[linux-2.6-block.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/aio.h>
#include <linux/uio.h>
#include <linux/bio.h>
#include <linux/workqueue.h>
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/ratelimit.h>

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

static struct kmem_cache *io_end_cachep;

int __init ext4_init_pageio(void)
{
	io_end_cachep = KMEM_CACHE(ext4_io_end, SLAB_RECLAIM_ACCOUNT);
	if (io_end_cachep == NULL)
		return -ENOMEM;
	return 0;
}

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

/*
 * 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_ratelimited(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_finish_bio(struct bio *bio)
{
	int i;
	int error = !test_bit(BIO_UPTODATE, &bio->bi_flags);
	struct bio_vec *bvec;

	bio_for_each_segment_all(bvec, bio, i) {
		struct page *page = bvec->bv_page;
		struct buffer_head *bh, *head;
		unsigned bio_start = bvec->bv_offset;
		unsigned bio_end = bio_start + bvec->bv_len;
		unsigned under_io = 0;
		unsigned long flags;

		if (!page)
			continue;

		if (error) {
			SetPageError(page);
			set_bit(AS_EIO, &page->mapping->flags);
		}
		bh = head = page_buffers(page);
		/*
		 * We check all buffers in the page under BH_Uptodate_Lock
		 * to avoid races with other end io clearing async_write flags
		 */
		local_irq_save(flags);
		bit_spin_lock(BH_Uptodate_Lock, &head->b_state);
		do {
			if (bh_offset(bh) < bio_start ||
			    bh_offset(bh) + bh->b_size > bio_end) {
				if (buffer_async_write(bh))
					under_io++;
				continue;
			}
			clear_buffer_async_write(bh);
			if (error)
				buffer_io_error(bh);
		} while ((bh = bh->b_this_page) != head);
		bit_spin_unlock(BH_Uptodate_Lock, &head->b_state);
		local_irq_restore(flags);
		if (!under_io)
			end_page_writeback(page);
	}
}

static void ext4_release_io_end(ext4_io_end_t *io_end)
{
	struct bio *bio, *next_bio;

	BUG_ON(!list_empty(&io_end->list));
	BUG_ON(io_end->flag & EXT4_IO_END_UNWRITTEN);
	WARN_ON(io_end->handle);

	if (atomic_dec_and_test(&EXT4_I(io_end->inode)->i_ioend_count))
		wake_up_all(ext4_ioend_wq(io_end->inode));

	for (bio = io_end->bio; bio; bio = next_bio) {
		next_bio = bio->bi_private;
		ext4_finish_bio(bio);
		bio_put(bio);
	}
	kmem_cache_free(io_end_cachep, io_end);
}

static void ext4_clear_io_unwritten_flag(ext4_io_end_t *io_end)
{
	struct inode *inode = io_end->inode;

	io_end->flag &= ~EXT4_IO_END_UNWRITTEN;
	/* Wake up anyone waiting on unwritten extent conversion */
	if (atomic_dec_and_test(&EXT4_I(inode)->i_unwritten))
		wake_up_all(ext4_ioend_wq(inode));
}

/*
 * Check a range of space and convert unwritten extents to written. Note that
 * we are protected from truncate touching same part of extent tree by the
 * fact that truncate code waits for all DIO to finish (thus exclusion from
 * direct IO is achieved) and also waits for PageWriteback bits. Thus we
 * cannot get to ext4_ext_truncate() before all IOs overlapping that range are
 * completed (happens from ext4_free_ioend()).
 */
static int ext4_end_io(ext4_io_end_t *io)
{
	struct inode *inode = io->inode;
	loff_t offset = io->offset;
	ssize_t size = io->size;
	handle_t *handle = io->handle;
	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);

	io->handle = NULL;	/* Following call will use up the handle */
	ret = ext4_convert_unwritten_extents(handle, 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);
	}
	ext4_clear_io_unwritten_flag(io);
	ext4_release_io_end(io);
	return ret;
}

static void dump_completed_IO(struct inode *inode, struct list_head *head)
{
#ifdef	EXT4FS_DEBUG
	struct list_head *cur, *before, *after;
	ext4_io_end_t *io, *io0, *io1;

	if (list_empty(head))
		return;

	ext4_debug("Dump inode %lu completed io list\n", inode->i_ino);
	list_for_each_entry(io, head, 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);
	}
#endif
}

/* Add the io_end to per-inode completed end_io list. */
static void ext4_add_complete_io(ext4_io_end_t *io_end)
{
	struct ext4_inode_info *ei = EXT4_I(io_end->inode);
	struct ext4_sb_info *sbi = EXT4_SB(io_end->inode->i_sb);
	struct workqueue_struct *wq;
	unsigned long flags;

	/* Only reserved conversions from writeback should enter here */
	WARN_ON(!(io_end->flag & EXT4_IO_END_UNWRITTEN));
	WARN_ON(!io_end->handle && sbi->s_journal);
	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	wq = sbi->rsv_conversion_wq;
	if (list_empty(&ei->i_rsv_conversion_list))
		queue_work(wq, &ei->i_rsv_conversion_work);
	list_add_tail(&io_end->list, &ei->i_rsv_conversion_list);
	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
}

static int ext4_do_flush_completed_IO(struct inode *inode,
				      struct list_head *head)
{
	ext4_io_end_t *io;
	struct list_head unwritten;
	unsigned long flags;
	struct ext4_inode_info *ei = EXT4_I(inode);
	int err, ret = 0;

	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	dump_completed_IO(inode, head);
	list_replace_init(head, &unwritten);
	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);

	while (!list_empty(&unwritten)) {
		io = list_entry(unwritten.next, ext4_io_end_t, list);
		BUG_ON(!(io->flag & EXT4_IO_END_UNWRITTEN));
		list_del_init(&io->list);

		err = ext4_end_io(io);
		if (unlikely(!ret && err))
			ret = err;
	}
	return ret;
}

/*
 * work on completed IO, to convert unwritten extents to extents
 */
void ext4_end_io_rsv_work(struct work_struct *work)
{
	struct ext4_inode_info *ei = container_of(work, struct ext4_inode_info,
						  i_rsv_conversion_work);
	ext4_do_flush_completed_IO(&ei->vfs_inode, &ei->i_rsv_conversion_list);
}

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_LIST_HEAD(&io->list);
		atomic_set(&io->count, 1);
	}
	return io;
}

void ext4_put_io_end_defer(ext4_io_end_t *io_end)
{
	if (atomic_dec_and_test(&io_end->count)) {
		if (!(io_end->flag & EXT4_IO_END_UNWRITTEN) || !io_end->size) {
			ext4_release_io_end(io_end);
			return;
		}
		ext4_add_complete_io(io_end);
	}
}

int ext4_put_io_end(ext4_io_end_t *io_end)
{
	int err = 0;

	if (atomic_dec_and_test(&io_end->count)) {
		if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
			err = ext4_convert_unwritten_extents(io_end->handle,
						io_end->inode, io_end->offset,
						io_end->size);
			io_end->handle = NULL;
			ext4_clear_io_unwritten_flag(io_end);
		}
		ext4_release_io_end(io_end);
	}
	return err;
}

ext4_io_end_t *ext4_get_io_end(ext4_io_end_t *io_end)
{
	atomic_inc(&io_end->count);
	return io_end;
}

/* BIO completion function for page writeback */
static void ext4_end_bio(struct bio *bio, int error)
{
	ext4_io_end_t *io_end = bio->bi_private;
	sector_t bi_sector = bio->bi_iter.bi_sector;

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

	if (error) {
		struct inode *inode = io_end->inode;

		ext4_warning(inode->i_sb, "I/O error %d writing to inode %lu "
			     "(offset %llu size %ld starting block %llu)",
			     error, inode->i_ino,
			     (unsigned long long) io_end->offset,
			     (long) io_end->size,
			     (unsigned long long)
			     bi_sector >> (inode->i_blkbits - 9));
		mapping_set_error(inode->i_mapping, error);
	}

	if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
		/*
		 * Link bio into list hanging from io_end. We have to do it
		 * atomically as bio completions can be racing against each
		 * other.
		 */
		bio->bi_private = xchg(&io_end->bio, bio);
		ext4_put_io_end_defer(io_end);
	} else {
		/*
		 * Drop io_end reference early. Inode can get freed once
		 * we finish the bio.
		 */
		ext4_put_io_end_defer(io_end);
		ext4_finish_bio(bio);
		bio_put(bio);
	}
}

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;
}

void ext4_io_submit_init(struct ext4_io_submit *io,
			 struct writeback_control *wbc)
{
	io->io_op = (wbc->sync_mode == WB_SYNC_ALL ?  WRITE_SYNC : WRITE);
	io->io_bio = NULL;
	io->io_end = NULL;
}

static int io_submit_init_bio(struct ext4_io_submit *io,
			      struct buffer_head *bh)
{
	int nvecs = bio_get_nr_vecs(bh->b_bdev);
	struct bio *bio;

	bio = bio_alloc(GFP_NOIO, min(nvecs, BIO_MAX_PAGES));
	if (!bio)
		return -ENOMEM;
	bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
	bio->bi_bdev = bh->b_bdev;
	bio->bi_end_io = ext4_end_bio;
	bio->bi_private = ext4_get_io_end(io->io_end);
	io->io_bio = bio;
	io->io_next_block = bh->b_blocknr;
	return 0;
}

static int io_submit_add_bh(struct ext4_io_submit *io,
			    struct inode *inode,
			    struct buffer_head *bh)
{
	int ret;

	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_bio(io, bh);
		if (ret)
			return ret;
	}
	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;
	io->io_next_block++;
	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, blocksize;
	struct buffer_head *bh, *head;
	int ret = 0;
	int nr_submitted = 0;

	blocksize = 1 << inode->i_blkbits;

	BUG_ON(!PageLocked(page));
	BUG_ON(PageWriteback(page));

	set_page_writeback(page);
	ClearPageError(page);

	/*
	 * In the first loop we prepare and mark buffers to submit. We have to
	 * mark all buffers in the page before submitting so that
	 * end_page_writeback() cannot be called from ext4_bio_end_io() when IO
	 * on the first buffer finishes and we are still working on submitting
	 * the second buffer.
	 */
	bh = head = page_buffers(page);
	do {
		block_start = bh_offset(bh);
		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_start + blocksize);
			clear_buffer_dirty(bh);
			set_buffer_uptodate(bh);
			continue;
		}
		if (!buffer_dirty(bh) || buffer_delay(bh) ||
		    !buffer_mapped(bh) || buffer_unwritten(bh)) {
			/* A hole? We can safely clear the dirty bit */
			if (!buffer_mapped(bh))
				clear_buffer_dirty(bh);
			if (io->io_bio)
				ext4_io_submit(io);
			continue;
		}
		if (buffer_new(bh)) {
			clear_buffer_new(bh);
			unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
		}
		set_buffer_async_write(bh);
	} while ((bh = bh->b_this_page) != head);

	/* Now submit buffers to write */
	bh = head = page_buffers(page);
	do {
		if (!buffer_async_write(bh))
			continue;
		ret = io_submit_add_bh(io, inode, 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.
			 */
			redirty_page_for_writepage(wbc, page);
			break;
		}
		nr_submitted++;
		clear_buffer_dirty(bh);
	} while ((bh = bh->b_this_page) != head);

	/* Error stopped previous loop? Clean up buffers... */
	if (ret) {
		do {
			clear_buffer_async_write(bh);
			bh = bh->b_this_page;
		} while (bh != head);
	}
	unlock_page(page);
	/* Nothing submitted - we have to end page writeback */
	if (!nr_submitted)
		end_page_writeback(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>
a27bb332	21	#include <linux/aio.h>
bd2d0210 TT	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>
1ae48a63	27	#include <linux/mm.h>
e8974c39	28	#include <linux/ratelimit.h>
bd2d0210 TT	29
	30	#include "ext4_jbd2.h"
	31	#include "xattr.h"
	32	#include "acl.h"
bd2d0210	33
0058f965	34	static struct kmem_cache *io_end_cachep;
bd2d0210	35
5dabfc78	36	int __init ext4_init_pageio(void)
bd2d0210	37	{
bd2d0210	38	io_end_cachep = KMEM_CACHE(ext4_io_end, SLAB_RECLAIM_ACCOUNT);
0058f965	39	if (io_end_cachep == NULL)
bd2d0210	40	return -ENOMEM;
bd2d0210 TT	41	return 0;
	42	}
	43
5dabfc78	44	void ext4_exit_pageio(void)
bd2d0210 TT	45	{
bd2d0210 TT	46	kmem_cache_destroy(io_end_cachep);
bd2d0210 TT	47	}
bd2d0210 TT	48
b0857d30 JK	49	/*
	50	* Print an buffer I/O error compatible with the fs/buffer.c. This
	51	* provides compatibility with dmesg scrapers that look for a specific
	52	* buffer I/O error message. We really need a unified error reporting
	53	* structure to userspace ala Digital Unix's uerf system, but it's
	54	* probably not going to happen in my lifetime, due to LKML politics...
	55	*/
	56	static void buffer_io_error(struct buffer_head *bh)
	57	{
	58	char b[BDEVNAME_SIZE];
e8974c39	59	printk_ratelimited(KERN_ERR "Buffer I/O error on device %s, logical block %llu\n",
b0857d30 JK	60	bdevname(bh->b_bdev, b),
	61	(unsigned long long)bh->b_blocknr);
	62	}
	63
	64	static void ext4_finish_bio(struct bio *bio)
	65	{
	66	int i;
	67	int error = !test_bit(BIO_UPTODATE, &bio->bi_flags);
2c30c71b	68	struct bio_vec *bvec;
b0857d30	69
2c30c71b	70	bio_for_each_segment_all(bvec, bio, i) {
b0857d30 JK	71	struct page *page = bvec->bv_page;
	72	struct buffer_head bh, head;
	73	unsigned bio_start = bvec->bv_offset;
	74	unsigned bio_end = bio_start + bvec->bv_len;
	75	unsigned under_io = 0;
	76	unsigned long flags;
	77
	78	if (!page)
	79	continue;
	80
	81	if (error) {
	82	SetPageError(page);
	83	set_bit(AS_EIO, &page->mapping->flags);
	84	}
	85	bh = head = page_buffers(page);
	86	/*
	87	* We check all buffers in the page under BH_Uptodate_Lock
	88	* to avoid races with other end io clearing async_write flags
	89	*/
	90	local_irq_save(flags);
	91	bit_spin_lock(BH_Uptodate_Lock, &head->b_state);
	92	do {
	93	if (bh_offset(bh) < bio_start \|\|
	94	bh_offset(bh) + bh->b_size > bio_end) {
	95	if (buffer_async_write(bh))
	96	under_io++;
	97	continue;
	98	}
	99	clear_buffer_async_write(bh);
	100	if (error)
	101	buffer_io_error(bh);
	102	} while ((bh = bh->b_this_page) != head);
	103	bit_spin_unlock(BH_Uptodate_Lock, &head->b_state);
	104	local_irq_restore(flags);
	105	if (!under_io)
	106	end_page_writeback(page);
	107	}
	108	}
	109
97a851ed	110	static void ext4_release_io_end(ext4_io_end_t *io_end)
bd2d0210	111	{
b0857d30 JK	112	struct bio bio, next_bio;
b0857d30 JK	113
97a851ed JK	114	BUG_ON(!list_empty(&io_end->list));
97a851ed JK	115	BUG_ON(io_end->flag & EXT4_IO_END_UNWRITTEN);
6b523df4	116	WARN_ON(io_end->handle);
97a851ed JK	117
	118	if (atomic_dec_and_test(&EXT4_I(io_end->inode)->i_ioend_count))
	119	wake_up_all(ext4_ioend_wq(io_end->inode));
b0857d30 JK	120
	121	for (bio = io_end->bio; bio; bio = next_bio) {
	122	next_bio = bio->bi_private;
	123	ext4_finish_bio(bio);
	124	bio_put(bio);
	125	}
97a851ed JK	126	kmem_cache_free(io_end_cachep, io_end);
	127	}
	128
	129	static void ext4_clear_io_unwritten_flag(ext4_io_end_t *io_end)
	130	{
	131	struct inode *inode = io_end->inode;
82e54229	132
97a851ed JK	133	io_end->flag &= ~EXT4_IO_END_UNWRITTEN;
	134	/* Wake up anyone waiting on unwritten extent conversion */
	135	if (atomic_dec_and_test(&EXT4_I(inode)->i_unwritten))
	136	wake_up_all(ext4_ioend_wq(inode));
bd2d0210 TT	137	}
bd2d0210 TT	138
a115f749 JK	139	/*
	140	* Check a range of space and convert unwritten extents to written. Note that
	141	* we are protected from truncate touching same part of extent tree by the
	142	* fact that truncate code waits for all DIO to finish (thus exclusion from
	143	* direct IO is achieved) and also waits for PageWriteback bits. Thus we
	144	* cannot get to ext4_ext_truncate() before all IOs overlapping that range are
	145	* completed (happens from ext4_free_ioend()).
	146	*/
28a535f9	147	static int ext4_end_io(ext4_io_end_t *io)
bd2d0210 TT	148	{
	149	struct inode *inode = io->inode;
	150	loff_t offset = io->offset;
	151	ssize_t size = io->size;
6b523df4	152	handle_t *handle = io->handle;
bd2d0210 TT	153	int ret = 0;
	154
	155	ext4_debug("ext4_end_io_nolock: io 0x%p from inode %lu,list->next 0x%p,"
	156	"list->prev 0x%p\n",
	157	io, inode->i_ino, io->list.next, io->list.prev);
	158
6b523df4 JK	159	io->handle = NULL; /* Following call will use up the handle */
6b523df4 JK	160	ret = ext4_convert_unwritten_extents(handle, inode, offset, size);
bd2d0210	161	if (ret < 0) {
b82e384c TT	162	ext4_msg(inode->i_sb, KERN_EMERG,
	163	"failed to convert unwritten extents to written "
	164	"extents -- potential data loss! "
	165	"(inode %lu, offset %llu, size %zd, error %d)",
	166	inode->i_ino, offset, size, ret);
bd2d0210	167	}
97a851ed JK	168	ext4_clear_io_unwritten_flag(io);
97a851ed JK	169	ext4_release_io_end(io);
bd2d0210 TT	170	return ret;
	171	}
	172
2e8fa54e	173	static void dump_completed_IO(struct inode inode, struct list_head head)
28a535f9 DM	174	{
	175	#ifdef EXT4FS_DEBUG
	176	struct list_head cur, before, *after;
	177	ext4_io_end_t io, io0, *io1;
28a535f9	178
2e8fa54e	179	if (list_empty(head))
28a535f9	180	return;
28a535f9	181
2e8fa54e JK	182	ext4_debug("Dump inode %lu completed io list\n", inode->i_ino);
2e8fa54e JK	183	list_for_each_entry(io, head, list) {
28a535f9 DM	184	cur = &io->list;
	185	before = cur->prev;
	186	io0 = container_of(before, ext4_io_end_t, list);
	187	after = cur->next;
	188	io1 = container_of(after, ext4_io_end_t, list);
	189
	190	ext4_debug("io 0x%p from inode %lu,prev 0x%p,next 0x%p\n",
	191	io, inode->i_ino, io0, io1);
	192	}
	193	#endif
	194	}
	195
	196	/* Add the io_end to per-inode completed end_io list. */
97a851ed	197	static void ext4_add_complete_io(ext4_io_end_t *io_end)
bd2d0210	198	{
28a535f9	199	struct ext4_inode_info *ei = EXT4_I(io_end->inode);
78371a45	200	struct ext4_sb_info *sbi = EXT4_SB(io_end->inode->i_sb);
28a535f9 DM	201	struct workqueue_struct *wq;
	202	unsigned long flags;
	203
7b7a8665 CH	204	/* Only reserved conversions from writeback should enter here */
7b7a8665 CH	205	WARN_ON(!(io_end->flag & EXT4_IO_END_UNWRITTEN));
78371a45	206	WARN_ON(!io_end->handle && sbi->s_journal);
d73d5046	207	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
78371a45	208	wq = sbi->rsv_conversion_wq;
7b7a8665 CH	209	if (list_empty(&ei->i_rsv_conversion_list))
	210	queue_work(wq, &ei->i_rsv_conversion_work);
	211	list_add_tail(&io_end->list, &ei->i_rsv_conversion_list);
28a535f9 DM	212	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
28a535f9 DM	213	}
d73d5046	214
2e8fa54e JK	215	static int ext4_do_flush_completed_IO(struct inode *inode,
2e8fa54e JK	216	struct list_head *head)
28a535f9 DM	217	{
28a535f9 DM	218	ext4_io_end_t *io;
002bd7fa	219	struct list_head unwritten;
28a535f9 DM	220	unsigned long flags;
	221	struct ext4_inode_info *ei = EXT4_I(inode);
	222	int err, ret = 0;
	223
28a535f9	224	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
2e8fa54e JK	225	dump_completed_IO(inode, head);
2e8fa54e JK	226	list_replace_init(head, &unwritten);
28a535f9 DM	227	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
	228
	229	while (!list_empty(&unwritten)) {
	230	io = list_entry(unwritten.next, ext4_io_end_t, list);
	231	BUG_ON(!(io->flag & EXT4_IO_END_UNWRITTEN));
	232	list_del_init(&io->list);
	233
	234	err = ext4_end_io(io);
	235	if (unlikely(!ret && err))
	236	ret = err;
28a535f9 DM	237	}
	238	return ret;
	239	}
	240
	241	/*
2e8fa54e	242	* work on completed IO, to convert unwritten extents to extents
28a535f9	243	*/
2e8fa54e JK	244	void ext4_end_io_rsv_work(struct work_struct *work)
	245	{
	246	struct ext4_inode_info *ei = container_of(work, struct ext4_inode_info,
	247	i_rsv_conversion_work);
	248	ext4_do_flush_completed_IO(&ei->vfs_inode, &ei->i_rsv_conversion_list);
	249	}
	250
bd2d0210 TT	251	ext4_io_end_t ext4_init_io_end(struct inode inode, gfp_t flags)
bd2d0210 TT	252	{
b17b35ec	253	ext4_io_end_t *io = kmem_cache_zalloc(io_end_cachep, flags);
bd2d0210	254	if (io) {
f7ad6d2e TT	255	atomic_inc(&EXT4_I(inode)->i_ioend_count);
f7ad6d2e TT	256	io->inode = inode;
bd2d0210	257	INIT_LIST_HEAD(&io->list);
97a851ed	258	atomic_set(&io->count, 1);
bd2d0210 TT	259	}
	260	return io;
	261	}
	262
97a851ed JK	263	void ext4_put_io_end_defer(ext4_io_end_t *io_end)
	264	{
	265	if (atomic_dec_and_test(&io_end->count)) {
	266	if (!(io_end->flag & EXT4_IO_END_UNWRITTEN) \|\| !io_end->size) {
	267	ext4_release_io_end(io_end);
	268	return;
	269	}
	270	ext4_add_complete_io(io_end);
	271	}
	272	}
	273
	274	int ext4_put_io_end(ext4_io_end_t *io_end)
	275	{
	276	int err = 0;
	277
	278	if (atomic_dec_and_test(&io_end->count)) {
	279	if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
6b523df4 JK	280	err = ext4_convert_unwritten_extents(io_end->handle,
	281	io_end->inode, io_end->offset,
	282	io_end->size);
	283	io_end->handle = NULL;
97a851ed JK	284	ext4_clear_io_unwritten_flag(io_end);
	285	}
	286	ext4_release_io_end(io_end);
	287	}
	288	return err;
	289	}
	290
	291	ext4_io_end_t ext4_get_io_end(ext4_io_end_t io_end)
	292	{
	293	atomic_inc(&io_end->count);
	294	return io_end;
	295	}
	296
822dbba3	297	/* BIO completion function for page writeback */
bd2d0210 TT	298	static void ext4_end_bio(struct bio *bio, int error)
	299	{
	300	ext4_io_end_t *io_end = bio->bi_private;
4f024f37	301	sector_t bi_sector = bio->bi_iter.bi_sector;
bd2d0210 TT	302
bd2d0210 TT	303	BUG_ON(!io_end);
bd2d0210 TT	304	bio->bi_end_io = NULL;
	305	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
	306	error = 0;
0058f965	307
f7ad6d2e	308	if (error) {
b0857d30 JK	309	struct inode *inode = io_end->inode;
b0857d30 JK	310
9503c67c	311	ext4_warning(inode->i_sb, "I/O error %d writing to inode %lu "
f7ad6d2e	312	"(offset %llu size %ld starting block %llu)",
9503c67c	313	error, inode->i_ino,
f7ad6d2e TT	314	(unsigned long long) io_end->offset,
	315	(long) io_end->size,
	316	(unsigned long long)
d50bdd5a	317	bi_sector >> (inode->i_blkbits - 9));
9503c67c	318	mapping_set_error(inode->i_mapping, error);
f7ad6d2e	319	}
822dbba3 JK	320
	321	if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
	322	/*
	323	* Link bio into list hanging from io_end. We have to do it
	324	* atomically as bio completions can be racing against each
	325	* other.
	326	*/
	327	bio->bi_private = xchg(&io_end->bio, bio);
	328	ext4_put_io_end_defer(io_end);
	329	} else {
	330	/*
	331	* Drop io_end reference early. Inode can get freed once
	332	* we finish the bio.
	333	*/
	334	ext4_put_io_end_defer(io_end);
	335	ext4_finish_bio(bio);
	336	bio_put(bio);
	337	}
bd2d0210 TT	338	}
	339
	340	void ext4_io_submit(struct ext4_io_submit *io)
	341	{
	342	struct bio *bio = io->io_bio;
	343
	344	if (bio) {
	345	bio_get(io->io_bio);
	346	submit_bio(io->io_op, io->io_bio);
	347	BUG_ON(bio_flagged(io->io_bio, BIO_EOPNOTSUPP));
	348	bio_put(io->io_bio);
	349	}
7dc57615	350	io->io_bio = NULL;
97a851ed JK	351	}
	352
	353	void ext4_io_submit_init(struct ext4_io_submit *io,
	354	struct writeback_control *wbc)
	355	{
	356	io->io_op = (wbc->sync_mode == WB_SYNC_ALL ? WRITE_SYNC : WRITE);
	357	io->io_bio = NULL;
7dc57615	358	io->io_end = NULL;
bd2d0210 TT	359	}
bd2d0210 TT	360
97a851ed JK	361	static int io_submit_init_bio(struct ext4_io_submit *io,
97a851ed JK	362	struct buffer_head *bh)
bd2d0210	363	{
bd2d0210 TT	364	int nvecs = bio_get_nr_vecs(bh->b_bdev);
	365	struct bio *bio;
	366
275d3ba6	367	bio = bio_alloc(GFP_NOIO, min(nvecs, BIO_MAX_PAGES));
a1d8d9a7 TT	368	if (!bio)
a1d8d9a7 TT	369	return -ENOMEM;
4f024f37	370	bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
bd2d0210	371	bio->bi_bdev = bh->b_bdev;
bd2d0210	372	bio->bi_end_io = ext4_end_bio;
97a851ed	373	bio->bi_private = ext4_get_io_end(io->io_end);
bd2d0210	374	io->io_bio = bio;
bd2d0210 TT	375	io->io_next_block = bh->b_blocknr;
	376	return 0;
	377	}
	378
	379	static int io_submit_add_bh(struct ext4_io_submit *io,
bd2d0210	380	struct inode *inode,
bd2d0210 TT	381	struct buffer_head *bh)
bd2d0210 TT	382	{
bd2d0210 TT	383	int ret;
bd2d0210 TT	384
bd2d0210 TT	385	if (io->io_bio && bh->b_blocknr != io->io_next_block) {
	386	submit_and_retry:
	387	ext4_io_submit(io);
	388	}
	389	if (io->io_bio == NULL) {
97a851ed	390	ret = io_submit_init_bio(io, bh);
bd2d0210 TT	391	if (ret)
	392	return ret;
	393	}
97a851ed JK	394	ret = bio_add_page(io->io_bio, bh->b_page, bh->b_size, bh_offset(bh));
	395	if (ret != bh->b_size)
	396	goto submit_and_retry;
bd2d0210	397	io->io_next_block++;
bd2d0210 TT	398	return 0;
	399	}
	400
	401	int ext4_bio_write_page(struct ext4_io_submit *io,
	402	struct page *page,
	403	int len,
	404	struct writeback_control *wbc)
	405	{
	406	struct inode *inode = page->mapping->host;
0058f965	407	unsigned block_start, blocksize;
bd2d0210 TT	408	struct buffer_head bh, head;
bd2d0210 TT	409	int ret = 0;
0058f965	410	int nr_submitted = 0;
bd2d0210 TT	411
	412	blocksize = 1 << inode->i_blkbits;
	413
d50bdd5a	414	BUG_ON(!PageLocked(page));
bd2d0210	415	BUG_ON(PageWriteback(page));
bd2d0210	416
a54aa761 TT	417	set_page_writeback(page);
a54aa761 TT	418	ClearPageError(page);
bd2d0210	419
0058f965 JK	420	/*
	421	* In the first loop we prepare and mark buffers to submit. We have to
	422	* mark all buffers in the page before submitting so that
	423	* end_page_writeback() cannot be called from ext4_bio_end_io() when IO
	424	* on the first buffer finishes and we are still working on submitting
	425	* the second buffer.
	426	*/
	427	bh = head = page_buffers(page);
	428	do {
	429	block_start = bh_offset(bh);
bd2d0210	430	if (block_start >= len) {
5a0dc736 YY	431	/*
	432	* Comments copied from block_write_full_page_endio:
	433	*
	434	* The page straddles i_size. It must be zeroed out on
	435	* each and every writepage invocation because it may
	436	* be mmapped. "A file is mapped in multiples of the
	437	* page size. For a file that is not a multiple of
	438	* the page size, the remaining memory is zeroed when
	439	* mapped, and writes to that region are not written
	440	* out to the file."
	441	*/
0058f965 JK	442	zero_user_segment(page, block_start,
0058f965 JK	443	block_start + blocksize);
bd2d0210 TT	444	clear_buffer_dirty(bh);
	445	set_buffer_uptodate(bh);
	446	continue;
	447	}
8a850c3f JK	448	if (!buffer_dirty(bh) \|\| buffer_delay(bh) \|\|
	449	!buffer_mapped(bh) \|\| buffer_unwritten(bh)) {
	450	/* A hole? We can safely clear the dirty bit */
	451	if (!buffer_mapped(bh))
	452	clear_buffer_dirty(bh);
	453	if (io->io_bio)
	454	ext4_io_submit(io);
	455	continue;
	456	}
0058f965 JK	457	if (buffer_new(bh)) {
	458	clear_buffer_new(bh);
	459	unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
	460	}
	461	set_buffer_async_write(bh);
	462	} while ((bh = bh->b_this_page) != head);
	463
	464	/* Now submit buffers to write */
	465	bh = head = page_buffers(page);
	466	do {
	467	if (!buffer_async_write(bh))
	468	continue;
97a851ed	469	ret = io_submit_add_bh(io, inode, bh);
bd2d0210 TT	470	if (ret) {
	471	/*
	472	* We only get here on ENOMEM. Not much else
	473	* we can do but mark the page as dirty, and
	474	* better luck next time.
	475	*/
1ae48a63	476	redirty_page_for_writepage(wbc, page);
bd2d0210 TT	477	break;
bd2d0210 TT	478	}
0058f965	479	nr_submitted++;
1ae48a63	480	clear_buffer_dirty(bh);
0058f965 JK	481	} while ((bh = bh->b_this_page) != head);
	482
	483	/* Error stopped previous loop? Clean up buffers... */
	484	if (ret) {
	485	do {
	486	clear_buffer_async_write(bh);
	487	bh = bh->b_this_page;
	488	} while (bh != head);
bd2d0210 TT	489	}
bd2d0210 TT	490	unlock_page(page);
0058f965 JK	491	/* Nothing submitted - we have to end page writeback */
	492	if (!nr_submitted)
	493	end_page_writeback(page);
bd2d0210 TT	494	return ret;
bd2d0210 TT	495	}