2005-04-17 02:20:36 +04:00
/*
* 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>
2008-06-06 09:46:27 +04:00
# include <linux/device_cgroup.h>
2005-04-17 02:20:36 +04:00
# include <linux/highmem.h>
# include <linux/blkdev.h>
2015-05-23 00:13:32 +03:00
# include <linux/backing-dev.h>
2005-04-17 02:20:36 +04:00
# include <linux/module.h>
# include <linux/blkpg.h>
2012-03-24 02:01:50 +04:00
# include <linux/magic.h>
2017-01-27 07:37:35 +03:00
# include <linux/dax.h>
2005-04-17 02:20:36 +04:00
# include <linux/buffer_head.h>
2011-09-16 10:31:11 +04:00
# include <linux/swap.h>
2009-02-25 12:44:19 +03:00
# include <linux/pagevec.h>
2006-08-29 22:06:09 +04:00
# include <linux/writeback.h>
2005-04-17 02:20:36 +04:00
# include <linux/mpage.h>
# include <linux/mount.h>
# include <linux/uio.h>
# include <linux/namei.h>
2007-05-08 11:24:32 +04:00
# include <linux/log2.h>
2011-09-16 10:31:11 +04:00
# include <linux/cleancache.h>
2015-09-09 00:58:40 +03:00
# include <linux/dax.h>
2016-05-07 21:40:28 +03:00
# include <linux/badblocks.h>
2016-10-31 20:59:25 +03:00
# include <linux/task_io_accounting_ops.h>
2016-10-11 23:51:11 +03:00
# include <linux/falloc.h>
2016-12-24 22:46:01 +03:00
# include <linux/uaccess.h>
2006-09-30 22:52:18 +04:00
# include "internal.h"
2005-04-17 02:20:36 +04:00
struct bdev_inode {
struct block_device bdev ;
struct inode vfs_inode ;
} ;
2008-02-18 15:48:31 +03:00
static const struct address_space_operations def_blk_aops ;
2005-04-17 02:20:36 +04:00
static inline struct bdev_inode * BDEV_I ( struct inode * inode )
{
return container_of ( inode , struct bdev_inode , vfs_inode ) ;
}
2015-06-26 14:58:32 +03:00
struct block_device * I_BDEV ( struct inode * inode )
2005-04-17 02:20:36 +04:00
{
return & BDEV_I ( inode ) - > bdev ;
}
EXPORT_SYMBOL ( I_BDEV ) ;
2015-11-09 19:23:40 +03:00
static void bdev_write_inode ( struct block_device * bdev )
2015-01-14 12:42:33 +03:00
{
2015-11-09 19:23:40 +03:00
struct inode * inode = bdev - > bd_inode ;
int ret ;
2015-01-14 12:42:33 +03:00
spin_lock ( & inode - > i_lock ) ;
while ( inode - > i_state & I_DIRTY ) {
spin_unlock ( & inode - > i_lock ) ;
2015-11-09 19:23:40 +03:00
ret = write_inode_now ( inode , true ) ;
if ( ret ) {
char name [ BDEVNAME_SIZE ] ;
pr_warn_ratelimited ( " VFS: Dirty inode writeback failed "
" for block device %s (err=%d). \n " ,
bdevname ( bdev , name ) , ret ) ;
}
2015-01-14 12:42:33 +03:00
spin_lock ( & inode - > i_lock ) ;
}
spin_unlock ( & inode - > i_lock ) ;
}
2007-05-07 01:49:55 +04:00
/* Kill _all_ buffers and pagecache , dirty or not.. */
2011-09-16 10:31:11 +04:00
void kill_bdev ( struct block_device * bdev )
2005-04-17 02:20:36 +04:00
{
2011-09-16 10:31:11 +04:00
struct address_space * mapping = bdev - > bd_inode - > i_mapping ;
2016-01-23 02:10:40 +03:00
if ( mapping - > nrpages = = 0 & & mapping - > nrexceptional = = 0 )
2007-05-07 01:49:55 +04:00
return ;
2011-09-16 10:31:11 +04:00
2007-05-07 01:49:55 +04:00
invalidate_bh_lrus ( ) ;
2011-09-16 10:31:11 +04:00
truncate_inode_pages ( mapping , 0 ) ;
2005-04-17 02:20:36 +04:00
}
2011-09-16 10:31:11 +04:00
EXPORT_SYMBOL ( kill_bdev ) ;
/* Invalidate clean unused buffers and pagecache. */
void invalidate_bdev ( struct block_device * bdev )
{
struct address_space * mapping = bdev - > bd_inode - > i_mapping ;
2017-05-04 00:56:02 +03:00
if ( mapping - > nrpages ) {
invalidate_bh_lrus ( ) ;
lru_add_drain_all ( ) ; /* make sure all lru add caches are flushed */
invalidate_mapping_pages ( mapping , 0 , - 1 ) ;
}
2011-09-16 10:31:11 +04:00
/* 99% of the time, we don't need to flush the cleancache on the bdev.
* But , for the strange corners , lets be cautious
*/
2011-09-21 19:56:28 +04:00
cleancache_invalidate_inode ( mapping ) ;
2011-09-16 10:31:11 +04:00
}
EXPORT_SYMBOL ( invalidate_bdev ) ;
2005-04-17 02:20:36 +04:00
int set_blocksize ( struct block_device * bdev , int size )
{
/* Size must be a power of two, and between 512 and PAGE_SIZE */
2007-05-08 11:24:32 +04:00
if ( size > PAGE_SIZE | | size < 512 | | ! is_power_of_2 ( size ) )
2005-04-17 02:20:36 +04:00
return - EINVAL ;
/* Size cannot be smaller than the size supported by the device */
2009-05-23 01:17:49 +04:00
if ( size < bdev_logical_block_size ( bdev ) )
2005-04-17 02:20:36 +04:00
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 ;
2006-03-24 14:18:05 +03:00
sb - > s_blocksize_bits = blksize_bits ( size ) ;
2005-04-17 02:20:36 +04:00
return sb - > s_blocksize ;
}
EXPORT_SYMBOL ( sb_set_blocksize ) ;
int sb_min_blocksize ( struct super_block * sb , int size )
{
2009-05-23 01:17:49 +04:00
int minsize = bdev_logical_block_size ( sb - > s_bdev ) ;
2005-04-17 02:20:36 +04:00
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 )
{
bh - > b_bdev = I_BDEV ( inode ) ;
bh - > b_blocknr = iblock ;
set_buffer_mapped ( bh ) ;
return 0 ;
}
2015-10-28 01:48:19 +03:00
static struct inode * bdev_file_inode ( struct file * file )
{
return file - > f_mapping - > host ;
}
2016-11-17 19:50:47 +03:00
static unsigned int dio_bio_write_op ( struct kiocb * iocb )
{
unsigned int op = REQ_OP_WRITE | REQ_SYNC | REQ_IDLE ;
/* avoid the need for a I/O completion work item */
if ( iocb - > ki_flags & IOCB_DSYNC )
op | = REQ_FUA ;
return op ;
}
2016-10-31 20:59:25 +03:00
# define DIO_INLINE_BIO_VECS 4
static void blkdev_bio_end_io_simple ( struct bio * bio )
{
struct task_struct * waiter = bio - > bi_private ;
WRITE_ONCE ( bio - > bi_private , NULL ) ;
wake_up_process ( waiter ) ;
}
static ssize_t
__blkdev_direct_IO_simple ( struct kiocb * iocb , struct iov_iter * iter ,
int nr_pages )
{
struct file * file = iocb - > ki_filp ;
struct block_device * bdev = I_BDEV ( bdev_file_inode ( file ) ) ;
2016-11-17 09:11:42 +03:00
struct bio_vec inline_vecs [ DIO_INLINE_BIO_VECS ] , * vecs , * bvec ;
2016-10-31 20:59:25 +03:00
loff_t pos = iocb - > ki_pos ;
bool should_dirty = false ;
struct bio bio ;
ssize_t ret ;
blk_qc_t qc ;
int i ;
2016-11-22 18:12:39 +03:00
if ( ( pos | iov_iter_alignment ( iter ) ) &
( bdev_logical_block_size ( bdev ) - 1 ) )
2016-10-31 20:59:25 +03:00
return - EINVAL ;
2016-11-17 09:11:42 +03:00
if ( nr_pages < = DIO_INLINE_BIO_VECS )
vecs = inline_vecs ;
else {
vecs = kmalloc ( nr_pages * sizeof ( struct bio_vec ) , GFP_KERNEL ) ;
if ( ! vecs )
return - ENOMEM ;
}
2016-11-22 18:57:21 +03:00
bio_init ( & bio , vecs , nr_pages ) ;
2017-08-23 20:10:32 +03:00
bio_set_dev ( & bio , bdev ) ;
2016-11-22 09:38:49 +03:00
bio . bi_iter . bi_sector = pos > > 9 ;
2017-06-27 20:01:22 +03:00
bio . bi_write_hint = iocb - > ki_hint ;
2016-10-31 20:59:25 +03:00
bio . bi_private = current ;
bio . bi_end_io = blkdev_bio_end_io_simple ;
ret = bio_iov_iter_get_pages ( & bio , iter ) ;
if ( unlikely ( ret ) )
return ret ;
ret = bio . bi_iter . bi_size ;
if ( iov_iter_rw ( iter ) = = READ ) {
2016-11-17 19:50:47 +03:00
bio . bi_opf = REQ_OP_READ ;
2016-10-31 20:59:25 +03:00
if ( iter_is_iovec ( iter ) )
should_dirty = true ;
} else {
2016-11-17 19:50:47 +03:00
bio . bi_opf = dio_bio_write_op ( iocb ) ;
2016-10-31 20:59:25 +03:00
task_io_account_write ( ret ) ;
}
qc = submit_bio ( & bio ) ;
for ( ; ; ) {
set_current_state ( TASK_UNINTERRUPTIBLE ) ;
if ( ! READ_ONCE ( bio . bi_private ) )
break ;
if ( ! ( iocb - > ki_flags & IOCB_HIPRI ) | |
2017-11-02 21:29:54 +03:00
! blk_poll ( bdev_get_queue ( bdev ) , qc ) )
2016-10-31 20:59:25 +03:00
io_schedule ( ) ;
}
__set_current_state ( TASK_RUNNING ) ;
bio_for_each_segment_all ( bvec , & bio , i ) {
if ( should_dirty & & ! PageCompound ( bvec - > bv_page ) )
set_page_dirty_lock ( bvec - > bv_page ) ;
put_page ( bvec - > bv_page ) ;
}
2016-11-17 09:11:42 +03:00
if ( vecs ! = inline_vecs )
kfree ( vecs ) ;
2017-06-03 10:38:06 +03:00
if ( unlikely ( bio . bi_status ) )
Merge branch 'for-4.13/block' of git://git.kernel.dk/linux-block
Pull core block/IO updates from Jens Axboe:
"This is the main pull request for the block layer for 4.13. Not a huge
round in terms of features, but there's a lot of churn related to some
core cleanups.
Note this depends on the UUID tree pull request, that Christoph
already sent out.
This pull request contains:
- A series from Christoph, unifying the error/stats codes in the
block layer. We now use blk_status_t everywhere, instead of using
different schemes for different places.
- Also from Christoph, some cleanups around request allocation and IO
scheduler interactions in blk-mq.
- And yet another series from Christoph, cleaning up how we handle
and do bounce buffering in the block layer.
- A blk-mq debugfs series from Bart, further improving on the support
we have for exporting internal information to aid debugging IO
hangs or stalls.
- Also from Bart, a series that cleans up the request initialization
differences across types of devices.
- A series from Goldwyn Rodrigues, allowing the block layer to return
failure if we will block and the user asked for non-blocking.
- Patch from Hannes for supporting setting loop devices block size to
that of the underlying device.
- Two series of patches from Javier, fixing various issues with
lightnvm, particular around pblk.
- A series from me, adding support for write hints. This comes with
NVMe support as well, so applications can help guide data placement
on flash to improve performance, latencies, and write
amplification.
- A series from Ming, improving and hardening blk-mq support for
stopping/starting and quiescing hardware queues.
- Two pull requests for NVMe updates. Nothing major on the feature
side, but lots of cleanups and bug fixes. From the usual crew.
- A series from Neil Brown, greatly improving the bio rescue set
support. Most notably, this kills the bio rescue work queues, if we
don't really need them.
- Lots of other little bug fixes that are all over the place"
* 'for-4.13/block' of git://git.kernel.dk/linux-block: (217 commits)
lightnvm: pblk: set line bitmap check under debug
lightnvm: pblk: verify that cache read is still valid
lightnvm: pblk: add initialization check
lightnvm: pblk: remove target using async. I/Os
lightnvm: pblk: use vmalloc for GC data buffer
lightnvm: pblk: use right metadata buffer for recovery
lightnvm: pblk: schedule if data is not ready
lightnvm: pblk: remove unused return variable
lightnvm: pblk: fix double-free on pblk init
lightnvm: pblk: fix bad le64 assignations
nvme: Makefile: remove dead build rule
blk-mq: map all HWQ also in hyperthreaded system
nvmet-rdma: register ib_client to not deadlock in device removal
nvme_fc: fix error recovery on link down.
nvmet_fc: fix crashes on bad opcodes
nvme_fc: Fix crash when nvme controller connection fails.
nvme_fc: replace ioabort msleep loop with completion
nvme_fc: fix double calls to nvme_cleanup_cmd()
nvme-fabrics: verify that a controller returns the correct NQN
nvme: simplify nvme_dev_attrs_are_visible
...
2017-07-03 20:34:51 +03:00
ret = blk_status_to_errno ( bio . bi_status ) ;
2017-06-29 00:30:13 +03:00
bio_uninit ( & bio ) ;
2016-10-31 20:59:25 +03:00
return ret ;
}
2016-11-17 09:14:22 +03:00
struct blkdev_dio {
union {
struct kiocb * iocb ;
struct task_struct * waiter ;
} ;
size_t size ;
atomic_t ref ;
bool multi_bio : 1 ;
bool should_dirty : 1 ;
bool is_sync : 1 ;
struct bio bio ;
} ;
static struct bio_set * blkdev_dio_pool __read_mostly ;
static void blkdev_bio_end_io ( struct bio * bio )
{
struct blkdev_dio * dio = bio - > bi_private ;
bool should_dirty = dio - > should_dirty ;
if ( dio - > multi_bio & & ! atomic_dec_and_test ( & dio - > ref ) ) {
2017-06-03 10:38:06 +03:00
if ( bio - > bi_status & & ! dio - > bio . bi_status )
dio - > bio . bi_status = bio - > bi_status ;
2016-11-17 09:14:22 +03:00
} else {
if ( ! dio - > is_sync ) {
struct kiocb * iocb = dio - > iocb ;
2017-06-03 10:38:06 +03:00
ssize_t ret ;
2016-11-17 09:14:22 +03:00
2017-06-03 10:38:06 +03:00
if ( likely ( ! dio - > bio . bi_status ) ) {
2016-11-17 09:14:22 +03:00
ret = dio - > size ;
iocb - > ki_pos + = ret ;
2017-06-03 10:38:06 +03:00
} else {
ret = blk_status_to_errno ( dio - > bio . bi_status ) ;
2016-11-17 09:14:22 +03:00
}
dio - > iocb - > ki_complete ( iocb , ret , 0 ) ;
bio_put ( & dio - > bio ) ;
} else {
struct task_struct * waiter = dio - > waiter ;
WRITE_ONCE ( dio - > waiter , NULL ) ;
wake_up_process ( waiter ) ;
}
}
if ( should_dirty ) {
bio_check_pages_dirty ( bio ) ;
} else {
struct bio_vec * bvec ;
int i ;
bio_for_each_segment_all ( bvec , bio , i )
put_page ( bvec - > bv_page ) ;
bio_put ( bio ) ;
}
}
2007-02-03 12:14:01 +03:00
static ssize_t
2016-11-17 09:14:22 +03:00
__blkdev_direct_IO ( struct kiocb * iocb , struct iov_iter * iter , int nr_pages )
2007-02-03 12:14:01 +03:00
{
struct file * file = iocb - > ki_filp ;
2015-10-28 01:48:19 +03:00
struct inode * inode = bdev_file_inode ( file ) ;
2016-11-17 09:14:22 +03:00
struct block_device * bdev = I_BDEV ( inode ) ;
2016-12-22 21:20:45 +03:00
struct blk_plug plug ;
2016-11-17 09:14:22 +03:00
struct blkdev_dio * dio ;
struct bio * bio ;
2017-01-24 16:50:19 +03:00
bool is_read = ( iov_iter_rw ( iter ) = = READ ) , is_sync ;
2016-11-17 09:14:22 +03:00
loff_t pos = iocb - > ki_pos ;
blk_qc_t qc = BLK_QC_T_NONE ;
2017-06-03 10:38:00 +03:00
int ret = 0 ;
2016-11-17 09:14:22 +03:00
2016-11-22 18:12:39 +03:00
if ( ( pos | iov_iter_alignment ( iter ) ) &
( bdev_logical_block_size ( bdev ) - 1 ) )
2016-11-17 09:14:22 +03:00
return - EINVAL ;
bio = bio_alloc_bioset ( GFP_KERNEL , nr_pages , blkdev_dio_pool ) ;
bio_get ( bio ) ; /* extra ref for the completion handler */
dio = container_of ( bio , struct blkdev_dio , bio ) ;
2017-01-24 16:50:19 +03:00
dio - > is_sync = is_sync = is_sync_kiocb ( iocb ) ;
2016-11-17 09:14:22 +03:00
if ( dio - > is_sync )
dio - > waiter = current ;
else
dio - > iocb = iocb ;
dio - > size = 0 ;
dio - > multi_bio = false ;
dio - > should_dirty = is_read & & ( iter - > type = = ITER_IOVEC ) ;
2016-12-22 21:20:45 +03:00
blk_start_plug ( & plug ) ;
2016-11-17 09:14:22 +03:00
for ( ; ; ) {
2017-08-23 20:10:32 +03:00
bio_set_dev ( bio , bdev ) ;
2016-11-22 09:38:49 +03:00
bio - > bi_iter . bi_sector = pos > > 9 ;
2017-06-27 20:01:22 +03:00
bio - > bi_write_hint = iocb - > ki_hint ;
2016-11-17 09:14:22 +03:00
bio - > bi_private = dio ;
bio - > bi_end_io = blkdev_bio_end_io ;
ret = bio_iov_iter_get_pages ( bio , iter ) ;
if ( unlikely ( ret ) ) {
2017-06-03 10:38:06 +03:00
bio - > bi_status = BLK_STS_IOERR ;
2016-11-17 09:14:22 +03:00
bio_endio ( bio ) ;
break ;
}
if ( is_read ) {
bio - > bi_opf = REQ_OP_READ ;
if ( dio - > should_dirty )
bio_set_pages_dirty ( bio ) ;
} else {
bio - > bi_opf = dio_bio_write_op ( iocb ) ;
task_io_account_write ( bio - > bi_iter . bi_size ) ;
}
dio - > size + = bio - > bi_iter . bi_size ;
pos + = bio - > bi_iter . bi_size ;
nr_pages = iov_iter_npages ( iter , BIO_MAX_PAGES ) ;
if ( ! nr_pages ) {
qc = submit_bio ( bio ) ;
break ;
}
if ( ! dio - > multi_bio ) {
dio - > multi_bio = true ;
atomic_set ( & dio - > ref , 2 ) ;
} else {
atomic_inc ( & dio - > ref ) ;
}
submit_bio ( bio ) ;
bio = bio_alloc ( GFP_KERNEL , nr_pages ) ;
}
2016-12-22 21:20:45 +03:00
blk_finish_plug ( & plug ) ;
2016-11-17 09:14:22 +03:00
2017-01-24 16:50:19 +03:00
if ( ! is_sync )
2016-11-17 09:14:22 +03:00
return - EIOCBQUEUED ;
for ( ; ; ) {
set_current_state ( TASK_UNINTERRUPTIBLE ) ;
if ( ! READ_ONCE ( dio - > waiter ) )
break ;
if ( ! ( iocb - > ki_flags & IOCB_HIPRI ) | |
2017-11-02 21:29:54 +03:00
! blk_poll ( bdev_get_queue ( bdev ) , qc ) )
2016-11-17 09:14:22 +03:00
io_schedule ( ) ;
}
__set_current_state ( TASK_RUNNING ) ;
2017-06-03 10:38:00 +03:00
if ( ! ret )
2017-06-03 10:38:06 +03:00
ret = blk_status_to_errno ( dio - > bio . bi_status ) ;
2016-12-14 06:07:55 +03:00
if ( likely ( ! ret ) )
2016-11-17 09:14:22 +03:00
ret = dio - > size ;
bio_put ( & dio - > bio ) ;
return ret ;
}
static ssize_t
blkdev_direct_IO ( struct kiocb * iocb , struct iov_iter * iter )
{
2016-10-31 20:59:25 +03:00
int nr_pages ;
2007-02-03 12:14:01 +03:00
2016-11-17 09:11:42 +03:00
nr_pages = iov_iter_npages ( iter , BIO_MAX_PAGES + 1 ) ;
2016-10-31 20:59:25 +03:00
if ( ! nr_pages )
return 0 ;
2016-11-17 09:11:42 +03:00
if ( is_sync_kiocb ( iocb ) & & nr_pages < = BIO_MAX_PAGES )
2016-10-31 20:59:25 +03:00
return __blkdev_direct_IO_simple ( iocb , iter , nr_pages ) ;
2016-11-17 09:14:22 +03:00
return __blkdev_direct_IO ( iocb , iter , min ( nr_pages , BIO_MAX_PAGES ) ) ;
}
static __init int blkdev_init ( void )
{
2017-06-18 07:38:57 +03:00
blkdev_dio_pool = bioset_create ( 4 , offsetof ( struct blkdev_dio , bio ) , BIOSET_NEED_BVECS ) ;
2016-11-17 09:14:22 +03:00
if ( ! blkdev_dio_pool )
return - ENOMEM ;
return 0 ;
2007-02-03 12:14:01 +03:00
}
2016-11-17 09:14:22 +03:00
module_init ( blkdev_init ) ;
2007-02-03 12:14:01 +03:00
2009-04-27 18:43:51 +04:00
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 ) ;
}
2009-02-25 12:44:19 +03:00
/*
* 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 )
{
2009-04-27 18:43:51 +04:00
return __sync_blockdev ( bdev , 1 ) ;
2009-02-25 12:44:19 +03:00
}
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 ) {
2009-04-27 18:43:53 +04:00
int res = sync_filesystem ( sb ) ;
2009-02-25 12:44:19 +03:00
drop_super ( sb ) ;
return res ;
}
return sync_blockdev ( bdev ) ;
}
2009-04-01 15:07:16 +04:00
EXPORT_SYMBOL ( fsync_bdev ) ;
2009-02-25 12:44:19 +03:00
/**
* 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 ) ;
2009-08-04 01:28:35 +04:00
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 .
*/
2009-02-25 12:44:19 +03:00
sb = get_super ( bdev ) ;
2016-08-23 18:55:31 +03:00
if ( sb )
drop_super ( sb ) ;
2009-08-04 01:28:35 +04:00
mutex_unlock ( & bdev - > bd_fsfreeze_mutex ) ;
return sb ;
}
sb = get_active_super ( bdev ) ;
if ( ! sb )
goto out ;
fs: add freeze_super/thaw_super fs hooks
Currently, freezing a filesystem involves calling freeze_super, which locks
sb->s_umount and then calls the fs-specific freeze_fs hook. This makes it
hard for gfs2 (and potentially other cluster filesystems) to use the vfs
freezing code to do freezes on all the cluster nodes.
In order to communicate that a freeze has been requested, and to make sure
that only one node is trying to freeze at a time, gfs2 uses a glock
(sd_freeze_gl). The problem is that there is no hook for gfs2 to acquire
this lock before calling freeze_super. This means that two nodes can
attempt to freeze the filesystem by both calling freeze_super, acquiring
the sb->s_umount lock, and then attempting to grab the cluster glock
sd_freeze_gl. Only one will succeed, and the other will be stuck in
freeze_super, making it impossible to finish freezing the node.
To solve this problem, this patch adds the freeze_super and thaw_super
hooks. If a filesystem implements these hooks, they are called instead of
the vfs freeze_super and thaw_super functions. This means that every
filesystem that implements these hooks must call the vfs freeze_super and
thaw_super functions itself within the hook function to make use of the vfs
freezing code.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2014-11-14 05:42:03 +03:00
if ( sb - > s_op - > freeze_super )
error = sb - > s_op - > freeze_super ( sb ) ;
else
error = freeze_super ( sb ) ;
2010-03-23 17:34:56 +03:00
if ( error ) {
deactivate_super ( sb ) ;
bdev - > bd_fsfreeze_count - - ;
2009-02-25 12:44:19 +03:00
mutex_unlock ( & bdev - > bd_fsfreeze_mutex ) ;
2010-03-23 17:34:56 +03:00
return ERR_PTR ( error ) ;
2009-02-25 12:44:19 +03:00
}
2010-03-23 17:34:56 +03:00
deactivate_super ( sb ) ;
2009-08-04 01:28:35 +04:00
out :
2009-02-25 12:44:19 +03:00
sync_blockdev ( bdev ) ;
mutex_unlock ( & bdev - > bd_fsfreeze_mutex ) ;
2009-08-04 01:28:06 +04:00
return sb ; /* thaw_bdev releases s->s_umount */
2009-02-25 12:44:19 +03:00
}
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 )
{
2009-08-04 01:28:35 +04:00
int error = - EINVAL ;
2009-02-25 12:44:19 +03:00
mutex_lock ( & bdev - > bd_fsfreeze_mutex ) ;
2009-08-04 01:28:35 +04:00
if ( ! bdev - > bd_fsfreeze_count )
2010-03-23 17:34:56 +03:00
goto out ;
2009-08-04 01:28:35 +04:00
error = 0 ;
if ( - - bdev - > bd_fsfreeze_count > 0 )
2010-03-23 17:34:56 +03:00
goto out ;
2009-08-04 01:28:35 +04:00
if ( ! sb )
2010-03-23 17:34:56 +03:00
goto out ;
2009-08-04 01:28:35 +04:00
fs: add freeze_super/thaw_super fs hooks
Currently, freezing a filesystem involves calling freeze_super, which locks
sb->s_umount and then calls the fs-specific freeze_fs hook. This makes it
hard for gfs2 (and potentially other cluster filesystems) to use the vfs
freezing code to do freezes on all the cluster nodes.
In order to communicate that a freeze has been requested, and to make sure
that only one node is trying to freeze at a time, gfs2 uses a glock
(sd_freeze_gl). The problem is that there is no hook for gfs2 to acquire
this lock before calling freeze_super. This means that two nodes can
attempt to freeze the filesystem by both calling freeze_super, acquiring
the sb->s_umount lock, and then attempting to grab the cluster glock
sd_freeze_gl. Only one will succeed, and the other will be stuck in
freeze_super, making it impossible to finish freezing the node.
To solve this problem, this patch adds the freeze_super and thaw_super
hooks. If a filesystem implements these hooks, they are called instead of
the vfs freeze_super and thaw_super functions. This means that every
filesystem that implements these hooks must call the vfs freeze_super and
thaw_super functions itself within the hook function to make use of the vfs
freezing code.
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2014-11-14 05:42:03 +03:00
if ( sb - > s_op - > thaw_super )
error = sb - > s_op - > thaw_super ( sb ) ;
else
error = thaw_super ( sb ) ;
2016-10-04 11:53:40 +03:00
if ( error )
2010-03-23 17:34:56 +03:00
bdev - > bd_fsfreeze_count + + ;
out :
2009-02-25 12:44:19 +03:00
mutex_unlock ( & bdev - > bd_fsfreeze_mutex ) ;
2016-10-04 11:53:40 +03:00
return error ;
2009-02-25 12:44:19 +03:00
}
EXPORT_SYMBOL ( thaw_bdev ) ;
2005-04-17 02:20:36 +04:00
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 ) ;
}
2014-10-10 02:26:58 +04:00
static int blkdev_readpages ( struct file * file , struct address_space * mapping ,
struct list_head * pages , unsigned nr_pages )
{
return mpage_readpages ( mapping , pages , nr_pages , blkdev_get_block ) ;
}
2007-10-16 12:25:04 +04:00
static int blkdev_write_begin ( struct file * file , struct address_space * mapping ,
loff_t pos , unsigned len , unsigned flags ,
struct page * * pagep , void * * fsdata )
2005-04-17 02:20:36 +04:00
{
2010-06-04 13:29:58 +04:00
return block_write_begin ( mapping , pos , len , flags , pagep ,
blkdev_get_block ) ;
2005-04-17 02:20:36 +04:00
}
2007-10-16 12:25:04 +04:00
static int blkdev_write_end ( struct file * file , struct address_space * mapping ,
loff_t pos , unsigned len , unsigned copied ,
struct page * page , void * fsdata )
2005-04-17 02:20:36 +04:00
{
2007-10-16 12:25:04 +04:00
int ret ;
ret = block_write_end ( file , mapping , pos , len , copied , page , fsdata ) ;
unlock_page ( page ) ;
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
put_page ( page ) ;
2007-10-16 12:25:04 +04:00
return ret ;
2005-04-17 02:20:36 +04:00
}
/*
* private llseek :
2013-01-24 02:07:38 +04:00
* for a block special file file_inode ( file ) - > i_size is zero
2005-04-17 02:20:36 +04:00
* so we compute the size by hand ( just as in block_read / write above )
*/
2012-12-18 03:59:39 +04:00
static loff_t block_llseek ( struct file * file , loff_t offset , int whence )
2005-04-17 02:20:36 +04:00
{
2015-10-28 01:48:19 +03:00
struct inode * bd_inode = bdev_file_inode ( file ) ;
2005-04-17 02:20:36 +04:00
loff_t retval ;
2016-01-22 23:40:57 +03:00
inode_lock ( bd_inode ) ;
2013-06-23 21:34:45 +04:00
retval = fixed_size_llseek ( file , offset , whence , i_size_read ( bd_inode ) ) ;
2016-01-22 23:40:57 +03:00
inode_unlock ( bd_inode ) ;
2005-04-17 02:20:36 +04:00
return retval ;
}
2011-07-17 04:44:56 +04:00
int blkdev_fsync ( struct file * filp , loff_t start , loff_t end , int datasync )
2005-04-17 02:20:36 +04:00
{
2015-10-28 01:48:19 +03:00
struct inode * bd_inode = bdev_file_inode ( filp ) ;
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 21:18:06 +04:00
struct block_device * bdev = I_BDEV ( bd_inode ) ;
2009-10-29 16:14:04 +03:00
int error ;
2011-08-02 04:17:48 +04:00
2017-07-06 14:02:28 +03:00
error = file_write_and_wait_range ( filp , start , end ) ;
2011-08-02 04:17:48 +04:00
if ( error )
return error ;
2009-10-29 16:14:04 +03:00
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 21:18:06 +04:00
/*
* 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 .
*/
2010-09-16 22:51:46 +04:00
error = blkdev_issue_flush ( bdev , GFP_KERNEL , NULL ) ;
2009-10-29 16:14:04 +03:00
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 21:18:06 +04:00
2009-10-29 16:14:04 +03:00
return error ;
2005-04-17 02:20:36 +04:00
}
2010-04-07 01:35:00 +04:00
EXPORT_SYMBOL ( blkdev_fsync ) ;
2005-04-17 02:20:36 +04:00
2014-06-05 03:07:46 +04:00
/**
* bdev_read_page ( ) - Start reading a page from a block device
* @ bdev : The device to read the page from
* @ sector : The offset on the device to read the page to ( need not be aligned )
* @ page : The page to read
*
* On entry , the page should be locked . It will be unlocked when the page
* has been read . If the block driver implements rw_page synchronously ,
* that will be true on exit from this function , but it need not be .
*
* Errors returned by this function are usually " soft " , eg out of memory , or
* queue full ; callers should try a different route to read this page rather
* than propagate an error back up the stack .
*
* Return : negative errno if an error occurs , 0 if submission was successful .
*/
int bdev_read_page ( struct block_device * bdev , sector_t sector ,
struct page * page )
{
const struct block_device_operations * ops = bdev - > bd_disk - > fops ;
2015-11-20 00:29:28 +03:00
int result = - EOPNOTSUPP ;
2015-05-12 20:48:53 +03:00
if ( ! ops - > rw_page | | bdev_get_integrity ( bdev ) )
2015-11-20 00:29:28 +03:00
return result ;
2017-11-09 21:49:58 +03:00
result = blk_queue_enter ( bdev - > bd_queue , 0 ) ;
2015-11-20 00:29:28 +03:00
if ( result )
return result ;
2016-08-05 17:11:04 +03:00
result = ops - > rw_page ( bdev , sector + get_start_sect ( bdev ) , page , false ) ;
2015-11-20 00:29:28 +03:00
blk_queue_exit ( bdev - > bd_queue ) ;
return result ;
2014-06-05 03:07:46 +04:00
}
EXPORT_SYMBOL_GPL ( bdev_read_page ) ;
/**
* bdev_write_page ( ) - Start writing a page to a block device
* @ bdev : The device to write the page to
* @ sector : The offset on the device to write the page to ( need not be aligned )
* @ page : The page to write
* @ wbc : The writeback_control for the write
*
* On entry , the page should be locked and not currently under writeback .
* On exit , if the write started successfully , the page will be unlocked and
* under writeback . If the write failed already ( eg the driver failed to
* queue the page to the device ) , the page will still be locked . If the
* caller is a - > writepage implementation , it will need to unlock the page .
*
* Errors returned by this function are usually " soft " , eg out of memory , or
* queue full ; callers should try a different route to write this page rather
* than propagate an error back up the stack .
*
* Return : negative errno if an error occurs , 0 if submission was successful .
*/
int bdev_write_page ( struct block_device * bdev , sector_t sector ,
struct page * page , struct writeback_control * wbc )
{
int result ;
const struct block_device_operations * ops = bdev - > bd_disk - > fops ;
2015-11-20 00:29:28 +03:00
2015-05-12 20:48:53 +03:00
if ( ! ops - > rw_page | | bdev_get_integrity ( bdev ) )
2014-06-05 03:07:46 +04:00
return - EOPNOTSUPP ;
2017-11-09 21:49:58 +03:00
result = blk_queue_enter ( bdev - > bd_queue , 0 ) ;
2015-11-20 00:29:28 +03:00
if ( result )
return result ;
2014-06-05 03:07:46 +04:00
set_page_writeback ( page ) ;
2016-08-05 17:11:04 +03:00
result = ops - > rw_page ( bdev , sector + get_start_sect ( bdev ) , page , true ) ;
2017-10-14 01:58:15 +03:00
if ( result ) {
2014-06-05 03:07:46 +04:00
end_page_writeback ( page ) ;
2017-10-14 01:58:15 +03:00
} else {
clean_page_buffers ( page ) ;
2014-06-05 03:07:46 +04:00
unlock_page ( page ) ;
2017-10-14 01:58:15 +03:00
}
2015-11-20 00:29:28 +03:00
blk_queue_exit ( bdev - > bd_queue ) ;
2014-06-05 03:07:46 +04:00
return result ;
}
EXPORT_SYMBOL_GPL ( bdev_write_page ) ;
2005-04-17 02:20:36 +04:00
/*
* pseudo - fs
*/
static __cacheline_aligned_in_smp DEFINE_SPINLOCK ( bdev_lock ) ;
2006-12-07 07:33:20 +03:00
static struct kmem_cache * bdev_cachep __read_mostly ;
2005-04-17 02:20:36 +04:00
static struct inode * bdev_alloc_inode ( struct super_block * sb )
{
2006-12-07 07:33:17 +03:00
struct bdev_inode * ei = kmem_cache_alloc ( bdev_cachep , GFP_KERNEL ) ;
2005-04-17 02:20:36 +04:00
if ( ! ei )
return NULL ;
return & ei - > vfs_inode ;
}
2011-01-07 09:49:49 +03:00
static void bdev_i_callback ( struct rcu_head * head )
2005-04-17 02:20:36 +04:00
{
2011-01-07 09:49:49 +03:00
struct inode * inode = container_of ( head , struct inode , i_rcu ) ;
2005-04-17 02:20:36 +04:00
struct bdev_inode * bdi = BDEV_I ( inode ) ;
kmem_cache_free ( bdev_cachep , bdi ) ;
}
2011-01-07 09:49:49 +03:00
static void bdev_destroy_inode ( struct inode * inode )
{
call_rcu ( & inode - > i_rcu , bdev_i_callback ) ;
}
2008-07-26 06:45:34 +04:00
static void init_once ( void * foo )
2005-04-17 02:20:36 +04:00
{
struct bdev_inode * ei = ( struct bdev_inode * ) foo ;
struct block_device * bdev = & ei - > bdev ;
2007-05-17 09:10:57 +04:00
memset ( bdev , 0 , sizeof ( * bdev ) ) ;
mutex_init ( & bdev - > bd_mutex ) ;
INIT_LIST_HEAD ( & bdev - > bd_list ) ;
2011-01-14 20:43:57 +03:00
# ifdef CONFIG_SYSFS
INIT_LIST_HEAD ( & bdev - > bd_holder_disks ) ;
# endif
2017-03-02 18:50:13 +03:00
bdev - > bd_bdi = & noop_backing_dev_info ;
2007-05-17 09:10:57 +04:00
inode_init_once ( & ei - > vfs_inode ) ;
2009-01-10 03:40:59 +03:00
/* Initialize mutex for freeze. */
mutex_init ( & bdev - > bd_fsfreeze_mutex ) ;
2005-04-17 02:20:36 +04:00
}
2010-06-07 22:34:48 +04:00
static void bdev_evict_inode ( struct inode * inode )
2005-04-17 02:20:36 +04:00
{
struct block_device * bdev = & BDEV_I ( inode ) - > bdev ;
2014-04-04 01:47:49 +04:00
truncate_inode_pages_final ( & inode - > i_data ) ;
2010-06-07 22:34:48 +04:00
invalidate_inode_buffers ( inode ) ; /* is it needed here? */
2012-05-03 16:48:02 +04:00
clear_inode ( inode ) ;
2005-04-17 02:20:36 +04:00
spin_lock ( & bdev_lock ) ;
list_del_init ( & bdev - > bd_list ) ;
spin_unlock ( & bdev_lock ) ;
2017-03-23 03:37:00 +03:00
/* Detach inode from wb early as bdi_put() may free bdi->wb */
inode_detach_wb ( inode ) ;
2017-03-02 18:50:13 +03:00
if ( bdev - > bd_bdi ! = & noop_backing_dev_info ) {
2017-02-02 17:56:52 +03:00
bdi_put ( bdev - > bd_bdi ) ;
2017-03-02 18:50:13 +03:00
bdev - > bd_bdi = & noop_backing_dev_info ;
}
2005-04-17 02:20:36 +04:00
}
2007-02-12 11:55:41 +03:00
static const struct super_operations bdev_sops = {
2005-04-17 02:20:36 +04:00
. statfs = simple_statfs ,
. alloc_inode = bdev_alloc_inode ,
. destroy_inode = bdev_destroy_inode ,
. drop_inode = generic_delete_inode ,
2010-06-07 22:34:48 +04:00
. evict_inode = bdev_evict_inode ,
2005-04-17 02:20:36 +04:00
} ;
2010-07-25 23:47:46 +04:00
static struct dentry * bd_mount ( struct file_system_type * fs_type ,
int flags , const char * dev_name , void * data )
2005-04-17 02:20:36 +04:00
{
2016-02-23 01:27:40 +03:00
struct dentry * dent ;
dent = mount_pseudo ( fs_type , " bdev: " , & bdev_sops , NULL , BDEVFS_MAGIC ) ;
2016-08-22 13:47:43 +03:00
if ( ! IS_ERR ( dent ) )
2016-02-23 01:27:40 +03:00
dent - > d_sb - > s_iflags | = SB_I_CGROUPWB ;
return dent ;
2005-04-17 02:20:36 +04:00
}
static struct file_system_type bd_type = {
. name = " bdev " ,
2010-07-25 23:47:46 +04:00
. mount = bd_mount ,
2005-04-17 02:20:36 +04:00
. kill_sb = kill_anon_super ,
} ;
2015-05-23 00:13:33 +03:00
struct super_block * blockdev_superblock __read_mostly ;
EXPORT_SYMBOL_GPL ( blockdev_superblock ) ;
2005-04-17 02:20:36 +04:00
void __init bdev_cache_init ( void )
{
int err ;
2012-01-10 03:43:59 +04:00
static struct vfsmount * bd_mnt ;
2008-12-02 01:34:56 +03:00
2005-04-17 02:20:36 +04:00
bdev_cachep = kmem_cache_create ( " bdev_cache " , sizeof ( struct bdev_inode ) ,
2006-03-24 14:16:06 +03:00
0 , ( SLAB_HWCACHE_ALIGN | SLAB_RECLAIM_ACCOUNT |
2016-01-15 02:18:21 +03:00
SLAB_MEM_SPREAD | SLAB_ACCOUNT | SLAB_PANIC ) ,
2007-07-20 05:11:58 +04:00
init_once ) ;
2005-04-17 02:20:36 +04:00
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 " ) ;
2012-01-10 03:43:59 +04:00
blockdev_superblock = bd_mnt - > mnt_sb ; /* For writeback */
2005-04-17 02:20:36 +04:00
}
/*
* 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 ) ;
2017-02-02 17:56:49 +03:00
/*
* If there is a bdev inode for this device , unhash it so that it gets evicted
* as soon as last inode reference is dropped .
*/
void bdev_unhash_inode ( dev_t dev )
{
struct inode * inode ;
inode = ilookup5 ( blockdev_superblock , hash ( dev ) , bdev_test , & dev ) ;
if ( inode ) {
remove_inode_hash ( inode ) ;
iput ( inode ) ;
}
}
2005-04-17 02:20:36 +04:00
struct block_device * bdget ( dev_t dev )
{
struct block_device * bdev ;
struct inode * inode ;
2008-12-02 01:34:56 +03:00
inode = iget5_locked ( blockdev_superblock , hash ( dev ) ,
2005-04-17 02:20:36 +04:00
bdev_test , bdev_set , & dev ) ;
if ( ! inode )
return NULL ;
bdev = & BDEV_I ( inode ) - > bdev ;
if ( inode - > i_state & I_NEW ) {
bdev - > bd_contains = NULL ;
2011-06-30 05:01:45 +04:00
bdev - > bd_super = NULL ;
2005-04-17 02:20:36 +04:00
bdev - > bd_inode = inode ;
2017-02-28 01:28:32 +03:00
bdev - > bd_block_size = i_blocksize ( inode ) ;
2005-04-17 02:20:36 +04:00
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 ) ;
spin_lock ( & bdev_lock ) ;
list_add ( & bdev - > bd_list , & all_bdevs ) ;
spin_unlock ( & bdev_lock ) ;
unlock_new_inode ( inode ) ;
}
return bdev ;
}
EXPORT_SYMBOL ( bdget ) ;
2009-07-29 23:07:55 +04:00
/**
* 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 )
{
2010-10-23 19:11:40 +04:00
ihold ( bdev - > bd_inode ) ;
2009-07-29 23:07:55 +04:00
return bdev ;
}
2013-04-01 20:47:56 +04:00
EXPORT_SYMBOL ( bdgrab ) ;
2009-07-29 23:07:55 +04:00
2005-04-17 02:20:36 +04:00
long nr_blockdev_pages ( void )
{
2007-07-16 10:40:23 +04:00
struct block_device * bdev ;
2005-04-17 02:20:36 +04:00
long ret = 0 ;
spin_lock ( & bdev_lock ) ;
2007-07-16 10:40:23 +04:00
list_for_each_entry ( bdev , & all_bdevs , bd_list ) {
2005-04-17 02:20:36 +04:00
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 ;
2006-06-23 01:47:21 +04:00
2005-04-17 02:20:36 +04:00
spin_lock ( & bdev_lock ) ;
bdev = inode - > i_bdev ;
2017-02-21 20:09:48 +03:00
if ( bdev & & ! inode_unhashed ( bdev - > bd_inode ) ) {
2015-11-21 00:18:43 +03:00
bdgrab ( bdev ) ;
2005-04-17 02:20:36 +04:00
spin_unlock ( & bdev_lock ) ;
return bdev ;
}
spin_unlock ( & bdev_lock ) ;
2006-06-23 01:47:21 +04:00
2017-02-21 20:09:48 +03:00
/*
* i_bdev references block device inode that was already shut down
* ( corresponding device got removed ) . Remove the reference and look
* up block device inode again just in case new device got
* reestablished under the same device number .
*/
if ( bdev )
bd_forget ( inode ) ;
2005-04-17 02:20:36 +04:00
bdev = bdget ( inode - > i_rdev ) ;
if ( bdev ) {
spin_lock ( & bdev_lock ) ;
2006-06-23 01:47:21 +04:00
if ( ! inode - > i_bdev ) {
/*
2010-10-23 19:11:40 +04:00
* We take an additional reference to bd_inode ,
2006-06-23 01:47:21 +04:00
* and it ' s released in clear_inode ( ) of inode .
* So , we can access it via - > i_mapping always
* without igrab ( ) .
*/
2015-11-21 00:18:43 +03:00
bdgrab ( bdev ) ;
2006-06-23 01:47:21 +04:00
inode - > i_bdev = bdev ;
inode - > i_mapping = bdev - > bd_inode - > i_mapping ;
}
2005-04-17 02:20:36 +04:00
spin_unlock ( & bdev_lock ) ;
}
return bdev ;
}
/* Call when you free inode */
void bd_forget ( struct inode * inode )
{
2006-06-23 01:47:21 +04:00
struct block_device * bdev = NULL ;
2005-04-17 02:20:36 +04:00
spin_lock ( & bdev_lock ) ;
2013-05-01 02:26:47 +04:00
if ( ! sb_is_blkdev_sb ( inode - > i_sb ) )
bdev = inode - > i_bdev ;
2016-07-19 20:16:52 +03:00
inode - > i_bdev = NULL ;
inode - > i_mapping = & inode - > i_data ;
2005-04-17 02:20:36 +04:00
spin_unlock ( & bdev_lock ) ;
2006-06-23 01:47:21 +04:00
if ( bdev )
2015-11-21 00:18:43 +03:00
bdput ( bdev ) ;
2005-04-17 02:20:36 +04:00
}
2010-04-07 13:52:29 +04:00
/**
* 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
*
2011-03-31 05:57:33 +04:00
* Test whether @ bdev can be claimed by @ holder .
2010-04-07 13:52:29 +04:00
*
* 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 )
2005-04-17 02:20:36 +04:00
{
if ( bdev - > bd_holder = = holder )
2010-04-07 13:52:29 +04:00
return true ; /* already a holder */
2005-04-17 02:20:36 +04:00
else if ( bdev - > bd_holder ! = NULL )
2010-04-07 13:52:29 +04:00
return false ; /* held by someone else */
block_dev: don't test bdev->bd_contains when it is not stable
bdev->bd_contains is not stable before calling __blkdev_get().
When __blkdev_get() is called on a parition with ->bd_openers == 0
it sets
bdev->bd_contains = bdev;
which is not correct for a partition.
After a call to __blkdev_get() succeeds, ->bd_openers will be > 0
and then ->bd_contains is stable.
When FMODE_EXCL is used, blkdev_get() calls
bd_start_claiming() -> bd_prepare_to_claim() -> bd_may_claim()
This call happens before __blkdev_get() is called, so ->bd_contains
is not stable. So bd_may_claim() cannot safely use ->bd_contains.
It currently tries to use it, and this can lead to a BUG_ON().
This happens when a whole device is already open with a bd_holder (in
use by dm in my particular example) and two threads race to open a
partition of that device for the first time, one opening with O_EXCL and
one without.
The thread that doesn't use O_EXCL gets through blkdev_get() to
__blkdev_get(), gains the ->bd_mutex, and sets bdev->bd_contains = bdev;
Immediately thereafter the other thread, using FMODE_EXCL, calls
bd_start_claiming() from blkdev_get(). This should fail because the
whole device has a holder, but because bdev->bd_contains == bdev
bd_may_claim() incorrectly reports success.
This thread continues and blocks on bd_mutex.
The first thread then sets bdev->bd_contains correctly and drops the mutex.
The thread using FMODE_EXCL then continues and when it calls bd_may_claim()
again in:
BUG_ON(!bd_may_claim(bdev, whole, holder));
The BUG_ON fires.
Fix this by removing the dependency on ->bd_contains in
bd_may_claim(). As bd_may_claim() has direct access to the whole
device, it can simply test if the target bdev is the whole device.
Fixes: 6b4517a7913a ("block: implement bd_claiming and claiming block")
Cc: stable@vger.kernel.org (v2.6.35+)
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2016-12-12 18:21:51 +03:00
else if ( whole = = bdev )
2010-04-07 13:52:29 +04:00
return true ; /* is a whole device which isn't held */
2005-04-17 02:20:36 +04:00
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 13:55:17 +03:00
else if ( whole - > bd_holder = = bd_may_claim )
2010-04-07 13:52:29 +04:00
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 */
2005-04-17 02:20:36 +04:00
else
2010-04-07 13:52:29 +04:00
return true ; /* is a partition of an un-held device */
}
2010-04-07 13:53:59 +04:00
/**
* 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 ;
2010-08-04 19:59:39 +04:00
/* if claiming is already in progress, wait for it to finish */
if ( whole - > bd_claiming ) {
2010-04-07 13:53:59 +04:00
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
2010-05-25 19:51:19 +04:00
* 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 .
2010-04-07 13:53:59 +04:00
*
* 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 ) ;
2011-06-13 14:45:48 +04:00
/*
* 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 ) ;
2010-05-25 19:50:21 +04:00
module_put ( disk - > fops - > owner ) ;
2010-04-07 13:53:59 +04:00
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 ) ;
}
}
2006-03-27 13:17:57 +04:00
# ifdef CONFIG_SYSFS
2011-01-14 20:43:57 +03:00
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 12:58:56 +04:00
static int add_symlink ( struct kobject * from , struct kobject * to )
2006-03-27 13:17:57 +04:00
{
[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 12:58:56 +04:00
return sysfs_create_link ( from , to , kobject_name ( to ) ) ;
2006-03-27 13:17:57 +04:00
}
static void del_symlink ( struct kobject * from , struct kobject * to )
{
sysfs_remove_link ( from , kobject_name ( to ) ) ;
}
2006-10-31 00:23:56 +03:00
/**
2010-11-13 13:55:17 +03:00
* bd_link_disk_holder - create symlinks between holding disk and slave bdev
* @ bdev : the claimed slave bdev
* @ disk : the holding disk
2006-10-31 00:23:56 +03:00
*
2011-01-14 20:43:57 +03:00
* DON ' T USE THIS UNLESS YOU ' RE ALREADY USING IT .
*
2010-11-13 13:55:17 +03:00
* This functions creates the following sysfs symlinks .
2006-03-27 13:17:57 +04:00
*
2010-11-13 13:55:17 +03:00
* - from " slaves " directory of the holder @ disk to the claimed @ bdev
* - from " holders " directory of the @ bdev to the holder @ disk
2006-03-27 13:17:57 +04:00
*
2010-11-13 13:55:17 +03:00
* For example , if / dev / dm - 0 maps to / dev / sda and disk for dm - 0 is
* passed to bd_link_disk_holder ( ) , then :
2006-03-27 13:17:57 +04:00
*
2010-11-13 13:55:17 +03:00
* / sys / block / dm - 0 / slaves / sda - - > / sys / block / sda
* / sys / block / sda / holders / dm - 0 - - > / sys / block / dm - 0
2006-03-27 13:17:57 +04:00
*
2010-11-13 13:55:17 +03:00
* 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 .
2006-03-27 13:17:57 +04:00
*
2010-11-13 13:55:17 +03:00
* CONTEXT :
* Might sleep .
2006-03-27 13:17:57 +04:00
*
2010-11-13 13:55:17 +03:00
* RETURNS :
* 0 on success , - errno on failure .
2006-03-27 13:17:57 +04:00
*/
2010-11-13 13:55:17 +03:00
int bd_link_disk_holder ( struct block_device * bdev , struct gendisk * disk )
2006-03-27 13:17:57 +04:00
{
2011-01-14 20:43:57 +03:00
struct bd_holder_disk * holder ;
2010-11-13 13:55:17 +03:00
int ret = 0 ;
2006-03-27 13:17:57 +04:00
2006-12-08 13:36:13 +03:00
mutex_lock ( & bdev - > bd_mutex ) ;
2006-10-31 00:23:56 +03:00
2011-01-14 20:43:57 +03:00
WARN_ON_ONCE ( ! bdev - > bd_holder ) ;
2007-07-16 10:41:25 +04:00
2010-11-13 13:55:17 +03:00
/* FIXME: remove the following once add_disk() handles errors */
if ( WARN_ON ( ! disk - > slave_dir | | ! bdev - > bd_part - > holder_dir ) )
goto out_unlock ;
2007-07-16 10:41:25 +04:00
2011-01-14 20:43:57 +03:00
holder = bd_find_holder_disk ( bdev , disk ) ;
if ( holder ) {
holder - > refcnt + + ;
2010-11-13 13:55:17 +03:00
goto out_unlock ;
2011-01-14 20:43:57 +03:00
}
2006-03-27 13:17:57 +04:00
2011-01-14 20:43:57 +03:00
holder = kzalloc ( sizeof ( * holder ) , GFP_KERNEL ) ;
if ( ! holder ) {
ret = - ENOMEM ;
2010-11-13 13:55:17 +03:00
goto out_unlock ;
}
2006-03-27 13:17:57 +04:00
2011-01-14 20:43:57 +03:00
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 ;
2011-02-24 11:56:32 +03:00
/*
* bdev could be deleted beneath us which would implicitly destroy
* the holder directory . Hold on to it .
*/
kobject_get ( bdev - > bd_part - > holder_dir ) ;
2011-01-14 20:43:57 +03:00
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 ) ;
2010-11-13 13:55:17 +03:00
out_unlock :
2006-03-27 13:18:00 +04:00
mutex_unlock ( & bdev - > bd_mutex ) ;
2010-11-13 13:55:17 +03:00
return ret ;
2006-03-27 13:17:57 +04:00
}
2010-11-13 13:55:17 +03:00
EXPORT_SYMBOL_GPL ( bd_link_disk_holder ) ;
2006-03-27 13:17:57 +04:00
2011-01-14 20:43:57 +03:00
/**
* 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 )
2006-03-27 13:17:57 +04:00
{
2011-01-14 20:43:57 +03:00
struct bd_holder_disk * holder ;
2006-03-27 13:17:57 +04:00
2011-01-14 20:43:57 +03:00
mutex_lock ( & bdev - > bd_mutex ) ;
2006-03-27 13:17:57 +04:00
2011-01-14 20:43:57 +03:00
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 ) ;
2011-02-24 11:56:32 +03:00
kobject_put ( bdev - > bd_part - > holder_dir ) ;
2011-01-14 20:43:57 +03:00
list_del_init ( & holder - > list ) ;
kfree ( holder ) ;
}
mutex_unlock ( & bdev - > bd_mutex ) ;
2005-04-17 02:20:36 +04:00
}
2011-01-14 20:43:57 +03:00
EXPORT_SYMBOL_GPL ( bd_unlink_disk_holder ) ;
2006-03-27 13:17:57 +04:00
# endif
2005-04-17 02:20:36 +04:00
2008-09-05 00:27:40 +04:00
/**
* flush_disk - invalidates all buffer - cache entries on a disk
*
* @ bdev : struct block device to be flushed
2011-02-26 21:54:00 +03:00
* @ kill_dirty : flag to guide handling of dirty inodes
2008-09-05 00:27:40 +04:00
*
* 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 .
*/
2011-02-24 09:25:47 +03:00
static void flush_disk ( struct block_device * bdev , bool kill_dirty )
2008-09-05 00:27:40 +04:00
{
2011-02-24 09:25:47 +03:00
if ( __invalidate_device ( bdev , kill_dirty ) ) {
2008-09-05 00:27:40 +04:00
printk ( KERN_WARNING " VFS: busy inodes on changed media or "
2015-04-13 15:31:34 +03:00
" resized disk %s \n " ,
bdev - > bd_disk ? bdev - > bd_disk - > disk_name : " " ) ;
2008-09-05 00:27:40 +04:00
}
if ( ! bdev - > bd_disk )
return ;
2011-08-23 22:01:04 +04:00
if ( disk_part_scan_enabled ( bdev - > bd_disk ) )
2008-09-05 00:27:40 +04:00
bdev - > bd_invalidated = 1 ;
}
2008-09-05 00:27:25 +04:00
/**
2008-10-09 12:42:38 +04:00
* check_disk_size_change - checks for disk size change and adjusts bdev size .
2008-09-05 00:27:25 +04:00
* @ 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 ) {
printk ( KERN_INFO
" %s: detected capacity change from %lld to %lld \n " ,
2015-04-13 15:31:34 +03:00
disk - > disk_name , bdev_size , disk_size ) ;
2008-09-05 00:27:25 +04:00
i_size_write ( bdev - > bd_inode , disk_size ) ;
2011-02-24 09:25:47 +03:00
flush_disk ( bdev , false ) ;
2008-09-05 00:27:25 +04:00
}
}
EXPORT_SYMBOL ( check_disk_size_change ) ;
2008-09-05 00:27:20 +04:00
/**
2008-10-09 12:42:38 +04:00
* revalidate_disk - wrapper for lower - level driver ' s revalidate_disk call - back
2008-09-05 00:27:20 +04:00
* @ 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 )
{
2008-09-05 00:27:25 +04:00
struct block_device * bdev ;
2008-09-05 00:27:20 +04:00
int ret = 0 ;
if ( disk - > fops - > revalidate_disk )
ret = disk - > fops - > revalidate_disk ( disk ) ;
2008-09-05 00:27:25 +04:00
bdev = bdget_disk ( disk , 0 ) ;
if ( ! bdev )
return ret ;
mutex_lock ( & bdev - > bd_mutex ) ;
check_disk_size_change ( disk , bdev ) ;
2013-02-22 04:42:01 +04:00
bdev - > bd_invalidated = 0 ;
2008-09-05 00:27:25 +04:00
mutex_unlock ( & bdev - > bd_mutex ) ;
bdput ( bdev ) ;
2008-09-05 00:27:20 +04:00
return ret ;
}
EXPORT_SYMBOL ( revalidate_disk ) ;
2005-04-17 02:20:36 +04:00
/*
* 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 ;
2009-09-22 04:01:13 +04:00
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 22:57:37 +03:00
unsigned int events ;
2005-04-17 02:20:36 +04:00
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 22:57:37 +03:00
events = disk_clear_events ( disk , DISK_EVENT_MEDIA_CHANGE |
DISK_EVENT_EJECT_REQUEST ) ;
if ( ! ( events & DISK_EVENT_MEDIA_CHANGE ) )
2005-04-17 02:20:36 +04:00
return 0 ;
2011-02-24 09:25:47 +03:00
flush_disk ( bdev , true ) ;
2005-04-17 02:20:36 +04:00
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 )
{
2009-05-23 01:17:49 +04:00
unsigned bsize = bdev_logical_block_size ( bdev ) ;
2005-04-17 02:20:36 +04:00
2016-01-22 23:40:57 +03:00
inode_lock ( bdev - > bd_inode ) ;
2013-02-22 03:16:42 +04:00
i_size_write ( bdev - > bd_inode , size ) ;
2016-01-22 23:40:57 +03:00
inode_unlock ( bdev - > bd_inode ) ;
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 15:29:47 +03:00
while ( bsize < PAGE_SIZE ) {
2005-04-17 02:20:36 +04:00
if ( size & bsize )
break ;
bsize < < = 1 ;
}
bdev - > bd_block_size = bsize ;
bdev - > bd_inode - > i_blkbits = blksize_bits ( bsize ) ;
}
EXPORT_SYMBOL ( bd_set_size ) ;
2013-05-06 06:11:03 +04:00
static void __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 13:36:16 +03:00
2007-02-21 00:58:18 +03:00
/*
* bd_mutex locking :
*
* mutex_lock ( part - > bd_mutex )
* mutex_lock_nested ( whole - > bd_mutex , 1 )
*/
2007-10-08 21:24:05 +04:00
static int __blkdev_get ( struct block_device * bdev , fmode_t mode , int for_part )
2005-04-17 02:20:36 +04:00
{
struct gendisk * disk ;
2011-10-19 16:31:07 +04:00
struct module * owner ;
2008-06-06 09:46:27 +04:00
int ret ;
2008-09-03 11:01:09 +04:00
int partno ;
2008-06-23 16:30:55 +04:00
int perm = 0 ;
2007-10-08 21:24:05 +04:00
if ( mode & FMODE_READ )
2008-06-23 16:30:55 +04:00
perm | = MAY_READ ;
2007-10-08 21:24:05 +04:00
if ( mode & FMODE_WRITE )
2008-06-23 16:30:55 +04:00
perm | = MAY_WRITE ;
/*
* hooks : / n / , see " layering violations " .
*/
2010-08-11 05:02:55 +04:00
if ( ! for_part ) {
ret = devcgroup_inode_permission ( bdev - > bd_inode , perm ) ;
if ( ret ! = 0 ) {
bdput ( bdev ) ;
return ret ;
}
2008-08-01 13:32:04 +04:00
}
2008-06-06 09:46:27 +04:00
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>
2009-01-09 00:31:10 +03:00
restart :
2008-08-25 14:56:12 +04:00
2008-11-05 12:21:06 +03:00
ret = - ENXIO ;
2008-09-03 11:01:09 +04:00
disk = get_gendisk ( bdev - > bd_dev , & partno ) ;
2008-08-25 14:56:12 +04:00
if ( ! disk )
2010-08-07 20:25:34 +04:00
goto out ;
2011-10-19 16:31:07 +04:00
owner = disk - > fops - > owner ;
2005-04-17 02:20:36 +04:00
2011-03-09 21:54:27 +03:00
disk_block_events ( disk ) ;
2006-12-08 13:36:16 +03:00
mutex_lock_nested ( & bdev - > bd_mutex , for_part ) ;
2005-04-17 02:20:36 +04:00
if ( ! bdev - > bd_openers ) {
bdev - > bd_disk = disk ;
2012-01-13 05:20:34 +04:00
bdev - > bd_queue = disk - > queue ;
2005-04-17 02:20:36 +04:00
bdev - > bd_contains = bdev ;
2017-08-23 20:10:31 +03:00
bdev - > bd_partno = partno ;
2016-02-27 02:19:43 +03:00
2008-09-03 11:01:09 +04:00
if ( ! partno ) {
2008-11-05 12:21:06 +03:00
ret = - ENXIO ;
bdev - > bd_part = disk_get_part ( disk , partno ) ;
if ( ! bdev - > bd_part )
goto out_clear ;
2011-04-21 22:54:45 +04:00
ret = 0 ;
2005-04-17 02:20:36 +04:00
if ( disk - > fops - > open ) {
2007-10-08 21:24:05 +04:00
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>
2009-01-09 00:31:10 +03:00
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 ;
2012-01-13 05:20:34 +04:00
bdev - > bd_queue = NULL ;
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>
2009-01-09 00:31:10 +03:00
mutex_unlock ( & bdev - > bd_mutex ) ;
2011-03-09 21:54:27 +03:00
disk_unblock_events ( disk ) ;
put_disk ( disk ) ;
2011-10-19 16:31:07 +04:00
module_put ( owner ) ;
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>
2009-01-09 00:31:10 +03:00
goto restart ;
}
2005-04-17 02:20:36 +04:00
}
2011-05-23 15:26:07 +04:00
2016-09-14 12:56:13 +03:00
if ( ! ret )
2011-05-23 15:26:07 +04:00
bd_set_size ( bdev , ( loff_t ) get_capacity ( disk ) < < 9 ) ;
2011-04-21 22:54:45 +04:00
/*
* 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 .
*/
2012-03-02 13:38:33 +04:00
if ( bdev - > bd_invalidated ) {
if ( ! ret )
rescan_partitions ( disk , bdev ) ;
else if ( ret = = - ENOMEDIUM )
invalidate_partitions ( disk , bdev ) ;
}
2015-11-30 21:20:29 +03:00
2011-04-21 22:54:45 +04:00
if ( ret )
goto out_clear ;
2005-04-17 02:20:36 +04:00
} else {
struct block_device * whole ;
whole = bdget_disk ( disk , 0 ) ;
ret = - ENOMEM ;
if ( ! whole )
2008-08-25 14:56:12 +04:00
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 13:36:16 +03:00
BUG_ON ( for_part ) ;
2007-10-08 21:24:05 +04:00
ret = __blkdev_get ( whole , mode , 1 ) ;
2005-04-17 02:20:36 +04:00
if ( ret )
2008-08-25 14:56:12 +04:00
goto out_clear ;
2005-04-17 02:20:36 +04:00
bdev - > bd_contains = whole ;
2008-11-05 12:21:06 +03:00
bdev - > bd_part = disk_get_part ( disk , partno ) ;
2008-09-03 11:03:02 +04:00
if ( ! ( disk - > flags & GENHD_FL_UP ) | |
2008-11-05 12:21:06 +03:00
! bdev - > bd_part | | ! bdev - > bd_part - > nr_sects ) {
2005-04-17 02:20:36 +04:00
ret = - ENXIO ;
2008-08-25 14:56:12 +04:00
goto out_clear ;
2005-04-17 02:20:36 +04:00
}
2008-11-05 12:21:06 +03:00
bd_set_size ( bdev , ( loff_t ) bdev - > bd_part - > nr_sects < < 9 ) ;
2005-04-17 02:20:36 +04:00
}
2017-03-23 03:36:53 +03:00
if ( bdev - > bd_bdi = = & noop_backing_dev_info )
bdev - > bd_bdi = bdi_get ( disk - > queue - > backing_dev_info ) ;
2005-04-17 02:20:36 +04:00
} else {
if ( bdev - > bd_contains = = bdev ) {
2011-04-21 22:54:45 +04:00
ret = 0 ;
if ( bdev - > bd_disk - > fops - > open )
2007-10-08 21:24:05 +04:00
ret = bdev - > bd_disk - > fops - > open ( bdev , mode ) ;
2011-04-21 22:54:45 +04:00
/* the same as first opener case, read comment there */
2012-03-02 13:38:33 +04:00
if ( bdev - > bd_invalidated ) {
if ( ! ret )
rescan_partitions ( bdev - > bd_disk , bdev ) ;
else if ( ret = = - ENOMEDIUM )
invalidate_partitions ( bdev - > bd_disk , bdev ) ;
}
2011-04-21 22:54:45 +04:00
if ( ret )
goto out_unlock_bdev ;
2005-04-17 02:20:36 +04:00
}
2011-03-09 21:54:27 +03:00
/* only one opener holds refs to the module and disk */
put_disk ( disk ) ;
2011-10-19 16:31:07 +04:00
module_put ( owner ) ;
2005-04-17 02:20:36 +04:00
}
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 13:36:16 +03:00
if ( for_part )
bdev - > bd_part_count + + ;
2006-03-23 14:00:28 +03:00
mutex_unlock ( & bdev - > bd_mutex ) ;
2011-03-09 21:54:27 +03:00
disk_unblock_events ( disk ) ;
2005-04-17 02:20:36 +04:00
return 0 ;
2008-08-25 14:56:12 +04:00
out_clear :
2008-11-05 12:21:06 +03:00
disk_put_part ( bdev - > bd_part ) ;
2005-04-17 02:20:36 +04:00
bdev - > bd_disk = NULL ;
2008-08-25 14:56:12 +04:00
bdev - > bd_part = NULL ;
2012-01-13 05:20:34 +04:00
bdev - > bd_queue = NULL ;
2005-04-17 02:20:36 +04:00
if ( bdev ! = bdev - > bd_contains )
2007-10-08 21:24:05 +04:00
__blkdev_put ( bdev - > bd_contains , mode , 1 ) ;
2005-04-17 02:20:36 +04:00
bdev - > bd_contains = NULL ;
2008-08-25 14:56:12 +04:00
out_unlock_bdev :
2006-03-23 14:00:28 +03:00
mutex_unlock ( & bdev - > bd_mutex ) ;
2011-03-09 21:54:27 +03:00
disk_unblock_events ( disk ) ;
2008-08-25 14:56:12 +04:00
put_disk ( disk ) ;
2011-10-19 16:31:07 +04:00
module_put ( owner ) ;
2011-03-19 15:53:31 +03:00
out :
2008-08-25 14:56:12 +04:00
bdput ( bdev ) ;
2005-04-17 02:20:36 +04:00
return ret ;
}
2010-11-13 13:55:18 +03:00
/**
* 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 13:55:17 +03:00
int blkdev_get ( struct block_device * bdev , fmode_t mode , void * holder )
2005-04-17 02:20:36 +04:00
{
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 13:55:17 +03:00
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 ) {
2011-04-21 22:54:46 +04:00
struct gendisk * disk = whole - > bd_disk ;
2010-11-13 13:55:17 +03:00
/* 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 22:57:37 +03:00
mutex_lock ( & bdev - > bd_mutex ) ;
2010-11-13 13:55:17 +03:00
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 22:57:37 +03:00
if ( ! res ) {
2010-11-13 13:55:17 +03:00
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 22:57:37 +03:00
/*
2011-04-21 22:54:46 +04:00
* 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 22:57:37 +03:00
*/
2011-06-01 10:27:41 +04:00
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 22:57:37 +03:00
bdev - > bd_write_holder = true ;
2011-04-21 22:54:46 +04:00
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 22:57:37 +03:00
}
mutex_unlock ( & bdev - > bd_mutex ) ;
2010-11-13 13:55:17 +03:00
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 13:55:17 +03:00
}
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 13:36:16 +03:00
}
2005-04-17 02:20:36 +04:00
EXPORT_SYMBOL ( blkdev_get ) ;
2010-11-13 13:55:18 +03:00
/**
* 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 ) ;
2011-02-17 02:11:53 +03:00
if ( ( mode & FMODE_WRITE ) & & bdev_read_only ( bdev ) ) {
blkdev_put ( bdev , mode ) ;
return ERR_PTR ( - EACCES ) ;
}
2010-11-13 13:55:18 +03:00
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 ) ;
2005-04-17 02:20:36 +04:00
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 ;
2017-08-29 17:13:21 +03:00
filp - > f_mode | = FMODE_NOWAIT ;
2007-10-08 21:24:05 +04:00
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 ;
2005-04-17 02:20:36 +04:00
bdev = bd_acquire ( inode ) ;
2006-10-28 21:38:33 +04:00
if ( bdev = = NULL )
return - ENOMEM ;
2005-04-17 02:20:36 +04:00
2007-10-08 21:24:05 +04:00
filp - > f_mapping = bdev - > bd_inode - > i_mapping ;
fs: new infrastructure for writeback error handling and reporting
Most filesystems currently use mapping_set_error and
filemap_check_errors for setting and reporting/clearing writeback errors
at the mapping level. filemap_check_errors is indirectly called from
most of the filemap_fdatawait_* functions and from
filemap_write_and_wait*. These functions are called from all sorts of
contexts to wait on writeback to finish -- e.g. mostly in fsync, but
also in truncate calls, getattr, etc.
The non-fsync callers are problematic. We should be reporting writeback
errors during fsync, but many places spread over the tree clear out
errors before they can be properly reported, or report errors at
nonsensical times.
If I get -EIO on a stat() call, there is no reason for me to assume that
it is because some previous writeback failed. The fact that it also
clears out the error such that a subsequent fsync returns 0 is a bug,
and a nasty one since that's potentially silent data corruption.
This patch adds a small bit of new infrastructure for setting and
reporting errors during address_space writeback. While the above was my
original impetus for adding this, I think it's also the case that
current fsync semantics are just problematic for userland. Most
applications that call fsync do so to ensure that the data they wrote
has hit the backing store.
In the case where there are multiple writers to the file at the same
time, this is really hard to determine. The first one to call fsync will
see any stored error, and the rest get back 0. The processes with open
fds may not be associated with one another in any way. They could even
be in different containers, so ensuring coordination between all fsync
callers is not really an option.
One way to remedy this would be to track what file descriptor was used
to dirty the file, but that's rather cumbersome and would likely be
slow. However, there is a simpler way to improve the semantics here
without incurring too much overhead.
This set adds an errseq_t to struct address_space, and a corresponding
one is added to struct file. Writeback errors are recorded in the
mapping's errseq_t, and the one in struct file is used as the "since"
value.
This changes the semantics of the Linux fsync implementation such that
applications can now use it to determine whether there were any
writeback errors since fsync(fd) was last called (or since the file was
opened in the case of fsync having never been called).
Note that those writeback errors may have occurred when writing data
that was dirtied via an entirely different fd, but that's the case now
with the current mapping_set_error/filemap_check_error infrastructure.
This will at least prevent you from getting a false report of success.
The new behavior is still consistent with the POSIX spec, and is more
reliable for application developers. This patch just adds some basic
infrastructure for doing this, and ensures that the f_wb_err "cursor"
is properly set when a file is opened. Later patches will change the
existing code to use this new infrastructure for reporting errors at
fsync time.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
2017-07-06 14:02:25 +03:00
filp - > f_wb_err = filemap_sample_wb_err ( filp - > f_mapping ) ;
2007-10-08 21:24:05 +04:00
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 13:55:17 +03:00
return blkdev_get ( bdev , filp - > f_mode , filp ) ;
2005-04-17 02:20:36 +04:00
}
2013-05-06 06:11:03 +04:00
static void __blkdev_put ( struct block_device * bdev , fmode_t mode , int for_part )
2006-12-08 13:36:13 +03:00
{
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 13:36:16 +03:00
struct block_device * victim = NULL ;
2006-12-08 13:36:13 +03:00
2006-12-08 13:36:16 +03:00
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 13:36:16 +03:00
if ( for_part )
bdev - > bd_part_count - - ;
2006-12-08 13:36:13 +03:00
if ( ! - - bdev - > bd_openers ) {
2010-11-13 13:55:17 +03:00
WARN_ON_ONCE ( bdev - > bd_holders ) ;
2006-12-08 13:36:13 +03:00
sync_blockdev ( bdev ) ;
kill_bdev ( bdev ) ;
block: detach bdev inode from its wb in __blkdev_put()
Since 52ebea749aae ("writeback: make backing_dev_info host
cgroup-specific bdi_writebacks") inode, at some point in its lifetime,
gets attached to a wb (struct bdi_writeback). Detaching happens on
evict, in inode_detach_wb() called from __destroy_inode(), and involves
updating wb.
However, detaching an internal bdev inode from its wb in
__destroy_inode() is too late. Its bdi and by extension root wb are
embedded into struct request_queue, which has different lifetime rules
and can be freed long before the final bdput() is called (can be from
__fput() of a corresponding /dev inode, through dput() - evict() -
bd_forget(). bdevs hold onto the underlying disk/queue pair only while
opened; as soon as bdev is closed all bets are off. In fact,
disk/queue can be gone before __blkdev_put() even returns:
1499 static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part)
1500 {
...
1518 if (bdev->bd_contains == bdev) {
1519 if (disk->fops->release)
1520 disk->fops->release(disk, mode);
[ Driver puts its references to disk/queue ]
1521 }
1522 if (!bdev->bd_openers) {
1523 struct module *owner = disk->fops->owner;
1524
1525 disk_put_part(bdev->bd_part);
1526 bdev->bd_part = NULL;
1527 bdev->bd_disk = NULL;
1528 if (bdev != bdev->bd_contains)
1529 victim = bdev->bd_contains;
1530 bdev->bd_contains = NULL;
1531
1532 put_disk(disk);
[ We put ours, the queue is gone
The last bdput() would result in a write to invalid memory ]
1533 module_put(owner);
...
1539 }
Since bdev inodes are special anyway, detach them in __blkdev_put()
after clearing inode's dirty bits, turning the problematic
inode_detach_wb() in __destroy_inode() into a noop.
add_disk() grabs its disk->queue since 523e1d399ce0 ("block: make
gendisk hold a reference to its queue"), so the old ->release comment
is removed in favor of the new inode_detach_wb() comment.
Cc: stable@vger.kernel.org # 4.2+, needs backporting
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Tested-by: Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-11-21 00:22:34 +03:00
bdev_write_inode ( bdev ) ;
2006-12-08 13:36:13 +03:00
}
if ( bdev - > bd_contains = = bdev ) {
if ( disk - > fops - > release )
2013-05-06 05:52:57 +04:00
disk - > fops - > release ( disk , mode ) ;
2006-12-08 13:36:13 +03:00
}
if ( ! bdev - > bd_openers ) {
struct module * owner = disk - > fops - > owner ;
2008-08-25 14:56:12 +04:00
disk_put_part ( bdev - > bd_part ) ;
bdev - > bd_part = NULL ;
2006-12-08 13:36:13 +03:00
bdev - > bd_disk = NULL ;
[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 13:36:16 +03:00
if ( bdev ! = bdev - > bd_contains )
victim = bdev - > bd_contains ;
2006-12-08 13:36:13 +03:00
bdev - > bd_contains = NULL ;
2011-10-19 16:31:07 +04:00
put_disk ( disk ) ;
module_put ( owner ) ;
2006-12-08 13:36:13 +03:00
}
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 13:36:16 +03:00
if ( victim )
2008-02-23 04:40:24 +03:00
__blkdev_put ( victim , mode , 1 ) ;
2006-12-08 13:36:13 +03:00
}
2013-05-06 06:11:03 +04:00
void 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 13:36:16 +03:00
{
2011-07-01 18:17:47 +04:00
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 13:55:17 +03:00
if ( mode & FMODE_EXCL ) {
2010-11-13 13:55:17 +03:00
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 22:57:37 +03:00
/*
* If this was the last claim , remove holder link and
* unblock evpoll if it was a write holder .
*/
2011-07-01 18:17:47 +04:00
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 22:57:37 +03:00
}
2011-03-09 21:54:27 +03:00
}
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 22:57:37 +03:00
2011-07-01 18:17:47 +04:00
/*
* 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 ) ;
2013-05-06 06:11:03 +04:00
__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 13:36:16 +03:00
}
2006-12-08 13:36:13 +03:00
EXPORT_SYMBOL ( blkdev_put ) ;
2005-04-17 02:20:36 +04:00
static int blkdev_close ( struct inode * inode , struct file * filp )
{
2015-10-28 01:48:19 +03:00
struct block_device * bdev = I_BDEV ( bdev_file_inode ( filp ) ) ;
2013-05-06 06:11:03 +04:00
blkdev_put ( bdev , filp - > f_mode ) ;
return 0 ;
2005-04-17 02:20:36 +04:00
}
2005-06-23 11:10:15 +04:00
static long block_ioctl ( struct file * file , unsigned cmd , unsigned long arg )
2005-04-17 02:20:36 +04:00
{
2015-10-28 01:48:19 +03:00
struct block_device * bdev = I_BDEV ( bdev_file_inode ( file ) ) ;
2008-09-19 11:17:36 +04:00
fmode_t mode = file - > f_mode ;
2008-11-05 16:58:42 +03:00
/*
* O_NDELAY can be altered using fcntl ( . . , F_SETFL , . . ) , so we have
* to updated it before every ioctl .
*/
2008-09-19 11:17:36 +04:00
if ( file - > f_flags & O_NDELAY )
2008-11-05 16:58:42 +03:00
mode | = FMODE_NDELAY ;
else
mode & = ~ FMODE_NDELAY ;
2008-09-19 11:17:36 +04:00
return blkdev_ioctl ( bdev , mode , cmd , arg ) ;
2005-04-17 02:20:36 +04:00
}
2009-08-20 19:43:41 +04:00
/*
* 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 .
*/
2014-04-03 11:21:50 +04:00
ssize_t blkdev_write_iter ( struct kiocb * iocb , struct iov_iter * from )
2009-08-20 19:43:41 +04:00
{
struct file * file = iocb - > ki_filp ;
2015-10-28 01:48:19 +03:00
struct inode * bd_inode = bdev_file_inode ( file ) ;
2015-04-07 18:35:14 +03:00
loff_t size = i_size_read ( bd_inode ) ;
2012-08-02 11:50:39 +04:00
struct blk_plug plug ;
2009-08-20 19:43:41 +04:00
ssize_t ret ;
2015-04-07 18:28:12 +03:00
2015-04-07 18:35:14 +03:00
if ( bdev_read_only ( I_BDEV ( bd_inode ) ) )
return - EPERM ;
2015-04-07 18:28:12 +03:00
2015-04-07 18:35:14 +03:00
if ( ! iov_iter_count ( from ) )
2015-04-07 18:28:12 +03:00
return 0 ;
2015-04-07 18:35:14 +03:00
if ( iocb - > ki_pos > = size )
return - ENOSPC ;
2017-08-29 17:13:21 +03:00
if ( ( iocb - > ki_flags & ( IOCB_NOWAIT | IOCB_DIRECT ) ) = = IOCB_NOWAIT )
return - EOPNOTSUPP ;
2015-04-07 18:35:14 +03:00
iov_iter_truncate ( from , size - iocb - > ki_pos ) ;
2009-08-20 19:43:41 +04:00
2012-08-02 11:50:39 +04:00
blk_start_plug ( & plug ) ;
2014-04-03 11:21:50 +04:00
ret = __generic_file_write_iter ( iocb , from ) ;
2016-04-07 18:52:01 +03:00
if ( ret > 0 )
ret = generic_write_sync ( iocb , ret ) ;
2012-08-02 11:50:39 +04:00
blk_finish_plug ( & plug ) ;
2009-08-20 19:43:41 +04:00
return ret ;
}
2014-04-03 11:21:50 +04:00
EXPORT_SYMBOL_GPL ( blkdev_write_iter ) ;
2009-08-20 19:43:41 +04:00
2014-09-29 18:21:10 +04:00
ssize_t blkdev_read_iter ( struct kiocb * iocb , struct iov_iter * to )
2012-12-08 04:48:39 +04:00
{
struct file * file = iocb - > ki_filp ;
2015-10-28 01:48:19 +03:00
struct inode * bd_inode = bdev_file_inode ( file ) ;
2012-12-08 04:48:39 +04:00
loff_t size = i_size_read ( bd_inode ) ;
2014-04-03 04:02:21 +04:00
loff_t pos = iocb - > ki_pos ;
2012-12-08 04:48:39 +04:00
if ( pos > = size )
return 0 ;
size - = pos ;
2014-04-03 04:02:21 +04:00
iov_iter_truncate ( to , size ) ;
return generic_file_read_iter ( iocb , to ) ;
2012-12-08 04:48:39 +04:00
}
2014-09-29 18:21:10 +04:00
EXPORT_SYMBOL_GPL ( blkdev_read_iter ) ;
2012-12-08 04:48:39 +04:00
2009-01-03 17:47:09 +03:00
/*
* 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 ) ;
}
2016-02-27 02:19:55 +03:00
static int blkdev_writepages ( struct address_space * mapping ,
struct writeback_control * wbc )
{
if ( dax_mapping ( mapping ) ) {
struct block_device * bdev = I_BDEV ( mapping - > host ) ;
return dax_writeback_mapping_range ( mapping , bdev , wbc ) ;
}
return generic_writepages ( mapping , wbc ) ;
}
2008-02-18 15:48:31 +03:00
static const struct address_space_operations def_blk_aops = {
2005-04-17 02:20:36 +04:00
. readpage = blkdev_readpage ,
2014-10-10 02:26:58 +04:00
. readpages = blkdev_readpages ,
2005-04-17 02:20:36 +04:00
. writepage = blkdev_writepage ,
2007-10-16 12:25:04 +04:00
. write_begin = blkdev_write_begin ,
. write_end = blkdev_write_end ,
2016-02-27 02:19:55 +03:00
. writepages = blkdev_writepages ,
2009-01-03 17:47:09 +03:00
. releasepage = blkdev_releasepage ,
2005-04-17 02:20:36 +04:00
. direct_IO = blkdev_direct_IO ,
2013-07-04 02:02:05 +04:00
. is_dirty_writeback = buffer_check_dirty_writeback ,
2005-04-17 02:20:36 +04:00
} ;
2016-10-11 23:51:11 +03:00
# define BLKDEV_FALLOC_FL_SUPPORTED \
( FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | \
FALLOC_FL_ZERO_RANGE | FALLOC_FL_NO_HIDE_STALE )
static long blkdev_fallocate ( struct file * file , int mode , loff_t start ,
loff_t len )
{
struct block_device * bdev = I_BDEV ( bdev_file_inode ( file ) ) ;
struct address_space * mapping ;
loff_t end = start + len - 1 ;
loff_t isize ;
int error ;
/* Fail if we don't recognize the flags. */
if ( mode & ~ BLKDEV_FALLOC_FL_SUPPORTED )
return - EOPNOTSUPP ;
/* Don't go off the end of the device. */
isize = i_size_read ( bdev - > bd_inode ) ;
if ( start > = isize )
return - EINVAL ;
if ( end > = isize ) {
if ( mode & FALLOC_FL_KEEP_SIZE ) {
len = isize - start ;
end = start + len - 1 ;
} else
return - EINVAL ;
}
/*
* Don ' t allow IO that isn ' t aligned to logical block size .
*/
if ( ( start | len ) & ( bdev_logical_block_size ( bdev ) - 1 ) )
return - EINVAL ;
/* Invalidate the page cache, including dirty pages. */
mapping = bdev - > bd_inode - > i_mapping ;
truncate_inode_pages_range ( mapping , start , end ) ;
switch ( mode ) {
case FALLOC_FL_ZERO_RANGE :
case FALLOC_FL_ZERO_RANGE | FALLOC_FL_KEEP_SIZE :
error = blkdev_issue_zeroout ( bdev , start > > 9 , len > > 9 ,
2017-04-05 20:21:08 +03:00
GFP_KERNEL , BLKDEV_ZERO_NOUNMAP ) ;
2016-10-11 23:51:11 +03:00
break ;
case FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE :
2017-04-05 20:21:11 +03:00
error = blkdev_issue_zeroout ( bdev , start > > 9 , len > > 9 ,
GFP_KERNEL , BLKDEV_ZERO_NOFALLBACK ) ;
2016-10-11 23:51:11 +03:00
break ;
case FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE | FALLOC_FL_NO_HIDE_STALE :
error = blkdev_issue_discard ( bdev , start > > 9 , len > > 9 ,
GFP_KERNEL , 0 ) ;
break ;
default :
return - EOPNOTSUPP ;
}
if ( error )
return error ;
/*
* Invalidate again ; if someone wandered in and dirtied a page ,
* the caller will be given - EBUSY . The third argument is
* inclusive , so the rounding here is safe .
*/
return invalidate_inode_pages2_range ( mapping ,
start > > PAGE_SHIFT ,
end > > PAGE_SHIFT ) ;
}
2006-03-28 13:56:42 +04:00
const struct file_operations def_blk_fops = {
2005-04-17 02:20:36 +04:00
. open = blkdev_open ,
. release = blkdev_close ,
. llseek = block_llseek ,
2014-04-03 04:02:21 +04:00
. read_iter = blkdev_read_iter ,
2014-04-03 11:21:50 +04:00
. write_iter = blkdev_write_iter ,
2016-05-07 21:40:28 +03:00
. mmap = generic_file_mmap ,
2010-04-07 01:35:00 +04:00
. fsync = blkdev_fsync ,
2005-06-23 11:10:15 +04:00
. unlocked_ioctl = block_ioctl ,
2005-04-17 02:20:36 +04:00
# ifdef CONFIG_COMPAT
. compat_ioctl = compat_blkdev_ioctl ,
# endif
2012-11-29 22:49:50 +04:00
. splice_read = generic_file_splice_read ,
2014-04-05 12:27:08 +04:00
. splice_write = iter_file_splice_write ,
2016-10-11 23:51:11 +03:00
. fallocate = blkdev_fallocate ,
2005-04-17 02:20:36 +04:00
} ;
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 ) ;
2008-09-19 11:17:36 +04:00
res = blkdev_ioctl ( bdev , 0 , cmd , arg ) ;
2005-04-17 02:20:36 +04:00
set_fs ( old_fs ) ;
return res ;
}
EXPORT_SYMBOL ( ioctl_by_bdev ) ;
/**
* lookup_bdev - lookup a struct block_device by name
2009-01-07 01:41:15 +03:00
* @ pathname : special file representing the block device
2005-04-17 02:20:36 +04:00
*
2008-10-09 12:42:38 +04:00
* Get a reference to the blockdevice at @ pathname in the current
2005-04-17 02:20:36 +04:00
* namespace if possible and return it . Return ERR_PTR ( error )
* otherwise .
*/
2008-08-02 09:04:36 +04:00
struct block_device * lookup_bdev ( const char * pathname )
2005-04-17 02:20:36 +04:00
{
struct block_device * bdev ;
struct inode * inode ;
2008-08-02 09:04:36 +04:00
struct path path ;
2005-04-17 02:20:36 +04:00
int error ;
2008-08-02 09:04:36 +04:00
if ( ! pathname | | ! * pathname )
2005-04-17 02:20:36 +04:00
return ERR_PTR ( - EINVAL ) ;
2008-08-02 09:04:36 +04:00
error = kern_path ( pathname , LOOKUP_FOLLOW , & path ) ;
2005-04-17 02:20:36 +04:00
if ( error )
return ERR_PTR ( error ) ;
2015-03-18 01:26:21 +03:00
inode = d_backing_inode ( path . dentry ) ;
2005-04-17 02:20:36 +04:00
error = - ENOTBLK ;
if ( ! S_ISBLK ( inode - > i_mode ) )
goto fail ;
error = - EACCES ;
2016-06-09 23:34:02 +03:00
if ( ! may_open_dev ( & path ) )
2005-04-17 02:20:36 +04:00
goto fail ;
error = - ENOMEM ;
bdev = bd_acquire ( inode ) ;
if ( ! bdev )
goto fail ;
out :
2008-08-02 09:04:36 +04:00
path_put ( & path ) ;
2005-04-17 02:20:36 +04:00
return bdev ;
fail :
bdev = ERR_PTR ( error ) ;
goto out ;
}
2008-08-01 13:00:11 +04:00
EXPORT_SYMBOL ( lookup_bdev ) ;
2005-04-17 02:20:36 +04:00
2011-02-24 09:25:47 +03:00
int __invalidate_device ( struct block_device * bdev , bool kill_dirty )
2006-08-29 22:06:11 +04:00
{
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 ) ;
2011-02-24 09:25:47 +03:00
res = invalidate_inodes ( sb , kill_dirty ) ;
2006-08-29 22:06:11 +04:00
drop_super ( sb ) ;
}
2007-05-07 01:49:54 +04:00
invalidate_bdev ( bdev ) ;
2006-08-29 22:06:11 +04:00
return res ;
}
EXPORT_SYMBOL ( __invalidate_device ) ;
2012-07-03 18:45:31 +04:00
void iterate_bdevs ( void ( * func ) ( struct block_device * , void * ) , void * arg )
{
struct inode * inode , * old_inode = NULL ;
2015-03-04 20:37:22 +03:00
spin_lock ( & blockdev_superblock - > s_inode_list_lock ) ;
2012-07-03 18:45:31 +04:00
list_for_each_entry ( inode , & blockdev_superblock - > s_inodes , i_sb_list ) {
struct address_space * mapping = inode - > i_mapping ;
2016-12-01 11:18:28 +03:00
struct block_device * bdev ;
2012-07-03 18:45:31 +04:00
spin_lock ( & inode - > i_lock ) ;
if ( inode - > i_state & ( I_FREEING | I_WILL_FREE | I_NEW ) | |
mapping - > nrpages = = 0 ) {
spin_unlock ( & inode - > i_lock ) ;
continue ;
}
__iget ( inode ) ;
spin_unlock ( & inode - > i_lock ) ;
2015-03-04 20:37:22 +03:00
spin_unlock ( & blockdev_superblock - > s_inode_list_lock ) ;
2012-07-03 18:45:31 +04:00
/*
* We hold a reference to ' inode ' so it couldn ' t have been
* removed from s_inodes list while we dropped the
2015-03-04 20:37:22 +03:00
* s_inode_list_lock We cannot iput the inode now as we can
2012-07-03 18:45:31 +04:00
* be holding the last reference and we cannot iput it under
2015-03-04 20:37:22 +03:00
* s_inode_list_lock . So we keep the reference and iput it
2012-07-03 18:45:31 +04:00
* later .
*/
iput ( old_inode ) ;
old_inode = inode ;
2016-12-01 11:18:28 +03:00
bdev = I_BDEV ( inode ) ;
2012-07-03 18:45:31 +04:00
2016-12-01 11:18:28 +03:00
mutex_lock ( & bdev - > bd_mutex ) ;
if ( bdev - > bd_openers )
func ( bdev , arg ) ;
mutex_unlock ( & bdev - > bd_mutex ) ;
2012-07-03 18:45:31 +04:00
2015-03-04 20:37:22 +03:00
spin_lock ( & blockdev_superblock - > s_inode_list_lock ) ;
2012-07-03 18:45:31 +04:00
}
2015-03-04 20:37:22 +03:00
spin_unlock ( & blockdev_superblock - > s_inode_list_lock ) ;
2012-07-03 18:45:31 +04:00
iput ( old_inode ) ;
}