linux/fs/block_dev.c

1668 lines
41 KiB
C
Raw Normal View History

/*
* linux/fs/block_dev.c
*
* Copyright (C) 1991, 1992 Linus Torvalds
* Copyright (C) 2001 Andrea Arcangeli <andrea@suse.de> SuSE
*/
#include <linux/init.h>
#include <linux/mm.h>
#include <linux/fcntl.h>
#include <linux/slab.h>
#include <linux/kmod.h>
#include <linux/major.h>
#include <linux/device_cgroup.h>
#include <linux/highmem.h>
#include <linux/blkdev.h>
#include <linux/module.h>
#include <linux/blkpg.h>
#include <linux/buffer_head.h>
#include <linux/pagevec.h>
#include <linux/writeback.h>
#include <linux/mpage.h>
#include <linux/mount.h>
#include <linux/uio.h>
#include <linux/namei.h>
#include <linux/log2.h>
#include <linux/kmemleak.h>
#include <asm/uaccess.h>
#include "internal.h"
struct bdev_inode {
struct block_device bdev;
struct inode vfs_inode;
};
static const struct address_space_operations def_blk_aops;
static inline struct bdev_inode *BDEV_I(struct inode *inode)
{
return container_of(inode, struct bdev_inode, vfs_inode);
}
inline struct block_device *I_BDEV(struct inode *inode)
{
return &BDEV_I(inode)->bdev;
}
EXPORT_SYMBOL(I_BDEV);
/*
writeback: split inode_wb_list_lock into bdi_writeback.list_lock Split the global inode_wb_list_lock into a per-bdi_writeback list_lock, as it's currently the most contended lock in the system for metadata heavy workloads. It won't help for single-filesystem workloads for which we'll need the I/O-less balance_dirty_pages, but at least we can dedicate a cpu to spinning on each bdi now for larger systems. Based on earlier patches from Nick Piggin and Dave Chinner. It reduces lock contentions to 1/4 in this test case: 10 HDD JBOD, 100 dd on each disk, XFS, 6GB ram lock_stat version 0.3 ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- class name con-bounces contentions waittime-min waittime-max waittime-total acq-bounces acquisitions holdtime-min holdtime-max holdtime-total ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- vanilla 2.6.39-rc3: inode_wb_list_lock: 42590 44433 0.12 147.74 144127.35 252274 886792 0.08 121.34 917211.23 ------------------ inode_wb_list_lock 2 [<ffffffff81165da5>] bdev_inode_switch_bdi+0x29/0x85 inode_wb_list_lock 34 [<ffffffff8115bd0b>] inode_wb_list_del+0x22/0x49 inode_wb_list_lock 12893 [<ffffffff8115bb53>] __mark_inode_dirty+0x170/0x1d0 inode_wb_list_lock 10702 [<ffffffff8115afef>] writeback_single_inode+0x16d/0x20a ------------------ inode_wb_list_lock 2 [<ffffffff81165da5>] bdev_inode_switch_bdi+0x29/0x85 inode_wb_list_lock 19 [<ffffffff8115bd0b>] inode_wb_list_del+0x22/0x49 inode_wb_list_lock 5550 [<ffffffff8115bb53>] __mark_inode_dirty+0x170/0x1d0 inode_wb_list_lock 8511 [<ffffffff8115b4ad>] writeback_sb_inodes+0x10f/0x157 2.6.39-rc3 + patch: &(&wb->list_lock)->rlock: 11383 11657 0.14 151.69 40429.51 90825 527918 0.11 145.90 556843.37 ------------------------ &(&wb->list_lock)->rlock 10 [<ffffffff8115b189>] inode_wb_list_del+0x5f/0x86 &(&wb->list_lock)->rlock 1493 [<ffffffff8115b1ed>] writeback_inodes_wb+0x3d/0x150 &(&wb->list_lock)->rlock 3652 [<ffffffff8115a8e9>] writeback_sb_inodes+0x123/0x16f &(&wb->list_lock)->rlock 1412 [<ffffffff8115a38e>] writeback_single_inode+0x17f/0x223 ------------------------ &(&wb->list_lock)->rlock 3 [<ffffffff8110b5af>] bdi_lock_two+0x46/0x4b &(&wb->list_lock)->rlock 6 [<ffffffff8115b189>] inode_wb_list_del+0x5f/0x86 &(&wb->list_lock)->rlock 2061 [<ffffffff8115af97>] __mark_inode_dirty+0x173/0x1cf &(&wb->list_lock)->rlock 2629 [<ffffffff8115a8e9>] writeback_sb_inodes+0x123/0x16f hughd@google.com: fix recursive lock when bdi_lock_two() is called with new the same as old akpm@linux-foundation.org: cleanup bdev_inode_switch_bdi() comment Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
2011-04-22 02:19:44 +02:00
* Move the inode from its current bdi to a new bdi. If the inode is dirty we
* need to move it onto the dirty list of @dst so that the inode is always on
* the right list.
*/
static void bdev_inode_switch_bdi(struct inode *inode,
struct backing_dev_info *dst)
{
writeback: split inode_wb_list_lock into bdi_writeback.list_lock Split the global inode_wb_list_lock into a per-bdi_writeback list_lock, as it's currently the most contended lock in the system for metadata heavy workloads. It won't help for single-filesystem workloads for which we'll need the I/O-less balance_dirty_pages, but at least we can dedicate a cpu to spinning on each bdi now for larger systems. Based on earlier patches from Nick Piggin and Dave Chinner. It reduces lock contentions to 1/4 in this test case: 10 HDD JBOD, 100 dd on each disk, XFS, 6GB ram lock_stat version 0.3 ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- class name con-bounces contentions waittime-min waittime-max waittime-total acq-bounces acquisitions holdtime-min holdtime-max holdtime-total ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- vanilla 2.6.39-rc3: inode_wb_list_lock: 42590 44433 0.12 147.74 144127.35 252274 886792 0.08 121.34 917211.23 ------------------ inode_wb_list_lock 2 [<ffffffff81165da5>] bdev_inode_switch_bdi+0x29/0x85 inode_wb_list_lock 34 [<ffffffff8115bd0b>] inode_wb_list_del+0x22/0x49 inode_wb_list_lock 12893 [<ffffffff8115bb53>] __mark_inode_dirty+0x170/0x1d0 inode_wb_list_lock 10702 [<ffffffff8115afef>] writeback_single_inode+0x16d/0x20a ------------------ inode_wb_list_lock 2 [<ffffffff81165da5>] bdev_inode_switch_bdi+0x29/0x85 inode_wb_list_lock 19 [<ffffffff8115bd0b>] inode_wb_list_del+0x22/0x49 inode_wb_list_lock 5550 [<ffffffff8115bb53>] __mark_inode_dirty+0x170/0x1d0 inode_wb_list_lock 8511 [<ffffffff8115b4ad>] writeback_sb_inodes+0x10f/0x157 2.6.39-rc3 + patch: &(&wb->list_lock)->rlock: 11383 11657 0.14 151.69 40429.51 90825 527918 0.11 145.90 556843.37 ------------------------ &(&wb->list_lock)->rlock 10 [<ffffffff8115b189>] inode_wb_list_del+0x5f/0x86 &(&wb->list_lock)->rlock 1493 [<ffffffff8115b1ed>] writeback_inodes_wb+0x3d/0x150 &(&wb->list_lock)->rlock 3652 [<ffffffff8115a8e9>] writeback_sb_inodes+0x123/0x16f &(&wb->list_lock)->rlock 1412 [<ffffffff8115a38e>] writeback_single_inode+0x17f/0x223 ------------------------ &(&wb->list_lock)->rlock 3 [<ffffffff8110b5af>] bdi_lock_two+0x46/0x4b &(&wb->list_lock)->rlock 6 [<ffffffff8115b189>] inode_wb_list_del+0x5f/0x86 &(&wb->list_lock)->rlock 2061 [<ffffffff8115af97>] __mark_inode_dirty+0x173/0x1cf &(&wb->list_lock)->rlock 2629 [<ffffffff8115a8e9>] writeback_sb_inodes+0x123/0x16f hughd@google.com: fix recursive lock when bdi_lock_two() is called with new the same as old akpm@linux-foundation.org: cleanup bdev_inode_switch_bdi() comment Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
2011-04-22 02:19:44 +02:00
struct backing_dev_info *old = inode->i_data.backing_dev_info;
if (unlikely(dst == old)) /* deadlock avoidance */
return;
bdi_lock_two(&old->wb, &dst->wb);
spin_lock(&inode->i_lock);
inode->i_data.backing_dev_info = dst;
if (inode->i_state & I_DIRTY)
list_move(&inode->i_wb_list, &dst->wb.b_dirty);
spin_unlock(&inode->i_lock);
writeback: split inode_wb_list_lock into bdi_writeback.list_lock Split the global inode_wb_list_lock into a per-bdi_writeback list_lock, as it's currently the most contended lock in the system for metadata heavy workloads. It won't help for single-filesystem workloads for which we'll need the I/O-less balance_dirty_pages, but at least we can dedicate a cpu to spinning on each bdi now for larger systems. Based on earlier patches from Nick Piggin and Dave Chinner. It reduces lock contentions to 1/4 in this test case: 10 HDD JBOD, 100 dd on each disk, XFS, 6GB ram lock_stat version 0.3 ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- class name con-bounces contentions waittime-min waittime-max waittime-total acq-bounces acquisitions holdtime-min holdtime-max holdtime-total ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- vanilla 2.6.39-rc3: inode_wb_list_lock: 42590 44433 0.12 147.74 144127.35 252274 886792 0.08 121.34 917211.23 ------------------ inode_wb_list_lock 2 [<ffffffff81165da5>] bdev_inode_switch_bdi+0x29/0x85 inode_wb_list_lock 34 [<ffffffff8115bd0b>] inode_wb_list_del+0x22/0x49 inode_wb_list_lock 12893 [<ffffffff8115bb53>] __mark_inode_dirty+0x170/0x1d0 inode_wb_list_lock 10702 [<ffffffff8115afef>] writeback_single_inode+0x16d/0x20a ------------------ inode_wb_list_lock 2 [<ffffffff81165da5>] bdev_inode_switch_bdi+0x29/0x85 inode_wb_list_lock 19 [<ffffffff8115bd0b>] inode_wb_list_del+0x22/0x49 inode_wb_list_lock 5550 [<ffffffff8115bb53>] __mark_inode_dirty+0x170/0x1d0 inode_wb_list_lock 8511 [<ffffffff8115b4ad>] writeback_sb_inodes+0x10f/0x157 2.6.39-rc3 + patch: &(&wb->list_lock)->rlock: 11383 11657 0.14 151.69 40429.51 90825 527918 0.11 145.90 556843.37 ------------------------ &(&wb->list_lock)->rlock 10 [<ffffffff8115b189>] inode_wb_list_del+0x5f/0x86 &(&wb->list_lock)->rlock 1493 [<ffffffff8115b1ed>] writeback_inodes_wb+0x3d/0x150 &(&wb->list_lock)->rlock 3652 [<ffffffff8115a8e9>] writeback_sb_inodes+0x123/0x16f &(&wb->list_lock)->rlock 1412 [<ffffffff8115a38e>] writeback_single_inode+0x17f/0x223 ------------------------ &(&wb->list_lock)->rlock 3 [<ffffffff8110b5af>] bdi_lock_two+0x46/0x4b &(&wb->list_lock)->rlock 6 [<ffffffff8115b189>] inode_wb_list_del+0x5f/0x86 &(&wb->list_lock)->rlock 2061 [<ffffffff8115af97>] __mark_inode_dirty+0x173/0x1cf &(&wb->list_lock)->rlock 2629 [<ffffffff8115a8e9>] writeback_sb_inodes+0x123/0x16f hughd@google.com: fix recursive lock when bdi_lock_two() is called with new the same as old akpm@linux-foundation.org: cleanup bdev_inode_switch_bdi() comment Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
2011-04-22 02:19:44 +02:00
spin_unlock(&old->wb.list_lock);
spin_unlock(&dst->wb.list_lock);
}
static sector_t max_block(struct block_device *bdev)
{
sector_t retval = ~((sector_t)0);
loff_t sz = i_size_read(bdev->bd_inode);
if (sz) {
unsigned int size = block_size(bdev);
unsigned int sizebits = blksize_bits(size);
retval = (sz >> sizebits);
}
return retval;
}
/* Kill _all_ buffers and pagecache , dirty or not.. */
static void kill_bdev(struct block_device *bdev)
{
if (bdev->bd_inode->i_mapping->nrpages == 0)
return;
invalidate_bh_lrus();
truncate_inode_pages(bdev->bd_inode->i_mapping, 0);
}
int set_blocksize(struct block_device *bdev, int size)
{
/* Size must be a power of two, and between 512 and PAGE_SIZE */
if (size > PAGE_SIZE || size < 512 || !is_power_of_2(size))
return -EINVAL;
/* Size cannot be smaller than the size supported by the device */
if (size < bdev_logical_block_size(bdev))
return -EINVAL;
/* Don't change the size if it is same as current */
if (bdev->bd_block_size != size) {
sync_blockdev(bdev);
bdev->bd_block_size = size;
bdev->bd_inode->i_blkbits = blksize_bits(size);
kill_bdev(bdev);
}
return 0;
}
EXPORT_SYMBOL(set_blocksize);
int sb_set_blocksize(struct super_block *sb, int size)
{
if (set_blocksize(sb->s_bdev, size))
return 0;
/* If we get here, we know size is power of two
* and it's value is between 512 and PAGE_SIZE */
sb->s_blocksize = size;
sb->s_blocksize_bits = blksize_bits(size);
return sb->s_blocksize;
}
EXPORT_SYMBOL(sb_set_blocksize);
int sb_min_blocksize(struct super_block *sb, int size)
{
int minsize = bdev_logical_block_size(sb->s_bdev);
if (size < minsize)
size = minsize;
return sb_set_blocksize(sb, size);
}
EXPORT_SYMBOL(sb_min_blocksize);
static int
blkdev_get_block(struct inode *inode, sector_t iblock,
struct buffer_head *bh, int create)
{
if (iblock >= max_block(I_BDEV(inode))) {
if (create)
return -EIO;
/*
* for reads, we're just trying to fill a partial page.
* return a hole, they will have to call get_block again
* before they can fill it, and they will get -EIO at that
* time
*/
return 0;
}
bh->b_bdev = I_BDEV(inode);
bh->b_blocknr = iblock;
set_buffer_mapped(bh);
return 0;
}
static int
blkdev_get_blocks(struct inode *inode, sector_t iblock,
struct buffer_head *bh, int create)
{
sector_t end_block = max_block(I_BDEV(inode));
unsigned long max_blocks = bh->b_size >> inode->i_blkbits;
if ((iblock + max_blocks) > end_block) {
max_blocks = end_block - iblock;
if ((long)max_blocks <= 0) {
if (create)
return -EIO; /* write fully beyond EOF */
/*
* It is a read which is fully beyond EOF. We return
* a !buffer_mapped buffer
*/
max_blocks = 0;
}
}
bh->b_bdev = I_BDEV(inode);
bh->b_blocknr = iblock;
bh->b_size = max_blocks << inode->i_blkbits;
if (max_blocks)
set_buffer_mapped(bh);
return 0;
}
static ssize_t
blkdev_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
loff_t offset, unsigned long nr_segs)
{
struct file *file = iocb->ki_filp;
struct inode *inode = file->f_mapping->host;
return __blockdev_direct_IO(rw, iocb, inode, I_BDEV(inode), iov, offset,
nr_segs, blkdev_get_blocks, NULL, NULL, 0);
}
int __sync_blockdev(struct block_device *bdev, int wait)
{
if (!bdev)
return 0;
if (!wait)
return filemap_flush(bdev->bd_inode->i_mapping);
return filemap_write_and_wait(bdev->bd_inode->i_mapping);
}
/*
* Write out and wait upon all the dirty data associated with a block
* device via its mapping. Does not take the superblock lock.
*/
int sync_blockdev(struct block_device *bdev)
{
return __sync_blockdev(bdev, 1);
}
EXPORT_SYMBOL(sync_blockdev);
/*
* Write out and wait upon all dirty data associated with this
* device. Filesystem data as well as the underlying block
* device. Takes the superblock lock.
*/
int fsync_bdev(struct block_device *bdev)
{
struct super_block *sb = get_super(bdev);
if (sb) {
int res = sync_filesystem(sb);
drop_super(sb);
return res;
}
return sync_blockdev(bdev);
}
EXPORT_SYMBOL(fsync_bdev);
/**
* freeze_bdev -- lock a filesystem and force it into a consistent state
* @bdev: blockdevice to lock
*
* If a superblock is found on this device, we take the s_umount semaphore
* on it to make sure nobody unmounts until the snapshot creation is done.
* The reference counter (bd_fsfreeze_count) guarantees that only the last
* unfreeze process can unfreeze the frozen filesystem actually when multiple
* freeze requests arrive simultaneously. It counts up in freeze_bdev() and
* count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze
* actually.
*/
struct super_block *freeze_bdev(struct block_device *bdev)
{
struct super_block *sb;
int error = 0;
mutex_lock(&bdev->bd_fsfreeze_mutex);
if (++bdev->bd_fsfreeze_count > 1) {
/*
* We don't even need to grab a reference - the first call
* to freeze_bdev grab an active reference and only the last
* thaw_bdev drops it.
*/
sb = get_super(bdev);
drop_super(sb);
mutex_unlock(&bdev->bd_fsfreeze_mutex);
return sb;
}
sb = get_active_super(bdev);
if (!sb)
goto out;
error = freeze_super(sb);
if (error) {
deactivate_super(sb);
bdev->bd_fsfreeze_count--;
mutex_unlock(&bdev->bd_fsfreeze_mutex);
return ERR_PTR(error);
}
deactivate_super(sb);
out:
sync_blockdev(bdev);
mutex_unlock(&bdev->bd_fsfreeze_mutex);
return sb; /* thaw_bdev releases s->s_umount */
}
EXPORT_SYMBOL(freeze_bdev);
/**
* thaw_bdev -- unlock filesystem
* @bdev: blockdevice to unlock
* @sb: associated superblock
*
* Unlocks the filesystem and marks it writeable again after freeze_bdev().
*/
int thaw_bdev(struct block_device *bdev, struct super_block *sb)
{
int error = -EINVAL;
mutex_lock(&bdev->bd_fsfreeze_mutex);
if (!bdev->bd_fsfreeze_count)
goto out;
error = 0;
if (--bdev->bd_fsfreeze_count > 0)
goto out;
if (!sb)
goto out;
error = thaw_super(sb);
if (error) {
bdev->bd_fsfreeze_count++;
mutex_unlock(&bdev->bd_fsfreeze_mutex);
return error;
}
out:
mutex_unlock(&bdev->bd_fsfreeze_mutex);
return 0;
}
EXPORT_SYMBOL(thaw_bdev);
static int blkdev_writepage(struct page *page, struct writeback_control *wbc)
{
return block_write_full_page(page, blkdev_get_block, wbc);
}
static int blkdev_readpage(struct file * file, struct page * page)
{
return block_read_full_page(page, blkdev_get_block);
}
static int blkdev_write_begin(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned flags,
struct page **pagep, void **fsdata)
{
return block_write_begin(mapping, pos, len, flags, pagep,
blkdev_get_block);
}
static int blkdev_write_end(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned copied,
struct page *page, void *fsdata)
{
int ret;
ret = block_write_end(file, mapping, pos, len, copied, page, fsdata);
unlock_page(page);
page_cache_release(page);
return ret;
}
/*
* private llseek:
* for a block special file file->f_path.dentry->d_inode->i_size is zero
* so we compute the size by hand (just as in block_read/write above)
*/
static loff_t block_llseek(struct file *file, loff_t offset, int origin)
{
struct inode *bd_inode = file->f_mapping->host;
loff_t size;
loff_t retval;
mutex_lock(&bd_inode->i_mutex);
size = i_size_read(bd_inode);
retval = -EINVAL;
switch (origin) {
case SEEK_END:
offset += size;
break;
case SEEK_CUR:
offset += file->f_pos;
case SEEK_SET:
break;
default:
goto out;
}
if (offset >= 0 && offset <= size) {
if (offset != file->f_pos) {
file->f_pos = offset;
}
retval = offset;
}
out:
mutex_unlock(&bd_inode->i_mutex);
return retval;
}
int blkdev_fsync(struct file *filp, loff_t start, loff_t end, int datasync)
{
fs/block_dev.c: fix performance regression in O_DIRECT|O_SYNC writes to block devices We are seeing a large regression in database performance on recent kernels. The database opens a block device with O_DIRECT|O_SYNC and a number of threads write to different regions of the file at the same time. A simple test case is below. I haven't defined DEVICE since getting it wrong will destroy your data :) On an 3 disk LVM with a 64k chunk size we see about 17MB/sec and only a few threads in IO wait: procs -----io---- -system-- -----cpu------ r b bi bo in cs us sy id wa st 0 3 0 16170 656 2259 0 0 86 14 0 0 2 0 16704 695 2408 0 0 92 8 0 0 2 0 17308 744 2653 0 0 86 14 0 0 2 0 17933 759 2777 0 0 89 10 0 Most threads are blocking in vfs_fsync_range, which has: mutex_lock(&mapping->host->i_mutex); err = fop->fsync(file, dentry, datasync); if (!ret) ret = err; mutex_unlock(&mapping->host->i_mutex); commit 148f948ba877f4d3cdef036b1ff6d9f68986706a (vfs: Introduce new helpers for syncing after writing to O_SYNC file or IS_SYNC inode) offers some explanation of what is going on: Use these new helpers for syncing from generic VFS functions. This makes O_SYNC writes to block devices acquire i_mutex for syncing. If we really care about this, we can make block_fsync() drop the i_mutex and reacquire it before it returns. Thanks Jan for such a good commit message! As well as dropping i_mutex, Christoph suggests we should remove the call to sync_blockdev(): > sync_blockdev is an overcomplicated alias for filemap_write_and_wait on > the block device inode, which is exactly what we did just before calling > into ->fsync The patch below incorporates both suggestions. With it the testcase improves from 17MB/s to 68M/sec: procs -----io---- -system-- -----cpu------ r b bi bo in cs us sy id wa st 0 7 0 65536 1000 3878 0 0 70 30 0 0 34 0 69632 1016 3921 0 1 46 53 0 0 57 0 69632 1000 3921 0 0 55 45 0 0 53 0 69640 754 4111 0 0 81 19 0 Testcase: #define _GNU_SOURCE #include <stdio.h> #include <pthread.h> #include <unistd.h> #include <stdlib.h> #include <string.h> #include <sys/types.h> #include <sys/stat.h> #include <fcntl.h> #define NR_THREADS 64 #define BUFSIZE (64 * 1024) #define DEVICE "/dev/mapper/XXXXXX" #define ALIGN(VAL, SIZE) (((VAL)+(SIZE)-1) & ~((SIZE)-1)) static int fd; static void *doit(void *arg) { unsigned long offset = (long)arg; char *b, *buf; b = malloc(BUFSIZE + 1024); buf = (char *)ALIGN((unsigned long)b, 1024); memset(buf, 0, BUFSIZE); while (1) pwrite(fd, buf, BUFSIZE, offset); } int main(int argc, char *argv[]) { int flags = O_RDWR|O_DIRECT; int i; unsigned long offset = 0; if (argc > 1 && !strcmp(argv[1], "O_SYNC")) flags |= O_SYNC; fd = open(DEVICE, flags); if (fd == -1) { perror("open"); exit(1); } for (i = 0; i < NR_THREADS-1; i++) { pthread_t tid; pthread_create(&tid, NULL, doit, (void *)offset); offset += BUFSIZE; } doit((void *)offset); return 0; } Signed-off-by: Anton Blanchard <anton@samba.org> Acked-by: Jan Kara <jack@suse.cz> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Jens Axboe <jens.axboe@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-04-23 19:18:06 +02:00
struct inode *bd_inode = filp->f_mapping->host;
struct block_device *bdev = I_BDEV(bd_inode);
int error;
fs/block_dev.c: fix performance regression in O_DIRECT|O_SYNC writes to block devices We are seeing a large regression in database performance on recent kernels. The database opens a block device with O_DIRECT|O_SYNC and a number of threads write to different regions of the file at the same time. A simple test case is below. I haven't defined DEVICE since getting it wrong will destroy your data :) On an 3 disk LVM with a 64k chunk size we see about 17MB/sec and only a few threads in IO wait: procs -----io---- -system-- -----cpu------ r b bi bo in cs us sy id wa st 0 3 0 16170 656 2259 0 0 86 14 0 0 2 0 16704 695 2408 0 0 92 8 0 0 2 0 17308 744 2653 0 0 86 14 0 0 2 0 17933 759 2777 0 0 89 10 0 Most threads are blocking in vfs_fsync_range, which has: mutex_lock(&mapping->host->i_mutex); err = fop->fsync(file, dentry, datasync); if (!ret) ret = err; mutex_unlock(&mapping->host->i_mutex); commit 148f948ba877f4d3cdef036b1ff6d9f68986706a (vfs: Introduce new helpers for syncing after writing to O_SYNC file or IS_SYNC inode) offers some explanation of what is going on: Use these new helpers for syncing from generic VFS functions. This makes O_SYNC writes to block devices acquire i_mutex for syncing. If we really care about this, we can make block_fsync() drop the i_mutex and reacquire it before it returns. Thanks Jan for such a good commit message! As well as dropping i_mutex, Christoph suggests we should remove the call to sync_blockdev(): > sync_blockdev is an overcomplicated alias for filemap_write_and_wait on > the block device inode, which is exactly what we did just before calling > into ->fsync The patch below incorporates both suggestions. With it the testcase improves from 17MB/s to 68M/sec: procs -----io---- -system-- -----cpu------ r b bi bo in cs us sy id wa st 0 7 0 65536 1000 3878 0 0 70 30 0 0 34 0 69632 1016 3921 0 1 46 53 0 0 57 0 69632 1000 3921 0 0 55 45 0 0 53 0 69640 754 4111 0 0 81 19 0 Testcase: #define _GNU_SOURCE #include <stdio.h> #include <pthread.h> #include <unistd.h> #include <stdlib.h> #include <string.h> #include <sys/types.h> #include <sys/stat.h> #include <fcntl.h> #define NR_THREADS 64 #define BUFSIZE (64 * 1024) #define DEVICE "/dev/mapper/XXXXXX" #define ALIGN(VAL, SIZE) (((VAL)+(SIZE)-1) & ~((SIZE)-1)) static int fd; static void *doit(void *arg) { unsigned long offset = (long)arg; char *b, *buf; b = malloc(BUFSIZE + 1024); buf = (char *)ALIGN((unsigned long)b, 1024); memset(buf, 0, BUFSIZE); while (1) pwrite(fd, buf, BUFSIZE, offset); } int main(int argc, char *argv[]) { int flags = O_RDWR|O_DIRECT; int i; unsigned long offset = 0; if (argc > 1 && !strcmp(argv[1], "O_SYNC")) flags |= O_SYNC; fd = open(DEVICE, flags); if (fd == -1) { perror("open"); exit(1); } for (i = 0; i < NR_THREADS-1; i++) { pthread_t tid; pthread_create(&tid, NULL, doit, (void *)offset); offset += BUFSIZE; } doit((void *)offset); return 0; } Signed-off-by: Anton Blanchard <anton@samba.org> Acked-by: Jan Kara <jack@suse.cz> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Jens Axboe <jens.axboe@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-04-23 19:18:06 +02:00
/*
* There is no need to serialise calls to blkdev_issue_flush with
* i_mutex and doing so causes performance issues with concurrent
* O_SYNC writers to a block device.
*/
error = blkdev_issue_flush(bdev, GFP_KERNEL, NULL);
if (error == -EOPNOTSUPP)
error = 0;
fs/block_dev.c: fix performance regression in O_DIRECT|O_SYNC writes to block devices We are seeing a large regression in database performance on recent kernels. The database opens a block device with O_DIRECT|O_SYNC and a number of threads write to different regions of the file at the same time. A simple test case is below. I haven't defined DEVICE since getting it wrong will destroy your data :) On an 3 disk LVM with a 64k chunk size we see about 17MB/sec and only a few threads in IO wait: procs -----io---- -system-- -----cpu------ r b bi bo in cs us sy id wa st 0 3 0 16170 656 2259 0 0 86 14 0 0 2 0 16704 695 2408 0 0 92 8 0 0 2 0 17308 744 2653 0 0 86 14 0 0 2 0 17933 759 2777 0 0 89 10 0 Most threads are blocking in vfs_fsync_range, which has: mutex_lock(&mapping->host->i_mutex); err = fop->fsync(file, dentry, datasync); if (!ret) ret = err; mutex_unlock(&mapping->host->i_mutex); commit 148f948ba877f4d3cdef036b1ff6d9f68986706a (vfs: Introduce new helpers for syncing after writing to O_SYNC file or IS_SYNC inode) offers some explanation of what is going on: Use these new helpers for syncing from generic VFS functions. This makes O_SYNC writes to block devices acquire i_mutex for syncing. If we really care about this, we can make block_fsync() drop the i_mutex and reacquire it before it returns. Thanks Jan for such a good commit message! As well as dropping i_mutex, Christoph suggests we should remove the call to sync_blockdev(): > sync_blockdev is an overcomplicated alias for filemap_write_and_wait on > the block device inode, which is exactly what we did just before calling > into ->fsync The patch below incorporates both suggestions. With it the testcase improves from 17MB/s to 68M/sec: procs -----io---- -system-- -----cpu------ r b bi bo in cs us sy id wa st 0 7 0 65536 1000 3878 0 0 70 30 0 0 34 0 69632 1016 3921 0 1 46 53 0 0 57 0 69632 1000 3921 0 0 55 45 0 0 53 0 69640 754 4111 0 0 81 19 0 Testcase: #define _GNU_SOURCE #include <stdio.h> #include <pthread.h> #include <unistd.h> #include <stdlib.h> #include <string.h> #include <sys/types.h> #include <sys/stat.h> #include <fcntl.h> #define NR_THREADS 64 #define BUFSIZE (64 * 1024) #define DEVICE "/dev/mapper/XXXXXX" #define ALIGN(VAL, SIZE) (((VAL)+(SIZE)-1) & ~((SIZE)-1)) static int fd; static void *doit(void *arg) { unsigned long offset = (long)arg; char *b, *buf; b = malloc(BUFSIZE + 1024); buf = (char *)ALIGN((unsigned long)b, 1024); memset(buf, 0, BUFSIZE); while (1) pwrite(fd, buf, BUFSIZE, offset); } int main(int argc, char *argv[]) { int flags = O_RDWR|O_DIRECT; int i; unsigned long offset = 0; if (argc > 1 && !strcmp(argv[1], "O_SYNC")) flags |= O_SYNC; fd = open(DEVICE, flags); if (fd == -1) { perror("open"); exit(1); } for (i = 0; i < NR_THREADS-1; i++) { pthread_t tid; pthread_create(&tid, NULL, doit, (void *)offset); offset += BUFSIZE; } doit((void *)offset); return 0; } Signed-off-by: Anton Blanchard <anton@samba.org> Acked-by: Jan Kara <jack@suse.cz> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Jens Axboe <jens.axboe@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-04-23 19:18:06 +02:00
return error;
}
EXPORT_SYMBOL(blkdev_fsync);
/*
* pseudo-fs
*/
static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock);
static struct kmem_cache * bdev_cachep __read_mostly;
static struct inode *bdev_alloc_inode(struct super_block *sb)
{
struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL);
if (!ei)
return NULL;
return &ei->vfs_inode;
}
2011-01-07 07:49:49 +01:00
static void bdev_i_callback(struct rcu_head *head)
{
2011-01-07 07:49:49 +01:00
struct inode *inode = container_of(head, struct inode, i_rcu);
struct bdev_inode *bdi = BDEV_I(inode);
2011-01-07 07:49:49 +01:00
INIT_LIST_HEAD(&inode->i_dentry);
kmem_cache_free(bdev_cachep, bdi);
}
2011-01-07 07:49:49 +01:00
static void bdev_destroy_inode(struct inode *inode)
{
call_rcu(&inode->i_rcu, bdev_i_callback);
}
static void init_once(void *foo)
{
struct bdev_inode *ei = (struct bdev_inode *) foo;
struct block_device *bdev = &ei->bdev;
memset(bdev, 0, sizeof(*bdev));
mutex_init(&bdev->bd_mutex);
INIT_LIST_HEAD(&bdev->bd_inodes);
INIT_LIST_HEAD(&bdev->bd_list);
#ifdef CONFIG_SYSFS
INIT_LIST_HEAD(&bdev->bd_holder_disks);
#endif
inode_init_once(&ei->vfs_inode);
/* Initialize mutex for freeze. */
mutex_init(&bdev->bd_fsfreeze_mutex);
}
static inline void __bd_forget(struct inode *inode)
{
list_del_init(&inode->i_devices);
inode->i_bdev = NULL;
inode->i_mapping = &inode->i_data;
}
static void bdev_evict_inode(struct inode *inode)
{
struct block_device *bdev = &BDEV_I(inode)->bdev;
struct list_head *p;
truncate_inode_pages(&inode->i_data, 0);
invalidate_inode_buffers(inode); /* is it needed here? */
end_writeback(inode);
spin_lock(&bdev_lock);
while ( (p = bdev->bd_inodes.next) != &bdev->bd_inodes ) {
__bd_forget(list_entry(p, struct inode, i_devices));
}
list_del_init(&bdev->bd_list);
spin_unlock(&bdev_lock);
}
static const struct super_operations bdev_sops = {
.statfs = simple_statfs,
.alloc_inode = bdev_alloc_inode,
.destroy_inode = bdev_destroy_inode,
.drop_inode = generic_delete_inode,
.evict_inode = bdev_evict_inode,
};
static struct dentry *bd_mount(struct file_system_type *fs_type,
int flags, const char *dev_name, void *data)
{
return mount_pseudo(fs_type, "bdev:", &bdev_sops, NULL, 0x62646576);
}
static struct file_system_type bd_type = {
.name = "bdev",
.mount = bd_mount,
.kill_sb = kill_anon_super,
};
struct super_block *blockdev_superblock __read_mostly;
void __init bdev_cache_init(void)
{
int err;
struct vfsmount *bd_mnt;
bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode),
0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
SLAB_MEM_SPREAD|SLAB_PANIC),
init_once);
err = register_filesystem(&bd_type);
if (err)
panic("Cannot register bdev pseudo-fs");
bd_mnt = kern_mount(&bd_type);
if (IS_ERR(bd_mnt))
panic("Cannot create bdev pseudo-fs");
/*
* This vfsmount structure is only used to obtain the
* blockdev_superblock, so tell kmemleak not to report it.
*/
kmemleak_not_leak(bd_mnt);
blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */
}
/*
* Most likely _very_ bad one - but then it's hardly critical for small
* /dev and can be fixed when somebody will need really large one.
* Keep in mind that it will be fed through icache hash function too.
*/
static inline unsigned long hash(dev_t dev)
{
return MAJOR(dev)+MINOR(dev);
}
static int bdev_test(struct inode *inode, void *data)
{
return BDEV_I(inode)->bdev.bd_dev == *(dev_t *)data;
}
static int bdev_set(struct inode *inode, void *data)
{
BDEV_I(inode)->bdev.bd_dev = *(dev_t *)data;
return 0;
}
static LIST_HEAD(all_bdevs);
struct block_device *bdget(dev_t dev)
{
struct block_device *bdev;
struct inode *inode;
inode = iget5_locked(blockdev_superblock, hash(dev),
bdev_test, bdev_set, &dev);
if (!inode)
return NULL;
bdev = &BDEV_I(inode)->bdev;
if (inode->i_state & I_NEW) {
bdev->bd_contains = NULL;
bdev->bd_inode = inode;
bdev->bd_block_size = (1 << inode->i_blkbits);
bdev->bd_part_count = 0;
bdev->bd_invalidated = 0;
inode->i_mode = S_IFBLK;
inode->i_rdev = dev;
inode->i_bdev = bdev;
inode->i_data.a_ops = &def_blk_aops;
mapping_set_gfp_mask(&inode->i_data, GFP_USER);
inode->i_data.backing_dev_info = &default_backing_dev_info;
spin_lock(&bdev_lock);
list_add(&bdev->bd_list, &all_bdevs);
spin_unlock(&bdev_lock);
unlock_new_inode(inode);
}
return bdev;
}
EXPORT_SYMBOL(bdget);
/**
* bdgrab -- Grab a reference to an already referenced block device
* @bdev: Block device to grab a reference to.
*/
struct block_device *bdgrab(struct block_device *bdev)
{
ihold(bdev->bd_inode);
return bdev;
}
long nr_blockdev_pages(void)
{
struct block_device *bdev;
long ret = 0;
spin_lock(&bdev_lock);
list_for_each_entry(bdev, &all_bdevs, bd_list) {
ret += bdev->bd_inode->i_mapping->nrpages;
}
spin_unlock(&bdev_lock);
return ret;
}
void bdput(struct block_device *bdev)
{
iput(bdev->bd_inode);
}
EXPORT_SYMBOL(bdput);
static struct block_device *bd_acquire(struct inode *inode)
{
struct block_device *bdev;
spin_lock(&bdev_lock);
bdev = inode->i_bdev;
if (bdev) {
ihold(bdev->bd_inode);
spin_unlock(&bdev_lock);
return bdev;
}
spin_unlock(&bdev_lock);
bdev = bdget(inode->i_rdev);
if (bdev) {
spin_lock(&bdev_lock);
if (!inode->i_bdev) {
/*
* We take an additional reference to bd_inode,
* and it's released in clear_inode() of inode.
* So, we can access it via ->i_mapping always
* without igrab().
*/
ihold(bdev->bd_inode);
inode->i_bdev = bdev;
inode->i_mapping = bdev->bd_inode->i_mapping;
list_add(&inode->i_devices, &bdev->bd_inodes);
}
spin_unlock(&bdev_lock);
}
return bdev;
}
/* Call when you free inode */
void bd_forget(struct inode *inode)
{
struct block_device *bdev = NULL;
spin_lock(&bdev_lock);
if (inode->i_bdev) {
if (!sb_is_blkdev_sb(inode->i_sb))
bdev = inode->i_bdev;
__bd_forget(inode);
}
spin_unlock(&bdev_lock);
if (bdev)
iput(bdev->bd_inode);
}
/**
* bd_may_claim - test whether a block device can be claimed
* @bdev: block device of interest
* @whole: whole block device containing @bdev, may equal @bdev
* @holder: holder trying to claim @bdev
*
* Test whether @bdev can be claimed by @holder.
*
* CONTEXT:
* spin_lock(&bdev_lock).
*
* RETURNS:
* %true if @bdev can be claimed, %false otherwise.
*/
static bool bd_may_claim(struct block_device *bdev, struct block_device *whole,
void *holder)
{
if (bdev->bd_holder == holder)
return true; /* already a holder */
else if (bdev->bd_holder != NULL)
return false; /* held by someone else */
else if (bdev->bd_contains == bdev)
return true; /* is a whole device which isn't held */
block: make blkdev_get/put() handle exclusive access Over time, block layer has accumulated a set of APIs dealing with bdev open, close, claim and release. * blkdev_get/put() are the primary open and close functions. * bd_claim/release() deal with exclusive open. * open/close_bdev_exclusive() are combination of open and claim and the other way around, respectively. * bd_link/unlink_disk_holder() to create and remove holder/slave symlinks. * open_by_devnum() wraps bdget() + blkdev_get(). The interface is a bit confusing and the decoupling of open and claim makes it impossible to properly guarantee exclusive access as in-kernel open + claim sequence can disturb the existing exclusive open even before the block layer knows the current open if for another exclusive access. Reorganize the interface such that, * blkdev_get() is extended to include exclusive access management. @holder argument is added and, if is @FMODE_EXCL specified, it will gain exclusive access atomically w.r.t. other exclusive accesses. * blkdev_put() is similarly extended. It now takes @mode argument and if @FMODE_EXCL is set, it releases an exclusive access. Also, when the last exclusive claim is released, the holder/slave symlinks are removed automatically. * bd_claim/release() and close_bdev_exclusive() are no longer necessary and either made static or removed. * bd_link_disk_holder() remains the same but bd_unlink_disk_holder() is no longer necessary and removed. * open_bdev_exclusive() becomes a simple wrapper around lookup_bdev() and blkdev_get(). It also has an unexpected extra bdev_read_only() test which probably should be moved into blkdev_get(). * open_by_devnum() is modified to take @holder argument and pass it to blkdev_get(). Most of bdev open/close operations are unified into blkdev_get/put() and most exclusive accesses are tested atomically at the open time (as it should). This cleans up code and removes some, both valid and invalid, but unnecessary all the same, corner cases. open_bdev_exclusive() and open_by_devnum() can use further cleanup - rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop special features. Well, let's leave them for another day. Most conversions are straight-forward. drbd conversion is a bit more involved as there was some reordering, but the logic should stay the same. Signed-off-by: Tejun Heo <tj@kernel.org> Acked-by: Neil Brown <neilb@suse.de> Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Acked-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Philipp Reisner <philipp.reisner@linbit.com> Cc: Peter Osterlund <petero2@telia.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Jan Kara <jack@suse.cz> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Cc: Alex Elder <aelder@sgi.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: dm-devel@redhat.com Cc: drbd-dev@lists.linbit.com Cc: Leo Chen <leochen@broadcom.com> Cc: Scott Branden <sbranden@broadcom.com> Cc: Chris Mason <chris.mason@oracle.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@linux.vnet.ibm.com> Cc: Joern Engel <joern@logfs.org> Cc: reiserfs-devel@vger.kernel.org Cc: Alexander Viro <viro@zeniv.linux.org.uk>
2010-11-13 11:55:17 +01:00
else if (whole->bd_holder == bd_may_claim)
return true; /* is a partition of a device that is being partitioned */
else if (whole->bd_holder != NULL)
return false; /* is a partition of a held device */
else
return true; /* is a partition of an un-held device */
}
block: implement bd_claiming and claiming block Currently, device claiming for exclusive open is done after low level open - disk->fops->open() - has completed successfully. This means that exclusive open attempts while a device is already exclusively open will fail only after disk->fops->open() is called. cdrom driver issues commands during open() which means that O_EXCL open attempt can unintentionally inject commands to in-progress command stream for burning thus disturbing burning process. In most cases, this doesn't cause problems because the first command to be issued is TUR which most devices can process in the middle of burning. However, depending on how a device replies to TUR during burning, cdrom driver may end up issuing further commands. This can't be resolved trivially by moving bd_claim() before doing actual open() because that means an open attempt which will end up failing could interfere other legit O_EXCL open attempts. ie. unconfirmed open attempts can fail others. This patch resolves the problem by introducing claiming block which is started by bd_start_claiming() and terminated either by bd_claim() or bd_abort_claiming(). bd_claim() from inside a claiming block is guaranteed to succeed and once a claiming block is started, other bd_start_claiming() or bd_claim() attempts block till the current claiming block is terminated. bd_claim() can still be used standalone although now it always synchronizes against claiming blocks, so the existing users will keep working without any change. blkdev_open() and open_bdev_exclusive() are converted to use claiming blocks so that exclusive open attempts from these functions don't interfere with the existing exclusive open. This problem was discovered while investigating bko#15403. https://bugzilla.kernel.org/show_bug.cgi?id=15403 The burning problem itself can be resolved by updating userspace probing tools to always open w/ O_EXCL. Signed-off-by: Tejun Heo <tj@kernel.org> Reported-by: Matthias-Christian Ott <ott@mirix.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2010-04-07 11:53:59 +02:00
/**
* bd_prepare_to_claim - prepare to claim a block device
* @bdev: block device of interest
* @whole: the whole device containing @bdev, may equal @bdev
* @holder: holder trying to claim @bdev
*
* Prepare to claim @bdev. This function fails if @bdev is already
* claimed by another holder and waits if another claiming is in
* progress. This function doesn't actually claim. On successful
* return, the caller has ownership of bd_claiming and bd_holder[s].
*
* CONTEXT:
* spin_lock(&bdev_lock). Might release bdev_lock, sleep and regrab
* it multiple times.
*
* RETURNS:
* 0 if @bdev can be claimed, -EBUSY otherwise.
*/
static int bd_prepare_to_claim(struct block_device *bdev,
struct block_device *whole, void *holder)
{
retry:
/* if someone else claimed, fail */
if (!bd_may_claim(bdev, whole, holder))
return -EBUSY;
/* if claiming is already in progress, wait for it to finish */
if (whole->bd_claiming) {
block: implement bd_claiming and claiming block Currently, device claiming for exclusive open is done after low level open - disk->fops->open() - has completed successfully. This means that exclusive open attempts while a device is already exclusively open will fail only after disk->fops->open() is called. cdrom driver issues commands during open() which means that O_EXCL open attempt can unintentionally inject commands to in-progress command stream for burning thus disturbing burning process. In most cases, this doesn't cause problems because the first command to be issued is TUR which most devices can process in the middle of burning. However, depending on how a device replies to TUR during burning, cdrom driver may end up issuing further commands. This can't be resolved trivially by moving bd_claim() before doing actual open() because that means an open attempt which will end up failing could interfere other legit O_EXCL open attempts. ie. unconfirmed open attempts can fail others. This patch resolves the problem by introducing claiming block which is started by bd_start_claiming() and terminated either by bd_claim() or bd_abort_claiming(). bd_claim() from inside a claiming block is guaranteed to succeed and once a claiming block is started, other bd_start_claiming() or bd_claim() attempts block till the current claiming block is terminated. bd_claim() can still be used standalone although now it always synchronizes against claiming blocks, so the existing users will keep working without any change. blkdev_open() and open_bdev_exclusive() are converted to use claiming blocks so that exclusive open attempts from these functions don't interfere with the existing exclusive open. This problem was discovered while investigating bko#15403. https://bugzilla.kernel.org/show_bug.cgi?id=15403 The burning problem itself can be resolved by updating userspace probing tools to always open w/ O_EXCL. Signed-off-by: Tejun Heo <tj@kernel.org> Reported-by: Matthias-Christian Ott <ott@mirix.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2010-04-07 11:53:59 +02:00
wait_queue_head_t *wq = bit_waitqueue(&whole->bd_claiming, 0);
DEFINE_WAIT(wait);
prepare_to_wait(wq, &wait, TASK_UNINTERRUPTIBLE);
spin_unlock(&bdev_lock);
schedule();
finish_wait(wq, &wait);
spin_lock(&bdev_lock);
goto retry;
}
/* yay, all mine */
return 0;
}
/**
* bd_start_claiming - start claiming a block device
* @bdev: block device of interest
* @holder: holder trying to claim @bdev
*
* @bdev is about to be opened exclusively. Check @bdev can be opened
* exclusively and mark that an exclusive open is in progress. Each
* successful call to this function must be matched with a call to
* either bd_finish_claiming() or bd_abort_claiming() (which do not
* fail).
*
* This function is used to gain exclusive access to the block device
* without actually causing other exclusive open attempts to fail. It
* should be used when the open sequence itself requires exclusive
* access but may subsequently fail.
block: implement bd_claiming and claiming block Currently, device claiming for exclusive open is done after low level open - disk->fops->open() - has completed successfully. This means that exclusive open attempts while a device is already exclusively open will fail only after disk->fops->open() is called. cdrom driver issues commands during open() which means that O_EXCL open attempt can unintentionally inject commands to in-progress command stream for burning thus disturbing burning process. In most cases, this doesn't cause problems because the first command to be issued is TUR which most devices can process in the middle of burning. However, depending on how a device replies to TUR during burning, cdrom driver may end up issuing further commands. This can't be resolved trivially by moving bd_claim() before doing actual open() because that means an open attempt which will end up failing could interfere other legit O_EXCL open attempts. ie. unconfirmed open attempts can fail others. This patch resolves the problem by introducing claiming block which is started by bd_start_claiming() and terminated either by bd_claim() or bd_abort_claiming(). bd_claim() from inside a claiming block is guaranteed to succeed and once a claiming block is started, other bd_start_claiming() or bd_claim() attempts block till the current claiming block is terminated. bd_claim() can still be used standalone although now it always synchronizes against claiming blocks, so the existing users will keep working without any change. blkdev_open() and open_bdev_exclusive() are converted to use claiming blocks so that exclusive open attempts from these functions don't interfere with the existing exclusive open. This problem was discovered while investigating bko#15403. https://bugzilla.kernel.org/show_bug.cgi?id=15403 The burning problem itself can be resolved by updating userspace probing tools to always open w/ O_EXCL. Signed-off-by: Tejun Heo <tj@kernel.org> Reported-by: Matthias-Christian Ott <ott@mirix.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2010-04-07 11:53:59 +02:00
*
* CONTEXT:
* Might sleep.
*
* RETURNS:
* Pointer to the block device containing @bdev on success, ERR_PTR()
* value on failure.
*/
static struct block_device *bd_start_claiming(struct block_device *bdev,
void *holder)
{
struct gendisk *disk;
struct block_device *whole;
int partno, err;
might_sleep();
/*
* @bdev might not have been initialized properly yet, look up
* and grab the outer block device the hard way.
*/
disk = get_gendisk(bdev->bd_dev, &partno);
if (!disk)
return ERR_PTR(-ENXIO);
/*
* Normally, @bdev should equal what's returned from bdget_disk()
* if partno is 0; however, some drivers (floppy) use multiple
* bdev's for the same physical device and @bdev may be one of the
* aliases. Keep @bdev if partno is 0. This means claimer
* tracking is broken for those devices but it has always been that
* way.
*/
if (partno)
whole = bdget_disk(disk, 0);
else
whole = bdgrab(bdev);
module_put(disk->fops->owner);
block: implement bd_claiming and claiming block Currently, device claiming for exclusive open is done after low level open - disk->fops->open() - has completed successfully. This means that exclusive open attempts while a device is already exclusively open will fail only after disk->fops->open() is called. cdrom driver issues commands during open() which means that O_EXCL open attempt can unintentionally inject commands to in-progress command stream for burning thus disturbing burning process. In most cases, this doesn't cause problems because the first command to be issued is TUR which most devices can process in the middle of burning. However, depending on how a device replies to TUR during burning, cdrom driver may end up issuing further commands. This can't be resolved trivially by moving bd_claim() before doing actual open() because that means an open attempt which will end up failing could interfere other legit O_EXCL open attempts. ie. unconfirmed open attempts can fail others. This patch resolves the problem by introducing claiming block which is started by bd_start_claiming() and terminated either by bd_claim() or bd_abort_claiming(). bd_claim() from inside a claiming block is guaranteed to succeed and once a claiming block is started, other bd_start_claiming() or bd_claim() attempts block till the current claiming block is terminated. bd_claim() can still be used standalone although now it always synchronizes against claiming blocks, so the existing users will keep working without any change. blkdev_open() and open_bdev_exclusive() are converted to use claiming blocks so that exclusive open attempts from these functions don't interfere with the existing exclusive open. This problem was discovered while investigating bko#15403. https://bugzilla.kernel.org/show_bug.cgi?id=15403 The burning problem itself can be resolved by updating userspace probing tools to always open w/ O_EXCL. Signed-off-by: Tejun Heo <tj@kernel.org> Reported-by: Matthias-Christian Ott <ott@mirix.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2010-04-07 11:53:59 +02:00
put_disk(disk);
if (!whole)
return ERR_PTR(-ENOMEM);
/* prepare to claim, if successful, mark claiming in progress */
spin_lock(&bdev_lock);
err = bd_prepare_to_claim(bdev, whole, holder);
if (err == 0) {
whole->bd_claiming = holder;
spin_unlock(&bdev_lock);
return whole;
} else {
spin_unlock(&bdev_lock);
bdput(whole);
return ERR_PTR(err);
}
}
#ifdef CONFIG_SYSFS
struct bd_holder_disk {
struct list_head list;
struct gendisk *disk;
int refcnt;
};
static struct bd_holder_disk *bd_find_holder_disk(struct block_device *bdev,
struct gendisk *disk)
{
struct bd_holder_disk *holder;
list_for_each_entry(holder, &bdev->bd_holder_disks, list)
if (holder->disk == disk)
return holder;
return NULL;
}
[PATCH] blockdev.c: check driver layer errors Check driver layer errors. Fix from: "Jun'ichi Nomura" <j-nomura@ce.jp.nec.com> In blockdevc-check-errors.patch, add_bd_holder() is modified to return error values when some of its operation failed. Among them, it returns -EEXIST when a given bd_holder object already exists in the list. However, in this case, the function completed its work successfully and need no action by its caller other than freeing unused bd_holder object. So I think it's better to return success after freeing by itself. Otherwise, bd_claim-ing with same claim pointer will fail. Typically, lvresize will fails with following message: device-mapper: reload ioctl failed: Invalid argument and you'll see messages like below in kernel log: device-mapper: table: 254:13: linear: dm-linear: Device lookup failed device-mapper: ioctl: error adding target to table Similarly, it should not add bd_holder to the list if either one of symlinking fails. I don't have a test case for this to happen but it should cause dereference of freed pointer. If a matching bd_holder is found in bd_holder_list, add_bd_holder() completes its job by just incrementing the reference count. In this case, it should be considered as success but it used to return 'fail' to let the caller free temporary bd_holder. Fixed it to return success and free given object by itself. Also, if either one of symlinking fails, the bd_holder should not be added to the list so that it can be discarded later. Otherwise, the caller will free bd_holder which is in the list. Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Cc: "Randy.Dunlap" <rdunlap@xenotime.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-29 10:58:56 +02:00
static int add_symlink(struct kobject *from, struct kobject *to)
{
[PATCH] blockdev.c: check driver layer errors Check driver layer errors. Fix from: "Jun'ichi Nomura" <j-nomura@ce.jp.nec.com> In blockdevc-check-errors.patch, add_bd_holder() is modified to return error values when some of its operation failed. Among them, it returns -EEXIST when a given bd_holder object already exists in the list. However, in this case, the function completed its work successfully and need no action by its caller other than freeing unused bd_holder object. So I think it's better to return success after freeing by itself. Otherwise, bd_claim-ing with same claim pointer will fail. Typically, lvresize will fails with following message: device-mapper: reload ioctl failed: Invalid argument and you'll see messages like below in kernel log: device-mapper: table: 254:13: linear: dm-linear: Device lookup failed device-mapper: ioctl: error adding target to table Similarly, it should not add bd_holder to the list if either one of symlinking fails. I don't have a test case for this to happen but it should cause dereference of freed pointer. If a matching bd_holder is found in bd_holder_list, add_bd_holder() completes its job by just incrementing the reference count. In this case, it should be considered as success but it used to return 'fail' to let the caller free temporary bd_holder. Fixed it to return success and free given object by itself. Also, if either one of symlinking fails, the bd_holder should not be added to the list so that it can be discarded later. Otherwise, the caller will free bd_holder which is in the list. Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Cc: "Randy.Dunlap" <rdunlap@xenotime.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-29 10:58:56 +02:00
return sysfs_create_link(from, to, kobject_name(to));
}
static void del_symlink(struct kobject *from, struct kobject *to)
{
sysfs_remove_link(from, kobject_name(to));
}
/**
* bd_link_disk_holder - create symlinks between holding disk and slave bdev
* @bdev: the claimed slave bdev
* @disk: the holding disk
*
* DON'T USE THIS UNLESS YOU'RE ALREADY USING IT.
*
* This functions creates the following sysfs symlinks.
*
* - from "slaves" directory of the holder @disk to the claimed @bdev
* - from "holders" directory of the @bdev to the holder @disk
*
* For example, if /dev/dm-0 maps to /dev/sda and disk for dm-0 is
* passed to bd_link_disk_holder(), then:
*
* /sys/block/dm-0/slaves/sda --> /sys/block/sda
* /sys/block/sda/holders/dm-0 --> /sys/block/dm-0
*
* The caller must have claimed @bdev before calling this function and
* ensure that both @bdev and @disk are valid during the creation and
* lifetime of these symlinks.
*
* CONTEXT:
* Might sleep.
*
* RETURNS:
* 0 on success, -errno on failure.
*/
int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk)
{
struct bd_holder_disk *holder;
int ret = 0;
mutex_lock(&bdev->bd_mutex);
WARN_ON_ONCE(!bdev->bd_holder);
/* FIXME: remove the following once add_disk() handles errors */
if (WARN_ON(!disk->slave_dir || !bdev->bd_part->holder_dir))
goto out_unlock;
holder = bd_find_holder_disk(bdev, disk);
if (holder) {
holder->refcnt++;
goto out_unlock;
}
holder = kzalloc(sizeof(*holder), GFP_KERNEL);
if (!holder) {
ret = -ENOMEM;
goto out_unlock;
}
INIT_LIST_HEAD(&holder->list);
holder->disk = disk;
holder->refcnt = 1;
ret = add_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj);
if (ret)
goto out_free;
ret = add_symlink(bdev->bd_part->holder_dir, &disk_to_dev(disk)->kobj);
if (ret)
goto out_del;
/*
* bdev could be deleted beneath us which would implicitly destroy
* the holder directory. Hold on to it.
*/
kobject_get(bdev->bd_part->holder_dir);
list_add(&holder->list, &bdev->bd_holder_disks);
goto out_unlock;
out_del:
del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj);
out_free:
kfree(holder);
out_unlock:
mutex_unlock(&bdev->bd_mutex);
return ret;
}
EXPORT_SYMBOL_GPL(bd_link_disk_holder);
/**
* bd_unlink_disk_holder - destroy symlinks created by bd_link_disk_holder()
* @bdev: the calimed slave bdev
* @disk: the holding disk
*
* DON'T USE THIS UNLESS YOU'RE ALREADY USING IT.
*
* CONTEXT:
* Might sleep.
*/
void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk)
{
struct bd_holder_disk *holder;
mutex_lock(&bdev->bd_mutex);
holder = bd_find_holder_disk(bdev, disk);
if (!WARN_ON_ONCE(holder == NULL) && !--holder->refcnt) {
del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj);
del_symlink(bdev->bd_part->holder_dir,
&disk_to_dev(disk)->kobj);
kobject_put(bdev->bd_part->holder_dir);
list_del_init(&holder->list);
kfree(holder);
}
mutex_unlock(&bdev->bd_mutex);
}
EXPORT_SYMBOL_GPL(bd_unlink_disk_holder);
#endif
/**
* flush_disk - invalidates all buffer-cache entries on a disk
*
* @bdev: struct block device to be flushed
* @kill_dirty: flag to guide handling of dirty inodes
*
* Invalidates all buffer-cache entries on a disk. It should be called
* when a disk has been changed -- either by a media change or online
* resize.
*/
Fix over-zealous flush_disk when changing device size. There are two cases when we call flush_disk. In one, the device has disappeared (check_disk_change) so any data will hold becomes irrelevant. In the oter, the device has changed size (check_disk_size_change) so data we hold may be irrelevant. In both cases it makes sense to discard any 'clean' buffers, so they will be read back from the device if needed. In the former case it makes sense to discard 'dirty' buffers as there will never be anywhere safe to write the data. In the second case it *does*not* make sense to discard dirty buffers as that will lead to file system corruption when you simply enlarge the containing devices. flush_disk calls __invalidate_devices. __invalidate_device calls both invalidate_inodes and invalidate_bdev. invalidate_inodes *does* discard I_DIRTY inodes and this does lead to fs corruption. invalidate_bev *does*not* discard dirty pages, but I don't really care about that at present. So this patch adds a flag to __invalidate_device (calling it __invalidate_device2) to indicate whether dirty buffers should be killed, and this is passed to invalidate_inodes which can choose to skip dirty inodes. flusk_disk then passes true from check_disk_change and false from check_disk_size_change. dm avoids tripping over this problem by calling i_size_write directly rathher than using check_disk_size_change. md does use check_disk_size_change and so is affected. This regression was introduced by commit 608aeef17a which causes check_disk_size_change to call flush_disk, so it is suitable for any kernel since 2.6.27. Cc: stable@kernel.org Acked-by: Jeff Moyer <jmoyer@redhat.com> Cc: Andrew Patterson <andrew.patterson@hp.com> Cc: Jens Axboe <axboe@kernel.dk> Signed-off-by: NeilBrown <neilb@suse.de>
2011-02-24 07:25:47 +01:00
static void flush_disk(struct block_device *bdev, bool kill_dirty)
{
Fix over-zealous flush_disk when changing device size. There are two cases when we call flush_disk. In one, the device has disappeared (check_disk_change) so any data will hold becomes irrelevant. In the oter, the device has changed size (check_disk_size_change) so data we hold may be irrelevant. In both cases it makes sense to discard any 'clean' buffers, so they will be read back from the device if needed. In the former case it makes sense to discard 'dirty' buffers as there will never be anywhere safe to write the data. In the second case it *does*not* make sense to discard dirty buffers as that will lead to file system corruption when you simply enlarge the containing devices. flush_disk calls __invalidate_devices. __invalidate_device calls both invalidate_inodes and invalidate_bdev. invalidate_inodes *does* discard I_DIRTY inodes and this does lead to fs corruption. invalidate_bev *does*not* discard dirty pages, but I don't really care about that at present. So this patch adds a flag to __invalidate_device (calling it __invalidate_device2) to indicate whether dirty buffers should be killed, and this is passed to invalidate_inodes which can choose to skip dirty inodes. flusk_disk then passes true from check_disk_change and false from check_disk_size_change. dm avoids tripping over this problem by calling i_size_write directly rathher than using check_disk_size_change. md does use check_disk_size_change and so is affected. This regression was introduced by commit 608aeef17a which causes check_disk_size_change to call flush_disk, so it is suitable for any kernel since 2.6.27. Cc: stable@kernel.org Acked-by: Jeff Moyer <jmoyer@redhat.com> Cc: Andrew Patterson <andrew.patterson@hp.com> Cc: Jens Axboe <axboe@kernel.dk> Signed-off-by: NeilBrown <neilb@suse.de>
2011-02-24 07:25:47 +01:00
if (__invalidate_device(bdev, kill_dirty)) {
char name[BDEVNAME_SIZE] = "";
if (bdev->bd_disk)
disk_name(bdev->bd_disk, 0, name);
printk(KERN_WARNING "VFS: busy inodes on changed media or "
"resized disk %s\n", name);
}
if (!bdev->bd_disk)
return;
if (disk_partitionable(bdev->bd_disk))
bdev->bd_invalidated = 1;
}
/**
* check_disk_size_change - checks for disk size change and adjusts bdev size.
* @disk: struct gendisk to check
* @bdev: struct bdev to adjust.
*
* This routine checks to see if the bdev size does not match the disk size
* and adjusts it if it differs.
*/
void check_disk_size_change(struct gendisk *disk, struct block_device *bdev)
{
loff_t disk_size, bdev_size;
disk_size = (loff_t)get_capacity(disk) << 9;
bdev_size = i_size_read(bdev->bd_inode);
if (disk_size != bdev_size) {
char name[BDEVNAME_SIZE];
disk_name(disk, 0, name);
printk(KERN_INFO
"%s: detected capacity change from %lld to %lld\n",
name, bdev_size, disk_size);
i_size_write(bdev->bd_inode, disk_size);
Fix over-zealous flush_disk when changing device size. There are two cases when we call flush_disk. In one, the device has disappeared (check_disk_change) so any data will hold becomes irrelevant. In the oter, the device has changed size (check_disk_size_change) so data we hold may be irrelevant. In both cases it makes sense to discard any 'clean' buffers, so they will be read back from the device if needed. In the former case it makes sense to discard 'dirty' buffers as there will never be anywhere safe to write the data. In the second case it *does*not* make sense to discard dirty buffers as that will lead to file system corruption when you simply enlarge the containing devices. flush_disk calls __invalidate_devices. __invalidate_device calls both invalidate_inodes and invalidate_bdev. invalidate_inodes *does* discard I_DIRTY inodes and this does lead to fs corruption. invalidate_bev *does*not* discard dirty pages, but I don't really care about that at present. So this patch adds a flag to __invalidate_device (calling it __invalidate_device2) to indicate whether dirty buffers should be killed, and this is passed to invalidate_inodes which can choose to skip dirty inodes. flusk_disk then passes true from check_disk_change and false from check_disk_size_change. dm avoids tripping over this problem by calling i_size_write directly rathher than using check_disk_size_change. md does use check_disk_size_change and so is affected. This regression was introduced by commit 608aeef17a which causes check_disk_size_change to call flush_disk, so it is suitable for any kernel since 2.6.27. Cc: stable@kernel.org Acked-by: Jeff Moyer <jmoyer@redhat.com> Cc: Andrew Patterson <andrew.patterson@hp.com> Cc: Jens Axboe <axboe@kernel.dk> Signed-off-by: NeilBrown <neilb@suse.de>
2011-02-24 07:25:47 +01:00
flush_disk(bdev, false);
}
}
EXPORT_SYMBOL(check_disk_size_change);
/**
* revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back
* @disk: struct gendisk to be revalidated
*
* This routine is a wrapper for lower-level driver's revalidate_disk
* call-backs. It is used to do common pre and post operations needed
* for all revalidate_disk operations.
*/
int revalidate_disk(struct gendisk *disk)
{
struct block_device *bdev;
int ret = 0;
if (disk->fops->revalidate_disk)
ret = disk->fops->revalidate_disk(disk);
bdev = bdget_disk(disk, 0);
if (!bdev)
return ret;
mutex_lock(&bdev->bd_mutex);
check_disk_size_change(disk, bdev);
mutex_unlock(&bdev->bd_mutex);
bdput(bdev);
return ret;
}
EXPORT_SYMBOL(revalidate_disk);
/*
* This routine checks whether a removable media has been changed,
* and invalidates all buffer-cache-entries in that case. This
* is a relatively slow routine, so we have to try to minimize using
* it. Thus it is called only upon a 'mount' or 'open'. This
* is the best way of combining speed and utility, I think.
* People changing diskettes in the middle of an operation deserve
* to lose :-)
*/
int check_disk_change(struct block_device *bdev)
{
struct gendisk *disk = bdev->bd_disk;
const struct block_device_operations *bdops = disk->fops;
implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-12-08 20:57:37 +01:00
unsigned int events;
implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-12-08 20:57:37 +01:00
events = disk_clear_events(disk, DISK_EVENT_MEDIA_CHANGE |
DISK_EVENT_EJECT_REQUEST);
if (!(events & DISK_EVENT_MEDIA_CHANGE))
return 0;
Fix over-zealous flush_disk when changing device size. There are two cases when we call flush_disk. In one, the device has disappeared (check_disk_change) so any data will hold becomes irrelevant. In the oter, the device has changed size (check_disk_size_change) so data we hold may be irrelevant. In both cases it makes sense to discard any 'clean' buffers, so they will be read back from the device if needed. In the former case it makes sense to discard 'dirty' buffers as there will never be anywhere safe to write the data. In the second case it *does*not* make sense to discard dirty buffers as that will lead to file system corruption when you simply enlarge the containing devices. flush_disk calls __invalidate_devices. __invalidate_device calls both invalidate_inodes and invalidate_bdev. invalidate_inodes *does* discard I_DIRTY inodes and this does lead to fs corruption. invalidate_bev *does*not* discard dirty pages, but I don't really care about that at present. So this patch adds a flag to __invalidate_device (calling it __invalidate_device2) to indicate whether dirty buffers should be killed, and this is passed to invalidate_inodes which can choose to skip dirty inodes. flusk_disk then passes true from check_disk_change and false from check_disk_size_change. dm avoids tripping over this problem by calling i_size_write directly rathher than using check_disk_size_change. md does use check_disk_size_change and so is affected. This regression was introduced by commit 608aeef17a which causes check_disk_size_change to call flush_disk, so it is suitable for any kernel since 2.6.27. Cc: stable@kernel.org Acked-by: Jeff Moyer <jmoyer@redhat.com> Cc: Andrew Patterson <andrew.patterson@hp.com> Cc: Jens Axboe <axboe@kernel.dk> Signed-off-by: NeilBrown <neilb@suse.de>
2011-02-24 07:25:47 +01:00
flush_disk(bdev, true);
if (bdops->revalidate_disk)
bdops->revalidate_disk(bdev->bd_disk);
return 1;
}
EXPORT_SYMBOL(check_disk_change);
void bd_set_size(struct block_device *bdev, loff_t size)
{
unsigned bsize = bdev_logical_block_size(bdev);
bdev->bd_inode->i_size = size;
while (bsize < PAGE_CACHE_SIZE) {
if (size & bsize)
break;
bsize <<= 1;
}
bdev->bd_block_size = bsize;
bdev->bd_inode->i_blkbits = blksize_bits(bsize);
}
EXPORT_SYMBOL(bd_set_size);
static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
[PATCH] lockdep: simplify some aspects of bd_mutex nesting When we open (actually blkdev_get) a partition we need to also open (get) the whole device that holds the partition. The involves some limited recursion. This patch tries to simplify some aspects of this. As well as opening the whole device, we need to increment ->bd_part_count when a partition is opened (this is used by rescan_partitions to avoid a rescan if any partition is active, as that would be confusing). The main change this patch makes is to move the inc/dec of bd_part_count into blkdev_{get,put} for the whole rather than doing it in blkdev_{get,put} for the partition. More specifically, we introduce __blkdev_get and __blkdev_put which do exactly what blkdev_{get,put} did, only with an extra "for_part" argument (blkget_{get,put} then call the __ version with a '0' for the extra argument). If for_part is 1, then the blkdev is being get(put) because a partition is being opened(closed) for the first(last) time, and so bd_part_count should be updated (on success). The particular advantage of pushing this function down is that the bd_mutex lock (which is needed to update bd_part_count) is already held at the lower level. Note that this slightly changes the semantics of bd_part_count. Instead of updating it whenever a partition is opened or released, it is now only updated on the first open or last release. This is an adequate semantic as it is only ever tested for "== 0". Having introduced these functions we remove the current bd_part_count updates from do_open (which is really the body of blkdev_get) and call __blkdev_get(... 1). Similarly in blkget_put we remove the old bd_part_count updates and call __blkget_put(..., 1). This call is moved to the end of __blkdev_put to avoid nested locks of bd_mutex. Finally the mutex_lock on whole->bd_mutex in do_open can be removed. It was only really needed to protect bd_part_count, and that is now managed (and protected) within the recursive call. The observation that bd_part_count is central to the locking issues, and the modifications to create __blkdev_put are from Peter Zijlstra. Cc: Ingo Molnar <mingo@elte.hu> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 11:36:16 +01:00
/*
* bd_mutex locking:
*
* mutex_lock(part->bd_mutex)
* mutex_lock_nested(whole->bd_mutex, 1)
*/
static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
{
struct gendisk *disk;
int ret;
int partno;
int perm = 0;
if (mode & FMODE_READ)
perm |= MAY_READ;
if (mode & FMODE_WRITE)
perm |= MAY_WRITE;
/*
* hooks: /n/, see "layering violations".
*/
if (!for_part) {
ret = devcgroup_inode_permission(bdev->bd_inode, perm);
if (ret != 0) {
bdput(bdev);
return ret;
}
}
md: make devices disappear when they are no longer needed. Currently md devices, once created, never disappear until the module is unloaded. This is essentially because the gendisk holds a reference to the mddev, and the mddev holds a reference to the gendisk, this a circular reference. If we drop the reference from mddev to gendisk, then we need to ensure that the mddev is destroyed when the gendisk is destroyed. However it is not possible to hook into the gendisk destruction process to enable this. So we drop the reference from the gendisk to the mddev and destroy the gendisk when the mddev gets destroyed. However this has a complication. Between the call __blkdev_get->get_gendisk->kobj_lookup->md_probe and the call __blkdev_get->md_open there is no obvious way to hold a reference on the mddev any more, so unless something is done, it will disappear and gendisk will be destroyed prematurely. Also, once we decide to destroy the mddev, there will be an unlockable moment before the gendisk is unlinked (blk_unregister_region) during which a new reference to the gendisk can be created. We need to ensure that this reference can not be used. i.e. the ->open must fail. So: 1/ in md_probe we set a flag in the mddev (hold_active) which indicates that the array should be treated as active, even though there are no references, and no appearance of activity. This is cleared by md_release when the device is closed if it is no longer needed. This ensures that the gendisk will survive between md_probe and md_open. 2/ In md_open we check if the mddev we expect to open matches the gendisk that we did open. If there is a mismatch we return -ERESTARTSYS and modify __blkdev_get to retry from the top in that case. In the -ERESTARTSYS sys case we make sure to wait until the old gendisk (that we succeeded in opening) is really gone so we loop at most once. Some udev configurations will always open an md device when it first appears. If we allow an md device that was just created by an open to disappear on an immediate close, then this can race with such udev configurations and result in an infinite loop the device being opened and closed, then re-open due to the 'ADD' even from the first open, and then close and so on. So we make sure an md device, once created by an open, remains active at least until some md 'ioctl' has been made on it. This means that all normal usage of md devices will allow them to disappear promptly when not needed, but the worst that an incorrect usage will do it cause an inactive md device to be left in existence (it can easily be removed). As an array can be stopped by writing to a sysfs attribute echo clear > /sys/block/mdXXX/md/array_state we need to use scheduled work for deleting the gendisk and other kobjects. This allows us to wait for any pending gendisk deletion to complete by simply calling flush_scheduled_work(). Signed-off-by: NeilBrown <neilb@suse.de>
2009-01-08 22:31:10 +01:00
restart:
ret = -ENXIO;
disk = get_gendisk(bdev->bd_dev, &partno);
if (!disk)
goto out;
disk_block_events(disk);
mutex_lock_nested(&bdev->bd_mutex, for_part);
if (!bdev->bd_openers) {
bdev->bd_disk = disk;
bdev->bd_contains = bdev;
if (!partno) {
struct backing_dev_info *bdi;
ret = -ENXIO;
bdev->bd_part = disk_get_part(disk, partno);
if (!bdev->bd_part)
goto out_clear;
ret = 0;
if (disk->fops->open) {
ret = disk->fops->open(bdev, mode);
md: make devices disappear when they are no longer needed. Currently md devices, once created, never disappear until the module is unloaded. This is essentially because the gendisk holds a reference to the mddev, and the mddev holds a reference to the gendisk, this a circular reference. If we drop the reference from mddev to gendisk, then we need to ensure that the mddev is destroyed when the gendisk is destroyed. However it is not possible to hook into the gendisk destruction process to enable this. So we drop the reference from the gendisk to the mddev and destroy the gendisk when the mddev gets destroyed. However this has a complication. Between the call __blkdev_get->get_gendisk->kobj_lookup->md_probe and the call __blkdev_get->md_open there is no obvious way to hold a reference on the mddev any more, so unless something is done, it will disappear and gendisk will be destroyed prematurely. Also, once we decide to destroy the mddev, there will be an unlockable moment before the gendisk is unlinked (blk_unregister_region) during which a new reference to the gendisk can be created. We need to ensure that this reference can not be used. i.e. the ->open must fail. So: 1/ in md_probe we set a flag in the mddev (hold_active) which indicates that the array should be treated as active, even though there are no references, and no appearance of activity. This is cleared by md_release when the device is closed if it is no longer needed. This ensures that the gendisk will survive between md_probe and md_open. 2/ In md_open we check if the mddev we expect to open matches the gendisk that we did open. If there is a mismatch we return -ERESTARTSYS and modify __blkdev_get to retry from the top in that case. In the -ERESTARTSYS sys case we make sure to wait until the old gendisk (that we succeeded in opening) is really gone so we loop at most once. Some udev configurations will always open an md device when it first appears. If we allow an md device that was just created by an open to disappear on an immediate close, then this can race with such udev configurations and result in an infinite loop the device being opened and closed, then re-open due to the 'ADD' even from the first open, and then close and so on. So we make sure an md device, once created by an open, remains active at least until some md 'ioctl' has been made on it. This means that all normal usage of md devices will allow them to disappear promptly when not needed, but the worst that an incorrect usage will do it cause an inactive md device to be left in existence (it can easily be removed). As an array can be stopped by writing to a sysfs attribute echo clear > /sys/block/mdXXX/md/array_state we need to use scheduled work for deleting the gendisk and other kobjects. This allows us to wait for any pending gendisk deletion to complete by simply calling flush_scheduled_work(). Signed-off-by: NeilBrown <neilb@suse.de>
2009-01-08 22:31:10 +01:00
if (ret == -ERESTARTSYS) {
/* Lost a race with 'disk' being
* deleted, try again.
* See md.c
*/
disk_put_part(bdev->bd_part);
bdev->bd_part = NULL;
bdev->bd_disk = NULL;
mutex_unlock(&bdev->bd_mutex);
disk_unblock_events(disk);
module_put(disk->fops->owner);
put_disk(disk);
md: make devices disappear when they are no longer needed. Currently md devices, once created, never disappear until the module is unloaded. This is essentially because the gendisk holds a reference to the mddev, and the mddev holds a reference to the gendisk, this a circular reference. If we drop the reference from mddev to gendisk, then we need to ensure that the mddev is destroyed when the gendisk is destroyed. However it is not possible to hook into the gendisk destruction process to enable this. So we drop the reference from the gendisk to the mddev and destroy the gendisk when the mddev gets destroyed. However this has a complication. Between the call __blkdev_get->get_gendisk->kobj_lookup->md_probe and the call __blkdev_get->md_open there is no obvious way to hold a reference on the mddev any more, so unless something is done, it will disappear and gendisk will be destroyed prematurely. Also, once we decide to destroy the mddev, there will be an unlockable moment before the gendisk is unlinked (blk_unregister_region) during which a new reference to the gendisk can be created. We need to ensure that this reference can not be used. i.e. the ->open must fail. So: 1/ in md_probe we set a flag in the mddev (hold_active) which indicates that the array should be treated as active, even though there are no references, and no appearance of activity. This is cleared by md_release when the device is closed if it is no longer needed. This ensures that the gendisk will survive between md_probe and md_open. 2/ In md_open we check if the mddev we expect to open matches the gendisk that we did open. If there is a mismatch we return -ERESTARTSYS and modify __blkdev_get to retry from the top in that case. In the -ERESTARTSYS sys case we make sure to wait until the old gendisk (that we succeeded in opening) is really gone so we loop at most once. Some udev configurations will always open an md device when it first appears. If we allow an md device that was just created by an open to disappear on an immediate close, then this can race with such udev configurations and result in an infinite loop the device being opened and closed, then re-open due to the 'ADD' even from the first open, and then close and so on. So we make sure an md device, once created by an open, remains active at least until some md 'ioctl' has been made on it. This means that all normal usage of md devices will allow them to disappear promptly when not needed, but the worst that an incorrect usage will do it cause an inactive md device to be left in existence (it can easily be removed). As an array can be stopped by writing to a sysfs attribute echo clear > /sys/block/mdXXX/md/array_state we need to use scheduled work for deleting the gendisk and other kobjects. This allows us to wait for any pending gendisk deletion to complete by simply calling flush_scheduled_work(). Signed-off-by: NeilBrown <neilb@suse.de>
2009-01-08 22:31:10 +01:00
goto restart;
}
}
if (!ret && !bdev->bd_openers) {
bd_set_size(bdev,(loff_t)get_capacity(disk)<<9);
bdi = blk_get_backing_dev_info(bdev);
if (bdi == NULL)
bdi = &default_backing_dev_info;
bdev_inode_switch_bdi(bdev->bd_inode, bdi);
}
/*
* If the device is invalidated, rescan partition
* if open succeeded or failed with -ENOMEDIUM.
* The latter is necessary to prevent ghost
* partitions on a removed medium.
*/
if (bdev->bd_invalidated && (!ret || ret == -ENOMEDIUM))
rescan_partitions(disk, bdev);
if (ret)
goto out_clear;
} else {
struct block_device *whole;
whole = bdget_disk(disk, 0);
ret = -ENOMEM;
if (!whole)
goto out_clear;
[PATCH] lockdep: simplify some aspects of bd_mutex nesting When we open (actually blkdev_get) a partition we need to also open (get) the whole device that holds the partition. The involves some limited recursion. This patch tries to simplify some aspects of this. As well as opening the whole device, we need to increment ->bd_part_count when a partition is opened (this is used by rescan_partitions to avoid a rescan if any partition is active, as that would be confusing). The main change this patch makes is to move the inc/dec of bd_part_count into blkdev_{get,put} for the whole rather than doing it in blkdev_{get,put} for the partition. More specifically, we introduce __blkdev_get and __blkdev_put which do exactly what blkdev_{get,put} did, only with an extra "for_part" argument (blkget_{get,put} then call the __ version with a '0' for the extra argument). If for_part is 1, then the blkdev is being get(put) because a partition is being opened(closed) for the first(last) time, and so bd_part_count should be updated (on success). The particular advantage of pushing this function down is that the bd_mutex lock (which is needed to update bd_part_count) is already held at the lower level. Note that this slightly changes the semantics of bd_part_count. Instead of updating it whenever a partition is opened or released, it is now only updated on the first open or last release. This is an adequate semantic as it is only ever tested for "== 0". Having introduced these functions we remove the current bd_part_count updates from do_open (which is really the body of blkdev_get) and call __blkdev_get(... 1). Similarly in blkget_put we remove the old bd_part_count updates and call __blkget_put(..., 1). This call is moved to the end of __blkdev_put to avoid nested locks of bd_mutex. Finally the mutex_lock on whole->bd_mutex in do_open can be removed. It was only really needed to protect bd_part_count, and that is now managed (and protected) within the recursive call. The observation that bd_part_count is central to the locking issues, and the modifications to create __blkdev_put are from Peter Zijlstra. Cc: Ingo Molnar <mingo@elte.hu> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 11:36:16 +01:00
BUG_ON(for_part);
ret = __blkdev_get(whole, mode, 1);
if (ret)
goto out_clear;
bdev->bd_contains = whole;
bdev_inode_switch_bdi(bdev->bd_inode,
whole->bd_inode->i_data.backing_dev_info);
bdev->bd_part = disk_get_part(disk, partno);
if (!(disk->flags & GENHD_FL_UP) ||
!bdev->bd_part || !bdev->bd_part->nr_sects) {
ret = -ENXIO;
goto out_clear;
}
bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9);
}
} else {
if (bdev->bd_contains == bdev) {
ret = 0;
if (bdev->bd_disk->fops->open)
ret = bdev->bd_disk->fops->open(bdev, mode);
/* the same as first opener case, read comment there */
if (bdev->bd_invalidated && (!ret || ret == -ENOMEDIUM))
rescan_partitions(bdev->bd_disk, bdev);
if (ret)
goto out_unlock_bdev;
}
/* only one opener holds refs to the module and disk */
module_put(disk->fops->owner);
put_disk(disk);
}
bdev->bd_openers++;
[PATCH] lockdep: simplify some aspects of bd_mutex nesting When we open (actually blkdev_get) a partition we need to also open (get) the whole device that holds the partition. The involves some limited recursion. This patch tries to simplify some aspects of this. As well as opening the whole device, we need to increment ->bd_part_count when a partition is opened (this is used by rescan_partitions to avoid a rescan if any partition is active, as that would be confusing). The main change this patch makes is to move the inc/dec of bd_part_count into blkdev_{get,put} for the whole rather than doing it in blkdev_{get,put} for the partition. More specifically, we introduce __blkdev_get and __blkdev_put which do exactly what blkdev_{get,put} did, only with an extra "for_part" argument (blkget_{get,put} then call the __ version with a '0' for the extra argument). If for_part is 1, then the blkdev is being get(put) because a partition is being opened(closed) for the first(last) time, and so bd_part_count should be updated (on success). The particular advantage of pushing this function down is that the bd_mutex lock (which is needed to update bd_part_count) is already held at the lower level. Note that this slightly changes the semantics of bd_part_count. Instead of updating it whenever a partition is opened or released, it is now only updated on the first open or last release. This is an adequate semantic as it is only ever tested for "== 0". Having introduced these functions we remove the current bd_part_count updates from do_open (which is really the body of blkdev_get) and call __blkdev_get(... 1). Similarly in blkget_put we remove the old bd_part_count updates and call __blkget_put(..., 1). This call is moved to the end of __blkdev_put to avoid nested locks of bd_mutex. Finally the mutex_lock on whole->bd_mutex in do_open can be removed. It was only really needed to protect bd_part_count, and that is now managed (and protected) within the recursive call. The observation that bd_part_count is central to the locking issues, and the modifications to create __blkdev_put are from Peter Zijlstra. Cc: Ingo Molnar <mingo@elte.hu> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 11:36:16 +01:00
if (for_part)
bdev->bd_part_count++;
mutex_unlock(&bdev->bd_mutex);
disk_unblock_events(disk);
return 0;
out_clear:
disk_put_part(bdev->bd_part);
bdev->bd_disk = NULL;
bdev->bd_part = NULL;
bdev_inode_switch_bdi(bdev->bd_inode, &default_backing_dev_info);
if (bdev != bdev->bd_contains)
__blkdev_put(bdev->bd_contains, mode, 1);
bdev->bd_contains = NULL;
out_unlock_bdev:
mutex_unlock(&bdev->bd_mutex);
disk_unblock_events(disk);
module_put(disk->fops->owner);
put_disk(disk);
out:
bdput(bdev);
return ret;
}
/**
* blkdev_get - open a block device
* @bdev: block_device to open
* @mode: FMODE_* mask
* @holder: exclusive holder identifier
*
* Open @bdev with @mode. If @mode includes %FMODE_EXCL, @bdev is
* open with exclusive access. Specifying %FMODE_EXCL with %NULL
* @holder is invalid. Exclusive opens may nest for the same @holder.
*
* On success, the reference count of @bdev is unchanged. On failure,
* @bdev is put.
*
* CONTEXT:
* Might sleep.
*
* RETURNS:
* 0 on success, -errno on failure.
*/
block: make blkdev_get/put() handle exclusive access Over time, block layer has accumulated a set of APIs dealing with bdev open, close, claim and release. * blkdev_get/put() are the primary open and close functions. * bd_claim/release() deal with exclusive open. * open/close_bdev_exclusive() are combination of open and claim and the other way around, respectively. * bd_link/unlink_disk_holder() to create and remove holder/slave symlinks. * open_by_devnum() wraps bdget() + blkdev_get(). The interface is a bit confusing and the decoupling of open and claim makes it impossible to properly guarantee exclusive access as in-kernel open + claim sequence can disturb the existing exclusive open even before the block layer knows the current open if for another exclusive access. Reorganize the interface such that, * blkdev_get() is extended to include exclusive access management. @holder argument is added and, if is @FMODE_EXCL specified, it will gain exclusive access atomically w.r.t. other exclusive accesses. * blkdev_put() is similarly extended. It now takes @mode argument and if @FMODE_EXCL is set, it releases an exclusive access. Also, when the last exclusive claim is released, the holder/slave symlinks are removed automatically. * bd_claim/release() and close_bdev_exclusive() are no longer necessary and either made static or removed. * bd_link_disk_holder() remains the same but bd_unlink_disk_holder() is no longer necessary and removed. * open_bdev_exclusive() becomes a simple wrapper around lookup_bdev() and blkdev_get(). It also has an unexpected extra bdev_read_only() test which probably should be moved into blkdev_get(). * open_by_devnum() is modified to take @holder argument and pass it to blkdev_get(). Most of bdev open/close operations are unified into blkdev_get/put() and most exclusive accesses are tested atomically at the open time (as it should). This cleans up code and removes some, both valid and invalid, but unnecessary all the same, corner cases. open_bdev_exclusive() and open_by_devnum() can use further cleanup - rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop special features. Well, let's leave them for another day. Most conversions are straight-forward. drbd conversion is a bit more involved as there was some reordering, but the logic should stay the same. Signed-off-by: Tejun Heo <tj@kernel.org> Acked-by: Neil Brown <neilb@suse.de> Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Acked-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Philipp Reisner <philipp.reisner@linbit.com> Cc: Peter Osterlund <petero2@telia.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Jan Kara <jack@suse.cz> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Cc: Alex Elder <aelder@sgi.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: dm-devel@redhat.com Cc: drbd-dev@lists.linbit.com Cc: Leo Chen <leochen@broadcom.com> Cc: Scott Branden <sbranden@broadcom.com> Cc: Chris Mason <chris.mason@oracle.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@linux.vnet.ibm.com> Cc: Joern Engel <joern@logfs.org> Cc: reiserfs-devel@vger.kernel.org Cc: Alexander Viro <viro@zeniv.linux.org.uk>
2010-11-13 11:55:17 +01:00
int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder)
{
block: make blkdev_get/put() handle exclusive access Over time, block layer has accumulated a set of APIs dealing with bdev open, close, claim and release. * blkdev_get/put() are the primary open and close functions. * bd_claim/release() deal with exclusive open. * open/close_bdev_exclusive() are combination of open and claim and the other way around, respectively. * bd_link/unlink_disk_holder() to create and remove holder/slave symlinks. * open_by_devnum() wraps bdget() + blkdev_get(). The interface is a bit confusing and the decoupling of open and claim makes it impossible to properly guarantee exclusive access as in-kernel open + claim sequence can disturb the existing exclusive open even before the block layer knows the current open if for another exclusive access. Reorganize the interface such that, * blkdev_get() is extended to include exclusive access management. @holder argument is added and, if is @FMODE_EXCL specified, it will gain exclusive access atomically w.r.t. other exclusive accesses. * blkdev_put() is similarly extended. It now takes @mode argument and if @FMODE_EXCL is set, it releases an exclusive access. Also, when the last exclusive claim is released, the holder/slave symlinks are removed automatically. * bd_claim/release() and close_bdev_exclusive() are no longer necessary and either made static or removed. * bd_link_disk_holder() remains the same but bd_unlink_disk_holder() is no longer necessary and removed. * open_bdev_exclusive() becomes a simple wrapper around lookup_bdev() and blkdev_get(). It also has an unexpected extra bdev_read_only() test which probably should be moved into blkdev_get(). * open_by_devnum() is modified to take @holder argument and pass it to blkdev_get(). Most of bdev open/close operations are unified into blkdev_get/put() and most exclusive accesses are tested atomically at the open time (as it should). This cleans up code and removes some, both valid and invalid, but unnecessary all the same, corner cases. open_bdev_exclusive() and open_by_devnum() can use further cleanup - rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop special features. Well, let's leave them for another day. Most conversions are straight-forward. drbd conversion is a bit more involved as there was some reordering, but the logic should stay the same. Signed-off-by: Tejun Heo <tj@kernel.org> Acked-by: Neil Brown <neilb@suse.de> Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Acked-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Philipp Reisner <philipp.reisner@linbit.com> Cc: Peter Osterlund <petero2@telia.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Jan Kara <jack@suse.cz> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Cc: Alex Elder <aelder@sgi.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: dm-devel@redhat.com Cc: drbd-dev@lists.linbit.com Cc: Leo Chen <leochen@broadcom.com> Cc: Scott Branden <sbranden@broadcom.com> Cc: Chris Mason <chris.mason@oracle.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@linux.vnet.ibm.com> Cc: Joern Engel <joern@logfs.org> Cc: reiserfs-devel@vger.kernel.org Cc: Alexander Viro <viro@zeniv.linux.org.uk>
2010-11-13 11:55:17 +01:00
struct block_device *whole = NULL;
int res;
WARN_ON_ONCE((mode & FMODE_EXCL) && !holder);
if ((mode & FMODE_EXCL) && holder) {
whole = bd_start_claiming(bdev, holder);
if (IS_ERR(whole)) {
bdput(bdev);
return PTR_ERR(whole);
}
}
res = __blkdev_get(bdev, mode, 0);
if (whole) {
struct gendisk *disk = whole->bd_disk;
/* finish claiming */
implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-12-08 20:57:37 +01:00
mutex_lock(&bdev->bd_mutex);
spin_lock(&bdev_lock);
implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-12-08 20:57:37 +01:00
if (!res) {
BUG_ON(!bd_may_claim(bdev, whole, holder));
/*
* Note that for a whole device bd_holders
* will be incremented twice, and bd_holder
* will be set to bd_may_claim before being
* set to holder
*/
whole->bd_holders++;
whole->bd_holder = bd_may_claim;
bdev->bd_holders++;
bdev->bd_holder = holder;
}
/* tell others that we're done */
BUG_ON(whole->bd_claiming != holder);
whole->bd_claiming = NULL;
wake_up_bit(&whole->bd_claiming, 0);
spin_unlock(&bdev_lock);
implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-12-08 20:57:37 +01:00
/*
* Block event polling for write claims if requested. Any
* write holder makes the write_holder state stick until
* all are released. This is good enough and tracking
* individual writeable reference is too fragile given the
* way @mode is used in blkdev_get/put().
implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-12-08 20:57:37 +01:00
*/
if (!res && (mode & FMODE_WRITE) && !bdev->bd_write_holder &&
(disk->flags & GENHD_FL_BLOCK_EVENTS_ON_EXCL_WRITE)) {
implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-12-08 20:57:37 +01:00
bdev->bd_write_holder = true;
disk_block_events(disk);
implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-12-08 20:57:37 +01:00
}
mutex_unlock(&bdev->bd_mutex);
bdput(whole);
block: make blkdev_get/put() handle exclusive access Over time, block layer has accumulated a set of APIs dealing with bdev open, close, claim and release. * blkdev_get/put() are the primary open and close functions. * bd_claim/release() deal with exclusive open. * open/close_bdev_exclusive() are combination of open and claim and the other way around, respectively. * bd_link/unlink_disk_holder() to create and remove holder/slave symlinks. * open_by_devnum() wraps bdget() + blkdev_get(). The interface is a bit confusing and the decoupling of open and claim makes it impossible to properly guarantee exclusive access as in-kernel open + claim sequence can disturb the existing exclusive open even before the block layer knows the current open if for another exclusive access. Reorganize the interface such that, * blkdev_get() is extended to include exclusive access management. @holder argument is added and, if is @FMODE_EXCL specified, it will gain exclusive access atomically w.r.t. other exclusive accesses. * blkdev_put() is similarly extended. It now takes @mode argument and if @FMODE_EXCL is set, it releases an exclusive access. Also, when the last exclusive claim is released, the holder/slave symlinks are removed automatically. * bd_claim/release() and close_bdev_exclusive() are no longer necessary and either made static or removed. * bd_link_disk_holder() remains the same but bd_unlink_disk_holder() is no longer necessary and removed. * open_bdev_exclusive() becomes a simple wrapper around lookup_bdev() and blkdev_get(). It also has an unexpected extra bdev_read_only() test which probably should be moved into blkdev_get(). * open_by_devnum() is modified to take @holder argument and pass it to blkdev_get(). Most of bdev open/close operations are unified into blkdev_get/put() and most exclusive accesses are tested atomically at the open time (as it should). This cleans up code and removes some, both valid and invalid, but unnecessary all the same, corner cases. open_bdev_exclusive() and open_by_devnum() can use further cleanup - rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop special features. Well, let's leave them for another day. Most conversions are straight-forward. drbd conversion is a bit more involved as there was some reordering, but the logic should stay the same. Signed-off-by: Tejun Heo <tj@kernel.org> Acked-by: Neil Brown <neilb@suse.de> Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Acked-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Philipp Reisner <philipp.reisner@linbit.com> Cc: Peter Osterlund <petero2@telia.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Jan Kara <jack@suse.cz> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Cc: Alex Elder <aelder@sgi.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: dm-devel@redhat.com Cc: drbd-dev@lists.linbit.com Cc: Leo Chen <leochen@broadcom.com> Cc: Scott Branden <sbranden@broadcom.com> Cc: Chris Mason <chris.mason@oracle.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@linux.vnet.ibm.com> Cc: Joern Engel <joern@logfs.org> Cc: reiserfs-devel@vger.kernel.org Cc: Alexander Viro <viro@zeniv.linux.org.uk>
2010-11-13 11:55:17 +01:00
}
return res;
[PATCH] lockdep: simplify some aspects of bd_mutex nesting When we open (actually blkdev_get) a partition we need to also open (get) the whole device that holds the partition. The involves some limited recursion. This patch tries to simplify some aspects of this. As well as opening the whole device, we need to increment ->bd_part_count when a partition is opened (this is used by rescan_partitions to avoid a rescan if any partition is active, as that would be confusing). The main change this patch makes is to move the inc/dec of bd_part_count into blkdev_{get,put} for the whole rather than doing it in blkdev_{get,put} for the partition. More specifically, we introduce __blkdev_get and __blkdev_put which do exactly what blkdev_{get,put} did, only with an extra "for_part" argument (blkget_{get,put} then call the __ version with a '0' for the extra argument). If for_part is 1, then the blkdev is being get(put) because a partition is being opened(closed) for the first(last) time, and so bd_part_count should be updated (on success). The particular advantage of pushing this function down is that the bd_mutex lock (which is needed to update bd_part_count) is already held at the lower level. Note that this slightly changes the semantics of bd_part_count. Instead of updating it whenever a partition is opened or released, it is now only updated on the first open or last release. This is an adequate semantic as it is only ever tested for "== 0". Having introduced these functions we remove the current bd_part_count updates from do_open (which is really the body of blkdev_get) and call __blkdev_get(... 1). Similarly in blkget_put we remove the old bd_part_count updates and call __blkget_put(..., 1). This call is moved to the end of __blkdev_put to avoid nested locks of bd_mutex. Finally the mutex_lock on whole->bd_mutex in do_open can be removed. It was only really needed to protect bd_part_count, and that is now managed (and protected) within the recursive call. The observation that bd_part_count is central to the locking issues, and the modifications to create __blkdev_put are from Peter Zijlstra. Cc: Ingo Molnar <mingo@elte.hu> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 11:36:16 +01:00
}
EXPORT_SYMBOL(blkdev_get);
/**
* blkdev_get_by_path - open a block device by name
* @path: path to the block device to open
* @mode: FMODE_* mask
* @holder: exclusive holder identifier
*
* Open the blockdevice described by the device file at @path. @mode
* and @holder are identical to blkdev_get().
*
* On success, the returned block_device has reference count of one.
*
* CONTEXT:
* Might sleep.
*
* RETURNS:
* Pointer to block_device on success, ERR_PTR(-errno) on failure.
*/
struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
void *holder)
{
struct block_device *bdev;
int err;
bdev = lookup_bdev(path);
if (IS_ERR(bdev))
return bdev;
err = blkdev_get(bdev, mode, holder);
if (err)
return ERR_PTR(err);
if ((mode & FMODE_WRITE) && bdev_read_only(bdev)) {
blkdev_put(bdev, mode);
return ERR_PTR(-EACCES);
}
return bdev;
}
EXPORT_SYMBOL(blkdev_get_by_path);
/**
* blkdev_get_by_dev - open a block device by device number
* @dev: device number of block device to open
* @mode: FMODE_* mask
* @holder: exclusive holder identifier
*
* Open the blockdevice described by device number @dev. @mode and
* @holder are identical to blkdev_get().
*
* Use it ONLY if you really do not have anything better - i.e. when
* you are behind a truly sucky interface and all you are given is a
* device number. _Never_ to be used for internal purposes. If you
* ever need it - reconsider your API.
*
* On success, the returned block_device has reference count of one.
*
* CONTEXT:
* Might sleep.
*
* RETURNS:
* Pointer to block_device on success, ERR_PTR(-errno) on failure.
*/
struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder)
{
struct block_device *bdev;
int err;
bdev = bdget(dev);
if (!bdev)
return ERR_PTR(-ENOMEM);
err = blkdev_get(bdev, mode, holder);
if (err)
return ERR_PTR(err);
return bdev;
}
EXPORT_SYMBOL(blkdev_get_by_dev);
static int blkdev_open(struct inode * inode, struct file * filp)
{
struct block_device *bdev;
/*
* Preserve backwards compatibility and allow large file access
* even if userspace doesn't ask for it explicitly. Some mkfs
* binary needs it. We might want to drop this workaround
* during an unstable branch.
*/
filp->f_flags |= O_LARGEFILE;
if (filp->f_flags & O_NDELAY)
filp->f_mode |= FMODE_NDELAY;
if (filp->f_flags & O_EXCL)
filp->f_mode |= FMODE_EXCL;
if ((filp->f_flags & O_ACCMODE) == 3)
filp->f_mode |= FMODE_WRITE_IOCTL;
bdev = bd_acquire(inode);
if (bdev == NULL)
return -ENOMEM;
filp->f_mapping = bdev->bd_inode->i_mapping;
block: make blkdev_get/put() handle exclusive access Over time, block layer has accumulated a set of APIs dealing with bdev open, close, claim and release. * blkdev_get/put() are the primary open and close functions. * bd_claim/release() deal with exclusive open. * open/close_bdev_exclusive() are combination of open and claim and the other way around, respectively. * bd_link/unlink_disk_holder() to create and remove holder/slave symlinks. * open_by_devnum() wraps bdget() + blkdev_get(). The interface is a bit confusing and the decoupling of open and claim makes it impossible to properly guarantee exclusive access as in-kernel open + claim sequence can disturb the existing exclusive open even before the block layer knows the current open if for another exclusive access. Reorganize the interface such that, * blkdev_get() is extended to include exclusive access management. @holder argument is added and, if is @FMODE_EXCL specified, it will gain exclusive access atomically w.r.t. other exclusive accesses. * blkdev_put() is similarly extended. It now takes @mode argument and if @FMODE_EXCL is set, it releases an exclusive access. Also, when the last exclusive claim is released, the holder/slave symlinks are removed automatically. * bd_claim/release() and close_bdev_exclusive() are no longer necessary and either made static or removed. * bd_link_disk_holder() remains the same but bd_unlink_disk_holder() is no longer necessary and removed. * open_bdev_exclusive() becomes a simple wrapper around lookup_bdev() and blkdev_get(). It also has an unexpected extra bdev_read_only() test which probably should be moved into blkdev_get(). * open_by_devnum() is modified to take @holder argument and pass it to blkdev_get(). Most of bdev open/close operations are unified into blkdev_get/put() and most exclusive accesses are tested atomically at the open time (as it should). This cleans up code and removes some, both valid and invalid, but unnecessary all the same, corner cases. open_bdev_exclusive() and open_by_devnum() can use further cleanup - rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop special features. Well, let's leave them for another day. Most conversions are straight-forward. drbd conversion is a bit more involved as there was some reordering, but the logic should stay the same. Signed-off-by: Tejun Heo <tj@kernel.org> Acked-by: Neil Brown <neilb@suse.de> Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Acked-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Philipp Reisner <philipp.reisner@linbit.com> Cc: Peter Osterlund <petero2@telia.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Jan Kara <jack@suse.cz> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Cc: Alex Elder <aelder@sgi.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: dm-devel@redhat.com Cc: drbd-dev@lists.linbit.com Cc: Leo Chen <leochen@broadcom.com> Cc: Scott Branden <sbranden@broadcom.com> Cc: Chris Mason <chris.mason@oracle.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@linux.vnet.ibm.com> Cc: Joern Engel <joern@logfs.org> Cc: reiserfs-devel@vger.kernel.org Cc: Alexander Viro <viro@zeniv.linux.org.uk>
2010-11-13 11:55:17 +01:00
return blkdev_get(bdev, filp->f_mode, filp);
}
static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part)
{
int ret = 0;
struct gendisk *disk = bdev->bd_disk;
[PATCH] lockdep: simplify some aspects of bd_mutex nesting When we open (actually blkdev_get) a partition we need to also open (get) the whole device that holds the partition. The involves some limited recursion. This patch tries to simplify some aspects of this. As well as opening the whole device, we need to increment ->bd_part_count when a partition is opened (this is used by rescan_partitions to avoid a rescan if any partition is active, as that would be confusing). The main change this patch makes is to move the inc/dec of bd_part_count into blkdev_{get,put} for the whole rather than doing it in blkdev_{get,put} for the partition. More specifically, we introduce __blkdev_get and __blkdev_put which do exactly what blkdev_{get,put} did, only with an extra "for_part" argument (blkget_{get,put} then call the __ version with a '0' for the extra argument). If for_part is 1, then the blkdev is being get(put) because a partition is being opened(closed) for the first(last) time, and so bd_part_count should be updated (on success). The particular advantage of pushing this function down is that the bd_mutex lock (which is needed to update bd_part_count) is already held at the lower level. Note that this slightly changes the semantics of bd_part_count. Instead of updating it whenever a partition is opened or released, it is now only updated on the first open or last release. This is an adequate semantic as it is only ever tested for "== 0". Having introduced these functions we remove the current bd_part_count updates from do_open (which is really the body of blkdev_get) and call __blkdev_get(... 1). Similarly in blkget_put we remove the old bd_part_count updates and call __blkget_put(..., 1). This call is moved to the end of __blkdev_put to avoid nested locks of bd_mutex. Finally the mutex_lock on whole->bd_mutex in do_open can be removed. It was only really needed to protect bd_part_count, and that is now managed (and protected) within the recursive call. The observation that bd_part_count is central to the locking issues, and the modifications to create __blkdev_put are from Peter Zijlstra. Cc: Ingo Molnar <mingo@elte.hu> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 11:36:16 +01:00
struct block_device *victim = NULL;
mutex_lock_nested(&bdev->bd_mutex, for_part);
[PATCH] lockdep: simplify some aspects of bd_mutex nesting When we open (actually blkdev_get) a partition we need to also open (get) the whole device that holds the partition. The involves some limited recursion. This patch tries to simplify some aspects of this. As well as opening the whole device, we need to increment ->bd_part_count when a partition is opened (this is used by rescan_partitions to avoid a rescan if any partition is active, as that would be confusing). The main change this patch makes is to move the inc/dec of bd_part_count into blkdev_{get,put} for the whole rather than doing it in blkdev_{get,put} for the partition. More specifically, we introduce __blkdev_get and __blkdev_put which do exactly what blkdev_{get,put} did, only with an extra "for_part" argument (blkget_{get,put} then call the __ version with a '0' for the extra argument). If for_part is 1, then the blkdev is being get(put) because a partition is being opened(closed) for the first(last) time, and so bd_part_count should be updated (on success). The particular advantage of pushing this function down is that the bd_mutex lock (which is needed to update bd_part_count) is already held at the lower level. Note that this slightly changes the semantics of bd_part_count. Instead of updating it whenever a partition is opened or released, it is now only updated on the first open or last release. This is an adequate semantic as it is only ever tested for "== 0". Having introduced these functions we remove the current bd_part_count updates from do_open (which is really the body of blkdev_get) and call __blkdev_get(... 1). Similarly in blkget_put we remove the old bd_part_count updates and call __blkget_put(..., 1). This call is moved to the end of __blkdev_put to avoid nested locks of bd_mutex. Finally the mutex_lock on whole->bd_mutex in do_open can be removed. It was only really needed to protect bd_part_count, and that is now managed (and protected) within the recursive call. The observation that bd_part_count is central to the locking issues, and the modifications to create __blkdev_put are from Peter Zijlstra. Cc: Ingo Molnar <mingo@elte.hu> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 11:36:16 +01:00
if (for_part)
bdev->bd_part_count--;
if (!--bdev->bd_openers) {
WARN_ON_ONCE(bdev->bd_holders);
sync_blockdev(bdev);
kill_bdev(bdev);
}
if (bdev->bd_contains == bdev) {
if (disk->fops->release)
ret = disk->fops->release(disk, mode);
}
if (!bdev->bd_openers) {
struct module *owner = disk->fops->owner;
put_disk(disk);
module_put(owner);
disk_put_part(bdev->bd_part);
bdev->bd_part = NULL;
bdev->bd_disk = NULL;
bdev_inode_switch_bdi(bdev->bd_inode,
&default_backing_dev_info);
[PATCH] lockdep: simplify some aspects of bd_mutex nesting When we open (actually blkdev_get) a partition we need to also open (get) the whole device that holds the partition. The involves some limited recursion. This patch tries to simplify some aspects of this. As well as opening the whole device, we need to increment ->bd_part_count when a partition is opened (this is used by rescan_partitions to avoid a rescan if any partition is active, as that would be confusing). The main change this patch makes is to move the inc/dec of bd_part_count into blkdev_{get,put} for the whole rather than doing it in blkdev_{get,put} for the partition. More specifically, we introduce __blkdev_get and __blkdev_put which do exactly what blkdev_{get,put} did, only with an extra "for_part" argument (blkget_{get,put} then call the __ version with a '0' for the extra argument). If for_part is 1, then the blkdev is being get(put) because a partition is being opened(closed) for the first(last) time, and so bd_part_count should be updated (on success). The particular advantage of pushing this function down is that the bd_mutex lock (which is needed to update bd_part_count) is already held at the lower level. Note that this slightly changes the semantics of bd_part_count. Instead of updating it whenever a partition is opened or released, it is now only updated on the first open or last release. This is an adequate semantic as it is only ever tested for "== 0". Having introduced these functions we remove the current bd_part_count updates from do_open (which is really the body of blkdev_get) and call __blkdev_get(... 1). Similarly in blkget_put we remove the old bd_part_count updates and call __blkget_put(..., 1). This call is moved to the end of __blkdev_put to avoid nested locks of bd_mutex. Finally the mutex_lock on whole->bd_mutex in do_open can be removed. It was only really needed to protect bd_part_count, and that is now managed (and protected) within the recursive call. The observation that bd_part_count is central to the locking issues, and the modifications to create __blkdev_put are from Peter Zijlstra. Cc: Ingo Molnar <mingo@elte.hu> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 11:36:16 +01:00
if (bdev != bdev->bd_contains)
victim = bdev->bd_contains;
bdev->bd_contains = NULL;
}
mutex_unlock(&bdev->bd_mutex);
bdput(bdev);
[PATCH] lockdep: simplify some aspects of bd_mutex nesting When we open (actually blkdev_get) a partition we need to also open (get) the whole device that holds the partition. The involves some limited recursion. This patch tries to simplify some aspects of this. As well as opening the whole device, we need to increment ->bd_part_count when a partition is opened (this is used by rescan_partitions to avoid a rescan if any partition is active, as that would be confusing). The main change this patch makes is to move the inc/dec of bd_part_count into blkdev_{get,put} for the whole rather than doing it in blkdev_{get,put} for the partition. More specifically, we introduce __blkdev_get and __blkdev_put which do exactly what blkdev_{get,put} did, only with an extra "for_part" argument (blkget_{get,put} then call the __ version with a '0' for the extra argument). If for_part is 1, then the blkdev is being get(put) because a partition is being opened(closed) for the first(last) time, and so bd_part_count should be updated (on success). The particular advantage of pushing this function down is that the bd_mutex lock (which is needed to update bd_part_count) is already held at the lower level. Note that this slightly changes the semantics of bd_part_count. Instead of updating it whenever a partition is opened or released, it is now only updated on the first open or last release. This is an adequate semantic as it is only ever tested for "== 0". Having introduced these functions we remove the current bd_part_count updates from do_open (which is really the body of blkdev_get) and call __blkdev_get(... 1). Similarly in blkget_put we remove the old bd_part_count updates and call __blkget_put(..., 1). This call is moved to the end of __blkdev_put to avoid nested locks of bd_mutex. Finally the mutex_lock on whole->bd_mutex in do_open can be removed. It was only really needed to protect bd_part_count, and that is now managed (and protected) within the recursive call. The observation that bd_part_count is central to the locking issues, and the modifications to create __blkdev_put are from Peter Zijlstra. Cc: Ingo Molnar <mingo@elte.hu> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 11:36:16 +01:00
if (victim)
__blkdev_put(victim, mode, 1);
return ret;
}
int blkdev_put(struct block_device *bdev, fmode_t mode)
[PATCH] lockdep: simplify some aspects of bd_mutex nesting When we open (actually blkdev_get) a partition we need to also open (get) the whole device that holds the partition. The involves some limited recursion. This patch tries to simplify some aspects of this. As well as opening the whole device, we need to increment ->bd_part_count when a partition is opened (this is used by rescan_partitions to avoid a rescan if any partition is active, as that would be confusing). The main change this patch makes is to move the inc/dec of bd_part_count into blkdev_{get,put} for the whole rather than doing it in blkdev_{get,put} for the partition. More specifically, we introduce __blkdev_get and __blkdev_put which do exactly what blkdev_{get,put} did, only with an extra "for_part" argument (blkget_{get,put} then call the __ version with a '0' for the extra argument). If for_part is 1, then the blkdev is being get(put) because a partition is being opened(closed) for the first(last) time, and so bd_part_count should be updated (on success). The particular advantage of pushing this function down is that the bd_mutex lock (which is needed to update bd_part_count) is already held at the lower level. Note that this slightly changes the semantics of bd_part_count. Instead of updating it whenever a partition is opened or released, it is now only updated on the first open or last release. This is an adequate semantic as it is only ever tested for "== 0". Having introduced these functions we remove the current bd_part_count updates from do_open (which is really the body of blkdev_get) and call __blkdev_get(... 1). Similarly in blkget_put we remove the old bd_part_count updates and call __blkget_put(..., 1). This call is moved to the end of __blkdev_put to avoid nested locks of bd_mutex. Finally the mutex_lock on whole->bd_mutex in do_open can be removed. It was only really needed to protect bd_part_count, and that is now managed (and protected) within the recursive call. The observation that bd_part_count is central to the locking issues, and the modifications to create __blkdev_put are from Peter Zijlstra. Cc: Ingo Molnar <mingo@elte.hu> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 11:36:16 +01:00
{
2011-07-01 16:17:47 +02:00
mutex_lock(&bdev->bd_mutex);
block: make blkdev_get/put() handle exclusive access Over time, block layer has accumulated a set of APIs dealing with bdev open, close, claim and release. * blkdev_get/put() are the primary open and close functions. * bd_claim/release() deal with exclusive open. * open/close_bdev_exclusive() are combination of open and claim and the other way around, respectively. * bd_link/unlink_disk_holder() to create and remove holder/slave symlinks. * open_by_devnum() wraps bdget() + blkdev_get(). The interface is a bit confusing and the decoupling of open and claim makes it impossible to properly guarantee exclusive access as in-kernel open + claim sequence can disturb the existing exclusive open even before the block layer knows the current open if for another exclusive access. Reorganize the interface such that, * blkdev_get() is extended to include exclusive access management. @holder argument is added and, if is @FMODE_EXCL specified, it will gain exclusive access atomically w.r.t. other exclusive accesses. * blkdev_put() is similarly extended. It now takes @mode argument and if @FMODE_EXCL is set, it releases an exclusive access. Also, when the last exclusive claim is released, the holder/slave symlinks are removed automatically. * bd_claim/release() and close_bdev_exclusive() are no longer necessary and either made static or removed. * bd_link_disk_holder() remains the same but bd_unlink_disk_holder() is no longer necessary and removed. * open_bdev_exclusive() becomes a simple wrapper around lookup_bdev() and blkdev_get(). It also has an unexpected extra bdev_read_only() test which probably should be moved into blkdev_get(). * open_by_devnum() is modified to take @holder argument and pass it to blkdev_get(). Most of bdev open/close operations are unified into blkdev_get/put() and most exclusive accesses are tested atomically at the open time (as it should). This cleans up code and removes some, both valid and invalid, but unnecessary all the same, corner cases. open_bdev_exclusive() and open_by_devnum() can use further cleanup - rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop special features. Well, let's leave them for another day. Most conversions are straight-forward. drbd conversion is a bit more involved as there was some reordering, but the logic should stay the same. Signed-off-by: Tejun Heo <tj@kernel.org> Acked-by: Neil Brown <neilb@suse.de> Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Acked-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Philipp Reisner <philipp.reisner@linbit.com> Cc: Peter Osterlund <petero2@telia.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Jan Kara <jack@suse.cz> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Cc: Alex Elder <aelder@sgi.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: dm-devel@redhat.com Cc: drbd-dev@lists.linbit.com Cc: Leo Chen <leochen@broadcom.com> Cc: Scott Branden <sbranden@broadcom.com> Cc: Chris Mason <chris.mason@oracle.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@linux.vnet.ibm.com> Cc: Joern Engel <joern@logfs.org> Cc: reiserfs-devel@vger.kernel.org Cc: Alexander Viro <viro@zeniv.linux.org.uk>
2010-11-13 11:55:17 +01:00
if (mode & FMODE_EXCL) {
bool bdev_free;
/*
* Release a claim on the device. The holder fields
* are protected with bdev_lock. bd_mutex is to
* synchronize disk_holder unlinking.
*/
spin_lock(&bdev_lock);
WARN_ON_ONCE(--bdev->bd_holders < 0);
WARN_ON_ONCE(--bdev->bd_contains->bd_holders < 0);
/* bd_contains might point to self, check in a separate step */
if ((bdev_free = !bdev->bd_holders))
bdev->bd_holder = NULL;
if (!bdev->bd_contains->bd_holders)
bdev->bd_contains->bd_holder = NULL;
spin_unlock(&bdev_lock);
implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-12-08 20:57:37 +01:00
/*
* If this was the last claim, remove holder link and
* unblock evpoll if it was a write holder.
*/
2011-07-01 16:17:47 +02:00
if (bdev_free && bdev->bd_write_holder) {
disk_unblock_events(bdev->bd_disk);
bdev->bd_write_holder = false;
implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-12-08 20:57:37 +01:00
}
}
implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Kay Sievers <kay.sievers@vrfy.org> Cc: Jan Kara <jack@suse.cz> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-12-08 20:57:37 +01:00
2011-07-01 16:17:47 +02:00
/*
* Trigger event checking and tell drivers to flush MEDIA_CHANGE
* event. This is to ensure detection of media removal commanded
* from userland - e.g. eject(1).
*/
disk_flush_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE);
mutex_unlock(&bdev->bd_mutex);
return __blkdev_put(bdev, mode, 0);
[PATCH] lockdep: simplify some aspects of bd_mutex nesting When we open (actually blkdev_get) a partition we need to also open (get) the whole device that holds the partition. The involves some limited recursion. This patch tries to simplify some aspects of this. As well as opening the whole device, we need to increment ->bd_part_count when a partition is opened (this is used by rescan_partitions to avoid a rescan if any partition is active, as that would be confusing). The main change this patch makes is to move the inc/dec of bd_part_count into blkdev_{get,put} for the whole rather than doing it in blkdev_{get,put} for the partition. More specifically, we introduce __blkdev_get and __blkdev_put which do exactly what blkdev_{get,put} did, only with an extra "for_part" argument (blkget_{get,put} then call the __ version with a '0' for the extra argument). If for_part is 1, then the blkdev is being get(put) because a partition is being opened(closed) for the first(last) time, and so bd_part_count should be updated (on success). The particular advantage of pushing this function down is that the bd_mutex lock (which is needed to update bd_part_count) is already held at the lower level. Note that this slightly changes the semantics of bd_part_count. Instead of updating it whenever a partition is opened or released, it is now only updated on the first open or last release. This is an adequate semantic as it is only ever tested for "== 0". Having introduced these functions we remove the current bd_part_count updates from do_open (which is really the body of blkdev_get) and call __blkdev_get(... 1). Similarly in blkget_put we remove the old bd_part_count updates and call __blkget_put(..., 1). This call is moved to the end of __blkdev_put to avoid nested locks of bd_mutex. Finally the mutex_lock on whole->bd_mutex in do_open can be removed. It was only really needed to protect bd_part_count, and that is now managed (and protected) within the recursive call. The observation that bd_part_count is central to the locking issues, and the modifications to create __blkdev_put are from Peter Zijlstra. Cc: Ingo Molnar <mingo@elte.hu> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 11:36:16 +01:00
}
EXPORT_SYMBOL(blkdev_put);
static int blkdev_close(struct inode * inode, struct file * filp)
{
struct block_device *bdev = I_BDEV(filp->f_mapping->host);
block: make blkdev_get/put() handle exclusive access Over time, block layer has accumulated a set of APIs dealing with bdev open, close, claim and release. * blkdev_get/put() are the primary open and close functions. * bd_claim/release() deal with exclusive open. * open/close_bdev_exclusive() are combination of open and claim and the other way around, respectively. * bd_link/unlink_disk_holder() to create and remove holder/slave symlinks. * open_by_devnum() wraps bdget() + blkdev_get(). The interface is a bit confusing and the decoupling of open and claim makes it impossible to properly guarantee exclusive access as in-kernel open + claim sequence can disturb the existing exclusive open even before the block layer knows the current open if for another exclusive access. Reorganize the interface such that, * blkdev_get() is extended to include exclusive access management. @holder argument is added and, if is @FMODE_EXCL specified, it will gain exclusive access atomically w.r.t. other exclusive accesses. * blkdev_put() is similarly extended. It now takes @mode argument and if @FMODE_EXCL is set, it releases an exclusive access. Also, when the last exclusive claim is released, the holder/slave symlinks are removed automatically. * bd_claim/release() and close_bdev_exclusive() are no longer necessary and either made static or removed. * bd_link_disk_holder() remains the same but bd_unlink_disk_holder() is no longer necessary and removed. * open_bdev_exclusive() becomes a simple wrapper around lookup_bdev() and blkdev_get(). It also has an unexpected extra bdev_read_only() test which probably should be moved into blkdev_get(). * open_by_devnum() is modified to take @holder argument and pass it to blkdev_get(). Most of bdev open/close operations are unified into blkdev_get/put() and most exclusive accesses are tested atomically at the open time (as it should). This cleans up code and removes some, both valid and invalid, but unnecessary all the same, corner cases. open_bdev_exclusive() and open_by_devnum() can use further cleanup - rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop special features. Well, let's leave them for another day. Most conversions are straight-forward. drbd conversion is a bit more involved as there was some reordering, but the logic should stay the same. Signed-off-by: Tejun Heo <tj@kernel.org> Acked-by: Neil Brown <neilb@suse.de> Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Acked-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Philipp Reisner <philipp.reisner@linbit.com> Cc: Peter Osterlund <petero2@telia.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Jan Kara <jack@suse.cz> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <joel.becker@oracle.com> Cc: Alex Elder <aelder@sgi.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: dm-devel@redhat.com Cc: drbd-dev@lists.linbit.com Cc: Leo Chen <leochen@broadcom.com> Cc: Scott Branden <sbranden@broadcom.com> Cc: Chris Mason <chris.mason@oracle.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@linux.vnet.ibm.com> Cc: Joern Engel <joern@logfs.org> Cc: reiserfs-devel@vger.kernel.org Cc: Alexander Viro <viro@zeniv.linux.org.uk>
2010-11-13 11:55:17 +01:00
return blkdev_put(bdev, filp->f_mode);
}
static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg)
{
struct block_device *bdev = I_BDEV(file->f_mapping->host);
fmode_t mode = file->f_mode;
/*
* O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have
* to updated it before every ioctl.
*/
if (file->f_flags & O_NDELAY)
mode |= FMODE_NDELAY;
else
mode &= ~FMODE_NDELAY;
return blkdev_ioctl(bdev, mode, cmd, arg);
}
/*
* Write data to the block device. Only intended for the block device itself
* and the raw driver which basically is a fake block device.
*
* Does not take i_mutex for the write and thus is not for general purpose
* use.
*/
ssize_t blkdev_aio_write(struct kiocb *iocb, const struct iovec *iov,
unsigned long nr_segs, loff_t pos)
{
struct file *file = iocb->ki_filp;
ssize_t ret;
BUG_ON(iocb->ki_pos != pos);
ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos);
if (ret > 0 || ret == -EIOCBQUEUED) {
ssize_t err;
err = generic_write_sync(file, pos, ret);
if (err < 0 && ret > 0)
ret = err;
}
return ret;
}
EXPORT_SYMBOL_GPL(blkdev_aio_write);
/*
* Try to release a page associated with block device when the system
* is under memory pressure.
*/
static int blkdev_releasepage(struct page *page, gfp_t wait)
{
struct super_block *super = BDEV_I(page->mapping->host)->bdev.bd_super;
if (super && super->s_op->bdev_try_to_free_page)
return super->s_op->bdev_try_to_free_page(super, page, wait);
return try_to_free_buffers(page);
}
static const struct address_space_operations def_blk_aops = {
.readpage = blkdev_readpage,
.writepage = blkdev_writepage,
.write_begin = blkdev_write_begin,
.write_end = blkdev_write_end,
.writepages = generic_writepages,
.releasepage = blkdev_releasepage,
.direct_IO = blkdev_direct_IO,
};
const struct file_operations def_blk_fops = {
.open = blkdev_open,
.release = blkdev_close,
.llseek = block_llseek,
.read = do_sync_read,
.write = do_sync_write,
.aio_read = generic_file_aio_read,
.aio_write = blkdev_aio_write,
.mmap = generic_file_mmap,
.fsync = blkdev_fsync,
.unlocked_ioctl = block_ioctl,
#ifdef CONFIG_COMPAT
.compat_ioctl = compat_blkdev_ioctl,
#endif
.splice_read = generic_file_splice_read,
.splice_write = generic_file_splice_write,
};
int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg)
{
int res;
mm_segment_t old_fs = get_fs();
set_fs(KERNEL_DS);
res = blkdev_ioctl(bdev, 0, cmd, arg);
set_fs(old_fs);
return res;
}
EXPORT_SYMBOL(ioctl_by_bdev);
/**
* lookup_bdev - lookup a struct block_device by name
* @pathname: special file representing the block device
*
* Get a reference to the blockdevice at @pathname in the current
* namespace if possible and return it. Return ERR_PTR(error)
* otherwise.
*/
struct block_device *lookup_bdev(const char *pathname)
{
struct block_device *bdev;
struct inode *inode;
struct path path;
int error;
if (!pathname || !*pathname)
return ERR_PTR(-EINVAL);
error = kern_path(pathname, LOOKUP_FOLLOW, &path);
if (error)
return ERR_PTR(error);
inode = path.dentry->d_inode;
error = -ENOTBLK;
if (!S_ISBLK(inode->i_mode))
goto fail;
error = -EACCES;
if (path.mnt->mnt_flags & MNT_NODEV)
goto fail;
error = -ENOMEM;
bdev = bd_acquire(inode);
if (!bdev)
goto fail;
out:
path_put(&path);
return bdev;
fail:
bdev = ERR_PTR(error);
goto out;
}
EXPORT_SYMBOL(lookup_bdev);
Fix over-zealous flush_disk when changing device size. There are two cases when we call flush_disk. In one, the device has disappeared (check_disk_change) so any data will hold becomes irrelevant. In the oter, the device has changed size (check_disk_size_change) so data we hold may be irrelevant. In both cases it makes sense to discard any 'clean' buffers, so they will be read back from the device if needed. In the former case it makes sense to discard 'dirty' buffers as there will never be anywhere safe to write the data. In the second case it *does*not* make sense to discard dirty buffers as that will lead to file system corruption when you simply enlarge the containing devices. flush_disk calls __invalidate_devices. __invalidate_device calls both invalidate_inodes and invalidate_bdev. invalidate_inodes *does* discard I_DIRTY inodes and this does lead to fs corruption. invalidate_bev *does*not* discard dirty pages, but I don't really care about that at present. So this patch adds a flag to __invalidate_device (calling it __invalidate_device2) to indicate whether dirty buffers should be killed, and this is passed to invalidate_inodes which can choose to skip dirty inodes. flusk_disk then passes true from check_disk_change and false from check_disk_size_change. dm avoids tripping over this problem by calling i_size_write directly rathher than using check_disk_size_change. md does use check_disk_size_change and so is affected. This regression was introduced by commit 608aeef17a which causes check_disk_size_change to call flush_disk, so it is suitable for any kernel since 2.6.27. Cc: stable@kernel.org Acked-by: Jeff Moyer <jmoyer@redhat.com> Cc: Andrew Patterson <andrew.patterson@hp.com> Cc: Jens Axboe <axboe@kernel.dk> Signed-off-by: NeilBrown <neilb@suse.de>
2011-02-24 07:25:47 +01:00
int __invalidate_device(struct block_device *bdev, bool kill_dirty)
{
struct super_block *sb = get_super(bdev);
int res = 0;
if (sb) {
/*
* no need to lock the super, get_super holds the
* read mutex so the filesystem cannot go away
* under us (->put_super runs with the write lock
* hold).
*/
shrink_dcache_sb(sb);
Fix over-zealous flush_disk when changing device size. There are two cases when we call flush_disk. In one, the device has disappeared (check_disk_change) so any data will hold becomes irrelevant. In the oter, the device has changed size (check_disk_size_change) so data we hold may be irrelevant. In both cases it makes sense to discard any 'clean' buffers, so they will be read back from the device if needed. In the former case it makes sense to discard 'dirty' buffers as there will never be anywhere safe to write the data. In the second case it *does*not* make sense to discard dirty buffers as that will lead to file system corruption when you simply enlarge the containing devices. flush_disk calls __invalidate_devices. __invalidate_device calls both invalidate_inodes and invalidate_bdev. invalidate_inodes *does* discard I_DIRTY inodes and this does lead to fs corruption. invalidate_bev *does*not* discard dirty pages, but I don't really care about that at present. So this patch adds a flag to __invalidate_device (calling it __invalidate_device2) to indicate whether dirty buffers should be killed, and this is passed to invalidate_inodes which can choose to skip dirty inodes. flusk_disk then passes true from check_disk_change and false from check_disk_size_change. dm avoids tripping over this problem by calling i_size_write directly rathher than using check_disk_size_change. md does use check_disk_size_change and so is affected. This regression was introduced by commit 608aeef17a which causes check_disk_size_change to call flush_disk, so it is suitable for any kernel since 2.6.27. Cc: stable@kernel.org Acked-by: Jeff Moyer <jmoyer@redhat.com> Cc: Andrew Patterson <andrew.patterson@hp.com> Cc: Jens Axboe <axboe@kernel.dk> Signed-off-by: NeilBrown <neilb@suse.de>
2011-02-24 07:25:47 +01:00
res = invalidate_inodes(sb, kill_dirty);
drop_super(sb);
}
invalidate_bdev(bdev);
return res;
}
EXPORT_SYMBOL(__invalidate_device);