[linux.git] / fs / ext4 / fsync.c

/*
 *  linux/fs/ext4/fsync.c
 *
 *  Copyright (C) 1993  Stephen Tweedie ([email protected])
 *  from
 *  Copyright (C) 1992  Remy Card ([email protected])
 *                      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 ([email protected]), 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	EXT4FS_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.
 */
extern int ext4_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
		 * conversion 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 int ext4_sync_parent(struct inode *inode)
{
	struct writeback_control wbc;
	struct dentry *dentry = NULL;
	struct inode *next;
	int ret = 0;

	if (!ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY))
		return 0;
	inode = igrab(inode);
	while (ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY)) {
		ext4_clear_inode_state(inode, EXT4_STATE_NEWENTRY);
		dentry = NULL;
		spin_lock(&inode->i_lock);
		if (!list_empty(&inode->i_dentry)) {
			dentry = list_first_entry(&inode->i_dentry,
						  struct dentry, d_alias);
			dget(dentry);
		}
		spin_unlock(&inode->i_lock);
		if (!dentry)
			break;
		next = igrab(dentry->d_parent->d_inode);
		dput(dentry);
		if (!next)
			break;
		iput(inode);
		inode = next;
		ret = sync_mapping_buffers(inode->i_mapping);
		if (ret)
			break;
		memset(&wbc, 0, sizeof(wbc));
		wbc.sync_mode = WB_SYNC_ALL;
		wbc.nr_to_write = 0;         /* only write out the inode */
		ret = sync_inode(inode, &wbc);
		if (ret)
			break;
	}
	iput(inode);
	return ret;
}

/**
 * __sync_file - generic_file_fsync without the locking and filemap_write
 * @inode:	inode to sync
 * @datasync:	only sync essential metadata if true
 *
 * This is just generic_file_fsync without the locking.  This is needed for
 * nojournal mode to make sure this inodes data/metadata makes it to disk
 * properly.  The i_mutex should be held already.
 */
static int __sync_inode(struct inode *inode, int datasync)
{
	int err;
	int ret;

	ret = sync_mapping_buffers(inode->i_mapping);
	if (!(inode->i_state & I_DIRTY))
		return ret;
	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
		return ret;

	err = sync_inode_metadata(inode, 1);
	if (ret == 0)
		ret = err;
	return ret;
}

/*
 * 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, loff_t start, loff_t end, 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;
	bool needs_barrier = false;

	J_ASSERT(ext4_journal_current_handle() == NULL);

	trace_ext4_sync_file_enter(file, datasync);

	ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
	if (ret)
		return ret;
	mutex_lock(&inode->i_mutex);

	if (inode->i_sb->s_flags & MS_RDONLY)
		goto out;

	ret = ext4_flush_completed_IO(inode);
	if (ret < 0)
		goto out;

	if (!journal) {
		ret = __sync_inode(inode, datasync);
		if (!ret && !list_empty(&inode->i_dentry))
			ret = ext4_sync_parent(inode);
		goto out;
	}

	/*
	 * 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)) {
		ret = ext4_force_commit(inode->i_sb);
		goto out;
	}

	commit_tid = datasync ? ei->i_datasync_tid : ei->i_sync_tid;
	if (journal->j_flags & JBD2_BARRIER &&
	    !jbd2_trans_will_send_data_barrier(journal, commit_tid))
		needs_barrier = true;
	jbd2_log_start_commit(journal, commit_tid);
	ret = jbd2_log_wait_commit(journal, commit_tid);
	if (needs_barrier)
		blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
 out:
	mutex_unlock(&inode->i_mutex);
	trace_ext4_sync_file_exit(inode, ret);
	return ret;
}
Commit	Line	Data
ac27a0ec	1	/*
617ba13b	2	* linux/fs/ext4/fsync.c
ac27a0ec DK	3	*
	4	* Copyright (C) 1993 Stephen Tweedie ([email protected])
	5	* from
	6	* Copyright (C) 1992 Remy Card ([email protected])
	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 ([email protected]), 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)
4a873a47 TT	38	{
e8bbe8c4	39	#ifdef EXT4FS_DEBUG
4a873a47 TT	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	*/
3889fd57	78	extern int ext4_flush_completed_IO(struct inode *inode)
4a873a47 TT	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
25985edc	104	* conversion finished. The io structure has a flag to
4a873a47 TT	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	*/
0893ed45	128	static int ext4_sync_parent(struct inode *inode)
14ece102	129	{
0893ed45	130	struct writeback_control wbc;
14ece102	131	struct dentry *dentry = NULL;
d59729f4	132	struct inode *next;
0893ed45	133	int ret = 0;
14ece102	134
d59729f4 TT	135	if (!ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY))
	136	return 0;
	137	inode = igrab(inode);
	138	while (ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY)) {
14ece102	139	ext4_clear_inode_state(inode, EXT4_STATE_NEWENTRY);
d59729f4 TT	140	dentry = NULL;
	141	spin_lock(&inode->i_lock);
	142	if (!list_empty(&inode->i_dentry)) {
	143	dentry = list_first_entry(&inode->i_dentry,
	144	struct dentry, d_alias);
	145	dget(dentry);
	146	}
	147	spin_unlock(&inode->i_lock);
	148	if (!dentry)
14ece102	149	break;
d59729f4 TT	150	next = igrab(dentry->d_parent->d_inode);
	151	dput(dentry);
	152	if (!next)
	153	break;
	154	iput(inode);
	155	inode = next;
0893ed45 CW	156	ret = sync_mapping_buffers(inode->i_mapping);
	157	if (ret)
	158	break;
	159	memset(&wbc, 0, sizeof(wbc));
	160	wbc.sync_mode = WB_SYNC_ALL;
	161	wbc.nr_to_write = 0; /* only write out the inode */
	162	ret = sync_inode(inode, &wbc);
	163	if (ret)
	164	break;
14ece102	165	}
d59729f4	166	iput(inode);
0893ed45	167	return ret;
14ece102 FM	168	}
14ece102 FM	169
02c24a82 JB	170	/**
	171	* __sync_file - generic_file_fsync without the locking and filemap_write
	172	* @inode: inode to sync
	173	* @datasync: only sync essential metadata if true
	174	*
	175	* This is just generic_file_fsync without the locking. This is needed for
	176	* nojournal mode to make sure this inodes data/metadata makes it to disk
	177	* properly. The i_mutex should be held already.
	178	*/
	179	static int __sync_inode(struct inode *inode, int datasync)
	180	{
	181	int err;
	182	int ret;
	183
	184	ret = sync_mapping_buffers(inode->i_mapping);
	185	if (!(inode->i_state & I_DIRTY))
	186	return ret;
	187	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
	188	return ret;
	189
	190	err = sync_inode_metadata(inode, 1);
	191	if (ret == 0)
	192	ret = err;
	193	return ret;
	194	}
	195
ac27a0ec	196	/*
617ba13b	197	* akpm: A new design for ext4_sync_file().
ac27a0ec DK	198	*
	199	* This is only called from sys_fsync(), sys_fdatasync() and sys_msync().
	200	* There cannot be a transaction open by this task.
	201	* Another task could have dirtied this inode. Its data can be in any
	202	* state in the journalling system.
	203	*
	204	* What we do is just kick off a commit and wait on it. This will snapshot the
	205	* inode to disk.
8d5d02e6 MC	206	*
8d5d02e6 MC	207	* i_mutex lock is held when entering and exiting this function
ac27a0ec DK	208	*/
ac27a0ec DK	209
02c24a82	210	int ext4_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
ac27a0ec	211	{
7ea80859	212	struct inode *inode = file->f_mapping->host;
b436b9be	213	struct ext4_inode_info *ei = EXT4_I(inode);
d755fb38	214	journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
b436b9be JK	215	int ret;
b436b9be JK	216	tid_t commit_tid;
93628ffb	217	bool needs_barrier = false;
ac27a0ec	218
ac39849d	219	J_ASSERT(ext4_journal_current_handle() == NULL);
ac27a0ec	220
0562e0ba	221	trace_ext4_sync_file_enter(file, datasync);
ede86cc4	222
02c24a82 JB	223	ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
	224	if (ret)
	225	return ret;
	226	mutex_lock(&inode->i_mutex);
	227
b436b9be	228	if (inode->i_sb->s_flags & MS_RDONLY)
02c24a82	229	goto out;
b436b9be	230
3889fd57	231	ret = ext4_flush_completed_IO(inode);
8d5d02e6	232	if (ret < 0)
0562e0ba	233	goto out;
60e6679e	234
14ece102	235	if (!journal) {
02c24a82	236	ret = __sync_inode(inode, datasync);
14ece102	237	if (!ret && !list_empty(&inode->i_dentry))
0893ed45	238	ret = ext4_sync_parent(inode);
0562e0ba	239	goto out;
14ece102	240	}
b436b9be	241
ac27a0ec	242	/*
b436b9be	243	* data=writeback,ordered:
ac27a0ec	244	* The caller's filemap_fdatawrite()/wait will sync the data.
b436b9be JK	245	* Metadata is in the journal, we wait for proper transaction to
b436b9be JK	246	* commit here.
ac27a0ec DK	247	*
	248	* data=journal:
	249	* filemap_fdatawrite won't do anything (the buffers are clean).
617ba13b	250	* ext4_force_commit will write the file data into the journal and
ac27a0ec DK	251	* will wait on that.
	252	* filemap_fdatawait() will encounter a ton of newly-dirtied pages
	253	* (they were dirtied by commit). But that's OK - the blocks are
	254	* safe in-journal, which is all fsync() needs to ensure.
	255	*/
0562e0ba JZ	256	if (ext4_should_journal_data(inode)) {
	257	ret = ext4_force_commit(inode->i_sb);
	258	goto out;
	259	}
ac27a0ec	260
b436b9be	261	commit_tid = datasync ? ei->i_datasync_tid : ei->i_sync_tid;
93628ffb JK	262	if (journal->j_flags & JBD2_BARRIER &&
	263	!jbd2_trans_will_send_data_barrier(journal, commit_tid))
	264	needs_barrier = true;
	265	jbd2_log_start_commit(journal, commit_tid);
	266	ret = jbd2_log_wait_commit(journal, commit_tid);
	267	if (needs_barrier)
dd3932ed	268	blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
0562e0ba	269	out:
02c24a82	270	mutex_unlock(&inode->i_mutex);
0562e0ba	271	trace_ext4_sync_file_exit(inode, ret);
ac27a0ec DK	272	return ret;
ac27a0ec DK	273	}