diff options
Diffstat (limited to 'fs/block_dev.c')
-rw-r--r-- | fs/block_dev.c | 166 |
1 files changed, 16 insertions, 150 deletions
diff --git a/fs/block_dev.c b/fs/block_dev.c index 1a1e5e3b1eaf..ab3a456f6650 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c | |||
@@ -70,19 +70,6 @@ static void bdev_inode_switch_bdi(struct inode *inode, | |||
70 | spin_unlock(&dst->wb.list_lock); | 70 | spin_unlock(&dst->wb.list_lock); |
71 | } | 71 | } |
72 | 72 | ||
73 | sector_t blkdev_max_block(struct block_device *bdev) | ||
74 | { | ||
75 | sector_t retval = ~((sector_t)0); | ||
76 | loff_t sz = i_size_read(bdev->bd_inode); | ||
77 | |||
78 | if (sz) { | ||
79 | unsigned int size = block_size(bdev); | ||
80 | unsigned int sizebits = blksize_bits(size); | ||
81 | retval = (sz >> sizebits); | ||
82 | } | ||
83 | return retval; | ||
84 | } | ||
85 | |||
86 | /* Kill _all_ buffers and pagecache , dirty or not.. */ | 73 | /* Kill _all_ buffers and pagecache , dirty or not.. */ |
87 | void kill_bdev(struct block_device *bdev) | 74 | void kill_bdev(struct block_device *bdev) |
88 | { | 75 | { |
@@ -116,8 +103,6 @@ EXPORT_SYMBOL(invalidate_bdev); | |||
116 | 103 | ||
117 | int set_blocksize(struct block_device *bdev, int size) | 104 | int set_blocksize(struct block_device *bdev, int size) |
118 | { | 105 | { |
119 | struct address_space *mapping; | ||
120 | |||
121 | /* Size must be a power of two, and between 512 and PAGE_SIZE */ | 106 | /* Size must be a power of two, and between 512 and PAGE_SIZE */ |
122 | if (size > PAGE_SIZE || size < 512 || !is_power_of_2(size)) | 107 | if (size > PAGE_SIZE || size < 512 || !is_power_of_2(size)) |
123 | return -EINVAL; | 108 | return -EINVAL; |
@@ -126,19 +111,6 @@ int set_blocksize(struct block_device *bdev, int size) | |||
126 | if (size < bdev_logical_block_size(bdev)) | 111 | if (size < bdev_logical_block_size(bdev)) |
127 | return -EINVAL; | 112 | return -EINVAL; |
128 | 113 | ||
129 | /* Prevent starting I/O or mapping the device */ | ||
130 | percpu_down_write(&bdev->bd_block_size_semaphore); | ||
131 | |||
132 | /* Check that the block device is not memory mapped */ | ||
133 | mapping = bdev->bd_inode->i_mapping; | ||
134 | mutex_lock(&mapping->i_mmap_mutex); | ||
135 | if (mapping_mapped(mapping)) { | ||
136 | mutex_unlock(&mapping->i_mmap_mutex); | ||
137 | percpu_up_write(&bdev->bd_block_size_semaphore); | ||
138 | return -EBUSY; | ||
139 | } | ||
140 | mutex_unlock(&mapping->i_mmap_mutex); | ||
141 | |||
142 | /* Don't change the size if it is same as current */ | 114 | /* Don't change the size if it is same as current */ |
143 | if (bdev->bd_block_size != size) { | 115 | if (bdev->bd_block_size != size) { |
144 | sync_blockdev(bdev); | 116 | sync_blockdev(bdev); |
@@ -146,9 +118,6 @@ int set_blocksize(struct block_device *bdev, int size) | |||
146 | bdev->bd_inode->i_blkbits = blksize_bits(size); | 118 | bdev->bd_inode->i_blkbits = blksize_bits(size); |
147 | kill_bdev(bdev); | 119 | kill_bdev(bdev); |
148 | } | 120 | } |
149 | |||
150 | percpu_up_write(&bdev->bd_block_size_semaphore); | ||
151 | |||
152 | return 0; | 121 | return 0; |
153 | } | 122 | } |
154 | 123 | ||
@@ -181,52 +150,12 @@ static int | |||
181 | blkdev_get_block(struct inode *inode, sector_t iblock, | 150 | blkdev_get_block(struct inode *inode, sector_t iblock, |
182 | struct buffer_head *bh, int create) | 151 | struct buffer_head *bh, int create) |
183 | { | 152 | { |
184 | if (iblock >= blkdev_max_block(I_BDEV(inode))) { | ||
185 | if (create) | ||
186 | return -EIO; | ||
187 | |||
188 | /* | ||
189 | * for reads, we're just trying to fill a partial page. | ||
190 | * return a hole, they will have to call get_block again | ||
191 | * before they can fill it, and they will get -EIO at that | ||
192 | * time | ||
193 | */ | ||
194 | return 0; | ||
195 | } | ||
196 | bh->b_bdev = I_BDEV(inode); | 153 | bh->b_bdev = I_BDEV(inode); |
197 | bh->b_blocknr = iblock; | 154 | bh->b_blocknr = iblock; |
198 | set_buffer_mapped(bh); | 155 | set_buffer_mapped(bh); |
199 | return 0; | 156 | return 0; |
200 | } | 157 | } |
201 | 158 | ||
202 | static int | ||
203 | blkdev_get_blocks(struct inode *inode, sector_t iblock, | ||
204 | struct buffer_head *bh, int create) | ||
205 | { | ||
206 | sector_t end_block = blkdev_max_block(I_BDEV(inode)); | ||
207 | unsigned long max_blocks = bh->b_size >> inode->i_blkbits; | ||
208 | |||
209 | if ((iblock + max_blocks) > end_block) { | ||
210 | max_blocks = end_block - iblock; | ||
211 | if ((long)max_blocks <= 0) { | ||
212 | if (create) | ||
213 | return -EIO; /* write fully beyond EOF */ | ||
214 | /* | ||
215 | * It is a read which is fully beyond EOF. We return | ||
216 | * a !buffer_mapped buffer | ||
217 | */ | ||
218 | max_blocks = 0; | ||
219 | } | ||
220 | } | ||
221 | |||
222 | bh->b_bdev = I_BDEV(inode); | ||
223 | bh->b_blocknr = iblock; | ||
224 | bh->b_size = max_blocks << inode->i_blkbits; | ||
225 | if (max_blocks) | ||
226 | set_buffer_mapped(bh); | ||
227 | return 0; | ||
228 | } | ||
229 | |||
230 | static ssize_t | 159 | static ssize_t |
231 | blkdev_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, | 160 | blkdev_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, |
232 | loff_t offset, unsigned long nr_segs) | 161 | loff_t offset, unsigned long nr_segs) |
@@ -235,7 +164,7 @@ blkdev_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, | |||
235 | struct inode *inode = file->f_mapping->host; | 164 | struct inode *inode = file->f_mapping->host; |
236 | 165 | ||
237 | return __blockdev_direct_IO(rw, iocb, inode, I_BDEV(inode), iov, offset, | 166 | return __blockdev_direct_IO(rw, iocb, inode, I_BDEV(inode), iov, offset, |
238 | nr_segs, blkdev_get_blocks, NULL, NULL, 0); | 167 | nr_segs, blkdev_get_block, NULL, NULL, 0); |
239 | } | 168 | } |
240 | 169 | ||
241 | int __sync_blockdev(struct block_device *bdev, int wait) | 170 | int __sync_blockdev(struct block_device *bdev, int wait) |
@@ -459,12 +388,6 @@ static struct inode *bdev_alloc_inode(struct super_block *sb) | |||
459 | struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL); | 388 | struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL); |
460 | if (!ei) | 389 | if (!ei) |
461 | return NULL; | 390 | return NULL; |
462 | |||
463 | if (unlikely(percpu_init_rwsem(&ei->bdev.bd_block_size_semaphore))) { | ||
464 | kmem_cache_free(bdev_cachep, ei); | ||
465 | return NULL; | ||
466 | } | ||
467 | |||
468 | return &ei->vfs_inode; | 391 | return &ei->vfs_inode; |
469 | } | 392 | } |
470 | 393 | ||
@@ -473,8 +396,6 @@ static void bdev_i_callback(struct rcu_head *head) | |||
473 | struct inode *inode = container_of(head, struct inode, i_rcu); | 396 | struct inode *inode = container_of(head, struct inode, i_rcu); |
474 | struct bdev_inode *bdi = BDEV_I(inode); | 397 | struct bdev_inode *bdi = BDEV_I(inode); |
475 | 398 | ||
476 | percpu_free_rwsem(&bdi->bdev.bd_block_size_semaphore); | ||
477 | |||
478 | kmem_cache_free(bdev_cachep, bdi); | 399 | kmem_cache_free(bdev_cachep, bdi); |
479 | } | 400 | } |
480 | 401 | ||
@@ -1593,22 +1514,6 @@ static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg) | |||
1593 | return blkdev_ioctl(bdev, mode, cmd, arg); | 1514 | return blkdev_ioctl(bdev, mode, cmd, arg); |
1594 | } | 1515 | } |
1595 | 1516 | ||
1596 | ssize_t blkdev_aio_read(struct kiocb *iocb, const struct iovec *iov, | ||
1597 | unsigned long nr_segs, loff_t pos) | ||
1598 | { | ||
1599 | ssize_t ret; | ||
1600 | struct block_device *bdev = I_BDEV(iocb->ki_filp->f_mapping->host); | ||
1601 | |||
1602 | percpu_down_read(&bdev->bd_block_size_semaphore); | ||
1603 | |||
1604 | ret = generic_file_aio_read(iocb, iov, nr_segs, pos); | ||
1605 | |||
1606 | percpu_up_read(&bdev->bd_block_size_semaphore); | ||
1607 | |||
1608 | return ret; | ||
1609 | } | ||
1610 | EXPORT_SYMBOL_GPL(blkdev_aio_read); | ||
1611 | |||
1612 | /* | 1517 | /* |
1613 | * Write data to the block device. Only intended for the block device itself | 1518 | * Write data to the block device. Only intended for the block device itself |
1614 | * and the raw driver which basically is a fake block device. | 1519 | * and the raw driver which basically is a fake block device. |
@@ -1620,16 +1525,12 @@ ssize_t blkdev_aio_write(struct kiocb *iocb, const struct iovec *iov, | |||
1620 | unsigned long nr_segs, loff_t pos) | 1525 | unsigned long nr_segs, loff_t pos) |
1621 | { | 1526 | { |
1622 | struct file *file = iocb->ki_filp; | 1527 | struct file *file = iocb->ki_filp; |
1623 | struct block_device *bdev = I_BDEV(file->f_mapping->host); | ||
1624 | struct blk_plug plug; | 1528 | struct blk_plug plug; |
1625 | ssize_t ret; | 1529 | ssize_t ret; |
1626 | 1530 | ||
1627 | BUG_ON(iocb->ki_pos != pos); | 1531 | BUG_ON(iocb->ki_pos != pos); |
1628 | 1532 | ||
1629 | blk_start_plug(&plug); | 1533 | blk_start_plug(&plug); |
1630 | |||
1631 | percpu_down_read(&bdev->bd_block_size_semaphore); | ||
1632 | |||
1633 | ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos); | 1534 | ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos); |
1634 | if (ret > 0 || ret == -EIOCBQUEUED) { | 1535 | if (ret > 0 || ret == -EIOCBQUEUED) { |
1635 | ssize_t err; | 1536 | ssize_t err; |
@@ -1638,62 +1539,27 @@ ssize_t blkdev_aio_write(struct kiocb *iocb, const struct iovec *iov, | |||
1638 | if (err < 0 && ret > 0) | 1539 | if (err < 0 && ret > 0) |
1639 | ret = err; | 1540 | ret = err; |
1640 | } | 1541 | } |
1641 | |||
1642 | percpu_up_read(&bdev->bd_block_size_semaphore); | ||
1643 | |||
1644 | blk_finish_plug(&plug); | 1542 | blk_finish_plug(&plug); |
1645 | |||
1646 | return ret; | 1543 | return ret; |
1647 | } | 1544 | } |
1648 | EXPORT_SYMBOL_GPL(blkdev_aio_write); | 1545 | EXPORT_SYMBOL_GPL(blkdev_aio_write); |
1649 | 1546 | ||
1650 | static int blkdev_mmap(struct file *file, struct vm_area_struct *vma) | 1547 | static ssize_t blkdev_aio_read(struct kiocb *iocb, const struct iovec *iov, |
1651 | { | 1548 | unsigned long nr_segs, loff_t pos) |
1652 | int ret; | ||
1653 | struct block_device *bdev = I_BDEV(file->f_mapping->host); | ||
1654 | |||
1655 | percpu_down_read(&bdev->bd_block_size_semaphore); | ||
1656 | |||
1657 | ret = generic_file_mmap(file, vma); | ||
1658 | |||
1659 | percpu_up_read(&bdev->bd_block_size_semaphore); | ||
1660 | |||
1661 | return ret; | ||
1662 | } | ||
1663 | |||
1664 | static ssize_t blkdev_splice_read(struct file *file, loff_t *ppos, | ||
1665 | struct pipe_inode_info *pipe, size_t len, | ||
1666 | unsigned int flags) | ||
1667 | { | ||
1668 | ssize_t ret; | ||
1669 | struct block_device *bdev = I_BDEV(file->f_mapping->host); | ||
1670 | |||
1671 | percpu_down_read(&bdev->bd_block_size_semaphore); | ||
1672 | |||
1673 | ret = generic_file_splice_read(file, ppos, pipe, len, flags); | ||
1674 | |||
1675 | percpu_up_read(&bdev->bd_block_size_semaphore); | ||
1676 | |||
1677 | return ret; | ||
1678 | } | ||
1679 | |||
1680 | static ssize_t blkdev_splice_write(struct pipe_inode_info *pipe, | ||
1681 | struct file *file, loff_t *ppos, size_t len, | ||
1682 | unsigned int flags) | ||
1683 | { | 1549 | { |
1684 | ssize_t ret; | 1550 | struct file *file = iocb->ki_filp; |
1685 | struct block_device *bdev = I_BDEV(file->f_mapping->host); | 1551 | struct inode *bd_inode = file->f_mapping->host; |
1686 | 1552 | loff_t size = i_size_read(bd_inode); | |
1687 | percpu_down_read(&bdev->bd_block_size_semaphore); | ||
1688 | |||
1689 | ret = generic_file_splice_write(pipe, file, ppos, len, flags); | ||
1690 | 1553 | ||
1691 | percpu_up_read(&bdev->bd_block_size_semaphore); | 1554 | if (pos >= size) |
1555 | return 0; | ||
1692 | 1556 | ||
1693 | return ret; | 1557 | size -= pos; |
1558 | if (size < INT_MAX) | ||
1559 | nr_segs = iov_shorten((struct iovec *)iov, nr_segs, size); | ||
1560 | return generic_file_aio_read(iocb, iov, nr_segs, pos); | ||
1694 | } | 1561 | } |
1695 | 1562 | ||
1696 | |||
1697 | /* | 1563 | /* |
1698 | * Try to release a page associated with block device when the system | 1564 | * Try to release a page associated with block device when the system |
1699 | * is under memory pressure. | 1565 | * is under memory pressure. |
@@ -1724,16 +1590,16 @@ const struct file_operations def_blk_fops = { | |||
1724 | .llseek = block_llseek, | 1590 | .llseek = block_llseek, |
1725 | .read = do_sync_read, | 1591 | .read = do_sync_read, |
1726 | .write = do_sync_write, | 1592 | .write = do_sync_write, |
1727 | .aio_read = blkdev_aio_read, | 1593 | .aio_read = blkdev_aio_read, |
1728 | .aio_write = blkdev_aio_write, | 1594 | .aio_write = blkdev_aio_write, |
1729 | .mmap = blkdev_mmap, | 1595 | .mmap = generic_file_mmap, |
1730 | .fsync = blkdev_fsync, | 1596 | .fsync = blkdev_fsync, |
1731 | .unlocked_ioctl = block_ioctl, | 1597 | .unlocked_ioctl = block_ioctl, |
1732 | #ifdef CONFIG_COMPAT | 1598 | #ifdef CONFIG_COMPAT |
1733 | .compat_ioctl = compat_blkdev_ioctl, | 1599 | .compat_ioctl = compat_blkdev_ioctl, |
1734 | #endif | 1600 | #endif |
1735 | .splice_read = blkdev_splice_read, | 1601 | .splice_read = generic_file_splice_read, |
1736 | .splice_write = blkdev_splice_write, | 1602 | .splice_write = generic_file_splice_write, |
1737 | }; | 1603 | }; |
1738 | 1604 | ||
1739 | int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg) | 1605 | int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg) |