blob: d11d0289f3d24bf9489acba6059c5d8ada9eda52 [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) {
Jun'ichi Nomura4b06e5b2010-01-29 09:56:22 +0900249 sb->s_frozen = SB_FREEZE_TRANS;
250 up_write(&sb->s_umount);
Christoph Hellwig45042302009-08-03 23:28:35 +0200251 mutex_unlock(&bdev->bd_fsfreeze_mutex);
252 return sb;
Nick Piggin585d3bc2009-02-25 10:44:19 +0100253 }
254
Christoph Hellwig45042302009-08-03 23:28:35 +0200255 sb->s_frozen = SB_FREEZE_WRITE;
256 smp_wmb();
257
258 sync_filesystem(sb);
259
260 sb->s_frozen = SB_FREEZE_TRANS;
261 smp_wmb();
262
263 sync_blockdev(sb->s_bdev);
264
265 if (sb->s_op->freeze_fs) {
266 error = sb->s_op->freeze_fs(sb);
267 if (error) {
268 printk(KERN_ERR
269 "VFS:Filesystem freeze failed\n");
270 sb->s_frozen = SB_UNFROZEN;
271 deactivate_locked_super(sb);
272 bdev->bd_fsfreeze_count--;
273 mutex_unlock(&bdev->bd_fsfreeze_mutex);
274 return ERR_PTR(error);
275 }
276 }
277 up_write(&sb->s_umount);
278
279 out:
Nick Piggin585d3bc2009-02-25 10:44:19 +0100280 sync_blockdev(bdev);
281 mutex_unlock(&bdev->bd_fsfreeze_mutex);
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200282 return sb; /* thaw_bdev releases s->s_umount */
Nick Piggin585d3bc2009-02-25 10:44:19 +0100283}
284EXPORT_SYMBOL(freeze_bdev);
285
286/**
287 * thaw_bdev -- unlock filesystem
288 * @bdev: blockdevice to unlock
289 * @sb: associated superblock
290 *
291 * Unlocks the filesystem and marks it writeable again after freeze_bdev().
292 */
293int thaw_bdev(struct block_device *bdev, struct super_block *sb)
294{
Christoph Hellwig45042302009-08-03 23:28:35 +0200295 int error = -EINVAL;
Nick Piggin585d3bc2009-02-25 10:44:19 +0100296
297 mutex_lock(&bdev->bd_fsfreeze_mutex);
Christoph Hellwig45042302009-08-03 23:28:35 +0200298 if (!bdev->bd_fsfreeze_count)
299 goto out_unlock;
Nick Piggin585d3bc2009-02-25 10:44:19 +0100300
Christoph Hellwig45042302009-08-03 23:28:35 +0200301 error = 0;
302 if (--bdev->bd_fsfreeze_count > 0)
303 goto out_unlock;
Nick Piggin585d3bc2009-02-25 10:44:19 +0100304
Christoph Hellwig45042302009-08-03 23:28:35 +0200305 if (!sb)
306 goto out_unlock;
307
308 BUG_ON(sb->s_bdev != bdev);
309 down_write(&sb->s_umount);
310 if (sb->s_flags & MS_RDONLY)
Jun'ichi Nomura4b06e5b2010-01-29 09:56:22 +0900311 goto out_unfrozen;
Christoph Hellwig45042302009-08-03 23:28:35 +0200312
313 if (sb->s_op->unfreeze_fs) {
314 error = sb->s_op->unfreeze_fs(sb);
315 if (error) {
316 printk(KERN_ERR
317 "VFS:Filesystem thaw failed\n");
318 sb->s_frozen = SB_FREEZE_TRANS;
319 bdev->bd_fsfreeze_count++;
320 mutex_unlock(&bdev->bd_fsfreeze_mutex);
321 return error;
Nick Piggin585d3bc2009-02-25 10:44:19 +0100322 }
Nick Piggin585d3bc2009-02-25 10:44:19 +0100323 }
324
Jun'ichi Nomura4b06e5b2010-01-29 09:56:22 +0900325out_unfrozen:
Christoph Hellwig45042302009-08-03 23:28:35 +0200326 sb->s_frozen = SB_UNFROZEN;
327 smp_wmb();
328 wake_up(&sb->s_wait_unfrozen);
329
Christoph Hellwig45042302009-08-03 23:28:35 +0200330 if (sb)
331 deactivate_locked_super(sb);
332out_unlock:
Nick Piggin585d3bc2009-02-25 10:44:19 +0100333 mutex_unlock(&bdev->bd_fsfreeze_mutex);
334 return 0;
335}
336EXPORT_SYMBOL(thaw_bdev);
337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338static int blkdev_writepage(struct page *page, struct writeback_control *wbc)
339{
340 return block_write_full_page(page, blkdev_get_block, wbc);
341}
342
343static int blkdev_readpage(struct file * file, struct page * page)
344{
345 return block_read_full_page(page, blkdev_get_block);
346}
347
Nick Piggin6272b5a2007-10-16 01:25:04 -0700348static int blkdev_write_begin(struct file *file, struct address_space *mapping,
349 loff_t pos, unsigned len, unsigned flags,
350 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351{
Nick Piggin6272b5a2007-10-16 01:25:04 -0700352 *pagep = NULL;
353 return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
354 blkdev_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355}
356
Nick Piggin6272b5a2007-10-16 01:25:04 -0700357static int blkdev_write_end(struct file *file, struct address_space *mapping,
358 loff_t pos, unsigned len, unsigned copied,
359 struct page *page, void *fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360{
Nick Piggin6272b5a2007-10-16 01:25:04 -0700361 int ret;
362 ret = block_write_end(file, mapping, pos, len, copied, page, fsdata);
363
364 unlock_page(page);
365 page_cache_release(page);
366
367 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368}
369
370/*
371 * private llseek:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800372 * for a block special file file->f_path.dentry->d_inode->i_size is zero
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 * so we compute the size by hand (just as in block_read/write above)
374 */
375static loff_t block_llseek(struct file *file, loff_t offset, int origin)
376{
377 struct inode *bd_inode = file->f_mapping->host;
378 loff_t size;
379 loff_t retval;
380
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800381 mutex_lock(&bd_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 size = i_size_read(bd_inode);
383
384 switch (origin) {
385 case 2:
386 offset += size;
387 break;
388 case 1:
389 offset += file->f_pos;
390 }
391 retval = -EINVAL;
392 if (offset >= 0 && offset <= size) {
393 if (offset != file->f_pos) {
394 file->f_pos = offset;
395 }
396 retval = offset;
397 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800398 mutex_unlock(&bd_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 return retval;
400}
401
402/*
403 * Filp is never NULL; the only case when ->fsync() is called with
404 * NULL first argument is nfsd_sync_dir() and that's not a directory.
405 */
406
407static int block_fsync(struct file *filp, struct dentry *dentry, int datasync)
408{
Christoph Hellwigab0a9732009-10-29 14:14:04 +0100409 struct block_device *bdev = I_BDEV(filp->f_mapping->host);
410 int error;
411
412 error = sync_blockdev(bdev);
413 if (error)
414 return error;
415
416 error = blkdev_issue_flush(bdev, NULL);
417 if (error == -EOPNOTSUPP)
418 error = 0;
419 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420}
421
422/*
423 * pseudo-fs
424 */
425
426static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock);
Christoph Lametere18b8902006-12-06 20:33:20 -0800427static struct kmem_cache * bdev_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428
429static struct inode *bdev_alloc_inode(struct super_block *sb)
430{
Christoph Lametere94b1762006-12-06 20:33:17 -0800431 struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 if (!ei)
433 return NULL;
434 return &ei->vfs_inode;
435}
436
437static void bdev_destroy_inode(struct inode *inode)
438{
439 struct bdev_inode *bdi = BDEV_I(inode);
440
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 kmem_cache_free(bdev_cachep, bdi);
442}
443
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700444static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445{
446 struct bdev_inode *ei = (struct bdev_inode *) foo;
447 struct block_device *bdev = &ei->bdev;
448
Christoph Lametera35afb82007-05-16 22:10:57 -0700449 memset(bdev, 0, sizeof(*bdev));
450 mutex_init(&bdev->bd_mutex);
Christoph Lametera35afb82007-05-16 22:10:57 -0700451 INIT_LIST_HEAD(&bdev->bd_inodes);
452 INIT_LIST_HEAD(&bdev->bd_list);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800453#ifdef CONFIG_SYSFS
Christoph Lametera35afb82007-05-16 22:10:57 -0700454 INIT_LIST_HEAD(&bdev->bd_holder_list);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800455#endif
Christoph Lametera35afb82007-05-16 22:10:57 -0700456 inode_init_once(&ei->vfs_inode);
Takashi Satofcccf502009-01-09 16:40:59 -0800457 /* Initialize mutex for freeze. */
458 mutex_init(&bdev->bd_fsfreeze_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459}
460
461static inline void __bd_forget(struct inode *inode)
462{
463 list_del_init(&inode->i_devices);
464 inode->i_bdev = NULL;
465 inode->i_mapping = &inode->i_data;
466}
467
468static void bdev_clear_inode(struct inode *inode)
469{
470 struct block_device *bdev = &BDEV_I(inode)->bdev;
471 struct list_head *p;
472 spin_lock(&bdev_lock);
473 while ( (p = bdev->bd_inodes.next) != &bdev->bd_inodes ) {
474 __bd_forget(list_entry(p, struct inode, i_devices));
475 }
476 list_del_init(&bdev->bd_list);
477 spin_unlock(&bdev_lock);
478}
479
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800480static const struct super_operations bdev_sops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 .statfs = simple_statfs,
482 .alloc_inode = bdev_alloc_inode,
483 .destroy_inode = bdev_destroy_inode,
484 .drop_inode = generic_delete_inode,
485 .clear_inode = bdev_clear_inode,
486};
487
David Howells454e2392006-06-23 02:02:57 -0700488static int bd_get_sb(struct file_system_type *fs_type,
489 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490{
David Howells454e2392006-06-23 02:02:57 -0700491 return get_sb_pseudo(fs_type, "bdev:", &bdev_sops, 0x62646576, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492}
493
494static struct file_system_type bd_type = {
495 .name = "bdev",
496 .get_sb = bd_get_sb,
497 .kill_sb = kill_anon_super,
498};
499
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800500struct super_block *blockdev_superblock __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501
502void __init bdev_cache_init(void)
503{
504 int err;
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800505 struct vfsmount *bd_mnt;
506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode),
Paul Jacksonfffb60f2006-03-24 03:16:06 -0800508 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
509 SLAB_MEM_SPREAD|SLAB_PANIC),
Paul Mundt20c2df82007-07-20 10:11:58 +0900510 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 err = register_filesystem(&bd_type);
512 if (err)
513 panic("Cannot register bdev pseudo-fs");
514 bd_mnt = kern_mount(&bd_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 if (IS_ERR(bd_mnt))
516 panic("Cannot create bdev pseudo-fs");
Catalin Marinas2e1483c2009-06-11 13:24:13 +0100517 /*
518 * This vfsmount structure is only used to obtain the
519 * blockdev_superblock, so tell kmemleak not to report it.
520 */
521 kmemleak_not_leak(bd_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */
523}
524
525/*
526 * Most likely _very_ bad one - but then it's hardly critical for small
527 * /dev and can be fixed when somebody will need really large one.
528 * Keep in mind that it will be fed through icache hash function too.
529 */
530static inline unsigned long hash(dev_t dev)
531{
532 return MAJOR(dev)+MINOR(dev);
533}
534
535static int bdev_test(struct inode *inode, void *data)
536{
537 return BDEV_I(inode)->bdev.bd_dev == *(dev_t *)data;
538}
539
540static int bdev_set(struct inode *inode, void *data)
541{
542 BDEV_I(inode)->bdev.bd_dev = *(dev_t *)data;
543 return 0;
544}
545
546static LIST_HEAD(all_bdevs);
547
548struct block_device *bdget(dev_t dev)
549{
550 struct block_device *bdev;
551 struct inode *inode;
552
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800553 inode = iget5_locked(blockdev_superblock, hash(dev),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 bdev_test, bdev_set, &dev);
555
556 if (!inode)
557 return NULL;
558
559 bdev = &BDEV_I(inode)->bdev;
560
561 if (inode->i_state & I_NEW) {
562 bdev->bd_contains = NULL;
563 bdev->bd_inode = inode;
564 bdev->bd_block_size = (1 << inode->i_blkbits);
565 bdev->bd_part_count = 0;
566 bdev->bd_invalidated = 0;
567 inode->i_mode = S_IFBLK;
568 inode->i_rdev = dev;
569 inode->i_bdev = bdev;
570 inode->i_data.a_ops = &def_blk_aops;
571 mapping_set_gfp_mask(&inode->i_data, GFP_USER);
572 inode->i_data.backing_dev_info = &default_backing_dev_info;
573 spin_lock(&bdev_lock);
574 list_add(&bdev->bd_list, &all_bdevs);
575 spin_unlock(&bdev_lock);
576 unlock_new_inode(inode);
577 }
578 return bdev;
579}
580
581EXPORT_SYMBOL(bdget);
582
Alan Jenkinsdddac6a2009-07-29 21:07:55 +0200583/**
584 * bdgrab -- Grab a reference to an already referenced block device
585 * @bdev: Block device to grab a reference to.
586 */
587struct block_device *bdgrab(struct block_device *bdev)
588{
589 atomic_inc(&bdev->bd_inode->i_count);
590 return bdev;
591}
592
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593long nr_blockdev_pages(void)
594{
Matthias Kaehlcke203a2932007-07-15 23:40:23 -0700595 struct block_device *bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 long ret = 0;
597 spin_lock(&bdev_lock);
Matthias Kaehlcke203a2932007-07-15 23:40:23 -0700598 list_for_each_entry(bdev, &all_bdevs, bd_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 ret += bdev->bd_inode->i_mapping->nrpages;
600 }
601 spin_unlock(&bdev_lock);
602 return ret;
603}
604
605void bdput(struct block_device *bdev)
606{
607 iput(bdev->bd_inode);
608}
609
610EXPORT_SYMBOL(bdput);
611
612static struct block_device *bd_acquire(struct inode *inode)
613{
614 struct block_device *bdev;
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700615
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 spin_lock(&bdev_lock);
617 bdev = inode->i_bdev;
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700618 if (bdev) {
619 atomic_inc(&bdev->bd_inode->i_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 spin_unlock(&bdev_lock);
621 return bdev;
622 }
623 spin_unlock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700624
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 bdev = bdget(inode->i_rdev);
626 if (bdev) {
627 spin_lock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700628 if (!inode->i_bdev) {
629 /*
630 * We take an additional bd_inode->i_count for inode,
631 * and it's released in clear_inode() of inode.
632 * So, we can access it via ->i_mapping always
633 * without igrab().
634 */
635 atomic_inc(&bdev->bd_inode->i_count);
636 inode->i_bdev = bdev;
637 inode->i_mapping = bdev->bd_inode->i_mapping;
638 list_add(&inode->i_devices, &bdev->bd_inodes);
639 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 spin_unlock(&bdev_lock);
641 }
642 return bdev;
643}
644
645/* Call when you free inode */
646
647void bd_forget(struct inode *inode)
648{
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700649 struct block_device *bdev = NULL;
650
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 spin_lock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700652 if (inode->i_bdev) {
Denis ChengRqc2acf7b2008-12-01 14:34:56 -0800653 if (!sb_is_blkdev_sb(inode->i_sb))
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700654 bdev = inode->i_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 __bd_forget(inode);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700656 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 spin_unlock(&bdev_lock);
OGAWA Hirofumi09d967c2006-06-22 14:47:21 -0700658
659 if (bdev)
660 iput(bdev->bd_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661}
662
663int bd_claim(struct block_device *bdev, void *holder)
664{
665 int res;
666 spin_lock(&bdev_lock);
667
668 /* first decide result */
669 if (bdev->bd_holder == holder)
670 res = 0; /* already a holder */
671 else if (bdev->bd_holder != NULL)
672 res = -EBUSY; /* held by someone else */
673 else if (bdev->bd_contains == bdev)
674 res = 0; /* is a whole device which isn't held */
675
676 else if (bdev->bd_contains->bd_holder == bd_claim)
677 res = 0; /* is a partition of a device that is being partitioned */
678 else if (bdev->bd_contains->bd_holder != NULL)
679 res = -EBUSY; /* is a partition of a held device */
680 else
681 res = 0; /* is a partition of an un-held device */
682
683 /* now impose change */
684 if (res==0) {
685 /* note that for a whole device bd_holders
686 * will be incremented twice, and bd_holder will
687 * be set to bd_claim before being set to holder
688 */
689 bdev->bd_contains->bd_holders ++;
690 bdev->bd_contains->bd_holder = bd_claim;
691 bdev->bd_holders++;
692 bdev->bd_holder = holder;
693 }
694 spin_unlock(&bdev_lock);
695 return res;
696}
697
698EXPORT_SYMBOL(bd_claim);
699
700void bd_release(struct block_device *bdev)
701{
702 spin_lock(&bdev_lock);
703 if (!--bdev->bd_contains->bd_holders)
704 bdev->bd_contains->bd_holder = NULL;
705 if (!--bdev->bd_holders)
706 bdev->bd_holder = NULL;
707 spin_unlock(&bdev_lock);
708}
709
710EXPORT_SYMBOL(bd_release);
711
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800712#ifdef CONFIG_SYSFS
713/*
714 * Functions for bd_claim_by_kobject / bd_release_from_kobject
715 *
716 * If a kobject is passed to bd_claim_by_kobject()
717 * and the kobject has a parent directory,
718 * following symlinks are created:
719 * o from the kobject to the claimed bdev
720 * o from "holders" directory of the bdev to the parent of the kobject
721 * bd_release_from_kobject() removes these symlinks.
722 *
723 * Example:
724 * If /dev/dm-0 maps to /dev/sda, kobject corresponding to
725 * /sys/block/dm-0/slaves is passed to bd_claim_by_kobject(), then:
726 * /sys/block/dm-0/slaves/sda --> /sys/block/sda
727 * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0
728 */
729
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700730static int add_symlink(struct kobject *from, struct kobject *to)
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800731{
732 if (!from || !to)
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700733 return 0;
734 return sysfs_create_link(from, to, kobject_name(to));
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800735}
736
737static void del_symlink(struct kobject *from, struct kobject *to)
738{
739 if (!from || !to)
740 return;
741 sysfs_remove_link(from, kobject_name(to));
742}
743
744/*
745 * 'struct bd_holder' contains pointers to kobjects symlinked by
746 * bd_claim_by_kobject.
747 * It's connected to bd_holder_list which is protected by bdev->bd_sem.
748 */
749struct bd_holder {
750 struct list_head list; /* chain of holders of the bdev */
751 int count; /* references from the holder */
752 struct kobject *sdir; /* holder object, e.g. "/block/dm-0/slaves" */
753 struct kobject *hdev; /* e.g. "/block/dm-0" */
754 struct kobject *hdir; /* e.g. "/block/sda/holders" */
755 struct kobject *sdev; /* e.g. "/block/sda" */
756};
757
758/*
759 * Get references of related kobjects at once.
760 * Returns 1 on success. 0 on failure.
761 *
762 * Should call bd_holder_release_dirs() after successful use.
763 */
764static int bd_holder_grab_dirs(struct block_device *bdev,
765 struct bd_holder *bo)
766{
767 if (!bdev || !bo)
768 return 0;
769
770 bo->sdir = kobject_get(bo->sdir);
771 if (!bo->sdir)
772 return 0;
773
774 bo->hdev = kobject_get(bo->sdir->parent);
775 if (!bo->hdev)
776 goto fail_put_sdir;
777
Tejun Heo0762b8b2008-08-25 19:56:12 +0900778 bo->sdev = kobject_get(&part_to_dev(bdev->bd_part)->kobj);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800779 if (!bo->sdev)
780 goto fail_put_hdev;
781
Tejun Heo4c465012008-08-25 19:56:11 +0900782 bo->hdir = kobject_get(bdev->bd_part->holder_dir);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800783 if (!bo->hdir)
784 goto fail_put_sdev;
785
786 return 1;
787
788fail_put_sdev:
789 kobject_put(bo->sdev);
790fail_put_hdev:
791 kobject_put(bo->hdev);
792fail_put_sdir:
793 kobject_put(bo->sdir);
794
795 return 0;
796}
797
798/* Put references of related kobjects at once. */
799static void bd_holder_release_dirs(struct bd_holder *bo)
800{
801 kobject_put(bo->hdir);
802 kobject_put(bo->sdev);
803 kobject_put(bo->hdev);
804 kobject_put(bo->sdir);
805}
806
807static struct bd_holder *alloc_bd_holder(struct kobject *kobj)
808{
809 struct bd_holder *bo;
810
811 bo = kzalloc(sizeof(*bo), GFP_KERNEL);
812 if (!bo)
813 return NULL;
814
815 bo->count = 1;
816 bo->sdir = kobj;
817
818 return bo;
819}
820
821static void free_bd_holder(struct bd_holder *bo)
822{
823 kfree(bo);
824}
825
826/**
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500827 * find_bd_holder - find matching struct bd_holder from the block device
828 *
829 * @bdev: struct block device to be searched
830 * @bo: target struct bd_holder
831 *
832 * Returns matching entry with @bo in @bdev->bd_holder_list.
833 * If found, increment the reference count and return the pointer.
834 * If not found, returns NULL.
835 */
Andrew Morton36a561d2006-10-30 22:07:03 -0800836static struct bd_holder *find_bd_holder(struct block_device *bdev,
837 struct bd_holder *bo)
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500838{
839 struct bd_holder *tmp;
840
841 list_for_each_entry(tmp, &bdev->bd_holder_list, list)
842 if (tmp->sdir == bo->sdir) {
843 tmp->count++;
844 return tmp;
845 }
846
847 return NULL;
848}
849
850/**
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800851 * add_bd_holder - create sysfs symlinks for bd_claim() relationship
852 *
853 * @bdev: block device to be bd_claimed
854 * @bo: preallocated and initialized by alloc_bd_holder()
855 *
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500856 * Add @bo to @bdev->bd_holder_list, create symlinks.
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800857 *
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500858 * Returns 0 if symlinks are created.
859 * Returns -ve if something fails.
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800860 */
861static int add_bd_holder(struct block_device *bdev, struct bd_holder *bo)
862{
Johannes Weiner4e916722007-07-15 23:41:25 -0700863 int err;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800864
865 if (!bo)
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700866 return -EINVAL;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800867
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800868 if (!bd_holder_grab_dirs(bdev, bo))
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700869 return -EBUSY;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800870
Johannes Weiner4e916722007-07-15 23:41:25 -0700871 err = add_symlink(bo->sdir, bo->sdev);
872 if (err)
873 return err;
874
875 err = add_symlink(bo->hdir, bo->hdev);
876 if (err) {
877 del_symlink(bo->sdir, bo->sdev);
878 return err;
Andrew Morton4d7dd8f2006-09-29 01:58:56 -0700879 }
Johannes Weiner4e916722007-07-15 23:41:25 -0700880
881 list_add_tail(&bo->list, &bdev->bd_holder_list);
882 return 0;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800883}
884
885/**
886 * del_bd_holder - delete sysfs symlinks for bd_claim() relationship
887 *
888 * @bdev: block device to be bd_claimed
889 * @kobj: holder's kobject
890 *
891 * If there is matching entry with @kobj in @bdev->bd_holder_list
892 * and no other bd_claim() from the same kobject,
893 * remove the struct bd_holder from the list, delete symlinks for it.
894 *
895 * Returns a pointer to the struct bd_holder when it's removed from the list
896 * and ready to be freed.
897 * Returns NULL if matching claim isn't found or there is other bd_claim()
898 * by the same kobject.
899 */
900static struct bd_holder *del_bd_holder(struct block_device *bdev,
901 struct kobject *kobj)
902{
903 struct bd_holder *bo;
904
905 list_for_each_entry(bo, &bdev->bd_holder_list, list) {
906 if (bo->sdir == kobj) {
907 bo->count--;
908 BUG_ON(bo->count < 0);
909 if (!bo->count) {
910 list_del(&bo->list);
911 del_symlink(bo->sdir, bo->sdev);
912 del_symlink(bo->hdir, bo->hdev);
913 bd_holder_release_dirs(bo);
914 return bo;
915 }
916 break;
917 }
918 }
919
920 return NULL;
921}
922
923/**
924 * bd_claim_by_kobject - bd_claim() with additional kobject signature
925 *
926 * @bdev: block device to be claimed
927 * @holder: holder's signature
928 * @kobj: holder's kobject
929 *
930 * Do bd_claim() and if it succeeds, create sysfs symlinks between
931 * the bdev and the holder's kobject.
932 * Use bd_release_from_kobject() when relesing the claimed bdev.
933 *
934 * Returns 0 on success. (same as bd_claim())
935 * Returns errno on failure.
936 */
937static int bd_claim_by_kobject(struct block_device *bdev, void *holder,
938 struct kobject *kobj)
939{
Johannes Weiner4e916722007-07-15 23:41:25 -0700940 int err;
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500941 struct bd_holder *bo, *found;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800942
943 if (!kobj)
944 return -EINVAL;
945
946 bo = alloc_bd_holder(kobj);
947 if (!bo)
948 return -ENOMEM;
949
Peter Zijlstra2e7b6512006-12-08 02:36:13 -0800950 mutex_lock(&bdev->bd_mutex);
Jun'ichi Nomuradf6c0cd2006-10-30 16:23:56 -0500951
Johannes Weiner4e916722007-07-15 23:41:25 -0700952 err = bd_claim(bdev, holder);
953 if (err)
Andrew Morton4210df22007-07-15 23:41:28 -0700954 goto fail;
Johannes Weiner4e916722007-07-15 23:41:25 -0700955
956 found = find_bd_holder(bdev, bo);
957 if (found)
Andrew Morton4210df22007-07-15 23:41:28 -0700958 goto fail;
Johannes Weiner4e916722007-07-15 23:41:25 -0700959
960 err = add_bd_holder(bdev, bo);
961 if (err)
962 bd_release(bdev);
Andrew Morton4210df22007-07-15 23:41:28 -0700963 else
964 bo = NULL;
965fail:
Jun'ichi Nomurab4cf1b72006-03-27 01:18:00 -0800966 mutex_unlock(&bdev->bd_mutex);
Andrew Morton4210df22007-07-15 23:41:28 -0700967 free_bd_holder(bo);
Johannes Weiner4e916722007-07-15 23:41:25 -0700968 return err;
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800969}
970
971/**
972 * bd_release_from_kobject - bd_release() with additional kobject signature
973 *
974 * @bdev: block device to be released
975 * @kobj: holder's kobject
976 *
977 * Do bd_release() and remove sysfs symlinks created by bd_claim_by_kobject().
978 */
979static void bd_release_from_kobject(struct block_device *bdev,
980 struct kobject *kobj)
981{
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800982 if (!kobj)
983 return;
984
Peter Zijlstra2e7b6512006-12-08 02:36:13 -0800985 mutex_lock(&bdev->bd_mutex);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800986 bd_release(bdev);
Andrew Morton4210df22007-07-15 23:41:28 -0700987 free_bd_holder(del_bd_holder(bdev, kobj));
Jun'ichi Nomurab4cf1b72006-03-27 01:18:00 -0800988 mutex_unlock(&bdev->bd_mutex);
Jun'ichi Nomura641dc632006-03-27 01:17:57 -0800989}
990
991/**
992 * bd_claim_by_disk - wrapper function for bd_claim_by_kobject()
993 *
994 * @bdev: block device to be claimed
995 * @holder: holder's signature
996 * @disk: holder's gendisk
997 *
998 * Call bd_claim_by_kobject() with getting @disk->slave_dir.
999 */
1000int bd_claim_by_disk(struct block_device *bdev, void *holder,
1001 struct gendisk *disk)
1002{
1003 return bd_claim_by_kobject(bdev, holder, kobject_get(disk->slave_dir));
1004}
1005EXPORT_SYMBOL_GPL(bd_claim_by_disk);
1006
1007/**
1008 * bd_release_from_disk - wrapper function for bd_release_from_kobject()
1009 *
1010 * @bdev: block device to be claimed
1011 * @disk: holder's gendisk
1012 *
1013 * Call bd_release_from_kobject() and put @disk->slave_dir.
1014 */
1015void bd_release_from_disk(struct block_device *bdev, struct gendisk *disk)
1016{
1017 bd_release_from_kobject(bdev, disk->slave_dir);
1018 kobject_put(disk->slave_dir);
1019}
1020EXPORT_SYMBOL_GPL(bd_release_from_disk);
1021#endif
1022
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023/*
1024 * Tries to open block device by device number. Use it ONLY if you
1025 * really do not have anything better - i.e. when you are behind a
1026 * truly sucky interface and all you are given is a device number. _Never_
1027 * to be used for internal purposes. If you ever need it - reconsider
1028 * your API.
1029 */
Al Viroaeb5d722008-09-02 15:28:45 -04001030struct block_device *open_by_devnum(dev_t dev, fmode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031{
1032 struct block_device *bdev = bdget(dev);
1033 int err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 if (bdev)
Al Viro572c4892007-10-08 13:24:05 -04001035 err = blkdev_get(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 return err ? ERR_PTR(err) : bdev;
1037}
1038
1039EXPORT_SYMBOL(open_by_devnum);
1040
Andrew Patterson0c002c22008-09-04 14:27:20 -06001041/**
Andrew Patterson56ade442008-09-04 14:27:40 -06001042 * flush_disk - invalidates all buffer-cache entries on a disk
1043 *
1044 * @bdev: struct block device to be flushed
1045 *
1046 * Invalidates all buffer-cache entries on a disk. It should be called
1047 * when a disk has been changed -- either by a media change or online
1048 * resize.
1049 */
1050static void flush_disk(struct block_device *bdev)
1051{
1052 if (__invalidate_device(bdev)) {
1053 char name[BDEVNAME_SIZE] = "";
1054
1055 if (bdev->bd_disk)
1056 disk_name(bdev->bd_disk, 0, name);
1057 printk(KERN_WARNING "VFS: busy inodes on changed media or "
1058 "resized disk %s\n", name);
1059 }
1060
1061 if (!bdev->bd_disk)
1062 return;
1063 if (disk_partitionable(bdev->bd_disk))
1064 bdev->bd_invalidated = 1;
1065}
1066
1067/**
Randy Dunlap57d1b532008-10-09 10:42:38 +02001068 * check_disk_size_change - checks for disk size change and adjusts bdev size.
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001069 * @disk: struct gendisk to check
1070 * @bdev: struct bdev to adjust.
1071 *
1072 * This routine checks to see if the bdev size does not match the disk size
1073 * and adjusts it if it differs.
1074 */
1075void check_disk_size_change(struct gendisk *disk, struct block_device *bdev)
1076{
1077 loff_t disk_size, bdev_size;
1078
1079 disk_size = (loff_t)get_capacity(disk) << 9;
1080 bdev_size = i_size_read(bdev->bd_inode);
1081 if (disk_size != bdev_size) {
1082 char name[BDEVNAME_SIZE];
1083
1084 disk_name(disk, 0, name);
1085 printk(KERN_INFO
1086 "%s: detected capacity change from %lld to %lld\n",
1087 name, bdev_size, disk_size);
1088 i_size_write(bdev->bd_inode, disk_size);
Andrew Patterson608aeef2008-09-04 14:27:45 -06001089 flush_disk(bdev);
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001090 }
1091}
1092EXPORT_SYMBOL(check_disk_size_change);
1093
1094/**
Randy Dunlap57d1b532008-10-09 10:42:38 +02001095 * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back
Andrew Patterson0c002c22008-09-04 14:27:20 -06001096 * @disk: struct gendisk to be revalidated
1097 *
1098 * This routine is a wrapper for lower-level driver's revalidate_disk
1099 * call-backs. It is used to do common pre and post operations needed
1100 * for all revalidate_disk operations.
1101 */
1102int revalidate_disk(struct gendisk *disk)
1103{
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001104 struct block_device *bdev;
Andrew Patterson0c002c22008-09-04 14:27:20 -06001105 int ret = 0;
1106
1107 if (disk->fops->revalidate_disk)
1108 ret = disk->fops->revalidate_disk(disk);
1109
Andrew Pattersonc3279d12008-09-04 14:27:25 -06001110 bdev = bdget_disk(disk, 0);
1111 if (!bdev)
1112 return ret;
1113
1114 mutex_lock(&bdev->bd_mutex);
1115 check_disk_size_change(disk, bdev);
1116 mutex_unlock(&bdev->bd_mutex);
1117 bdput(bdev);
Andrew Patterson0c002c22008-09-04 14:27:20 -06001118 return ret;
1119}
1120EXPORT_SYMBOL(revalidate_disk);
1121
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122/*
1123 * This routine checks whether a removable media has been changed,
1124 * and invalidates all buffer-cache-entries in that case. This
1125 * is a relatively slow routine, so we have to try to minimize using
1126 * it. Thus it is called only upon a 'mount' or 'open'. This
1127 * is the best way of combining speed and utility, I think.
1128 * People changing diskettes in the middle of an operation deserve
1129 * to lose :-)
1130 */
1131int check_disk_change(struct block_device *bdev)
1132{
1133 struct gendisk *disk = bdev->bd_disk;
Alexey Dobriyan83d5cde2009-09-21 17:01:13 -07001134 const struct block_device_operations *bdops = disk->fops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135
1136 if (!bdops->media_changed)
1137 return 0;
1138 if (!bdops->media_changed(bdev->bd_disk))
1139 return 0;
1140
Andrew Patterson56ade442008-09-04 14:27:40 -06001141 flush_disk(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 if (bdops->revalidate_disk)
1143 bdops->revalidate_disk(bdev->bd_disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 return 1;
1145}
1146
1147EXPORT_SYMBOL(check_disk_change);
1148
1149void bd_set_size(struct block_device *bdev, loff_t size)
1150{
Martin K. Petersene1defc42009-05-22 17:17:49 -04001151 unsigned bsize = bdev_logical_block_size(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152
1153 bdev->bd_inode->i_size = size;
1154 while (bsize < PAGE_CACHE_SIZE) {
1155 if (size & bsize)
1156 break;
1157 bsize <<= 1;
1158 }
1159 bdev->bd_block_size = bsize;
1160 bdev->bd_inode->i_blkbits = blksize_bits(bsize);
1161}
1162EXPORT_SYMBOL(bd_set_size);
1163
Al Viro9a1c3542008-02-22 20:40:24 -05001164static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
NeilBrown37be4122006-12-08 02:36:16 -08001165
Peter Zijlstra6d740cd2007-02-20 13:58:18 -08001166/*
1167 * bd_mutex locking:
1168 *
1169 * mutex_lock(part->bd_mutex)
1170 * mutex_lock_nested(whole->bd_mutex, 1)
1171 */
1172
Al Viro572c4892007-10-08 13:24:05 -04001173static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 struct gendisk *disk;
Pavel Emelyanov7db9cfd2008-06-05 22:46:27 -07001176 int ret;
Tejun Heocf771cb2008-09-03 09:01:09 +02001177 int partno;
Al Virofe6e9c12008-06-23 08:30:55 -04001178 int perm = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179
Al Viro572c4892007-10-08 13:24:05 -04001180 if (mode & FMODE_READ)
Al Virofe6e9c12008-06-23 08:30:55 -04001181 perm |= MAY_READ;
Al Viro572c4892007-10-08 13:24:05 -04001182 if (mode & FMODE_WRITE)
Al Virofe6e9c12008-06-23 08:30:55 -04001183 perm |= MAY_WRITE;
1184 /*
1185 * hooks: /n/, see "layering violations".
1186 */
1187 ret = devcgroup_inode_permission(bdev->bd_inode, perm);
Al Viro82666022008-08-01 05:32:04 -04001188 if (ret != 0) {
1189 bdput(bdev);
Pavel Emelyanov7db9cfd2008-06-05 22:46:27 -07001190 return ret;
Al Viro82666022008-08-01 05:32:04 -04001191 }
Pavel Emelyanov7db9cfd2008-06-05 22:46:27 -07001192
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193 lock_kernel();
NeilBrownd3374822009-01-09 08:31:10 +11001194 restart:
Tejun Heo0762b8b2008-08-25 19:56:12 +09001195
Tejun Heo89f97492008-11-05 10:21:06 +01001196 ret = -ENXIO;
Tejun Heocf771cb2008-09-03 09:01:09 +02001197 disk = get_gendisk(bdev->bd_dev, &partno);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001198 if (!disk)
1199 goto out_unlock_kernel;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200
NeilBrown6796bf52006-12-08 02:36:16 -08001201 mutex_lock_nested(&bdev->bd_mutex, for_part);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 if (!bdev->bd_openers) {
1203 bdev->bd_disk = disk;
1204 bdev->bd_contains = bdev;
Tejun Heocf771cb2008-09-03 09:01:09 +02001205 if (!partno) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 struct backing_dev_info *bdi;
Tejun Heo89f97492008-11-05 10:21:06 +01001207
1208 ret = -ENXIO;
1209 bdev->bd_part = disk_get_part(disk, partno);
1210 if (!bdev->bd_part)
1211 goto out_clear;
1212
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 if (disk->fops->open) {
Al Viro572c4892007-10-08 13:24:05 -04001214 ret = disk->fops->open(bdev, mode);
NeilBrownd3374822009-01-09 08:31:10 +11001215 if (ret == -ERESTARTSYS) {
1216 /* Lost a race with 'disk' being
1217 * deleted, try again.
1218 * See md.c
1219 */
1220 disk_put_part(bdev->bd_part);
1221 bdev->bd_part = NULL;
1222 module_put(disk->fops->owner);
1223 put_disk(disk);
1224 bdev->bd_disk = NULL;
1225 mutex_unlock(&bdev->bd_mutex);
1226 goto restart;
1227 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228 if (ret)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001229 goto out_clear;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 }
1231 if (!bdev->bd_openers) {
1232 bd_set_size(bdev,(loff_t)get_capacity(disk)<<9);
1233 bdi = blk_get_backing_dev_info(bdev);
1234 if (bdi == NULL)
1235 bdi = &default_backing_dev_info;
1236 bdev->bd_inode->i_data.backing_dev_info = bdi;
1237 }
1238 if (bdev->bd_invalidated)
1239 rescan_partitions(disk, bdev);
1240 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 struct block_device *whole;
1242 whole = bdget_disk(disk, 0);
1243 ret = -ENOMEM;
1244 if (!whole)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001245 goto out_clear;
NeilBrown37be4122006-12-08 02:36:16 -08001246 BUG_ON(for_part);
Al Viro572c4892007-10-08 13:24:05 -04001247 ret = __blkdev_get(whole, mode, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 if (ret)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001249 goto out_clear;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 bdev->bd_contains = whole;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 bdev->bd_inode->i_data.backing_dev_info =
1252 whole->bd_inode->i_data.backing_dev_info;
Tejun Heo89f97492008-11-05 10:21:06 +01001253 bdev->bd_part = disk_get_part(disk, partno);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001254 if (!(disk->flags & GENHD_FL_UP) ||
Tejun Heo89f97492008-11-05 10:21:06 +01001255 !bdev->bd_part || !bdev->bd_part->nr_sects) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 ret = -ENXIO;
Tejun Heo0762b8b2008-08-25 19:56:12 +09001257 goto out_clear;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 }
Tejun Heo89f97492008-11-05 10:21:06 +01001259 bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 }
1261 } else {
Tejun Heo0762b8b2008-08-25 19:56:12 +09001262 module_put(disk->fops->owner);
Neil Brown960cc0f2009-10-26 08:59:17 +01001263 put_disk(disk);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001264 disk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265 if (bdev->bd_contains == bdev) {
1266 if (bdev->bd_disk->fops->open) {
Al Viro572c4892007-10-08 13:24:05 -04001267 ret = bdev->bd_disk->fops->open(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 if (ret)
Tejun Heo0762b8b2008-08-25 19:56:12 +09001269 goto out_unlock_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 }
1271 if (bdev->bd_invalidated)
1272 rescan_partitions(bdev->bd_disk, bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 }
1274 }
1275 bdev->bd_openers++;
NeilBrown37be4122006-12-08 02:36:16 -08001276 if (for_part)
1277 bdev->bd_part_count++;
Arjan van de Venc039e312006-03-23 03:00:28 -08001278 mutex_unlock(&bdev->bd_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 unlock_kernel();
1280 return 0;
1281
Tejun Heo0762b8b2008-08-25 19:56:12 +09001282 out_clear:
Tejun Heo89f97492008-11-05 10:21:06 +01001283 disk_put_part(bdev->bd_part);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 bdev->bd_disk = NULL;
Tejun Heo0762b8b2008-08-25 19:56:12 +09001285 bdev->bd_part = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286 bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info;
1287 if (bdev != bdev->bd_contains)
Al Viro572c4892007-10-08 13:24:05 -04001288 __blkdev_put(bdev->bd_contains, mode, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 bdev->bd_contains = NULL;
Tejun Heo0762b8b2008-08-25 19:56:12 +09001290 out_unlock_bdev:
Arjan van de Venc039e312006-03-23 03:00:28 -08001291 mutex_unlock(&bdev->bd_mutex);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001292 out_unlock_kernel:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 unlock_kernel();
Tejun Heo0762b8b2008-08-25 19:56:12 +09001294
Tejun Heo0762b8b2008-08-25 19:56:12 +09001295 if (disk)
1296 module_put(disk->fops->owner);
1297 put_disk(disk);
1298 bdput(bdev);
1299
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 return ret;
1301}
1302
Al Viro572c4892007-10-08 13:24:05 -04001303int blkdev_get(struct block_device *bdev, fmode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304{
Al Viro572c4892007-10-08 13:24:05 -04001305 return __blkdev_get(bdev, mode, 0);
NeilBrown37be4122006-12-08 02:36:16 -08001306}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307EXPORT_SYMBOL(blkdev_get);
1308
1309static int blkdev_open(struct inode * inode, struct file * filp)
1310{
1311 struct block_device *bdev;
1312 int res;
1313
1314 /*
1315 * Preserve backwards compatibility and allow large file access
1316 * even if userspace doesn't ask for it explicitly. Some mkfs
1317 * binary needs it. We might want to drop this workaround
1318 * during an unstable branch.
1319 */
1320 filp->f_flags |= O_LARGEFILE;
1321
Al Viro572c4892007-10-08 13:24:05 -04001322 if (filp->f_flags & O_NDELAY)
1323 filp->f_mode |= FMODE_NDELAY;
1324 if (filp->f_flags & O_EXCL)
1325 filp->f_mode |= FMODE_EXCL;
1326 if ((filp->f_flags & O_ACCMODE) == 3)
1327 filp->f_mode |= FMODE_WRITE_IOCTL;
1328
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 bdev = bd_acquire(inode);
Pavel Emelianov6a2aae02006-10-28 10:38:33 -07001330 if (bdev == NULL)
1331 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332
Al Viro572c4892007-10-08 13:24:05 -04001333 filp->f_mapping = bdev->bd_inode->i_mapping;
1334
1335 res = blkdev_get(bdev, filp->f_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 if (res)
1337 return res;
1338
Christoph Hellwigebbefc02008-11-05 14:54:41 +01001339 if (filp->f_mode & FMODE_EXCL) {
1340 res = bd_claim(bdev, filp);
1341 if (res)
1342 goto out_blkdev_put;
1343 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344
Christoph Hellwigebbefc02008-11-05 14:54:41 +01001345 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346
Christoph Hellwigebbefc02008-11-05 14:54:41 +01001347 out_blkdev_put:
Al Viro9a1c3542008-02-22 20:40:24 -05001348 blkdev_put(bdev, filp->f_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 return res;
1350}
1351
Al Viro9a1c3542008-02-22 20:40:24 -05001352static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part)
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001353{
1354 int ret = 0;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001355 struct gendisk *disk = bdev->bd_disk;
NeilBrown37be4122006-12-08 02:36:16 -08001356 struct block_device *victim = NULL;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001357
NeilBrown6796bf52006-12-08 02:36:16 -08001358 mutex_lock_nested(&bdev->bd_mutex, for_part);
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001359 lock_kernel();
NeilBrown37be4122006-12-08 02:36:16 -08001360 if (for_part)
1361 bdev->bd_part_count--;
1362
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001363 if (!--bdev->bd_openers) {
1364 sync_blockdev(bdev);
1365 kill_bdev(bdev);
1366 }
1367 if (bdev->bd_contains == bdev) {
1368 if (disk->fops->release)
Al Viro9a1c3542008-02-22 20:40:24 -05001369 ret = disk->fops->release(disk, mode);
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001370 }
1371 if (!bdev->bd_openers) {
1372 struct module *owner = disk->fops->owner;
1373
1374 put_disk(disk);
1375 module_put(owner);
Tejun Heo0762b8b2008-08-25 19:56:12 +09001376 disk_put_part(bdev->bd_part);
1377 bdev->bd_part = NULL;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001378 bdev->bd_disk = NULL;
1379 bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info;
NeilBrown37be4122006-12-08 02:36:16 -08001380 if (bdev != bdev->bd_contains)
1381 victim = bdev->bd_contains;
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001382 bdev->bd_contains = NULL;
1383 }
1384 unlock_kernel();
1385 mutex_unlock(&bdev->bd_mutex);
1386 bdput(bdev);
NeilBrown37be4122006-12-08 02:36:16 -08001387 if (victim)
Al Viro9a1c3542008-02-22 20:40:24 -05001388 __blkdev_put(victim, mode, 1);
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001389 return ret;
1390}
1391
Al Viro9a1c3542008-02-22 20:40:24 -05001392int blkdev_put(struct block_device *bdev, fmode_t mode)
NeilBrown37be4122006-12-08 02:36:16 -08001393{
Al Viro9a1c3542008-02-22 20:40:24 -05001394 return __blkdev_put(bdev, mode, 0);
NeilBrown37be4122006-12-08 02:36:16 -08001395}
Peter Zijlstra2e7b6512006-12-08 02:36:13 -08001396EXPORT_SYMBOL(blkdev_put);
1397
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398static int blkdev_close(struct inode * inode, struct file * filp)
1399{
1400 struct block_device *bdev = I_BDEV(filp->f_mapping->host);
1401 if (bdev->bd_holder == filp)
1402 bd_release(bdev);
Al Viro9a1c3542008-02-22 20:40:24 -05001403 return blkdev_put(bdev, filp->f_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404}
1405
Arnd Bergmannbb93e3a2005-06-23 00:10:15 -07001406static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407{
Al Viro56b26ad2008-09-19 03:17:36 -04001408 struct block_device *bdev = I_BDEV(file->f_mapping->host);
1409 fmode_t mode = file->f_mode;
Christoph Hellwigfd4ce1a2008-11-05 14:58:42 +01001410
1411 /*
1412 * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have
1413 * to updated it before every ioctl.
1414 */
Al Viro56b26ad2008-09-19 03:17:36 -04001415 if (file->f_flags & O_NDELAY)
Christoph Hellwigfd4ce1a2008-11-05 14:58:42 +01001416 mode |= FMODE_NDELAY;
1417 else
1418 mode &= ~FMODE_NDELAY;
1419
Al Viro56b26ad2008-09-19 03:17:36 -04001420 return blkdev_ioctl(bdev, mode, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421}
1422
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001423/*
Christoph Hellwigeef99382009-08-20 17:43:41 +02001424 * Write data to the block device. Only intended for the block device itself
1425 * and the raw driver which basically is a fake block device.
1426 *
1427 * Does not take i_mutex for the write and thus is not for general purpose
1428 * use.
1429 */
1430ssize_t blkdev_aio_write(struct kiocb *iocb, const struct iovec *iov,
1431 unsigned long nr_segs, loff_t pos)
1432{
1433 struct file *file = iocb->ki_filp;
1434 ssize_t ret;
1435
1436 BUG_ON(iocb->ki_pos != pos);
1437
1438 ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos);
1439 if (ret > 0 || ret == -EIOCBQUEUED) {
1440 ssize_t err;
1441
1442 err = generic_write_sync(file, pos, ret);
1443 if (err < 0 && ret > 0)
1444 ret = err;
1445 }
1446 return ret;
1447}
1448EXPORT_SYMBOL_GPL(blkdev_aio_write);
1449
1450/*
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001451 * Try to release a page associated with block device when the system
1452 * is under memory pressure.
1453 */
1454static int blkdev_releasepage(struct page *page, gfp_t wait)
1455{
1456 struct super_block *super = BDEV_I(page->mapping->host)->bdev.bd_super;
1457
1458 if (super && super->s_op->bdev_try_to_free_page)
1459 return super->s_op->bdev_try_to_free_page(super, page, wait);
1460
1461 return try_to_free_buffers(page);
1462}
1463
Adrian Bunk4c54ac62008-02-18 13:48:31 +01001464static const struct address_space_operations def_blk_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465 .readpage = blkdev_readpage,
1466 .writepage = blkdev_writepage,
1467 .sync_page = block_sync_page,
Nick Piggin6272b5a2007-10-16 01:25:04 -07001468 .write_begin = blkdev_write_begin,
1469 .write_end = blkdev_write_end,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 .writepages = generic_writepages,
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001471 .releasepage = blkdev_releasepage,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 .direct_IO = blkdev_direct_IO,
1473};
1474
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -08001475const struct file_operations def_blk_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 .open = blkdev_open,
1477 .release = blkdev_close,
1478 .llseek = block_llseek,
Badari Pulavarty543ade12006-09-30 23:28:48 -07001479 .read = do_sync_read,
1480 .write = do_sync_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 .aio_read = generic_file_aio_read,
Christoph Hellwigeef99382009-08-20 17:43:41 +02001482 .aio_write = blkdev_aio_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 .mmap = generic_file_mmap,
1484 .fsync = block_fsync,
Arnd Bergmannbb93e3a2005-06-23 00:10:15 -07001485 .unlocked_ioctl = block_ioctl,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486#ifdef CONFIG_COMPAT
1487 .compat_ioctl = compat_blkdev_ioctl,
1488#endif
Jens Axboe7f9c51f2006-05-01 19:59:32 +02001489 .splice_read = generic_file_splice_read,
1490 .splice_write = generic_file_splice_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491};
1492
1493int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg)
1494{
1495 int res;
1496 mm_segment_t old_fs = get_fs();
1497 set_fs(KERNEL_DS);
Al Viro56b26ad2008-09-19 03:17:36 -04001498 res = blkdev_ioctl(bdev, 0, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 set_fs(old_fs);
1500 return res;
1501}
1502
1503EXPORT_SYMBOL(ioctl_by_bdev);
1504
1505/**
1506 * lookup_bdev - lookup a struct block_device by name
Randy Dunlap94e29592009-01-06 14:41:15 -08001507 * @pathname: special file representing the block device
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508 *
Randy Dunlap57d1b532008-10-09 10:42:38 +02001509 * Get a reference to the blockdevice at @pathname in the current
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 * namespace if possible and return it. Return ERR_PTR(error)
1511 * otherwise.
1512 */
Al Viro421748e2008-08-02 01:04:36 -04001513struct block_device *lookup_bdev(const char *pathname)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514{
1515 struct block_device *bdev;
1516 struct inode *inode;
Al Viro421748e2008-08-02 01:04:36 -04001517 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 int error;
1519
Al Viro421748e2008-08-02 01:04:36 -04001520 if (!pathname || !*pathname)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 return ERR_PTR(-EINVAL);
1522
Al Viro421748e2008-08-02 01:04:36 -04001523 error = kern_path(pathname, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 if (error)
1525 return ERR_PTR(error);
1526
Al Viro421748e2008-08-02 01:04:36 -04001527 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 error = -ENOTBLK;
1529 if (!S_ISBLK(inode->i_mode))
1530 goto fail;
1531 error = -EACCES;
Al Viro421748e2008-08-02 01:04:36 -04001532 if (path.mnt->mnt_flags & MNT_NODEV)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 goto fail;
1534 error = -ENOMEM;
1535 bdev = bd_acquire(inode);
1536 if (!bdev)
1537 goto fail;
1538out:
Al Viro421748e2008-08-02 01:04:36 -04001539 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 return bdev;
1541fail:
1542 bdev = ERR_PTR(error);
1543 goto out;
1544}
Al Virod5686b42008-08-01 05:00:11 -04001545EXPORT_SYMBOL(lookup_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546
1547/**
Al Viro30c40d22008-02-22 19:50:45 -05001548 * open_bdev_exclusive - open a block device by name and set it up for use
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 *
1550 * @path: special file representing the block device
Al Viro30c40d22008-02-22 19:50:45 -05001551 * @mode: FMODE_... combination to pass be used
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 * @holder: owner for exclusion
1553 *
1554 * Open the blockdevice described by the special file at @path, claim it
1555 * for the @holder.
1556 */
Al Viro30c40d22008-02-22 19:50:45 -05001557struct block_device *open_bdev_exclusive(const char *path, fmode_t mode, void *holder)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558{
1559 struct block_device *bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 int error = 0;
1561
1562 bdev = lookup_bdev(path);
1563 if (IS_ERR(bdev))
1564 return bdev;
1565
Al Viro572c4892007-10-08 13:24:05 -04001566 error = blkdev_get(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 if (error)
1568 return ERR_PTR(error);
1569 error = -EACCES;
Al Viro30c40d22008-02-22 19:50:45 -05001570 if ((mode & FMODE_WRITE) && bdev_read_only(bdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 goto blkdev_put;
1572 error = bd_claim(bdev, holder);
1573 if (error)
1574 goto blkdev_put;
1575
1576 return bdev;
1577
1578blkdev_put:
Al Viro9a1c3542008-02-22 20:40:24 -05001579 blkdev_put(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 return ERR_PTR(error);
1581}
1582
Al Viro30c40d22008-02-22 19:50:45 -05001583EXPORT_SYMBOL(open_bdev_exclusive);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584
1585/**
Al Viro30c40d22008-02-22 19:50:45 -05001586 * close_bdev_exclusive - close a blockdevice opened by open_bdev_exclusive()
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 *
1588 * @bdev: blockdevice to close
Al Viro30c40d22008-02-22 19:50:45 -05001589 * @mode: mode, must match that used to open.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590 *
Al Viro30c40d22008-02-22 19:50:45 -05001591 * This is the counterpart to open_bdev_exclusive().
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 */
Al Viro30c40d22008-02-22 19:50:45 -05001593void close_bdev_exclusive(struct block_device *bdev, fmode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594{
1595 bd_release(bdev);
Al Viro30c40d22008-02-22 19:50:45 -05001596 blkdev_put(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597}
1598
Al Viro30c40d22008-02-22 19:50:45 -05001599EXPORT_SYMBOL(close_bdev_exclusive);
David Howellsb71e8a42006-08-29 19:06:11 +01001600
1601int __invalidate_device(struct block_device *bdev)
1602{
1603 struct super_block *sb = get_super(bdev);
1604 int res = 0;
1605
1606 if (sb) {
1607 /*
1608 * no need to lock the super, get_super holds the
1609 * read mutex so the filesystem cannot go away
1610 * under us (->put_super runs with the write lock
1611 * hold).
1612 */
1613 shrink_dcache_sb(sb);
1614 res = invalidate_inodes(sb);
1615 drop_super(sb);
1616 }
Peter Zijlstraf98393a2007-05-06 14:49:54 -07001617 invalidate_bdev(bdev);
David Howellsb71e8a42006-08-29 19:06:11 +01001618 return res;
1619}
1620EXPORT_SYMBOL(__invalidate_device);