[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/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;

int __init ext4_init_pageio(void)
{
	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_end_cachep == NULL) {
		kmem_cache_destroy(io_page_cachep);
		return -ENOMEM;
	}
	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 = ext4_ioend_wq(inode);

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

static void put_io_page(struct ext4_io_page *io_page)
{
	if (atomic_dec_and_test(&io_page->p_count)) {
		put_page(io_page->p_page);
		kmem_cache_free(io_page_cachep, io_page);
	}
}

void ext4_free_io_end(ext4_io_end_t *io)
{
	int i;

	BUG_ON(!io);
	if (io->page)
		put_page(io->page);
	for (i = 0; i < io->num_io_pages; i++)
		put_io_page(io->pages[i]);
	io->num_io_pages = 0;
	if (atomic_dec_and_test(&EXT4_I(io->inode)->i_ioend_count))
		wake_up_all(ext4_ioend_wq(io->inode));
	kmem_cache_free(io_end_cachep, io);
}

/*
 * check a range of space and convert unwritten extents to written.
 *
 * Called with inode->i_mutex; we depend on this when we manipulate
 * io->flag, since we could otherwise race with ext4_flush_completed_IO()
 */
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);

	ret = ext4_convert_unwritten_extents(inode, offset, size);
	if (ret < 0) {
		ext4_msg(inode->i_sb, KERN_EMERG,
			 "failed to convert unwritten extents to written "
			 "extents -- potential data loss!  "
			 "(inode %lu, offset %llu, size %zd, error %d)",
			 inode->i_ino, offset, size, ret);
	}

	if (io->iocb)
		aio_complete(io->iocb, io->result, 0);

	if (io->flag & EXT4_IO_END_DIRECT)
		inode_dio_done(inode);
	/* Wake up anyone waiting on unwritten extent conversion */
	if (atomic_dec_and_test(&EXT4_I(inode)->i_aiodio_unwritten))
		wake_up_all(ext4_ioend_wq(io->inode));
	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;

	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	if (io->flag & EXT4_IO_END_IN_FSYNC)
		goto requeue;
	if (list_empty(&io->list)) {
		spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
		goto free;
	}

	if (!mutex_trylock(&inode->i_mutex)) {
		bool was_queued;
requeue:
		was_queued = !!(io->flag & EXT4_IO_END_QUEUED);
		io->flag |= EXT4_IO_END_QUEUED;
		spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
		/*
		 * Requeue the work instead of waiting so that the work
		 * items queued after this can be processed.
		 */
		queue_work(EXT4_SB(inode->i_sb)->dio_unwritten_wq, &io->work);
		/*
		 * To prevent the ext4-dio-unwritten thread from keeping
		 * requeueing end_io requests and occupying cpu for too long,
		 * yield the cpu if it sees an end_io request that has already
		 * been requeued.
		 */
		if (was_queued)
			yield();
		return;
	}
	list_del_init(&io->list);
	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
	(void) ext4_end_io_nolock(io);
	mutex_unlock(&inode->i_mutex);
free:
	ext4_free_io_end(io);
}

ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags)
{
	ext4_io_end_t *io = kmem_cache_zalloc(io_end_cachep, flags);
	if (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;
	sector_t bi_sector = bio->bi_sector;

	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;
		loff_t offset;
		loff_t io_end_offset;

		if (error) {
			SetPageError(page);
			set_bit(AS_EIO, &page->mapping->flags);
			head = page_buffers(page);
			BUG_ON(!head);

			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))
					buffer_io_error(bh);

				offset += bh->b_size;
				bh = bh->b_this_page;
			} while (bh != head);
		}

		if (atomic_read(&io_end->pages[i]->p_count) == 1)
			end_page_writeback(io_end->pages[i]->p_page);
	}
	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)
			     bi_sector >> (inode->i_blkbits - 9));
	}

	if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
		ext4_free_io_end(io_end);
		return;
	}

	/* 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 = NULL;
	io->io_op = 0;
	io->io_end = NULL;
}

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;
	bio = bio_alloc(GFP_NOIO, min(nvecs, BIO_MAX_PAGES));
	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 : 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))
		ext4_set_io_unwritten_flag(inode, io_end);
	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;
		atomic_inc(&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(!PageLocked(page));
	BUG_ON(PageWriteback(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;
	atomic_set(&io_page->p_count, 1);
	get_page(page);
	set_page_writeback(page);
	ClearPageError(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) {
			/*
			 * Comments copied from block_write_full_page_endio:
			 *
			 * The page straddles i_size.  It must be zeroed out on
			 * each and every writepage invocation because it may
			 * be mmapped.  "A file is mapped in multiples of the
			 * page size.  For a file that is not a multiple of
			 * the  page size, the remaining memory is zeroed when
			 * mapped, and writes to that region are not written
			 * out to the file."
			 */
			zero_user_segment(page, block_start, block_end);
			clear_buffer_dirty(bh);
			set_buffer_uptodate(bh);
			continue;
		}
		clear_buffer_dirty(bh);
		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 (atomic_read(&io_page->p_count) == 1)
		end_page_writeback(page);
	put_io_page(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
bd2d0210 TT	9	#include <linux/fs.h>
	10	#include <linux/time.h>
	11	#include <linux/jbd2.h>
	12	#include <linux/highuid.h>
	13	#include <linux/pagemap.h>
	14	#include <linux/quotaops.h>
	15	#include <linux/string.h>
	16	#include <linux/buffer_head.h>
	17	#include <linux/writeback.h>
	18	#include <linux/pagevec.h>
	19	#include <linux/mpage.h>
	20	#include <linux/namei.h>
	21	#include <linux/uio.h>
	22	#include <linux/bio.h>
	23	#include <linux/workqueue.h>
	24	#include <linux/kernel.h>
	25	#include <linux/slab.h>
	26
	27	#include "ext4_jbd2.h"
	28	#include "xattr.h"
	29	#include "acl.h"
	30	#include "ext4_extents.h"
	31
	32	static struct kmem_cache io_page_cachep, io_end_cachep;
	33
5dabfc78	34	int __init ext4_init_pageio(void)
bd2d0210 TT	35	{
	36	io_page_cachep = KMEM_CACHE(ext4_io_page, SLAB_RECLAIM_ACCOUNT);
	37	if (io_page_cachep == NULL)
	38	return -ENOMEM;
	39	io_end_cachep = KMEM_CACHE(ext4_io_end, SLAB_RECLAIM_ACCOUNT);
13195184	40	if (io_end_cachep == NULL) {
bd2d0210 TT	41	kmem_cache_destroy(io_page_cachep);
	42	return -ENOMEM;
	43	}
bd2d0210 TT	44	return 0;
	45	}
	46
5dabfc78	47	void ext4_exit_pageio(void)
bd2d0210 TT	48	{
	49	kmem_cache_destroy(io_end_cachep);
	50	kmem_cache_destroy(io_page_cachep);
	51	}
	52
f7ad6d2e TT	53	void ext4_ioend_wait(struct inode *inode)
f7ad6d2e TT	54	{
e9e3bcec	55	wait_queue_head_t *wq = ext4_ioend_wq(inode);
f7ad6d2e TT	56
	57	wait_event(*wq, (atomic_read(&EXT4_I(inode)->i_ioend_count) == 0));
	58	}
	59
83668e71 TT	60	static void put_io_page(struct ext4_io_page *io_page)
	61	{
	62	if (atomic_dec_and_test(&io_page->p_count)) {
83668e71 TT	63	put_page(io_page->p_page);
	64	kmem_cache_free(io_page_cachep, io_page);
	65	}
	66	}
	67
bd2d0210 TT	68	void ext4_free_io_end(ext4_io_end_t *io)
	69	{
	70	int i;
	71
	72	BUG_ON(!io);
	73	if (io->page)
	74	put_page(io->page);
83668e71 TT	75	for (i = 0; i < io->num_io_pages; i++)
83668e71 TT	76	put_io_page(io->pages[i]);
bd2d0210	77	io->num_io_pages = 0;
4e298021 TT	78	if (atomic_dec_and_test(&EXT4_I(io->inode)->i_ioend_count))
4e298021 TT	79	wake_up_all(ext4_ioend_wq(io->inode));
bd2d0210 TT	80	kmem_cache_free(io_end_cachep, io);
	81	}
	82
	83	/*
	84	* check a range of space and convert unwritten extents to written.
d73d5046 TM	85	*
	86	* Called with inode->i_mutex; we depend on this when we manipulate
	87	* io->flag, since we could otherwise race with ext4_flush_completed_IO()
bd2d0210 TT	88	*/
	89	int ext4_end_io_nolock(ext4_io_end_t *io)
	90	{
	91	struct inode *inode = io->inode;
	92	loff_t offset = io->offset;
	93	ssize_t size = io->size;
	94	int ret = 0;
	95
	96	ext4_debug("ext4_end_io_nolock: io 0x%p from inode %lu,list->next 0x%p,"
	97	"list->prev 0x%p\n",
	98	io, inode->i_ino, io->list.next, io->list.prev);
	99
bd2d0210 TT	100	ret = ext4_convert_unwritten_extents(inode, offset, size);
bd2d0210 TT	101	if (ret < 0) {
b82e384c TT	102	ext4_msg(inode->i_sb, KERN_EMERG,
	103	"failed to convert unwritten extents to written "
	104	"extents -- potential data loss! "
	105	"(inode %lu, offset %llu, size %zd, error %d)",
	106	inode->i_ino, offset, size, ret);
bd2d0210 TT	107	}
	108
	109	if (io->iocb)
	110	aio_complete(io->iocb, io->result, 0);
e9e3bcec	111
266991b1 JM	112	if (io->flag & EXT4_IO_END_DIRECT)
266991b1 JM	113	inode_dio_done(inode);
b82e384c TT	114	/* Wake up anyone waiting on unwritten extent conversion */
	115	if (atomic_dec_and_test(&EXT4_I(inode)->i_aiodio_unwritten))
	116	wake_up_all(ext4_ioend_wq(io->inode));
bd2d0210 TT	117	return ret;
	118	}
	119
	120	/*
	121	* work on completed aio dio IO, to convert unwritten extents to extents
	122	*/
	123	static void ext4_end_io_work(struct work_struct *work)
	124	{
	125	ext4_io_end_t *io = container_of(work, ext4_io_end_t, work);
	126	struct inode *inode = io->inode;
	127	struct ext4_inode_info *ei = EXT4_I(inode);
	128	unsigned long flags;
bd2d0210	129
d73d5046	130	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
491caa43 JM	131	if (io->flag & EXT4_IO_END_IN_FSYNC)
491caa43 JM	132	goto requeue;
d73d5046 TM	133	if (list_empty(&io->list)) {
	134	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
	135	goto free;
	136	}
d73d5046	137
8c0bec21	138	if (!mutex_trylock(&inode->i_mutex)) {
491caa43 JM	139	bool was_queued;
	140	requeue:
	141	was_queued = !!(io->flag & EXT4_IO_END_QUEUED);
	142	io->flag \|= EXT4_IO_END_QUEUED;
b82e384c	143	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
8c0bec21 JZ	144	/*
	145	* Requeue the work instead of waiting so that the work
	146	* items queued after this can be processed.
	147	*/
	148	queue_work(EXT4_SB(inode->i_sb)->dio_unwritten_wq, &io->work);
	149	/*
	150	* To prevent the ext4-dio-unwritten thread from keeping
	151	* requeueing end_io requests and occupying cpu for too long,
	152	* yield the cpu if it sees an end_io request that has already
	153	* been requeued.
	154	*/
491caa43	155	if (was_queued)
8c0bec21	156	yield();
8c0bec21 JZ	157	return;
8c0bec21 JZ	158	}
b82e384c	159	list_del_init(&io->list);
bd2d0210	160	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
b82e384c	161	(void) ext4_end_io_nolock(io);
bd2d0210	162	mutex_unlock(&inode->i_mutex);
d73d5046	163	free:
bd2d0210 TT	164	ext4_free_io_end(io);
	165	}
	166
	167	ext4_io_end_t ext4_init_io_end(struct inode inode, gfp_t flags)
	168	{
b17b35ec	169	ext4_io_end_t *io = kmem_cache_zalloc(io_end_cachep, flags);
bd2d0210	170	if (io) {
f7ad6d2e TT	171	atomic_inc(&EXT4_I(inode)->i_ioend_count);
f7ad6d2e TT	172	io->inode = inode;
bd2d0210 TT	173	INIT_WORK(&io->work, ext4_end_io_work);
	174	INIT_LIST_HEAD(&io->list);
	175	}
	176	return io;
	177	}
	178
	179	/*
	180	* Print an buffer I/O error compatible with the fs/buffer.c. This
	181	* provides compatibility with dmesg scrapers that look for a specific
	182	* buffer I/O error message. We really need a unified error reporting
	183	* structure to userspace ala Digital Unix's uerf system, but it's
	184	* probably not going to happen in my lifetime, due to LKML politics...
	185	*/
	186	static void buffer_io_error(struct buffer_head *bh)
	187	{
	188	char b[BDEVNAME_SIZE];
	189	printk(KERN_ERR "Buffer I/O error on device %s, logical block %llu\n",
	190	bdevname(bh->b_bdev, b),
	191	(unsigned long long)bh->b_blocknr);
	192	}
	193
	194	static void ext4_end_bio(struct bio *bio, int error)
	195	{
	196	ext4_io_end_t *io_end = bio->bi_private;
	197	struct workqueue_struct *wq;
	198	struct inode *inode;
	199	unsigned long flags;
bd2d0210	200	int i;
d50bdd5a	201	sector_t bi_sector = bio->bi_sector;
bd2d0210 TT	202
bd2d0210 TT	203	BUG_ON(!io_end);
bd2d0210 TT	204	bio->bi_private = NULL;
	205	bio->bi_end_io = NULL;
	206	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
	207	error = 0;
bd2d0210 TT	208	bio_put(bio);
bd2d0210 TT	209
bd2d0210 TT	210	for (i = 0; i < io_end->num_io_pages; i++) {
	211	struct page *page = io_end->pages[i]->p_page;
	212	struct buffer_head bh, head;
39db00f1 CW	213	loff_t offset;
39db00f1 CW	214	loff_t io_end_offset;
bd2d0210	215
39db00f1	216	if (error) {
bd2d0210	217	SetPageError(page);
39db00f1 CW	218	set_bit(AS_EIO, &page->mapping->flags);
	219	head = page_buffers(page);
	220	BUG_ON(!head);
	221
	222	io_end_offset = io_end->offset + io_end->size;
bd2d0210 TT	223
	224	offset = (sector_t) page->index << PAGE_CACHE_SHIFT;
	225	bh = head;
	226	do {
	227	if ((offset >= io_end->offset) &&
39db00f1 CW	228	(offset+bh->b_size <= io_end_offset))
	229	buffer_io_error(bh);
	230
bd2d0210 TT	231	offset += bh->b_size;
	232	bh = bh->b_this_page;
	233	} while (bh != head);
	234	}
	235
b43d17f3 CW	236	if (atomic_read(&io_end->pages[i]->p_count) == 1)
b43d17f3 CW	237	end_page_writeback(io_end->pages[i]->p_page);
bd2d0210	238	}
f7ad6d2e TT	239	inode = io_end->inode;
	240
	241	if (error) {
	242	io_end->flag \|= EXT4_IO_END_ERROR;
	243	ext4_warning(inode->i_sb, "I/O error writing to inode %lu "
	244	"(offset %llu size %ld starting block %llu)",
	245	inode->i_ino,
	246	(unsigned long long) io_end->offset,
	247	(long) io_end->size,
	248	(unsigned long long)
d50bdd5a	249	bi_sector >> (inode->i_blkbits - 9));
f7ad6d2e	250	}
bd2d0210	251
b6168443 TT	252	if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
	253	ext4_free_io_end(io_end);
	254	return;
	255	}
	256
bd2d0210 TT	257	/* Add the io_end to per-inode completed io list*/
	258	spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
	259	list_add_tail(&io_end->list, &EXT4_I(inode)->i_completed_io_list);
	260	spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
	261
	262	wq = EXT4_SB(inode->i_sb)->dio_unwritten_wq;
	263	/* queue the work to convert unwritten extents to written */
	264	queue_work(wq, &io_end->work);
	265	}
	266
	267	void ext4_io_submit(struct ext4_io_submit *io)
	268	{
	269	struct bio *bio = io->io_bio;
	270
	271	if (bio) {
	272	bio_get(io->io_bio);
	273	submit_bio(io->io_op, io->io_bio);
	274	BUG_ON(bio_flagged(io->io_bio, BIO_EOPNOTSUPP));
	275	bio_put(io->io_bio);
	276	}
7dc57615	277	io->io_bio = NULL;
bd2d0210	278	io->io_op = 0;
7dc57615	279	io->io_end = NULL;
bd2d0210 TT	280	}
	281
	282	static int io_submit_init(struct ext4_io_submit *io,
	283	struct inode *inode,
	284	struct writeback_control *wbc,
	285	struct buffer_head *bh)
	286	{
	287	ext4_io_end_t *io_end;
	288	struct page *page = bh->b_page;
	289	int nvecs = bio_get_nr_vecs(bh->b_bdev);
	290	struct bio *bio;
	291
	292	io_end = ext4_init_io_end(inode, GFP_NOFS);
	293	if (!io_end)
	294	return -ENOMEM;
275d3ba6	295	bio = bio_alloc(GFP_NOIO, min(nvecs, BIO_MAX_PAGES));
bd2d0210 TT	296	bio->bi_sector = bh->b_blocknr * (bh->b_size >> 9);
	297	bio->bi_bdev = bh->b_bdev;
	298	bio->bi_private = io->io_end = io_end;
	299	bio->bi_end_io = ext4_end_bio;
	300
bd2d0210 TT	301	io_end->offset = (page->index << PAGE_CACHE_SHIFT) + bh_offset(bh);
	302
	303	io->io_bio = bio;
721a9602	304	io->io_op = (wbc->sync_mode == WB_SYNC_ALL ? WRITE_SYNC : WRITE);
bd2d0210 TT	305	io->io_next_block = bh->b_blocknr;
	306	return 0;
	307	}
	308
	309	static int io_submit_add_bh(struct ext4_io_submit *io,
	310	struct ext4_io_page *io_page,
	311	struct inode *inode,
	312	struct writeback_control *wbc,
	313	struct buffer_head *bh)
	314	{
	315	ext4_io_end_t *io_end;
	316	int ret;
	317
	318	if (buffer_new(bh)) {
	319	clear_buffer_new(bh);
	320	unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
	321	}
	322
	323	if (!buffer_mapped(bh) \|\| buffer_delay(bh)) {
	324	if (!buffer_mapped(bh))
	325	clear_buffer_dirty(bh);
	326	if (io->io_bio)
	327	ext4_io_submit(io);
	328	return 0;
	329	}
	330
	331	if (io->io_bio && bh->b_blocknr != io->io_next_block) {
	332	submit_and_retry:
	333	ext4_io_submit(io);
	334	}
	335	if (io->io_bio == NULL) {
	336	ret = io_submit_init(io, inode, wbc, bh);
	337	if (ret)
	338	return ret;
	339	}
	340	io_end = io->io_end;
	341	if ((io_end->num_io_pages >= MAX_IO_PAGES) &&
	342	(io_end->pages[io_end->num_io_pages-1] != io_page))
	343	goto submit_and_retry;
0edeb71d TM	344	if (buffer_uninit(bh))
0edeb71d TM	345	ext4_set_io_unwritten_flag(inode, io_end);
bd2d0210 TT	346	io->io_end->size += bh->b_size;
	347	io->io_next_block++;
	348	ret = bio_add_page(io->io_bio, bh->b_page, bh->b_size, bh_offset(bh));
	349	if (ret != bh->b_size)
	350	goto submit_and_retry;
	351	if ((io_end->num_io_pages == 0) \|\|
	352	(io_end->pages[io_end->num_io_pages-1] != io_page)) {
	353	io_end->pages[io_end->num_io_pages++] = io_page;
83668e71	354	atomic_inc(&io_page->p_count);
bd2d0210 TT	355	}
	356	return 0;
	357	}
	358
	359	int ext4_bio_write_page(struct ext4_io_submit *io,
	360	struct page *page,
	361	int len,
	362	struct writeback_control *wbc)
	363	{
	364	struct inode *inode = page->mapping->host;
	365	unsigned block_start, block_end, blocksize;
	366	struct ext4_io_page *io_page;
	367	struct buffer_head bh, head;
	368	int ret = 0;
	369
	370	blocksize = 1 << inode->i_blkbits;
	371
d50bdd5a	372	BUG_ON(!PageLocked(page));
bd2d0210	373	BUG_ON(PageWriteback(page));
bd2d0210 TT	374
	375	io_page = kmem_cache_alloc(io_page_cachep, GFP_NOFS);
	376	if (!io_page) {
	377	set_page_dirty(page);
	378	unlock_page(page);
	379	return -ENOMEM;
	380	}
	381	io_page->p_page = page;
83668e71	382	atomic_set(&io_page->p_count, 1);
bd2d0210	383	get_page(page);
a54aa761 TT	384	set_page_writeback(page);
a54aa761 TT	385	ClearPageError(page);
bd2d0210 TT	386
	387	for (bh = head = page_buffers(page), block_start = 0;
	388	bh != head \|\| !block_start;
	389	block_start = block_end, bh = bh->b_this_page) {
d50bdd5a	390
bd2d0210 TT	391	block_end = block_start + blocksize;
bd2d0210 TT	392	if (block_start >= len) {
5a0dc736 YY	393	/*
	394	* Comments copied from block_write_full_page_endio:
	395	*
	396	* The page straddles i_size. It must be zeroed out on
	397	* each and every writepage invocation because it may
	398	* be mmapped. "A file is mapped in multiples of the
	399	* page size. For a file that is not a multiple of
	400	* the page size, the remaining memory is zeroed when
	401	* mapped, and writes to that region are not written
	402	* out to the file."
	403	*/
	404	zero_user_segment(page, block_start, block_end);
bd2d0210 TT	405	clear_buffer_dirty(bh);
	406	set_buffer_uptodate(bh);
	407	continue;
	408	}
d50bdd5a	409	clear_buffer_dirty(bh);
bd2d0210 TT	410	ret = io_submit_add_bh(io, io_page, inode, wbc, bh);
	411	if (ret) {
	412	/*
	413	* We only get here on ENOMEM. Not much else
	414	* we can do but mark the page as dirty, and
	415	* better luck next time.
	416	*/
	417	set_page_dirty(page);
	418	break;
	419	}
	420	}
	421	unlock_page(page);
	422	/*
	423	* If the page was truncated before we could do the writeback,
	424	* or we had a memory allocation error while trying to write
	425	* the first buffer head, we won't have submitted any pages for
	426	* I/O. In that case we need to make sure we've cleared the
	427	* PageWriteback bit from the page to prevent the system from
	428	* wedging later on.
	429	*/
b43d17f3 CW	430	if (atomic_read(&io_page->p_count) == 1)
b43d17f3 CW	431	end_page_writeback(page);
83668e71	432	put_io_page(io_page);
bd2d0210 TT	433	return ret;
bd2d0210 TT	434	}