[deliverable/linux.git] / fs / ext4 / page-io.c

/*
 * linux/fs/ext4/page-io.c
 *
 * This contains the new page_io functions for ext4
 *
 * Written by Theodore Ts'o, 2010.
 */

#include <linux/module.h>
#include <linux/fs.h>
#include <linux/time.h>
#include <linux/jbd2.h>
#include <linux/highuid.h>
#include <linux/pagemap.h>
#include <linux/quotaops.h>
#include <linux/string.h>
#include <linux/buffer_head.h>
#include <linux/writeback.h>
#include <linux/pagevec.h>
#include <linux/mpage.h>
#include <linux/namei.h>
#include <linux/uio.h>
#include <linux/bio.h>
#include <linux/workqueue.h>
#include <linux/kernel.h>
#include <linux/slab.h>

#include "ext4_jbd2.h"
#include "xattr.h"
#include "acl.h"
#include "ext4_extents.h"

static struct kmem_cache *io_page_cachep, *io_end_cachep;

#define WQ_HASH_SZ		37
#define to_ioend_wq(v)	(&ioend_wq[((unsigned long)v) % WQ_HASH_SZ])
static wait_queue_head_t ioend_wq[WQ_HASH_SZ];

int __init ext4_init_pageio(void)
{
	int i;

	io_page_cachep = KMEM_CACHE(ext4_io_page, SLAB_RECLAIM_ACCOUNT);
	if (io_page_cachep == NULL)
		return -ENOMEM;
	io_end_cachep = KMEM_CACHE(ext4_io_end, SLAB_RECLAIM_ACCOUNT);
	if (io_page_cachep == NULL) {
		kmem_cache_destroy(io_page_cachep);
		return -ENOMEM;
	}
	for (i = 0; i < WQ_HASH_SZ; i++)
		init_waitqueue_head(&ioend_wq[i]);

	return 0;
}

void ext4_exit_pageio(void)
{
	kmem_cache_destroy(io_end_cachep);
	kmem_cache_destroy(io_page_cachep);
}

void ext4_ioend_wait(struct inode *inode)
{
	wait_queue_head_t *wq = to_ioend_wq(inode);

	wait_event(*wq, (atomic_read(&EXT4_I(inode)->i_ioend_count) == 0));
}

void ext4_free_io_end(ext4_io_end_t *io)
{
	int i;
	wait_queue_head_t *wq;

	BUG_ON(!io);
	if (io->page)
		put_page(io->page);
	for (i = 0; i < io->num_io_pages; i++) {
		if (--io->pages[i]->p_count == 0) {
			struct page *page = io->pages[i]->p_page;

			end_page_writeback(page);
			put_page(page);
			kmem_cache_free(io_page_cachep, io->pages[i]);
		}
	}
	io->num_io_pages = 0;
	wq = to_ioend_wq(io->inode);
	if (atomic_dec_and_test(&EXT4_I(io->inode)->i_ioend_count) &&
	    waitqueue_active(wq))
		wake_up_all(wq);
	kmem_cache_free(io_end_cachep, io);
}

/*
 * check a range of space and convert unwritten extents to written.
 */
int ext4_end_io_nolock(ext4_io_end_t *io)
{
	struct inode *inode = io->inode;
	loff_t offset = io->offset;
	ssize_t size = io->size;
	int ret = 0;

	ext4_debug("ext4_end_io_nolock: io 0x%p from inode %lu,list->next 0x%p,"
		   "list->prev 0x%p\n",
		   io, inode->i_ino, io->list.next, io->list.prev);

	if (list_empty(&io->list))
		return ret;

	if (!(io->flag & EXT4_IO_END_UNWRITTEN))
		return ret;

	ret = ext4_convert_unwritten_extents(inode, offset, size);
	if (ret < 0) {
		printk(KERN_EMERG "%s: failed to convert unwritten "
			"extents to written extents, error is %d "
			"io is still on inode %lu aio dio list\n",
		       __func__, ret, inode->i_ino);
		return ret;
	}

	if (io->iocb)
		aio_complete(io->iocb, io->result, 0);
	/* clear the DIO AIO unwritten flag */
	io->flag &= ~EXT4_IO_END_UNWRITTEN;
	return ret;
}

/*
 * work on completed aio dio IO, to convert unwritten extents to extents
 */
static void ext4_end_io_work(struct work_struct *work)
{
	ext4_io_end_t		*io = container_of(work, ext4_io_end_t, work);
	struct inode		*inode = io->inode;
	struct ext4_inode_info	*ei = EXT4_I(inode);
	unsigned long		flags;
	int			ret;

	mutex_lock(&inode->i_mutex);
	ret = ext4_end_io_nolock(io);
	if (ret < 0) {
		mutex_unlock(&inode->i_mutex);
		return;
	}

	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	if (!list_empty(&io->list))
		list_del_init(&io->list);
	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
	mutex_unlock(&inode->i_mutex);
	ext4_free_io_end(io);
}

ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags)
{
	ext4_io_end_t *io = NULL;

	io = kmem_cache_alloc(io_end_cachep, flags);
	if (io) {
		memset(io, 0, sizeof(*io));
		atomic_inc(&EXT4_I(inode)->i_ioend_count);
		io->inode = inode;
		INIT_WORK(&io->work, ext4_end_io_work);
		INIT_LIST_HEAD(&io->list);
	}
	return io;
}

/*
 * Print an buffer I/O error compatible with the fs/buffer.c.  This
 * provides compatibility with dmesg scrapers that look for a specific
 * buffer I/O error message.  We really need a unified error reporting
 * structure to userspace ala Digital Unix's uerf system, but it's
 * probably not going to happen in my lifetime, due to LKML politics...
 */
static void buffer_io_error(struct buffer_head *bh)
{
	char b[BDEVNAME_SIZE];
	printk(KERN_ERR "Buffer I/O error on device %s, logical block %llu\n",
			bdevname(bh->b_bdev, b),
			(unsigned long long)bh->b_blocknr);
}

static void ext4_end_bio(struct bio *bio, int error)
{
	ext4_io_end_t *io_end = bio->bi_private;
	struct workqueue_struct *wq;
	struct inode *inode;
	unsigned long flags;
	int i;

	BUG_ON(!io_end);
	bio->bi_private = NULL;
	bio->bi_end_io = NULL;
	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
		error = 0;
	bio_put(bio);

	for (i = 0; i < io_end->num_io_pages; i++) {
		struct page *page = io_end->pages[i]->p_page;
		struct buffer_head *bh, *head;
		int partial_write = 0;

		head = page_buffers(page);
		if (error)
			SetPageError(page);
		BUG_ON(!head);
		if (head->b_size == PAGE_CACHE_SIZE)
			clear_buffer_dirty(head);
		else {
			loff_t offset;
			loff_t io_end_offset = io_end->offset + io_end->size;

			offset = (sector_t) page->index << PAGE_CACHE_SHIFT;
			bh = head;
			do {
				if ((offset >= io_end->offset) &&
				    (offset+bh->b_size <= io_end_offset)) {
					if (error)
						buffer_io_error(bh);

					clear_buffer_dirty(bh);
				}
				if (buffer_delay(bh))
					partial_write = 1;
				else if (!buffer_mapped(bh))
					clear_buffer_dirty(bh);
				else if (buffer_dirty(bh))
					partial_write = 1;
				offset += bh->b_size;
				bh = bh->b_this_page;
			} while (bh != head);
		}

		if (--io_end->pages[i]->p_count == 0) {
			struct page *page = io_end->pages[i]->p_page;

			end_page_writeback(page);
			put_page(page);
			kmem_cache_free(io_page_cachep, io_end->pages[i]);
		}

		/*
		 * If this is a partial write which happened to make
		 * all buffers uptodate then we can optimize away a
		 * bogus readpage() for the next read(). Here we
		 * 'discover' whether the page went uptodate as a
		 * result of this (potentially partial) write.
		 */
		if (!partial_write)
			SetPageUptodate(page);
	}
	io_end->num_io_pages = 0;
	inode = io_end->inode;

	if (error) {
		io_end->flag |= EXT4_IO_END_ERROR;
		ext4_warning(inode->i_sb, "I/O error writing to inode %lu "
			     "(offset %llu size %ld starting block %llu)",
			     inode->i_ino,
			     (unsigned long long) io_end->offset,
			     (long) io_end->size,
			     (unsigned long long)
			     bio->bi_sector >> (inode->i_blkbits - 9));
	}

	/* Add the io_end to per-inode completed io list*/
	spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
	list_add_tail(&io_end->list, &EXT4_I(inode)->i_completed_io_list);
	spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);

	wq = EXT4_SB(inode->i_sb)->dio_unwritten_wq;
	/* queue the work to convert unwritten extents to written */
	queue_work(wq, &io_end->work);
}

void ext4_io_submit(struct ext4_io_submit *io)
{
	struct bio *bio = io->io_bio;

	if (bio) {
		bio_get(io->io_bio);
		submit_bio(io->io_op, io->io_bio);
		BUG_ON(bio_flagged(io->io_bio, BIO_EOPNOTSUPP));
		bio_put(io->io_bio);
	}
	io->io_bio = 0;
	io->io_op = 0;
	io->io_end = 0;
}

static int io_submit_init(struct ext4_io_submit *io,
			  struct inode *inode,
			  struct writeback_control *wbc,
			  struct buffer_head *bh)
{
	ext4_io_end_t *io_end;
	struct page *page = bh->b_page;
	int nvecs = bio_get_nr_vecs(bh->b_bdev);
	struct bio *bio;

	io_end = ext4_init_io_end(inode, GFP_NOFS);
	if (!io_end)
		return -ENOMEM;
	do {
		bio = bio_alloc(GFP_NOIO, nvecs);
		nvecs >>= 1;
	} while (bio == NULL);

	bio->bi_sector = bh->b_blocknr * (bh->b_size >> 9);
	bio->bi_bdev = bh->b_bdev;
	bio->bi_private = io->io_end = io_end;
	bio->bi_end_io = ext4_end_bio;

	io_end->offset = (page->index << PAGE_CACHE_SHIFT) + bh_offset(bh);

	io->io_bio = bio;
	io->io_op = (wbc->sync_mode == WB_SYNC_ALL ?
			WRITE_SYNC_PLUG : WRITE);
	io->io_next_block = bh->b_blocknr;
	return 0;
}

static int io_submit_add_bh(struct ext4_io_submit *io,
			    struct ext4_io_page *io_page,
			    struct inode *inode,
			    struct writeback_control *wbc,
			    struct buffer_head *bh)
{
	ext4_io_end_t *io_end;
	int ret;

	if (buffer_new(bh)) {
		clear_buffer_new(bh);
		unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
	}

	if (!buffer_mapped(bh) || buffer_delay(bh)) {
		if (!buffer_mapped(bh))
			clear_buffer_dirty(bh);
		if (io->io_bio)
			ext4_io_submit(io);
		return 0;
	}

	if (io->io_bio && bh->b_blocknr != io->io_next_block) {
submit_and_retry:
		ext4_io_submit(io);
	}
	if (io->io_bio == NULL) {
		ret = io_submit_init(io, inode, wbc, bh);
		if (ret)
			return ret;
	}
	io_end = io->io_end;
	if ((io_end->num_io_pages >= MAX_IO_PAGES) &&
	    (io_end->pages[io_end->num_io_pages-1] != io_page))
		goto submit_and_retry;
	if (buffer_uninit(bh))
		io->io_end->flag |= EXT4_IO_END_UNWRITTEN;
	io->io_end->size += bh->b_size;
	io->io_next_block++;
	ret = bio_add_page(io->io_bio, bh->b_page, bh->b_size, bh_offset(bh));
	if (ret != bh->b_size)
		goto submit_and_retry;
	if ((io_end->num_io_pages == 0) ||
	    (io_end->pages[io_end->num_io_pages-1] != io_page)) {
		io_end->pages[io_end->num_io_pages++] = io_page;
		io_page->p_count++;
	}
	return 0;
}

int ext4_bio_write_page(struct ext4_io_submit *io,
			struct page *page,
			int len,
			struct writeback_control *wbc)
{
	struct inode *inode = page->mapping->host;
	unsigned block_start, block_end, blocksize;
	struct ext4_io_page *io_page;
	struct buffer_head *bh, *head;
	int ret = 0;

	blocksize = 1 << inode->i_blkbits;

	BUG_ON(PageWriteback(page));
	set_page_writeback(page);
	ClearPageError(page);

	io_page = kmem_cache_alloc(io_page_cachep, GFP_NOFS);
	if (!io_page) {
		set_page_dirty(page);
		unlock_page(page);
		return -ENOMEM;
	}
	io_page->p_page = page;
	io_page->p_count = 0;
	get_page(page);

	for (bh = head = page_buffers(page), block_start = 0;
	     bh != head || !block_start;
	     block_start = block_end, bh = bh->b_this_page) {
		block_end = block_start + blocksize;
		if (block_start >= len) {
			clear_buffer_dirty(bh);
			set_buffer_uptodate(bh);
			continue;
		}
		ret = io_submit_add_bh(io, io_page, inode, wbc, bh);
		if (ret) {
			/*
			 * We only get here on ENOMEM.  Not much else
			 * we can do but mark the page as dirty, and
			 * better luck next time.
			 */
			set_page_dirty(page);
			break;
		}
	}
	unlock_page(page);
	/*
	 * If the page was truncated before we could do the writeback,
	 * or we had a memory allocation error while trying to write
	 * the first buffer head, we won't have submitted any pages for
	 * I/O.  In that case we need to make sure we've cleared the
	 * PageWriteback bit from the page to prevent the system from
	 * wedging later on.
	 */
	if (io_page->p_count == 0) {
		put_page(page);
		end_page_writeback(page);
		kmem_cache_free(io_page_cachep, io_page);
	}
	return ret;
}
Commit	Line	Data
bd2d0210 TT	1	/*
	2	* linux/fs/ext4/page-io.c
	3	*
	4	* This contains the new page_io functions for ext4
	5	*
	6	* Written by Theodore Ts'o, 2010.
	7	*/
	8
	9	#include <linux/module.h>
	10	#include <linux/fs.h>
	11	#include <linux/time.h>
	12	#include <linux/jbd2.h>
	13	#include <linux/highuid.h>
	14	#include <linux/pagemap.h>
	15	#include <linux/quotaops.h>
	16	#include <linux/string.h>
	17	#include <linux/buffer_head.h>
	18	#include <linux/writeback.h>
	19	#include <linux/pagevec.h>
	20	#include <linux/mpage.h>
	21	#include <linux/namei.h>
	22	#include <linux/uio.h>
	23	#include <linux/bio.h>
	24	#include <linux/workqueue.h>
	25	#include <linux/kernel.h>
	26	#include <linux/slab.h>
	27
	28	#include "ext4_jbd2.h"
	29	#include "xattr.h"
	30	#include "acl.h"
	31	#include "ext4_extents.h"
	32
	33	static struct kmem_cache io_page_cachep, io_end_cachep;
	34
f7ad6d2e TT	35	#define WQ_HASH_SZ 37
	36	#define to_ioend_wq(v) (&ioend_wq[((unsigned long)v) % WQ_HASH_SZ])
	37	static wait_queue_head_t ioend_wq[WQ_HASH_SZ];
	38
5dabfc78	39	int __init ext4_init_pageio(void)
bd2d0210	40	{
f7ad6d2e TT	41	int i;
f7ad6d2e TT	42
bd2d0210 TT	43	io_page_cachep = KMEM_CACHE(ext4_io_page, SLAB_RECLAIM_ACCOUNT);
	44	if (io_page_cachep == NULL)
	45	return -ENOMEM;
	46	io_end_cachep = KMEM_CACHE(ext4_io_end, SLAB_RECLAIM_ACCOUNT);
	47	if (io_page_cachep == NULL) {
	48	kmem_cache_destroy(io_page_cachep);
	49	return -ENOMEM;
	50	}
f7ad6d2e TT	51	for (i = 0; i < WQ_HASH_SZ; i++)
f7ad6d2e TT	52	init_waitqueue_head(&ioend_wq[i]);
bd2d0210 TT	53
	54	return 0;
	55	}
	56
5dabfc78	57	void ext4_exit_pageio(void)
bd2d0210 TT	58	{
	59	kmem_cache_destroy(io_end_cachep);
	60	kmem_cache_destroy(io_page_cachep);
	61	}
	62
f7ad6d2e TT	63	void ext4_ioend_wait(struct inode *inode)
	64	{
	65	wait_queue_head_t *wq = to_ioend_wq(inode);
	66
	67	wait_event(*wq, (atomic_read(&EXT4_I(inode)->i_ioend_count) == 0));
	68	}
	69
bd2d0210 TT	70	void ext4_free_io_end(ext4_io_end_t *io)
	71	{
	72	int i;
f7ad6d2e	73	wait_queue_head_t *wq;
bd2d0210 TT	74
	75	BUG_ON(!io);
	76	if (io->page)
	77	put_page(io->page);
	78	for (i = 0; i < io->num_io_pages; i++) {
	79	if (--io->pages[i]->p_count == 0) {
	80	struct page *page = io->pages[i]->p_page;
	81
	82	end_page_writeback(page);
	83	put_page(page);
	84	kmem_cache_free(io_page_cachep, io->pages[i]);
	85	}
	86	}
	87	io->num_io_pages = 0;
f7ad6d2e TT	88	wq = to_ioend_wq(io->inode);
	89	if (atomic_dec_and_test(&EXT4_I(io->inode)->i_ioend_count) &&
	90	waitqueue_active(wq))
	91	wake_up_all(wq);
bd2d0210 TT	92	kmem_cache_free(io_end_cachep, io);
	93	}
	94
	95	/*
	96	* check a range of space and convert unwritten extents to written.
	97	*/
	98	int ext4_end_io_nolock(ext4_io_end_t *io)
	99	{
	100	struct inode *inode = io->inode;
	101	loff_t offset = io->offset;
	102	ssize_t size = io->size;
	103	int ret = 0;
	104
	105	ext4_debug("ext4_end_io_nolock: io 0x%p from inode %lu,list->next 0x%p,"
	106	"list->prev 0x%p\n",
	107	io, inode->i_ino, io->list.next, io->list.prev);
	108
	109	if (list_empty(&io->list))
	110	return ret;
	111
	112	if (!(io->flag & EXT4_IO_END_UNWRITTEN))
	113	return ret;
	114
	115	ret = ext4_convert_unwritten_extents(inode, offset, size);
	116	if (ret < 0) {
	117	printk(KERN_EMERG "%s: failed to convert unwritten "
	118	"extents to written extents, error is %d "
	119	"io is still on inode %lu aio dio list\n",
	120	__func__, ret, inode->i_ino);
	121	return ret;
	122	}
	123
	124	if (io->iocb)
	125	aio_complete(io->iocb, io->result, 0);
	126	/* clear the DIO AIO unwritten flag */
	127	io->flag &= ~EXT4_IO_END_UNWRITTEN;
	128	return ret;
	129	}
	130
	131	/*
	132	* work on completed aio dio IO, to convert unwritten extents to extents
	133	*/
	134	static void ext4_end_io_work(struct work_struct *work)
	135	{
	136	ext4_io_end_t *io = container_of(work, ext4_io_end_t, work);
	137	struct inode *inode = io->inode;
	138	struct ext4_inode_info *ei = EXT4_I(inode);
	139	unsigned long flags;
	140	int ret;
	141
	142	mutex_lock(&inode->i_mutex);
	143	ret = ext4_end_io_nolock(io);
	144	if (ret < 0) {
	145	mutex_unlock(&inode->i_mutex);
	146	return;
	147	}
	148
	149	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	150	if (!list_empty(&io->list))
	151	list_del_init(&io->list);
	152	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
	153	mutex_unlock(&inode->i_mutex);
	154	ext4_free_io_end(io);
	155	}
156
157	ext4_io_end_t ext4_init_io_end(struct inode inode, gfp_t flags)
158	{
159	ext4_io_end_t *io = NULL;
160
161	io = kmem_cache_alloc(io_end_cachep, flags);
162	if (io) {
163	memset(io, 0, sizeof(*io));
f7ad6d2e TT	164	atomic_inc(&EXT4_I(inode)->i_ioend_count);
f7ad6d2e TT	165	io->inode = inode;
bd2d0210 TT	166	INIT_WORK(&io->work, ext4_end_io_work);
	167	INIT_LIST_HEAD(&io->list);
	168	}
	169	return io;
	170	}
	171
	172	/*
	173	* Print an buffer I/O error compatible with the fs/buffer.c. This
	174	* provides compatibility with dmesg scrapers that look for a specific
	175	* buffer I/O error message. We really need a unified error reporting
	176	* structure to userspace ala Digital Unix's uerf system, but it's
	177	* probably not going to happen in my lifetime, due to LKML politics...
	178	*/
	179	static void buffer_io_error(struct buffer_head *bh)
	180	{
	181	char b[BDEVNAME_SIZE];
	182	printk(KERN_ERR "Buffer I/O error on device %s, logical block %llu\n",
	183	bdevname(bh->b_bdev, b),
	184	(unsigned long long)bh->b_blocknr);
	185	}
	186
	187	static void ext4_end_bio(struct bio *bio, int error)
	188	{
	189	ext4_io_end_t *io_end = bio->bi_private;
	190	struct workqueue_struct *wq;
	191	struct inode *inode;
	192	unsigned long flags;
bd2d0210 TT	193	int i;
	194
	195	BUG_ON(!io_end);
bd2d0210 TT	196	bio->bi_private = NULL;
	197	bio->bi_end_io = NULL;
	198	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
	199	error = 0;
bd2d0210 TT	200	bio_put(bio);
bd2d0210 TT	201
bd2d0210 TT	202	for (i = 0; i < io_end->num_io_pages; i++) {
	203	struct page *page = io_end->pages[i]->p_page;
	204	struct buffer_head bh, head;
	205	int partial_write = 0;
	206
	207	head = page_buffers(page);
	208	if (error)
	209	SetPageError(page);
	210	BUG_ON(!head);
	211	if (head->b_size == PAGE_CACHE_SIZE)
	212	clear_buffer_dirty(head);
	213	else {
	214	loff_t offset;
	215	loff_t io_end_offset = io_end->offset + io_end->size;
	216
	217	offset = (sector_t) page->index << PAGE_CACHE_SHIFT;
	218	bh = head;
	219	do {
	220	if ((offset >= io_end->offset) &&
	221	(offset+bh->b_size <= io_end_offset)) {
	222	if (error)
	223	buffer_io_error(bh);
	224
	225	clear_buffer_dirty(bh);
	226	}
	227	if (buffer_delay(bh))
	228	partial_write = 1;
	229	else if (!buffer_mapped(bh))
	230	clear_buffer_dirty(bh);
	231	else if (buffer_dirty(bh))
	232	partial_write = 1;
	233	offset += bh->b_size;
	234	bh = bh->b_this_page;
	235	} while (bh != head);
	236	}
	237
	238	if (--io_end->pages[i]->p_count == 0) {
	239	struct page *page = io_end->pages[i]->p_page;
	240
	241	end_page_writeback(page);
	242	put_page(page);
	243	kmem_cache_free(io_page_cachep, io_end->pages[i]);
	244	}
	245
	246	/*
	247	* If this is a partial write which happened to make
	248	* all buffers uptodate then we can optimize away a
	249	* bogus readpage() for the next read(). Here we
	250	* 'discover' whether the page went uptodate as a
	251	* result of this (potentially partial) write.
	252	*/
	253	if (!partial_write)
	254	SetPageUptodate(page);
	255	}
bd2d0210	256	io_end->num_io_pages = 0;
f7ad6d2e TT	257	inode = io_end->inode;
	258
	259	if (error) {
	260	io_end->flag \|= EXT4_IO_END_ERROR;
	261	ext4_warning(inode->i_sb, "I/O error writing to inode %lu "
	262	"(offset %llu size %ld starting block %llu)",
	263	inode->i_ino,
	264	(unsigned long long) io_end->offset,
	265	(long) io_end->size,
	266	(unsigned long long)
	267	bio->bi_sector >> (inode->i_blkbits - 9));
	268	}
bd2d0210 TT	269
	270	/* Add the io_end to per-inode completed io list*/
	271	spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
	272	list_add_tail(&io_end->list, &EXT4_I(inode)->i_completed_io_list);
	273	spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
	274
	275	wq = EXT4_SB(inode->i_sb)->dio_unwritten_wq;
	276	/* queue the work to convert unwritten extents to written */
	277	queue_work(wq, &io_end->work);
	278	}
	279
	280	void ext4_io_submit(struct ext4_io_submit *io)
	281	{
	282	struct bio *bio = io->io_bio;
	283
	284	if (bio) {
	285	bio_get(io->io_bio);
	286	submit_bio(io->io_op, io->io_bio);
	287	BUG_ON(bio_flagged(io->io_bio, BIO_EOPNOTSUPP));
	288	bio_put(io->io_bio);
	289	}
	290	io->io_bio = 0;
	291	io->io_op = 0;
	292	io->io_end = 0;
	293	}
	294
	295	static int io_submit_init(struct ext4_io_submit *io,
	296	struct inode *inode,
	297	struct writeback_control *wbc,
	298	struct buffer_head *bh)
	299	{
	300	ext4_io_end_t *io_end;
	301	struct page *page = bh->b_page;
	302	int nvecs = bio_get_nr_vecs(bh->b_bdev);
	303	struct bio *bio;
	304
	305	io_end = ext4_init_io_end(inode, GFP_NOFS);
	306	if (!io_end)
	307	return -ENOMEM;
	308	do {
	309	bio = bio_alloc(GFP_NOIO, nvecs);
	310	nvecs >>= 1;
	311	} while (bio == NULL);
	312
	313	bio->bi_sector = bh->b_blocknr * (bh->b_size >> 9);
	314	bio->bi_bdev = bh->b_bdev;
	315	bio->bi_private = io->io_end = io_end;
	316	bio->bi_end_io = ext4_end_bio;
	317
bd2d0210 TT	318	io_end->offset = (page->index << PAGE_CACHE_SHIFT) + bh_offset(bh);
	319
	320	io->io_bio = bio;
	321	io->io_op = (wbc->sync_mode == WB_SYNC_ALL ?
	322	WRITE_SYNC_PLUG : WRITE);
	323	io->io_next_block = bh->b_blocknr;
	324	return 0;
	325	}
	326
	327	static int io_submit_add_bh(struct ext4_io_submit *io,
	328	struct ext4_io_page *io_page,
	329	struct inode *inode,
	330	struct writeback_control *wbc,
	331	struct buffer_head *bh)
	332	{
	333	ext4_io_end_t *io_end;
	334	int ret;
	335
	336	if (buffer_new(bh)) {
	337	clear_buffer_new(bh);
	338	unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
	339	}
	340
	341	if (!buffer_mapped(bh) \|\| buffer_delay(bh)) {
	342	if (!buffer_mapped(bh))
	343	clear_buffer_dirty(bh);
	344	if (io->io_bio)
	345	ext4_io_submit(io);
	346	return 0;
	347	}
	348
	349	if (io->io_bio && bh->b_blocknr != io->io_next_block) {
	350	submit_and_retry:
	351	ext4_io_submit(io);
	352	}
	353	if (io->io_bio == NULL) {
	354	ret = io_submit_init(io, inode, wbc, bh);
	355	if (ret)
	356	return ret;
	357	}
	358	io_end = io->io_end;
	359	if ((io_end->num_io_pages >= MAX_IO_PAGES) &&
	360	(io_end->pages[io_end->num_io_pages-1] != io_page))
	361	goto submit_and_retry;
	362	if (buffer_uninit(bh))
	363	io->io_end->flag \|= EXT4_IO_END_UNWRITTEN;
	364	io->io_end->size += bh->b_size;
	365	io->io_next_block++;
	366	ret = bio_add_page(io->io_bio, bh->b_page, bh->b_size, bh_offset(bh));
	367	if (ret != bh->b_size)
	368	goto submit_and_retry;
	369	if ((io_end->num_io_pages == 0) \|\|
	370	(io_end->pages[io_end->num_io_pages-1] != io_page)) {
	371	io_end->pages[io_end->num_io_pages++] = io_page;
	372	io_page->p_count++;
	373	}
	374	return 0;
	375	}
	376
	377	int ext4_bio_write_page(struct ext4_io_submit *io,
	378	struct page *page,
	379	int len,
	380	struct writeback_control *wbc)
	381	{
382	struct inode *inode = page->mapping->host;
383	unsigned block_start, block_end, blocksize;
384	struct ext4_io_page *io_page;
385	struct buffer_head bh, head;
386	int ret = 0;
387
388	blocksize = 1 << inode->i_blkbits;
389
390	BUG_ON(PageWriteback(page));
391	set_page_writeback(page);
392	ClearPageError(page);
393
394	io_page = kmem_cache_alloc(io_page_cachep, GFP_NOFS);
395	if (!io_page) {
396	set_page_dirty(page);
397	unlock_page(page);
398	return -ENOMEM;
399	}
400	io_page->p_page = page;
401	io_page->p_count = 0;
402	get_page(page);
403
404	for (bh = head = page_buffers(page), block_start = 0;
405	bh != head \|\| !block_start;
406	block_start = block_end, bh = bh->b_this_page) {
407	block_end = block_start + blocksize;
408	if (block_start >= len) {
409	clear_buffer_dirty(bh);
410	set_buffer_uptodate(bh);
411	continue;
412	}
413	ret = io_submit_add_bh(io, io_page, inode, wbc, bh);
414	if (ret) {
415	/*
416	* We only get here on ENOMEM. Not much else
417	* we can do but mark the page as dirty, and
418	* better luck next time.
419	*/
420	set_page_dirty(page);
421	break;
422	}
423	}
424	unlock_page(page);
425	/*
426	* If the page was truncated before we could do the writeback,
427	* or we had a memory allocation error while trying to write
428	* the first buffer head, we won't have submitted any pages for
429	* I/O. In that case we need to make sure we've cleared the
430	* PageWriteback bit from the page to prevent the system from
431	* wedging later on.
432	*/
433	if (io_page->p_count == 0) {
434	put_page(page);
435	end_page_writeback(page);
436	kmem_cache_free(io_page_cachep, io_page);
437	}
438	return ret;
439	}