2005-04-17 00:20:36 +02:00
|
|
|
/*
|
2005-11-02 04:58:39 +01:00
|
|
|
* Copyright (c) 2000-2005 Silicon Graphics, Inc.
|
|
|
|
* All Rights Reserved.
|
2005-04-17 00:20:36 +02:00
|
|
|
*
|
2005-11-02 04:58:39 +01:00
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public License as
|
2005-04-17 00:20:36 +02:00
|
|
|
* published by the Free Software Foundation.
|
|
|
|
*
|
2005-11-02 04:58:39 +01:00
|
|
|
* This program is distributed in the hope that it would be useful,
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
|
|
* GNU General Public License for more details.
|
2005-04-17 00:20:36 +02:00
|
|
|
*
|
2005-11-02 04:58:39 +01:00
|
|
|
* You should have received a copy of the GNU General Public License
|
|
|
|
* along with this program; if not, write the Free Software Foundation,
|
|
|
|
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
#include "xfs.h"
|
2010-02-15 10:44:46 +01:00
|
|
|
#include "xfs_fs.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_log.h"
|
|
|
|
#include "xfs_sb.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_ag.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_trans.h"
|
|
|
|
#include "xfs_mount.h"
|
|
|
|
#include "xfs_bmap_btree.h"
|
|
|
|
#include "xfs_alloc.h"
|
|
|
|
#include "xfs_dinode.h"
|
|
|
|
#include "xfs_inode.h"
|
2010-02-15 10:44:48 +01:00
|
|
|
#include "xfs_inode_item.h"
|
2010-02-15 10:44:46 +01:00
|
|
|
#include "xfs_bmap.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_error.h"
|
2007-08-29 02:58:01 +02:00
|
|
|
#include "xfs_vnodeops.h"
|
2008-11-28 04:23:32 +01:00
|
|
|
#include "xfs_da_btree.h"
|
2008-12-03 13:55:34 +01:00
|
|
|
#include "xfs_ioctl.h"
|
2010-02-15 10:44:46 +01:00
|
|
|
#include "xfs_trace.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
#include <linux/dcache.h>
|
2011-01-14 13:07:43 +01:00
|
|
|
#include <linux/falloc.h>
|
2012-08-21 11:11:57 +02:00
|
|
|
#include <linux/pagevec.h>
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2009-09-27 20:29:37 +02:00
|
|
|
static const struct vm_operations_struct xfs_file_vm_ops;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2011-01-12 01:37:10 +01:00
|
|
|
/*
|
|
|
|
* Locking primitives for read and write IO paths to ensure we consistently use
|
|
|
|
* and order the inode->i_mutex, ip->i_lock and ip->i_iolock.
|
|
|
|
*/
|
|
|
|
static inline void
|
|
|
|
xfs_rw_ilock(
|
|
|
|
struct xfs_inode *ip,
|
|
|
|
int type)
|
|
|
|
{
|
|
|
|
if (type & XFS_IOLOCK_EXCL)
|
|
|
|
mutex_lock(&VFS_I(ip)->i_mutex);
|
|
|
|
xfs_ilock(ip, type);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
xfs_rw_iunlock(
|
|
|
|
struct xfs_inode *ip,
|
|
|
|
int type)
|
|
|
|
{
|
|
|
|
xfs_iunlock(ip, type);
|
|
|
|
if (type & XFS_IOLOCK_EXCL)
|
|
|
|
mutex_unlock(&VFS_I(ip)->i_mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
xfs_rw_ilock_demote(
|
|
|
|
struct xfs_inode *ip,
|
|
|
|
int type)
|
|
|
|
{
|
|
|
|
xfs_ilock_demote(ip, type);
|
|
|
|
if (type & XFS_IOLOCK_EXCL)
|
|
|
|
mutex_unlock(&VFS_I(ip)->i_mutex);
|
|
|
|
}
|
|
|
|
|
2010-02-15 10:44:46 +01:00
|
|
|
/*
|
|
|
|
* xfs_iozero
|
|
|
|
*
|
|
|
|
* xfs_iozero clears the specified range of buffer supplied,
|
|
|
|
* and marks all the affected blocks as valid and modified. If
|
|
|
|
* an affected block is not allocated, it will be allocated. If
|
|
|
|
* an affected block is not completely overwritten, and is not
|
|
|
|
* valid before the operation, it will be read from disk before
|
|
|
|
* being partially zeroed.
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_iozero(
|
|
|
|
struct xfs_inode *ip, /* inode */
|
|
|
|
loff_t pos, /* offset in file */
|
|
|
|
size_t count) /* size of data to zero */
|
|
|
|
{
|
|
|
|
struct page *page;
|
|
|
|
struct address_space *mapping;
|
|
|
|
int status;
|
|
|
|
|
|
|
|
mapping = VFS_I(ip)->i_mapping;
|
|
|
|
do {
|
|
|
|
unsigned offset, bytes;
|
|
|
|
void *fsdata;
|
|
|
|
|
|
|
|
offset = (pos & (PAGE_CACHE_SIZE -1)); /* Within page */
|
|
|
|
bytes = PAGE_CACHE_SIZE - offset;
|
|
|
|
if (bytes > count)
|
|
|
|
bytes = count;
|
|
|
|
|
|
|
|
status = pagecache_write_begin(NULL, mapping, pos, bytes,
|
|
|
|
AOP_FLAG_UNINTERRUPTIBLE,
|
|
|
|
&page, &fsdata);
|
|
|
|
if (status)
|
|
|
|
break;
|
|
|
|
|
|
|
|
zero_user(page, offset, bytes);
|
|
|
|
|
|
|
|
status = pagecache_write_end(NULL, mapping, pos, bytes, bytes,
|
|
|
|
page, fsdata);
|
|
|
|
WARN_ON(status <= 0); /* can't return less than zero! */
|
|
|
|
pos += bytes;
|
|
|
|
count -= bytes;
|
|
|
|
status = 0;
|
|
|
|
} while (count);
|
|
|
|
|
|
|
|
return (-status);
|
|
|
|
}
|
|
|
|
|
2011-10-02 16:25:16 +02:00
|
|
|
/*
|
|
|
|
* Fsync operations on directories are much simpler than on regular files,
|
|
|
|
* as there is no file data to flush, and thus also no need for explicit
|
|
|
|
* cache flush operations, and there are no non-transaction metadata updates
|
|
|
|
* on directories either.
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_dir_fsync(
|
|
|
|
struct file *file,
|
|
|
|
loff_t start,
|
|
|
|
loff_t end,
|
|
|
|
int datasync)
|
|
|
|
{
|
|
|
|
struct xfs_inode *ip = XFS_I(file->f_mapping->host);
|
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
|
|
|
xfs_lsn_t lsn = 0;
|
|
|
|
|
|
|
|
trace_xfs_dir_fsync(ip);
|
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_SHARED);
|
|
|
|
if (xfs_ipincount(ip))
|
|
|
|
lsn = ip->i_itemp->ili_last_lsn;
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_SHARED);
|
|
|
|
|
|
|
|
if (!lsn)
|
|
|
|
return 0;
|
|
|
|
return _xfs_log_force_lsn(mp, lsn, XFS_LOG_SYNC, NULL);
|
|
|
|
}
|
|
|
|
|
2010-02-15 10:44:48 +01:00
|
|
|
STATIC int
|
|
|
|
xfs_file_fsync(
|
|
|
|
struct file *file,
|
2011-07-17 02:44:56 +02:00
|
|
|
loff_t start,
|
|
|
|
loff_t end,
|
2010-02-15 10:44:48 +01:00
|
|
|
int datasync)
|
|
|
|
{
|
2010-05-26 17:53:25 +02:00
|
|
|
struct inode *inode = file->f_mapping->host;
|
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
2011-06-16 14:02:23 +02:00
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
2010-02-15 10:44:48 +01:00
|
|
|
int error = 0;
|
|
|
|
int log_flushed = 0;
|
2011-09-19 16:55:51 +02:00
|
|
|
xfs_lsn_t lsn = 0;
|
2010-02-15 10:44:48 +01:00
|
|
|
|
2010-06-24 03:57:09 +02:00
|
|
|
trace_xfs_file_fsync(ip);
|
2010-02-15 10:44:48 +01:00
|
|
|
|
2011-07-17 02:44:56 +02:00
|
|
|
error = filemap_write_and_wait_range(inode->i_mapping, start, end);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
2011-06-16 14:02:23 +02:00
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
2010-02-15 10:44:48 +01:00
|
|
|
return -XFS_ERROR(EIO);
|
|
|
|
|
|
|
|
xfs_iflags_clear(ip, XFS_ITRUNCATED);
|
|
|
|
|
2011-06-16 14:02:23 +02:00
|
|
|
if (mp->m_flags & XFS_MOUNT_BARRIER) {
|
|
|
|
/*
|
|
|
|
* If we have an RT and/or log subvolume we need to make sure
|
|
|
|
* to flush the write cache the device used for file data
|
|
|
|
* first. This is to ensure newly written file data make
|
|
|
|
* it to disk before logging the new inode size in case of
|
|
|
|
* an extending write.
|
|
|
|
*/
|
|
|
|
if (XFS_IS_REALTIME_INODE(ip))
|
|
|
|
xfs_blkdev_issue_flush(mp->m_rtdev_targp);
|
|
|
|
else if (mp->m_logdev_targp != mp->m_ddev_targp)
|
|
|
|
xfs_blkdev_issue_flush(mp->m_ddev_targp);
|
|
|
|
}
|
|
|
|
|
2010-02-15 10:44:48 +01:00
|
|
|
/*
|
2012-02-29 10:53:52 +01:00
|
|
|
* All metadata updates are logged, which means that we just have
|
|
|
|
* to flush the log up to the latest LSN that touched the inode.
|
2010-02-15 10:44:48 +01:00
|
|
|
*/
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_SHARED);
|
2012-02-29 10:53:55 +01:00
|
|
|
if (xfs_ipincount(ip)) {
|
|
|
|
if (!datasync ||
|
|
|
|
(ip->i_itemp->ili_fields & ~XFS_ILOG_TIMESTAMP))
|
|
|
|
lsn = ip->i_itemp->ili_last_lsn;
|
|
|
|
}
|
2012-02-29 10:53:52 +01:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_SHARED);
|
2010-02-15 10:44:48 +01:00
|
|
|
|
2012-02-29 10:53:52 +01:00
|
|
|
if (lsn)
|
2011-09-19 16:55:51 +02:00
|
|
|
error = _xfs_log_force_lsn(mp, lsn, XFS_LOG_SYNC, &log_flushed);
|
|
|
|
|
2011-06-16 14:02:23 +02:00
|
|
|
/*
|
|
|
|
* If we only have a single device, and the log force about was
|
|
|
|
* a no-op we might have to flush the data device cache here.
|
|
|
|
* This can only happen for fdatasync/O_DSYNC if we were overwriting
|
|
|
|
* an already allocated file and thus do not have any metadata to
|
|
|
|
* commit.
|
|
|
|
*/
|
|
|
|
if ((mp->m_flags & XFS_MOUNT_BARRIER) &&
|
|
|
|
mp->m_logdev_targp == mp->m_ddev_targp &&
|
|
|
|
!XFS_IS_REALTIME_INODE(ip) &&
|
|
|
|
!log_flushed)
|
|
|
|
xfs_blkdev_issue_flush(mp->m_ddev_targp);
|
2010-02-15 10:44:48 +01:00
|
|
|
|
|
|
|
return -error;
|
|
|
|
}
|
|
|
|
|
2010-02-15 10:44:47 +01:00
|
|
|
STATIC ssize_t
|
|
|
|
xfs_file_aio_read(
|
2010-02-15 10:44:46 +01:00
|
|
|
struct kiocb *iocb,
|
|
|
|
const struct iovec *iovp,
|
2010-02-15 10:44:47 +01:00
|
|
|
unsigned long nr_segs,
|
|
|
|
loff_t pos)
|
2010-02-15 10:44:46 +01:00
|
|
|
{
|
|
|
|
struct file *file = iocb->ki_filp;
|
|
|
|
struct inode *inode = file->f_mapping->host;
|
2010-02-15 10:44:47 +01:00
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
2010-02-15 10:44:46 +01:00
|
|
|
size_t size = 0;
|
|
|
|
ssize_t ret = 0;
|
2010-02-15 10:44:47 +01:00
|
|
|
int ioflags = 0;
|
2010-02-15 10:44:46 +01:00
|
|
|
xfs_fsize_t n;
|
|
|
|
|
|
|
|
XFS_STATS_INC(xs_read_calls);
|
|
|
|
|
2010-02-15 10:44:47 +01:00
|
|
|
BUG_ON(iocb->ki_pos != pos);
|
|
|
|
|
|
|
|
if (unlikely(file->f_flags & O_DIRECT))
|
|
|
|
ioflags |= IO_ISDIRECT;
|
|
|
|
if (file->f_mode & FMODE_NOCMTIME)
|
|
|
|
ioflags |= IO_INVIS;
|
|
|
|
|
2012-06-08 07:45:44 +02:00
|
|
|
ret = generic_segment_checks(iovp, &nr_segs, &size, VERIFY_WRITE);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
2010-02-15 10:44:46 +01:00
|
|
|
|
|
|
|
if (unlikely(ioflags & IO_ISDIRECT)) {
|
|
|
|
xfs_buftarg_t *target =
|
|
|
|
XFS_IS_REALTIME_INODE(ip) ?
|
|
|
|
mp->m_rtdev_targp : mp->m_ddev_targp;
|
2010-02-15 10:44:47 +01:00
|
|
|
if ((iocb->ki_pos & target->bt_smask) ||
|
2010-02-15 10:44:46 +01:00
|
|
|
(size & target->bt_smask)) {
|
2011-12-18 21:00:11 +01:00
|
|
|
if (iocb->ki_pos == i_size_read(inode))
|
2010-02-15 10:44:47 +01:00
|
|
|
return 0;
|
2010-02-15 10:44:46 +01:00
|
|
|
return -XFS_ERROR(EINVAL);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-06-08 07:44:54 +02:00
|
|
|
n = mp->m_super->s_maxbytes - iocb->ki_pos;
|
2010-02-15 10:44:47 +01:00
|
|
|
if (n <= 0 || size == 0)
|
2010-02-15 10:44:46 +01:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (n < size)
|
|
|
|
size = n;
|
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
|
|
|
return -EIO;
|
|
|
|
|
2011-08-25 09:17:01 +02:00
|
|
|
/*
|
|
|
|
* Locking is a bit tricky here. If we take an exclusive lock
|
|
|
|
* for direct IO, we effectively serialise all new concurrent
|
|
|
|
* read IO to this file and block it behind IO that is currently in
|
|
|
|
* progress because IO in progress holds the IO lock shared. We only
|
|
|
|
* need to hold the lock exclusive to blow away the page cache, so
|
|
|
|
* only take lock exclusively if the page cache needs invalidation.
|
|
|
|
* This allows the normal direct IO case of no page cache pages to
|
|
|
|
* proceeed concurrently without serialisation.
|
|
|
|
*/
|
|
|
|
xfs_rw_ilock(ip, XFS_IOLOCK_SHARED);
|
|
|
|
if ((ioflags & IO_ISDIRECT) && inode->i_mapping->nrpages) {
|
|
|
|
xfs_rw_iunlock(ip, XFS_IOLOCK_SHARED);
|
2011-01-12 01:37:10 +01:00
|
|
|
xfs_rw_ilock(ip, XFS_IOLOCK_EXCL);
|
|
|
|
|
2010-02-15 10:44:47 +01:00
|
|
|
if (inode->i_mapping->nrpages) {
|
|
|
|
ret = -xfs_flushinval_pages(ip,
|
|
|
|
(iocb->ki_pos & PAGE_CACHE_MASK),
|
|
|
|
-1, FI_REMAPF_LOCKED);
|
2011-01-12 01:37:10 +01:00
|
|
|
if (ret) {
|
|
|
|
xfs_rw_iunlock(ip, XFS_IOLOCK_EXCL);
|
|
|
|
return ret;
|
|
|
|
}
|
2010-02-15 10:44:47 +01:00
|
|
|
}
|
2011-01-12 01:37:10 +01:00
|
|
|
xfs_rw_ilock_demote(ip, XFS_IOLOCK_EXCL);
|
2011-08-25 09:17:01 +02:00
|
|
|
}
|
2010-02-15 10:44:46 +01:00
|
|
|
|
2010-02-15 10:44:47 +01:00
|
|
|
trace_xfs_file_read(ip, size, iocb->ki_pos, ioflags);
|
2010-02-15 10:44:46 +01:00
|
|
|
|
2010-02-15 10:44:47 +01:00
|
|
|
ret = generic_file_aio_read(iocb, iovp, nr_segs, iocb->ki_pos);
|
2010-02-15 10:44:46 +01:00
|
|
|
if (ret > 0)
|
|
|
|
XFS_STATS_ADD(xs_read_bytes, ret);
|
|
|
|
|
2011-01-12 01:37:10 +01:00
|
|
|
xfs_rw_iunlock(ip, XFS_IOLOCK_SHARED);
|
2010-02-15 10:44:46 +01:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2010-02-15 10:44:47 +01:00
|
|
|
STATIC ssize_t
|
|
|
|
xfs_file_splice_read(
|
2010-02-15 10:44:46 +01:00
|
|
|
struct file *infilp,
|
|
|
|
loff_t *ppos,
|
|
|
|
struct pipe_inode_info *pipe,
|
|
|
|
size_t count,
|
2010-02-15 10:44:47 +01:00
|
|
|
unsigned int flags)
|
2010-02-15 10:44:46 +01:00
|
|
|
{
|
2010-02-15 10:44:47 +01:00
|
|
|
struct xfs_inode *ip = XFS_I(infilp->f_mapping->host);
|
|
|
|
int ioflags = 0;
|
2010-02-15 10:44:46 +01:00
|
|
|
ssize_t ret;
|
|
|
|
|
|
|
|
XFS_STATS_INC(xs_read_calls);
|
2010-02-15 10:44:47 +01:00
|
|
|
|
|
|
|
if (infilp->f_mode & FMODE_NOCMTIME)
|
|
|
|
ioflags |= IO_INVIS;
|
|
|
|
|
2010-02-15 10:44:46 +01:00
|
|
|
if (XFS_FORCED_SHUTDOWN(ip->i_mount))
|
|
|
|
return -EIO;
|
|
|
|
|
2011-01-12 01:37:10 +01:00
|
|
|
xfs_rw_ilock(ip, XFS_IOLOCK_SHARED);
|
2010-02-15 10:44:46 +01:00
|
|
|
|
|
|
|
trace_xfs_file_splice_read(ip, count, *ppos, ioflags);
|
|
|
|
|
|
|
|
ret = generic_file_splice_read(infilp, ppos, pipe, count, flags);
|
|
|
|
if (ret > 0)
|
|
|
|
XFS_STATS_ADD(xs_read_bytes, ret);
|
|
|
|
|
2011-01-12 01:37:10 +01:00
|
|
|
xfs_rw_iunlock(ip, XFS_IOLOCK_SHARED);
|
2010-02-15 10:44:46 +01:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2011-01-12 01:37:10 +01:00
|
|
|
/*
|
|
|
|
* xfs_file_splice_write() does not use xfs_rw_ilock() because
|
|
|
|
* generic_file_splice_write() takes the i_mutex itself. This, in theory,
|
|
|
|
* couuld cause lock inversions between the aio_write path and the splice path
|
|
|
|
* if someone is doing concurrent splice(2) based writes and write(2) based
|
|
|
|
* writes to the same inode. The only real way to fix this is to re-implement
|
|
|
|
* the generic code here with correct locking orders.
|
|
|
|
*/
|
2010-02-15 10:44:47 +01:00
|
|
|
STATIC ssize_t
|
|
|
|
xfs_file_splice_write(
|
2010-02-15 10:44:46 +01:00
|
|
|
struct pipe_inode_info *pipe,
|
|
|
|
struct file *outfilp,
|
|
|
|
loff_t *ppos,
|
|
|
|
size_t count,
|
2010-02-15 10:44:47 +01:00
|
|
|
unsigned int flags)
|
2010-02-15 10:44:46 +01:00
|
|
|
{
|
|
|
|
struct inode *inode = outfilp->f_mapping->host;
|
2010-02-15 10:44:47 +01:00
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
|
|
|
int ioflags = 0;
|
|
|
|
ssize_t ret;
|
2010-02-15 10:44:46 +01:00
|
|
|
|
|
|
|
XFS_STATS_INC(xs_write_calls);
|
2010-02-15 10:44:47 +01:00
|
|
|
|
|
|
|
if (outfilp->f_mode & FMODE_NOCMTIME)
|
|
|
|
ioflags |= IO_INVIS;
|
|
|
|
|
2010-02-15 10:44:46 +01:00
|
|
|
if (XFS_FORCED_SHUTDOWN(ip->i_mount))
|
|
|
|
return -EIO;
|
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_IOLOCK_EXCL);
|
|
|
|
|
|
|
|
trace_xfs_file_splice_write(ip, count, *ppos, ioflags);
|
|
|
|
|
|
|
|
ret = generic_file_splice_write(pipe, outfilp, ppos, count, flags);
|
2011-12-18 21:00:11 +01:00
|
|
|
if (ret > 0)
|
|
|
|
XFS_STATS_ADD(xs_write_bytes, ret);
|
2010-02-15 10:44:46 +01:00
|
|
|
|
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2012-03-27 16:34:49 +02:00
|
|
|
* This routine is called to handle zeroing any space in the last block of the
|
|
|
|
* file that is beyond the EOF. We do this since the size is being increased
|
|
|
|
* without writing anything to that block and we don't want to read the
|
|
|
|
* garbage on the disk.
|
2010-02-15 10:44:46 +01:00
|
|
|
*/
|
|
|
|
STATIC int /* error (positive) */
|
|
|
|
xfs_zero_last_block(
|
2012-03-27 16:34:49 +02:00
|
|
|
struct xfs_inode *ip,
|
|
|
|
xfs_fsize_t offset,
|
|
|
|
xfs_fsize_t isize)
|
2010-02-15 10:44:46 +01:00
|
|
|
{
|
2012-03-27 16:34:49 +02:00
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
|
|
|
xfs_fileoff_t last_fsb = XFS_B_TO_FSBT(mp, isize);
|
|
|
|
int zero_offset = XFS_B_FSB_OFFSET(mp, isize);
|
|
|
|
int zero_len;
|
|
|
|
int nimaps = 1;
|
|
|
|
int error = 0;
|
|
|
|
struct xfs_bmbt_irec imap;
|
2010-02-15 10:44:46 +01:00
|
|
|
|
2012-03-27 16:34:49 +02:00
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
2011-09-18 22:40:45 +02:00
|
|
|
error = xfs_bmapi_read(ip, last_fsb, 1, &imap, &nimaps, 0);
|
2012-03-27 16:34:49 +02:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
2011-09-18 22:40:45 +02:00
|
|
|
if (error)
|
2010-02-15 10:44:46 +01:00
|
|
|
return error;
|
2012-03-27 16:34:49 +02:00
|
|
|
|
2010-02-15 10:44:46 +01:00
|
|
|
ASSERT(nimaps > 0);
|
2012-03-27 16:34:49 +02:00
|
|
|
|
2010-02-15 10:44:46 +01:00
|
|
|
/*
|
|
|
|
* If the block underlying isize is just a hole, then there
|
|
|
|
* is nothing to zero.
|
|
|
|
*/
|
2012-03-27 16:34:49 +02:00
|
|
|
if (imap.br_startblock == HOLESTARTBLOCK)
|
2010-02-15 10:44:46 +01:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
zero_len = mp->m_sb.sb_blocksize - zero_offset;
|
|
|
|
if (isize + zero_len > offset)
|
|
|
|
zero_len = offset - isize;
|
2012-03-27 16:34:49 +02:00
|
|
|
return xfs_iozero(ip, isize, zero_len);
|
2010-02-15 10:44:46 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2012-03-27 16:34:49 +02:00
|
|
|
* Zero any on disk space between the current EOF and the new, larger EOF.
|
|
|
|
*
|
|
|
|
* This handles the normal case of zeroing the remainder of the last block in
|
|
|
|
* the file and the unusual case of zeroing blocks out beyond the size of the
|
|
|
|
* file. This second case only happens with fixed size extents and when the
|
|
|
|
* system crashes before the inode size was updated but after blocks were
|
|
|
|
* allocated.
|
|
|
|
*
|
|
|
|
* Expects the iolock to be held exclusive, and will take the ilock internally.
|
2010-02-15 10:44:46 +01:00
|
|
|
*/
|
|
|
|
int /* error (positive) */
|
|
|
|
xfs_zero_eof(
|
2012-03-27 16:34:49 +02:00
|
|
|
struct xfs_inode *ip,
|
|
|
|
xfs_off_t offset, /* starting I/O offset */
|
|
|
|
xfs_fsize_t isize) /* current inode size */
|
2010-02-15 10:44:46 +01:00
|
|
|
{
|
2012-03-27 16:34:49 +02:00
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
|
|
|
xfs_fileoff_t start_zero_fsb;
|
|
|
|
xfs_fileoff_t end_zero_fsb;
|
|
|
|
xfs_fileoff_t zero_count_fsb;
|
|
|
|
xfs_fileoff_t last_fsb;
|
|
|
|
xfs_fileoff_t zero_off;
|
|
|
|
xfs_fsize_t zero_len;
|
|
|
|
int nimaps;
|
|
|
|
int error = 0;
|
|
|
|
struct xfs_bmbt_irec imap;
|
|
|
|
|
|
|
|
ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL));
|
2010-02-15 10:44:46 +01:00
|
|
|
ASSERT(offset > isize);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* First handle zeroing the block on which isize resides.
|
2012-03-27 16:34:49 +02:00
|
|
|
*
|
2010-02-15 10:44:46 +01:00
|
|
|
* We only zero a part of that block so it is handled specially.
|
|
|
|
*/
|
2012-03-27 16:34:49 +02:00
|
|
|
if (XFS_B_FSB_OFFSET(mp, isize) != 0) {
|
|
|
|
error = xfs_zero_last_block(ip, offset, isize);
|
|
|
|
if (error)
|
|
|
|
return error;
|
2010-02-15 10:44:46 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2012-03-27 16:34:49 +02:00
|
|
|
* Calculate the range between the new size and the old where blocks
|
|
|
|
* needing to be zeroed may exist.
|
|
|
|
*
|
|
|
|
* To get the block where the last byte in the file currently resides,
|
|
|
|
* we need to subtract one from the size and truncate back to a block
|
|
|
|
* boundary. We subtract 1 in case the size is exactly on a block
|
|
|
|
* boundary.
|
2010-02-15 10:44:46 +01:00
|
|
|
*/
|
|
|
|
last_fsb = isize ? XFS_B_TO_FSBT(mp, isize - 1) : (xfs_fileoff_t)-1;
|
|
|
|
start_zero_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)isize);
|
|
|
|
end_zero_fsb = XFS_B_TO_FSBT(mp, offset - 1);
|
|
|
|
ASSERT((xfs_sfiloff_t)last_fsb < (xfs_sfiloff_t)start_zero_fsb);
|
|
|
|
if (last_fsb == end_zero_fsb) {
|
|
|
|
/*
|
|
|
|
* The size was only incremented on its last block.
|
|
|
|
* We took care of that above, so just return.
|
|
|
|
*/
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
ASSERT(start_zero_fsb <= end_zero_fsb);
|
|
|
|
while (start_zero_fsb <= end_zero_fsb) {
|
|
|
|
nimaps = 1;
|
|
|
|
zero_count_fsb = end_zero_fsb - start_zero_fsb + 1;
|
2012-03-27 16:34:49 +02:00
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
2011-09-18 22:40:45 +02:00
|
|
|
error = xfs_bmapi_read(ip, start_zero_fsb, zero_count_fsb,
|
|
|
|
&imap, &nimaps, 0);
|
2012-03-27 16:34:49 +02:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
|
|
|
if (error)
|
2010-02-15 10:44:46 +01:00
|
|
|
return error;
|
2012-03-27 16:34:49 +02:00
|
|
|
|
2010-02-15 10:44:46 +01:00
|
|
|
ASSERT(nimaps > 0);
|
|
|
|
|
|
|
|
if (imap.br_state == XFS_EXT_UNWRITTEN ||
|
|
|
|
imap.br_startblock == HOLESTARTBLOCK) {
|
|
|
|
start_zero_fsb = imap.br_startoff + imap.br_blockcount;
|
|
|
|
ASSERT(start_zero_fsb <= (end_zero_fsb + 1));
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* There are blocks we need to zero.
|
|
|
|
*/
|
|
|
|
zero_off = XFS_FSB_TO_B(mp, start_zero_fsb);
|
|
|
|
zero_len = XFS_FSB_TO_B(mp, imap.br_blockcount);
|
|
|
|
|
|
|
|
if ((zero_off + zero_len) > offset)
|
|
|
|
zero_len = offset - zero_off;
|
|
|
|
|
|
|
|
error = xfs_iozero(ip, zero_off, zero_len);
|
2012-03-27 16:34:49 +02:00
|
|
|
if (error)
|
|
|
|
return error;
|
2010-02-15 10:44:46 +01:00
|
|
|
|
|
|
|
start_zero_fsb = imap.br_startoff + imap.br_blockcount;
|
|
|
|
ASSERT(start_zero_fsb <= (end_zero_fsb + 1));
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-01-11 00:23:42 +01:00
|
|
|
/*
|
|
|
|
* Common pre-write limit and setup checks.
|
|
|
|
*
|
2011-12-18 21:00:13 +01:00
|
|
|
* Called with the iolocked held either shared and exclusive according to
|
|
|
|
* @iolock, and returns with it held. Might upgrade the iolock to exclusive
|
|
|
|
* if called for a direct write beyond i_size.
|
2011-01-11 00:23:42 +01:00
|
|
|
*/
|
|
|
|
STATIC ssize_t
|
|
|
|
xfs_file_aio_write_checks(
|
|
|
|
struct file *file,
|
|
|
|
loff_t *pos,
|
|
|
|
size_t *count,
|
|
|
|
int *iolock)
|
|
|
|
{
|
|
|
|
struct inode *inode = file->f_mapping->host;
|
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
|
|
|
int error = 0;
|
|
|
|
|
xfs: don't serialise adjacent concurrent direct IO appending writes
For append write workloads, extending the file requires a certain
amount of exclusive locking to be done up front to ensure sanity in
things like ensuring that we've zeroed any allocated regions
between the old EOF and the start of the new IO.
For single threads, this typically isn't a problem, and for large
IOs we don't serialise enough for it to be a problem for two
threads on really fast block devices. However for smaller IO and
larger thread counts we have a problem.
Take 4 concurrent sequential, single block sized and aligned IOs.
After the first IO is submitted but before it completes, we end up
with this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
And the IO is done without exclusive locking because offset <=
ip->i_size. When we submit IO 2, we see offset > ip->i_size, and
grab the IO lock exclusive, because there is a chance we need to do
EOF zeroing. However, there is already an IO in progress that avoids
the need for IO zeroing because offset <= ip->i_new_size. hence we
could avoid holding the IO lock exlcusive for this. Hence after
submission of the second IO, we'd end up this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
There is no need to grab the i_mutex of the IO lock in exclusive
mode if we don't need to invalidate the page cache. Taking these
locks on every direct IO effective serialises them as taking the IO
lock in exclusive mode has to wait for all shared holders to drop
the lock. That only happens when IO is complete, so effective it
prevents dispatch of concurrent direct IO writes to the same inode.
And so you can see that for the third concurrent IO, we'd avoid
exclusive locking for the same reason we avoided the exclusive lock
for the second IO.
Fixing this is a bit more complex than that, because we need to hold
a write-submission local value of ip->i_new_size to that clearing
the value is only done if no other thread has updated it before our
IO completes.....
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
2011-08-25 09:17:02 +02:00
|
|
|
restart:
|
2011-01-11 00:23:42 +01:00
|
|
|
error = generic_write_checks(file, pos, count, S_ISBLK(inode->i_mode));
|
2012-03-27 16:34:47 +02:00
|
|
|
if (error)
|
2011-01-11 00:23:42 +01:00
|
|
|
return error;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the offset is beyond the size of the file, we need to zero any
|
|
|
|
* blocks that fall between the existing EOF and the start of this
|
2011-12-18 21:00:12 +01:00
|
|
|
* write. If zeroing is needed and we are currently holding the
|
2012-03-27 16:34:47 +02:00
|
|
|
* iolock shared, we need to update it to exclusive which implies
|
|
|
|
* having to redo all checks before.
|
2011-01-11 00:23:42 +01:00
|
|
|
*/
|
2011-12-18 21:00:12 +01:00
|
|
|
if (*pos > i_size_read(inode)) {
|
xfs: don't serialise adjacent concurrent direct IO appending writes
For append write workloads, extending the file requires a certain
amount of exclusive locking to be done up front to ensure sanity in
things like ensuring that we've zeroed any allocated regions
between the old EOF and the start of the new IO.
For single threads, this typically isn't a problem, and for large
IOs we don't serialise enough for it to be a problem for two
threads on really fast block devices. However for smaller IO and
larger thread counts we have a problem.
Take 4 concurrent sequential, single block sized and aligned IOs.
After the first IO is submitted but before it completes, we end up
with this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
And the IO is done without exclusive locking because offset <=
ip->i_size. When we submit IO 2, we see offset > ip->i_size, and
grab the IO lock exclusive, because there is a chance we need to do
EOF zeroing. However, there is already an IO in progress that avoids
the need for IO zeroing because offset <= ip->i_new_size. hence we
could avoid holding the IO lock exlcusive for this. Hence after
submission of the second IO, we'd end up this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
There is no need to grab the i_mutex of the IO lock in exclusive
mode if we don't need to invalidate the page cache. Taking these
locks on every direct IO effective serialises them as taking the IO
lock in exclusive mode has to wait for all shared holders to drop
the lock. That only happens when IO is complete, so effective it
prevents dispatch of concurrent direct IO writes to the same inode.
And so you can see that for the third concurrent IO, we'd avoid
exclusive locking for the same reason we avoided the exclusive lock
for the second IO.
Fixing this is a bit more complex than that, because we need to hold
a write-submission local value of ip->i_new_size to that clearing
the value is only done if no other thread has updated it before our
IO completes.....
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
2011-08-25 09:17:02 +02:00
|
|
|
if (*iolock == XFS_IOLOCK_SHARED) {
|
2012-03-27 16:34:47 +02:00
|
|
|
xfs_rw_iunlock(ip, *iolock);
|
xfs: don't serialise adjacent concurrent direct IO appending writes
For append write workloads, extending the file requires a certain
amount of exclusive locking to be done up front to ensure sanity in
things like ensuring that we've zeroed any allocated regions
between the old EOF and the start of the new IO.
For single threads, this typically isn't a problem, and for large
IOs we don't serialise enough for it to be a problem for two
threads on really fast block devices. However for smaller IO and
larger thread counts we have a problem.
Take 4 concurrent sequential, single block sized and aligned IOs.
After the first IO is submitted but before it completes, we end up
with this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
And the IO is done without exclusive locking because offset <=
ip->i_size. When we submit IO 2, we see offset > ip->i_size, and
grab the IO lock exclusive, because there is a chance we need to do
EOF zeroing. However, there is already an IO in progress that avoids
the need for IO zeroing because offset <= ip->i_new_size. hence we
could avoid holding the IO lock exlcusive for this. Hence after
submission of the second IO, we'd end up this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
There is no need to grab the i_mutex of the IO lock in exclusive
mode if we don't need to invalidate the page cache. Taking these
locks on every direct IO effective serialises them as taking the IO
lock in exclusive mode has to wait for all shared holders to drop
the lock. That only happens when IO is complete, so effective it
prevents dispatch of concurrent direct IO writes to the same inode.
And so you can see that for the third concurrent IO, we'd avoid
exclusive locking for the same reason we avoided the exclusive lock
for the second IO.
Fixing this is a bit more complex than that, because we need to hold
a write-submission local value of ip->i_new_size to that clearing
the value is only done if no other thread has updated it before our
IO completes.....
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
2011-08-25 09:17:02 +02:00
|
|
|
*iolock = XFS_IOLOCK_EXCL;
|
2012-03-27 16:34:47 +02:00
|
|
|
xfs_rw_ilock(ip, *iolock);
|
xfs: don't serialise adjacent concurrent direct IO appending writes
For append write workloads, extending the file requires a certain
amount of exclusive locking to be done up front to ensure sanity in
things like ensuring that we've zeroed any allocated regions
between the old EOF and the start of the new IO.
For single threads, this typically isn't a problem, and for large
IOs we don't serialise enough for it to be a problem for two
threads on really fast block devices. However for smaller IO and
larger thread counts we have a problem.
Take 4 concurrent sequential, single block sized and aligned IOs.
After the first IO is submitted but before it completes, we end up
with this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
And the IO is done without exclusive locking because offset <=
ip->i_size. When we submit IO 2, we see offset > ip->i_size, and
grab the IO lock exclusive, because there is a chance we need to do
EOF zeroing. However, there is already an IO in progress that avoids
the need for IO zeroing because offset <= ip->i_new_size. hence we
could avoid holding the IO lock exlcusive for this. Hence after
submission of the second IO, we'd end up this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
There is no need to grab the i_mutex of the IO lock in exclusive
mode if we don't need to invalidate the page cache. Taking these
locks on every direct IO effective serialises them as taking the IO
lock in exclusive mode has to wait for all shared holders to drop
the lock. That only happens when IO is complete, so effective it
prevents dispatch of concurrent direct IO writes to the same inode.
And so you can see that for the third concurrent IO, we'd avoid
exclusive locking for the same reason we avoided the exclusive lock
for the second IO.
Fixing this is a bit more complex than that, because we need to hold
a write-submission local value of ip->i_new_size to that clearing
the value is only done if no other thread has updated it before our
IO completes.....
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
2011-08-25 09:17:02 +02:00
|
|
|
goto restart;
|
|
|
|
}
|
2011-12-18 21:00:11 +01:00
|
|
|
error = -xfs_zero_eof(ip, *pos, i_size_read(inode));
|
2012-03-27 16:34:47 +02:00
|
|
|
if (error)
|
|
|
|
return error;
|
xfs: don't serialise adjacent concurrent direct IO appending writes
For append write workloads, extending the file requires a certain
amount of exclusive locking to be done up front to ensure sanity in
things like ensuring that we've zeroed any allocated regions
between the old EOF and the start of the new IO.
For single threads, this typically isn't a problem, and for large
IOs we don't serialise enough for it to be a problem for two
threads on really fast block devices. However for smaller IO and
larger thread counts we have a problem.
Take 4 concurrent sequential, single block sized and aligned IOs.
After the first IO is submitted but before it completes, we end up
with this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
And the IO is done without exclusive locking because offset <=
ip->i_size. When we submit IO 2, we see offset > ip->i_size, and
grab the IO lock exclusive, because there is a chance we need to do
EOF zeroing. However, there is already an IO in progress that avoids
the need for IO zeroing because offset <= ip->i_new_size. hence we
could avoid holding the IO lock exlcusive for this. Hence after
submission of the second IO, we'd end up this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
There is no need to grab the i_mutex of the IO lock in exclusive
mode if we don't need to invalidate the page cache. Taking these
locks on every direct IO effective serialises them as taking the IO
lock in exclusive mode has to wait for all shared holders to drop
the lock. That only happens when IO is complete, so effective it
prevents dispatch of concurrent direct IO writes to the same inode.
And so you can see that for the third concurrent IO, we'd avoid
exclusive locking for the same reason we avoided the exclusive lock
for the second IO.
Fixing this is a bit more complex than that, because we need to hold
a write-submission local value of ip->i_new_size to that clearing
the value is only done if no other thread has updated it before our
IO completes.....
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
2011-08-25 09:17:02 +02:00
|
|
|
}
|
2011-01-11 00:23:42 +01:00
|
|
|
|
2012-02-29 10:53:52 +01:00
|
|
|
/*
|
|
|
|
* Updating the timestamps will grab the ilock again from
|
|
|
|
* xfs_fs_dirty_inode, so we have to call it after dropping the
|
|
|
|
* lock above. Eventually we should look into a way to avoid
|
|
|
|
* the pointless lock roundtrip.
|
|
|
|
*/
|
2012-03-26 15:59:21 +02:00
|
|
|
if (likely(!(file->f_mode & FMODE_NOCMTIME))) {
|
|
|
|
error = file_update_time(file);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
2012-02-29 10:53:52 +01:00
|
|
|
|
2011-01-11 00:23:42 +01:00
|
|
|
/*
|
|
|
|
* If we're writing the file then make sure to clear the setuid and
|
|
|
|
* setgid bits if the process is not being run by root. This keeps
|
|
|
|
* people from modifying setuid and setgid binaries.
|
|
|
|
*/
|
|
|
|
return file_remove_suid(file);
|
|
|
|
}
|
|
|
|
|
2011-01-11 00:15:36 +01:00
|
|
|
/*
|
|
|
|
* xfs_file_dio_aio_write - handle direct IO writes
|
|
|
|
*
|
|
|
|
* Lock the inode appropriately to prepare for and issue a direct IO write.
|
2011-01-11 00:22:40 +01:00
|
|
|
* By separating it from the buffered write path we remove all the tricky to
|
2011-01-11 00:15:36 +01:00
|
|
|
* follow locking changes and looping.
|
|
|
|
*
|
2011-01-11 00:22:40 +01:00
|
|
|
* If there are cached pages or we're extending the file, we need IOLOCK_EXCL
|
|
|
|
* until we're sure the bytes at the new EOF have been zeroed and/or the cached
|
|
|
|
* pages are flushed out.
|
|
|
|
*
|
|
|
|
* In most cases the direct IO writes will be done holding IOLOCK_SHARED
|
|
|
|
* allowing them to be done in parallel with reads and other direct IO writes.
|
|
|
|
* However, if the IO is not aligned to filesystem blocks, the direct IO layer
|
|
|
|
* needs to do sub-block zeroing and that requires serialisation against other
|
|
|
|
* direct IOs to the same block. In this case we need to serialise the
|
|
|
|
* submission of the unaligned IOs so that we don't get racing block zeroing in
|
|
|
|
* the dio layer. To avoid the problem with aio, we also need to wait for
|
|
|
|
* outstanding IOs to complete so that unwritten extent conversion is completed
|
|
|
|
* before we try to map the overlapping block. This is currently implemented by
|
2011-08-23 10:28:13 +02:00
|
|
|
* hitting it with a big hammer (i.e. inode_dio_wait()).
|
2011-01-11 00:22:40 +01:00
|
|
|
*
|
2011-01-11 00:15:36 +01:00
|
|
|
* Returns with locks held indicated by @iolock and errors indicated by
|
|
|
|
* negative return values.
|
|
|
|
*/
|
|
|
|
STATIC ssize_t
|
|
|
|
xfs_file_dio_aio_write(
|
|
|
|
struct kiocb *iocb,
|
|
|
|
const struct iovec *iovp,
|
|
|
|
unsigned long nr_segs,
|
|
|
|
loff_t pos,
|
2011-12-18 21:00:14 +01:00
|
|
|
size_t ocount)
|
2011-01-11 00:15:36 +01:00
|
|
|
{
|
|
|
|
struct file *file = iocb->ki_filp;
|
|
|
|
struct address_space *mapping = file->f_mapping;
|
|
|
|
struct inode *inode = mapping->host;
|
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
|
|
|
ssize_t ret = 0;
|
|
|
|
size_t count = ocount;
|
2011-01-11 00:22:40 +01:00
|
|
|
int unaligned_io = 0;
|
2011-12-18 21:00:14 +01:00
|
|
|
int iolock;
|
2011-01-11 00:15:36 +01:00
|
|
|
struct xfs_buftarg *target = XFS_IS_REALTIME_INODE(ip) ?
|
|
|
|
mp->m_rtdev_targp : mp->m_ddev_targp;
|
|
|
|
|
|
|
|
if ((pos & target->bt_smask) || (count & target->bt_smask))
|
|
|
|
return -XFS_ERROR(EINVAL);
|
|
|
|
|
2011-01-11 00:22:40 +01:00
|
|
|
if ((pos & mp->m_blockmask) || ((pos + count) & mp->m_blockmask))
|
|
|
|
unaligned_io = 1;
|
|
|
|
|
xfs: don't serialise adjacent concurrent direct IO appending writes
For append write workloads, extending the file requires a certain
amount of exclusive locking to be done up front to ensure sanity in
things like ensuring that we've zeroed any allocated regions
between the old EOF and the start of the new IO.
For single threads, this typically isn't a problem, and for large
IOs we don't serialise enough for it to be a problem for two
threads on really fast block devices. However for smaller IO and
larger thread counts we have a problem.
Take 4 concurrent sequential, single block sized and aligned IOs.
After the first IO is submitted but before it completes, we end up
with this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
And the IO is done without exclusive locking because offset <=
ip->i_size. When we submit IO 2, we see offset > ip->i_size, and
grab the IO lock exclusive, because there is a chance we need to do
EOF zeroing. However, there is already an IO in progress that avoids
the need for IO zeroing because offset <= ip->i_new_size. hence we
could avoid holding the IO lock exlcusive for this. Hence after
submission of the second IO, we'd end up this state:
IO 1 IO 2 IO 3 IO 4
+-------+-------+-------+-------+
^ ^
| |
| |
| |
| \- ip->i_new_size
\- ip->i_size
There is no need to grab the i_mutex of the IO lock in exclusive
mode if we don't need to invalidate the page cache. Taking these
locks on every direct IO effective serialises them as taking the IO
lock in exclusive mode has to wait for all shared holders to drop
the lock. That only happens when IO is complete, so effective it
prevents dispatch of concurrent direct IO writes to the same inode.
And so you can see that for the third concurrent IO, we'd avoid
exclusive locking for the same reason we avoided the exclusive lock
for the second IO.
Fixing this is a bit more complex than that, because we need to hold
a write-submission local value of ip->i_new_size to that clearing
the value is only done if no other thread has updated it before our
IO completes.....
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
2011-08-25 09:17:02 +02:00
|
|
|
/*
|
|
|
|
* We don't need to take an exclusive lock unless there page cache needs
|
|
|
|
* to be invalidated or unaligned IO is being executed. We don't need to
|
|
|
|
* consider the EOF extension case here because
|
|
|
|
* xfs_file_aio_write_checks() will relock the inode as necessary for
|
|
|
|
* EOF zeroing cases and fill out the new inode size as appropriate.
|
|
|
|
*/
|
|
|
|
if (unaligned_io || mapping->nrpages)
|
2011-12-18 21:00:14 +01:00
|
|
|
iolock = XFS_IOLOCK_EXCL;
|
2011-01-11 00:15:36 +01:00
|
|
|
else
|
2011-12-18 21:00:14 +01:00
|
|
|
iolock = XFS_IOLOCK_SHARED;
|
|
|
|
xfs_rw_ilock(ip, iolock);
|
2011-08-27 16:42:53 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Recheck if there are cached pages that need invalidate after we got
|
|
|
|
* the iolock to protect against other threads adding new pages while
|
|
|
|
* we were waiting for the iolock.
|
|
|
|
*/
|
2011-12-18 21:00:14 +01:00
|
|
|
if (mapping->nrpages && iolock == XFS_IOLOCK_SHARED) {
|
|
|
|
xfs_rw_iunlock(ip, iolock);
|
|
|
|
iolock = XFS_IOLOCK_EXCL;
|
|
|
|
xfs_rw_ilock(ip, iolock);
|
2011-08-27 16:42:53 +02:00
|
|
|
}
|
2011-01-11 00:15:36 +01:00
|
|
|
|
2011-12-18 21:00:14 +01:00
|
|
|
ret = xfs_file_aio_write_checks(file, &pos, &count, &iolock);
|
2011-01-11 00:23:42 +01:00
|
|
|
if (ret)
|
2011-12-18 21:00:14 +01:00
|
|
|
goto out;
|
2011-01-11 00:15:36 +01:00
|
|
|
|
|
|
|
if (mapping->nrpages) {
|
|
|
|
ret = -xfs_flushinval_pages(ip, (pos & PAGE_CACHE_MASK), -1,
|
|
|
|
FI_REMAPF_LOCKED);
|
|
|
|
if (ret)
|
2011-12-18 21:00:14 +01:00
|
|
|
goto out;
|
2011-01-11 00:15:36 +01:00
|
|
|
}
|
|
|
|
|
2011-01-11 00:22:40 +01:00
|
|
|
/*
|
|
|
|
* If we are doing unaligned IO, wait for all other IO to drain,
|
|
|
|
* otherwise demote the lock if we had to flush cached pages
|
|
|
|
*/
|
|
|
|
if (unaligned_io)
|
2011-08-23 10:28:13 +02:00
|
|
|
inode_dio_wait(inode);
|
2011-12-18 21:00:14 +01:00
|
|
|
else if (iolock == XFS_IOLOCK_EXCL) {
|
2011-01-11 00:15:36 +01:00
|
|
|
xfs_rw_ilock_demote(ip, XFS_IOLOCK_EXCL);
|
2011-12-18 21:00:14 +01:00
|
|
|
iolock = XFS_IOLOCK_SHARED;
|
2011-01-11 00:15:36 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
trace_xfs_file_direct_write(ip, count, iocb->ki_pos, 0);
|
|
|
|
ret = generic_file_direct_write(iocb, iovp,
|
|
|
|
&nr_segs, pos, &iocb->ki_pos, count, ocount);
|
|
|
|
|
2011-12-18 21:00:14 +01:00
|
|
|
out:
|
|
|
|
xfs_rw_iunlock(ip, iolock);
|
|
|
|
|
2011-01-11 00:15:36 +01:00
|
|
|
/* No fallback to buffered IO on errors for XFS. */
|
|
|
|
ASSERT(ret < 0 || ret == count);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2010-02-15 10:44:47 +01:00
|
|
|
STATIC ssize_t
|
2011-01-11 00:17:30 +01:00
|
|
|
xfs_file_buffered_aio_write(
|
2010-02-15 10:44:46 +01:00
|
|
|
struct kiocb *iocb,
|
|
|
|
const struct iovec *iovp,
|
2010-02-15 10:44:47 +01:00
|
|
|
unsigned long nr_segs,
|
2011-01-11 00:17:30 +01:00
|
|
|
loff_t pos,
|
2011-12-18 21:00:14 +01:00
|
|
|
size_t ocount)
|
2010-02-15 10:44:46 +01:00
|
|
|
{
|
|
|
|
struct file *file = iocb->ki_filp;
|
|
|
|
struct address_space *mapping = file->f_mapping;
|
|
|
|
struct inode *inode = mapping->host;
|
2010-02-15 10:44:47 +01:00
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
2011-01-11 00:17:30 +01:00
|
|
|
ssize_t ret;
|
|
|
|
int enospc = 0;
|
2011-12-18 21:00:14 +01:00
|
|
|
int iolock = XFS_IOLOCK_EXCL;
|
2011-01-11 00:17:30 +01:00
|
|
|
size_t count = ocount;
|
2010-02-15 10:44:46 +01:00
|
|
|
|
2011-12-18 21:00:14 +01:00
|
|
|
xfs_rw_ilock(ip, iolock);
|
2010-02-15 10:44:46 +01:00
|
|
|
|
2011-12-18 21:00:14 +01:00
|
|
|
ret = xfs_file_aio_write_checks(file, &pos, &count, &iolock);
|
2011-01-11 00:23:42 +01:00
|
|
|
if (ret)
|
2011-12-18 21:00:14 +01:00
|
|
|
goto out;
|
2010-02-15 10:44:46 +01:00
|
|
|
|
|
|
|
/* We can write back this queue in page reclaim */
|
|
|
|
current->backing_dev_info = mapping->backing_dev_info;
|
|
|
|
|
|
|
|
write_retry:
|
2011-01-11 00:17:30 +01:00
|
|
|
trace_xfs_file_buffered_write(ip, count, iocb->ki_pos, 0);
|
|
|
|
ret = generic_file_buffered_write(iocb, iovp, nr_segs,
|
2012-10-08 12:56:04 +02:00
|
|
|
pos, &iocb->ki_pos, count, 0);
|
|
|
|
|
2011-01-11 00:17:30 +01:00
|
|
|
/*
|
2012-10-08 12:56:04 +02:00
|
|
|
* If we just got an ENOSPC, try to write back all dirty inodes to
|
|
|
|
* convert delalloc space to free up some of the excess reserved
|
|
|
|
* metadata space.
|
2011-01-11 00:17:30 +01:00
|
|
|
*/
|
|
|
|
if (ret == -ENOSPC && !enospc) {
|
|
|
|
enospc = 1;
|
2012-10-08 12:56:04 +02:00
|
|
|
xfs_flush_inodes(ip->i_mount);
|
|
|
|
goto write_retry;
|
2010-02-15 10:44:46 +01:00
|
|
|
}
|
2011-12-18 21:00:14 +01:00
|
|
|
|
2010-02-15 10:44:46 +01:00
|
|
|
current->backing_dev_info = NULL;
|
2011-12-18 21:00:14 +01:00
|
|
|
out:
|
|
|
|
xfs_rw_iunlock(ip, iolock);
|
2011-01-11 00:17:30 +01:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
STATIC ssize_t
|
|
|
|
xfs_file_aio_write(
|
|
|
|
struct kiocb *iocb,
|
|
|
|
const struct iovec *iovp,
|
|
|
|
unsigned long nr_segs,
|
|
|
|
loff_t pos)
|
|
|
|
{
|
|
|
|
struct file *file = iocb->ki_filp;
|
|
|
|
struct address_space *mapping = file->f_mapping;
|
|
|
|
struct inode *inode = mapping->host;
|
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
|
|
|
ssize_t ret;
|
|
|
|
size_t ocount = 0;
|
|
|
|
|
|
|
|
XFS_STATS_INC(xs_write_calls);
|
|
|
|
|
|
|
|
BUG_ON(iocb->ki_pos != pos);
|
|
|
|
|
|
|
|
ret = generic_segment_checks(iovp, &nr_segs, &ocount, VERIFY_READ);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
if (ocount == 0)
|
|
|
|
return 0;
|
|
|
|
|
2012-06-12 16:20:39 +02:00
|
|
|
sb_start_write(inode->i_sb);
|
2011-01-11 00:17:30 +01:00
|
|
|
|
2012-06-12 16:20:39 +02:00
|
|
|
if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
|
|
|
|
ret = -EIO;
|
|
|
|
goto out;
|
|
|
|
}
|
2011-01-11 00:17:30 +01:00
|
|
|
|
|
|
|
if (unlikely(file->f_flags & O_DIRECT))
|
2011-12-18 21:00:14 +01:00
|
|
|
ret = xfs_file_dio_aio_write(iocb, iovp, nr_segs, pos, ocount);
|
2011-01-11 00:17:30 +01:00
|
|
|
else
|
|
|
|
ret = xfs_file_buffered_aio_write(iocb, iovp, nr_segs, pos,
|
2011-12-18 21:00:14 +01:00
|
|
|
ocount);
|
2010-02-15 10:44:46 +01:00
|
|
|
|
2011-12-18 21:00:14 +01:00
|
|
|
if (ret > 0) {
|
|
|
|
ssize_t err;
|
2010-02-15 10:44:46 +01:00
|
|
|
|
2011-12-18 21:00:14 +01:00
|
|
|
XFS_STATS_ADD(xs_write_bytes, ret);
|
2010-02-15 10:44:46 +01:00
|
|
|
|
2011-12-18 21:00:14 +01:00
|
|
|
/* Handle various SYNC-type writes */
|
|
|
|
err = generic_write_sync(file, pos, ret);
|
|
|
|
if (err < 0)
|
|
|
|
ret = err;
|
2010-02-15 10:44:46 +01:00
|
|
|
}
|
|
|
|
|
2012-06-12 16:20:39 +02:00
|
|
|
out:
|
|
|
|
sb_end_write(inode->i_sb);
|
2011-01-11 00:13:53 +01:00
|
|
|
return ret;
|
2010-02-15 10:44:46 +01:00
|
|
|
}
|
|
|
|
|
2011-01-14 13:07:43 +01:00
|
|
|
STATIC long
|
|
|
|
xfs_file_fallocate(
|
|
|
|
struct file *file,
|
|
|
|
int mode,
|
|
|
|
loff_t offset,
|
|
|
|
loff_t len)
|
|
|
|
{
|
|
|
|
struct inode *inode = file->f_path.dentry->d_inode;
|
|
|
|
long error;
|
|
|
|
loff_t new_size = 0;
|
|
|
|
xfs_flock64_t bf;
|
|
|
|
xfs_inode_t *ip = XFS_I(inode);
|
|
|
|
int cmd = XFS_IOC_RESVSP;
|
2011-03-25 23:13:08 +01:00
|
|
|
int attr_flags = XFS_ATTR_NOLOCK;
|
2011-01-14 13:07:43 +01:00
|
|
|
|
|
|
|
if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
|
|
|
|
return -EOPNOTSUPP;
|
|
|
|
|
|
|
|
bf.l_whence = 0;
|
|
|
|
bf.l_start = offset;
|
|
|
|
bf.l_len = len;
|
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_IOLOCK_EXCL);
|
|
|
|
|
|
|
|
if (mode & FALLOC_FL_PUNCH_HOLE)
|
|
|
|
cmd = XFS_IOC_UNRESVSP;
|
|
|
|
|
|
|
|
/* check the new inode size is valid before allocating */
|
|
|
|
if (!(mode & FALLOC_FL_KEEP_SIZE) &&
|
|
|
|
offset + len > i_size_read(inode)) {
|
|
|
|
new_size = offset + len;
|
|
|
|
error = inode_newsize_ok(inode, new_size);
|
|
|
|
if (error)
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
2011-03-25 23:13:08 +01:00
|
|
|
if (file->f_flags & O_DSYNC)
|
|
|
|
attr_flags |= XFS_ATTR_SYNC;
|
|
|
|
|
|
|
|
error = -xfs_change_file_space(ip, cmd, &bf, 0, attr_flags);
|
2011-01-14 13:07:43 +01:00
|
|
|
if (error)
|
|
|
|
goto out_unlock;
|
|
|
|
|
|
|
|
/* Change file size if needed */
|
|
|
|
if (new_size) {
|
|
|
|
struct iattr iattr;
|
|
|
|
|
|
|
|
iattr.ia_valid = ATTR_SIZE;
|
|
|
|
iattr.ia_size = new_size;
|
2011-07-08 14:34:23 +02:00
|
|
|
error = -xfs_setattr_size(ip, &iattr, XFS_ATTR_NOLOCK);
|
2011-01-14 13:07:43 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
STATIC int
|
2006-03-14 04:00:35 +01:00
|
|
|
xfs_file_open(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct inode *inode,
|
2008-11-28 04:23:32 +01:00
|
|
|
struct file *file)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2008-11-28 04:23:32 +01:00
|
|
|
if (!(file->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
|
2005-04-17 00:20:36 +02:00
|
|
|
return -EFBIG;
|
2008-11-28 04:23:32 +01:00
|
|
|
if (XFS_FORCED_SHUTDOWN(XFS_M(inode->i_sb)))
|
|
|
|
return -EIO;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
STATIC int
|
|
|
|
xfs_dir_open(
|
|
|
|
struct inode *inode,
|
|
|
|
struct file *file)
|
|
|
|
{
|
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
|
|
|
int mode;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
error = xfs_file_open(inode, file);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If there are any blocks, read-ahead block 0 as we're almost
|
|
|
|
* certain to have the next operation be a read there.
|
|
|
|
*/
|
|
|
|
mode = xfs_ilock_map_shared(ip);
|
|
|
|
if (ip->i_d.di_nextents > 0)
|
|
|
|
xfs_da_reada_buf(NULL, ip, 0, XFS_DATA_FORK);
|
|
|
|
xfs_iunlock(ip, mode);
|
|
|
|
return 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
STATIC int
|
2006-03-14 04:00:35 +01:00
|
|
|
xfs_file_release(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct inode *inode,
|
|
|
|
struct file *filp)
|
|
|
|
{
|
2007-08-29 02:58:01 +02:00
|
|
|
return -xfs_release(XFS_I(inode));
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
STATIC int
|
2006-03-14 04:00:35 +01:00
|
|
|
xfs_file_readdir(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct file *filp,
|
|
|
|
void *dirent,
|
|
|
|
filldir_t filldir)
|
|
|
|
{
|
2007-08-28 05:58:24 +02:00
|
|
|
struct inode *inode = filp->f_path.dentry->d_inode;
|
2007-08-29 02:58:01 +02:00
|
|
|
xfs_inode_t *ip = XFS_I(inode);
|
2007-08-28 05:58:24 +02:00
|
|
|
int error;
|
|
|
|
size_t bufsize;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The Linux API doesn't pass down the total size of the buffer
|
|
|
|
* we read into down to the filesystem. With the filldir concept
|
|
|
|
* it's not needed for correct information, but the XFS dir2 leaf
|
|
|
|
* code wants an estimate of the buffer size to calculate it's
|
|
|
|
* readahead window and size the buffers used for mapping to
|
|
|
|
* physical blocks.
|
|
|
|
*
|
|
|
|
* Try to give it an estimate that's good enough, maybe at some
|
|
|
|
* point we can change the ->readdir prototype to include the
|
2010-02-03 18:50:13 +01:00
|
|
|
* buffer size. For now we use the current glibc buffer size.
|
2007-08-28 05:58:24 +02:00
|
|
|
*/
|
2010-02-03 18:50:13 +01:00
|
|
|
bufsize = (size_t)min_t(loff_t, 32768, ip->i_d.di_size);
|
2007-08-28 05:58:24 +02:00
|
|
|
|
2007-08-29 02:58:01 +02:00
|
|
|
error = xfs_readdir(ip, dirent, bufsize,
|
2007-08-28 05:58:24 +02:00
|
|
|
(xfs_off_t *)&filp->f_pos, filldir);
|
|
|
|
if (error)
|
|
|
|
return -error;
|
|
|
|
return 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
STATIC int
|
2006-03-14 04:00:35 +01:00
|
|
|
xfs_file_mmap(
|
2005-04-17 00:20:36 +02:00
|
|
|
struct file *filp,
|
|
|
|
struct vm_area_struct *vma)
|
|
|
|
{
|
2006-03-14 04:00:35 +01:00
|
|
|
vma->vm_ops = &xfs_file_vm_ops;
|
2005-06-21 06:07:45 +02:00
|
|
|
|
2006-06-09 06:52:13 +02:00
|
|
|
file_accessed(filp);
|
2005-04-17 00:20:36 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2007-07-19 08:28:17 +02:00
|
|
|
/*
|
|
|
|
* mmap()d file has taken write protection fault and is being made
|
|
|
|
* writable. We can set the page state up correctly for a writable
|
|
|
|
* page, which means we can do correct delalloc accounting (ENOSPC
|
|
|
|
* checking!) and unwritten extent mapping.
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_vm_page_mkwrite(
|
|
|
|
struct vm_area_struct *vma,
|
2009-04-01 00:23:21 +02:00
|
|
|
struct vm_fault *vmf)
|
2007-07-19 08:28:17 +02:00
|
|
|
{
|
2009-04-01 00:23:21 +02:00
|
|
|
return block_page_mkwrite(vma, vmf, xfs_get_blocks);
|
2007-07-19 08:28:17 +02:00
|
|
|
}
|
|
|
|
|
2012-08-21 11:11:57 +02:00
|
|
|
/*
|
|
|
|
* This type is designed to indicate the type of offset we would like
|
|
|
|
* to search from page cache for either xfs_seek_data() or xfs_seek_hole().
|
|
|
|
*/
|
|
|
|
enum {
|
|
|
|
HOLE_OFF = 0,
|
|
|
|
DATA_OFF,
|
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lookup the desired type of offset from the given page.
|
|
|
|
*
|
|
|
|
* On success, return true and the offset argument will point to the
|
|
|
|
* start of the region that was found. Otherwise this function will
|
|
|
|
* return false and keep the offset argument unchanged.
|
|
|
|
*/
|
|
|
|
STATIC bool
|
|
|
|
xfs_lookup_buffer_offset(
|
|
|
|
struct page *page,
|
|
|
|
loff_t *offset,
|
|
|
|
unsigned int type)
|
|
|
|
{
|
|
|
|
loff_t lastoff = page_offset(page);
|
|
|
|
bool found = false;
|
|
|
|
struct buffer_head *bh, *head;
|
|
|
|
|
|
|
|
bh = head = page_buffers(page);
|
|
|
|
do {
|
|
|
|
/*
|
|
|
|
* Unwritten extents that have data in the page
|
|
|
|
* cache covering them can be identified by the
|
|
|
|
* BH_Unwritten state flag. Pages with multiple
|
|
|
|
* buffers might have a mix of holes, data and
|
|
|
|
* unwritten extents - any buffer with valid
|
|
|
|
* data in it should have BH_Uptodate flag set
|
|
|
|
* on it.
|
|
|
|
*/
|
|
|
|
if (buffer_unwritten(bh) ||
|
|
|
|
buffer_uptodate(bh)) {
|
|
|
|
if (type == DATA_OFF)
|
|
|
|
found = true;
|
|
|
|
} else {
|
|
|
|
if (type == HOLE_OFF)
|
|
|
|
found = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (found) {
|
|
|
|
*offset = lastoff;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
lastoff += bh->b_size;
|
|
|
|
} while ((bh = bh->b_this_page) != head);
|
|
|
|
|
|
|
|
return found;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This routine is called to find out and return a data or hole offset
|
|
|
|
* from the page cache for unwritten extents according to the desired
|
|
|
|
* type for xfs_seek_data() or xfs_seek_hole().
|
|
|
|
*
|
|
|
|
* The argument offset is used to tell where we start to search from the
|
|
|
|
* page cache. Map is used to figure out the end points of the range to
|
|
|
|
* lookup pages.
|
|
|
|
*
|
|
|
|
* Return true if the desired type of offset was found, and the argument
|
|
|
|
* offset is filled with that address. Otherwise, return false and keep
|
|
|
|
* offset unchanged.
|
|
|
|
*/
|
|
|
|
STATIC bool
|
|
|
|
xfs_find_get_desired_pgoff(
|
|
|
|
struct inode *inode,
|
|
|
|
struct xfs_bmbt_irec *map,
|
|
|
|
unsigned int type,
|
|
|
|
loff_t *offset)
|
|
|
|
{
|
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
|
|
|
struct pagevec pvec;
|
|
|
|
pgoff_t index;
|
|
|
|
pgoff_t end;
|
|
|
|
loff_t endoff;
|
|
|
|
loff_t startoff = *offset;
|
|
|
|
loff_t lastoff = startoff;
|
|
|
|
bool found = false;
|
|
|
|
|
|
|
|
pagevec_init(&pvec, 0);
|
|
|
|
|
|
|
|
index = startoff >> PAGE_CACHE_SHIFT;
|
|
|
|
endoff = XFS_FSB_TO_B(mp, map->br_startoff + map->br_blockcount);
|
|
|
|
end = endoff >> PAGE_CACHE_SHIFT;
|
|
|
|
do {
|
|
|
|
int want;
|
|
|
|
unsigned nr_pages;
|
|
|
|
unsigned int i;
|
|
|
|
|
|
|
|
want = min_t(pgoff_t, end - index, PAGEVEC_SIZE);
|
|
|
|
nr_pages = pagevec_lookup(&pvec, inode->i_mapping, index,
|
|
|
|
want);
|
|
|
|
/*
|
|
|
|
* No page mapped into given range. If we are searching holes
|
|
|
|
* and if this is the first time we got into the loop, it means
|
|
|
|
* that the given offset is landed in a hole, return it.
|
|
|
|
*
|
|
|
|
* If we have already stepped through some block buffers to find
|
|
|
|
* holes but they all contains data. In this case, the last
|
|
|
|
* offset is already updated and pointed to the end of the last
|
|
|
|
* mapped page, if it does not reach the endpoint to search,
|
|
|
|
* that means there should be a hole between them.
|
|
|
|
*/
|
|
|
|
if (nr_pages == 0) {
|
|
|
|
/* Data search found nothing */
|
|
|
|
if (type == DATA_OFF)
|
|
|
|
break;
|
|
|
|
|
|
|
|
ASSERT(type == HOLE_OFF);
|
|
|
|
if (lastoff == startoff || lastoff < endoff) {
|
|
|
|
found = true;
|
|
|
|
*offset = lastoff;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* At lease we found one page. If this is the first time we
|
|
|
|
* step into the loop, and if the first page index offset is
|
|
|
|
* greater than the given search offset, a hole was found.
|
|
|
|
*/
|
|
|
|
if (type == HOLE_OFF && lastoff == startoff &&
|
|
|
|
lastoff < page_offset(pvec.pages[0])) {
|
|
|
|
found = true;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < nr_pages; i++) {
|
|
|
|
struct page *page = pvec.pages[i];
|
|
|
|
loff_t b_offset;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* At this point, the page may be truncated or
|
|
|
|
* invalidated (changing page->mapping to NULL),
|
|
|
|
* or even swizzled back from swapper_space to tmpfs
|
|
|
|
* file mapping. However, page->index will not change
|
|
|
|
* because we have a reference on the page.
|
|
|
|
*
|
|
|
|
* Searching done if the page index is out of range.
|
|
|
|
* If the current offset is not reaches the end of
|
|
|
|
* the specified search range, there should be a hole
|
|
|
|
* between them.
|
|
|
|
*/
|
|
|
|
if (page->index > end) {
|
|
|
|
if (type == HOLE_OFF && lastoff < endoff) {
|
|
|
|
*offset = lastoff;
|
|
|
|
found = true;
|
|
|
|
}
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
lock_page(page);
|
|
|
|
/*
|
|
|
|
* Page truncated or invalidated(page->mapping == NULL).
|
|
|
|
* We can freely skip it and proceed to check the next
|
|
|
|
* page.
|
|
|
|
*/
|
|
|
|
if (unlikely(page->mapping != inode->i_mapping)) {
|
|
|
|
unlock_page(page);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!page_has_buffers(page)) {
|
|
|
|
unlock_page(page);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
found = xfs_lookup_buffer_offset(page, &b_offset, type);
|
|
|
|
if (found) {
|
|
|
|
/*
|
|
|
|
* The found offset may be less than the start
|
|
|
|
* point to search if this is the first time to
|
|
|
|
* come here.
|
|
|
|
*/
|
|
|
|
*offset = max_t(loff_t, startoff, b_offset);
|
|
|
|
unlock_page(page);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We either searching data but nothing was found, or
|
|
|
|
* searching hole but found a data buffer. In either
|
|
|
|
* case, probably the next page contains the desired
|
|
|
|
* things, update the last offset to it so.
|
|
|
|
*/
|
|
|
|
lastoff = page_offset(page) + PAGE_SIZE;
|
|
|
|
unlock_page(page);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The number of returned pages less than our desired, search
|
|
|
|
* done. In this case, nothing was found for searching data,
|
|
|
|
* but we found a hole behind the last offset.
|
|
|
|
*/
|
|
|
|
if (nr_pages < want) {
|
|
|
|
if (type == HOLE_OFF) {
|
|
|
|
*offset = lastoff;
|
|
|
|
found = true;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
index = pvec.pages[i - 1]->index + 1;
|
|
|
|
pagevec_release(&pvec);
|
|
|
|
} while (index <= end);
|
|
|
|
|
|
|
|
out:
|
|
|
|
pagevec_release(&pvec);
|
|
|
|
return found;
|
|
|
|
}
|
|
|
|
|
2012-05-10 15:29:17 +02:00
|
|
|
STATIC loff_t
|
|
|
|
xfs_seek_data(
|
|
|
|
struct file *file,
|
2012-08-21 11:11:45 +02:00
|
|
|
loff_t start)
|
2012-05-10 15:29:17 +02:00
|
|
|
{
|
|
|
|
struct inode *inode = file->f_mapping->host;
|
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
|
|
|
loff_t uninitialized_var(offset);
|
|
|
|
xfs_fsize_t isize;
|
|
|
|
xfs_fileoff_t fsbno;
|
|
|
|
xfs_filblks_t end;
|
|
|
|
uint lock;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
lock = xfs_ilock_map_shared(ip);
|
|
|
|
|
|
|
|
isize = i_size_read(inode);
|
|
|
|
if (start >= isize) {
|
|
|
|
error = ENXIO;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Try to read extents from the first block indicated
|
|
|
|
* by fsbno to the end block of the file.
|
|
|
|
*/
|
2012-08-21 11:12:07 +02:00
|
|
|
fsbno = XFS_B_TO_FSBT(mp, start);
|
2012-05-10 15:29:17 +02:00
|
|
|
end = XFS_B_TO_FSB(mp, isize);
|
2012-08-21 11:12:07 +02:00
|
|
|
for (;;) {
|
|
|
|
struct xfs_bmbt_irec map[2];
|
|
|
|
int nmap = 2;
|
|
|
|
unsigned int i;
|
2012-05-10 15:29:17 +02:00
|
|
|
|
2012-08-21 11:12:07 +02:00
|
|
|
error = xfs_bmapi_read(ip, fsbno, end - fsbno, map, &nmap,
|
|
|
|
XFS_BMAPI_ENTIRE);
|
|
|
|
if (error)
|
|
|
|
goto out_unlock;
|
2012-05-10 15:29:17 +02:00
|
|
|
|
2012-08-21 11:12:07 +02:00
|
|
|
/* No extents at given offset, must be beyond EOF */
|
|
|
|
if (nmap == 0) {
|
|
|
|
error = ENXIO;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < nmap; i++) {
|
|
|
|
offset = max_t(loff_t, start,
|
|
|
|
XFS_FSB_TO_B(mp, map[i].br_startoff));
|
|
|
|
|
|
|
|
/* Landed in a data extent */
|
|
|
|
if (map[i].br_startblock == DELAYSTARTBLOCK ||
|
|
|
|
(map[i].br_state == XFS_EXT_NORM &&
|
|
|
|
!isnullstartblock(map[i].br_startblock)))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Landed in an unwritten extent, try to search data
|
|
|
|
* from page cache.
|
|
|
|
*/
|
|
|
|
if (map[i].br_state == XFS_EXT_UNWRITTEN) {
|
|
|
|
if (xfs_find_get_desired_pgoff(inode, &map[i],
|
|
|
|
DATA_OFF, &offset))
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* map[0] is hole or its an unwritten extent but
|
|
|
|
* without data in page cache. Probably means that
|
|
|
|
* we are reading after EOF if nothing in map[1].
|
|
|
|
*/
|
2012-05-10 15:29:17 +02:00
|
|
|
if (nmap == 1) {
|
|
|
|
error = ENXIO;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
2012-08-21 11:12:07 +02:00
|
|
|
ASSERT(i > 1);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Nothing was found, proceed to the next round of search
|
|
|
|
* if reading offset not beyond or hit EOF.
|
|
|
|
*/
|
|
|
|
fsbno = map[i - 1].br_startoff + map[i - 1].br_blockcount;
|
|
|
|
start = XFS_FSB_TO_B(mp, fsbno);
|
|
|
|
if (start >= isize) {
|
|
|
|
error = ENXIO;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
2012-05-10 15:29:17 +02:00
|
|
|
}
|
|
|
|
|
2012-08-21 11:12:07 +02:00
|
|
|
out:
|
2012-05-10 15:29:17 +02:00
|
|
|
if (offset != file->f_pos)
|
|
|
|
file->f_pos = offset;
|
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
xfs_iunlock_map_shared(ip, lock);
|
|
|
|
|
|
|
|
if (error)
|
|
|
|
return -error;
|
|
|
|
return offset;
|
|
|
|
}
|
|
|
|
|
|
|
|
STATIC loff_t
|
|
|
|
xfs_seek_hole(
|
|
|
|
struct file *file,
|
2012-08-21 11:11:45 +02:00
|
|
|
loff_t start)
|
2012-05-10 15:29:17 +02:00
|
|
|
{
|
|
|
|
struct inode *inode = file->f_mapping->host;
|
|
|
|
struct xfs_inode *ip = XFS_I(inode);
|
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
|
|
|
loff_t uninitialized_var(offset);
|
|
|
|
xfs_fsize_t isize;
|
|
|
|
xfs_fileoff_t fsbno;
|
2012-08-21 11:12:18 +02:00
|
|
|
xfs_filblks_t end;
|
2012-05-10 15:29:17 +02:00
|
|
|
uint lock;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
|
|
|
return -XFS_ERROR(EIO);
|
|
|
|
|
|
|
|
lock = xfs_ilock_map_shared(ip);
|
|
|
|
|
|
|
|
isize = i_size_read(inode);
|
|
|
|
if (start >= isize) {
|
|
|
|
error = ENXIO;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
fsbno = XFS_B_TO_FSBT(mp, start);
|
2012-08-21 11:12:18 +02:00
|
|
|
end = XFS_B_TO_FSB(mp, isize);
|
|
|
|
|
|
|
|
for (;;) {
|
|
|
|
struct xfs_bmbt_irec map[2];
|
|
|
|
int nmap = 2;
|
|
|
|
unsigned int i;
|
|
|
|
|
|
|
|
error = xfs_bmapi_read(ip, fsbno, end - fsbno, map, &nmap,
|
|
|
|
XFS_BMAPI_ENTIRE);
|
|
|
|
if (error)
|
|
|
|
goto out_unlock;
|
|
|
|
|
|
|
|
/* No extents at given offset, must be beyond EOF */
|
|
|
|
if (nmap == 0) {
|
|
|
|
error = ENXIO;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < nmap; i++) {
|
|
|
|
offset = max_t(loff_t, start,
|
|
|
|
XFS_FSB_TO_B(mp, map[i].br_startoff));
|
|
|
|
|
|
|
|
/* Landed in a hole */
|
|
|
|
if (map[i].br_startblock == HOLESTARTBLOCK)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Landed in an unwritten extent, try to search hole
|
|
|
|
* from page cache.
|
|
|
|
*/
|
|
|
|
if (map[i].br_state == XFS_EXT_UNWRITTEN) {
|
|
|
|
if (xfs_find_get_desired_pgoff(inode, &map[i],
|
|
|
|
HOLE_OFF, &offset))
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
}
|
2012-05-10 15:29:17 +02:00
|
|
|
|
|
|
|
/*
|
2012-08-21 11:12:18 +02:00
|
|
|
* map[0] contains data or its unwritten but contains
|
|
|
|
* data in page cache, probably means that we are
|
|
|
|
* reading after EOF. We should fix offset to point
|
|
|
|
* to the end of the file(i.e., there is an implicit
|
|
|
|
* hole at the end of any file).
|
2012-05-10 15:29:17 +02:00
|
|
|
*/
|
2012-08-21 11:12:18 +02:00
|
|
|
if (nmap == 1) {
|
|
|
|
offset = isize;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
ASSERT(i > 1);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Both mappings contains data, proceed to the next round of
|
|
|
|
* search if the current reading offset not beyond or hit EOF.
|
|
|
|
*/
|
|
|
|
fsbno = map[i - 1].br_startoff + map[i - 1].br_blockcount;
|
|
|
|
start = XFS_FSB_TO_B(mp, fsbno);
|
|
|
|
if (start >= isize) {
|
|
|
|
offset = isize;
|
|
|
|
break;
|
|
|
|
}
|
2012-05-10 15:29:17 +02:00
|
|
|
}
|
|
|
|
|
2012-08-21 11:12:18 +02:00
|
|
|
out:
|
|
|
|
/*
|
|
|
|
* At this point, we must have found a hole. However, the returned
|
|
|
|
* offset may be bigger than the file size as it may be aligned to
|
|
|
|
* page boundary for unwritten extents, we need to deal with this
|
|
|
|
* situation in particular.
|
|
|
|
*/
|
|
|
|
offset = min_t(loff_t, offset, isize);
|
2012-05-10 15:29:17 +02:00
|
|
|
if (offset != file->f_pos)
|
|
|
|
file->f_pos = offset;
|
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
xfs_iunlock_map_shared(ip, lock);
|
|
|
|
|
|
|
|
if (error)
|
|
|
|
return -error;
|
|
|
|
return offset;
|
|
|
|
}
|
|
|
|
|
|
|
|
STATIC loff_t
|
|
|
|
xfs_file_llseek(
|
|
|
|
struct file *file,
|
|
|
|
loff_t offset,
|
|
|
|
int origin)
|
|
|
|
{
|
|
|
|
switch (origin) {
|
|
|
|
case SEEK_END:
|
|
|
|
case SEEK_CUR:
|
|
|
|
case SEEK_SET:
|
|
|
|
return generic_file_llseek(file, offset, origin);
|
|
|
|
case SEEK_DATA:
|
2012-08-21 11:11:45 +02:00
|
|
|
return xfs_seek_data(file, offset);
|
2012-05-10 15:29:17 +02:00
|
|
|
case SEEK_HOLE:
|
2012-08-21 11:11:45 +02:00
|
|
|
return xfs_seek_hole(file, offset);
|
2012-05-10 15:29:17 +02:00
|
|
|
default:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2006-03-28 11:56:42 +02:00
|
|
|
const struct file_operations xfs_file_operations = {
|
2012-05-10 15:29:17 +02:00
|
|
|
.llseek = xfs_file_llseek,
|
2005-04-17 00:20:36 +02:00
|
|
|
.read = do_sync_read,
|
2005-09-02 07:43:05 +02:00
|
|
|
.write = do_sync_write,
|
2006-03-14 04:00:35 +01:00
|
|
|
.aio_read = xfs_file_aio_read,
|
|
|
|
.aio_write = xfs_file_aio_write,
|
2006-03-31 05:08:59 +02:00
|
|
|
.splice_read = xfs_file_splice_read,
|
|
|
|
.splice_write = xfs_file_splice_write,
|
2006-03-14 04:00:35 +01:00
|
|
|
.unlocked_ioctl = xfs_file_ioctl,
|
2005-04-17 00:20:36 +02:00
|
|
|
#ifdef CONFIG_COMPAT
|
2006-03-14 04:00:35 +01:00
|
|
|
.compat_ioctl = xfs_file_compat_ioctl,
|
2005-04-17 00:20:36 +02:00
|
|
|
#endif
|
2006-03-14 04:00:35 +01:00
|
|
|
.mmap = xfs_file_mmap,
|
|
|
|
.open = xfs_file_open,
|
|
|
|
.release = xfs_file_release,
|
|
|
|
.fsync = xfs_file_fsync,
|
2011-01-14 13:07:43 +01:00
|
|
|
.fallocate = xfs_file_fallocate,
|
2005-04-17 00:20:36 +02:00
|
|
|
};
|
|
|
|
|
2006-03-28 11:56:42 +02:00
|
|
|
const struct file_operations xfs_dir_file_operations = {
|
2008-11-28 04:23:32 +01:00
|
|
|
.open = xfs_dir_open,
|
2005-04-17 00:20:36 +02:00
|
|
|
.read = generic_read_dir,
|
2006-03-14 04:00:35 +01:00
|
|
|
.readdir = xfs_file_readdir,
|
2008-08-24 13:24:41 +02:00
|
|
|
.llseek = generic_file_llseek,
|
2006-03-14 04:00:35 +01:00
|
|
|
.unlocked_ioctl = xfs_file_ioctl,
|
2005-05-06 15:44:46 +02:00
|
|
|
#ifdef CONFIG_COMPAT
|
2006-03-14 04:00:35 +01:00
|
|
|
.compat_ioctl = xfs_file_compat_ioctl,
|
2005-05-06 15:44:46 +02:00
|
|
|
#endif
|
2011-10-02 16:25:16 +02:00
|
|
|
.fsync = xfs_dir_fsync,
|
2005-04-17 00:20:36 +02:00
|
|
|
};
|
|
|
|
|
2009-09-27 20:29:37 +02:00
|
|
|
static const struct vm_operations_struct xfs_file_vm_ops = {
|
2007-07-19 10:46:59 +02:00
|
|
|
.fault = filemap_fault,
|
2007-07-19 08:28:17 +02:00
|
|
|
.page_mkwrite = xfs_vm_page_mkwrite,
|
2012-10-09 01:28:46 +02:00
|
|
|
.remap_pages = generic_file_remap_pages,
|
2005-06-21 06:07:45 +02:00
|
|
|
};
|