[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)) {
		end_page_writeback(io_page->p_page);
		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);
		}

		put_io_page(io_end->pages[i]);
	}
	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)
			     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.
	 */
	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)) {
6268b325	63	end_page_writeback(io_page->p_page);
83668e71 TT	64	put_page(io_page->p_page);
	65	kmem_cache_free(io_page_cachep, io_page);
	66	}
	67	}
	68
bd2d0210 TT	69	void ext4_free_io_end(ext4_io_end_t *io)
	70	{
	71	int i;
	72
	73	BUG_ON(!io);
	74	if (io->page)
	75	put_page(io->page);
83668e71 TT	76	for (i = 0; i < io->num_io_pages; i++)
83668e71 TT	77	put_io_page(io->pages[i]);
bd2d0210	78	io->num_io_pages = 0;
4e298021 TT	79	if (atomic_dec_and_test(&EXT4_I(io->inode)->i_ioend_count))
4e298021 TT	80	wake_up_all(ext4_ioend_wq(io->inode));
bd2d0210 TT	81	kmem_cache_free(io_end_cachep, io);
	82	}
	83
	84	/*
	85	* check a range of space and convert unwritten extents to written.
d73d5046 TM	86	*
	87	* Called with inode->i_mutex; we depend on this when we manipulate
	88	* io->flag, since we could otherwise race with ext4_flush_completed_IO()
bd2d0210 TT	89	*/
	90	int ext4_end_io_nolock(ext4_io_end_t *io)
	91	{
	92	struct inode *inode = io->inode;
	93	loff_t offset = io->offset;
	94	ssize_t size = io->size;
	95	int ret = 0;
	96
	97	ext4_debug("ext4_end_io_nolock: io 0x%p from inode %lu,list->next 0x%p,"
	98	"list->prev 0x%p\n",
	99	io, inode->i_ino, io->list.next, io->list.prev);
	100
bd2d0210 TT	101	ret = ext4_convert_unwritten_extents(inode, offset, size);
bd2d0210 TT	102	if (ret < 0) {
b82e384c TT	103	ext4_msg(inode->i_sb, KERN_EMERG,
	104	"failed to convert unwritten extents to written "
	105	"extents -- potential data loss! "
	106	"(inode %lu, offset %llu, size %zd, error %d)",
	107	inode->i_ino, offset, size, ret);
bd2d0210 TT	108	}
	109
	110	if (io->iocb)
	111	aio_complete(io->iocb, io->result, 0);
e9e3bcec	112
266991b1 JM	113	if (io->flag & EXT4_IO_END_DIRECT)
266991b1 JM	114	inode_dio_done(inode);
b82e384c TT	115	/* Wake up anyone waiting on unwritten extent conversion */
	116	if (atomic_dec_and_test(&EXT4_I(inode)->i_aiodio_unwritten))
	117	wake_up_all(ext4_ioend_wq(io->inode));
bd2d0210 TT	118	return ret;
	119	}
	120
	121	/*
	122	* work on completed aio dio IO, to convert unwritten extents to extents
	123	*/
	124	static void ext4_end_io_work(struct work_struct *work)
	125	{
	126	ext4_io_end_t *io = container_of(work, ext4_io_end_t, work);
	127	struct inode *inode = io->inode;
	128	struct ext4_inode_info *ei = EXT4_I(inode);
	129	unsigned long flags;
bd2d0210	130
d73d5046	131	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
491caa43 JM	132	if (io->flag & EXT4_IO_END_IN_FSYNC)
491caa43 JM	133	goto requeue;
d73d5046 TM	134	if (list_empty(&io->list)) {
	135	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
	136	goto free;
	137	}
d73d5046	138
8c0bec21	139	if (!mutex_trylock(&inode->i_mutex)) {
491caa43 JM	140	bool was_queued;
	141	requeue:
	142	was_queued = !!(io->flag & EXT4_IO_END_QUEUED);
	143	io->flag \|= EXT4_IO_END_QUEUED;
b82e384c	144	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
8c0bec21 JZ	145	/*
	146	* Requeue the work instead of waiting so that the work
	147	* items queued after this can be processed.
	148	*/
	149	queue_work(EXT4_SB(inode->i_sb)->dio_unwritten_wq, &io->work);
	150	/*
	151	* To prevent the ext4-dio-unwritten thread from keeping
	152	* requeueing end_io requests and occupying cpu for too long,
	153	* yield the cpu if it sees an end_io request that has already
	154	* been requeued.
	155	*/
491caa43	156	if (was_queued)
8c0bec21	157	yield();
8c0bec21 JZ	158	return;
8c0bec21 JZ	159	}
b82e384c	160	list_del_init(&io->list);
bd2d0210	161	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
b82e384c	162	(void) ext4_end_io_nolock(io);
bd2d0210	163	mutex_unlock(&inode->i_mutex);
d73d5046	164	free:
bd2d0210 TT	165	ext4_free_io_end(io);
	166	}
	167
	168	ext4_io_end_t ext4_init_io_end(struct inode inode, gfp_t flags)
	169	{
b17b35ec	170	ext4_io_end_t *io = kmem_cache_zalloc(io_end_cachep, flags);
bd2d0210	171	if (io) {
f7ad6d2e TT	172	atomic_inc(&EXT4_I(inode)->i_ioend_count);
f7ad6d2e TT	173	io->inode = inode;
bd2d0210 TT	174	INIT_WORK(&io->work, ext4_end_io_work);
	175	INIT_LIST_HEAD(&io->list);
	176	}
	177	return io;
	178	}
	179
	180	/*
	181	* Print an buffer I/O error compatible with the fs/buffer.c. This
	182	* provides compatibility with dmesg scrapers that look for a specific
	183	* buffer I/O error message. We really need a unified error reporting
	184	* structure to userspace ala Digital Unix's uerf system, but it's
	185	* probably not going to happen in my lifetime, due to LKML politics...
	186	*/
	187	static void buffer_io_error(struct buffer_head *bh)
	188	{
	189	char b[BDEVNAME_SIZE];
	190	printk(KERN_ERR "Buffer I/O error on device %s, logical block %llu\n",
	191	bdevname(bh->b_bdev, b),
	192	(unsigned long long)bh->b_blocknr);
	193	}
	194
	195	static void ext4_end_bio(struct bio *bio, int error)
	196	{
	197	ext4_io_end_t *io_end = bio->bi_private;
	198	struct workqueue_struct *wq;
	199	struct inode *inode;
	200	unsigned long flags;
bd2d0210	201	int i;
d50bdd5a	202	sector_t bi_sector = bio->bi_sector;
bd2d0210 TT	203
bd2d0210 TT	204	BUG_ON(!io_end);
bd2d0210 TT	205	bio->bi_private = NULL;
	206	bio->bi_end_io = NULL;
	207	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
	208	error = 0;
bd2d0210 TT	209	bio_put(bio);
bd2d0210 TT	210
bd2d0210 TT	211	for (i = 0; i < io_end->num_io_pages; i++) {
	212	struct page *page = io_end->pages[i]->p_page;
	213	struct buffer_head bh, head;
39db00f1 CW	214	loff_t offset;
39db00f1 CW	215	loff_t io_end_offset;
bd2d0210	216
39db00f1	217	if (error) {
bd2d0210	218	SetPageError(page);
39db00f1 CW	219	set_bit(AS_EIO, &page->mapping->flags);
	220	head = page_buffers(page);
	221	BUG_ON(!head);
	222
	223	io_end_offset = io_end->offset + io_end->size;
bd2d0210 TT	224
	225	offset = (sector_t) page->index << PAGE_CACHE_SHIFT;
	226	bh = head;
	227	do {
	228	if ((offset >= io_end->offset) &&
39db00f1 CW	229	(offset+bh->b_size <= io_end_offset))
	230	buffer_io_error(bh);
	231
bd2d0210 TT	232	offset += bh->b_size;
	233	bh = bh->b_this_page;
	234	} while (bh != head);
	235	}
	236
6268b325	237	put_io_page(io_end->pages[i]);
bd2d0210	238	}
6268b325	239	io_end->num_io_pages = 0;
f7ad6d2e TT	240	inode = io_end->inode;
	241
	242	if (error) {
	243	io_end->flag \|= EXT4_IO_END_ERROR;
	244	ext4_warning(inode->i_sb, "I/O error writing to inode %lu "
	245	"(offset %llu size %ld starting block %llu)",
	246	inode->i_ino,
	247	(unsigned long long) io_end->offset,
	248	(long) io_end->size,
	249	(unsigned long long)
d50bdd5a	250	bi_sector >> (inode->i_blkbits - 9));
f7ad6d2e	251	}
bd2d0210	252
b6168443 TT	253	if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
	254	ext4_free_io_end(io_end);
	255	return;
	256	}
	257
bd2d0210 TT	258	/* Add the io_end to per-inode completed io list*/
	259	spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
	260	list_add_tail(&io_end->list, &EXT4_I(inode)->i_completed_io_list);
	261	spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
	262
	263	wq = EXT4_SB(inode->i_sb)->dio_unwritten_wq;
	264	/* queue the work to convert unwritten extents to written */
	265	queue_work(wq, &io_end->work);
	266	}
	267
	268	void ext4_io_submit(struct ext4_io_submit *io)
	269	{
	270	struct bio *bio = io->io_bio;
	271
	272	if (bio) {
	273	bio_get(io->io_bio);
	274	submit_bio(io->io_op, io->io_bio);
	275	BUG_ON(bio_flagged(io->io_bio, BIO_EOPNOTSUPP));
	276	bio_put(io->io_bio);
	277	}
7dc57615	278	io->io_bio = NULL;
bd2d0210	279	io->io_op = 0;
7dc57615	280	io->io_end = NULL;
bd2d0210 TT	281	}
	282
	283	static int io_submit_init(struct ext4_io_submit *io,
	284	struct inode *inode,
	285	struct writeback_control *wbc,
	286	struct buffer_head *bh)
	287	{
	288	ext4_io_end_t *io_end;
	289	struct page *page = bh->b_page;
	290	int nvecs = bio_get_nr_vecs(bh->b_bdev);
	291	struct bio *bio;
	292
	293	io_end = ext4_init_io_end(inode, GFP_NOFS);
	294	if (!io_end)
	295	return -ENOMEM;
275d3ba6	296	bio = bio_alloc(GFP_NOIO, min(nvecs, BIO_MAX_PAGES));
bd2d0210 TT	297	bio->bi_sector = bh->b_blocknr * (bh->b_size >> 9);
	298	bio->bi_bdev = bh->b_bdev;
	299	bio->bi_private = io->io_end = io_end;
	300	bio->bi_end_io = ext4_end_bio;
	301
bd2d0210 TT	302	io_end->offset = (page->index << PAGE_CACHE_SHIFT) + bh_offset(bh);
	303
	304	io->io_bio = bio;
721a9602	305	io->io_op = (wbc->sync_mode == WB_SYNC_ALL ? WRITE_SYNC : WRITE);
bd2d0210 TT	306	io->io_next_block = bh->b_blocknr;
	307	return 0;
	308	}
	309
	310	static int io_submit_add_bh(struct ext4_io_submit *io,
	311	struct ext4_io_page *io_page,
	312	struct inode *inode,
	313	struct writeback_control *wbc,
	314	struct buffer_head *bh)
	315	{
	316	ext4_io_end_t *io_end;
	317	int ret;
	318
	319	if (buffer_new(bh)) {
	320	clear_buffer_new(bh);
	321	unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
	322	}
	323
	324	if (!buffer_mapped(bh) \|\| buffer_delay(bh)) {
	325	if (!buffer_mapped(bh))
	326	clear_buffer_dirty(bh);
	327	if (io->io_bio)
	328	ext4_io_submit(io);
	329	return 0;
	330	}
	331
	332	if (io->io_bio && bh->b_blocknr != io->io_next_block) {
	333	submit_and_retry:
	334	ext4_io_submit(io);
	335	}
	336	if (io->io_bio == NULL) {
	337	ret = io_submit_init(io, inode, wbc, bh);
	338	if (ret)
	339	return ret;
	340	}
	341	io_end = io->io_end;
	342	if ((io_end->num_io_pages >= MAX_IO_PAGES) &&
	343	(io_end->pages[io_end->num_io_pages-1] != io_page))
	344	goto submit_and_retry;
0edeb71d TM	345	if (buffer_uninit(bh))
0edeb71d TM	346	ext4_set_io_unwritten_flag(inode, io_end);
bd2d0210 TT	347	io->io_end->size += bh->b_size;
	348	io->io_next_block++;
	349	ret = bio_add_page(io->io_bio, bh->b_page, bh->b_size, bh_offset(bh));
	350	if (ret != bh->b_size)
	351	goto submit_and_retry;
	352	if ((io_end->num_io_pages == 0) \|\|
	353	(io_end->pages[io_end->num_io_pages-1] != io_page)) {
	354	io_end->pages[io_end->num_io_pages++] = io_page;
83668e71	355	atomic_inc(&io_page->p_count);
bd2d0210 TT	356	}
	357	return 0;
	358	}
	359
	360	int ext4_bio_write_page(struct ext4_io_submit *io,
	361	struct page *page,
	362	int len,
	363	struct writeback_control *wbc)
	364	{
	365	struct inode *inode = page->mapping->host;
	366	unsigned block_start, block_end, blocksize;
	367	struct ext4_io_page *io_page;
	368	struct buffer_head bh, head;
	369	int ret = 0;
	370
	371	blocksize = 1 << inode->i_blkbits;
	372
d50bdd5a	373	BUG_ON(!PageLocked(page));
bd2d0210	374	BUG_ON(PageWriteback(page));
bd2d0210 TT	375
	376	io_page = kmem_cache_alloc(io_page_cachep, GFP_NOFS);
	377	if (!io_page) {
	378	set_page_dirty(page);
	379	unlock_page(page);
	380	return -ENOMEM;
	381	}
	382	io_page->p_page = page;
83668e71	383	atomic_set(&io_page->p_count, 1);
bd2d0210	384	get_page(page);
a54aa761 TT	385	set_page_writeback(page);
a54aa761 TT	386	ClearPageError(page);
bd2d0210 TT	387
	388	for (bh = head = page_buffers(page), block_start = 0;
	389	bh != head \|\| !block_start;
	390	block_start = block_end, bh = bh->b_this_page) {
d50bdd5a	391
bd2d0210 TT	392	block_end = block_start + blocksize;
bd2d0210 TT	393	if (block_start >= len) {
5a0dc736 YY	394	/*
	395	* Comments copied from block_write_full_page_endio:
	396	*
	397	* The page straddles i_size. It must be zeroed out on
	398	* each and every writepage invocation because it may
	399	* be mmapped. "A file is mapped in multiples of the
	400	* page size. For a file that is not a multiple of
	401	* the page size, the remaining memory is zeroed when
	402	* mapped, and writes to that region are not written
	403	* out to the file."
	404	*/
	405	zero_user_segment(page, block_start, block_end);
bd2d0210 TT	406	clear_buffer_dirty(bh);
	407	set_buffer_uptodate(bh);
	408	continue;
	409	}
d50bdd5a	410	clear_buffer_dirty(bh);
bd2d0210 TT	411	ret = io_submit_add_bh(io, io_page, inode, wbc, bh);
	412	if (ret) {
	413	/*
	414	* We only get here on ENOMEM. Not much else
	415	* we can do but mark the page as dirty, and
	416	* better luck next time.
	417	*/
	418	set_page_dirty(page);
	419	break;
	420	}
	421	}
	422	unlock_page(page);
	423	/*
	424	* If the page was truncated before we could do the writeback,
	425	* or we had a memory allocation error while trying to write
	426	* the first buffer head, we won't have submitted any pages for
	427	* I/O. In that case we need to make sure we've cleared the
	428	* PageWriteback bit from the page to prevent the system from
	429	* wedging later on.
	430	*/
83668e71	431	put_io_page(io_page);
bd2d0210 TT	432	return ret;
bd2d0210 TT	433	}