blob: 73d6a735b8f311cc62fe5b22c626a7a4baa9c452 [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 *
Nick Piggin585d3bc2009-02-25 10:44:19 +0100219 * If a superblock is found on this device, we take the s_umount semaphore
220 * on it to make sure nobody unmounts until the snapshot creation is done.
221 * The reference counter (bd_fsfreeze_count) guarantees that only the last
222 * unfreeze process can unfreeze the frozen filesystem actually when multiple
223 * freeze requests arrive simultaneously. It counts up in freeze_bdev() and
224 * count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze
225 * actually.
226 */
227struct super_block *freeze_bdev(struct block_device *bdev)
228{
229 struct super_block *sb;
230 int error = 0;
231
232 mutex_lock(&bdev->bd_fsfreeze_mutex);
Christoph Hellwig45042302009-08-03 23:28:35 +0200233 if (++bdev->bd_fsfreeze_count > 1) {
234 /*
235 * We don't even need to grab a reference - the first call
236 * to freeze_bdev grab an active reference and only the last
237 * thaw_bdev drops it.
238 */
Nick Piggin585d3bc2009-02-25 10:44:19 +0100239 sb = get_super(bdev);
Christoph Hellwig45042302009-08-03 23:28:35 +0200240 drop_super(sb);
Nick Piggin585d3bc2009-02-25 10:44:19 +0100241 mutex_unlock(&bdev->bd_fsfreeze_mutex);
242 return sb;
243 }
Nick Piggin585d3bc2009-02-25 10:44:19 +0100244
Christoph Hellwig45042302009-08-03 23:28:35 +0200245 sb = get_active_super(bdev);
246 if (!sb)
247 goto out;
248 if (sb->s_flags & MS_RDONLY) {
249 deactivate_locked_super(sb);
250 mutex_unlock(&bdev->bd_fsfreeze_mutex);
251 return sb;
Nick Piggin585d3bc2009-02-25 10:44:19 +0100252 }
253
Christoph Hellwig45042302009-08-03 23:28:35 +0200254 sb->s_frozen = SB_FREEZE_WRITE;
255 smp_wmb();
256
257 sync_filesystem(sb);
258
259 sb->s_frozen = SB_FREEZE_TRANS;
260 smp_wmb();
261
262 sync_blockdev(sb->s_bdev);
263
264 if (sb->s_op->freeze_fs) {
265 error = sb->s_op->freeze_fs(sb);
266 if (error) {
267 printk(KERN_ERR
268 "VFS:Filesystem freeze failed\n");
269 sb->s_frozen = SB_UNFROZEN;
270 deactivate_locked_super(sb);
271 bdev->bd_fsfreeze_count--;
272 mutex_unlock(&bdev->bd_fsfreeze_mutex);
273 return ERR_PTR(error);
274 }
275 }
276 up_write(&sb->s_umount);
277
278 out:
Nick Piggin585d3bc2009-02-25 10:44:19 +0100279 sync_blockdev(bdev);
280 mutex_unlock(&bdev->bd_fsfreeze_mutex);
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200281 return sb; /* thaw_bdev releases s->s_umount */
Nick Piggin585d3bc2009-02-25 10:44:19 +0100282}
283EXPORT_SYMBOL(freeze_bdev);
284
285/**
286 * thaw_bdev -- unlock filesystem
287 * @bdev: blockdevice to unlock
288 * @sb: associated superblock
289 *
290 * Unlocks the filesystem and marks it writeable again after freeze_bdev().
291 */
292int thaw_bdev(struct block_device *bdev, struct super_block *sb)
293{
Christoph Hellwig45042302009-08-03 23:28:35 +0200294 int error = -EINVAL;
Nick Piggin585d3bc2009-02-25 10:44:19 +0100295
296 mutex_lock(&bdev->bd_fsfreeze_mutex);
Christoph Hellwig45042302009-08-03 23:28:35 +0200297 if (!bdev->bd_fsfreeze_count)
298 goto out_unlock;
Nick Piggin585d3bc2009-02-25 10:44:19 +0100299
Christoph Hellwig45042302009-08-03 23:28:35 +0200300 error = 0;
301 if (--bdev->bd_fsfreeze_count > 0)
302 goto out_unlock;
Nick Piggin585d3bc2009-02-25 10:44:19 +0100303
Christoph Hellwig45042302009-08-03 23:28:35 +0200304 if (!sb)
305 goto out_unlock;
306
307 BUG_ON(sb->s_bdev != bdev);
308 down_write(&sb->s_umount);
309 if (sb->s_flags & MS_RDONLY)
310 goto out_deactivate;
311
312 if (sb->s_op->unfreeze_fs) {
313 error = sb->s_op->unfreeze_fs(sb);
314 if (error) {
315 printk(KERN_ERR
316 "VFS:Filesystem thaw failed\n");
317 sb->s_frozen = SB_FREEZE_TRANS;
318 bdev->bd_fsfreeze_count++;
319 mutex_unlock(&bdev->bd_fsfreeze_mutex);
320 return error;
Nick Piggin585d3bc2009-02-25 10:44:19 +0100321 }
Nick Piggin585d3bc2009-02-25 10:44:19 +0100322 }
323
Christoph Hellwig45042302009-08-03 23:28:35 +0200324 sb->s_frozen = SB_UNFROZEN;
325 smp_wmb();
326 wake_up(&sb->s_wait_unfrozen);
327
328out_deactivate:
329 if (sb)
330 deactivate_locked_super(sb);
331out_unlock:
Nick Piggin585d3bc2009-02-25 10:44:19 +0100332 mutex_unlock(&bdev->bd_fsfreeze_mutex);
333 return 0;
334}
335EXPORT_SYMBOL(thaw_bdev);
336
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337static int blkdev_writepage(struct page *page, struct writeback_control *wbc)
338{
339 return block_write_full_page(page, blkdev_get_block, wbc);
340}
341
342static int blkdev_readpage(struct file * file, struct page * page)
343{
344 return block_read_full_page(page, blkdev_get_block);
345}
346
Nick Piggin6272b5a2007-10-16 01:25:04 -0700347static int blkdev_write_begin(struct file *file, struct address_space *mapping,
348 loff_t pos, unsigned len, unsigned flags,
349 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350{
Nick Piggin6272b5a2007-10-16 01:25:04 -0700351 *pagep = NULL;
352 return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
353 blkdev_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354}
355
Nick Piggin6272b5a2007-10-16 01:25:04 -0700356static int blkdev_write_end(struct file *file, struct address_space *mapping,
357 loff_t pos, unsigned len, unsigned copied,
358 struct page *page, void *fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359{
Nick Piggin6272b5a2007-10-16 01:25:04 -0700360 int ret;
361 ret = block_write_end(file, mapping, pos, len, copied, page, fsdata);
362
363 unlock_page(page);
364 page_cache_release(page);
365
366 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367}
368
369/*
370 * private llseek:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800371 * for a block special file file->f_path.dentry->d_inode->i_size is zero
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 * so we compute the size by hand (just as in block_read/write above)
373 */
374static loff_t block_llseek(struct file *file, loff_t offset, int origin)
375{
376 struct inode *bd_inode = file->f_mapping->host;
377 loff_t size;
378 loff_t retval;
379
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800380 mutex_lock(&bd_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 size = i_size_read(bd_inode);
382
383 switch (origin) {
384 case 2:
385 offset += size;
386 break;
387 case 1:
388 offset += file->f_pos;
389 }
390 retval = -EINVAL;
391 if (offset >= 0 && offset <= size) {
392 if (offset != file->f_pos) {
393 file->f_pos = offset;
394 }
395 retval = offset;
396 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800397 mutex_unlock(&bd_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 return retval;
399}
400
401/*
402 * Filp is never NULL; the only case when ->fsync() is called with
403 * NULL first argument is nfsd_sync_dir() and that's not a directory.
404 */
405
406static int block_fsync(struct file *filp, struct dentry *dentry, int datasync)
407{
Christoph Hellwigab0a9732009-10-29 14:14:04 +0100408 struct block_device *bdev = I_BDEV(filp->f_mapping->host);
409 int error;
410
411 error = sync_blockdev(bdev);
412 if (error)
413 return error;
414
415 error = blkdev_issue_flush(bdev, NULL);
416 if (error == -EOPNOTSUPP)
417 error = 0;
418 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419}
420
421/*
422 * pseudo-fs
423 */
424
425static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock);
Christoph Lametere18b8902006-12-06 20:33:20 -0800426static struct kmem_cache * bdev_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427
428static struct inode *bdev_alloc_inode(struct super_block *sb)
429{
Christoph Lametere94b1762006-12-06 20:33:17 -0800430 struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 if (!ei)
432 return NULL;
433 return &ei->vfs_inode;
434}
435
436static void bdev_destroy_inode(struct inode *inode)
437{
438 struct bdev_inode *bdi = BDEV_I(inode);
439
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 kmem_cache_free(bdev_cachep, bdi);
441}
442
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700443static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444{
445 struct bdev_inode *ei = (struct bdev_inode *) foo;
446 struct block_device *bdev = &ei->bdev;
447
Christoph Lametera35afb82007-05-16 22:10:57 -0700448 memset(bdev, 0, sizeof(*bdev));
449 mutex_init(&bdev->bd_mutex);
Christoph Lametera35afb82007-05-16 22:10:57 -0700450 INIT_LIST_HEAD(&bdev->bd_inodes);
451 INIT_LIST_HEAD(&bdev->bd_list);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800452#ifdef CONFIG_SYSFS
Christoph Lametera35afb82007-05-16 22:10:57 -0700453 INIT_LIST_HEAD(&bdev->bd_holder_list);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800454#endif
Christoph Lametera35afb82007-05-16 22:10:57 -0700455 inode_init_once(&ei->vfs_inode);
Takashi Satofcccf502009-01-09 16:40:59 -0800456 /* Initialize mutex for freeze. */
457 mutex_init(&bdev->bd_fsfreeze_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458}
459
460static inline void __bd_forget(struct inode *inode)
461{
462 list_del_init(&inode->i_devices);
463 inode->i_bdev = NULL;
464 inode->i_mapping = &inode->i_data;
465}
466
467static void bdev_clear_inode(struct inode *inode)
468{
469 struct block_device *bdev = &BDEV_I(inode)->bdev;
470 struct list_head *p;
471 spin_lock(&bdev_lock);
472 while ( (p = bdev->bd_inodes.next) != &bdev->bd_inodes ) {
473 __bd_forget(list_entry(p, struct inode, i_devices));
474 }
475 list_del_init(&bdev->bd_list);
476 spin_unlock(&bdev_lock);
477}
478
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800479static const struct super_operations bdev_sops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 .statfs = simple_statfs,
481 .alloc_inode = bdev_alloc_inode,
482 .destroy_inode = bdev_destroy_inode,
483 .drop_inode = generic_delete_inode,
484 .clear_inode = bdev_clear_inode,
485};
486
David Howells454e2392006-06-23 02:02:57 -0700487static int bd_get_sb(struct file_system_type *fs_type,
488 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
David Howells454e2392006-06-23 02:02:57 -0700490 return get_sb_pseudo(fs_type, "bdev:", &bdev_sops, 0x62646576, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491}
492
493static struct file_system_type bd_type = {
494 .name = "bdev",
495 .get_sb = bd_get_sb,
496 .kill_sb = kill_anon_super,
497};
498
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800499struct super_block *blockdev_superblock __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500
501void __init bdev_cache_init(void)
502{
503 int err;
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800504 struct vfsmount *bd_mnt;
505
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode),
Paul Jacksonfffb60f2006-03-24 03:16:06 -0800507 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
508 SLAB_MEM_SPREAD|SLAB_PANIC),
Paul Mundt20c2df82007-07-20 10:11:58 +0900509 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 err = register_filesystem(&bd_type);
511 if (err)
512 panic("Cannot register bdev pseudo-fs");
513 bd_mnt = kern_mount(&bd_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 if (IS_ERR(bd_mnt))
515 panic("Cannot create bdev pseudo-fs");
Catalin Marinas2e1483c2009-06-11 13:24:13 +0100516 /*
517 * This vfsmount structure is only used to obtain the
518 * blockdev_superblock, so tell kmemleak not to report it.
519 */
520 kmemleak_not_leak(bd_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */
522}
523
524/*
525 * Most likely _very_ bad one - but then it's hardly critical for small
526 * /dev and can be fixed when somebody will need really large one.
527 * Keep in mind that it will be fed through icache hash function too.
528 */
529static inline unsigned long hash(dev_t dev)
530{
531 return MAJOR(dev)+MINOR(dev);
532}
533
534static int bdev_test(struct inode *inode, void *data)
535{
536 return BDEV_I(inode)->bdev.bd_dev == *(dev_t *)data;
537}
538
539static int bdev_set(struct inode *inode, void *data)
540{
541 BDEV_I(inode)->bdev.bd_dev = *(dev_t *)data;
542 return 0;
543}
544
545static LIST_HEAD(all_bdevs);
546
547struct block_device *bdget(dev_t dev)
548{
549 struct block_device *bdev;
550 struct inode *inode;
551
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800552 inode = iget5_locked(blockdev_superblock, hash(dev),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 bdev_test, bdev_set, &dev);
554
555 if (!inode)
556 return NULL;
557
558 bdev = &BDEV_I(inode)->bdev;
559
560 if (inode->i_state & I_NEW) {
561 bdev->bd_contains = NULL;
562 bdev->bd_inode = inode;
563 bdev->bd_block_size = (1 << inode->i_blkbits);
564 bdev->bd_part_count = 0;
565 bdev->bd_invalidated = 0;
566 inode->i_mode = S_IFBLK;
567 inode->i_rdev = dev;
568 inode->i_bdev = bdev;
569 inode->i_data.a_ops = &def_blk_aops;
570 mapping_set_gfp_mask(&inode->i_data, GFP_USER);
571 inode->i_data.backing_dev_info = &default_backing_dev_info;
572 spin_lock(&bdev_lock);
573 list_add(&bdev->bd_list, &all_bdevs);
574 spin_unlock(&bdev_lock);
575 unlock_new_inode(inode);
576 }
577 return bdev;
578}
579
580EXPORT_SYMBOL(bdget);
581
Alan Jenkinsdddac6a2009-07-29 21:07:55 +0200582/**
583 * bdgrab -- Grab a reference to an already referenced block device
584 * @bdev: Block device to grab a reference to.
585 */
586struct block_device *bdgrab(struct block_device *bdev)
587{
588 atomic_inc(&bdev->bd_inode->i_count);
589 return bdev;
590}
591
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592long nr_blockdev_pages(void)
593{
Matthias Kaehlcke203a2932007-07-15 23:40:23 -0700594 struct block_device *bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 long ret = 0;
596 spin_lock(&bdev_lock);
Matthias Kaehlcke203a2932007-07-15 23:40:23 -0700597 list_for_each_entry(bdev, &all_bdevs, bd_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 ret += bdev->bd_inode->i_mapping->nrpages;
599 }
600 spin_unlock(&bdev_lock);
601 return ret;
602}
603
604void bdput(struct block_device *bdev)
605{
606 iput(bdev->bd_inode);
607}
608
609EXPORT_SYMBOL(bdput);
610
611static struct block_device *bd_acquire(struct inode *inode)
612{
613 struct block_device *bdev;
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700614
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 spin_lock(&bdev_lock);
616 bdev = inode->i_bdev;
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700617 if (bdev) {
618 atomic_inc(&bdev->bd_inode->i_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 spin_unlock(&bdev_lock);
620 return bdev;
621 }
622 spin_unlock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700623
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 bdev = bdget(inode->i_rdev);
625 if (bdev) {
626 spin_lock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700627 if (!inode->i_bdev) {
628 /*
629 * We take an additional bd_inode->i_count for inode,
630 * and it's released in clear_inode() of inode.
631 * So, we can access it via ->i_mapping always
632 * without igrab().
633 */
634 atomic_inc(&bdev->bd_inode->i_count);
635 inode->i_bdev = bdev;
636 inode->i_mapping = bdev->bd_inode->i_mapping;
637 list_add(&inode->i_devices, &bdev->bd_inodes);
638 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 spin_unlock(&bdev_lock);
640 }
641 return bdev;
642}
643
644/* Call when you free inode */
645
646void bd_forget(struct inode *inode)
647{
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700648 struct block_device *bdev = NULL;
649
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 spin_lock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700651 if (inode->i_bdev) {
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800652 if (!sb_is_blkdev_sb(inode->i_sb))
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700653 bdev = inode->i_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 __bd_forget(inode);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700655 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 spin_unlock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700657
658 if (bdev)
659 iput(bdev->bd_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660}
661
662int bd_claim(struct block_device *bdev, void *holder)
663{
664 int res;
665 spin_lock(&bdev_lock);
666
667 /* first decide result */
668 if (bdev->bd_holder == holder)
669 res = 0; /* already a holder */
670 else if (bdev->bd_holder != NULL)
671 res = -EBUSY; /* held by someone else */
672 else if (bdev->bd_contains == bdev)
673 res = 0; /* is a whole device which isn't held */
674
675 else if (bdev->bd_contains->bd_holder == bd_claim)
676 res = 0; /* is a partition of a device that is being partitioned */
677 else if (bdev->bd_contains->bd_holder != NULL)
678 res = -EBUSY; /* is a partition of a held device */
679 else
680 res = 0; /* is a partition of an un-held device */
681
682 /* now impose change */
683 if (res==0) {
684 /* note that for a whole device bd_holders
685 * will be incremented twice, and bd_holder will
686 * be set to bd_claim before being set to holder
687 */
688 bdev->bd_contains->bd_holders ++;
689 bdev->bd_contains->bd_holder = bd_claim;
690 bdev->bd_holders++;
691 bdev->bd_holder = holder;
692 }
693 spin_unlock(&bdev_lock);
694 return res;
695}
696
697EXPORT_SYMBOL(bd_claim);
698
699void bd_release(struct block_device *bdev)
700{
701 spin_lock(&bdev_lock);
702 if (!--bdev->bd_contains->bd_holders)
703 bdev->bd_contains->bd_holder = NULL;
704 if (!--bdev->bd_holders)
705 bdev->bd_holder = NULL;
706 spin_unlock(&bdev_lock);
707}
708
709EXPORT_SYMBOL(bd_release);
710
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800711#ifdef CONFIG_SYSFS
712/*
713 * Functions for bd_claim_by_kobject / bd_release_from_kobject
714 *
715 * If a kobject is passed to bd_claim_by_kobject()
716 * and the kobject has a parent directory,
717 * following symlinks are created:
718 * o from the kobject to the claimed bdev
719 * o from "holders" directory of the bdev to the parent of the kobject
720 * bd_release_from_kobject() removes these symlinks.
721 *
722 * Example:
723 * If /dev/dm-0 maps to /dev/sda, kobject corresponding to
724 * /sys/block/dm-0/slaves is passed to bd_claim_by_kobject(), then:
725 * /sys/block/dm-0/slaves/sda --> /sys/block/sda
726 * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0
727 */
728
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700729static int add_symlink(struct kobject *from, struct kobject *to)
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800730{
731 if (!from || !to)
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700732 return 0;
733 return sysfs_create_link(from, to, kobject_name(to));
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800734}
735
736static void del_symlink(struct kobject *from, struct kobject *to)
737{
738 if (!from || !to)
739 return;
740 sysfs_remove_link(from, kobject_name(to));
741}
742
743/*
744 * 'struct bd_holder' contains pointers to kobjects symlinked by
745 * bd_claim_by_kobject.
746 * It's connected to bd_holder_list which is protected by bdev->bd_sem.
747 */
748struct bd_holder {
749 struct list_head list; /* chain of holders of the bdev */
750 int count; /* references from the holder */
751 struct kobject *sdir; /* holder object, e.g. "/block/dm-0/slaves" */
752 struct kobject *hdev; /* e.g. "/block/dm-0" */
753 struct kobject *hdir; /* e.g. "/block/sda/holders" */
754 struct kobject *sdev; /* e.g. "/block/sda" */
755};
756
757/*
758 * Get references of related kobjects at once.
759 * Returns 1 on success. 0 on failure.
760 *
761 * Should call bd_holder_release_dirs() after successful use.
762 */
763static int bd_holder_grab_dirs(struct block_device *bdev,
764 struct bd_holder *bo)
765{
766 if (!bdev || !bo)
767 return 0;
768
769 bo->sdir = kobject_get(bo->sdir);
770 if (!bo->sdir)
771 return 0;
772
773 bo->hdev = kobject_get(bo->sdir->parent);
774 if (!bo->hdev)
775 goto fail_put_sdir;
776
Tejun Heo0762b8b2008-08-25 19:56:12 +0900777 bo->sdev = kobject_get(&part_to_dev(bdev->bd_part)->kobj);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800778 if (!bo->sdev)
779 goto fail_put_hdev;
780
Tejun Heo4c465012008-08-25 19:56:11 +0900781 bo->hdir = kobject_get(bdev->bd_part->holder_dir);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800782 if (!bo->hdir)
783 goto fail_put_sdev;
784
785 return 1;
786
787fail_put_sdev:
788 kobject_put(bo->sdev);
789fail_put_hdev:
790 kobject_put(bo->hdev);
791fail_put_sdir:
792 kobject_put(bo->sdir);
793
794 return 0;
795}
796
797/* Put references of related kobjects at once. */
798static void bd_holder_release_dirs(struct bd_holder *bo)
799{
800 kobject_put(bo->hdir);
801 kobject_put(bo->sdev);
802 kobject_put(bo->hdev);
803 kobject_put(bo->sdir);
804}
805
806static struct bd_holder *alloc_bd_holder(struct kobject *kobj)
807{
808 struct bd_holder *bo;
809
810 bo = kzalloc(sizeof(*bo), GFP_KERNEL);
811 if (!bo)
812 return NULL;
813
814 bo->count = 1;
815 bo->sdir = kobj;
816
817 return bo;
818}
819
820static void free_bd_holder(struct bd_holder *bo)
821{
822 kfree(bo);
823}
824
825/**
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500826 * find_bd_holder - find matching struct bd_holder from the block device
827 *
828 * @bdev: struct block device to be searched
829 * @bo: target struct bd_holder
830 *
831 * Returns matching entry with @bo in @bdev->bd_holder_list.
832 * If found, increment the reference count and return the pointer.
833 * If not found, returns NULL.
834 */
Andrew Morton36a561d2006-10-30 22:07:03 -0800835static struct bd_holder *find_bd_holder(struct block_device *bdev,
836 struct bd_holder *bo)
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500837{
838 struct bd_holder *tmp;
839
840 list_for_each_entry(tmp, &bdev->bd_holder_list, list)
841 if (tmp->sdir == bo->sdir) {
842 tmp->count++;
843 return tmp;
844 }
845
846 return NULL;
847}
848
849/**
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800850 * add_bd_holder - create sysfs symlinks for bd_claim() relationship
851 *
852 * @bdev: block device to be bd_claimed
853 * @bo: preallocated and initialized by alloc_bd_holder()
854 *
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500855 * Add @bo to @bdev->bd_holder_list, create symlinks.
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800856 *
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500857 * Returns 0 if symlinks are created.
858 * Returns -ve if something fails.
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800859 */
860static int add_bd_holder(struct block_device *bdev, struct bd_holder *bo)
861{
Johannes Weiner4e916722007-07-15 23:41:25 -0700862 int err;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800863
864 if (!bo)
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700865 return -EINVAL;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800866
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800867 if (!bd_holder_grab_dirs(bdev, bo))
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700868 return -EBUSY;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800869
Johannes Weiner4e916722007-07-15 23:41:25 -0700870 err = add_symlink(bo->sdir, bo->sdev);
871 if (err)
872 return err;
873
874 err = add_symlink(bo->hdir, bo->hdev);
875 if (err) {
876 del_symlink(bo->sdir, bo->sdev);
877 return err;
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700878 }
Johannes Weiner4e916722007-07-15 23:41:25 -0700879
880 list_add_tail(&bo->list, &bdev->bd_holder_list);
881 return 0;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800882}
883
884/**
885 * del_bd_holder - delete sysfs symlinks for bd_claim() relationship
886 *
887 * @bdev: block device to be bd_claimed
888 * @kobj: holder's kobject
889 *
890 * If there is matching entry with @kobj in @bdev->bd_holder_list
891 * and no other bd_claim() from the same kobject,
892 * remove the struct bd_holder from the list, delete symlinks for it.
893 *
894 * Returns a pointer to the struct bd_holder when it's removed from the list
895 * and ready to be freed.
896 * Returns NULL if matching claim isn't found or there is other bd_claim()
897 * by the same kobject.
898 */
899static struct bd_holder *del_bd_holder(struct block_device *bdev,
900 struct kobject *kobj)
901{
902 struct bd_holder *bo;
903
904 list_for_each_entry(bo, &bdev->bd_holder_list, list) {
905 if (bo->sdir == kobj) {
906 bo->count--;
907 BUG_ON(bo->count < 0);
908 if (!bo->count) {
909 list_del(&bo->list);
910 del_symlink(bo->sdir, bo->sdev);
911 del_symlink(bo->hdir, bo->hdev);
912 bd_holder_release_dirs(bo);
913 return bo;
914 }
915 break;
916 }
917 }
918
919 return NULL;
920}
921
922/**
923 * bd_claim_by_kobject - bd_claim() with additional kobject signature
924 *
925 * @bdev: block device to be claimed
926 * @holder: holder's signature
927 * @kobj: holder's kobject
928 *
929 * Do bd_claim() and if it succeeds, create sysfs symlinks between
930 * the bdev and the holder's kobject.
931 * Use bd_release_from_kobject() when relesing the claimed bdev.
932 *
933 * Returns 0 on success. (same as bd_claim())
934 * Returns errno on failure.
935 */
936static int bd_claim_by_kobject(struct block_device *bdev, void *holder,
937 struct kobject *kobj)
938{
Johannes Weiner4e916722007-07-15 23:41:25 -0700939 int err;
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500940 struct bd_holder *bo, *found;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800941
942 if (!kobj)
943 return -EINVAL;
944
945 bo = alloc_bd_holder(kobj);
946 if (!bo)
947 return -ENOMEM;
948
Peter Zijlstra2e7b6512006-12-08 02:36:13 -0800949 mutex_lock(&bdev->bd_mutex);
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500950
Johannes Weiner4e916722007-07-15 23:41:25 -0700951 err = bd_claim(bdev, holder);
952 if (err)
Andrew Morton4210df22007-07-15 23:41:28 -0700953 goto fail;
Johannes Weiner4e916722007-07-15 23:41:25 -0700954
955 found = find_bd_holder(bdev, bo);
956 if (found)
Andrew Morton4210df22007-07-15 23:41:28 -0700957 goto fail;
Johannes Weiner4e916722007-07-15 23:41:25 -0700958
959 err = add_bd_holder(bdev, bo);
960 if (err)
961 bd_release(bdev);
Andrew Morton4210df22007-07-15 23:41:28 -0700962 else
963 bo = NULL;
964fail:
Jun'ichi Nomurab4cf1b72006-03-27 01:18:00 -0800965 mutex_unlock(&bdev->bd_mutex);
Andrew Morton4210df22007-07-15 23:41:28 -0700966 free_bd_holder(bo);
Johannes Weiner4e916722007-07-15 23:41:25 -0700967 return err;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800968}
969
970/**
971 * bd_release_from_kobject - bd_release() with additional kobject signature
972 *
973 * @bdev: block device to be released
974 * @kobj: holder's kobject
975 *
976 * Do bd_release() and remove sysfs symlinks created by bd_claim_by_kobject().
977 */
978static void bd_release_from_kobject(struct block_device *bdev,
979 struct kobject *kobj)
980{
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800981 if (!kobj)
982 return;
983
Peter Zijlstra2e7b6512006-12-08 02:36:13 -0800984 mutex_lock(&bdev->bd_mutex);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800985 bd_release(bdev);
Andrew Morton4210df22007-07-15 23:41:28 -0700986 free_bd_holder(del_bd_holder(bdev, kobj));
Jun'ichi Nomurab4cf1b72006-03-27 01:18:00 -0800987 mutex_unlock(&bdev->bd_mutex);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800988}
989
990/**
991 * bd_claim_by_disk - wrapper function for bd_claim_by_kobject()
992 *
993 * @bdev: block device to be claimed
994 * @holder: holder's signature
995 * @disk: holder's gendisk
996 *
997 * Call bd_claim_by_kobject() with getting @disk->slave_dir.
998 */
999int bd_claim_by_disk(struct block_device *bdev, void *holder,
1000 struct gendisk *disk)
1001{
1002 return bd_claim_by_kobject(bdev, holder, kobject_get(disk->slave_dir));
1003}
1004EXPORT_SYMBOL_GPL(bd_claim_by_disk);
1005
1006/**
1007 * bd_release_from_disk - wrapper function for bd_release_from_kobject()
1008 *
1009 * @bdev: block device to be claimed
1010 * @disk: holder's gendisk
1011 *
1012 * Call bd_release_from_kobject() and put @disk->slave_dir.
1013 */
1014void bd_release_from_disk(struct block_device *bdev, struct gendisk *disk)
1015{
1016 bd_release_from_kobject(bdev, disk->slave_dir);
1017 kobject_put(disk->slave_dir);
1018}
1019EXPORT_SYMBOL_GPL(bd_release_from_disk);
1020#endif
1021
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022/*
1023 * Tries to open block device by device number. Use it ONLY if you
1024 * really do not have anything better - i.e. when you are behind a
1025 * truly sucky interface and all you are given is a device number. _Never_
1026 * to be used for internal purposes. If you ever need it - reconsider
1027 * your API.
1028 */
Al Viroaeb5d722008-09-02 15:28:45 -04001029struct block_device *open_by_devnum(dev_t dev, fmode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030{
1031 struct block_device *bdev = bdget(dev);
1032 int err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 if (bdev)
Al Viro572c4892007-10-08 13:24:05 -04001034 err = blkdev_get(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 return err ? ERR_PTR(err) : bdev;
1036}
1037
1038EXPORT_SYMBOL(open_by_devnum);
1039
Andrew Patterson0c002c22008-09-04 14:27:20 -06001040/**
Andrew Patterson56ade442008-09-04 14:27:40 -06001041 * flush_disk - invalidates all buffer-cache entries on a disk
1042 *
1043 * @bdev: struct block device to be flushed
1044 *
1045 * Invalidates all buffer-cache entries on a disk. It should be called
1046 * when a disk has been changed -- either by a media change or online
1047 * resize.
1048 */
1049static void flush_disk(struct block_device *bdev)
1050{
1051 if (__invalidate_device(bdev)) {
1052 char name[BDEVNAME_SIZE] = "";
1053
1054 if (bdev->bd_disk)
1055 disk_name(bdev->bd_disk, 0, name);
1056 printk(KERN_WARNING "VFS: busy inodes on changed media or "
1057 "resized disk %s\n", name);
1058 }
1059
1060 if (!bdev->bd_disk)
1061 return;
1062 if (disk_partitionable(bdev->bd_disk))
1063 bdev->bd_invalidated = 1;
1064}
1065
1066/**
Randy Dunlap57d1b532008-10-09 10:42:38 +02001067 * check_disk_size_change - checks for disk size change and adjusts bdev size.
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001068 * @disk: struct gendisk to check
1069 * @bdev: struct bdev to adjust.
1070 *
1071 * This routine checks to see if the bdev size does not match the disk size
1072 * and adjusts it if it differs.
1073 */
1074void check_disk_size_change(struct gendisk *disk, struct block_device *bdev)
1075{
1076 loff_t disk_size, bdev_size;
1077
1078 disk_size = (loff_t)get_capacity(disk) << 9;
1079 bdev_size = i_size_read(bdev->bd_inode);
1080 if (disk_size != bdev_size) {
1081 char name[BDEVNAME_SIZE];
1082
1083 disk_name(disk, 0, name);
1084 printk(KERN_INFO
1085 "%s: detected capacity change from %lld to %lld\n",
1086 name, bdev_size, disk_size);
1087 i_size_write(bdev->bd_inode, disk_size);
Andrew Patterson608aeef2008-09-04 14:27:45 -06001088 flush_disk(bdev);
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001089 }
1090}
1091EXPORT_SYMBOL(check_disk_size_change);
1092
1093/**
Randy Dunlap57d1b532008-10-09 10:42:38 +02001094 * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back
Andrew Patterson0c002c22008-09-04 14:27:20 -06001095 * @disk: struct gendisk to be revalidated
1096 *
1097 * This routine is a wrapper for lower-level driver's revalidate_disk
1098 * call-backs. It is used to do common pre and post operations needed
1099 * for all revalidate_disk operations.
1100 */
1101int revalidate_disk(struct gendisk *disk)
1102{
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001103 struct block_device *bdev;
Andrew Patterson0c002c22008-09-04 14:27:20 -06001104 int ret = 0;
1105
1106 if (disk->fops->revalidate_disk)
1107 ret = disk->fops->revalidate_disk(disk);
1108
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001109 bdev = bdget_disk(disk, 0);
1110 if (!bdev)
1111 return ret;
1112
1113 mutex_lock(&bdev->bd_mutex);
1114 check_disk_size_change(disk, bdev);
1115 mutex_unlock(&bdev->bd_mutex);
1116 bdput(bdev);
Andrew Patterson0c002c22008-09-04 14:27:20 -06001117 return ret;
1118}
1119EXPORT_SYMBOL(revalidate_disk);
1120
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121/*
1122 * This routine checks whether a removable media has been changed,
1123 * and invalidates all buffer-cache-entries in that case. This
1124 * is a relatively slow routine, so we have to try to minimize using
1125 * it. Thus it is called only upon a 'mount' or 'open'. This
1126 * is the best way of combining speed and utility, I think.
1127 * People changing diskettes in the middle of an operation deserve
1128 * to lose :-)
1129 */
1130int check_disk_change(struct block_device *bdev)
1131{
1132 struct gendisk *disk = bdev->bd_disk;
Alexey Dobriyan83d5cde2009-09-21 17:01:13 -07001133 const struct block_device_operations *bdops = disk->fops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134
1135 if (!bdops->media_changed)
1136 return 0;
1137 if (!bdops->media_changed(bdev->bd_disk))
1138 return 0;
1139
Andrew Patterson56ade442008-09-04 14:27:40 -06001140 flush_disk(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 if (bdops->revalidate_disk)
1142 bdops->revalidate_disk(bdev->bd_disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 return 1;
1144}
1145
1146EXPORT_SYMBOL(check_disk_change);
1147
1148void bd_set_size(struct block_device *bdev, loff_t size)
1149{
Martin K. Petersene1defc42009-05-22 17:17:49 -04001150 unsigned bsize = bdev_logical_block_size(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151
1152 bdev->bd_inode->i_size = size;
1153 while (bsize < PAGE_CACHE_SIZE) {
1154 if (size & bsize)
1155 break;
1156 bsize <<= 1;
1157 }
1158 bdev->bd_block_size = bsize;
1159 bdev->bd_inode->i_blkbits = blksize_bits(bsize);
1160}
1161EXPORT_SYMBOL(bd_set_size);
1162
Al Viro9a1c3542008-02-22 20:40:24 -05001163static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
NeilBrown37be4122006-12-08 02:36:16 -08001164
Peter Zijlstra6d740cd2007-02-20 13:58:18 -08001165/*
1166 * bd_mutex locking:
1167 *
1168 * mutex_lock(part->bd_mutex)
1169 * mutex_lock_nested(whole->bd_mutex, 1)
1170 */
1171
Al Viro572c4892007-10-08 13:24:05 -04001172static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174 struct gendisk *disk;
Pavel Emelyanov7db9cfd2008-06-05 22:46:27 -07001175 int ret;
Tejun Heocf771cb2008-09-03 09:01:09 +02001176 int partno;
Al Virofe6e9c12008-06-23 08:30:55 -04001177 int perm = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178
Al Viro572c4892007-10-08 13:24:05 -04001179 if (mode & FMODE_READ)
Al Virofe6e9c12008-06-23 08:30:55 -04001180 perm |= MAY_READ;
Al Viro572c4892007-10-08 13:24:05 -04001181 if (mode & FMODE_WRITE)
Al Virofe6e9c12008-06-23 08:30:55 -04001182 perm |= MAY_WRITE;
1183 /*
1184 * hooks: /n/, see "layering violations".
1185 */
1186 ret = devcgroup_inode_permission(bdev->bd_inode, perm);
Al Viro82666022008-08-01 05:32:04 -04001187 if (ret != 0) {
1188 bdput(bdev);
Pavel Emelyanov7db9cfd2008-06-05 22:46:27 -07001189 return ret;
Al Viro82666022008-08-01 05:32:04 -04001190 }
Pavel Emelyanov7db9cfd2008-06-05 22:46:27 -07001191
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 lock_kernel();
NeilBrownd3374822009-01-09 08:31:10 +11001193 restart:
Tejun Heo0762b8b2008-08-25 19:56:12 +09001194
Tejun Heo89f97492008-11-05 10:21:06 +01001195 ret = -ENXIO;
Tejun Heocf771cb2008-09-03 09:01:09 +02001196 disk = get_gendisk(bdev->bd_dev, &partno);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001197 if (!disk)
1198 goto out_unlock_kernel;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199
NeilBrown6796bf52006-12-08 02:36:16 -08001200 mutex_lock_nested(&bdev->bd_mutex, for_part);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 if (!bdev->bd_openers) {
1202 bdev->bd_disk = disk;
1203 bdev->bd_contains = bdev;
Tejun Heocf771cb2008-09-03 09:01:09 +02001204 if (!partno) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 struct backing_dev_info *bdi;
Tejun Heo89f97492008-11-05 10:21:06 +01001206
1207 ret = -ENXIO;
1208 bdev->bd_part = disk_get_part(disk, partno);
1209 if (!bdev->bd_part)
1210 goto out_clear;
1211
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 if (disk->fops->open) {
Al Viro572c4892007-10-08 13:24:05 -04001213 ret = disk->fops->open(bdev, mode);
NeilBrownd3374822009-01-09 08:31:10 +11001214 if (ret == -ERESTARTSYS) {
1215 /* Lost a race with 'disk' being
1216 * deleted, try again.
1217 * See md.c
1218 */
1219 disk_put_part(bdev->bd_part);
1220 bdev->bd_part = NULL;
1221 module_put(disk->fops->owner);
1222 put_disk(disk);
1223 bdev->bd_disk = NULL;
1224 mutex_unlock(&bdev->bd_mutex);
1225 goto restart;
1226 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 if (ret)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001228 goto out_clear;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 }
1230 if (!bdev->bd_openers) {
1231 bd_set_size(bdev,(loff_t)get_capacity(disk)<<9);
1232 bdi = blk_get_backing_dev_info(bdev);
1233 if (bdi == NULL)
1234 bdi = &default_backing_dev_info;
1235 bdev->bd_inode->i_data.backing_dev_info = bdi;
1236 }
1237 if (bdev->bd_invalidated)
1238 rescan_partitions(disk, bdev);
1239 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 struct block_device *whole;
1241 whole = bdget_disk(disk, 0);
1242 ret = -ENOMEM;
1243 if (!whole)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001244 goto out_clear;
NeilBrown37be4122006-12-08 02:36:16 -08001245 BUG_ON(for_part);
Al Viro572c4892007-10-08 13:24:05 -04001246 ret = __blkdev_get(whole, mode, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247 if (ret)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001248 goto out_clear;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 bdev->bd_contains = whole;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 bdev->bd_inode->i_data.backing_dev_info =
1251 whole->bd_inode->i_data.backing_dev_info;
Tejun Heo89f97492008-11-05 10:21:06 +01001252 bdev->bd_part = disk_get_part(disk, partno);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001253 if (!(disk->flags & GENHD_FL_UP) ||
Tejun Heo89f97492008-11-05 10:21:06 +01001254 !bdev->bd_part || !bdev->bd_part->nr_sects) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 ret = -ENXIO;
Tejun Heo0762b8b2008-08-25 19:56:12 +09001256 goto out_clear;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 }
Tejun Heo89f97492008-11-05 10:21:06 +01001258 bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259 }
1260 } else {
Tejun Heo0762b8b2008-08-25 19:56:12 +09001261 module_put(disk->fops->owner);
Neil Brown960cc0f2009-10-26 08:59:17 +01001262 put_disk(disk);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001263 disk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264 if (bdev->bd_contains == bdev) {
1265 if (bdev->bd_disk->fops->open) {
Al Viro572c4892007-10-08 13:24:05 -04001266 ret = bdev->bd_disk->fops->open(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 if (ret)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001268 goto out_unlock_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 }
1270 if (bdev->bd_invalidated)
1271 rescan_partitions(bdev->bd_disk, bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272 }
1273 }
1274 bdev->bd_openers++;
NeilBrown37be4122006-12-08 02:36:16 -08001275 if (for_part)
1276 bdev->bd_part_count++;
Arjan van de Venc039e312006-03-23 03:00:28 -08001277 mutex_unlock(&bdev->bd_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 unlock_kernel();
1279 return 0;
1280
Tejun Heo0762b8b2008-08-25 19:56:12 +09001281 out_clear:
Tejun Heo89f97492008-11-05 10:21:06 +01001282 disk_put_part(bdev->bd_part);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 bdev->bd_disk = NULL;
Tejun Heo0762b8b2008-08-25 19:56:12 +09001284 bdev->bd_part = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info;
1286 if (bdev != bdev->bd_contains)
Al Viro572c4892007-10-08 13:24:05 -04001287 __blkdev_put(bdev->bd_contains, mode, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 bdev->bd_contains = NULL;
Tejun Heo0762b8b2008-08-25 19:56:12 +09001289 out_unlock_bdev:
Arjan van de Venc039e312006-03-23 03:00:28 -08001290 mutex_unlock(&bdev->bd_mutex);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001291 out_unlock_kernel:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 unlock_kernel();
Tejun Heo0762b8b2008-08-25 19:56:12 +09001293
Tejun Heo0762b8b2008-08-25 19:56:12 +09001294 if (disk)
1295 module_put(disk->fops->owner);
1296 put_disk(disk);
1297 bdput(bdev);
1298
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299 return ret;
1300}
1301
Al Viro572c4892007-10-08 13:24:05 -04001302int blkdev_get(struct block_device *bdev, fmode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303{
Al Viro572c4892007-10-08 13:24:05 -04001304 return __blkdev_get(bdev, mode, 0);
NeilBrown37be4122006-12-08 02:36:16 -08001305}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306EXPORT_SYMBOL(blkdev_get);
1307
1308static int blkdev_open(struct inode * inode, struct file * filp)
1309{
1310 struct block_device *bdev;
1311 int res;
1312
1313 /*
1314 * Preserve backwards compatibility and allow large file access
1315 * even if userspace doesn't ask for it explicitly. Some mkfs
1316 * binary needs it. We might want to drop this workaround
1317 * during an unstable branch.
1318 */
1319 filp->f_flags |= O_LARGEFILE;
1320
Al Viro572c4892007-10-08 13:24:05 -04001321 if (filp->f_flags & O_NDELAY)
1322 filp->f_mode |= FMODE_NDELAY;
1323 if (filp->f_flags & O_EXCL)
1324 filp->f_mode |= FMODE_EXCL;
1325 if ((filp->f_flags & O_ACCMODE) == 3)
1326 filp->f_mode |= FMODE_WRITE_IOCTL;
1327
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 bdev = bd_acquire(inode);
Pavel Emelianov6a2aae02006-10-28 10:38:33 -07001329 if (bdev == NULL)
1330 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331
Al Viro572c4892007-10-08 13:24:05 -04001332 filp->f_mapping = bdev->bd_inode->i_mapping;
1333
1334 res = blkdev_get(bdev, filp->f_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 if (res)
1336 return res;
1337
Christoph Hellwigebbefc02008-11-05 14:54:41 +01001338 if (filp->f_mode & FMODE_EXCL) {
1339 res = bd_claim(bdev, filp);
1340 if (res)
1341 goto out_blkdev_put;
1342 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343
Christoph Hellwigebbefc02008-11-05 14:54:41 +01001344 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345
Christoph Hellwigebbefc02008-11-05 14:54:41 +01001346 out_blkdev_put:
Al Viro9a1c3542008-02-22 20:40:24 -05001347 blkdev_put(bdev, filp->f_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 return res;
1349}
1350
Al Viro9a1c3542008-02-22 20:40:24 -05001351static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part)
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001352{
1353 int ret = 0;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001354 struct gendisk *disk = bdev->bd_disk;
NeilBrown37be4122006-12-08 02:36:16 -08001355 struct block_device *victim = NULL;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001356
NeilBrown6796bf52006-12-08 02:36:16 -08001357 mutex_lock_nested(&bdev->bd_mutex, for_part);
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001358 lock_kernel();
NeilBrown37be4122006-12-08 02:36:16 -08001359 if (for_part)
1360 bdev->bd_part_count--;
1361
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001362 if (!--bdev->bd_openers) {
1363 sync_blockdev(bdev);
1364 kill_bdev(bdev);
1365 }
1366 if (bdev->bd_contains == bdev) {
1367 if (disk->fops->release)
Al Viro9a1c3542008-02-22 20:40:24 -05001368 ret = disk->fops->release(disk, mode);
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001369 }
1370 if (!bdev->bd_openers) {
1371 struct module *owner = disk->fops->owner;
1372
1373 put_disk(disk);
1374 module_put(owner);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001375 disk_put_part(bdev->bd_part);
1376 bdev->bd_part = NULL;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001377 bdev->bd_disk = NULL;
1378 bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info;
NeilBrown37be4122006-12-08 02:36:16 -08001379 if (bdev != bdev->bd_contains)
1380 victim = bdev->bd_contains;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001381 bdev->bd_contains = NULL;
1382 }
1383 unlock_kernel();
1384 mutex_unlock(&bdev->bd_mutex);
1385 bdput(bdev);
NeilBrown37be4122006-12-08 02:36:16 -08001386 if (victim)
Al Viro9a1c3542008-02-22 20:40:24 -05001387 __blkdev_put(victim, mode, 1);
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001388 return ret;
1389}
1390
Al Viro9a1c3542008-02-22 20:40:24 -05001391int blkdev_put(struct block_device *bdev, fmode_t mode)
NeilBrown37be4122006-12-08 02:36:16 -08001392{
Al Viro9a1c3542008-02-22 20:40:24 -05001393 return __blkdev_put(bdev, mode, 0);
NeilBrown37be4122006-12-08 02:36:16 -08001394}
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001395EXPORT_SYMBOL(blkdev_put);
1396
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397static int blkdev_close(struct inode * inode, struct file * filp)
1398{
1399 struct block_device *bdev = I_BDEV(filp->f_mapping->host);
1400 if (bdev->bd_holder == filp)
1401 bd_release(bdev);
Al Viro9a1c3542008-02-22 20:40:24 -05001402 return blkdev_put(bdev, filp->f_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403}
1404
Arnd Bergmannbb93e3a2005-06-23 00:10:15 -07001405static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406{
Al Viro56b26ad2008-09-19 03:17:36 -04001407 struct block_device *bdev = I_BDEV(file->f_mapping->host);
1408 fmode_t mode = file->f_mode;
Christoph Hellwigfd4ce1a2008-11-05 14:58:42 +01001409
1410 /*
1411 * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have
1412 * to updated it before every ioctl.
1413 */
Al Viro56b26ad2008-09-19 03:17:36 -04001414 if (file->f_flags & O_NDELAY)
Christoph Hellwigfd4ce1a2008-11-05 14:58:42 +01001415 mode |= FMODE_NDELAY;
1416 else
1417 mode &= ~FMODE_NDELAY;
1418
Al Viro56b26ad2008-09-19 03:17:36 -04001419 return blkdev_ioctl(bdev, mode, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420}
1421
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001422/*
Christoph Hellwigeef99382009-08-20 17:43:41 +02001423 * Write data to the block device. Only intended for the block device itself
1424 * and the raw driver which basically is a fake block device.
1425 *
1426 * Does not take i_mutex for the write and thus is not for general purpose
1427 * use.
1428 */
1429ssize_t blkdev_aio_write(struct kiocb *iocb, const struct iovec *iov,
1430 unsigned long nr_segs, loff_t pos)
1431{
1432 struct file *file = iocb->ki_filp;
1433 ssize_t ret;
1434
1435 BUG_ON(iocb->ki_pos != pos);
1436
1437 ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos);
1438 if (ret > 0 || ret == -EIOCBQUEUED) {
1439 ssize_t err;
1440
1441 err = generic_write_sync(file, pos, ret);
1442 if (err < 0 && ret > 0)
1443 ret = err;
1444 }
1445 return ret;
1446}
1447EXPORT_SYMBOL_GPL(blkdev_aio_write);
1448
1449/*
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001450 * Try to release a page associated with block device when the system
1451 * is under memory pressure.
1452 */
1453static int blkdev_releasepage(struct page *page, gfp_t wait)
1454{
1455 struct super_block *super = BDEV_I(page->mapping->host)->bdev.bd_super;
1456
1457 if (super && super->s_op->bdev_try_to_free_page)
1458 return super->s_op->bdev_try_to_free_page(super, page, wait);
1459
1460 return try_to_free_buffers(page);
1461}
1462
Adrian Bunk4c54ac62008-02-18 13:48:31 +01001463static const struct address_space_operations def_blk_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 .readpage = blkdev_readpage,
1465 .writepage = blkdev_writepage,
1466 .sync_page = block_sync_page,
Nick Piggin6272b5a2007-10-16 01:25:04 -07001467 .write_begin = blkdev_write_begin,
1468 .write_end = blkdev_write_end,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 .writepages = generic_writepages,
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001470 .releasepage = blkdev_releasepage,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 .direct_IO = blkdev_direct_IO,
1472};
1473
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -08001474const struct file_operations def_blk_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 .open = blkdev_open,
1476 .release = blkdev_close,
1477 .llseek = block_llseek,
Badari Pulavarty543ade12006-09-30 23:28:48 -07001478 .read = do_sync_read,
1479 .write = do_sync_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 .aio_read = generic_file_aio_read,
Christoph Hellwigeef99382009-08-20 17:43:41 +02001481 .aio_write = blkdev_aio_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 .mmap = generic_file_mmap,
1483 .fsync = block_fsync,
Arnd Bergmannbb93e3a2005-06-23 00:10:15 -07001484 .unlocked_ioctl = block_ioctl,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485#ifdef CONFIG_COMPAT
1486 .compat_ioctl = compat_blkdev_ioctl,
1487#endif
Jens Axboe7f9c51f2006-05-01 19:59:32 +02001488 .splice_read = generic_file_splice_read,
1489 .splice_write = generic_file_splice_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490};
1491
1492int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg)
1493{
1494 int res;
1495 mm_segment_t old_fs = get_fs();
1496 set_fs(KERNEL_DS);
Al Viro56b26ad2008-09-19 03:17:36 -04001497 res = blkdev_ioctl(bdev, 0, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 set_fs(old_fs);
1499 return res;
1500}
1501
1502EXPORT_SYMBOL(ioctl_by_bdev);
1503
1504/**
1505 * lookup_bdev - lookup a struct block_device by name
Randy Dunlap94e29592009-01-06 14:41:15 -08001506 * @pathname: special file representing the block device
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 *
Randy Dunlap57d1b532008-10-09 10:42:38 +02001508 * Get a reference to the blockdevice at @pathname in the current
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 * namespace if possible and return it. Return ERR_PTR(error)
1510 * otherwise.
1511 */
Al Viro421748e2008-08-02 01:04:36 -04001512struct block_device *lookup_bdev(const char *pathname)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513{
1514 struct block_device *bdev;
1515 struct inode *inode;
Al Viro421748e2008-08-02 01:04:36 -04001516 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517 int error;
1518
Al Viro421748e2008-08-02 01:04:36 -04001519 if (!pathname || !*pathname)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 return ERR_PTR(-EINVAL);
1521
Al Viro421748e2008-08-02 01:04:36 -04001522 error = kern_path(pathname, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 if (error)
1524 return ERR_PTR(error);
1525
Al Viro421748e2008-08-02 01:04:36 -04001526 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 error = -ENOTBLK;
1528 if (!S_ISBLK(inode->i_mode))
1529 goto fail;
1530 error = -EACCES;
Al Viro421748e2008-08-02 01:04:36 -04001531 if (path.mnt->mnt_flags & MNT_NODEV)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 goto fail;
1533 error = -ENOMEM;
1534 bdev = bd_acquire(inode);
1535 if (!bdev)
1536 goto fail;
1537out:
Al Viro421748e2008-08-02 01:04:36 -04001538 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 return bdev;
1540fail:
1541 bdev = ERR_PTR(error);
1542 goto out;
1543}
Al Virod5686b42008-08-01 05:00:11 -04001544EXPORT_SYMBOL(lookup_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545
1546/**
Al Viro30c40d22008-02-22 19:50:45 -05001547 * open_bdev_exclusive - open a block device by name and set it up for use
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 *
1549 * @path: special file representing the block device
Al Viro30c40d22008-02-22 19:50:45 -05001550 * @mode: FMODE_... combination to pass be used
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 * @holder: owner for exclusion
1552 *
1553 * Open the blockdevice described by the special file at @path, claim it
1554 * for the @holder.
1555 */
Al Viro30c40d22008-02-22 19:50:45 -05001556struct block_device *open_bdev_exclusive(const char *path, fmode_t mode, void *holder)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557{
1558 struct block_device *bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 int error = 0;
1560
1561 bdev = lookup_bdev(path);
1562 if (IS_ERR(bdev))
1563 return bdev;
1564
Al Viro572c4892007-10-08 13:24:05 -04001565 error = blkdev_get(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566 if (error)
1567 return ERR_PTR(error);
1568 error = -EACCES;
Al Viro30c40d22008-02-22 19:50:45 -05001569 if ((mode & FMODE_WRITE) && bdev_read_only(bdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 goto blkdev_put;
1571 error = bd_claim(bdev, holder);
1572 if (error)
1573 goto blkdev_put;
1574
1575 return bdev;
1576
1577blkdev_put:
Al Viro9a1c3542008-02-22 20:40:24 -05001578 blkdev_put(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 return ERR_PTR(error);
1580}
1581
Al Viro30c40d22008-02-22 19:50:45 -05001582EXPORT_SYMBOL(open_bdev_exclusive);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583
1584/**
Al Viro30c40d22008-02-22 19:50:45 -05001585 * close_bdev_exclusive - close a blockdevice opened by open_bdev_exclusive()
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586 *
1587 * @bdev: blockdevice to close
Al Viro30c40d22008-02-22 19:50:45 -05001588 * @mode: mode, must match that used to open.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 *
Al Viro30c40d22008-02-22 19:50:45 -05001590 * This is the counterpart to open_bdev_exclusive().
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 */
Al Viro30c40d22008-02-22 19:50:45 -05001592void close_bdev_exclusive(struct block_device *bdev, fmode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593{
1594 bd_release(bdev);
Al Viro30c40d22008-02-22 19:50:45 -05001595 blkdev_put(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596}
1597
Al Viro30c40d22008-02-22 19:50:45 -05001598EXPORT_SYMBOL(close_bdev_exclusive);
David Howellsb71e8a42006-08-29 19:06:11 +01001599
1600int __invalidate_device(struct block_device *bdev)
1601{
1602 struct super_block *sb = get_super(bdev);
1603 int res = 0;
1604
1605 if (sb) {
1606 /*
1607 * no need to lock the super, get_super holds the
1608 * read mutex so the filesystem cannot go away
1609 * under us (->put_super runs with the write lock
1610 * hold).
1611 */
1612 shrink_dcache_sb(sb);
1613 res = invalidate_inodes(sb);
1614 drop_super(sb);
1615 }
Peter Zijlstraf98393a2007-05-06 14:49:54 -07001616 invalidate_bdev(bdev);
David Howellsb71e8a42006-08-29 19:06:11 +01001617 return res;
1618}
1619EXPORT_SYMBOL(__invalidate_device);