[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/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 <linux/mm.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)
{
	printk_ratelimited(KERN_ERR "Buffer I/O error on device %pg, logical block %llu\n",
		       bh->b_bdev,
			(unsigned long long)bh->b_blocknr);
}

static void ext4_finish_bio(struct bio *bio)
{
	int i;
	struct bio_vec *bvec;

	bio_for_each_segment_all(bvec, bio, i) {
		struct page *page = bvec->bv_page;
#ifdef CONFIG_EXT4_FS_ENCRYPTION
		struct page *data_page = NULL;
		struct ext4_crypto_ctx *ctx = NULL;
#endif
		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;

#ifdef CONFIG_EXT4_FS_ENCRYPTION
		if (!page->mapping) {
			/* The bounce data pages are unmapped. */
			data_page = page;
			ctx = (struct ext4_crypto_ctx *)page_private(data_page);
			page = ctx->w.control_page;
		}
#endif

		if (bio->bi_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 (bio->bi_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) {
#ifdef CONFIG_EXT4_FS_ENCRYPTION
			if (ctx)
				ext4_restore_control_page(data_page);
#endif
			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);

	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);
}

/*
 * 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) {
		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)
{
	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 (bio->bi_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)",
			     bio->bi_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, bio->bi_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) {
		int io_op = io->io_wbc->sync_mode == WB_SYNC_ALL ?
			    WRITE_SYNC : WRITE;
		bio_get(io->io_bio);
		submit_bio(io_op, io->io_bio);
		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_wbc = wbc;
	io->io_bio = NULL;
	io->io_end = NULL;
}

static int io_submit_init_bio(struct ext4_io_submit *io,
			      struct buffer_head *bh)
{
	struct bio *bio;

	bio = bio_alloc(GFP_NOIO, BIO_MAX_PAGES);
	if (!bio)
		return -ENOMEM;
	wbc_init_bio(io->io_wbc, bio);
	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 page *page,
			    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, page, bh->b_size, bh_offset(bh));
	if (ret != bh->b_size)
		goto submit_and_retry;
	wbc_account_io(io->io_wbc, page, bh->b_size);
	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,
			bool keep_towrite)
{
	struct page *data_page = NULL;
	struct inode *inode = page->mapping->host;
	unsigned block_start, blocksize;
	struct buffer_head *bh, *head;
	int ret = 0;
	int nr_submitted = 0;
	int nr_to_submit = 0;

	blocksize = 1 << inode->i_blkbits;

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

	if (keep_towrite)
		set_page_writeback_keepwrite(page);
	else
		set_page_writeback(page);
	ClearPageError(page);

	/*
	 * Comments copied from block_write_full_page:
	 *
	 * 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."
	 */
	if (len < PAGE_CACHE_SIZE)
		zero_user_segment(page, len, PAGE_CACHE_SIZE);
	/*
	 * 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) {
			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);
		nr_to_submit++;
	} while ((bh = bh->b_this_page) != head);

	bh = head = page_buffers(page);

	if (ext4_encrypted_inode(inode) && S_ISREG(inode->i_mode) &&
	    nr_to_submit) {
		data_page = ext4_encrypt(inode, page);
		if (IS_ERR(data_page)) {
			ret = PTR_ERR(data_page);
			data_page = NULL;
			goto out;
		}
	}

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

	/* Error stopped previous loop? Clean up buffers... */
	if (ret) {
	out:
		if (data_page)
			ext4_restore_control_page(data_page);
		printk_ratelimited(KERN_ERR "%s: ret = %d\n", __func__, ret);
		redirty_page_for_writepage(wbc, page);
		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>
bd2d0210 TT	10	#include <linux/time.h>
bd2d0210 TT	11	#include <linux/highuid.h>
	12	#include <linux/pagemap.h>
	13	#include <linux/quotaops.h>
	14	#include <linux/string.h>
	15	#include <linux/buffer_head.h>
	16	#include <linux/writeback.h>
	17	#include <linux/pagevec.h>
	18	#include <linux/mpage.h>
	19	#include <linux/namei.h>
	20	#include <linux/uio.h>
	21	#include <linux/bio.h>
	22	#include <linux/workqueue.h>
	23	#include <linux/kernel.h>
	24	#include <linux/slab.h>
1ae48a63	25	#include <linux/mm.h>
bd2d0210 TT	26
	27	#include "ext4_jbd2.h"
	28	#include "xattr.h"
	29	#include "acl.h"
bd2d0210	30
0058f965	31	static struct kmem_cache *io_end_cachep;
bd2d0210	32
5dabfc78	33	int __init ext4_init_pageio(void)
bd2d0210	34	{
bd2d0210	35	io_end_cachep = KMEM_CACHE(ext4_io_end, SLAB_RECLAIM_ACCOUNT);
0058f965	36	if (io_end_cachep == NULL)
bd2d0210	37	return -ENOMEM;
bd2d0210 TT	38	return 0;
	39	}
	40
5dabfc78	41	void ext4_exit_pageio(void)
bd2d0210 TT	42	{
bd2d0210 TT	43	kmem_cache_destroy(io_end_cachep);
bd2d0210 TT	44	}
bd2d0210 TT	45
b0857d30 JK	46	/*
	47	* Print an buffer I/O error compatible with the fs/buffer.c. This
	48	* provides compatibility with dmesg scrapers that look for a specific
	49	* buffer I/O error message. We really need a unified error reporting
	50	* structure to userspace ala Digital Unix's uerf system, but it's
	51	* probably not going to happen in my lifetime, due to LKML politics...
	52	*/
	53	static void buffer_io_error(struct buffer_head *bh)
	54	{
a1c6f057 DM	55	printk_ratelimited(KERN_ERR "Buffer I/O error on device %pg, logical block %llu\n",
a1c6f057 DM	56	bh->b_bdev,
b0857d30 JK	57	(unsigned long long)bh->b_blocknr);
	58	}
	59
	60	static void ext4_finish_bio(struct bio *bio)
	61	{
	62	int i;
2c30c71b	63	struct bio_vec *bvec;
b0857d30	64
2c30c71b	65	bio_for_each_segment_all(bvec, bio, i) {
b0857d30	66	struct page *page = bvec->bv_page;
2058f83a MH	67	#ifdef CONFIG_EXT4_FS_ENCRYPTION
	68	struct page *data_page = NULL;
	69	struct ext4_crypto_ctx *ctx = NULL;
	70	#endif
b0857d30 JK	71	struct buffer_head bh, head;
	72	unsigned bio_start = bvec->bv_offset;
	73	unsigned bio_end = bio_start + bvec->bv_len;
	74	unsigned under_io = 0;
	75	unsigned long flags;
	76
	77	if (!page)
	78	continue;
	79
2058f83a MH	80	#ifdef CONFIG_EXT4_FS_ENCRYPTION
	81	if (!page->mapping) {
	82	/* The bounce data pages are unmapped. */
	83	data_page = page;
	84	ctx = (struct ext4_crypto_ctx *)page_private(data_page);
614def70	85	page = ctx->w.control_page;
2058f83a MH	86	}
	87	#endif
	88
4246a0b6	89	if (bio->bi_error) {
b0857d30 JK	90	SetPageError(page);
	91	set_bit(AS_EIO, &page->mapping->flags);
	92	}
	93	bh = head = page_buffers(page);
	94	/*
	95	* We check all buffers in the page under BH_Uptodate_Lock
	96	* to avoid races with other end io clearing async_write flags
	97	*/
	98	local_irq_save(flags);
	99	bit_spin_lock(BH_Uptodate_Lock, &head->b_state);
	100	do {
	101	if (bh_offset(bh) < bio_start \|\|
	102	bh_offset(bh) + bh->b_size > bio_end) {
	103	if (buffer_async_write(bh))
	104	under_io++;
	105	continue;
	106	}
	107	clear_buffer_async_write(bh);
4246a0b6	108	if (bio->bi_error)
b0857d30 JK	109	buffer_io_error(bh);
	110	} while ((bh = bh->b_this_page) != head);
	111	bit_spin_unlock(BH_Uptodate_Lock, &head->b_state);
	112	local_irq_restore(flags);
2058f83a MH	113	if (!under_io) {
	114	#ifdef CONFIG_EXT4_FS_ENCRYPTION
	115	if (ctx)
	116	ext4_restore_control_page(data_page);
	117	#endif
b0857d30	118	end_page_writeback(page);
2058f83a	119	}
b0857d30 JK	120	}
	121	}
	122
97a851ed	123	static void ext4_release_io_end(ext4_io_end_t *io_end)
bd2d0210	124	{
b0857d30 JK	125	struct bio bio, next_bio;
b0857d30 JK	126
97a851ed JK	127	BUG_ON(!list_empty(&io_end->list));
97a851ed JK	128	BUG_ON(io_end->flag & EXT4_IO_END_UNWRITTEN);
6b523df4	129	WARN_ON(io_end->handle);
97a851ed	130
b0857d30 JK	131	for (bio = io_end->bio; bio; bio = next_bio) {
	132	next_bio = bio->bi_private;
	133	ext4_finish_bio(bio);
	134	bio_put(bio);
	135	}
97a851ed JK	136	kmem_cache_free(io_end_cachep, io_end);
	137	}
	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	255	io->inode = inode;
bd2d0210	256	INIT_LIST_HEAD(&io->list);
97a851ed	257	atomic_set(&io->count, 1);
bd2d0210 TT	258	}
	259	return io;
	260	}
	261
97a851ed JK	262	void ext4_put_io_end_defer(ext4_io_end_t *io_end)
	263	{
	264	if (atomic_dec_and_test(&io_end->count)) {
	265	if (!(io_end->flag & EXT4_IO_END_UNWRITTEN) \|\| !io_end->size) {
	266	ext4_release_io_end(io_end);
	267	return;
	268	}
	269	ext4_add_complete_io(io_end);
	270	}
	271	}
	272
	273	int ext4_put_io_end(ext4_io_end_t *io_end)
	274	{
	275	int err = 0;
	276
	277	if (atomic_dec_and_test(&io_end->count)) {
	278	if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
6b523df4 JK	279	err = ext4_convert_unwritten_extents(io_end->handle,
	280	io_end->inode, io_end->offset,
	281	io_end->size);
	282	io_end->handle = NULL;
97a851ed JK	283	ext4_clear_io_unwritten_flag(io_end);
	284	}
	285	ext4_release_io_end(io_end);
	286	}
	287	return err;
	288	}
	289
	290	ext4_io_end_t ext4_get_io_end(ext4_io_end_t io_end)
	291	{
	292	atomic_inc(&io_end->count);
	293	return io_end;
	294	}
	295
822dbba3	296	/* BIO completion function for page writeback */
4246a0b6	297	static void ext4_end_bio(struct bio *bio)
bd2d0210 TT	298	{
bd2d0210 TT	299	ext4_io_end_t *io_end = bio->bi_private;
4f024f37	300	sector_t bi_sector = bio->bi_iter.bi_sector;
bd2d0210 TT	301
bd2d0210 TT	302	BUG_ON(!io_end);
bd2d0210	303	bio->bi_end_io = NULL;
0058f965	304
4246a0b6	305	if (bio->bi_error) {
b0857d30 JK	306	struct inode *inode = io_end->inode;
b0857d30 JK	307
9503c67c	308	ext4_warning(inode->i_sb, "I/O error %d writing to inode %lu "
f7ad6d2e	309	"(offset %llu size %ld starting block %llu)",
4246a0b6	310	bio->bi_error, inode->i_ino,
f7ad6d2e TT	311	(unsigned long long) io_end->offset,
	312	(long) io_end->size,
	313	(unsigned long long)
d50bdd5a	314	bi_sector >> (inode->i_blkbits - 9));
4246a0b6	315	mapping_set_error(inode->i_mapping, bio->bi_error);
f7ad6d2e	316	}
822dbba3 JK	317
	318	if (io_end->flag & EXT4_IO_END_UNWRITTEN) {
	319	/*
	320	* Link bio into list hanging from io_end. We have to do it
	321	* atomically as bio completions can be racing against each
	322	* other.
	323	*/
	324	bio->bi_private = xchg(&io_end->bio, bio);
	325	ext4_put_io_end_defer(io_end);
	326	} else {
	327	/*
	328	* Drop io_end reference early. Inode can get freed once
	329	* we finish the bio.
	330	*/
	331	ext4_put_io_end_defer(io_end);
	332	ext4_finish_bio(bio);
	333	bio_put(bio);
	334	}
bd2d0210 TT	335	}
	336
	337	void ext4_io_submit(struct ext4_io_submit *io)
	338	{
	339	struct bio *bio = io->io_bio;
	340
	341	if (bio) {
5a33911f TH	342	int io_op = io->io_wbc->sync_mode == WB_SYNC_ALL ?
5a33911f TH	343	WRITE_SYNC : WRITE;
bd2d0210	344	bio_get(io->io_bio);
5a33911f	345	submit_bio(io_op, io->io_bio);
bd2d0210 TT	346	bio_put(io->io_bio);
bd2d0210 TT	347	}
7dc57615	348	io->io_bio = NULL;
97a851ed JK	349	}
	350
	351	void ext4_io_submit_init(struct ext4_io_submit *io,
	352	struct writeback_control *wbc)
	353	{
5a33911f	354	io->io_wbc = wbc;
97a851ed	355	io->io_bio = NULL;
7dc57615	356	io->io_end = NULL;
bd2d0210 TT	357	}
bd2d0210 TT	358
97a851ed JK	359	static int io_submit_init_bio(struct ext4_io_submit *io,
97a851ed JK	360	struct buffer_head *bh)
bd2d0210	361	{
bd2d0210 TT	362	struct bio *bio;
bd2d0210 TT	363
b54ffb73	364	bio = bio_alloc(GFP_NOIO, BIO_MAX_PAGES);
a1d8d9a7 TT	365	if (!bio)
a1d8d9a7 TT	366	return -ENOMEM;
001e4a87	367	wbc_init_bio(io->io_wbc, bio);
4f024f37	368	bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
bd2d0210	369	bio->bi_bdev = bh->b_bdev;
bd2d0210	370	bio->bi_end_io = ext4_end_bio;
97a851ed	371	bio->bi_private = ext4_get_io_end(io->io_end);
bd2d0210	372	io->io_bio = bio;
bd2d0210 TT	373	io->io_next_block = bh->b_blocknr;
	374	return 0;
	375	}
	376
	377	static int io_submit_add_bh(struct ext4_io_submit *io,
bd2d0210	378	struct inode *inode,
2058f83a	379	struct page *page,
bd2d0210 TT	380	struct buffer_head *bh)
bd2d0210 TT	381	{
bd2d0210 TT	382	int ret;
bd2d0210 TT	383
bd2d0210 TT	384	if (io->io_bio && bh->b_blocknr != io->io_next_block) {
	385	submit_and_retry:
	386	ext4_io_submit(io);
	387	}
	388	if (io->io_bio == NULL) {
97a851ed	389	ret = io_submit_init_bio(io, bh);
bd2d0210 TT	390	if (ret)
	391	return ret;
	392	}
2058f83a	393	ret = bio_add_page(io->io_bio, page, bh->b_size, bh_offset(bh));
97a851ed JK	394	if (ret != bh->b_size)
97a851ed JK	395	goto submit_and_retry;
001e4a87	396	wbc_account_io(io->io_wbc, page, bh->b_size);
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,
1c8349a1 NJ	404	struct writeback_control *wbc,
1c8349a1 NJ	405	bool keep_towrite)
bd2d0210	406	{
2058f83a	407	struct page *data_page = NULL;
bd2d0210	408	struct inode *inode = page->mapping->host;
0058f965	409	unsigned block_start, blocksize;
bd2d0210 TT	410	struct buffer_head bh, head;
bd2d0210 TT	411	int ret = 0;
0058f965	412	int nr_submitted = 0;
937d7b84	413	int nr_to_submit = 0;
bd2d0210 TT	414
	415	blocksize = 1 << inode->i_blkbits;
	416
d50bdd5a	417	BUG_ON(!PageLocked(page));
bd2d0210	418	BUG_ON(PageWriteback(page));
bd2d0210	419
1c8349a1 NJ	420	if (keep_towrite)
	421	set_page_writeback_keepwrite(page);
	422	else
	423	set_page_writeback(page);
a54aa761	424	ClearPageError(page);
bd2d0210	425
eeece469	426	/*
f8409abd	427	* Comments copied from block_write_full_page:
eeece469 JK	428	*
	429	* The page straddles i_size. It must be zeroed out on each and every
	430	* writepage invocation because it may be mmapped. "A file is mapped
	431	* in multiples of the page size. For a file that is not a multiple of
	432	* the page size, the remaining memory is zeroed when mapped, and
	433	* writes to that region are not written out to the file."
	434	*/
	435	if (len < PAGE_CACHE_SIZE)
	436	zero_user_segment(page, len, PAGE_CACHE_SIZE);
0058f965 JK	437	/*
	438	* In the first loop we prepare and mark buffers to submit. We have to
	439	* mark all buffers in the page before submitting so that
	440	* end_page_writeback() cannot be called from ext4_bio_end_io() when IO
	441	* on the first buffer finishes and we are still working on submitting
	442	* the second buffer.
	443	*/
	444	bh = head = page_buffers(page);
	445	do {
	446	block_start = bh_offset(bh);
bd2d0210 TT	447	if (block_start >= len) {
	448	clear_buffer_dirty(bh);
	449	set_buffer_uptodate(bh);
	450	continue;
	451	}
8a850c3f JK	452	if (!buffer_dirty(bh) \|\| buffer_delay(bh) \|\|
	453	!buffer_mapped(bh) \|\| buffer_unwritten(bh)) {
	454	/* A hole? We can safely clear the dirty bit */
	455	if (!buffer_mapped(bh))
	456	clear_buffer_dirty(bh);
	457	if (io->io_bio)
	458	ext4_io_submit(io);
	459	continue;
	460	}
0058f965 JK	461	if (buffer_new(bh)) {
	462	clear_buffer_new(bh);
	463	unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
	464	}
	465	set_buffer_async_write(bh);
937d7b84	466	nr_to_submit++;
0058f965 JK	467	} while ((bh = bh->b_this_page) != head);
0058f965 JK	468
0058f965	469	bh = head = page_buffers(page);
2058f83a	470
937d7b84 TT	471	if (ext4_encrypted_inode(inode) && S_ISREG(inode->i_mode) &&
937d7b84 TT	472	nr_to_submit) {
2058f83a MH	473	data_page = ext4_encrypt(inode, page);
	474	if (IS_ERR(data_page)) {
	475	ret = PTR_ERR(data_page);
	476	data_page = NULL;
	477	goto out;
	478	}
	479	}
	480
	481	/* Now submit buffers to write */
0058f965 JK	482	do {
	483	if (!buffer_async_write(bh))
	484	continue;
2058f83a MH	485	ret = io_submit_add_bh(io, inode,
2058f83a MH	486	data_page ? data_page : page, bh);
bd2d0210 TT	487	if (ret) {
	488	/*
	489	* We only get here on ENOMEM. Not much else
	490	* we can do but mark the page as dirty, and
	491	* better luck next time.
	492	*/
bd2d0210 TT	493	break;
bd2d0210 TT	494	}
0058f965	495	nr_submitted++;
1ae48a63	496	clear_buffer_dirty(bh);
0058f965 JK	497	} while ((bh = bh->b_this_page) != head);
	498
	499	/* Error stopped previous loop? Clean up buffers... */
	500	if (ret) {
2058f83a MH	501	out:
	502	if (data_page)
	503	ext4_restore_control_page(data_page);
	504	printk_ratelimited(KERN_ERR "%s: ret = %d\n", __func__, ret);
	505	redirty_page_for_writepage(wbc, page);
0058f965 JK	506	do {
	507	clear_buffer_async_write(bh);
	508	bh = bh->b_this_page;
	509	} while (bh != head);
bd2d0210 TT	510	}
bd2d0210 TT	511	unlock_page(page);
0058f965 JK	512	/* Nothing submitted - we have to end page writeback */
	513	if (!nr_submitted)
	514	end_page_writeback(page);
bd2d0210 TT	515	return ret;
bd2d0210 TT	516	}