xfs_aops.c

linux 内核源代码

				if (startio) {
					xfs_add_to_ioend(inode, bh, offset,
							type, &ioend,
							new_ioend);
				} else {
					set_buffer_dirty(bh);
					unlock_buffer(bh);
					mark_buffer_dirty(bh);
				}
				page_dirty--;
				count++;
			}
		} else if (buffer_uptodate(bh) && startio) {
			/*
			 * we got here because the buffer is already mapped.
			 * That means it must already have extents allocated
			 * underneath it. Map the extent by reading it.
			 */
			if (!iomap_valid || flags != BMAPI_READ) {
				flags = BMAPI_READ;
				size = xfs_probe_cluster(inode, page, bh,
								head, 1);
				err = xfs_map_blocks(inode, offset, size,
						&iomap, flags);
				if (err)
					goto error;
				iomap_valid = xfs_iomap_valid(&iomap, offset);
			}

			/*
			 * We set the type to IOMAP_NEW in case we are doing a
			 * small write at EOF that is extending the file but
			 * without needing an allocation. We need to update the
			 * file size on I/O completion in this case so it is
			 * the same case as having just allocated a new extent
			 * that we are writing into for the first time.
			 */
			type = IOMAP_NEW;
			if (!test_and_set_bit(BH_Lock, &bh->b_state)) {
				ASSERT(buffer_mapped(bh));
				if (iomap_valid)
					all_bh = 1;
				xfs_add_to_ioend(inode, bh, offset, type,
						&ioend, !iomap_valid);
				page_dirty--;
				count++;
			} else {
				iomap_valid = 0;
			}
		} else if ((buffer_uptodate(bh) || PageUptodate(page)) &&
			   (unmapped || startio)) {
			iomap_valid = 0;
		}

		if (!iohead)
			iohead = ioend;

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

	if (uptodate && bh == head)
		SetPageUptodate(page);

	if (startio)
		xfs_start_page_writeback(page, wbc, 1, count);

	if (ioend && iomap_valid) {
		offset = (iomap.iomap_offset + iomap.iomap_bsize - 1) >>
					PAGE_CACHE_SHIFT;
		tlast = min_t(pgoff_t, offset, last_index);
		xfs_cluster_write(inode, page->index + 1, &iomap, &ioend,
					wbc, startio, all_bh, tlast);
	}

	if (iohead)
		xfs_submit_ioend(iohead);

	return page_dirty;

error:
	if (iohead)
		xfs_cancel_ioend(iohead);

	/*
	 * If it's delalloc and we have nowhere to put it,
	 * throw it away, unless the lower layers told
	 * us to try again.
	 */
	if (err != -EAGAIN) {
		if (!unmapped)
			block_invalidatepage(page, 0);
		ClearPageUptodate(page);
	}
	return err;
}

/*
 * writepage: Called from one of two places:
 *
 * 1. we are flushing a delalloc buffer head.
 *
 * 2. we are writing out a dirty page. Typically the page dirty
 *    state is cleared before we get here. In this case is it
 *    conceivable we have no buffer heads.
 *
 * For delalloc space on the page we need to allocate space and
 * flush it. For unmapped buffer heads on the page we should
 * allocate space if the page is uptodate. For any other dirty
 * buffer heads on the page we should flush them.
 *
 * If we detect that a transaction would be required to flush
 * the page, we have to check the process flags first, if we
 * are already in a transaction or disk I/O during allocations
 * is off, we need to fail the writepage and redirty the page.
 */

STATIC int
xfs_vm_writepage(
	struct page		*page,
	struct writeback_control *wbc)
{
	int			error;
	int			need_trans;
	int			delalloc, unmapped, unwritten;
	struct inode		*inode = page->mapping->host;

	xfs_page_trace(XFS_WRITEPAGE_ENTER, inode, page, 0);

	/*
	 * We need a transaction if:
	 *  1. There are delalloc buffers on the page
	 *  2. The page is uptodate and we have unmapped buffers
	 *  3. The page is uptodate and we have no buffers
	 *  4. There are unwritten buffers on the page
	 */

	if (!page_has_buffers(page)) {
		unmapped = 1;
		need_trans = 1;
	} else {
		xfs_count_page_state(page, &delalloc, &unmapped, &unwritten);
		if (!PageUptodate(page))
			unmapped = 0;
		need_trans = delalloc + unmapped + unwritten;
	}

	/*
	 * If we need a transaction and the process flags say
	 * we are already in a transaction, or no IO is allowed
	 * then mark the page dirty again and leave the page
	 * as is.
	 */
	if (current_test_flags(PF_FSTRANS) && need_trans)
		goto out_fail;

	/*
	 * Delay hooking up buffer heads until we have
	 * made our go/no-go decision.
	 */
	if (!page_has_buffers(page))
		create_empty_buffers(page, 1 << inode->i_blkbits, 0);

	/*
	 * Convert delayed allocate, unwritten or unmapped space
	 * to real space and flush out to disk.
	 */
	error = xfs_page_state_convert(inode, page, wbc, 1, unmapped);
	if (error == -EAGAIN)
		goto out_fail;
	if (unlikely(error < 0))
		goto out_unlock;

	return 0;

out_fail:
	redirty_page_for_writepage(wbc, page);
	unlock_page(page);
	return 0;
out_unlock:
	unlock_page(page);
	return error;
}

STATIC int
xfs_vm_writepages(
	struct address_space	*mapping,
	struct writeback_control *wbc)
{
	xfs_iflags_clear(XFS_I(mapping->host), XFS_ITRUNCATED);
	return generic_writepages(mapping, wbc);
}

/*
 * Called to move a page into cleanable state - and from there
 * to be released. Possibly the page is already clean. We always
 * have buffer heads in this call.
 *
 * Returns 0 if the page is ok to release, 1 otherwise.
 *
 * Possible scenarios are:
 *
 * 1. We are being called to release a page which has been written
 *    to via regular I/O. buffer heads will be dirty and possibly
 *    delalloc. If no delalloc buffer heads in this case then we
 *    can just return zero.
 *
 * 2. We are called to release a page which has been written via
 *    mmap, all we need to do is ensure there is no delalloc
 *    state in the buffer heads, if not we can let the caller
 *    free them and we should come back later via writepage.
 */
STATIC int
xfs_vm_releasepage(
	struct page		*page,
	gfp_t			gfp_mask)
{
	struct inode		*inode = page->mapping->host;
	int			dirty, delalloc, unmapped, unwritten;
	struct writeback_control wbc = {
		.sync_mode = WB_SYNC_ALL,
		.nr_to_write = 1,
	};

	xfs_page_trace(XFS_RELEASEPAGE_ENTER, inode, page, 0);

	if (!page_has_buffers(page))
		return 0;

	xfs_count_page_state(page, &delalloc, &unmapped, &unwritten);
	if (!delalloc && !unwritten)
		goto free_buffers;

	if (!(gfp_mask & __GFP_FS))
		return 0;

	/* If we are already inside a transaction or the thread cannot
	 * do I/O, we cannot release this page.
	 */
	if (current_test_flags(PF_FSTRANS))
		return 0;

	/*
	 * Convert delalloc space to real space, do not flush the
	 * data out to disk, that will be done by the caller.
	 * Never need to allocate space here - we will always
	 * come back to writepage in that case.
	 */
	dirty = xfs_page_state_convert(inode, page, &wbc, 0, 0);
	if (dirty == 0 && !unwritten)
		goto free_buffers;
	return 0;

free_buffers:
	return try_to_free_buffers(page);
}

STATIC int
__xfs_get_blocks(
	struct inode		*inode,
	sector_t		iblock,
	struct buffer_head	*bh_result,
	int			create,
	int			direct,
	bmapi_flags_t		flags)
{
	xfs_iomap_t		iomap;
	xfs_off_t		offset;
	ssize_t			size;
	int			niomap = 1;
	int			error;

	offset = (xfs_off_t)iblock << inode->i_blkbits;
	ASSERT(bh_result->b_size >= (1 << inode->i_blkbits));
	size = bh_result->b_size;
	error = xfs_bmap(XFS_I(inode), offset, size,
			     create ? flags : BMAPI_READ, &iomap, &niomap);
	if (error)
		return -error;
	if (niomap == 0)
		return 0;

	if (iomap.iomap_bn != IOMAP_DADDR_NULL) {
		/*
		 * For unwritten extents do not report a disk address on
		 * the read case (treat as if we're reading into a hole).
		 */
		if (create || !(iomap.iomap_flags & IOMAP_UNWRITTEN)) {
			xfs_map_buffer(bh_result, &iomap, offset,
				       inode->i_blkbits);
		}
		if (create && (iomap.iomap_flags & IOMAP_UNWRITTEN)) {
			if (direct)
				bh_result->b_private = inode;
			set_buffer_unwritten(bh_result);
		}
	}

	/*
	 * If this is a realtime file, data may be on a different device.
	 * to that pointed to from the buffer_head b_bdev currently.
	 */
	bh_result->b_bdev = iomap.iomap_target->bt_bdev;

	/*
	 * If we previously allocated a block out beyond eof and we are now
	 * coming back to use it then we will need to flag it as new even if it
	 * has a disk address.
	 *
	 * With sub-block writes into unwritten extents we also need to mark
	 * the buffer as new so that the unwritten parts of the buffer gets
	 * correctly zeroed.
	 */
	if (create &&
	    ((!buffer_mapped(bh_result) && !buffer_uptodate(bh_result)) ||
	     (offset >= i_size_read(inode)) ||
	     (iomap.iomap_flags & (IOMAP_NEW|IOMAP_UNWRITTEN))))
		set_buffer_new(bh_result);

	if (iomap.iomap_flags & IOMAP_DELAY) {
		BUG_ON(direct);
		if (create) {
			set_buffer_uptodate(bh_result);
			set_buffer_mapped(bh_result);
			set_buffer_delay(bh_result);
		}
	}

	if (direct || size > (1 << inode->i_blkbits)) {
		ASSERT(iomap.iomap_bsize - iomap.iomap_delta > 0);
		offset = min_t(xfs_off_t,
				iomap.iomap_bsize - iomap.iomap_delta, size);
		bh_result->b_size = (ssize_t)min_t(xfs_off_t, LONG_MAX, offset);
	}

	return 0;
}

int
xfs_get_blocks(
	struct inode		*inode,
	sector_t		iblock,
	struct buffer_head	*bh_result,
	int			create)
{
	return __xfs_get_blocks(inode, iblock,
				bh_result, create, 0, BMAPI_WRITE);
}

STATIC int
xfs_get_blocks_direct(
	struct inode		*inode,
	sector_t		iblock,
	struct buffer_head	*bh_result,
	int			create)
{
	return __xfs_get_blocks(inode, iblock,
				bh_result, create, 1, BMAPI_WRITE|BMAPI_DIRECT);
}

STATIC void
xfs_end_io_direct(
	struct kiocb	*iocb,
	loff_t		offset,
	ssize_t		size,
	void		*private)
{
	xfs_ioend_t	*ioend = iocb->private;

	/*
	 * Non-NULL private data means we need to issue a transaction to
	 * convert a range from unwritten to written extents.  This needs
	 * to happen from process context but aio+dio I/O completion
	 * happens from irq context so we need to defer it to a workqueue.
	 * This is not necessary for synchronous direct I/O, but we do
	 * it anyway to keep the code uniform and simpler.
	 *
	 * Well, if only it were that simple. Because synchronous direct I/O
	 * requires extent conversion to occur *before* we return to userspace,
	 * we have to wait for extent conversion to complete. Look at the
	 * iocb that has been passed to us to determine if this is AIO or
	 * not. If it is synchronous, tell xfs_finish_ioend() to kick the
	 * workqueue and wait for it to complete.
	 *
	 * The core direct I/O code might be changed to always call the
	 * completion handler in the future, in which case all this can
	 * go away.
	 */
	ioend->io_offset = offset;
	ioend->io_size = size;
	if (ioend->io_type == IOMAP_READ) {
		xfs_finish_ioend(ioend, 0);
	} else if (private && size > 0) {
		xfs_finish_ioend(ioend, is_sync_kiocb(iocb));
	} else {
		/*
		 * A direct I/O write ioend starts it's life in unwritten
		 * state in case they map an unwritten extent.  This write
		 * didn't map an unwritten extent so switch it's completion
		 * handler.
		 */
		INIT_WORK(&ioend->io_work, xfs_end_bio_written);
		xfs_finish_ioend(ioend, 0);
	}

	/*
	 * blockdev_direct_IO can return an error even after the I/O
	 * completion handler was called.  Thus we need to protect
	 * against double-freeing.
	 */
	iocb->private = NULL;
}

STATIC ssize_t
xfs_vm_direct_IO(
	int			rw,
	struct kiocb		*iocb,
	const struct iovec	*iov,
	loff_t			offset,
	unsigned long		nr_segs)
{
	struct file	*file = iocb->ki_filp;
	struct inode	*inode = file->f_mapping->host;
	xfs_iomap_t	iomap;
	int		maps = 1;
	int		error;
	ssize_t		ret;

	error = xfs_bmap(XFS_I(inode), offset, 0,
				BMAPI_DEVICE, &iomap, &maps);
	if (error)
		return -error;

	if (rw == WRITE) {
		iocb->private = xfs_alloc_ioend(inode, IOMAP_UNWRITTEN);
		ret = blockdev_direct_IO_own_locking(rw, iocb, inode,
			iomap.iomap_target->bt_bdev,
			iov, offset, nr_segs,
			xfs_get_blocks_direct,
			xfs_end_io_direct);
	} else {
		iocb->private = xfs_alloc_ioend(inode, IOMAP_READ);
		ret = blockdev_direct_IO_no_locking(rw, iocb, inode,
			iomap.iomap_target->bt_bdev,
			iov, offset, nr_segs,
			xfs_get_blocks_direct,
			xfs_end_io_direct);
	}

	if (unlikely(ret != -EIOCBQUEUED && iocb->private))
		xfs_destroy_ioend(iocb->private);
	return ret;
}

STATIC int
xfs_vm_write_begin(
	struct file		*file,
	struct address_space	*mapping,
	loff_t			pos,
	unsigned		len,
	unsigned		flags,
	struct page		**pagep,
	void			**fsdata)
{
	*pagep = NULL;
	return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
								xfs_get_blocks);
}

STATIC sector_t
xfs_vm_bmap(
	struct address_space	*mapping,
	sector_t		block)
{
	struct inode		*inode = (struct inode *)mapping->host;
	struct xfs_inode	*ip = XFS_I(inode);

	vn_trace_entry(XFS_I(inode), __FUNCTION__,
			(inst_t *)__return_address);
	xfs_rwlock(ip, VRWLOCK_READ);
	xfs_flush_pages(ip, (xfs_off_t)0, -1, 0, FI_REMAPF);
	xfs_rwunlock(ip, VRWLOCK_READ);
	return generic_block_bmap(mapping, block, xfs_get_blocks);
}

STATIC int
xfs_vm_readpage(
	struct file		*unused,
	struct page		*page)
{
	return mpage_readpage(page, xfs_get_blocks);
}

STATIC int
xfs_vm_readpages(
	struct file		*unused,
	struct address_space	*mapping,
	struct list_head	*pages,
	unsigned		nr_pages)
{
	return mpage_readpages(mapping, pages, nr_pages, xfs_get_blocks);
}

STATIC void
xfs_vm_invalidatepage(
	struct page		*page,
	unsigned long		offset)
{
	xfs_page_trace(XFS_INVALIDPAGE_ENTER,
			page->mapping->host, page, offset);
	block_invalidatepage(page, offset);
}

const struct address_space_operations xfs_address_space_operations = {
	.readpage		= xfs_vm_readpage,
	.readpages		= xfs_vm_readpages,
	.writepage		= xfs_vm_writepage,
	.writepages		= xfs_vm_writepages,
	.sync_page		= block_sync_page,
	.releasepage		= xfs_vm_releasepage,
	.invalidatepage		= xfs_vm_invalidatepage,
	.write_begin		= xfs_vm_write_begin,
	.write_end		= generic_write_end,
	.bmap			= xfs_vm_bmap,
	.direct_IO		= xfs_vm_direct_IO,
	.migratepage		= buffer_migrate_page,
};