[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);
	BUG_ON(!list_empty(&io->list));
	BUG_ON(io->flag & EXT4_IO_END_UNWRITTEN);

	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. */
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;
	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_unwritten))
		wake_up_all(ext4_ioend_wq(io->inode));
	return ret;
}

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

	if (list_empty(&EXT4_I(inode)->i_completed_io_list)) {
		ext4_debug("inode %lu completed_io list is empty\n",
			   inode->i_ino);
		return;
	}

	ext4_debug("Dump inode %lu completed_io list\n", inode->i_ino);
	list_for_each_entry(io, &EXT4_I(inode)->i_completed_io_list, list) {
		cur = &io->list;
		before = cur->prev;
		io0 = container_of(before, ext4_io_end_t, list);
		after = cur->next;
		io1 = container_of(after, ext4_io_end_t, list);

		ext4_debug("io 0x%p from inode %lu,prev 0x%p,next 0x%p\n",
			    io, inode->i_ino, io0, io1);
	}
#endif
}

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

	BUG_ON(!(io_end->flag & EXT4_IO_END_UNWRITTEN));
	wq = EXT4_SB(io_end->inode->i_sb)->dio_unwritten_wq;

	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	if (list_empty(&ei->i_completed_io_list)) {
		io_end->flag |= EXT4_IO_END_QUEUED;
		queue_work(wq, &io_end->work);
	}
	list_add_tail(&io_end->list, &ei->i_completed_io_list);
	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
}

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

	INIT_LIST_HEAD(&complete);
	INIT_LIST_HEAD(&to_free);

	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	dump_completed_IO(inode);
	list_replace_init(&ei->i_completed_io_list, &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;

		list_add_tail(&io->list, &complete);
	}
	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	while (!list_empty(&complete)) {
		io = list_entry(complete.next, ext4_io_end_t, list);
		io->flag &= ~EXT4_IO_END_UNWRITTEN;
		/* end_io context can not be destroyed now because it still
		 * used by queued worker. Worker thread will destroy it later */
		if (io->flag & EXT4_IO_END_QUEUED)
			list_del_init(&io->list);
		else
			list_move(&io->list, &to_free);
	}
	/* If we are called from worker context, it is time to clear queued
	 * flag, and destroy it's end_io if it was converted already */
	if (work_io) {
		work_io->flag &= ~EXT4_IO_END_QUEUED;
		if (!(work_io->flag & EXT4_IO_END_UNWRITTEN))
			list_add_tail(&work_io->list, &to_free);
	}
	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);

	while (!list_empty(&to_free)) {
		io = list_entry(to_free.next, ext4_io_end_t, list);
		list_del_init(&io->list);
		ext4_free_io_end(io);
	}
	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);
	ext4_do_flush_completed_IO(io->inode, io);
}

int ext4_flush_unwritten_io(struct inode *inode)
{
	int ret;
	WARN_ON_ONCE(!mutex_is_locked(&inode->i_mutex) &&
		     !(inode->i_state & I_FREEING));
	ret = ext4_do_flush_completed_IO(inode, NULL);
	ext4_unwritten_wait(inode);
	return ret;
}

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 inode *inode;
	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;
	}

	ext4_add_complete_io(io_end);
}

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);
28a535f9	74	BUG_ON(!list_empty(&io->list));
82e54229 DM	75	BUG_ON(io->flag & EXT4_IO_END_UNWRITTEN);
82e54229 DM	76
bd2d0210 TT	77	if (io->page)
bd2d0210 TT	78	put_page(io->page);
83668e71 TT	79	for (i = 0; i < io->num_io_pages; i++)
83668e71 TT	80	put_io_page(io->pages[i]);
bd2d0210	81	io->num_io_pages = 0;
4e298021 TT	82	if (atomic_dec_and_test(&EXT4_I(io->inode)->i_ioend_count))
4e298021 TT	83	wake_up_all(ext4_ioend_wq(io->inode));
bd2d0210 TT	84	kmem_cache_free(io_end_cachep, io);
	85	}
	86
28a535f9 DM	87	/* check a range of space and convert unwritten extents to written. */
28a535f9 DM	88	static int ext4_end_io(ext4_io_end_t *io)
bd2d0210 TT	89	{
	90	struct inode *inode = io->inode;
	91	loff_t offset = io->offset;
	92	ssize_t size = io->size;
	93	int ret = 0;
	94
	95	ext4_debug("ext4_end_io_nolock: io 0x%p from inode %lu,list->next 0x%p,"
	96	"list->prev 0x%p\n",
	97	io, inode->i_ino, io->list.next, io->list.prev);
	98
bd2d0210 TT	99	ret = ext4_convert_unwritten_extents(inode, offset, size);
bd2d0210 TT	100	if (ret < 0) {
b82e384c TT	101	ext4_msg(inode->i_sb, KERN_EMERG,
	102	"failed to convert unwritten extents to written "
	103	"extents -- potential data loss! "
	104	"(inode %lu, offset %llu, size %zd, error %d)",
	105	inode->i_ino, offset, size, ret);
bd2d0210	106	}
bd2d0210 TT	107	if (io->iocb)
bd2d0210 TT	108	aio_complete(io->iocb, io->result, 0);
e9e3bcec	109
266991b1 JM	110	if (io->flag & EXT4_IO_END_DIRECT)
266991b1 JM	111	inode_dio_done(inode);
b82e384c	112	/* Wake up anyone waiting on unwritten extent conversion */
e27f41e1	113	if (atomic_dec_and_test(&EXT4_I(inode)->i_unwritten))
b82e384c	114	wake_up_all(ext4_ioend_wq(io->inode));
bd2d0210 TT	115	return ret;
	116	}
	117
28a535f9 DM	118	static void dump_completed_IO(struct inode *inode)
	119	{
	120	#ifdef EXT4FS_DEBUG
	121	struct list_head cur, before, *after;
	122	ext4_io_end_t io, io0, *io1;
	123	unsigned long flags;
	124
	125	if (list_empty(&EXT4_I(inode)->i_completed_io_list)) {
	126	ext4_debug("inode %lu completed_io list is empty\n",
	127	inode->i_ino);
	128	return;
	129	}
	130
	131	ext4_debug("Dump inode %lu completed_io list\n", inode->i_ino);
	132	list_for_each_entry(io, &EXT4_I(inode)->i_completed_io_list, list) {
	133	cur = &io->list;
	134	before = cur->prev;
	135	io0 = container_of(before, ext4_io_end_t, list);
	136	after = cur->next;
	137	io1 = container_of(after, ext4_io_end_t, list);
	138
	139	ext4_debug("io 0x%p from inode %lu,prev 0x%p,next 0x%p\n",
	140	io, inode->i_ino, io0, io1);
	141	}
	142	#endif
	143	}
	144
	145	/* Add the io_end to per-inode completed end_io list. */
	146	void ext4_add_complete_io(ext4_io_end_t *io_end)
bd2d0210	147	{
28a535f9 DM	148	struct ext4_inode_info *ei = EXT4_I(io_end->inode);
	149	struct workqueue_struct *wq;
	150	unsigned long flags;
	151
	152	BUG_ON(!(io_end->flag & EXT4_IO_END_UNWRITTEN));
	153	wq = EXT4_SB(io_end->inode->i_sb)->dio_unwritten_wq;
bd2d0210	154
d73d5046	155	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
28a535f9 DM	156	if (list_empty(&ei->i_completed_io_list)) {
	157	io_end->flag \|= EXT4_IO_END_QUEUED;
	158	queue_work(wq, &io_end->work);
d73d5046	159	}
28a535f9 DM	160	list_add_tail(&io_end->list, &ei->i_completed_io_list);
	161	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
	162	}
d73d5046	163
28a535f9 DM	164	static int ext4_do_flush_completed_IO(struct inode *inode,
	165	ext4_io_end_t *work_io)
	166	{
	167	ext4_io_end_t *io;
	168	struct list_head unwritten, complete, to_free;
	169	unsigned long flags;
	170	struct ext4_inode_info *ei = EXT4_I(inode);
	171	int err, ret = 0;
	172
	173	INIT_LIST_HEAD(&complete);
	174	INIT_LIST_HEAD(&to_free);
	175
	176	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	177	dump_completed_IO(inode);
	178	list_replace_init(&ei->i_completed_io_list, &unwritten);
	179	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
	180
	181	while (!list_empty(&unwritten)) {
	182	io = list_entry(unwritten.next, ext4_io_end_t, list);
	183	BUG_ON(!(io->flag & EXT4_IO_END_UNWRITTEN));
	184	list_del_init(&io->list);
	185
	186	err = ext4_end_io(io);
	187	if (unlikely(!ret && err))
	188	ret = err;
	189
	190	list_add_tail(&io->list, &complete);
	191	}
28a535f9 DM	192	spin_lock_irqsave(&ei->i_completed_io_lock, flags);
	193	while (!list_empty(&complete)) {
	194	io = list_entry(complete.next, ext4_io_end_t, list);
	195	io->flag &= ~EXT4_IO_END_UNWRITTEN;
	196	/* end_io context can not be destroyed now because it still
	197	* used by queued worker. Worker thread will destroy it later */
	198	if (io->flag & EXT4_IO_END_QUEUED)
	199	list_del_init(&io->list);
	200	else
	201	list_move(&io->list, &to_free);
	202	}
	203	/* If we are called from worker context, it is time to clear queued
	204	* flag, and destroy it's end_io if it was converted already */
	205	if (work_io) {
	206	work_io->flag &= ~EXT4_IO_END_QUEUED;
	207	if (!(work_io->flag & EXT4_IO_END_UNWRITTEN))
	208	list_add_tail(&work_io->list, &to_free);
8c0bec21	209	}
bd2d0210	210	spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
28a535f9 DM	211
	212	while (!list_empty(&to_free)) {
	213	io = list_entry(to_free.next, ext4_io_end_t, list);
	214	list_del_init(&io->list);
	215	ext4_free_io_end(io);
	216	}
	217	return ret;
	218	}
	219
	220	/*
	221	* work on completed aio dio IO, to convert unwritten extents to extents
	222	*/
	223	static void ext4_end_io_work(struct work_struct *work)
	224	{
	225	ext4_io_end_t *io = container_of(work, ext4_io_end_t, work);
	226	ext4_do_flush_completed_IO(io->inode, io);
	227	}
	228
c278531d	229	int ext4_flush_unwritten_io(struct inode *inode)
28a535f9	230	{
c278531d DM	231	int ret;
	232	WARN_ON_ONCE(!mutex_is_locked(&inode->i_mutex) &&
	233	!(inode->i_state & I_FREEING));
	234	ret = ext4_do_flush_completed_IO(inode, NULL);
	235	ext4_unwritten_wait(inode);
	236	return ret;
bd2d0210 TT	237	}
	238
	239	ext4_io_end_t ext4_init_io_end(struct inode inode, gfp_t flags)
	240	{
b17b35ec	241	ext4_io_end_t *io = kmem_cache_zalloc(io_end_cachep, flags);
bd2d0210	242	if (io) {
f7ad6d2e TT	243	atomic_inc(&EXT4_I(inode)->i_ioend_count);
f7ad6d2e TT	244	io->inode = inode;
bd2d0210 TT	245	INIT_WORK(&io->work, ext4_end_io_work);
	246	INIT_LIST_HEAD(&io->list);
	247	}
	248	return io;
	249	}
	250
	251	/*
	252	* Print an buffer I/O error compatible with the fs/buffer.c. This
	253	* provides compatibility with dmesg scrapers that look for a specific
	254	* buffer I/O error message. We really need a unified error reporting
	255	* structure to userspace ala Digital Unix's uerf system, but it's
	256	* probably not going to happen in my lifetime, due to LKML politics...
	257	*/
	258	static void buffer_io_error(struct buffer_head *bh)
	259	{
	260	char b[BDEVNAME_SIZE];
	261	printk(KERN_ERR "Buffer I/O error on device %s, logical block %llu\n",
	262	bdevname(bh->b_bdev, b),
	263	(unsigned long long)bh->b_blocknr);
	264	}
	265
	266	static void ext4_end_bio(struct bio *bio, int error)
	267	{
	268	ext4_io_end_t *io_end = bio->bi_private;
bd2d0210	269	struct inode *inode;
bd2d0210	270	int i;
d50bdd5a	271	sector_t bi_sector = bio->bi_sector;
bd2d0210 TT	272
bd2d0210 TT	273	BUG_ON(!io_end);
bd2d0210 TT	274	bio->bi_private = NULL;
	275	bio->bi_end_io = NULL;
	276	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
	277	error = 0;
bd2d0210 TT	278	bio_put(bio);
bd2d0210 TT	279
bd2d0210 TT	280	for (i = 0; i < io_end->num_io_pages; i++) {
	281	struct page *page = io_end->pages[i]->p_page;
	282	struct buffer_head bh, head;
39db00f1 CW	283	loff_t offset;
39db00f1 CW	284	loff_t io_end_offset;
bd2d0210	285
39db00f1	286	if (error) {
bd2d0210	287	SetPageError(page);
39db00f1 CW	288	set_bit(AS_EIO, &page->mapping->flags);
	289	head = page_buffers(page);
	290	BUG_ON(!head);
	291
	292	io_end_offset = io_end->offset + io_end->size;
bd2d0210 TT	293
	294	offset = (sector_t) page->index << PAGE_CACHE_SHIFT;
	295	bh = head;
	296	do {
	297	if ((offset >= io_end->offset) &&
39db00f1 CW	298	(offset+bh->b_size <= io_end_offset))
	299	buffer_io_error(bh);
	300
bd2d0210 TT	301	offset += bh->b_size;
	302	bh = bh->b_this_page;
	303	} while (bh != head);
	304	}
	305
6268b325	306	put_io_page(io_end->pages[i]);
bd2d0210	307	}
6268b325	308	io_end->num_io_pages = 0;
f7ad6d2e TT	309	inode = io_end->inode;
	310
	311	if (error) {
	312	io_end->flag \|= EXT4_IO_END_ERROR;
	313	ext4_warning(inode->i_sb, "I/O error writing to inode %lu "
	314	"(offset %llu size %ld starting block %llu)",
	315	inode->i_ino,
	316	(unsigned long long) io_end->offset,
	317	(long) io_end->size,
	318	(unsigned long long)
d50bdd5a	319	bi_sector >> (inode->i_blkbits - 9));
f7ad6d2e	320	}
bd2d0210	321
b6168443 TT	322	if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
	323	ext4_free_io_end(io_end);
	324	return;
	325	}
	326
28a535f9	327	ext4_add_complete_io(io_end);
bd2d0210 TT	328	}
	329
	330	void ext4_io_submit(struct ext4_io_submit *io)
	331	{
	332	struct bio *bio = io->io_bio;
	333
	334	if (bio) {
	335	bio_get(io->io_bio);
	336	submit_bio(io->io_op, io->io_bio);
	337	BUG_ON(bio_flagged(io->io_bio, BIO_EOPNOTSUPP));
	338	bio_put(io->io_bio);
	339	}
7dc57615	340	io->io_bio = NULL;
bd2d0210	341	io->io_op = 0;
7dc57615	342	io->io_end = NULL;
bd2d0210 TT	343	}
	344
	345	static int io_submit_init(struct ext4_io_submit *io,
	346	struct inode *inode,
	347	struct writeback_control *wbc,
	348	struct buffer_head *bh)
	349	{
	350	ext4_io_end_t *io_end;
	351	struct page *page = bh->b_page;
	352	int nvecs = bio_get_nr_vecs(bh->b_bdev);
	353	struct bio *bio;
	354
	355	io_end = ext4_init_io_end(inode, GFP_NOFS);
	356	if (!io_end)
	357	return -ENOMEM;
275d3ba6	358	bio = bio_alloc(GFP_NOIO, min(nvecs, BIO_MAX_PAGES));
bd2d0210 TT	359	bio->bi_sector = bh->b_blocknr * (bh->b_size >> 9);
	360	bio->bi_bdev = bh->b_bdev;
	361	bio->bi_private = io->io_end = io_end;
	362	bio->bi_end_io = ext4_end_bio;
	363
bd2d0210 TT	364	io_end->offset = (page->index << PAGE_CACHE_SHIFT) + bh_offset(bh);
	365
	366	io->io_bio = bio;
721a9602	367	io->io_op = (wbc->sync_mode == WB_SYNC_ALL ? WRITE_SYNC : WRITE);
bd2d0210 TT	368	io->io_next_block = bh->b_blocknr;
	369	return 0;
	370	}
	371
	372	static int io_submit_add_bh(struct ext4_io_submit *io,
	373	struct ext4_io_page *io_page,
	374	struct inode *inode,
	375	struct writeback_control *wbc,
	376	struct buffer_head *bh)
	377	{
	378	ext4_io_end_t *io_end;
	379	int ret;
	380
	381	if (buffer_new(bh)) {
	382	clear_buffer_new(bh);
	383	unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
	384	}
	385
	386	if (!buffer_mapped(bh) \|\| buffer_delay(bh)) {
	387	if (!buffer_mapped(bh))
	388	clear_buffer_dirty(bh);
	389	if (io->io_bio)
	390	ext4_io_submit(io);
	391	return 0;
	392	}
	393
	394	if (io->io_bio && bh->b_blocknr != io->io_next_block) {
	395	submit_and_retry:
	396	ext4_io_submit(io);
	397	}
	398	if (io->io_bio == NULL) {
	399	ret = io_submit_init(io, inode, wbc, bh);
	400	if (ret)
	401	return ret;
	402	}
	403	io_end = io->io_end;
	404	if ((io_end->num_io_pages >= MAX_IO_PAGES) &&
	405	(io_end->pages[io_end->num_io_pages-1] != io_page))
	406	goto submit_and_retry;
0edeb71d TM	407	if (buffer_uninit(bh))
0edeb71d TM	408	ext4_set_io_unwritten_flag(inode, io_end);
bd2d0210 TT	409	io->io_end->size += bh->b_size;
	410	io->io_next_block++;
	411	ret = bio_add_page(io->io_bio, bh->b_page, bh->b_size, bh_offset(bh));
	412	if (ret != bh->b_size)
	413	goto submit_and_retry;
	414	if ((io_end->num_io_pages == 0) \|\|
	415	(io_end->pages[io_end->num_io_pages-1] != io_page)) {
	416	io_end->pages[io_end->num_io_pages++] = io_page;
83668e71	417	atomic_inc(&io_page->p_count);
bd2d0210 TT	418	}
	419	return 0;
	420	}
	421
	422	int ext4_bio_write_page(struct ext4_io_submit *io,
	423	struct page *page,
	424	int len,
	425	struct writeback_control *wbc)
	426	{
	427	struct inode *inode = page->mapping->host;
	428	unsigned block_start, block_end, blocksize;
	429	struct ext4_io_page *io_page;
	430	struct buffer_head bh, head;
	431	int ret = 0;
	432
	433	blocksize = 1 << inode->i_blkbits;
	434
d50bdd5a	435	BUG_ON(!PageLocked(page));
bd2d0210	436	BUG_ON(PageWriteback(page));
bd2d0210 TT	437
	438	io_page = kmem_cache_alloc(io_page_cachep, GFP_NOFS);
	439	if (!io_page) {
	440	set_page_dirty(page);
	441	unlock_page(page);
	442	return -ENOMEM;
	443	}
	444	io_page->p_page = page;
83668e71	445	atomic_set(&io_page->p_count, 1);
bd2d0210	446	get_page(page);
a54aa761 TT	447	set_page_writeback(page);
a54aa761 TT	448	ClearPageError(page);
bd2d0210 TT	449
	450	for (bh = head = page_buffers(page), block_start = 0;
	451	bh != head \|\| !block_start;
	452	block_start = block_end, bh = bh->b_this_page) {
d50bdd5a	453
bd2d0210 TT	454	block_end = block_start + blocksize;
bd2d0210 TT	455	if (block_start >= len) {
5a0dc736 YY	456	/*
	457	* Comments copied from block_write_full_page_endio:
	458	*
	459	* The page straddles i_size. It must be zeroed out on
	460	* each and every writepage invocation because it may
	461	* be mmapped. "A file is mapped in multiples of the
	462	* page size. For a file that is not a multiple of
	463	* the page size, the remaining memory is zeroed when
	464	* mapped, and writes to that region are not written
	465	* out to the file."
	466	*/
	467	zero_user_segment(page, block_start, block_end);
bd2d0210 TT	468	clear_buffer_dirty(bh);
	469	set_buffer_uptodate(bh);
	470	continue;
	471	}
d50bdd5a	472	clear_buffer_dirty(bh);
bd2d0210 TT	473	ret = io_submit_add_bh(io, io_page, inode, wbc, bh);
	474	if (ret) {
	475	/*
	476	* We only get here on ENOMEM. Not much else
	477	* we can do but mark the page as dirty, and
	478	* better luck next time.
	479	*/
	480	set_page_dirty(page);
	481	break;
	482	}
	483	}
	484	unlock_page(page);
	485	/*
	486	* If the page was truncated before we could do the writeback,
	487	* or we had a memory allocation error while trying to write
	488	* the first buffer head, we won't have submitted any pages for
	489	* I/O. In that case we need to make sure we've cleared the
	490	* PageWriteback bit from the page to prevent the system from
	491	* wedging later on.
	492	*/
83668e71	493	put_io_page(io_page);
bd2d0210 TT	494	return ret;
bd2d0210 TT	495	}