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 <asm/uaccess.h>
34 struct block_device bdev
;
35 struct inode vfs_inode
;
38 static const struct address_space_operations def_blk_aops
;
40 static inline struct bdev_inode
*BDEV_I(struct inode
*inode
)
42 return container_of(inode
, struct bdev_inode
, vfs_inode
);
45 inline struct block_device
*I_BDEV(struct inode
*inode
)
47 return &BDEV_I(inode
)->bdev
;
49 EXPORT_SYMBOL(I_BDEV
);
51 static void bdev_write_inode(struct inode
*inode
)
53 spin_lock(&inode
->i_lock
);
54 while (inode
->i_state
& I_DIRTY
) {
55 spin_unlock(&inode
->i_lock
);
56 WARN_ON_ONCE(write_inode_now(inode
, true));
57 spin_lock(&inode
->i_lock
);
59 spin_unlock(&inode
->i_lock
);
62 /* Kill _all_ buffers and pagecache , dirty or not.. */
63 void kill_bdev(struct block_device
*bdev
)
65 struct address_space
*mapping
= bdev
->bd_inode
->i_mapping
;
67 if (mapping
->nrpages
== 0 && mapping
->nrshadows
== 0)
71 truncate_inode_pages(mapping
, 0);
73 EXPORT_SYMBOL(kill_bdev
);
75 /* Invalidate clean unused buffers and pagecache. */
76 void invalidate_bdev(struct block_device
*bdev
)
78 struct address_space
*mapping
= bdev
->bd_inode
->i_mapping
;
80 if (mapping
->nrpages
== 0)
84 lru_add_drain_all(); /* make sure all lru add caches are flushed */
85 invalidate_mapping_pages(mapping
, 0, -1);
86 /* 99% of the time, we don't need to flush the cleancache on the bdev.
87 * But, for the strange corners, lets be cautious
89 cleancache_invalidate_inode(mapping
);
91 EXPORT_SYMBOL(invalidate_bdev
);
93 int set_blocksize(struct block_device
*bdev
, int size
)
95 /* Size must be a power of two, and between 512 and PAGE_SIZE */
96 if (size
> PAGE_SIZE
|| size
< 512 || !is_power_of_2(size
))
99 /* Size cannot be smaller than the size supported by the device */
100 if (size
< bdev_logical_block_size(bdev
))
103 /* Don't change the size if it is same as current */
104 if (bdev
->bd_block_size
!= size
) {
106 bdev
->bd_block_size
= size
;
107 bdev
->bd_inode
->i_blkbits
= blksize_bits(size
);
113 EXPORT_SYMBOL(set_blocksize
);
115 int sb_set_blocksize(struct super_block
*sb
, int size
)
117 if (set_blocksize(sb
->s_bdev
, size
))
119 /* If we get here, we know size is power of two
120 * and it's value is between 512 and PAGE_SIZE */
121 sb
->s_blocksize
= size
;
122 sb
->s_blocksize_bits
= blksize_bits(size
);
123 return sb
->s_blocksize
;
126 EXPORT_SYMBOL(sb_set_blocksize
);
128 int sb_min_blocksize(struct super_block
*sb
, int size
)
130 int minsize
= bdev_logical_block_size(sb
->s_bdev
);
133 return sb_set_blocksize(sb
, size
);
136 EXPORT_SYMBOL(sb_min_blocksize
);
139 blkdev_get_block(struct inode
*inode
, sector_t iblock
,
140 struct buffer_head
*bh
, int create
)
142 bh
->b_bdev
= I_BDEV(inode
);
143 bh
->b_blocknr
= iblock
;
144 set_buffer_mapped(bh
);
149 blkdev_direct_IO(struct kiocb
*iocb
, struct iov_iter
*iter
, loff_t offset
)
151 struct file
*file
= iocb
->ki_filp
;
152 struct inode
*inode
= file
->f_mapping
->host
;
154 return __blockdev_direct_IO(iocb
, inode
, I_BDEV(inode
), iter
, offset
,
155 blkdev_get_block
, NULL
, NULL
, 0);
158 int __sync_blockdev(struct block_device
*bdev
, int wait
)
163 return filemap_flush(bdev
->bd_inode
->i_mapping
);
164 return filemap_write_and_wait(bdev
->bd_inode
->i_mapping
);
168 * Write out and wait upon all the dirty data associated with a block
169 * device via its mapping. Does not take the superblock lock.
171 int sync_blockdev(struct block_device
*bdev
)
173 return __sync_blockdev(bdev
, 1);
175 EXPORT_SYMBOL(sync_blockdev
);
178 * Write out and wait upon all dirty data associated with this
179 * device. Filesystem data as well as the underlying block
180 * device. Takes the superblock lock.
182 int fsync_bdev(struct block_device
*bdev
)
184 struct super_block
*sb
= get_super(bdev
);
186 int res
= sync_filesystem(sb
);
190 return sync_blockdev(bdev
);
192 EXPORT_SYMBOL(fsync_bdev
);
195 * freeze_bdev -- lock a filesystem and force it into a consistent state
196 * @bdev: blockdevice to lock
198 * If a superblock is found on this device, we take the s_umount semaphore
199 * on it to make sure nobody unmounts until the snapshot creation is done.
200 * The reference counter (bd_fsfreeze_count) guarantees that only the last
201 * unfreeze process can unfreeze the frozen filesystem actually when multiple
202 * freeze requests arrive simultaneously. It counts up in freeze_bdev() and
203 * count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze
206 struct super_block
*freeze_bdev(struct block_device
*bdev
)
208 struct super_block
*sb
;
211 mutex_lock(&bdev
->bd_fsfreeze_mutex
);
212 if (++bdev
->bd_fsfreeze_count
> 1) {
214 * We don't even need to grab a reference - the first call
215 * to freeze_bdev grab an active reference and only the last
216 * thaw_bdev drops it.
218 sb
= get_super(bdev
);
220 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
224 sb
= get_active_super(bdev
);
227 if (sb
->s_op
->freeze_super
)
228 error
= sb
->s_op
->freeze_super(sb
);
230 error
= freeze_super(sb
);
232 deactivate_super(sb
);
233 bdev
->bd_fsfreeze_count
--;
234 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
235 return ERR_PTR(error
);
237 deactivate_super(sb
);
240 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
241 return sb
; /* thaw_bdev releases s->s_umount */
243 EXPORT_SYMBOL(freeze_bdev
);
246 * thaw_bdev -- unlock filesystem
247 * @bdev: blockdevice to unlock
248 * @sb: associated superblock
250 * Unlocks the filesystem and marks it writeable again after freeze_bdev().
252 int thaw_bdev(struct block_device
*bdev
, struct super_block
*sb
)
256 mutex_lock(&bdev
->bd_fsfreeze_mutex
);
257 if (!bdev
->bd_fsfreeze_count
)
261 if (--bdev
->bd_fsfreeze_count
> 0)
267 if (sb
->s_op
->thaw_super
)
268 error
= sb
->s_op
->thaw_super(sb
);
270 error
= thaw_super(sb
);
272 bdev
->bd_fsfreeze_count
++;
273 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
277 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
280 EXPORT_SYMBOL(thaw_bdev
);
282 static int blkdev_writepage(struct page
*page
, struct writeback_control
*wbc
)
284 return block_write_full_page(page
, blkdev_get_block
, wbc
);
287 static int blkdev_readpage(struct file
* file
, struct page
* page
)
289 return block_read_full_page(page
, blkdev_get_block
);
292 static int blkdev_readpages(struct file
*file
, struct address_space
*mapping
,
293 struct list_head
*pages
, unsigned nr_pages
)
295 return mpage_readpages(mapping
, pages
, nr_pages
, blkdev_get_block
);
298 static int blkdev_write_begin(struct file
*file
, struct address_space
*mapping
,
299 loff_t pos
, unsigned len
, unsigned flags
,
300 struct page
**pagep
, void **fsdata
)
302 return block_write_begin(mapping
, pos
, len
, flags
, pagep
,
306 static int blkdev_write_end(struct file
*file
, struct address_space
*mapping
,
307 loff_t pos
, unsigned len
, unsigned copied
,
308 struct page
*page
, void *fsdata
)
311 ret
= block_write_end(file
, mapping
, pos
, len
, copied
, page
, fsdata
);
314 page_cache_release(page
);
321 * for a block special file file_inode(file)->i_size is zero
322 * so we compute the size by hand (just as in block_read/write above)
324 static loff_t
block_llseek(struct file
*file
, loff_t offset
, int whence
)
326 struct inode
*bd_inode
= file
->f_mapping
->host
;
329 mutex_lock(&bd_inode
->i_mutex
);
330 retval
= fixed_size_llseek(file
, offset
, whence
, i_size_read(bd_inode
));
331 mutex_unlock(&bd_inode
->i_mutex
);
335 int blkdev_fsync(struct file
*filp
, loff_t start
, loff_t end
, int datasync
)
337 struct inode
*bd_inode
= filp
->f_mapping
->host
;
338 struct block_device
*bdev
= I_BDEV(bd_inode
);
341 error
= filemap_write_and_wait_range(filp
->f_mapping
, start
, end
);
346 * There is no need to serialise calls to blkdev_issue_flush with
347 * i_mutex and doing so causes performance issues with concurrent
348 * O_SYNC writers to a block device.
350 error
= blkdev_issue_flush(bdev
, GFP_KERNEL
, NULL
);
351 if (error
== -EOPNOTSUPP
)
356 EXPORT_SYMBOL(blkdev_fsync
);
359 * bdev_read_page() - Start reading a page from a block device
360 * @bdev: The device to read the page from
361 * @sector: The offset on the device to read the page to (need not be aligned)
362 * @page: The page to read
364 * On entry, the page should be locked. It will be unlocked when the page
365 * has been read. If the block driver implements rw_page synchronously,
366 * that will be true on exit from this function, but it need not be.
368 * Errors returned by this function are usually "soft", eg out of memory, or
369 * queue full; callers should try a different route to read this page rather
370 * than propagate an error back up the stack.
372 * Return: negative errno if an error occurs, 0 if submission was successful.
374 int bdev_read_page(struct block_device
*bdev
, sector_t sector
,
377 const struct block_device_operations
*ops
= bdev
->bd_disk
->fops
;
380 return ops
->rw_page(bdev
, sector
+ get_start_sect(bdev
), page
, READ
);
382 EXPORT_SYMBOL_GPL(bdev_read_page
);
385 * bdev_write_page() - Start writing a page to a block device
386 * @bdev: The device to write the page to
387 * @sector: The offset on the device to write the page to (need not be aligned)
388 * @page: The page to write
389 * @wbc: The writeback_control for the write
391 * On entry, the page should be locked and not currently under writeback.
392 * On exit, if the write started successfully, the page will be unlocked and
393 * under writeback. If the write failed already (eg the driver failed to
394 * queue the page to the device), the page will still be locked. If the
395 * caller is a ->writepage implementation, it will need to unlock the page.
397 * Errors returned by this function are usually "soft", eg out of memory, or
398 * queue full; callers should try a different route to write this page rather
399 * than propagate an error back up the stack.
401 * Return: negative errno if an error occurs, 0 if submission was successful.
403 int bdev_write_page(struct block_device
*bdev
, sector_t sector
,
404 struct page
*page
, struct writeback_control
*wbc
)
407 int rw
= (wbc
->sync_mode
== WB_SYNC_ALL
) ? WRITE_SYNC
: WRITE
;
408 const struct block_device_operations
*ops
= bdev
->bd_disk
->fops
;
411 set_page_writeback(page
);
412 result
= ops
->rw_page(bdev
, sector
+ get_start_sect(bdev
), page
, rw
);
414 end_page_writeback(page
);
419 EXPORT_SYMBOL_GPL(bdev_write_page
);
422 * bdev_direct_access() - Get the address for directly-accessibly memory
423 * @bdev: The device containing the memory
424 * @sector: The offset within the device
425 * @addr: Where to put the address of the memory
426 * @pfn: The Page Frame Number for the memory
427 * @size: The number of bytes requested
429 * If a block device is made up of directly addressable memory, this function
430 * will tell the caller the PFN and the address of the memory. The address
431 * may be directly dereferenced within the kernel without the need to call
432 * ioremap(), kmap() or similar. The PFN is suitable for inserting into
435 * Return: negative errno if an error occurs, otherwise the number of bytes
436 * accessible at this address.
438 long bdev_direct_access(struct block_device
*bdev
, sector_t sector
,
439 void **addr
, unsigned long *pfn
, long size
)
442 const struct block_device_operations
*ops
= bdev
->bd_disk
->fops
;
446 if (!ops
->direct_access
)
448 if ((sector
+ DIV_ROUND_UP(size
, 512)) >
449 part_nr_sects_read(bdev
->bd_part
))
451 sector
+= get_start_sect(bdev
);
452 if (sector
% (PAGE_SIZE
/ 512))
454 avail
= ops
->direct_access(bdev
, sector
, addr
, pfn
, size
);
457 return min(avail
, size
);
459 EXPORT_SYMBOL_GPL(bdev_direct_access
);
465 static __cacheline_aligned_in_smp
DEFINE_SPINLOCK(bdev_lock
);
466 static struct kmem_cache
* bdev_cachep __read_mostly
;
468 static struct inode
*bdev_alloc_inode(struct super_block
*sb
)
470 struct bdev_inode
*ei
= kmem_cache_alloc(bdev_cachep
, GFP_KERNEL
);
473 return &ei
->vfs_inode
;
476 static void bdev_i_callback(struct rcu_head
*head
)
478 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
479 struct bdev_inode
*bdi
= BDEV_I(inode
);
481 kmem_cache_free(bdev_cachep
, bdi
);
484 static void bdev_destroy_inode(struct inode
*inode
)
486 call_rcu(&inode
->i_rcu
, bdev_i_callback
);
489 static void init_once(void *foo
)
491 struct bdev_inode
*ei
= (struct bdev_inode
*) foo
;
492 struct block_device
*bdev
= &ei
->bdev
;
494 memset(bdev
, 0, sizeof(*bdev
));
495 mutex_init(&bdev
->bd_mutex
);
496 INIT_LIST_HEAD(&bdev
->bd_inodes
);
497 INIT_LIST_HEAD(&bdev
->bd_list
);
499 INIT_LIST_HEAD(&bdev
->bd_holder_disks
);
501 inode_init_once(&ei
->vfs_inode
);
502 /* Initialize mutex for freeze. */
503 mutex_init(&bdev
->bd_fsfreeze_mutex
);
506 static inline void __bd_forget(struct inode
*inode
)
508 list_del_init(&inode
->i_devices
);
509 inode
->i_bdev
= NULL
;
510 inode
->i_mapping
= &inode
->i_data
;
513 static void bdev_evict_inode(struct inode
*inode
)
515 struct block_device
*bdev
= &BDEV_I(inode
)->bdev
;
517 truncate_inode_pages_final(&inode
->i_data
);
518 invalidate_inode_buffers(inode
); /* is it needed here? */
520 spin_lock(&bdev_lock
);
521 while ( (p
= bdev
->bd_inodes
.next
) != &bdev
->bd_inodes
) {
522 __bd_forget(list_entry(p
, struct inode
, i_devices
));
524 list_del_init(&bdev
->bd_list
);
525 spin_unlock(&bdev_lock
);
528 static const struct super_operations bdev_sops
= {
529 .statfs
= simple_statfs
,
530 .alloc_inode
= bdev_alloc_inode
,
531 .destroy_inode
= bdev_destroy_inode
,
532 .drop_inode
= generic_delete_inode
,
533 .evict_inode
= bdev_evict_inode
,
536 static struct dentry
*bd_mount(struct file_system_type
*fs_type
,
537 int flags
, const char *dev_name
, void *data
)
539 return mount_pseudo(fs_type
, "bdev:", &bdev_sops
, NULL
, BDEVFS_MAGIC
);
542 static struct file_system_type bd_type
= {
545 .kill_sb
= kill_anon_super
,
548 static struct super_block
*blockdev_superblock __read_mostly
;
550 void __init
bdev_cache_init(void)
553 static struct vfsmount
*bd_mnt
;
555 bdev_cachep
= kmem_cache_create("bdev_cache", sizeof(struct bdev_inode
),
556 0, (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
557 SLAB_MEM_SPREAD
|SLAB_PANIC
),
559 err
= register_filesystem(&bd_type
);
561 panic("Cannot register bdev pseudo-fs");
562 bd_mnt
= kern_mount(&bd_type
);
564 panic("Cannot create bdev pseudo-fs");
565 blockdev_superblock
= bd_mnt
->mnt_sb
; /* For writeback */
569 * Most likely _very_ bad one - but then it's hardly critical for small
570 * /dev and can be fixed when somebody will need really large one.
571 * Keep in mind that it will be fed through icache hash function too.
573 static inline unsigned long hash(dev_t dev
)
575 return MAJOR(dev
)+MINOR(dev
);
578 static int bdev_test(struct inode
*inode
, void *data
)
580 return BDEV_I(inode
)->bdev
.bd_dev
== *(dev_t
*)data
;
583 static int bdev_set(struct inode
*inode
, void *data
)
585 BDEV_I(inode
)->bdev
.bd_dev
= *(dev_t
*)data
;
589 static LIST_HEAD(all_bdevs
);
591 struct block_device
*bdget(dev_t dev
)
593 struct block_device
*bdev
;
596 inode
= iget5_locked(blockdev_superblock
, hash(dev
),
597 bdev_test
, bdev_set
, &dev
);
602 bdev
= &BDEV_I(inode
)->bdev
;
604 if (inode
->i_state
& I_NEW
) {
605 bdev
->bd_contains
= NULL
;
606 bdev
->bd_super
= NULL
;
607 bdev
->bd_inode
= inode
;
608 bdev
->bd_block_size
= (1 << inode
->i_blkbits
);
609 bdev
->bd_part_count
= 0;
610 bdev
->bd_invalidated
= 0;
611 inode
->i_mode
= S_IFBLK
;
613 inode
->i_bdev
= bdev
;
614 inode
->i_data
.a_ops
= &def_blk_aops
;
615 mapping_set_gfp_mask(&inode
->i_data
, GFP_USER
);
616 spin_lock(&bdev_lock
);
617 list_add(&bdev
->bd_list
, &all_bdevs
);
618 spin_unlock(&bdev_lock
);
619 unlock_new_inode(inode
);
624 EXPORT_SYMBOL(bdget
);
627 * bdgrab -- Grab a reference to an already referenced block device
628 * @bdev: Block device to grab a reference to.
630 struct block_device
*bdgrab(struct block_device
*bdev
)
632 ihold(bdev
->bd_inode
);
635 EXPORT_SYMBOL(bdgrab
);
637 long nr_blockdev_pages(void)
639 struct block_device
*bdev
;
641 spin_lock(&bdev_lock
);
642 list_for_each_entry(bdev
, &all_bdevs
, bd_list
) {
643 ret
+= bdev
->bd_inode
->i_mapping
->nrpages
;
645 spin_unlock(&bdev_lock
);
649 void bdput(struct block_device
*bdev
)
651 iput(bdev
->bd_inode
);
654 EXPORT_SYMBOL(bdput
);
656 static struct block_device
*bd_acquire(struct inode
*inode
)
658 struct block_device
*bdev
;
660 spin_lock(&bdev_lock
);
661 bdev
= inode
->i_bdev
;
663 ihold(bdev
->bd_inode
);
664 spin_unlock(&bdev_lock
);
667 spin_unlock(&bdev_lock
);
669 bdev
= bdget(inode
->i_rdev
);
671 spin_lock(&bdev_lock
);
672 if (!inode
->i_bdev
) {
674 * We take an additional reference to bd_inode,
675 * and it's released in clear_inode() of inode.
676 * So, we can access it via ->i_mapping always
679 ihold(bdev
->bd_inode
);
680 inode
->i_bdev
= bdev
;
681 inode
->i_mapping
= bdev
->bd_inode
->i_mapping
;
682 list_add(&inode
->i_devices
, &bdev
->bd_inodes
);
684 spin_unlock(&bdev_lock
);
689 int sb_is_blkdev_sb(struct super_block
*sb
)
691 return sb
== blockdev_superblock
;
694 /* Call when you free inode */
696 void bd_forget(struct inode
*inode
)
698 struct block_device
*bdev
= NULL
;
700 spin_lock(&bdev_lock
);
701 if (!sb_is_blkdev_sb(inode
->i_sb
))
702 bdev
= inode
->i_bdev
;
704 spin_unlock(&bdev_lock
);
707 iput(bdev
->bd_inode
);
711 * bd_may_claim - test whether a block device can be claimed
712 * @bdev: block device of interest
713 * @whole: whole block device containing @bdev, may equal @bdev
714 * @holder: holder trying to claim @bdev
716 * Test whether @bdev can be claimed by @holder.
719 * spin_lock(&bdev_lock).
722 * %true if @bdev can be claimed, %false otherwise.
724 static bool bd_may_claim(struct block_device
*bdev
, struct block_device
*whole
,
727 if (bdev
->bd_holder
== holder
)
728 return true; /* already a holder */
729 else if (bdev
->bd_holder
!= NULL
)
730 return false; /* held by someone else */
731 else if (bdev
->bd_contains
== bdev
)
732 return true; /* is a whole device which isn't held */
734 else if (whole
->bd_holder
== bd_may_claim
)
735 return true; /* is a partition of a device that is being partitioned */
736 else if (whole
->bd_holder
!= NULL
)
737 return false; /* is a partition of a held device */
739 return true; /* is a partition of an un-held device */
743 * bd_prepare_to_claim - prepare to claim a block device
744 * @bdev: block device of interest
745 * @whole: the whole device containing @bdev, may equal @bdev
746 * @holder: holder trying to claim @bdev
748 * Prepare to claim @bdev. This function fails if @bdev is already
749 * claimed by another holder and waits if another claiming is in
750 * progress. This function doesn't actually claim. On successful
751 * return, the caller has ownership of bd_claiming and bd_holder[s].
754 * spin_lock(&bdev_lock). Might release bdev_lock, sleep and regrab
758 * 0 if @bdev can be claimed, -EBUSY otherwise.
760 static int bd_prepare_to_claim(struct block_device
*bdev
,
761 struct block_device
*whole
, void *holder
)
764 /* if someone else claimed, fail */
765 if (!bd_may_claim(bdev
, whole
, holder
))
768 /* if claiming is already in progress, wait for it to finish */
769 if (whole
->bd_claiming
) {
770 wait_queue_head_t
*wq
= bit_waitqueue(&whole
->bd_claiming
, 0);
773 prepare_to_wait(wq
, &wait
, TASK_UNINTERRUPTIBLE
);
774 spin_unlock(&bdev_lock
);
776 finish_wait(wq
, &wait
);
777 spin_lock(&bdev_lock
);
786 * bd_start_claiming - start claiming a block device
787 * @bdev: block device of interest
788 * @holder: holder trying to claim @bdev
790 * @bdev is about to be opened exclusively. Check @bdev can be opened
791 * exclusively and mark that an exclusive open is in progress. Each
792 * successful call to this function must be matched with a call to
793 * either bd_finish_claiming() or bd_abort_claiming() (which do not
796 * This function is used to gain exclusive access to the block device
797 * without actually causing other exclusive open attempts to fail. It
798 * should be used when the open sequence itself requires exclusive
799 * access but may subsequently fail.
805 * Pointer to the block device containing @bdev on success, ERR_PTR()
808 static struct block_device
*bd_start_claiming(struct block_device
*bdev
,
811 struct gendisk
*disk
;
812 struct block_device
*whole
;
818 * @bdev might not have been initialized properly yet, look up
819 * and grab the outer block device the hard way.
821 disk
= get_gendisk(bdev
->bd_dev
, &partno
);
823 return ERR_PTR(-ENXIO
);
826 * Normally, @bdev should equal what's returned from bdget_disk()
827 * if partno is 0; however, some drivers (floppy) use multiple
828 * bdev's for the same physical device and @bdev may be one of the
829 * aliases. Keep @bdev if partno is 0. This means claimer
830 * tracking is broken for those devices but it has always been that
834 whole
= bdget_disk(disk
, 0);
836 whole
= bdgrab(bdev
);
838 module_put(disk
->fops
->owner
);
841 return ERR_PTR(-ENOMEM
);
843 /* prepare to claim, if successful, mark claiming in progress */
844 spin_lock(&bdev_lock
);
846 err
= bd_prepare_to_claim(bdev
, whole
, holder
);
848 whole
->bd_claiming
= holder
;
849 spin_unlock(&bdev_lock
);
852 spin_unlock(&bdev_lock
);
859 struct bd_holder_disk
{
860 struct list_head list
;
861 struct gendisk
*disk
;
865 static struct bd_holder_disk
*bd_find_holder_disk(struct block_device
*bdev
,
866 struct gendisk
*disk
)
868 struct bd_holder_disk
*holder
;
870 list_for_each_entry(holder
, &bdev
->bd_holder_disks
, list
)
871 if (holder
->disk
== disk
)
876 static int add_symlink(struct kobject
*from
, struct kobject
*to
)
878 return sysfs_create_link(from
, to
, kobject_name(to
));
881 static void del_symlink(struct kobject
*from
, struct kobject
*to
)
883 sysfs_remove_link(from
, kobject_name(to
));
887 * bd_link_disk_holder - create symlinks between holding disk and slave bdev
888 * @bdev: the claimed slave bdev
889 * @disk: the holding disk
891 * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT.
893 * This functions creates the following sysfs symlinks.
895 * - from "slaves" directory of the holder @disk to the claimed @bdev
896 * - from "holders" directory of the @bdev to the holder @disk
898 * For example, if /dev/dm-0 maps to /dev/sda and disk for dm-0 is
899 * passed to bd_link_disk_holder(), then:
901 * /sys/block/dm-0/slaves/sda --> /sys/block/sda
902 * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0
904 * The caller must have claimed @bdev before calling this function and
905 * ensure that both @bdev and @disk are valid during the creation and
906 * lifetime of these symlinks.
912 * 0 on success, -errno on failure.
914 int bd_link_disk_holder(struct block_device
*bdev
, struct gendisk
*disk
)
916 struct bd_holder_disk
*holder
;
919 mutex_lock(&bdev
->bd_mutex
);
921 WARN_ON_ONCE(!bdev
->bd_holder
);
923 /* FIXME: remove the following once add_disk() handles errors */
924 if (WARN_ON(!disk
->slave_dir
|| !bdev
->bd_part
->holder_dir
))
927 holder
= bd_find_holder_disk(bdev
, disk
);
933 holder
= kzalloc(sizeof(*holder
), GFP_KERNEL
);
939 INIT_LIST_HEAD(&holder
->list
);
943 ret
= add_symlink(disk
->slave_dir
, &part_to_dev(bdev
->bd_part
)->kobj
);
947 ret
= add_symlink(bdev
->bd_part
->holder_dir
, &disk_to_dev(disk
)->kobj
);
951 * bdev could be deleted beneath us which would implicitly destroy
952 * the holder directory. Hold on to it.
954 kobject_get(bdev
->bd_part
->holder_dir
);
956 list_add(&holder
->list
, &bdev
->bd_holder_disks
);
960 del_symlink(disk
->slave_dir
, &part_to_dev(bdev
->bd_part
)->kobj
);
964 mutex_unlock(&bdev
->bd_mutex
);
967 EXPORT_SYMBOL_GPL(bd_link_disk_holder
);
970 * bd_unlink_disk_holder - destroy symlinks created by bd_link_disk_holder()
971 * @bdev: the calimed slave bdev
972 * @disk: the holding disk
974 * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT.
979 void bd_unlink_disk_holder(struct block_device
*bdev
, struct gendisk
*disk
)
981 struct bd_holder_disk
*holder
;
983 mutex_lock(&bdev
->bd_mutex
);
985 holder
= bd_find_holder_disk(bdev
, disk
);
987 if (!WARN_ON_ONCE(holder
== NULL
) && !--holder
->refcnt
) {
988 del_symlink(disk
->slave_dir
, &part_to_dev(bdev
->bd_part
)->kobj
);
989 del_symlink(bdev
->bd_part
->holder_dir
,
990 &disk_to_dev(disk
)->kobj
);
991 kobject_put(bdev
->bd_part
->holder_dir
);
992 list_del_init(&holder
->list
);
996 mutex_unlock(&bdev
->bd_mutex
);
998 EXPORT_SYMBOL_GPL(bd_unlink_disk_holder
);
1002 * flush_disk - invalidates all buffer-cache entries on a disk
1004 * @bdev: struct block device to be flushed
1005 * @kill_dirty: flag to guide handling of dirty inodes
1007 * Invalidates all buffer-cache entries on a disk. It should be called
1008 * when a disk has been changed -- either by a media change or online
1011 static void flush_disk(struct block_device
*bdev
, bool kill_dirty
)
1013 if (__invalidate_device(bdev
, kill_dirty
)) {
1014 char name
[BDEVNAME_SIZE
] = "";
1017 disk_name(bdev
->bd_disk
, 0, name
);
1018 printk(KERN_WARNING
"VFS: busy inodes on changed media or "
1019 "resized disk %s\n", name
);
1024 if (disk_part_scan_enabled(bdev
->bd_disk
))
1025 bdev
->bd_invalidated
= 1;
1029 * check_disk_size_change - checks for disk size change and adjusts bdev size.
1030 * @disk: struct gendisk to check
1031 * @bdev: struct bdev to adjust.
1033 * This routine checks to see if the bdev size does not match the disk size
1034 * and adjusts it if it differs.
1036 void check_disk_size_change(struct gendisk
*disk
, struct block_device
*bdev
)
1038 loff_t disk_size
, bdev_size
;
1040 disk_size
= (loff_t
)get_capacity(disk
) << 9;
1041 bdev_size
= i_size_read(bdev
->bd_inode
);
1042 if (disk_size
!= bdev_size
) {
1043 char name
[BDEVNAME_SIZE
];
1045 disk_name(disk
, 0, name
);
1047 "%s: detected capacity change from %lld to %lld\n",
1048 name
, bdev_size
, disk_size
);
1049 i_size_write(bdev
->bd_inode
, disk_size
);
1050 flush_disk(bdev
, false);
1053 EXPORT_SYMBOL(check_disk_size_change
);
1056 * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back
1057 * @disk: struct gendisk to be revalidated
1059 * This routine is a wrapper for lower-level driver's revalidate_disk
1060 * call-backs. It is used to do common pre and post operations needed
1061 * for all revalidate_disk operations.
1063 int revalidate_disk(struct gendisk
*disk
)
1065 struct block_device
*bdev
;
1068 if (disk
->fops
->revalidate_disk
)
1069 ret
= disk
->fops
->revalidate_disk(disk
);
1071 bdev
= bdget_disk(disk
, 0);
1075 mutex_lock(&bdev
->bd_mutex
);
1076 check_disk_size_change(disk
, bdev
);
1077 bdev
->bd_invalidated
= 0;
1078 mutex_unlock(&bdev
->bd_mutex
);
1082 EXPORT_SYMBOL(revalidate_disk
);
1085 * This routine checks whether a removable media has been changed,
1086 * and invalidates all buffer-cache-entries in that case. This
1087 * is a relatively slow routine, so we have to try to minimize using
1088 * it. Thus it is called only upon a 'mount' or 'open'. This
1089 * is the best way of combining speed and utility, I think.
1090 * People changing diskettes in the middle of an operation deserve
1093 int check_disk_change(struct block_device
*bdev
)
1095 struct gendisk
*disk
= bdev
->bd_disk
;
1096 const struct block_device_operations
*bdops
= disk
->fops
;
1097 unsigned int events
;
1099 events
= disk_clear_events(disk
, DISK_EVENT_MEDIA_CHANGE
|
1100 DISK_EVENT_EJECT_REQUEST
);
1101 if (!(events
& DISK_EVENT_MEDIA_CHANGE
))
1104 flush_disk(bdev
, true);
1105 if (bdops
->revalidate_disk
)
1106 bdops
->revalidate_disk(bdev
->bd_disk
);
1110 EXPORT_SYMBOL(check_disk_change
);
1112 void bd_set_size(struct block_device
*bdev
, loff_t size
)
1114 unsigned bsize
= bdev_logical_block_size(bdev
);
1116 mutex_lock(&bdev
->bd_inode
->i_mutex
);
1117 i_size_write(bdev
->bd_inode
, size
);
1118 mutex_unlock(&bdev
->bd_inode
->i_mutex
);
1119 while (bsize
< PAGE_CACHE_SIZE
) {
1124 bdev
->bd_block_size
= bsize
;
1125 bdev
->bd_inode
->i_blkbits
= blksize_bits(bsize
);
1127 EXPORT_SYMBOL(bd_set_size
);
1129 static void __blkdev_put(struct block_device
*bdev
, fmode_t mode
, int for_part
);
1134 * mutex_lock(part->bd_mutex)
1135 * mutex_lock_nested(whole->bd_mutex, 1)
1138 static int __blkdev_get(struct block_device
*bdev
, fmode_t mode
, int for_part
)
1140 struct gendisk
*disk
;
1141 struct module
*owner
;
1146 if (mode
& FMODE_READ
)
1148 if (mode
& FMODE_WRITE
)
1151 * hooks: /n/, see "layering violations".
1154 ret
= devcgroup_inode_permission(bdev
->bd_inode
, perm
);
1164 disk
= get_gendisk(bdev
->bd_dev
, &partno
);
1167 owner
= disk
->fops
->owner
;
1169 disk_block_events(disk
);
1170 mutex_lock_nested(&bdev
->bd_mutex
, for_part
);
1171 if (!bdev
->bd_openers
) {
1172 bdev
->bd_disk
= disk
;
1173 bdev
->bd_queue
= disk
->queue
;
1174 bdev
->bd_contains
= bdev
;
1177 bdev
->bd_part
= disk_get_part(disk
, partno
);
1182 if (disk
->fops
->open
) {
1183 ret
= disk
->fops
->open(bdev
, mode
);
1184 if (ret
== -ERESTARTSYS
) {
1185 /* Lost a race with 'disk' being
1186 * deleted, try again.
1189 disk_put_part(bdev
->bd_part
);
1190 bdev
->bd_part
= NULL
;
1191 bdev
->bd_disk
= NULL
;
1192 bdev
->bd_queue
= NULL
;
1193 mutex_unlock(&bdev
->bd_mutex
);
1194 disk_unblock_events(disk
);
1202 bd_set_size(bdev
,(loff_t
)get_capacity(disk
)<<9);
1205 * If the device is invalidated, rescan partition
1206 * if open succeeded or failed with -ENOMEDIUM.
1207 * The latter is necessary to prevent ghost
1208 * partitions on a removed medium.
1210 if (bdev
->bd_invalidated
) {
1212 rescan_partitions(disk
, bdev
);
1213 else if (ret
== -ENOMEDIUM
)
1214 invalidate_partitions(disk
, bdev
);
1219 struct block_device
*whole
;
1220 whole
= bdget_disk(disk
, 0);
1225 ret
= __blkdev_get(whole
, mode
, 1);
1228 bdev
->bd_contains
= whole
;
1229 bdev
->bd_part
= disk_get_part(disk
, partno
);
1230 if (!(disk
->flags
& GENHD_FL_UP
) ||
1231 !bdev
->bd_part
|| !bdev
->bd_part
->nr_sects
) {
1235 bd_set_size(bdev
, (loff_t
)bdev
->bd_part
->nr_sects
<< 9);
1238 if (bdev
->bd_contains
== bdev
) {
1240 if (bdev
->bd_disk
->fops
->open
)
1241 ret
= bdev
->bd_disk
->fops
->open(bdev
, mode
);
1242 /* the same as first opener case, read comment there */
1243 if (bdev
->bd_invalidated
) {
1245 rescan_partitions(bdev
->bd_disk
, bdev
);
1246 else if (ret
== -ENOMEDIUM
)
1247 invalidate_partitions(bdev
->bd_disk
, bdev
);
1250 goto out_unlock_bdev
;
1252 /* only one opener holds refs to the module and disk */
1258 bdev
->bd_part_count
++;
1259 mutex_unlock(&bdev
->bd_mutex
);
1260 disk_unblock_events(disk
);
1264 disk_put_part(bdev
->bd_part
);
1265 bdev
->bd_disk
= NULL
;
1266 bdev
->bd_part
= NULL
;
1267 bdev
->bd_queue
= NULL
;
1268 if (bdev
!= bdev
->bd_contains
)
1269 __blkdev_put(bdev
->bd_contains
, mode
, 1);
1270 bdev
->bd_contains
= NULL
;
1272 mutex_unlock(&bdev
->bd_mutex
);
1273 disk_unblock_events(disk
);
1283 * blkdev_get - open a block device
1284 * @bdev: block_device to open
1285 * @mode: FMODE_* mask
1286 * @holder: exclusive holder identifier
1288 * Open @bdev with @mode. If @mode includes %FMODE_EXCL, @bdev is
1289 * open with exclusive access. Specifying %FMODE_EXCL with %NULL
1290 * @holder is invalid. Exclusive opens may nest for the same @holder.
1292 * On success, the reference count of @bdev is unchanged. On failure,
1299 * 0 on success, -errno on failure.
1301 int blkdev_get(struct block_device
*bdev
, fmode_t mode
, void *holder
)
1303 struct block_device
*whole
= NULL
;
1306 WARN_ON_ONCE((mode
& FMODE_EXCL
) && !holder
);
1308 if ((mode
& FMODE_EXCL
) && holder
) {
1309 whole
= bd_start_claiming(bdev
, holder
);
1310 if (IS_ERR(whole
)) {
1312 return PTR_ERR(whole
);
1316 res
= __blkdev_get(bdev
, mode
, 0);
1319 struct gendisk
*disk
= whole
->bd_disk
;
1321 /* finish claiming */
1322 mutex_lock(&bdev
->bd_mutex
);
1323 spin_lock(&bdev_lock
);
1326 BUG_ON(!bd_may_claim(bdev
, whole
, holder
));
1328 * Note that for a whole device bd_holders
1329 * will be incremented twice, and bd_holder
1330 * will be set to bd_may_claim before being
1333 whole
->bd_holders
++;
1334 whole
->bd_holder
= bd_may_claim
;
1336 bdev
->bd_holder
= holder
;
1339 /* tell others that we're done */
1340 BUG_ON(whole
->bd_claiming
!= holder
);
1341 whole
->bd_claiming
= NULL
;
1342 wake_up_bit(&whole
->bd_claiming
, 0);
1344 spin_unlock(&bdev_lock
);
1347 * Block event polling for write claims if requested. Any
1348 * write holder makes the write_holder state stick until
1349 * all are released. This is good enough and tracking
1350 * individual writeable reference is too fragile given the
1351 * way @mode is used in blkdev_get/put().
1353 if (!res
&& (mode
& FMODE_WRITE
) && !bdev
->bd_write_holder
&&
1354 (disk
->flags
& GENHD_FL_BLOCK_EVENTS_ON_EXCL_WRITE
)) {
1355 bdev
->bd_write_holder
= true;
1356 disk_block_events(disk
);
1359 mutex_unlock(&bdev
->bd_mutex
);
1365 EXPORT_SYMBOL(blkdev_get
);
1368 * blkdev_get_by_path - open a block device by name
1369 * @path: path to the block device to open
1370 * @mode: FMODE_* mask
1371 * @holder: exclusive holder identifier
1373 * Open the blockdevice described by the device file at @path. @mode
1374 * and @holder are identical to blkdev_get().
1376 * On success, the returned block_device has reference count of one.
1382 * Pointer to block_device on success, ERR_PTR(-errno) on failure.
1384 struct block_device
*blkdev_get_by_path(const char *path
, fmode_t mode
,
1387 struct block_device
*bdev
;
1390 bdev
= lookup_bdev(path
);
1394 err
= blkdev_get(bdev
, mode
, holder
);
1396 return ERR_PTR(err
);
1398 if ((mode
& FMODE_WRITE
) && bdev_read_only(bdev
)) {
1399 blkdev_put(bdev
, mode
);
1400 return ERR_PTR(-EACCES
);
1405 EXPORT_SYMBOL(blkdev_get_by_path
);
1408 * blkdev_get_by_dev - open a block device by device number
1409 * @dev: device number of block device to open
1410 * @mode: FMODE_* mask
1411 * @holder: exclusive holder identifier
1413 * Open the blockdevice described by device number @dev. @mode and
1414 * @holder are identical to blkdev_get().
1416 * Use it ONLY if you really do not have anything better - i.e. when
1417 * you are behind a truly sucky interface and all you are given is a
1418 * device number. _Never_ to be used for internal purposes. If you
1419 * ever need it - reconsider your API.
1421 * On success, the returned block_device has reference count of one.
1427 * Pointer to block_device on success, ERR_PTR(-errno) on failure.
1429 struct block_device
*blkdev_get_by_dev(dev_t dev
, fmode_t mode
, void *holder
)
1431 struct block_device
*bdev
;
1436 return ERR_PTR(-ENOMEM
);
1438 err
= blkdev_get(bdev
, mode
, holder
);
1440 return ERR_PTR(err
);
1444 EXPORT_SYMBOL(blkdev_get_by_dev
);
1446 static int blkdev_open(struct inode
* inode
, struct file
* filp
)
1448 struct block_device
*bdev
;
1451 * Preserve backwards compatibility and allow large file access
1452 * even if userspace doesn't ask for it explicitly. Some mkfs
1453 * binary needs it. We might want to drop this workaround
1454 * during an unstable branch.
1456 filp
->f_flags
|= O_LARGEFILE
;
1458 if (filp
->f_flags
& O_NDELAY
)
1459 filp
->f_mode
|= FMODE_NDELAY
;
1460 if (filp
->f_flags
& O_EXCL
)
1461 filp
->f_mode
|= FMODE_EXCL
;
1462 if ((filp
->f_flags
& O_ACCMODE
) == 3)
1463 filp
->f_mode
|= FMODE_WRITE_IOCTL
;
1465 bdev
= bd_acquire(inode
);
1469 filp
->f_mapping
= bdev
->bd_inode
->i_mapping
;
1471 return blkdev_get(bdev
, filp
->f_mode
, filp
);
1474 static void __blkdev_put(struct block_device
*bdev
, fmode_t mode
, int for_part
)
1476 struct gendisk
*disk
= bdev
->bd_disk
;
1477 struct block_device
*victim
= NULL
;
1479 mutex_lock_nested(&bdev
->bd_mutex
, for_part
);
1481 bdev
->bd_part_count
--;
1483 if (!--bdev
->bd_openers
) {
1484 WARN_ON_ONCE(bdev
->bd_holders
);
1485 sync_blockdev(bdev
);
1488 * ->release can cause the queue to disappear, so flush all
1489 * dirty data before.
1491 bdev_write_inode(bdev
->bd_inode
);
1493 if (bdev
->bd_contains
== bdev
) {
1494 if (disk
->fops
->release
)
1495 disk
->fops
->release(disk
, mode
);
1497 if (!bdev
->bd_openers
) {
1498 struct module
*owner
= disk
->fops
->owner
;
1500 disk_put_part(bdev
->bd_part
);
1501 bdev
->bd_part
= NULL
;
1502 bdev
->bd_disk
= NULL
;
1503 if (bdev
!= bdev
->bd_contains
)
1504 victim
= bdev
->bd_contains
;
1505 bdev
->bd_contains
= NULL
;
1510 mutex_unlock(&bdev
->bd_mutex
);
1513 __blkdev_put(victim
, mode
, 1);
1516 void blkdev_put(struct block_device
*bdev
, fmode_t mode
)
1518 mutex_lock(&bdev
->bd_mutex
);
1520 if (mode
& FMODE_EXCL
) {
1524 * Release a claim on the device. The holder fields
1525 * are protected with bdev_lock. bd_mutex is to
1526 * synchronize disk_holder unlinking.
1528 spin_lock(&bdev_lock
);
1530 WARN_ON_ONCE(--bdev
->bd_holders
< 0);
1531 WARN_ON_ONCE(--bdev
->bd_contains
->bd_holders
< 0);
1533 /* bd_contains might point to self, check in a separate step */
1534 if ((bdev_free
= !bdev
->bd_holders
))
1535 bdev
->bd_holder
= NULL
;
1536 if (!bdev
->bd_contains
->bd_holders
)
1537 bdev
->bd_contains
->bd_holder
= NULL
;
1539 spin_unlock(&bdev_lock
);
1542 * If this was the last claim, remove holder link and
1543 * unblock evpoll if it was a write holder.
1545 if (bdev_free
&& bdev
->bd_write_holder
) {
1546 disk_unblock_events(bdev
->bd_disk
);
1547 bdev
->bd_write_holder
= false;
1552 * Trigger event checking and tell drivers to flush MEDIA_CHANGE
1553 * event. This is to ensure detection of media removal commanded
1554 * from userland - e.g. eject(1).
1556 disk_flush_events(bdev
->bd_disk
, DISK_EVENT_MEDIA_CHANGE
);
1558 mutex_unlock(&bdev
->bd_mutex
);
1560 __blkdev_put(bdev
, mode
, 0);
1562 EXPORT_SYMBOL(blkdev_put
);
1564 static int blkdev_close(struct inode
* inode
, struct file
* filp
)
1566 struct block_device
*bdev
= I_BDEV(filp
->f_mapping
->host
);
1567 blkdev_put(bdev
, filp
->f_mode
);
1571 static long block_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1573 struct block_device
*bdev
= I_BDEV(file
->f_mapping
->host
);
1574 fmode_t mode
= file
->f_mode
;
1577 * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have
1578 * to updated it before every ioctl.
1580 if (file
->f_flags
& O_NDELAY
)
1581 mode
|= FMODE_NDELAY
;
1583 mode
&= ~FMODE_NDELAY
;
1585 return blkdev_ioctl(bdev
, mode
, cmd
, arg
);
1589 * Write data to the block device. Only intended for the block device itself
1590 * and the raw driver which basically is a fake block device.
1592 * Does not take i_mutex for the write and thus is not for general purpose
1595 ssize_t
blkdev_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
1597 struct file
*file
= iocb
->ki_filp
;
1598 struct inode
*bd_inode
= file
->f_mapping
->host
;
1599 loff_t size
= i_size_read(bd_inode
);
1600 struct blk_plug plug
;
1603 if (bdev_read_only(I_BDEV(bd_inode
)))
1606 if (!iov_iter_count(from
))
1609 if (iocb
->ki_pos
>= size
)
1612 iov_iter_truncate(from
, size
- iocb
->ki_pos
);
1614 blk_start_plug(&plug
);
1615 ret
= __generic_file_write_iter(iocb
, from
);
1618 err
= generic_write_sync(file
, iocb
->ki_pos
- ret
, ret
);
1622 blk_finish_plug(&plug
);
1625 EXPORT_SYMBOL_GPL(blkdev_write_iter
);
1627 ssize_t
blkdev_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
1629 struct file
*file
= iocb
->ki_filp
;
1630 struct inode
*bd_inode
= file
->f_mapping
->host
;
1631 loff_t size
= i_size_read(bd_inode
);
1632 loff_t pos
= iocb
->ki_pos
;
1638 iov_iter_truncate(to
, size
);
1639 return generic_file_read_iter(iocb
, to
);
1641 EXPORT_SYMBOL_GPL(blkdev_read_iter
);
1644 * Try to release a page associated with block device when the system
1645 * is under memory pressure.
1647 static int blkdev_releasepage(struct page
*page
, gfp_t wait
)
1649 struct super_block
*super
= BDEV_I(page
->mapping
->host
)->bdev
.bd_super
;
1651 if (super
&& super
->s_op
->bdev_try_to_free_page
)
1652 return super
->s_op
->bdev_try_to_free_page(super
, page
, wait
);
1654 return try_to_free_buffers(page
);
1657 static const struct address_space_operations def_blk_aops
= {
1658 .readpage
= blkdev_readpage
,
1659 .readpages
= blkdev_readpages
,
1660 .writepage
= blkdev_writepage
,
1661 .write_begin
= blkdev_write_begin
,
1662 .write_end
= blkdev_write_end
,
1663 .writepages
= generic_writepages
,
1664 .releasepage
= blkdev_releasepage
,
1665 .direct_IO
= blkdev_direct_IO
,
1666 .is_dirty_writeback
= buffer_check_dirty_writeback
,
1669 const struct file_operations def_blk_fops
= {
1670 .open
= blkdev_open
,
1671 .release
= blkdev_close
,
1672 .llseek
= block_llseek
,
1673 .read_iter
= blkdev_read_iter
,
1674 .write_iter
= blkdev_write_iter
,
1675 .mmap
= generic_file_mmap
,
1676 .fsync
= blkdev_fsync
,
1677 .unlocked_ioctl
= block_ioctl
,
1678 #ifdef CONFIG_COMPAT
1679 .compat_ioctl
= compat_blkdev_ioctl
,
1681 .splice_read
= generic_file_splice_read
,
1682 .splice_write
= iter_file_splice_write
,
1685 int ioctl_by_bdev(struct block_device
*bdev
, unsigned cmd
, unsigned long arg
)
1688 mm_segment_t old_fs
= get_fs();
1690 res
= blkdev_ioctl(bdev
, 0, cmd
, arg
);
1695 EXPORT_SYMBOL(ioctl_by_bdev
);
1698 * lookup_bdev - lookup a struct block_device by name
1699 * @pathname: special file representing the block device
1701 * Get a reference to the blockdevice at @pathname in the current
1702 * namespace if possible and return it. Return ERR_PTR(error)
1705 struct block_device
*lookup_bdev(const char *pathname
)
1707 struct block_device
*bdev
;
1708 struct inode
*inode
;
1712 if (!pathname
|| !*pathname
)
1713 return ERR_PTR(-EINVAL
);
1715 error
= kern_path(pathname
, LOOKUP_FOLLOW
, &path
);
1717 return ERR_PTR(error
);
1719 inode
= path
.dentry
->d_inode
;
1721 if (!S_ISBLK(inode
->i_mode
))
1724 if (path
.mnt
->mnt_flags
& MNT_NODEV
)
1727 bdev
= bd_acquire(inode
);
1734 bdev
= ERR_PTR(error
);
1737 EXPORT_SYMBOL(lookup_bdev
);
1739 int __invalidate_device(struct block_device
*bdev
, bool kill_dirty
)
1741 struct super_block
*sb
= get_super(bdev
);
1746 * no need to lock the super, get_super holds the
1747 * read mutex so the filesystem cannot go away
1748 * under us (->put_super runs with the write lock
1751 shrink_dcache_sb(sb
);
1752 res
= invalidate_inodes(sb
, kill_dirty
);
1755 invalidate_bdev(bdev
);
1758 EXPORT_SYMBOL(__invalidate_device
);
1760 void iterate_bdevs(void (*func
)(struct block_device
*, void *), void *arg
)
1762 struct inode
*inode
, *old_inode
= NULL
;
1764 spin_lock(&inode_sb_list_lock
);
1765 list_for_each_entry(inode
, &blockdev_superblock
->s_inodes
, i_sb_list
) {
1766 struct address_space
*mapping
= inode
->i_mapping
;
1768 spin_lock(&inode
->i_lock
);
1769 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
|I_NEW
) ||
1770 mapping
->nrpages
== 0) {
1771 spin_unlock(&inode
->i_lock
);
1775 spin_unlock(&inode
->i_lock
);
1776 spin_unlock(&inode_sb_list_lock
);
1778 * We hold a reference to 'inode' so it couldn't have been
1779 * removed from s_inodes list while we dropped the
1780 * inode_sb_list_lock. We cannot iput the inode now as we can
1781 * be holding the last reference and we cannot iput it under
1782 * inode_sb_list_lock. So we keep the reference and iput it
1788 func(I_BDEV(inode
), arg
);
1790 spin_lock(&inode_sb_list_lock
);
1792 spin_unlock(&inode_sb_list_lock
);