blob: 71e7e03ac34313ed2fcb890d727bf1e47e006317 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/block_dev.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 * Copyright (C) 2001 Andrea Arcangeli <andrea@suse.de> SuSE
6 */
7
Linus Torvalds1da177e2005-04-16 15:20:36 -07008#include <linux/init.h>
9#include <linux/mm.h>
10#include <linux/fcntl.h>
11#include <linux/slab.h>
12#include <linux/kmod.h>
13#include <linux/major.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/smp_lock.h>
Pavel Emelyanov7db9cfd2008-06-05 22:46:27 -070015#include <linux/device_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/highmem.h>
17#include <linux/blkdev.h>
18#include <linux/module.h>
19#include <linux/blkpg.h>
20#include <linux/buffer_head.h>
Nick Piggin585d3bc2009-02-25 10:44:19 +010021#include <linux/pagevec.h>
David Howells811d7362006-08-29 19:06:09 +010022#include <linux/writeback.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#include <linux/mpage.h>
24#include <linux/mount.h>
25#include <linux/uio.h>
26#include <linux/namei.h>
Vignesh Babu BM1368c4f2007-05-08 00:24:32 -070027#include <linux/log2.h>
Catalin Marinas2e1483c2009-06-11 13:24:13 +010028#include <linux/kmemleak.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <asm/uaccess.h>
David Howells07f3f052006-09-30 20:52:18 +020030#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070031
32struct bdev_inode {
33 struct block_device bdev;
34 struct inode vfs_inode;
35};
36
Adrian Bunk4c54ac62008-02-18 13:48:31 +010037static const struct address_space_operations def_blk_aops;
38
Linus Torvalds1da177e2005-04-16 15:20:36 -070039static inline struct bdev_inode *BDEV_I(struct inode *inode)
40{
41 return container_of(inode, struct bdev_inode, vfs_inode);
42}
43
44inline struct block_device *I_BDEV(struct inode *inode)
45{
46 return &BDEV_I(inode)->bdev;
47}
48
49EXPORT_SYMBOL(I_BDEV);
50
51static sector_t max_block(struct block_device *bdev)
52{
53 sector_t retval = ~((sector_t)0);
54 loff_t sz = i_size_read(bdev->bd_inode);
55
56 if (sz) {
57 unsigned int size = block_size(bdev);
58 unsigned int sizebits = blksize_bits(size);
59 retval = (sz >> sizebits);
60 }
61 return retval;
62}
63
Peter Zijlstraf9a14392007-05-06 14:49:55 -070064/* Kill _all_ buffers and pagecache , dirty or not.. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070065static void kill_bdev(struct block_device *bdev)
66{
Peter Zijlstraf9a14392007-05-06 14:49:55 -070067 if (bdev->bd_inode->i_mapping->nrpages == 0)
68 return;
69 invalidate_bh_lrus();
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 truncate_inode_pages(bdev->bd_inode->i_mapping, 0);
71}
72
73int set_blocksize(struct block_device *bdev, int size)
74{
75 /* Size must be a power of two, and between 512 and PAGE_SIZE */
Vignesh Babu BM1368c4f2007-05-08 00:24:32 -070076 if (size > PAGE_SIZE || size < 512 || !is_power_of_2(size))
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 return -EINVAL;
78
79 /* Size cannot be smaller than the size supported by the device */
Martin K. Petersene1defc42009-05-22 17:17:49 -040080 if (size < bdev_logical_block_size(bdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 return -EINVAL;
82
83 /* Don't change the size if it is same as current */
84 if (bdev->bd_block_size != size) {
85 sync_blockdev(bdev);
86 bdev->bd_block_size = size;
87 bdev->bd_inode->i_blkbits = blksize_bits(size);
88 kill_bdev(bdev);
89 }
90 return 0;
91}
92
93EXPORT_SYMBOL(set_blocksize);
94
95int sb_set_blocksize(struct super_block *sb, int size)
96{
Linus Torvalds1da177e2005-04-16 15:20:36 -070097 if (set_blocksize(sb->s_bdev, size))
98 return 0;
99 /* If we get here, we know size is power of two
100 * and it's value is between 512 and PAGE_SIZE */
101 sb->s_blocksize = size;
Coywolf Qi Hunt38885bd2006-03-24 03:18:05 -0800102 sb->s_blocksize_bits = blksize_bits(size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 return sb->s_blocksize;
104}
105
106EXPORT_SYMBOL(sb_set_blocksize);
107
108int sb_min_blocksize(struct super_block *sb, int size)
109{
Martin K. Petersene1defc42009-05-22 17:17:49 -0400110 int minsize = bdev_logical_block_size(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 if (size < minsize)
112 size = minsize;
113 return sb_set_blocksize(sb, size);
114}
115
116EXPORT_SYMBOL(sb_min_blocksize);
117
118static int
119blkdev_get_block(struct inode *inode, sector_t iblock,
120 struct buffer_head *bh, int create)
121{
122 if (iblock >= max_block(I_BDEV(inode))) {
123 if (create)
124 return -EIO;
125
126 /*
127 * for reads, we're just trying to fill a partial page.
128 * return a hole, they will have to call get_block again
129 * before they can fill it, and they will get -EIO at that
130 * time
131 */
132 return 0;
133 }
134 bh->b_bdev = I_BDEV(inode);
135 bh->b_blocknr = iblock;
136 set_buffer_mapped(bh);
137 return 0;
138}
139
Andrew Mortonb2e895d2007-02-03 01:14:01 -0800140static int
141blkdev_get_blocks(struct inode *inode, sector_t iblock,
142 struct buffer_head *bh, int create)
143{
144 sector_t end_block = max_block(I_BDEV(inode));
145 unsigned long max_blocks = bh->b_size >> inode->i_blkbits;
146
147 if ((iblock + max_blocks) > end_block) {
148 max_blocks = end_block - iblock;
149 if ((long)max_blocks <= 0) {
150 if (create)
151 return -EIO; /* write fully beyond EOF */
152 /*
153 * It is a read which is fully beyond EOF. We return
154 * a !buffer_mapped buffer
155 */
156 max_blocks = 0;
157 }
158 }
159
160 bh->b_bdev = I_BDEV(inode);
161 bh->b_blocknr = iblock;
162 bh->b_size = max_blocks << inode->i_blkbits;
163 if (max_blocks)
164 set_buffer_mapped(bh);
165 return 0;
166}
167
168static ssize_t
169blkdev_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
170 loff_t offset, unsigned long nr_segs)
171{
172 struct file *file = iocb->ki_filp;
173 struct inode *inode = file->f_mapping->host;
174
175 return blockdev_direct_IO_no_locking(rw, iocb, inode, I_BDEV(inode),
176 iov, offset, nr_segs, blkdev_get_blocks, NULL);
177}
178
Jan Kara5cee5812009-04-27 16:43:51 +0200179int __sync_blockdev(struct block_device *bdev, int wait)
180{
181 if (!bdev)
182 return 0;
183 if (!wait)
184 return filemap_flush(bdev->bd_inode->i_mapping);
185 return filemap_write_and_wait(bdev->bd_inode->i_mapping);
186}
187
Nick Piggin585d3bc2009-02-25 10:44:19 +0100188/*
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.
191 */
192int sync_blockdev(struct block_device *bdev)
193{
Jan Kara5cee5812009-04-27 16:43:51 +0200194 return __sync_blockdev(bdev, 1);
Nick Piggin585d3bc2009-02-25 10:44:19 +0100195}
196EXPORT_SYMBOL(sync_blockdev);
197
198/*
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.
202 */
203int fsync_bdev(struct block_device *bdev)
204{
205 struct super_block *sb = get_super(bdev);
206 if (sb) {
Jan Kara60b06802009-04-27 16:43:53 +0200207 int res = sync_filesystem(sb);
Nick Piggin585d3bc2009-02-25 10:44:19 +0100208 drop_super(sb);
209 return res;
210 }
211 return sync_blockdev(bdev);
212}
Al Viro47e44912009-04-01 07:07:16 -0400213EXPORT_SYMBOL(fsync_bdev);
Nick Piggin585d3bc2009-02-25 10:44:19 +0100214
215/**
216 * freeze_bdev -- lock a filesystem and force it into a consistent state
217 * @bdev: blockdevice to lock
218 *
219 * This takes the block device bd_mount_sem to make sure no new mounts
220 * happen on bdev until thaw_bdev() is called.
221 * If a superblock is found on this device, we take the s_umount semaphore
222 * on it to make sure nobody unmounts until the snapshot creation is done.
223 * The reference counter (bd_fsfreeze_count) guarantees that only the last
224 * unfreeze process can unfreeze the frozen filesystem actually when multiple
225 * freeze requests arrive simultaneously. It counts up in freeze_bdev() and
226 * count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze
227 * actually.
228 */
229struct super_block *freeze_bdev(struct block_device *bdev)
230{
231 struct super_block *sb;
232 int error = 0;
233
234 mutex_lock(&bdev->bd_fsfreeze_mutex);
235 if (bdev->bd_fsfreeze_count > 0) {
236 bdev->bd_fsfreeze_count++;
237 sb = get_super(bdev);
238 mutex_unlock(&bdev->bd_fsfreeze_mutex);
239 return sb;
240 }
241 bdev->bd_fsfreeze_count++;
242
243 down(&bdev->bd_mount_sem);
244 sb = get_super(bdev);
245 if (sb && !(sb->s_flags & MS_RDONLY)) {
246 sb->s_frozen = SB_FREEZE_WRITE;
247 smp_wmb();
248
Jan Kara60b06802009-04-27 16:43:53 +0200249 sync_filesystem(sb);
Nick Piggin585d3bc2009-02-25 10:44:19 +0100250
251 sb->s_frozen = SB_FREEZE_TRANS;
252 smp_wmb();
253
254 sync_blockdev(sb->s_bdev);
255
256 if (sb->s_op->freeze_fs) {
257 error = sb->s_op->freeze_fs(sb);
258 if (error) {
259 printk(KERN_ERR
260 "VFS:Filesystem freeze failed\n");
261 sb->s_frozen = SB_UNFROZEN;
262 drop_super(sb);
263 up(&bdev->bd_mount_sem);
264 bdev->bd_fsfreeze_count--;
265 mutex_unlock(&bdev->bd_fsfreeze_mutex);
266 return ERR_PTR(error);
267 }
268 }
269 }
270
271 sync_blockdev(bdev);
272 mutex_unlock(&bdev->bd_fsfreeze_mutex);
273
274 return sb; /* thaw_bdev releases s->s_umount and bd_mount_sem */
275}
276EXPORT_SYMBOL(freeze_bdev);
277
278/**
279 * thaw_bdev -- unlock filesystem
280 * @bdev: blockdevice to unlock
281 * @sb: associated superblock
282 *
283 * Unlocks the filesystem and marks it writeable again after freeze_bdev().
284 */
285int thaw_bdev(struct block_device *bdev, struct super_block *sb)
286{
287 int error = 0;
288
289 mutex_lock(&bdev->bd_fsfreeze_mutex);
290 if (!bdev->bd_fsfreeze_count) {
291 mutex_unlock(&bdev->bd_fsfreeze_mutex);
292 return -EINVAL;
293 }
294
295 bdev->bd_fsfreeze_count--;
296 if (bdev->bd_fsfreeze_count > 0) {
297 if (sb)
298 drop_super(sb);
299 mutex_unlock(&bdev->bd_fsfreeze_mutex);
300 return 0;
301 }
302
303 if (sb) {
304 BUG_ON(sb->s_bdev != bdev);
305 if (!(sb->s_flags & MS_RDONLY)) {
306 if (sb->s_op->unfreeze_fs) {
307 error = sb->s_op->unfreeze_fs(sb);
308 if (error) {
309 printk(KERN_ERR
310 "VFS:Filesystem thaw failed\n");
311 sb->s_frozen = SB_FREEZE_TRANS;
312 bdev->bd_fsfreeze_count++;
313 mutex_unlock(&bdev->bd_fsfreeze_mutex);
314 return error;
315 }
316 }
317 sb->s_frozen = SB_UNFROZEN;
318 smp_wmb();
319 wake_up(&sb->s_wait_unfrozen);
320 }
321 drop_super(sb);
322 }
323
324 up(&bdev->bd_mount_sem);
325 mutex_unlock(&bdev->bd_fsfreeze_mutex);
326 return 0;
327}
328EXPORT_SYMBOL(thaw_bdev);
329
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330static int blkdev_writepage(struct page *page, struct writeback_control *wbc)
331{
332 return block_write_full_page(page, blkdev_get_block, wbc);
333}
334
335static int blkdev_readpage(struct file * file, struct page * page)
336{
337 return block_read_full_page(page, blkdev_get_block);
338}
339
Nick Piggin6272b5a2007-10-16 01:25:04 -0700340static int blkdev_write_begin(struct file *file, struct address_space *mapping,
341 loff_t pos, unsigned len, unsigned flags,
342 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343{
Nick Piggin6272b5a2007-10-16 01:25:04 -0700344 *pagep = NULL;
345 return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
346 blkdev_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347}
348
Nick Piggin6272b5a2007-10-16 01:25:04 -0700349static int blkdev_write_end(struct file *file, struct address_space *mapping,
350 loff_t pos, unsigned len, unsigned copied,
351 struct page *page, void *fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352{
Nick Piggin6272b5a2007-10-16 01:25:04 -0700353 int ret;
354 ret = block_write_end(file, mapping, pos, len, copied, page, fsdata);
355
356 unlock_page(page);
357 page_cache_release(page);
358
359 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360}
361
362/*
363 * private llseek:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800364 * for a block special file file->f_path.dentry->d_inode->i_size is zero
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 * so we compute the size by hand (just as in block_read/write above)
366 */
367static loff_t block_llseek(struct file *file, loff_t offset, int origin)
368{
369 struct inode *bd_inode = file->f_mapping->host;
370 loff_t size;
371 loff_t retval;
372
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800373 mutex_lock(&bd_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 size = i_size_read(bd_inode);
375
376 switch (origin) {
377 case 2:
378 offset += size;
379 break;
380 case 1:
381 offset += file->f_pos;
382 }
383 retval = -EINVAL;
384 if (offset >= 0 && offset <= size) {
385 if (offset != file->f_pos) {
386 file->f_pos = offset;
387 }
388 retval = offset;
389 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800390 mutex_unlock(&bd_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 return retval;
392}
393
394/*
395 * Filp is never NULL; the only case when ->fsync() is called with
396 * NULL first argument is nfsd_sync_dir() and that's not a directory.
397 */
398
399static int block_fsync(struct file *filp, struct dentry *dentry, int datasync)
400{
401 return sync_blockdev(I_BDEV(filp->f_mapping->host));
402}
403
404/*
405 * pseudo-fs
406 */
407
408static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock);
Christoph Lametere18b8902006-12-06 20:33:20 -0800409static struct kmem_cache * bdev_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410
411static struct inode *bdev_alloc_inode(struct super_block *sb)
412{
Christoph Lametere94b1762006-12-06 20:33:17 -0800413 struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 if (!ei)
415 return NULL;
416 return &ei->vfs_inode;
417}
418
419static void bdev_destroy_inode(struct inode *inode)
420{
421 struct bdev_inode *bdi = BDEV_I(inode);
422
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 kmem_cache_free(bdev_cachep, bdi);
424}
425
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700426static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427{
428 struct bdev_inode *ei = (struct bdev_inode *) foo;
429 struct block_device *bdev = &ei->bdev;
430
Christoph Lametera35afb82007-05-16 22:10:57 -0700431 memset(bdev, 0, sizeof(*bdev));
432 mutex_init(&bdev->bd_mutex);
433 sema_init(&bdev->bd_mount_sem, 1);
434 INIT_LIST_HEAD(&bdev->bd_inodes);
435 INIT_LIST_HEAD(&bdev->bd_list);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800436#ifdef CONFIG_SYSFS
Christoph Lametera35afb82007-05-16 22:10:57 -0700437 INIT_LIST_HEAD(&bdev->bd_holder_list);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800438#endif
Christoph Lametera35afb82007-05-16 22:10:57 -0700439 inode_init_once(&ei->vfs_inode);
Takashi Satofcccf502009-01-09 16:40:59 -0800440 /* Initialize mutex for freeze. */
441 mutex_init(&bdev->bd_fsfreeze_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442}
443
444static inline void __bd_forget(struct inode *inode)
445{
446 list_del_init(&inode->i_devices);
447 inode->i_bdev = NULL;
448 inode->i_mapping = &inode->i_data;
449}
450
451static void bdev_clear_inode(struct inode *inode)
452{
453 struct block_device *bdev = &BDEV_I(inode)->bdev;
454 struct list_head *p;
455 spin_lock(&bdev_lock);
456 while ( (p = bdev->bd_inodes.next) != &bdev->bd_inodes ) {
457 __bd_forget(list_entry(p, struct inode, i_devices));
458 }
459 list_del_init(&bdev->bd_list);
460 spin_unlock(&bdev_lock);
461}
462
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800463static const struct super_operations bdev_sops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 .statfs = simple_statfs,
465 .alloc_inode = bdev_alloc_inode,
466 .destroy_inode = bdev_destroy_inode,
467 .drop_inode = generic_delete_inode,
468 .clear_inode = bdev_clear_inode,
469};
470
David Howells454e2392006-06-23 02:02:57 -0700471static int bd_get_sb(struct file_system_type *fs_type,
472 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473{
David Howells454e2392006-06-23 02:02:57 -0700474 return get_sb_pseudo(fs_type, "bdev:", &bdev_sops, 0x62646576, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475}
476
477static struct file_system_type bd_type = {
478 .name = "bdev",
479 .get_sb = bd_get_sb,
480 .kill_sb = kill_anon_super,
481};
482
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800483struct super_block *blockdev_superblock __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
485void __init bdev_cache_init(void)
486{
487 int err;
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800488 struct vfsmount *bd_mnt;
489
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode),
Paul Jacksonfffb60f2006-03-24 03:16:06 -0800491 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
492 SLAB_MEM_SPREAD|SLAB_PANIC),
Paul Mundt20c2df82007-07-20 10:11:58 +0900493 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 err = register_filesystem(&bd_type);
495 if (err)
496 panic("Cannot register bdev pseudo-fs");
497 bd_mnt = kern_mount(&bd_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 if (IS_ERR(bd_mnt))
499 panic("Cannot create bdev pseudo-fs");
Catalin Marinas2e1483c2009-06-11 13:24:13 +0100500 /*
501 * This vfsmount structure is only used to obtain the
502 * blockdev_superblock, so tell kmemleak not to report it.
503 */
504 kmemleak_not_leak(bd_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */
506}
507
508/*
509 * Most likely _very_ bad one - but then it's hardly critical for small
510 * /dev and can be fixed when somebody will need really large one.
511 * Keep in mind that it will be fed through icache hash function too.
512 */
513static inline unsigned long hash(dev_t dev)
514{
515 return MAJOR(dev)+MINOR(dev);
516}
517
518static int bdev_test(struct inode *inode, void *data)
519{
520 return BDEV_I(inode)->bdev.bd_dev == *(dev_t *)data;
521}
522
523static int bdev_set(struct inode *inode, void *data)
524{
525 BDEV_I(inode)->bdev.bd_dev = *(dev_t *)data;
526 return 0;
527}
528
529static LIST_HEAD(all_bdevs);
530
531struct block_device *bdget(dev_t dev)
532{
533 struct block_device *bdev;
534 struct inode *inode;
535
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800536 inode = iget5_locked(blockdev_superblock, hash(dev),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 bdev_test, bdev_set, &dev);
538
539 if (!inode)
540 return NULL;
541
542 bdev = &BDEV_I(inode)->bdev;
543
544 if (inode->i_state & I_NEW) {
545 bdev->bd_contains = NULL;
546 bdev->bd_inode = inode;
547 bdev->bd_block_size = (1 << inode->i_blkbits);
548 bdev->bd_part_count = 0;
549 bdev->bd_invalidated = 0;
550 inode->i_mode = S_IFBLK;
551 inode->i_rdev = dev;
552 inode->i_bdev = bdev;
553 inode->i_data.a_ops = &def_blk_aops;
554 mapping_set_gfp_mask(&inode->i_data, GFP_USER);
555 inode->i_data.backing_dev_info = &default_backing_dev_info;
556 spin_lock(&bdev_lock);
557 list_add(&bdev->bd_list, &all_bdevs);
558 spin_unlock(&bdev_lock);
559 unlock_new_inode(inode);
560 }
561 return bdev;
562}
563
564EXPORT_SYMBOL(bdget);
565
Alan Jenkinsdddac6a2009-07-29 21:07:55 +0200566/**
567 * bdgrab -- Grab a reference to an already referenced block device
568 * @bdev: Block device to grab a reference to.
569 */
570struct block_device *bdgrab(struct block_device *bdev)
571{
572 atomic_inc(&bdev->bd_inode->i_count);
573 return bdev;
574}
575
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576long nr_blockdev_pages(void)
577{
Matthias Kaehlcke203a2932007-07-15 23:40:23 -0700578 struct block_device *bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 long ret = 0;
580 spin_lock(&bdev_lock);
Matthias Kaehlcke203a2932007-07-15 23:40:23 -0700581 list_for_each_entry(bdev, &all_bdevs, bd_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 ret += bdev->bd_inode->i_mapping->nrpages;
583 }
584 spin_unlock(&bdev_lock);
585 return ret;
586}
587
588void bdput(struct block_device *bdev)
589{
590 iput(bdev->bd_inode);
591}
592
593EXPORT_SYMBOL(bdput);
594
595static struct block_device *bd_acquire(struct inode *inode)
596{
597 struct block_device *bdev;
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700598
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 spin_lock(&bdev_lock);
600 bdev = inode->i_bdev;
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700601 if (bdev) {
602 atomic_inc(&bdev->bd_inode->i_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 spin_unlock(&bdev_lock);
604 return bdev;
605 }
606 spin_unlock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700607
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 bdev = bdget(inode->i_rdev);
609 if (bdev) {
610 spin_lock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700611 if (!inode->i_bdev) {
612 /*
613 * We take an additional bd_inode->i_count for inode,
614 * and it's released in clear_inode() of inode.
615 * So, we can access it via ->i_mapping always
616 * without igrab().
617 */
618 atomic_inc(&bdev->bd_inode->i_count);
619 inode->i_bdev = bdev;
620 inode->i_mapping = bdev->bd_inode->i_mapping;
621 list_add(&inode->i_devices, &bdev->bd_inodes);
622 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 spin_unlock(&bdev_lock);
624 }
625 return bdev;
626}
627
628/* Call when you free inode */
629
630void bd_forget(struct inode *inode)
631{
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700632 struct block_device *bdev = NULL;
633
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 spin_lock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700635 if (inode->i_bdev) {
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800636 if (!sb_is_blkdev_sb(inode->i_sb))
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700637 bdev = inode->i_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 __bd_forget(inode);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700639 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 spin_unlock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700641
642 if (bdev)
643 iput(bdev->bd_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644}
645
646int bd_claim(struct block_device *bdev, void *holder)
647{
648 int res;
649 spin_lock(&bdev_lock);
650
651 /* first decide result */
652 if (bdev->bd_holder == holder)
653 res = 0; /* already a holder */
654 else if (bdev->bd_holder != NULL)
655 res = -EBUSY; /* held by someone else */
656 else if (bdev->bd_contains == bdev)
657 res = 0; /* is a whole device which isn't held */
658
659 else if (bdev->bd_contains->bd_holder == bd_claim)
660 res = 0; /* is a partition of a device that is being partitioned */
661 else if (bdev->bd_contains->bd_holder != NULL)
662 res = -EBUSY; /* is a partition of a held device */
663 else
664 res = 0; /* is a partition of an un-held device */
665
666 /* now impose change */
667 if (res==0) {
668 /* note that for a whole device bd_holders
669 * will be incremented twice, and bd_holder will
670 * be set to bd_claim before being set to holder
671 */
672 bdev->bd_contains->bd_holders ++;
673 bdev->bd_contains->bd_holder = bd_claim;
674 bdev->bd_holders++;
675 bdev->bd_holder = holder;
676 }
677 spin_unlock(&bdev_lock);
678 return res;
679}
680
681EXPORT_SYMBOL(bd_claim);
682
683void bd_release(struct block_device *bdev)
684{
685 spin_lock(&bdev_lock);
686 if (!--bdev->bd_contains->bd_holders)
687 bdev->bd_contains->bd_holder = NULL;
688 if (!--bdev->bd_holders)
689 bdev->bd_holder = NULL;
690 spin_unlock(&bdev_lock);
691}
692
693EXPORT_SYMBOL(bd_release);
694
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800695#ifdef CONFIG_SYSFS
696/*
697 * Functions for bd_claim_by_kobject / bd_release_from_kobject
698 *
699 * If a kobject is passed to bd_claim_by_kobject()
700 * and the kobject has a parent directory,
701 * following symlinks are created:
702 * o from the kobject to the claimed bdev
703 * o from "holders" directory of the bdev to the parent of the kobject
704 * bd_release_from_kobject() removes these symlinks.
705 *
706 * Example:
707 * If /dev/dm-0 maps to /dev/sda, kobject corresponding to
708 * /sys/block/dm-0/slaves is passed to bd_claim_by_kobject(), then:
709 * /sys/block/dm-0/slaves/sda --> /sys/block/sda
710 * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0
711 */
712
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700713static int add_symlink(struct kobject *from, struct kobject *to)
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800714{
715 if (!from || !to)
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700716 return 0;
717 return sysfs_create_link(from, to, kobject_name(to));
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800718}
719
720static void del_symlink(struct kobject *from, struct kobject *to)
721{
722 if (!from || !to)
723 return;
724 sysfs_remove_link(from, kobject_name(to));
725}
726
727/*
728 * 'struct bd_holder' contains pointers to kobjects symlinked by
729 * bd_claim_by_kobject.
730 * It's connected to bd_holder_list which is protected by bdev->bd_sem.
731 */
732struct bd_holder {
733 struct list_head list; /* chain of holders of the bdev */
734 int count; /* references from the holder */
735 struct kobject *sdir; /* holder object, e.g. "/block/dm-0/slaves" */
736 struct kobject *hdev; /* e.g. "/block/dm-0" */
737 struct kobject *hdir; /* e.g. "/block/sda/holders" */
738 struct kobject *sdev; /* e.g. "/block/sda" */
739};
740
741/*
742 * Get references of related kobjects at once.
743 * Returns 1 on success. 0 on failure.
744 *
745 * Should call bd_holder_release_dirs() after successful use.
746 */
747static int bd_holder_grab_dirs(struct block_device *bdev,
748 struct bd_holder *bo)
749{
750 if (!bdev || !bo)
751 return 0;
752
753 bo->sdir = kobject_get(bo->sdir);
754 if (!bo->sdir)
755 return 0;
756
757 bo->hdev = kobject_get(bo->sdir->parent);
758 if (!bo->hdev)
759 goto fail_put_sdir;
760
Tejun Heo0762b8b2008-08-25 19:56:12 +0900761 bo->sdev = kobject_get(&part_to_dev(bdev->bd_part)->kobj);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800762 if (!bo->sdev)
763 goto fail_put_hdev;
764
Tejun Heo4c465012008-08-25 19:56:11 +0900765 bo->hdir = kobject_get(bdev->bd_part->holder_dir);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800766 if (!bo->hdir)
767 goto fail_put_sdev;
768
769 return 1;
770
771fail_put_sdev:
772 kobject_put(bo->sdev);
773fail_put_hdev:
774 kobject_put(bo->hdev);
775fail_put_sdir:
776 kobject_put(bo->sdir);
777
778 return 0;
779}
780
781/* Put references of related kobjects at once. */
782static void bd_holder_release_dirs(struct bd_holder *bo)
783{
784 kobject_put(bo->hdir);
785 kobject_put(bo->sdev);
786 kobject_put(bo->hdev);
787 kobject_put(bo->sdir);
788}
789
790static struct bd_holder *alloc_bd_holder(struct kobject *kobj)
791{
792 struct bd_holder *bo;
793
794 bo = kzalloc(sizeof(*bo), GFP_KERNEL);
795 if (!bo)
796 return NULL;
797
798 bo->count = 1;
799 bo->sdir = kobj;
800
801 return bo;
802}
803
804static void free_bd_holder(struct bd_holder *bo)
805{
806 kfree(bo);
807}
808
809/**
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500810 * find_bd_holder - find matching struct bd_holder from the block device
811 *
812 * @bdev: struct block device to be searched
813 * @bo: target struct bd_holder
814 *
815 * Returns matching entry with @bo in @bdev->bd_holder_list.
816 * If found, increment the reference count and return the pointer.
817 * If not found, returns NULL.
818 */
Andrew Morton36a561d2006-10-30 22:07:03 -0800819static struct bd_holder *find_bd_holder(struct block_device *bdev,
820 struct bd_holder *bo)
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500821{
822 struct bd_holder *tmp;
823
824 list_for_each_entry(tmp, &bdev->bd_holder_list, list)
825 if (tmp->sdir == bo->sdir) {
826 tmp->count++;
827 return tmp;
828 }
829
830 return NULL;
831}
832
833/**
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800834 * add_bd_holder - create sysfs symlinks for bd_claim() relationship
835 *
836 * @bdev: block device to be bd_claimed
837 * @bo: preallocated and initialized by alloc_bd_holder()
838 *
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500839 * Add @bo to @bdev->bd_holder_list, create symlinks.
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800840 *
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500841 * Returns 0 if symlinks are created.
842 * Returns -ve if something fails.
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800843 */
844static int add_bd_holder(struct block_device *bdev, struct bd_holder *bo)
845{
Johannes Weiner4e916722007-07-15 23:41:25 -0700846 int err;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800847
848 if (!bo)
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700849 return -EINVAL;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800850
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800851 if (!bd_holder_grab_dirs(bdev, bo))
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700852 return -EBUSY;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800853
Johannes Weiner4e916722007-07-15 23:41:25 -0700854 err = add_symlink(bo->sdir, bo->sdev);
855 if (err)
856 return err;
857
858 err = add_symlink(bo->hdir, bo->hdev);
859 if (err) {
860 del_symlink(bo->sdir, bo->sdev);
861 return err;
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700862 }
Johannes Weiner4e916722007-07-15 23:41:25 -0700863
864 list_add_tail(&bo->list, &bdev->bd_holder_list);
865 return 0;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800866}
867
868/**
869 * del_bd_holder - delete sysfs symlinks for bd_claim() relationship
870 *
871 * @bdev: block device to be bd_claimed
872 * @kobj: holder's kobject
873 *
874 * If there is matching entry with @kobj in @bdev->bd_holder_list
875 * and no other bd_claim() from the same kobject,
876 * remove the struct bd_holder from the list, delete symlinks for it.
877 *
878 * Returns a pointer to the struct bd_holder when it's removed from the list
879 * and ready to be freed.
880 * Returns NULL if matching claim isn't found or there is other bd_claim()
881 * by the same kobject.
882 */
883static struct bd_holder *del_bd_holder(struct block_device *bdev,
884 struct kobject *kobj)
885{
886 struct bd_holder *bo;
887
888 list_for_each_entry(bo, &bdev->bd_holder_list, list) {
889 if (bo->sdir == kobj) {
890 bo->count--;
891 BUG_ON(bo->count < 0);
892 if (!bo->count) {
893 list_del(&bo->list);
894 del_symlink(bo->sdir, bo->sdev);
895 del_symlink(bo->hdir, bo->hdev);
896 bd_holder_release_dirs(bo);
897 return bo;
898 }
899 break;
900 }
901 }
902
903 return NULL;
904}
905
906/**
907 * bd_claim_by_kobject - bd_claim() with additional kobject signature
908 *
909 * @bdev: block device to be claimed
910 * @holder: holder's signature
911 * @kobj: holder's kobject
912 *
913 * Do bd_claim() and if it succeeds, create sysfs symlinks between
914 * the bdev and the holder's kobject.
915 * Use bd_release_from_kobject() when relesing the claimed bdev.
916 *
917 * Returns 0 on success. (same as bd_claim())
918 * Returns errno on failure.
919 */
920static int bd_claim_by_kobject(struct block_device *bdev, void *holder,
921 struct kobject *kobj)
922{
Johannes Weiner4e916722007-07-15 23:41:25 -0700923 int err;
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500924 struct bd_holder *bo, *found;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800925
926 if (!kobj)
927 return -EINVAL;
928
929 bo = alloc_bd_holder(kobj);
930 if (!bo)
931 return -ENOMEM;
932
Peter Zijlstra2e7b6512006-12-08 02:36:13 -0800933 mutex_lock(&bdev->bd_mutex);
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500934
Johannes Weiner4e916722007-07-15 23:41:25 -0700935 err = bd_claim(bdev, holder);
936 if (err)
Andrew Morton4210df22007-07-15 23:41:28 -0700937 goto fail;
Johannes Weiner4e916722007-07-15 23:41:25 -0700938
939 found = find_bd_holder(bdev, bo);
940 if (found)
Andrew Morton4210df22007-07-15 23:41:28 -0700941 goto fail;
Johannes Weiner4e916722007-07-15 23:41:25 -0700942
943 err = add_bd_holder(bdev, bo);
944 if (err)
945 bd_release(bdev);
Andrew Morton4210df22007-07-15 23:41:28 -0700946 else
947 bo = NULL;
948fail:
Jun'ichi Nomurab4cf1b72006-03-27 01:18:00 -0800949 mutex_unlock(&bdev->bd_mutex);
Andrew Morton4210df22007-07-15 23:41:28 -0700950 free_bd_holder(bo);
Johannes Weiner4e916722007-07-15 23:41:25 -0700951 return err;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800952}
953
954/**
955 * bd_release_from_kobject - bd_release() with additional kobject signature
956 *
957 * @bdev: block device to be released
958 * @kobj: holder's kobject
959 *
960 * Do bd_release() and remove sysfs symlinks created by bd_claim_by_kobject().
961 */
962static void bd_release_from_kobject(struct block_device *bdev,
963 struct kobject *kobj)
964{
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800965 if (!kobj)
966 return;
967
Peter Zijlstra2e7b6512006-12-08 02:36:13 -0800968 mutex_lock(&bdev->bd_mutex);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800969 bd_release(bdev);
Andrew Morton4210df22007-07-15 23:41:28 -0700970 free_bd_holder(del_bd_holder(bdev, kobj));
Jun'ichi Nomurab4cf1b72006-03-27 01:18:00 -0800971 mutex_unlock(&bdev->bd_mutex);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800972}
973
974/**
975 * bd_claim_by_disk - wrapper function for bd_claim_by_kobject()
976 *
977 * @bdev: block device to be claimed
978 * @holder: holder's signature
979 * @disk: holder's gendisk
980 *
981 * Call bd_claim_by_kobject() with getting @disk->slave_dir.
982 */
983int bd_claim_by_disk(struct block_device *bdev, void *holder,
984 struct gendisk *disk)
985{
986 return bd_claim_by_kobject(bdev, holder, kobject_get(disk->slave_dir));
987}
988EXPORT_SYMBOL_GPL(bd_claim_by_disk);
989
990/**
991 * bd_release_from_disk - wrapper function for bd_release_from_kobject()
992 *
993 * @bdev: block device to be claimed
994 * @disk: holder's gendisk
995 *
996 * Call bd_release_from_kobject() and put @disk->slave_dir.
997 */
998void bd_release_from_disk(struct block_device *bdev, struct gendisk *disk)
999{
1000 bd_release_from_kobject(bdev, disk->slave_dir);
1001 kobject_put(disk->slave_dir);
1002}
1003EXPORT_SYMBOL_GPL(bd_release_from_disk);
1004#endif
1005
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006/*
1007 * Tries to open block device by device number. Use it ONLY if you
1008 * really do not have anything better - i.e. when you are behind a
1009 * truly sucky interface and all you are given is a device number. _Never_
1010 * to be used for internal purposes. If you ever need it - reconsider
1011 * your API.
1012 */
Al Viroaeb5d722008-09-02 15:28:45 -04001013struct block_device *open_by_devnum(dev_t dev, fmode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014{
1015 struct block_device *bdev = bdget(dev);
1016 int err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 if (bdev)
Al Viro572c4892007-10-08 13:24:05 -04001018 err = blkdev_get(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 return err ? ERR_PTR(err) : bdev;
1020}
1021
1022EXPORT_SYMBOL(open_by_devnum);
1023
Andrew Patterson0c002c22008-09-04 14:27:20 -06001024/**
Andrew Patterson56ade442008-09-04 14:27:40 -06001025 * flush_disk - invalidates all buffer-cache entries on a disk
1026 *
1027 * @bdev: struct block device to be flushed
1028 *
1029 * Invalidates all buffer-cache entries on a disk. It should be called
1030 * when a disk has been changed -- either by a media change or online
1031 * resize.
1032 */
1033static void flush_disk(struct block_device *bdev)
1034{
1035 if (__invalidate_device(bdev)) {
1036 char name[BDEVNAME_SIZE] = "";
1037
1038 if (bdev->bd_disk)
1039 disk_name(bdev->bd_disk, 0, name);
1040 printk(KERN_WARNING "VFS: busy inodes on changed media or "
1041 "resized disk %s\n", name);
1042 }
1043
1044 if (!bdev->bd_disk)
1045 return;
1046 if (disk_partitionable(bdev->bd_disk))
1047 bdev->bd_invalidated = 1;
1048}
1049
1050/**
Randy Dunlap57d1b532008-10-09 10:42:38 +02001051 * check_disk_size_change - checks for disk size change and adjusts bdev size.
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001052 * @disk: struct gendisk to check
1053 * @bdev: struct bdev to adjust.
1054 *
1055 * This routine checks to see if the bdev size does not match the disk size
1056 * and adjusts it if it differs.
1057 */
1058void check_disk_size_change(struct gendisk *disk, struct block_device *bdev)
1059{
1060 loff_t disk_size, bdev_size;
1061
1062 disk_size = (loff_t)get_capacity(disk) << 9;
1063 bdev_size = i_size_read(bdev->bd_inode);
1064 if (disk_size != bdev_size) {
1065 char name[BDEVNAME_SIZE];
1066
1067 disk_name(disk, 0, name);
1068 printk(KERN_INFO
1069 "%s: detected capacity change from %lld to %lld\n",
1070 name, bdev_size, disk_size);
1071 i_size_write(bdev->bd_inode, disk_size);
Andrew Patterson608aeef2008-09-04 14:27:45 -06001072 flush_disk(bdev);
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001073 }
1074}
1075EXPORT_SYMBOL(check_disk_size_change);
1076
1077/**
Randy Dunlap57d1b532008-10-09 10:42:38 +02001078 * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back
Andrew Patterson0c002c22008-09-04 14:27:20 -06001079 * @disk: struct gendisk to be revalidated
1080 *
1081 * This routine is a wrapper for lower-level driver's revalidate_disk
1082 * call-backs. It is used to do common pre and post operations needed
1083 * for all revalidate_disk operations.
1084 */
1085int revalidate_disk(struct gendisk *disk)
1086{
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001087 struct block_device *bdev;
Andrew Patterson0c002c22008-09-04 14:27:20 -06001088 int ret = 0;
1089
1090 if (disk->fops->revalidate_disk)
1091 ret = disk->fops->revalidate_disk(disk);
1092
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001093 bdev = bdget_disk(disk, 0);
1094 if (!bdev)
1095 return ret;
1096
1097 mutex_lock(&bdev->bd_mutex);
1098 check_disk_size_change(disk, bdev);
1099 mutex_unlock(&bdev->bd_mutex);
1100 bdput(bdev);
Andrew Patterson0c002c22008-09-04 14:27:20 -06001101 return ret;
1102}
1103EXPORT_SYMBOL(revalidate_disk);
1104
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105/*
1106 * This routine checks whether a removable media has been changed,
1107 * and invalidates all buffer-cache-entries in that case. This
1108 * is a relatively slow routine, so we have to try to minimize using
1109 * it. Thus it is called only upon a 'mount' or 'open'. This
1110 * is the best way of combining speed and utility, I think.
1111 * People changing diskettes in the middle of an operation deserve
1112 * to lose :-)
1113 */
1114int check_disk_change(struct block_device *bdev)
1115{
1116 struct gendisk *disk = bdev->bd_disk;
1117 struct block_device_operations * bdops = disk->fops;
1118
1119 if (!bdops->media_changed)
1120 return 0;
1121 if (!bdops->media_changed(bdev->bd_disk))
1122 return 0;
1123
Andrew Patterson56ade442008-09-04 14:27:40 -06001124 flush_disk(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 if (bdops->revalidate_disk)
1126 bdops->revalidate_disk(bdev->bd_disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 return 1;
1128}
1129
1130EXPORT_SYMBOL(check_disk_change);
1131
1132void bd_set_size(struct block_device *bdev, loff_t size)
1133{
Martin K. Petersene1defc42009-05-22 17:17:49 -04001134 unsigned bsize = bdev_logical_block_size(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135
1136 bdev->bd_inode->i_size = size;
1137 while (bsize < PAGE_CACHE_SIZE) {
1138 if (size & bsize)
1139 break;
1140 bsize <<= 1;
1141 }
1142 bdev->bd_block_size = bsize;
1143 bdev->bd_inode->i_blkbits = blksize_bits(bsize);
1144}
1145EXPORT_SYMBOL(bd_set_size);
1146
Al Viro9a1c3542008-02-22 20:40:24 -05001147static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
NeilBrown37be4122006-12-08 02:36:16 -08001148
Peter Zijlstra6d740cd2007-02-20 13:58:18 -08001149/*
1150 * bd_mutex locking:
1151 *
1152 * mutex_lock(part->bd_mutex)
1153 * mutex_lock_nested(whole->bd_mutex, 1)
1154 */
1155
Al Viro572c4892007-10-08 13:24:05 -04001156static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 struct gendisk *disk;
Pavel Emelyanov7db9cfd2008-06-05 22:46:27 -07001159 int ret;
Tejun Heocf771cb2008-09-03 09:01:09 +02001160 int partno;
Al Virofe6e9c12008-06-23 08:30:55 -04001161 int perm = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162
Al Viro572c4892007-10-08 13:24:05 -04001163 if (mode & FMODE_READ)
Al Virofe6e9c12008-06-23 08:30:55 -04001164 perm |= MAY_READ;
Al Viro572c4892007-10-08 13:24:05 -04001165 if (mode & FMODE_WRITE)
Al Virofe6e9c12008-06-23 08:30:55 -04001166 perm |= MAY_WRITE;
1167 /*
1168 * hooks: /n/, see "layering violations".
1169 */
1170 ret = devcgroup_inode_permission(bdev->bd_inode, perm);
Al Viro82666022008-08-01 05:32:04 -04001171 if (ret != 0) {
1172 bdput(bdev);
Pavel Emelyanov7db9cfd2008-06-05 22:46:27 -07001173 return ret;
Al Viro82666022008-08-01 05:32:04 -04001174 }
Pavel Emelyanov7db9cfd2008-06-05 22:46:27 -07001175
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 lock_kernel();
NeilBrownd3374822009-01-09 08:31:10 +11001177 restart:
Tejun Heo0762b8b2008-08-25 19:56:12 +09001178
Tejun Heo89f97492008-11-05 10:21:06 +01001179 ret = -ENXIO;
Tejun Heocf771cb2008-09-03 09:01:09 +02001180 disk = get_gendisk(bdev->bd_dev, &partno);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001181 if (!disk)
1182 goto out_unlock_kernel;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
NeilBrown6796bf52006-12-08 02:36:16 -08001184 mutex_lock_nested(&bdev->bd_mutex, for_part);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 if (!bdev->bd_openers) {
1186 bdev->bd_disk = disk;
1187 bdev->bd_contains = bdev;
Tejun Heocf771cb2008-09-03 09:01:09 +02001188 if (!partno) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 struct backing_dev_info *bdi;
Tejun Heo89f97492008-11-05 10:21:06 +01001190
1191 ret = -ENXIO;
1192 bdev->bd_part = disk_get_part(disk, partno);
1193 if (!bdev->bd_part)
1194 goto out_clear;
1195
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 if (disk->fops->open) {
Al Viro572c4892007-10-08 13:24:05 -04001197 ret = disk->fops->open(bdev, mode);
NeilBrownd3374822009-01-09 08:31:10 +11001198 if (ret == -ERESTARTSYS) {
1199 /* Lost a race with 'disk' being
1200 * deleted, try again.
1201 * See md.c
1202 */
1203 disk_put_part(bdev->bd_part);
1204 bdev->bd_part = NULL;
1205 module_put(disk->fops->owner);
1206 put_disk(disk);
1207 bdev->bd_disk = NULL;
1208 mutex_unlock(&bdev->bd_mutex);
1209 goto restart;
1210 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 if (ret)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001212 goto out_clear;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 }
1214 if (!bdev->bd_openers) {
1215 bd_set_size(bdev,(loff_t)get_capacity(disk)<<9);
1216 bdi = blk_get_backing_dev_info(bdev);
1217 if (bdi == NULL)
1218 bdi = &default_backing_dev_info;
1219 bdev->bd_inode->i_data.backing_dev_info = bdi;
1220 }
1221 if (bdev->bd_invalidated)
1222 rescan_partitions(disk, bdev);
1223 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 struct block_device *whole;
1225 whole = bdget_disk(disk, 0);
1226 ret = -ENOMEM;
1227 if (!whole)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001228 goto out_clear;
NeilBrown37be4122006-12-08 02:36:16 -08001229 BUG_ON(for_part);
Al Viro572c4892007-10-08 13:24:05 -04001230 ret = __blkdev_get(whole, mode, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 if (ret)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001232 goto out_clear;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 bdev->bd_contains = whole;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 bdev->bd_inode->i_data.backing_dev_info =
1235 whole->bd_inode->i_data.backing_dev_info;
Tejun Heo89f97492008-11-05 10:21:06 +01001236 bdev->bd_part = disk_get_part(disk, partno);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001237 if (!(disk->flags & GENHD_FL_UP) ||
Tejun Heo89f97492008-11-05 10:21:06 +01001238 !bdev->bd_part || !bdev->bd_part->nr_sects) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 ret = -ENXIO;
Tejun Heo0762b8b2008-08-25 19:56:12 +09001240 goto out_clear;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 }
Tejun Heo89f97492008-11-05 10:21:06 +01001242 bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 }
1244 } else {
1245 put_disk(disk);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001246 module_put(disk->fops->owner);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001247 disk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 if (bdev->bd_contains == bdev) {
1249 if (bdev->bd_disk->fops->open) {
Al Viro572c4892007-10-08 13:24:05 -04001250 ret = bdev->bd_disk->fops->open(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 if (ret)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001252 goto out_unlock_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 }
1254 if (bdev->bd_invalidated)
1255 rescan_partitions(bdev->bd_disk, bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 }
1257 }
1258 bdev->bd_openers++;
NeilBrown37be4122006-12-08 02:36:16 -08001259 if (for_part)
1260 bdev->bd_part_count++;
Arjan van de Venc039e312006-03-23 03:00:28 -08001261 mutex_unlock(&bdev->bd_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 unlock_kernel();
1263 return 0;
1264
Tejun Heo0762b8b2008-08-25 19:56:12 +09001265 out_clear:
Tejun Heo89f97492008-11-05 10:21:06 +01001266 disk_put_part(bdev->bd_part);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 bdev->bd_disk = NULL;
Tejun Heo0762b8b2008-08-25 19:56:12 +09001268 bdev->bd_part = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info;
1270 if (bdev != bdev->bd_contains)
Al Viro572c4892007-10-08 13:24:05 -04001271 __blkdev_put(bdev->bd_contains, mode, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272 bdev->bd_contains = NULL;
Tejun Heo0762b8b2008-08-25 19:56:12 +09001273 out_unlock_bdev:
Arjan van de Venc039e312006-03-23 03:00:28 -08001274 mutex_unlock(&bdev->bd_mutex);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001275 out_unlock_kernel:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 unlock_kernel();
Tejun Heo0762b8b2008-08-25 19:56:12 +09001277
Tejun Heo0762b8b2008-08-25 19:56:12 +09001278 if (disk)
1279 module_put(disk->fops->owner);
1280 put_disk(disk);
1281 bdput(bdev);
1282
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 return ret;
1284}
1285
Al Viro572c4892007-10-08 13:24:05 -04001286int blkdev_get(struct block_device *bdev, fmode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287{
Al Viro572c4892007-10-08 13:24:05 -04001288 return __blkdev_get(bdev, mode, 0);
NeilBrown37be4122006-12-08 02:36:16 -08001289}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290EXPORT_SYMBOL(blkdev_get);
1291
1292static int blkdev_open(struct inode * inode, struct file * filp)
1293{
1294 struct block_device *bdev;
1295 int res;
1296
1297 /*
1298 * Preserve backwards compatibility and allow large file access
1299 * even if userspace doesn't ask for it explicitly. Some mkfs
1300 * binary needs it. We might want to drop this workaround
1301 * during an unstable branch.
1302 */
1303 filp->f_flags |= O_LARGEFILE;
1304
Al Viro572c4892007-10-08 13:24:05 -04001305 if (filp->f_flags & O_NDELAY)
1306 filp->f_mode |= FMODE_NDELAY;
1307 if (filp->f_flags & O_EXCL)
1308 filp->f_mode |= FMODE_EXCL;
1309 if ((filp->f_flags & O_ACCMODE) == 3)
1310 filp->f_mode |= FMODE_WRITE_IOCTL;
1311
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 bdev = bd_acquire(inode);
Pavel Emelianov6a2aae02006-10-28 10:38:33 -07001313 if (bdev == NULL)
1314 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315
Al Viro572c4892007-10-08 13:24:05 -04001316 filp->f_mapping = bdev->bd_inode->i_mapping;
1317
1318 res = blkdev_get(bdev, filp->f_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 if (res)
1320 return res;
1321
Christoph Hellwigebbefc02008-11-05 14:54:41 +01001322 if (filp->f_mode & FMODE_EXCL) {
1323 res = bd_claim(bdev, filp);
1324 if (res)
1325 goto out_blkdev_put;
1326 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327
Christoph Hellwigebbefc02008-11-05 14:54:41 +01001328 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329
Christoph Hellwigebbefc02008-11-05 14:54:41 +01001330 out_blkdev_put:
Al Viro9a1c3542008-02-22 20:40:24 -05001331 blkdev_put(bdev, filp->f_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 return res;
1333}
1334
Al Viro9a1c3542008-02-22 20:40:24 -05001335static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part)
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001336{
1337 int ret = 0;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001338 struct gendisk *disk = bdev->bd_disk;
NeilBrown37be4122006-12-08 02:36:16 -08001339 struct block_device *victim = NULL;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001340
NeilBrown6796bf52006-12-08 02:36:16 -08001341 mutex_lock_nested(&bdev->bd_mutex, for_part);
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001342 lock_kernel();
NeilBrown37be4122006-12-08 02:36:16 -08001343 if (for_part)
1344 bdev->bd_part_count--;
1345
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001346 if (!--bdev->bd_openers) {
1347 sync_blockdev(bdev);
1348 kill_bdev(bdev);
1349 }
1350 if (bdev->bd_contains == bdev) {
1351 if (disk->fops->release)
Al Viro9a1c3542008-02-22 20:40:24 -05001352 ret = disk->fops->release(disk, mode);
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001353 }
1354 if (!bdev->bd_openers) {
1355 struct module *owner = disk->fops->owner;
1356
1357 put_disk(disk);
1358 module_put(owner);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001359 disk_put_part(bdev->bd_part);
1360 bdev->bd_part = NULL;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001361 bdev->bd_disk = NULL;
1362 bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info;
NeilBrown37be4122006-12-08 02:36:16 -08001363 if (bdev != bdev->bd_contains)
1364 victim = bdev->bd_contains;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001365 bdev->bd_contains = NULL;
1366 }
1367 unlock_kernel();
1368 mutex_unlock(&bdev->bd_mutex);
1369 bdput(bdev);
NeilBrown37be4122006-12-08 02:36:16 -08001370 if (victim)
Al Viro9a1c3542008-02-22 20:40:24 -05001371 __blkdev_put(victim, mode, 1);
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001372 return ret;
1373}
1374
Al Viro9a1c3542008-02-22 20:40:24 -05001375int blkdev_put(struct block_device *bdev, fmode_t mode)
NeilBrown37be4122006-12-08 02:36:16 -08001376{
Al Viro9a1c3542008-02-22 20:40:24 -05001377 return __blkdev_put(bdev, mode, 0);
NeilBrown37be4122006-12-08 02:36:16 -08001378}
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001379EXPORT_SYMBOL(blkdev_put);
1380
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381static int blkdev_close(struct inode * inode, struct file * filp)
1382{
1383 struct block_device *bdev = I_BDEV(filp->f_mapping->host);
1384 if (bdev->bd_holder == filp)
1385 bd_release(bdev);
Al Viro9a1c3542008-02-22 20:40:24 -05001386 return blkdev_put(bdev, filp->f_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387}
1388
Arnd Bergmannbb93e3a2005-06-23 00:10:15 -07001389static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390{
Al Viro56b26ad2008-09-19 03:17:36 -04001391 struct block_device *bdev = I_BDEV(file->f_mapping->host);
1392 fmode_t mode = file->f_mode;
Christoph Hellwigfd4ce1a2008-11-05 14:58:42 +01001393
1394 /*
1395 * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have
1396 * to updated it before every ioctl.
1397 */
Al Viro56b26ad2008-09-19 03:17:36 -04001398 if (file->f_flags & O_NDELAY)
Christoph Hellwigfd4ce1a2008-11-05 14:58:42 +01001399 mode |= FMODE_NDELAY;
1400 else
1401 mode &= ~FMODE_NDELAY;
1402
Al Viro56b26ad2008-09-19 03:17:36 -04001403 return blkdev_ioctl(bdev, mode, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404}
1405
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001406/*
Christoph Hellwigeef99382009-08-20 17:43:41 +02001407 * Write data to the block device. Only intended for the block device itself
1408 * and the raw driver which basically is a fake block device.
1409 *
1410 * Does not take i_mutex for the write and thus is not for general purpose
1411 * use.
1412 */
1413ssize_t blkdev_aio_write(struct kiocb *iocb, const struct iovec *iov,
1414 unsigned long nr_segs, loff_t pos)
1415{
1416 struct file *file = iocb->ki_filp;
1417 ssize_t ret;
1418
1419 BUG_ON(iocb->ki_pos != pos);
1420
1421 ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos);
1422 if (ret > 0 || ret == -EIOCBQUEUED) {
1423 ssize_t err;
1424
1425 err = generic_write_sync(file, pos, ret);
1426 if (err < 0 && ret > 0)
1427 ret = err;
1428 }
1429 return ret;
1430}
1431EXPORT_SYMBOL_GPL(blkdev_aio_write);
1432
1433/*
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001434 * Try to release a page associated with block device when the system
1435 * is under memory pressure.
1436 */
1437static int blkdev_releasepage(struct page *page, gfp_t wait)
1438{
1439 struct super_block *super = BDEV_I(page->mapping->host)->bdev.bd_super;
1440
1441 if (super && super->s_op->bdev_try_to_free_page)
1442 return super->s_op->bdev_try_to_free_page(super, page, wait);
1443
1444 return try_to_free_buffers(page);
1445}
1446
Adrian Bunk4c54ac62008-02-18 13:48:31 +01001447static const struct address_space_operations def_blk_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 .readpage = blkdev_readpage,
1449 .writepage = blkdev_writepage,
1450 .sync_page = block_sync_page,
Nick Piggin6272b5a2007-10-16 01:25:04 -07001451 .write_begin = blkdev_write_begin,
1452 .write_end = blkdev_write_end,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 .writepages = generic_writepages,
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001454 .releasepage = blkdev_releasepage,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 .direct_IO = blkdev_direct_IO,
1456};
1457
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -08001458const struct file_operations def_blk_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 .open = blkdev_open,
1460 .release = blkdev_close,
1461 .llseek = block_llseek,
Badari Pulavarty543ade12006-09-30 23:28:48 -07001462 .read = do_sync_read,
1463 .write = do_sync_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 .aio_read = generic_file_aio_read,
Christoph Hellwigeef99382009-08-20 17:43:41 +02001465 .aio_write = blkdev_aio_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466 .mmap = generic_file_mmap,
1467 .fsync = block_fsync,
Arnd Bergmannbb93e3a2005-06-23 00:10:15 -07001468 .unlocked_ioctl = block_ioctl,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469#ifdef CONFIG_COMPAT
1470 .compat_ioctl = compat_blkdev_ioctl,
1471#endif
Jens Axboe7f9c51f2006-05-01 19:59:32 +02001472 .splice_read = generic_file_splice_read,
1473 .splice_write = generic_file_splice_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474};
1475
1476int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg)
1477{
1478 int res;
1479 mm_segment_t old_fs = get_fs();
1480 set_fs(KERNEL_DS);
Al Viro56b26ad2008-09-19 03:17:36 -04001481 res = blkdev_ioctl(bdev, 0, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 set_fs(old_fs);
1483 return res;
1484}
1485
1486EXPORT_SYMBOL(ioctl_by_bdev);
1487
1488/**
1489 * lookup_bdev - lookup a struct block_device by name
Randy Dunlap94e29592009-01-06 14:41:15 -08001490 * @pathname: special file representing the block device
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 *
Randy Dunlap57d1b532008-10-09 10:42:38 +02001492 * Get a reference to the blockdevice at @pathname in the current
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 * namespace if possible and return it. Return ERR_PTR(error)
1494 * otherwise.
1495 */
Al Viro421748e2008-08-02 01:04:36 -04001496struct block_device *lookup_bdev(const char *pathname)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497{
1498 struct block_device *bdev;
1499 struct inode *inode;
Al Viro421748e2008-08-02 01:04:36 -04001500 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 int error;
1502
Al Viro421748e2008-08-02 01:04:36 -04001503 if (!pathname || !*pathname)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 return ERR_PTR(-EINVAL);
1505
Al Viro421748e2008-08-02 01:04:36 -04001506 error = kern_path(pathname, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 if (error)
1508 return ERR_PTR(error);
1509
Al Viro421748e2008-08-02 01:04:36 -04001510 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 error = -ENOTBLK;
1512 if (!S_ISBLK(inode->i_mode))
1513 goto fail;
1514 error = -EACCES;
Al Viro421748e2008-08-02 01:04:36 -04001515 if (path.mnt->mnt_flags & MNT_NODEV)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 goto fail;
1517 error = -ENOMEM;
1518 bdev = bd_acquire(inode);
1519 if (!bdev)
1520 goto fail;
1521out:
Al Viro421748e2008-08-02 01:04:36 -04001522 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 return bdev;
1524fail:
1525 bdev = ERR_PTR(error);
1526 goto out;
1527}
Al Virod5686b42008-08-01 05:00:11 -04001528EXPORT_SYMBOL(lookup_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529
1530/**
Al Viro30c40d22008-02-22 19:50:45 -05001531 * open_bdev_exclusive - open a block device by name and set it up for use
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 *
1533 * @path: special file representing the block device
Al Viro30c40d22008-02-22 19:50:45 -05001534 * @mode: FMODE_... combination to pass be used
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 * @holder: owner for exclusion
1536 *
1537 * Open the blockdevice described by the special file at @path, claim it
1538 * for the @holder.
1539 */
Al Viro30c40d22008-02-22 19:50:45 -05001540struct block_device *open_bdev_exclusive(const char *path, fmode_t mode, void *holder)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541{
1542 struct block_device *bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 int error = 0;
1544
1545 bdev = lookup_bdev(path);
1546 if (IS_ERR(bdev))
1547 return bdev;
1548
Al Viro572c4892007-10-08 13:24:05 -04001549 error = blkdev_get(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 if (error)
1551 return ERR_PTR(error);
1552 error = -EACCES;
Al Viro30c40d22008-02-22 19:50:45 -05001553 if ((mode & FMODE_WRITE) && bdev_read_only(bdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554 goto blkdev_put;
1555 error = bd_claim(bdev, holder);
1556 if (error)
1557 goto blkdev_put;
1558
1559 return bdev;
1560
1561blkdev_put:
Al Viro9a1c3542008-02-22 20:40:24 -05001562 blkdev_put(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 return ERR_PTR(error);
1564}
1565
Al Viro30c40d22008-02-22 19:50:45 -05001566EXPORT_SYMBOL(open_bdev_exclusive);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567
1568/**
Al Viro30c40d22008-02-22 19:50:45 -05001569 * close_bdev_exclusive - close a blockdevice opened by open_bdev_exclusive()
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 *
1571 * @bdev: blockdevice to close
Al Viro30c40d22008-02-22 19:50:45 -05001572 * @mode: mode, must match that used to open.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 *
Al Viro30c40d22008-02-22 19:50:45 -05001574 * This is the counterpart to open_bdev_exclusive().
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575 */
Al Viro30c40d22008-02-22 19:50:45 -05001576void close_bdev_exclusive(struct block_device *bdev, fmode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577{
1578 bd_release(bdev);
Al Viro30c40d22008-02-22 19:50:45 -05001579 blkdev_put(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580}
1581
Al Viro30c40d22008-02-22 19:50:45 -05001582EXPORT_SYMBOL(close_bdev_exclusive);
David Howellsb71e8a42006-08-29 19:06:11 +01001583
1584int __invalidate_device(struct block_device *bdev)
1585{
1586 struct super_block *sb = get_super(bdev);
1587 int res = 0;
1588
1589 if (sb) {
1590 /*
1591 * no need to lock the super, get_super holds the
1592 * read mutex so the filesystem cannot go away
1593 * under us (->put_super runs with the write lock
1594 * hold).
1595 */
1596 shrink_dcache_sb(sb);
1597 res = invalidate_inodes(sb);
1598 drop_super(sb);
1599 }
Peter Zijlstraf98393a2007-05-06 14:49:54 -07001600 invalidate_bdev(bdev);
David Howellsb71e8a42006-08-29 19:06:11 +01001601 return res;
1602}
1603EXPORT_SYMBOL(__invalidate_device);