[net-next-2.6.git] / fs / ext4 / fsync.c

/*
 *  linux/fs/ext4/fsync.c
 *
 *  Copyright (C) 1993  Stephen Tweedie (sct@redhat.com)
 *  from
 *  Copyright (C) 1992  Remy Card (card@masi.ibp.fr)
 *                      Laboratoire MASI - Institut Blaise Pascal
 *                      Universite Pierre et Marie Curie (Paris VI)
 *  from
 *  linux/fs/minix/truncate.c   Copyright (C) 1991, 1992  Linus Torvalds
 *
 *  ext4fs fsync primitive
 *
 *  Big-endian to little-endian byte-swapping/bitmaps by
 *        David S. Miller (davem@caip.rutgers.edu), 1995
 *
 *  Removed unnecessary code duplication for little endian machines
 *  and excessive __inline__s.
 *        Andi Kleen, 1997
 *
 * Major simplications and cleanup - we only need to do the metadata, because
 * we can depend on generic_block_fdatasync() to sync the data blocks.
 */

#include <linux/time.h>
#include <linux/fs.h>
#include <linux/sched.h>
#include <linux/writeback.h>
#include <linux/jbd2.h>
#include <linux/blkdev.h>

#include "ext4.h"
#include "ext4_jbd2.h"

#include <trace/events/ext4.h>

static void dump_completed_IO(struct inode * inode)
{
#ifdef	EXT4_DEBUG
	struct list_head *cur, *before, *after;
	ext4_io_end_t *io, *io0, *io1;
	unsigned long flags;

	if (list_empty(&EXT4_I(inode)->i_completed_io_list)){
		ext4_debug("inode %lu completed_io list is empty\n", inode->i_ino);
		return;
	}

	ext4_debug("Dump inode %lu completed_io list \n", inode->i_ino);
	spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
	list_for_each_entry(io, &EXT4_I(inode)->i_completed_io_list, list){
		cur = &io->list;
		before = cur->prev;
		io0 = container_of(before, ext4_io_end_t, list);
		after = cur->next;
		io1 = container_of(after, ext4_io_end_t, list);

		ext4_debug("io 0x%p from inode %lu,prev 0x%p,next 0x%p\n",
			    io, inode->i_ino, io0, io1);
	}
	spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
#endif
}

/*
 * This function is called from ext4_sync_file().
 *
 * When IO is completed, the work to convert unwritten extents to
 * written is queued on workqueue but may not get immediately
 * scheduled. When fsync is called, we need to ensure the
 * conversion is complete before fsync returns.
 * The inode keeps track of a list of pending/completed IO that
 * might needs to do the conversion. This function walks through
 * the list and convert the related unwritten extents for completed IO
 * to written.
 * The function return the number of pending IOs on success.
 */
static int flush_completed_IO(struct inode *inode)
{
	ext4_io_end_t *io;
	struct ext4_inode_info *ei = EXT4_I(inode);
	unsigned long flags;
	int ret = 0;
	int ret2 = 0;

	if (list_empty(&ei->i_completed_io_list))
		return ret;

	dump_completed_IO(inode);
	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	while (!list_empty(&ei->i_completed_io_list)){
		io = list_entry(ei->i_completed_io_list.next,
				ext4_io_end_t, list);
		/*
		 * Calling ext4_end_io_nolock() to convert completed
		 * IO to written.
		 *
		 * When ext4_sync_file() is called, run_queue() may already
		 * about to flush the work corresponding to this io structure.
		 * It will be upset if it founds the io structure related
		 * to the work-to-be schedule is freed.
		 *
		 * Thus we need to keep the io structure still valid here after
		 * convertion finished. The io structure has a flag to
		 * avoid double converting from both fsync and background work
		 * queue work.
		 */
		spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
		ret = ext4_end_io_nolock(io);
		spin_lock_irqsave(&ei->i_completed_io_lock, flags);
		if (ret < 0)
			ret2 = ret;
		else
			list_del_init(&io->list);
	}
	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
	return (ret2 < 0) ? ret2 : 0;
}

/*
 * If we're not journaling and this is a just-created file, we have to
 * sync our parent directory (if it was freshly created) since
 * otherwise it will only be written by writeback, leaving a huge
 * window during which a crash may lose the file.  This may apply for
 * the parent directory's parent as well, and so on recursively, if
 * they are also freshly created.
 */
static void ext4_sync_parent(struct inode *inode)
{
	struct dentry *dentry = NULL;

	while (inode && ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY)) {
		ext4_clear_inode_state(inode, EXT4_STATE_NEWENTRY);
		dentry = list_entry(inode->i_dentry.next,
				    struct dentry, d_alias);
		if (!dentry || !dentry->d_parent || !dentry->d_parent->d_inode)
			break;
		inode = dentry->d_parent->d_inode;
		sync_mapping_buffers(inode->i_mapping);
	}
}

/*
 * akpm: A new design for ext4_sync_file().
 *
 * This is only called from sys_fsync(), sys_fdatasync() and sys_msync().
 * There cannot be a transaction open by this task.
 * Another task could have dirtied this inode.  Its data can be in any
 * state in the journalling system.
 *
 * What we do is just kick off a commit and wait on it.  This will snapshot the
 * inode to disk.
 *
 * i_mutex lock is held when entering and exiting this function
 */

int ext4_sync_file(struct file *file, int datasync)
{
	struct inode *inode = file->f_mapping->host;
	struct ext4_inode_info *ei = EXT4_I(inode);
	journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
	int ret;
	tid_t commit_tid;

	J_ASSERT(ext4_journal_current_handle() == NULL);

	trace_ext4_sync_file(file, datasync);

	if (inode->i_sb->s_flags & MS_RDONLY)
		return 0;

	ret = flush_completed_IO(inode);
	if (ret < 0)
		return ret;

	if (!journal) {
		ret = generic_file_fsync(file, datasync);
		if (!ret && !list_empty(&inode->i_dentry))
			ext4_sync_parent(inode);
		return ret;
	}

	/*
	 * data=writeback,ordered:
	 *  The caller's filemap_fdatawrite()/wait will sync the data.
	 *  Metadata is in the journal, we wait for proper transaction to
	 *  commit here.
	 *
	 * data=journal:
	 *  filemap_fdatawrite won't do anything (the buffers are clean).
	 *  ext4_force_commit will write the file data into the journal and
	 *  will wait on that.
	 *  filemap_fdatawait() will encounter a ton of newly-dirtied pages
	 *  (they were dirtied by commit).  But that's OK - the blocks are
	 *  safe in-journal, which is all fsync() needs to ensure.
	 */
	if (ext4_should_journal_data(inode))
		return ext4_force_commit(inode->i_sb);

	commit_tid = datasync ? ei->i_datasync_tid : ei->i_sync_tid;
	if (jbd2_log_start_commit(journal, commit_tid)) {
		/*
		 * When the journal is on a different device than the
		 * fs data disk, we need to issue the barrier in
		 * writeback mode.  (In ordered mode, the jbd2 layer
		 * will take care of issuing the barrier.  In
		 * data=journal, all of the data blocks are written to
		 * the journal device.)
		 */
		if (ext4_should_writeback_data(inode) &&
		    (journal->j_fs_dev != journal->j_dev) &&
		    (journal->j_flags & JBD2_BARRIER))
			blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL,
					NULL);
		ret = jbd2_log_wait_commit(journal, commit_tid);
	} else if (journal->j_flags & JBD2_BARRIER)
		blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
	return ret;
}
Commit	Line	Data
ac27a0ec	1	/*
617ba13b	2	* linux/fs/ext4/fsync.c
ac27a0ec DK	3	*
	4	* Copyright (C) 1993 Stephen Tweedie (sct@redhat.com)
	5	* from
	6	* Copyright (C) 1992 Remy Card (card@masi.ibp.fr)
	7	* Laboratoire MASI - Institut Blaise Pascal
	8	* Universite Pierre et Marie Curie (Paris VI)
	9	* from
	10	* linux/fs/minix/truncate.c Copyright (C) 1991, 1992 Linus Torvalds
	11	*
617ba13b	12	* ext4fs fsync primitive
ac27a0ec DK	13	*
	14	* Big-endian to little-endian byte-swapping/bitmaps by
	15	* David S. Miller (davem@caip.rutgers.edu), 1995
	16	*
	17	* Removed unnecessary code duplication for little endian machines
	18	* and excessive __inline__s.
	19	* Andi Kleen, 1997
	20	*
	21	* Major simplications and cleanup - we only need to do the metadata, because
	22	* we can depend on generic_block_fdatasync() to sync the data blocks.
	23	*/
	24
	25	#include <linux/time.h>
	26	#include <linux/fs.h>
	27	#include <linux/sched.h>
	28	#include <linux/writeback.h>
dab291af	29	#include <linux/jbd2.h>
d755fb38	30	#include <linux/blkdev.h>
9bffad1e	31
3dcf5451 CH	32	#include "ext4.h"
3dcf5451 CH	33	#include "ext4_jbd2.h"
ac27a0ec	34
9bffad1e TT	35	#include <trace/events/ext4.h>
9bffad1e TT	36
4a873a47 TT	37	static void dump_completed_IO(struct inode * inode)
	38	{
	39	#ifdef EXT4_DEBUG
	40	struct list_head cur, before, *after;
	41	ext4_io_end_t io, io0, *io1;
	42	unsigned long flags;
	43
	44	if (list_empty(&EXT4_I(inode)->i_completed_io_list)){
	45	ext4_debug("inode %lu completed_io list is empty\n", inode->i_ino);
	46	return;
	47	}
	48
	49	ext4_debug("Dump inode %lu completed_io list \n", inode->i_ino);
	50	spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
	51	list_for_each_entry(io, &EXT4_I(inode)->i_completed_io_list, list){
	52	cur = &io->list;
	53	before = cur->prev;
	54	io0 = container_of(before, ext4_io_end_t, list);
	55	after = cur->next;
	56	io1 = container_of(after, ext4_io_end_t, list);
	57
	58	ext4_debug("io 0x%p from inode %lu,prev 0x%p,next 0x%p\n",
	59	io, inode->i_ino, io0, io1);
	60	}
	61	spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
	62	#endif
	63	}
	64
	65	/*
	66	* This function is called from ext4_sync_file().
	67	*
	68	* When IO is completed, the work to convert unwritten extents to
	69	* written is queued on workqueue but may not get immediately
	70	* scheduled. When fsync is called, we need to ensure the
	71	* conversion is complete before fsync returns.
	72	* The inode keeps track of a list of pending/completed IO that
	73	* might needs to do the conversion. This function walks through
	74	* the list and convert the related unwritten extents for completed IO
	75	* to written.
	76	* The function return the number of pending IOs on success.
	77	*/
	78	static int flush_completed_IO(struct inode *inode)
	79	{
	80	ext4_io_end_t *io;
	81	struct ext4_inode_info *ei = EXT4_I(inode);
	82	unsigned long flags;
	83	int ret = 0;
	84	int ret2 = 0;
	85
	86	if (list_empty(&ei->i_completed_io_list))
	87	return ret;
	88
	89	dump_completed_IO(inode);
	90	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	91	while (!list_empty(&ei->i_completed_io_list)){
	92	io = list_entry(ei->i_completed_io_list.next,
	93	ext4_io_end_t, list);
	94	/*
	95	* Calling ext4_end_io_nolock() to convert completed
	96	* IO to written.
	97	*
	98	* When ext4_sync_file() is called, run_queue() may already
	99	* about to flush the work corresponding to this io structure.
	100	* It will be upset if it founds the io structure related
101	* to the work-to-be schedule is freed.
102	*
103	* Thus we need to keep the io structure still valid here after
104	* convertion finished. The io structure has a flag to
105	* avoid double converting from both fsync and background work
106	* queue work.
107	*/
108	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
109	ret = ext4_end_io_nolock(io);
110	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
111	if (ret < 0)
112	ret2 = ret;
113	else
114	list_del_init(&io->list);
115	}
116	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
117	return (ret2 < 0) ? ret2 : 0;
118	}
119
14ece102 FM	120	/*
	121	* If we're not journaling and this is a just-created file, we have to
	122	* sync our parent directory (if it was freshly created) since
	123	* otherwise it will only be written by writeback, leaving a huge
	124	* window during which a crash may lose the file. This may apply for
	125	* the parent directory's parent as well, and so on recursively, if
	126	* they are also freshly created.
	127	*/
	128	static void ext4_sync_parent(struct inode *inode)
	129	{
	130	struct dentry *dentry = NULL;
	131
	132	while (inode && ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY)) {
	133	ext4_clear_inode_state(inode, EXT4_STATE_NEWENTRY);
	134	dentry = list_entry(inode->i_dentry.next,
	135	struct dentry, d_alias);
	136	if (!dentry \|\| !dentry->d_parent \|\| !dentry->d_parent->d_inode)
	137	break;
	138	inode = dentry->d_parent->d_inode;
	139	sync_mapping_buffers(inode->i_mapping);
	140	}
	141	}
	142
ac27a0ec	143	/*
617ba13b	144	* akpm: A new design for ext4_sync_file().
ac27a0ec DK	145	*
	146	* This is only called from sys_fsync(), sys_fdatasync() and sys_msync().
	147	* There cannot be a transaction open by this task.
	148	* Another task could have dirtied this inode. Its data can be in any
	149	* state in the journalling system.
	150	*
	151	* What we do is just kick off a commit and wait on it. This will snapshot the
	152	* inode to disk.
8d5d02e6 MC	153	*
8d5d02e6 MC	154	* i_mutex lock is held when entering and exiting this function
ac27a0ec DK	155	*/
ac27a0ec DK	156
7ea80859	157	int ext4_sync_file(struct file *file, int datasync)
ac27a0ec	158	{
7ea80859	159	struct inode *inode = file->f_mapping->host;
b436b9be	160	struct ext4_inode_info *ei = EXT4_I(inode);
d755fb38	161	journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
b436b9be JK	162	int ret;
b436b9be JK	163	tid_t commit_tid;
ac27a0ec	164
ac39849d	165	J_ASSERT(ext4_journal_current_handle() == NULL);
ac27a0ec	166
7ea80859	167	trace_ext4_sync_file(file, datasync);
ede86cc4	168
b436b9be JK	169	if (inode->i_sb->s_flags & MS_RDONLY)
	170	return 0;
	171
c7064ef1	172	ret = flush_completed_IO(inode);
8d5d02e6	173	if (ret < 0)
6b17d902	174	return ret;
60e6679e	175
14ece102	176	if (!journal) {
1b061d92	177	ret = generic_file_fsync(file, datasync);
14ece102 FM	178	if (!ret && !list_empty(&inode->i_dentry))
	179	ext4_sync_parent(inode);
	180	return ret;
	181	}
b436b9be	182
ac27a0ec	183	/*
b436b9be	184	* data=writeback,ordered:
ac27a0ec	185	* The caller's filemap_fdatawrite()/wait will sync the data.
b436b9be JK	186	* Metadata is in the journal, we wait for proper transaction to
b436b9be JK	187	* commit here.
ac27a0ec DK	188	*
	189	* data=journal:
	190	* filemap_fdatawrite won't do anything (the buffers are clean).
617ba13b	191	* ext4_force_commit will write the file data into the journal and
ac27a0ec DK	192	* will wait on that.
	193	* filemap_fdatawait() will encounter a ton of newly-dirtied pages
	194	* (they were dirtied by commit). But that's OK - the blocks are
	195	* safe in-journal, which is all fsync() needs to ensure.
	196	*/
6b17d902 TT	197	if (ext4_should_journal_data(inode))
6b17d902 TT	198	return ext4_force_commit(inode->i_sb);
ac27a0ec	199
b436b9be	200	commit_tid = datasync ? ei->i_datasync_tid : ei->i_sync_tid;
cc3e1bea TT	201	if (jbd2_log_start_commit(journal, commit_tid)) {
	202	/*
	203	* When the journal is on a different device than the
	204	* fs data disk, we need to issue the barrier in
	205	* writeback mode. (In ordered mode, the jbd2 layer
	206	* will take care of issuing the barrier. In
	207	* data=journal, all of the data blocks are written to
	208	* the journal device.)
	209	*/
	210	if (ext4_should_writeback_data(inode) &&
	211	(journal->j_fs_dev != journal->j_dev) &&
	212	(journal->j_flags & JBD2_BARRIER))
fbd9b09a	213	blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL,
dd3932ed	214	NULL);
0671e704	215	ret = jbd2_log_wait_commit(journal, commit_tid);
cc3e1bea	216	} else if (journal->j_flags & JBD2_BARRIER)
dd3932ed	217	blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
ac27a0ec DK	218	return ret;
ac27a0ec DK	219	}