) {
1163 /* Lost a race with 'disk' being
1164 * deleted, try again.
1167 disk_put_part(bdev
->bd_part
);
1168 bdev
->bd_part
= NULL
;
1169 bdev
->bd_disk
= NULL
;
1170 bdev
->bd_queue
= NULL
;
1171 mutex_unlock(&bdev
->bd_mutex
);
1172 disk_unblock_events(disk
);
1180 bd_set_size(bdev
,(loff_t
)get_capacity(disk
)<<9);
1181 bdi
= blk_get_backing_dev_info(bdev
);
1183 bdi
= &default_backing_dev_info
;
1184 bdev_inode_switch_bdi(bdev
->bd_inode
, bdi
);
1188 * If the device is invalidated, rescan partition
1189 * if open succeeded or failed with -ENOMEDIUM.
1190 * The latter is necessary to prevent ghost
1191 * partitions on a removed medium.
1193 if (bdev
->bd_invalidated
) {
1195 rescan_partitions(disk
, bdev
);
1196 else if (ret
== -ENOMEDIUM
)
1197 invalidate_partitions(disk
, bdev
);
1202 struct block_device
*whole
;
1203 whole
= bdget_disk(disk
, 0);
1208 ret
= __blkdev_get(whole
, mode
, 1);
1211 bdev
->bd_contains
= whole
;
1212 bdev_inode_switch_bdi(bdev
->bd_inode
,
1213 whole
->bd_inode
->i_data
.backing_dev_info
);
1214 bdev
->bd_part
= disk_get_part(disk
, partno
);
1215 if (!(disk
->flags
& GENHD_FL_UP
) ||
1216 !bdev
->bd_part
|| !bdev
->bd_part
->nr_sects
) {
1220 bd_set_size(bdev
, (loff_t
)bdev
->bd_part
->nr_sects
<< 9);
1223 if (bdev
->bd_contains
== bdev
) {
1225 if (bdev
->bd_disk
->fops
->open
)
1226 ret
= bdev
->bd_disk
->fops
->open(bdev
, mode
);
1227 /* the same as first opener case, read comment there */
1228 if (bdev
->bd_invalidated
) {
1230 rescan_partitions(bdev
->bd_disk
, bdev
);
1231 else if (ret
== -ENOMEDIUM
)
1232 invalidate_partitions(bdev
->bd_disk
, bdev
);
1235 goto out_unlock_bdev
;
1237 /* only one opener holds refs to the module and disk */
1243 bdev
->bd_part_count
++;
1244 mutex_unlock(&bdev
->bd_mutex
);
1245 disk_unblock_events(disk
);
1249 disk_put_part(bdev
->bd_part
);
1250 bdev
->bd_disk
= NULL
;
1251 bdev
->bd_part
= NULL
;
1252 bdev
->bd_queue
= NULL
;
1253 bdev_inode_switch_bdi(bdev
->bd_inode
, &default_backing_dev_info
);
1254 if (bdev
!= bdev
->bd_contains
)
1255 __blkdev_put(bdev
->bd_contains
, mode
, 1);
1256 bdev
->bd_contains
= NULL
;
1258 mutex_unlock(&bdev
->bd_mutex
);
1259 disk_unblock_events(disk
);
1269 * blkdev_get - open a block device
1270 * @bdev: block_device to open
1271 * @mode: FMODE_* mask
1272 * @holder: exclusive holder identifier
1274 * Open @bdev with @mode. If @mode includes %FMODE_EXCL, @bdev is
1275 * open with exclusive access. Specifying %FMODE_EXCL with %NULL
1276 * @holder is invalid. Exclusive opens may nest for the same @holder.
1278 * On success, the reference count of @bdev is unchanged. On failure,
1285 * 0 on success, -errno on failure.
1287 int blkdev_get(struct block_device
*bdev
, fmode_t mode
, void *holder
)
1289 struct block_device
*whole
= NULL
;
1292 WARN_ON_ONCE((mode
& FMODE_EXCL
) && !holder
);
1294 if ((mode
& FMODE_EXCL
) && holder
) {
1295 whole
= bd_start_claiming(bdev
, holder
);
1296 if (IS_ERR(whole
)) {
1298 return PTR_ERR(whole
);
1302 res
= __blkdev_get(bdev
, mode
, 0);
1305 struct gendisk
*disk
= whole
->bd_disk
;
1307 /* finish claiming */
1308 mutex_lock(&bdev
->bd_mutex
);
1309 spin_lock(&bdev_lock
);
1312 BUG_ON(!bd_may_claim(bdev
, whole
, holder
));
1314 * Note that for a whole device bd_holders
1315 * will be incremented twice, and bd_holder
1316 * will be set to bd_may_claim before being
1319 whole
->bd_holders
++;
1320 whole
->bd_holder
= bd_may_claim
;
1322 bdev
->bd_holder
= holder
;
1325 /* tell others that we're done */
1326 BUG_ON(whole
->bd_claiming
!= holder
);
1327 whole
->bd_claiming
= NULL
;
1328 wake_up_bit(&whole
->bd_claiming
, 0);
1330 spin_unlock(&bdev_lock
);
1333 * Block event polling for write claims if requested. Any
1334 * write holder makes the write_holder state stick until
1335 * all are released. This is good enough and tracking
1336 * individual writeable reference is too fragile given the
1337 * way @mode is used in blkdev_get/put().
1339 if (!res
&& (mode
& FMODE_WRITE
) && !bdev
->bd_write_holder
&&
1340 (disk
->flags
& GENHD_FL_BLOCK_EVENTS_ON_EXCL_WRITE
)) {
1341 bdev
->bd_write_holder
= true;
1342 disk_block_events(disk
);
1345 mutex_unlock(&bdev
->bd_mutex
);
1351 EXPORT_SYMBOL(blkdev_get
);
1354 * blkdev_get_by_path - open a block device by name
1355 * @path: path to the block device to open
1356 * @mode: FMODE_* mask
1357 * @holder: exclusive holder identifier
1359 * Open the blockdevice described by the device file at @path. @mode
1360 * and @holder are identical to blkdev_get().
1362 * On success, the returned block_device has reference count of one.
1368 * Pointer to block_device on success, ERR_PTR(-errno) on failure.
1370 struct block_device
*blkdev_get_by_path(const char *path
, fmode_t mode
,
1373 struct block_device
*bdev
;
1376 bdev
= lookup_bdev(path
);
1380 err
= blkdev_get(bdev
, mode
, holder
);
1382 return ERR_PTR(err
);
1384 if ((mode
& FMODE_WRITE
) && bdev_read_only(bdev
)) {
1385 blkdev_put(bdev
, mode
);
1386 return ERR_PTR(-EACCES
);
1391 EXPORT_SYMBOL(blkdev_get_by_path
);
1394 * blkdev_get_by_dev - open a block device by device number
1395 * @dev: device number of block device to open
1396 * @mode: FMODE_* mask
1397 * @holder: exclusive holder identifier
1399 * Open the blockdevice described by device number @dev. @mode and
1400 * @holder are identical to blkdev_get().
1402 * Use it ONLY if you really do not have anything better - i.e. when
1403 * you are behind a truly sucky interface and all you are given is a
1404 * device number. _Never_ to be used for internal purposes. If you
1405 * ever need it - reconsider your API.
1407 * On success, the returned block_device has reference count of one.
1413 * Pointer to block_device on success, ERR_PTR(-errno) on failure.
1415 struct block_device
*blkdev_get_by_dev(dev_t dev
, fmode_t mode
, void *holder
)
1417 struct block_device
*bdev
;
1422 return ERR_PTR(-ENOMEM
);
1424 err
= blkdev_get(bdev
, mode
, holder
);
1426 return ERR_PTR(err
);
1430 EXPORT_SYMBOL(blkdev_get_by_dev
);
1432 static int blkdev_open(struct inode
* inode
, struct file
* filp
)
1434 struct block_device
*bdev
;
1437 * Preserve backwards compatibility and allow large file access
1438 * even if userspace doesn't ask for it explicitly. Some mkfs
1439 * binary needs it. We might want to drop this workaround
1440 * during an unstable branch.
1442 filp
->f_flags
|= O_LARGEFILE
;
1444 if (filp
->f_flags
& O_NDELAY
)
1445 filp
->f_mode
|= FMODE_NDELAY
;
1446 if (filp
->f_flags
& O_EXCL
)
1447 filp
->f_mode
|= FMODE_EXCL
;
1448 if ((filp
->f_flags
& O_ACCMODE
) == 3)
1449 filp
->f_mode
|= FMODE_WRITE_IOCTL
;
1451 bdev
= bd_acquire(inode
);
1455 filp
->f_mapping
= bdev
->bd_inode
->i_mapping
;
1457 return blkdev_get(bdev
, filp
->f_mode
, filp
);
1460 static void __blkdev_put(struct block_device
*bdev
, fmode_t mode
, int for_part
)
1462 struct gendisk
*disk
= bdev
->bd_disk
;
1463 struct block_device
*victim
= NULL
;
1465 mutex_lock_nested(&bdev
->bd_mutex
, for_part
);
1467 bdev
->bd_part_count
--;
1469 if (!--bdev
->bd_openers
) {
1470 WARN_ON_ONCE(bdev
->bd_holders
);
1471 sync_blockdev(bdev
);
1473 /* ->release can cause the old bdi to disappear,
1474 * so must switch it out first
1476 bdev_inode_switch_bdi(bdev
->bd_inode
,
1477 &default_backing_dev_info
);
1479 if (bdev
->bd_contains
== bdev
) {
1480 if (disk
->fops
->release
)
1481 disk
->fops
->release(disk
, mode
);
1483 if (!bdev
->bd_openers
) {
1484 struct module
*owner
= disk
->fops
->owner
;
1486 disk_put_part(bdev
->bd_part
);
1487 bdev
->bd_part
= NULL
;
1488 bdev
->bd_disk
= NULL
;
1489 if (bdev
!= bdev
->bd_contains
)
1490 victim
= bdev
->bd_contains
;
1491 bdev
->bd_contains
= NULL
;
1496 mutex_unlock(&bdev
->bd_mutex
);
1499 __blkdev_put(victim
, mode
, 1);
1502 void blkdev_put(struct block_device
*bdev
, fmode_t mode
)
1504 mutex_lock(&bdev
->bd_mutex
);
1506 if (mode
& FMODE_EXCL
) {
1510 * Release a claim on the device. The holder fields
1511 * are protected with bdev_lock. bd_mutex is to
1512 * synchronize disk_holder unlinking.
1514 spin_lock(&bdev_lock
);
1516 WARN_ON_ONCE(--bdev
->bd_holders
< 0);
1517 WARN_ON_ONCE(--bdev
->bd_contains
->bd_holders
< 0);
1519 /* bd_contains might point to self, check in a separate step */
1520 if ((bdev_free
= !bdev
->bd_holders
))
1521 bdev
->bd_holder
= NULL
;
1522 if (!bdev
->bd_contains
->bd_holders
)
1523 bdev
->bd_contains
->bd_holder
= NULL
;
1525 spin_unlock(&bdev_lock
);
1528 * If this was the last claim, remove holder link and
1529 * unblock evpoll if it was a write holder.
1531 if (bdev_free
&& bdev
->bd_write_holder
) {
1532 disk_unblock_events(bdev
->bd_disk
);
1533 bdev
->bd_write_holder
= false;
1538 * Trigger event checking and tell drivers to flush MEDIA_CHANGE
1539 * event. This is to ensure detection of media removal commanded
1540 * from userland - e.g. eject(1).
1542 disk_flush_events(bdev
->bd_disk
, DISK_EVENT_MEDIA_CHANGE
);
1544 mutex_unlock(&bdev
->bd_mutex
);
1546 __blkdev_put(bdev
, mode
, 0);
1548 EXPORT_SYMBOL(blkdev_put
);
1550 static int blkdev_close(struct inode
* inode
, struct file
* filp
)
1552 struct block_device
*bdev
= I_BDEV(filp
->f_mapping
->host
);
1553 blkdev_put(bdev
, filp
->f_mode
);
1557 static long block_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1559 struct block_device
*bdev
= I_BDEV(file
->f_mapping
->host
);
1560 fmode_t mode
= file
->f_mode
;
1563 * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have
1564 * to updated it before every ioctl.
1566 if (file
->f_flags
& O_NDELAY
)
1567 mode
|= FMODE_NDELAY
;
1569 mode
&= ~FMODE_NDELAY
;
1571 return blkdev_ioctl(bdev
, mode
, cmd
, arg
);
1575 * Write data to the block device. Only intended for the block device itself
1576 * and the raw driver which basically is a fake block device.
1578 * Does not take i_mutex for the write and thus is not for general purpose
1581 ssize_t
blkdev_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
1583 struct file
*file
= iocb
->ki_filp
;
1584 struct blk_plug plug
;
1587 blk_start_plug(&plug
);
1588 ret
= __generic_file_write_iter(iocb
, from
);
1591 err
= generic_write_sync(file
, iocb
->ki_pos
- ret
, ret
);
1595 blk_finish_plug(&plug
);
1598 EXPORT_SYMBOL_GPL(blkdev_write_iter
);
1600 static ssize_t
blkdev_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
1602 struct file
*file
= iocb
->ki_filp
;
1603 struct inode
*bd_inode
= file
->f_mapping
->host
;
1604 loff_t size
= i_size_read(bd_inode
);
1605 loff_t pos
= iocb
->ki_pos
;
1611 iov_iter_truncate(to
, size
);
1612 return generic_file_read_iter(iocb
, to
);
1616 * Try to release a page associated with block device when the system
1617 * is under memory pressure.
1619 static int blkdev_releasepage(struct page
*page
, gfp_t wait
)
1621 struct super_block
*super
= BDEV_I(page
->mapping
->host
)->bdev
.bd_super
;
1623 if (super
&& super
->s_op
->bdev_try_to_free_page
)
1624 return super
->s_op
->bdev_try_to_free_page(super
, page
, wait
);
1626 return try_to_free_buffers(page
);
1629 static const struct address_space_operations def_blk_aops
= {
1630 .readpage
= blkdev_readpage
,
1631 .readpages
= blkdev_readpages
,
1632 .writepage
= blkdev_writepage
,
1633 .write_begin
= blkdev_write_begin
,
1634 .write_end
= blkdev_write_end
,
1635 .writepages
= generic_writepages
,
1636 .releasepage
= blkdev_releasepage
,
1637 .direct_IO
= blkdev_direct_IO
,
1638 .is_dirty_writeback
= buffer_check_dirty_writeback
,
1641 const struct file_operations def_blk_fops
= {
1642 .open
= blkdev_open
,
1643 .release
= blkdev_close
,
1644 .llseek
= block_llseek
,
1645 .read
= new_sync_read
,
1646 .write
= new_sync_write
,
1647 .read_iter
= blkdev_read_iter
,
1648 .write_iter
= blkdev_write_iter
,
1649 .mmap
= generic_file_mmap
,
1650 .fsync
= blkdev_fsync
,
1651 .unlocked_ioctl
= block_ioctl
,
1652 #ifdef CONFIG_COMPAT
1653 .compat_ioctl
= compat_blkdev_ioctl
,
1655 .splice_read
= generic_file_splice_read
,
1656 .splice_write
= iter_file_splice_write
,
1659 int ioctl_by_bdev(struct block_device
*bdev
, unsigned cmd
, unsigned long arg
)
1662 mm_segment_t old_fs
= get_fs();
1664 res
= blkdev_ioctl(bdev
, 0, cmd
, arg
);
1669 EXPORT_SYMBOL(ioctl_by_bdev
);
1672 * lookup_bdev - lookup a struct block_device by name
1673 * @pathname: special file representing the block device
1675 * Get a reference to the blockdevice at @pathname in the current
1676 * namespace if possible and return it. Return ERR_PTR(error)
1679 struct block_device
*lookup_bdev(const char *pathname
)
1681 struct block_device
*bdev
;
1682 struct inode
*inode
;
1686 if (!pathname
|| !*pathname
)
1687 return ERR_PTR(-EINVAL
);
1689 error
= kern_path(pathname
, LOOKUP_FOLLOW
, &path
);
1691 return ERR_PTR(error
);
1693 inode
= path
.dentry
->d_inode
;
1695 if (!S_ISBLK(inode
->i_mode
))
1698 if (path
.mnt
->mnt_flags
& MNT_NODEV
)
1701 bdev
= bd_acquire(inode
);
1708 bdev
= ERR_PTR(error
);
1711 EXPORT_SYMBOL(lookup_bdev
);
1713 int __invalidate_device(struct block_device
*bdev
, bool kill_dirty
)
1715 struct super_block
*sb
= get_super(bdev
);
1720 * no need to lock the super, get_super holds the
1721 * read mutex so the filesystem cannot go away
1722 * under us (->put_super runs with the write lock
1725 shrink_dcache_sb(sb
);
1726 res
= invalidate_inodes(sb
, kill_dirty
);
1729 invalidate_bdev(bdev
);
1732 EXPORT_SYMBOL(__invalidate_device
);
1734 void iterate_bdevs(void (*func
)(struct block_device
*, void *), void *arg
)
1736 struct inode
*inode
, *old_inode
= NULL
;
1738 spin_lock(&inode_sb_list_lock
);
1739 list_for_each_entry(inode
, &blockdev_superblock
->s_inodes
, i_sb_list
) {
1740 struct address_space
*mapping
= inode
->i_mapping
;
1742 spin_lock(&inode
->i_lock
);
1743 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
|I_NEW
) ||
1744 mapping
->nrpages
== 0) {
1745 spin_unlock(&inode
->i_lock
);
1749 spin_unlock(&inode
->i_lock
);
1750 spin_unlock(&inode_sb_list_lock
);
1752 * We hold a reference to 'inode' so it couldn't have been
1753 * removed from s_inodes list while we dropped the
1754 * inode_sb_list_lock. We cannot iput the inode now as we can
1755 * be holding the last reference and we cannot iput it under
1756 * inode_sb_list_lock. So we keep the reference and iput it
1762 func(I_BDEV(inode
), arg
);
1764 spin_lock(&inode_sb_list_lock
);
1766 spin_unlock(&inode_sb_list_lock
);